The Anatomy of Discretion - An Analysis of Prosecutorial Decision Making, Frederick and Stemen, 2012

• Aug. 1, 2016 • Locations: United States of America • Topics: Crime/Demographics, Criminal Prosecution

Download original document:
Brief thumbnail

This text is machine-read, and may contain errors. Check the original document to verify accuracy.

The author(s) shown below used Federal funds provided by the U.S.
Department of Justice and prepared the following final report:
Document Title:

The Anatomy of Discretion: An Analysis of
Prosecutorial Decision Making – Technical
Report

Author:

Bruce Frederick, Don Stemen

Document No.:

240334

Date Received:

December 2012

Award Number:

2009-IJ-CX-0040

This report has not been published by the U.S. Department of Justice.
To provide better customer service, NCJRS has made this Federallyfunded grant final report available electronically in addition to
traditional paper copies.
Opinions or points of view expressed are those
of the author(s) and do not necessarily reflect
the official position or policies of the U.S.
Department of Justice.

This document is a research report submitted to the U.S. Department of Justice. This report has not
been published by the Department. Opinions or points of view expressed are those of the author(s)
and do not necessarily reflect the official position or policies of the U.S. Department of Justice.

The Anatomy of Discretion: An Analysis of Prosecutorial
Decision Making – Technical Report
Final Report to the National Institute of Justice
Grant No: 2009-IJ-CX-0040

Bruce Frederick, Principal Investigator
Vera Institute of Justice
Don Stemen, Principal Investigator
Loyola University Chicago

December 2012

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making

© 2012 Vera Institute of Justice. All rights reserved.
Requests for additional information about the research described in this report should be directed to
contactvera@vera.org.
This study was supported by grant #2009-IJ-CX-0040 from the National Institute of Justice, Office
of Justice Programs, U.S. Department of Justice. Points of view expressed in this report are those of
the authors and do not necessarily represent the official position of the U.S. Department of Justice
or the views of the district attorneys in participating jurisdictions.
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making i

Abstract
Prosecuting attorneys enjoy broader discretion in making decisions that influence criminal case
outcomes than any other actors in the American justice system. That they do so with little or no
public scrutiny suggests questions about justice and fairness. This study examines the impact of
legal, quasi-legal, and extra-legal factors on case outcomes throughout the prosecutorial process. It
then examines how prosecutors weigh these factors in their decision making and explores the
formal and informal mechanisms that constrain or regulate prosecutors’ decision-making.
The study examines case screening decisions, charging decisions, plea offers, sentence
recommendations, and dismissals in two moderately large county prosecutors’ offices. It includes
statistical analyses of actual case outcomes, responses to a standardized set of hypothetical cases,
and responses to a survey of prosecutors’ opinions and priorities, as well as qualitative analyses of
two waves of individual interviews and focus groups. It addresses the following questions:


How did prosecutors define and apply the concepts of justice and fairness?



What factors were associated with prosecutorial outcomes at each stage?



How did prosecutors interpret and weigh different case-specific factors in making decisions
at each stage?



How did contextual factors constrain or regulate prosecutorial decision making?



How consistent were prosecutors’ decisions across similar cases? What case-level and
contextual factors influenced the degree of consistency?

Two county prosecutors’ offices participated in the study—labeled Northern County and
Southern County in project reports. Analyses of administrative data for Northern County examined
76,721 felony and misdemeanor cases screened between January 2009 and June 2011 that involved
person, property, drug, public order, domestic violence, weapons, or driving under the influence
offenses. Analyses of administrative data for Southern County examined 4,890 felony drug cases
screened between May 2007 and July 2009 and 1,164 felony person and property cases screened
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making ii

between January 2007 and June 2007. Analyses of a 76-item opinion survey conducted in the fall
of 2010 examined responses from 62 Northern County prosecutors (a 67 percent response rate) and
65 Southern County prosecutors (a 94 percent response rate). Analyses of decisions in hypothetical
cases examined responses to 10 vignettes by each of 62 prosecutors in Southern County in the fall
of 2011. Finally, two waves of interviews and focus groups were conducted in Southern County
during October 2010 and December 2010, and in Northern County during November 2010 and
March 2011.
Researchers found that prosecutors’ decisions were guided by two basic questions: “Can I prove
the case?” and “Should I prove the case?” The relative influence of these questions was found to
shift over the course of a case. The first question was most influential at the outset of a case, where
the objective strength of evidence was the determining factor in most screening decisions. Later,
factors such as the seriousness of the offense, the defendant’s criminal history, characteristics of the
defendant and victim, and contextual factors became increasingly influential, as prosecutors
evaluated whether a case should go forward.
While prosecutorial discretion is generally seen as very broad and unconstrained, prosecutors
often rely on a fairly limited array of legal and quasi-legal factors to make decisions, and their
decision making is further constrained by several contextual factors.. These contextual constraints—
rules, resources, and relationships—sometimes trump evaluations of the strength of the evidence,
the seriousness of the offense, and the defendant’s criminal history. Future evaluations of
prosecutorial outcomes should consider these contextual constraints when assessing the impact of
case-level factors, and chief prosecutors and criminal justice policy makers should be alert to the
potential for contextual factors to influence and possibly distort the exercise of prosecutorial
discretion.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making iii

Table of Contents
Acknowledgements ....................................................................................................................... vii
List of Figures ................................................................................................................................ ix
List of Tables ................................................................................................................................. xv
List of Appendices ....................................................................................................................... xix
Part 1. Introduction and Research Questions .................................................................................. 1
1.1 Introduction ................................................................................................................................ 1
1.2 Review of Relevant Literature .................................................................................................... 2
1.3 Research Questions................................................................................................................... 11
1.4 Site Selection ............................................................................................................................ 13
1.5 Methods Overview ................................................................................................................... 18
1.6 Report Overview....................................................................................................................... 20
Part 2. The Prosecutor’s Perspective ............................................................................................. 22
2.1 Introduction .............................................................................................................................. 22
2.2 Summary of Research Methods: Focus Groups and General Survey ..................................... 22
2.3 The Goals of Prosecution: Justice, Consistency, and Efficiency ............................................. 25
2.3.1 Justice ................................................................................................................................ 25
2.3.2 Consistency and Flexibility ............................................................................................... 47
2.3.3 Efficiency .......................................................................................................................... 54
2.3.4 Prosecutors’ Perspective on the Goals of Prosecution ...................................................... 58
2.4 Strength of the Evidence and Severity of the Case: Case-Specific Factors in Decision Making
........................................................................................................................................................ 59
2.4.1 Strength of the evidence .................................................................................................... 60
2.4.2 Severity of the offense, defendant criminal history, and other case-specific factors ........ 66
2.4.3 Improving the evaluation of evidence ............................................................................... 70
2.5 Rules, Resources, and Relationships: Outside Influences on Decision Making ...................... 71
2.5.1 Rules .................................................................................................................................. 71
2.5.2 Resources .......................................................................................................................... 88
2.5.3 Relationships ..................................................................................................................... 98
2.6 Understanding the Prosecutor’s Perspective .......................................................................... 116
Part 3. The Stages of the Prosecutorial Process: Analyzing Case Outcomes.............................. 117
3.1. Introduction ........................................................................................................................... 117

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making iv

3.2. Summary of Research Methods: Administrative Data, Factorial Survey, Focus Groups, and
Surveys ......................................................................................................................................... 118
3.2.1 Examining Case Outcomes in Southern County ............................................................. 118
3.2.2 Examining Case Outcomes in Northern County ............................................................. 121
3.2.3 Factorial Survey .............................................................................................................. 123
3.2.4 Focus Groups and General Survey .................................................................................. 127
3.3 Case Processing Overview ..................................................................................................... 127
3.3.1 Case Processing Overview for Southern County ............................................................ 127
3.3.2 Case Processing Overview for Northern County ............................................................ 130
3.3.3 Comparison and Summary .............................................................................................. 132
3.4 Screening ................................................................................................................................ 133
3.4.1 Screening Decisions in Southern County ........................................................................ 137
3.4.2 Screening Decisions in Northern County .................................................................... 145
3.4.3 Factors Influencing the Screening Decision: Hypothetical Cases............................... 169
3.4.4 Factors Influencing the Screening Decision: The Primacy of Evidence and the Severity
of the Case ............................................................................................................................ 175
3.5 Charging ................................................................................................................................. 176
3.5.1 Factors Influencing the Charging Decision: Administrative Data from Actual Cases ... 181
3.5.2 Factors Influencing the Charging Decision: Hypothetical Cases .................................... 182
3.6 Plea Offers .............................................................................................................................. 195
3.6.1 Factors Influencing the Plea Offer Decision: Administrative Data from Actual Cases .. 201
3.6.2. Factors Influencing Plea Offers: Hypothetical Cases..................................................... 222
3.7 Dismissals ............................................................................................................................... 244
3.7.1 Dismissal Decisions in Southern County ........................................................................ 246
3.7.2 Dismissal Decisions in Northern County ........................................................................ 255
3.7.3 Reasons for Dismissal in Southern County ..................................................................... 265
3.8 Understanding Case Outcomes across the Prosecutorial Process .......................................... 270
Part 4. Conclusions ...................................................................................................................... 275
4.1 Key Findings........................................................................................................................... 275
4.1.1 Prosecutors Are Guided by Two Questions: ‘Can I prove the case?’ and ‘Should I prove
the case?’ .................................................................................................................................. 275
4.1.2 Prosecutorial Decision Making Is Often Affected by External Constraints ................... 277
4.1.3 Inconsistency and Variation in Approach Are Common ................................................ 280
4.2 Continuing Challenges ........................................................................................................... 281
4.2.1 External Resources .......................................................................................................... 281
4.2.2 Training ........................................................................................................................... 282
4.2.3 Balancing Consistency and Flexibility ............................................................................ 283
4.3 Limitations .............................................................................................................................. 284
4.4 Conclusion .............................................................................................................................. 286
References ................................................................................................................................... 287
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making v

Dissemination of Research Findings ........................................................................................... 292
Appendices .................................................................................................................................. 294

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making vi

Acknowledgements
Many people helped make this study possible. First and foremost, we would like to thank the
district attorneys in Southern County and Northern County for providing us access to their offices
and for consistently supporting and facilitating our efforts. We would especially like to thank the
staff in both counties for generously contributing their time and thoughtful insights throughout two
years of research.
At the National Institute of Justice, our grant officer, Linda Truitt, provided support,
encouragement, and critical assistance—and displayed extraordinary patience in helping us work
through unanticipated difficulties in the management of the project. We are also grateful to Vera
Institute Director Michael Jacobson, Associate Research Director Siobhan Carney, and the many
Vera staff who supported us throughout this project. We would especially like to thank Christine
Scott-Hayward and Kathryne Brewer, who helped write the original proposal; April Bang and
Olivia Sideman who helped enormously with project administration and logistics; Stephen O’Neil
and Kevin Alligood, who assisted with on-site coordination and original data collection; General
Counsel Karen Goldstein and Legal Department staff for assistance with contracts, data sharing
agreements, and human subjects research protocols; Maureen Christensen and Shelley Azumbrado
for invaluable assistance with budgeting and fiscal management for the project; Michael Mehler for
recording and editing video podcasts; and Director of Communications Robin Campbell, writer and
editor Alice Chasan, and Communications Department support staff for their hard work and
commitment to the difficult task of making this research accessible to a broad audience.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making vii

Most importantly, we would like to thank Neil Weiner, who as the Research Director at the
Vera Institute of Justice was the primary creative force behind the development and design of this
project. Sadly, he passed away before the project could begin. We dedicate this project to him.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making viii

List of Figures
Figure 1.5-1 Overview of Research Methods ...................................................................................................20
Figure 2.3.1-1 Responses to the question “The most important function of the criminal justice system is to
prevent and repress crime,” by research site ....................................................................................................29
Figure 2.3.1-2 Responses to question “The most important role of the criminal justice system is to prevent
and repress crime,” by research site and ADA experience ...............................................................................29
Figure 2.3.1-3 Responses to the question “The most important function of the criminal justice system is
protecting the rights of the accused,” by research site......................................................................................30
Figure 2.3.1-4 Responses to the question “Sanctioning offenders should involve punishment rather than
rehabilitation,” by research site ........................................................................................................................31
Figure 2.3.1-5 Responses to the question “Many community-based programs do not provide sufficient
punishment for offenders,” by research site .....................................................................................................32
Figure 2.3.1-6 Responses to the question “Many offenders currently imprisoned could be adequately handled
in non-prison sanctions,” by research site ........................................................................................................33
Figure 2.3.1-7 Responses to the question: “Offenders do not need to be punished in order to be rehabilitated,”
by research site and ADA experience...............................................................................................................33
Figure 2.3.1-8 Responses to the question: “Low dismissal rates after charges are filed is important for
individual success,” by research site ................................................................................................................35
Figure 2.3.1-9 Responses to the question: “Low dismissal rates after charges are filed is important for
individual success,” by research site and ADA experience ..............................................................................36
Figure 2.3.1-10 Responses to the question: “High deferral/diversion rates for eligible defendants are
important for individual success,” by research site ..........................................................................................37
Figure 2.3.1-11 Responses to the question: “High success/completion rates for defendants deferred/diverted
are important for individual success,” by research site ....................................................................................38
Figure 2.3.1-12 Responses to the question: “High deferral/diversion rates for eligible defendants is important
for individual success,” by research site and ADA experience ........................................................................38
Figure 2.3.1-13 Responses to the question: “High conviction rates is important for individual success,” by
research site ......................................................................................................................................................39
Figure 2.3.1-14 Responses to the question: “High rate of guilty pleas to most serious charge(s) is important
for individual success,” by research site ...........................................................................................................40
Figure 2.3.1-15 Responses to the question: “High rate of imprisonment for serious crimes is important for
individual success,” by research site ................................................................................................................40
Figure 2.3.1-16 Responses to the question: “High rate of guilty pleas to most serious charge(s) is important
for individual success,” by research site and ADA experience ........................................................................41
Figure 2.3.1-17 Responses to the question: “Fair treatment of defendants is important for individual success,”
by research site and ADA experience...............................................................................................................42
Figure 2.3.1-18 Responses to the question: “A high rate of citizen satisfaction with the DA’s office is
important for office success,” by research site .................................................................................................44
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making ix

Figure 2.3.1-19 Responses to the question: “Fewer defendants re-arrested after prosecution is important for
office success,” by research site .......................................................................................................................46
Figure 2.3.1-20 Responses to the question: “Lower crime rates is important for office success,” by research
site.....................................................................................................................................................................47
Figure 2.3.2-1 Responses to the question: “Similar outcomes for similar cases within units is important for
office success,” by research site .......................................................................................................................52
Figure 2.3.2-2 Responses to the question: “Similar outcomes for similar cases across units is important for
office success,” by research site .......................................................................................................................53
Figure 2.3.2-3 Responses to the question: “For similar cases there should be a great deal of consistency
across prosecutors in the factors that influence the decision on a case,” by research site ................................53
Figure 2.3.3-1 Responses to the question: “How important is it to examine cases at screening in terms of their
plea bargaining potential, their potential for early disposition, and their potential to reduce the number of
cases in the system,” by research site ...............................................................................................................56
Figure 2.3.3-2 Responses to question: “I examine cases at screening in terms of their plea bargaining
potential, their potential for early disposition, and their potential to reduce the number of cases in the
system,” by research site and ADA experience ................................................................................................57
Figure 2.3.3-3 Responses to the question: “I am willing to adjust my decisions to increase courtroom
efficiency,” by research site..............................................................................................................................58
Figure 2.4.1-1 Responses to question: “The most important consideration when screening a case for
prosecution,” by research site ...........................................................................................................................61
Figure 2.5.1-1 Response to question: “I feel constrained by office policies and practices about when to accept
or decline cases for prosecution,” by research site ...........................................................................................76
Figure 2.5.1-2 Response to question: “Office policies compel me to decline cases that I would prefer to
prosecute,” by research site ..............................................................................................................................76
Figure 2.5.1-3 Response to question: “Office priorities require case outcomes that I often disagree with,” by
research site ......................................................................................................................................................77
Figure 2.5.1-4 Response to question: “Prosecutors in my office receive adequate training when they are
transferred to a new unit,” by research site.......................................................................................................78
Figure 2.5.1-5 Response to question: “Office goals and priorities are clearly communicated to staff,” by
research site ......................................................................................................................................................78
Figure 2.5.1-6 Response to question: “I often compare my cases to those of my colleagues to ensure that I am
getting similar outcomes,” by research site ......................................................................................................82
Figure 2.5.1-7 Response to question: “It is important to routinely review cases as a group,” by research site83
Figure 2.5.1-8 Response to question: “There needs to be more communication among staff to ensure
consistency of outcomes,” by research site ......................................................................................................83
Figure 2.5.1-9 Response to question: “I often discuss how to handle cases with my colleagues,” by research
site and ADA experience ..................................................................................................................................85
Figure 2.5.1-10 Response to question: “I often compare my cases to those of my colleagues to ensure that I
am getting similar outcomes,” by research site and ADA experience..............................................................86

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making x

Figure 2.5.1-11 Response to question: “It is important to routinely review cases as a group,” by research site
and ADA experience ........................................................................................................................................86
Figure 2.5.1-12 Response to question: “There needs to be more communication among staff to ensure
consistency of outcomes,” by research site and ADA experience ....................................................................87
Figure 2.5.2-1 Response to question: “The lack of investigators in my office affects the outcomes of my
cases,” by research site .....................................................................................................................................94
Figure 2.5.2-2 Response to question: “The lack of support staff to coordinate contacts with victims,
witnesses, and defense counsel adversely affects my cases,” by research site .................................................94
Figure 2.5.2-3 Response to question: “There is enough clerical staff to assist me with paperwork on all my
cases,” by research site .....................................................................................................................................95
Figure 2.5.2-4 Response to question: “The lack of technology in the office makes work difficult,” by research
site.....................................................................................................................................................................95
Figure 2.5.2-5 Response to question: “Caseloads prevent me from devoting enough time to all my cases,” by
research site ......................................................................................................................................................96
Figure 2.5.2-6 Response to question: “I decline or dismiss cases when the amount of time and effort need to
obtain a conviction exceeds the benefits of the potential sentence,” by research site ......................................97
Figure 2.5.3-1 Response to question: “Good relations with defense attorneys is important to individual
success,” by research site................................................................................................................................104
Figure 2.5.3-2 Response to question: “Good relations with defense bar is important to office success,” by
research site ....................................................................................................................................................104
Figure 2.5.3-3 Response to question: “I would consider altering my decisions for defense attorneys who I
respect” by research site .................................................................................................................................105
Figure 2.5.3-4 Response to question: “I tailor my decisions to gain or to maintain the trust of defense
attorneys,” by research site .............................................................................................................................105
Figure 2.5.3-5 Response to question: “Good relationships with law enforcement officers is important to
individual success,” by research site ..............................................................................................................108
Figure 2.5.3-6 Response to question: “Good relations with law enforcement agencies is important to office
success,” by research site................................................................................................................................109
Figure 2.5.3-7 Response to question: “I feel pressure from law enforcement officers to accept cases for
prosecution,” by research site .........................................................................................................................109
Figure 2.5.3-8 Response to question: “I feel pressure from law enforcement officers to accept cases for
prosecution,” by research site and ADA experience ......................................................................................110
Figure 2.5.3-9 Response to question: “I make decisions based on how my colleagues will perceive my
performance,” by research site .......................................................................................................................114
Figure 2.5.3-10 Response to question: “The decisions I make are affected by how I think they will be
perceived by my supervisor,” by research site ...............................................................................................114
Figure 2.5.3-11 Response to question: “I make decisions based on how my colleagues perceive my
performance,” by research site and ADA experience .....................................................................................115
Figure 2.5.3-12 Response to question: “The decisions I make are affected by how I think they will be
perceived by my supervisor,” by research site and ADA experience .............................................................115
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making xi

Figure 3.4-1 Percentage of cases accepted for prosecution in Southern and Northern Counties, by crime type
and offense level .............................................................................................................................................137
Figure 3.4.1-1 Acceptance rates for cases screened in Southern County .......................................................138
Figure 3.4.2-1 Number of cases screened in Northern County, by offense type and offense severity level ..146
Figure 3.4.2-2 Ten most frequently screened cases in Northern County, by offense type and offense severity
level ................................................................................................................................................................147
Figure 3.4.2-3 Percent of cases screened in Northern County involving single or multiple arrest charges, by
offense type ....................................................................................................................................................148
Figure 3.4.2-4 Percent of cases screened in Northern County involving primary municipal police department
or other law enforcement agency, by offense type .........................................................................................149
Figure 3.4.2-5 Percent of cases screened in Northern County, by defendant race/gender .............................150
Figure 3.4.2-6 Percent of cases screened in Northern County involving defendants in custody or not in
custody, by offense type .................................................................................................................................151
Figure 3.4.2-7 Percent of cases screened in Northern County, by victim race/gender ...................................152
Figure 3.4.2-8 Percent of cases screened in Northern County by a specialized or general crimes unit, by
offense type ....................................................................................................................................................153
Figure 3.4.2-9 Percent of cases screened in Northern County by inexperienced, mid-career, and experienced
ADAs, by offense type ...................................................................................................................................154
Figure 3.4.3-1 Proportion of hypothetical cases rejected at screening, by vignette .......................................170
Figure 3.4.3-2 Proportion of hypothetical cases rejected at screening, by offense seriousness and prosecutor’s
rating of strength of evidence .........................................................................................................................173
Figure 3.5-1 Response to question: “Realizing that each case is unique, which of the following best describes
your general approach to charging?” by research site ....................................................................................178
Figure 3.5.1-1 Percent of cases in which most serious arrest charge was issued ...........................................182
Figure 3.5.2-1 Mean number of charges to be filed per hypothetical case, by vignette .................................184
Figure 3.5.2-2 Mean statutory rank of top filing charge, by vignette .............................................................185
Figure 3.5.2-3 Mean number of charges to be filed per hypothetical case by offense seriousness and
prosecutor’s rating of strength of evidence (among cases accepted for prosecution) ....................................189
Figure 3.5.2-4 Mean number of charges to be filed per hypothetical case by offense seriousness and
prosecutor’s rating of defendant’s criminal history (among cases accepted for prosecution) ........................190
Figure 3.5.2-5 Mean statutory rank of top filing charge by offense seriousness and prosecutor’s rating of
strength of evidence (among cases accepted for prosecution) ........................................................................194
Figure 3.5.2-6 Mean statutory rank of top filing charge by offense seriousness and prosecutor’s rating of
criminal history (among cases accepted for prosecution) ...............................................................................194
Figure 3.6-1 Response to question: “Plea offer should focus primarily on length of sentence,” by research site
........................................................................................................................................................................197
Figure 3.6-2 Response to question: “Plea offer should focus primarily on offense severity,” by research site
........................................................................................................................................................................197
Figure 3.6-3 Response to question: “Plea offer should include all charges filed with an offer to forego
additional charges if the plea offer is accepted,” by research site ..................................................................199
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making xii

Figure 3.6-4 Response to question: “A plea offer should include only the charges the defendant should plead
guilty to with a threat to add additional charges if the plea offer is not accepted,” by research site ..............199
Figure 3.6.2-1 Mean number of charges requiring guilty plea per plea offer, by vignette .............................223
Figure 3.6.2-2 Range of number of charges requiring guilty plea per plea offer, by vignette, across
prosecutors......................................................................................................................................................224
Figure 3.6.2-3 Mean statutory rank of top plea offer charge, by vignette ......................................................225
Figure 3.6.2-4 Range of statutory rank of top plea offer charge, by vignette, across prosecutors .................225
Figure 3.6.2-5 Proportion of accepted hypothetical cases for which plea offer would recommend a period of
incarceration ...................................................................................................................................................227
Figure 3.6.2-6 Mean number of plea offer charges requiring guilty plea by offense seriousness and
prosecutor’s rating of strength of evidence (among cases accepted for prosecution) ....................................229
Figure 3.6.2-7 Mean number of plea offer charges requiring guilty plea by offense seriousness and
prosecutor’s rating of defendant’s criminal history (among cases accepted for prosecution) ........................230
Figure 3.6.2-8 Mean number of plea offer charges requiring guilty plea by offense seriousness and
prosecutor’s rating of strength of evidence (for prosecutors with 2 to 5 years’ experience in the Southern
County DA’s office) .......................................................................................................................................231
Figure 3.6.2-9 Mean number of plea offer charges requiring guilty plea by offense seriousness and
prosecutor’s rating of strength of evidence (for prosecutors with more than 5 years’ experience in the
Southern County DA’s office) ........................................................................................................................231
Figure 3.6.2-10 Mean statutory rank of top plea offer charge per hypothetical case by offense seriousness and
prosecutor’s rating of strength of evidence (among cases accepted for prosecution) ....................................235
Figure 3.6.2-11 Mean statutory rank of top plea offer charge per hypothetical case by offense seriousness and
prosecutor’s rating of criminal history (among cases accepted for prosecution) ...........................................236
Figure 3.6.2-12 Mean statutory rank of top plea offer charge by offense seriousness and prosecutor’s rating
of criminal history (for prosecutors who disagreed with General Survey Item Q8a) ....................................237
Figure 3.6.2-13 Mean statutory rank of top plea offer charge by offense seriousness and prosecutor’s rating
of criminal history (for prosecutors who agreed with General Survey Item Q8a) .........................................238
Figure 3.6.2-14 Proportion of accepted hypothetical cases with incarceration recommendations by offense
seriousness and prosecutor’s rating of strength of evidence...........................................................................242
Figure 3.6.2-15 Proportion of accepted hypothetical cases with incarceration recommendations by offense
seriousness and prosecutor’s rating of defendant’s criminal history ..............................................................242
Figure 3.7.1-1 Acceptance rates and non-conviction rates among accepted cases by prosecutor unit in
Southern County .............................................................................................................................................247
Figure A.3-1 Hypothetical cases, by arrest offense seriousness and researchers’ classification of strength of
evidence ..........................................................................................................................................................308
Figure A.3-2 Hypothetical cases, by arrest seriousness and prosecutors’ ratings of strength of evidence.....308
Figure A.3-3 Percent of hypothetical cases by offense seriousness and prosecutors’ rating of defendant
criminal history ...............................................................................................................................................310
Figure A.3-4 Percentage of hypothetical cases, by defendant race and offense seriousness ..........................311
Figure A.3-5 Hypothetical cases, by offense seriousness and defendant’s approximate age at arrest ...........312
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making xiii

Figure A.3-6 Prosecutor race and gender .......................................................................................................313
Figure A.3-7 Distribution of prosecutors’ years of experience in the Southern County prosecutor’s office .314
Figure A.3-8 Distribution of prosecutors’ scores on a 4-item severity scale incorporated in the factorial
survey analyses ...............................................................................................................................................315
Figure A.3-9 Distribution of prosecutors’ scores on a 3-item fairness scale incorporated in the factorial
survey analyses ...............................................................................................................................................317
Figure A.3-10 Distribution of prosecutors’ scores on a 4 item consistency scale incorporated in the factorial
survey analyses ...............................................................................................................................................318

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making xiv

List of Tables
Table 1.4-1 Characteristics of the population served, by research site.............................................................14
Table 1.4-2 Selected characteristics of participating prosecutors’ offices .......................................................17
Table 1.4-3 Characteristics of prosecuting attorneys, by jurisdiction ..............................................................18
Table 3.2.3-1 Arrest charges in factorial survey vignettes, classified by strength of evidence and seriousness
of top arrest charge .........................................................................................................................................124
Table 3.3.1-1 Selected outcome measures by crime type and selected processing stages in Southern County
........................................................................................................................................................................129
Table 3.3.2-1 Selected outcome measures by crime type and selected processing stages in Northern County
........................................................................................................................................................................131
Table 3.4.1-1 Percentage of screened cases accepted for prosecution in Southern County, by selected case
characteristics .................................................................................................................................................138
Table 3.4.1-2 Final logistic regression models for effects of selected factors on the odds of acceptance at
screening in Southern County.........................................................................................................................143
Table 3.4.2-1 Percentage of screened cases accepted for prosecution in Northern County, by selected case
characteristics .................................................................................................................................................155
Table 3.4.2-2 HLM unconditional models predicting cases accepted at screening in Northern County........159
Table 3.4.2-3 Individual-level factors predicting cases accepted at screening in Northern County...............161
Table 3.4.2-4 Two-stage HLM models predicting cases accepted at screening in Northern County,
prosecutor-level predictors .............................................................................................................................167
Table 3.4.3-1 Hierarchical logistic regression model for the decision to reject at screening .........................172
Table 3.5.2-1 Hierarchical regression model for the number of charges to file .............................................188
Table 3.5.2-2 Hierarchical regression model for statutory rank of top filing charge .....................................193
Table 3.6.1-1 Number of charges filed and requiring a guilty plea in plea offers ..........................................203
Table 3.6.1-2 Final linear regression models for effects of selected factors on change in number of charges
from initial filing to plea offer in Southern County ........................................................................................204
Table 3.6.1-3 Average statutory rank of top filed charge and top charges requiring a guilty plea in plea offers
........................................................................................................................................................................206
Table 3.6.1-4 Final linear regression models for effects of selected factors on change in statutory class of top
charge from initial filing to plea offer in Southern County ............................................................................208
Table 3.6.1-5 Average months of exposure for top filed charge and for top charges requiring a guilty plea in
plea offers .......................................................................................................................................................209
Table 3.6.1-6 Final linear regression models for effects of selected factors on change in top charge
incarceration exposure from initial filing to plea offer in Southern County...................................................211
Table 3.6.1-7 Average months of exposure for top filed charge and for top charges requiring a guilty plea in
plea offers .......................................................................................................................................................213
Table 3.6.1-8 Final linear regression models for effects of selected factors on change in aggregate
incarceration exposure from initial filing to plea offer in Southern County...................................................214
Table 3.6.1-9 Final logistic regression models for effects of selected factors on the probability of an
incarceration recommendation in Southern County .......................................................................................217
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making xv

Table 3.6.2-1 Hierarchical regression model for the number of plea offer charges requiring guilty plea......228
Table 3.6.2-2 Hierarchical regression model for statutory rank of top plea offer charge ...............................234
Table 3.6.2-3 Hierarchical logistic regression model for the probability of an incarceration recommendation
........................................................................................................................................................................240
Table 3.7.1-1 Percentage of filed cases with non-conviction dispositions in Southern County, by selected
case characteristics .........................................................................................................................................248
Table 3.7.1-2 Logistic regression models for predicting non-conviction given filing in Southern County ...252
Table 3.7.2-1 Percentage of accepted cases dismissed in Northern County, by selected case characteristics
........................................................................................................................................................................257
Table 3.7.2-2 Logistic regression predicting case dismissal in Northern County ..........................................261
Table 3.7.3-1 Distribution of reasons for charge dismissal by processing stage for the Person and Property
Teams in Southern County .............................................................................................................................266
Table 3.7.3-2 Distribution of reasons for charge dismissal by processing stage for the Drug Unit in Southern
County ............................................................................................................................................................267
Table A.3-1 Factorial survey vignette descriptions ........................................................................................304
Table A.3-2 Descriptions of the component items for severity, fairness, and consistency scales ..................316
Table A.3-3 Distributions of prosecutors’ responses to general survey items incorporated in the factorial
survey analyses ...............................................................................................................................................320
Table A.3-4 Distribution of prosecutors’ responses to general survey question 7 – “Realizing that each case is
unique, which of the following best describes your general approach to charging?:” ...................................320
Table A.6-1 Administrative data available for analysis, by participating jurisdiction ...................................335
Table B1: Response distributions for General Survey Question 1 – “Realizing that each case is
unique, please indicate how important the following outcomes are in how you evaluate your own
success as a prosecutor:” ...........................................................................................................................336
Table B2: Response distributions for General Survey Question 2 – “Please indicate how important
you believe the following outcomes are to the District Attorney in evaluating the overall success of
the office in which you currently work:” ....................................................................................................337
Table B3: Response distributions for General Survey Question 3 – “Please indicate how frequently
you experience the following situations/feelings in your work:” ............................................................338
Table B4: Response distributions for General Survey Question 4 – “Realizing that each case is
unique, please indicate how frequently you find yourself encountering the following circumstances:”
........................................................................................................................................................................339
Table B5: Response distributions for General Survey Question 5 – “In addition to examining cases
in terms of the legal elements of the alleged offense, please circle the number that best
corresponds to how important the following considerations are in screening a case for
prosecution:” .................................................................................................................................................340
Table B6: Response distributions for General Survey Question 6 – “Of the considerations listed
below, which is the most important when screening a case for prosecution?:” .................................341
Table B7: Response distributions for General Survey Question 7 – “Realizing that each case is
unique, which of the following best describes your general approach to charging?:” .......................342
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making xvi

Table B8: Response distributions for General Survey Question 8– “Please circle the number that
best corresponds to your level of agreement with the following statements:” ....................................343
Table B9: Response distributions for General Survey Question 9– “Please indicate how much you
generally agree or disagree with each statement:”.................................................................................344
Table B10: Response distributions for General Survey Question 10– “Please circle the number
that best corresponds to your level of agreement with the following statements:” ............................345
Table B11: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 1 – “Realizing
that each case is unique, please indicate ................................................................................................346
Table B12: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 2 – “Please
indicate how important you believe the following outcomes are to the District Attorney in evaluating
the overall success of the office in which you currently work :” ............................................................347
Table B13: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 3– “Please
indicate how frequently you experience the following situations/feelings in your work:” ..................348
Table B14: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 4– “Realizing
that each case is unique, please indicate how frequently you find yourself encountering the
following circumstances:” ...........................................................................................................................349
Table B15: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 5– “In addition to
examining cases in terms of the legal elements of the alleged offense, please circle the number
that best corresponds to how important the following considerations are in screening a case for
prosecution:” .................................................................................................................................................350
Table B16: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 8– “ Please
circle the number that best corresponds to your level of agreement with the following statements:”
........................................................................................................................................................................351
Table B17: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 9– “Please
indicate how much you generally agree or disagree with each statement:” .......................................352
Table B18: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 10– “Please
indicate how much you generally agree or disagree with each statement:” .......................................353
Table B19: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 1 – “Realizing
that each case is unique, please indicate how important the following outcomes are in how you
evaluate your own success as a prosecutor:” .........................................................................................354
Table B20: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 2 – “Please
indicate how important you believe the following outcomes are to the DA in evaluating the overall
success of the office in which you currently work:”.................................................................................355
Table B21: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 3 – “Please
indicate how frequently you experience the following situations/feelings in your work:” ..................356
Table B22: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 4 – “Realizing
that each case is unique, please indicate how frequently you find yourself encountering the
following circumstances:” ...........................................................................................................................357
Table B23: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 5 – “In
addition to examining cases in terms of the legal elements of the alleged offense, please circle the
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making xvii

number that best corresponds to how important the following considerations are in screening a
case for prosecution:” ..................................................................................................................................358
Table B24: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 8 – “ Please
circle the number that best corresponds to your level of agreement with the following statements:”
........................................................................................................................................................................359
Table B25: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 9 – “Please
indicate how much you generally agree or disagree with each statement:” .......................................360
Table B26: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 10 – “Please
circle the number that best corresponds to your level of agreement with the following statements:”
........................................................................................................................................................................361
Table B27: Dimensions of Variation in Survey Responses............................................................................362
Table C1: Descriptive Statistics for Individual- and Prosecutor-Level Predictors, Person Cases .................365
Table C2: Descriptive Statistics for Individual- and Prosecutor-Level Predictors, Property Cases without
Victim Characteristics ....................................................................................................................................366
Table C3: Descriptive Statistics for Individual- and Prosecutor-Level Predictors, Property Cases with Victim
Characteristics ................................................................................................................................................367
Table C4: Descriptive Statistics for Individual- and Prosecutor-Level Predictors, Drug Cases ....................368
Table C5: Descriptive Statistics for Individual- and Prosecutor-Level Predictors, Public Order Cases ........369
Table C6: Descriptive Statistics for Individual- and Prosecutor-Level Predictors, Domestic Violence Cases
........................................................................................................................................................................370
Table C7: Descriptive Statistics for Individual- and Prosecutor-Level Predictors, Weapons Cases..............371

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making xviii

List of Appendices
Appendix A: Methods ...................................................................................................................... 295
Appendix B: Supplemental Tables for General Survey Responses ................................................. 336
Appendix C: Supplemental Tables for Case Outcomes in Northern County ................................... 365
Appendix D: General Survey Instrument ......................................................................................... 372
Appendix E: Factorial Survey Instrument ........................................................................................ 382

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making xix

Part 1. Introduction and Research Questions
1.1 Introduction
In the American criminal justice system, discretion is perhaps broader, more often available, and
less constrained in the hands of the prosecuting attorney than in the hands of any other system actor
(see, e.g., Davis, 2008). While discretion plays an important role in other parts of the criminal
justice system – police discretion at arrest, judicial discretion at sentencing, parole board discretion
at release – prosecutors have not been subject to the same level of public and scientific scrutiny and
formal regulation as their law enforcement, judicial, or parole board colleagues. Moreover, how
prosecutors utilize their discretion, and what goes into prosecutorial decision-making, is little
understood outside of the community of prosecutors.
In recent years, a growing body of scholarship has examined those factors that affect case
outcomes throughout the prosecutorial process, from the initial screening decision to final sentence
recommendation (see, e.g., Free, 2002; Hartley, Maddam, & Spohn, 2007; Kingsnorth &
MacIntosh, 2004; Spohn and Holleran, 2001). Research has shown that outcomes are affected by
legal factors (e.g., strength of the evidence, type and seriousness of the offense, and defendant’s
culpability), quasi-legal factors (e.g., legally non-relevant though potentially influential factors,
such as defendant-victim relationship, victim age, and defendant age), and extra-legal factors (e.g.,
legally impermissible factors pertaining to defendant and victim, such as race, ethnicity, or gender).
While this literature has begun to illuminate variables that may impact prosecutorial decisionmaking, it is limited in three critical ways. First, prior studies tend to analyze a single decision point
– primarily the charging or screening decision – and are unable to determine the impact of different
factors across the prosecutorial process. Second, while studies have examined factors that relate to
characteristics of the defendant, offense, and victim, few have looked at the impact of prosecutor
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 1

characteristics or contextual characteristics on these decisions (see, e.g., Franklin, 2010). Finally,
much of the prior research has focused on quantitative examination of the factors affecting case
outcomes; but little attention has been devoted to the qualitative study of how and when prosecutors
weigh these factors.
This project expands what is known about prosecutorial decision-making in several ways. Using
data from two large urban/suburban county prosecutors’ offices – Southern County and Northern
County – the project examines case outcomes and prosecutors’ decision-making processes through
a sequence of mutually reinforcing qualitative and quantitative research approaches, including
administrative data analysis, surveys of prosecutors, and focus groups with prosecutors and
managers. The study goes beyond the existing literature by looking at a wide variety of individual
and organizational factors that might affect prosecutors’ decisions throughout the entire
prosecutorial process. Furthermore, it looks at how prosecutors weigh both legal and extra-legal
factors and when these factors enter into their decisions in the course of a case. Finally, the study
examines internal and external, formal and informal mechanisms that regulate prosecutors’ decision
making, including office policies, office and court resources, and relationships with other actors in
the criminal justice system. Such information is central to facilitating and expanding the principled
use of prosecutorial discretion and to identifying and intervening in conditions conducive to its
unprincipled use.

1.2 Review of Relevant Literature
1.2.1 Theoretical Perspectives on Prosecutorial Decision-Making
Law may be seen as a balance between formally rational law and substantively rational law.
Formally rational law occurs when courtroom decision making is based only on legally relevant
factors and consistent rules of action; in other words, outcomes under a formally rational system
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 2

“are primarily the result of legal rules and criteria applied equally” to all cases (Dixon, 1995, p. 61).
In contrast, substantively rational law occurs when courtroom decision making is based on factors
outside the law (e.g. defendant’s characteristics, needs, or circumstances) individually applied to
particular cases; moreover, substantive rationality is not guided solely by adherence to processes but
also by reference to “extralegal” goals and outcomes, such as social equality and justice or the
practical consequences of decisions for individuals (e.g. the defendant, the victim, etc.) and
organizations (e.g. the court, the prosecutor, etc.) (for a review see, e.g., Mears 1998).
Several theoretical perspectives seek to explain how substantively rational criteria interact with
formally rational criteria in criminal justice decision making. According to Albonetti (1991),
decision making reflects the use of bounded rationality, with courtroom actors making decisions
based on limited information about a defendant’s character or a particular case. This limited access
to information produces uncertainty that courtroom actors seek to minimize by engaging in
“uncertainty management” behavior (see, e.g., Ulmer et al. 2007). Albonetti (1991) combines the
uncertainty avoidance perspective with causal attribution, arguing that courtroom actors make
subjective attributions from stereotypes of defendant and case characteristics to reduce decisionmaking uncertainty, linking these characteristics to evaluations of the likelihood of future
criminality or the potential impact of sentences. In the end, according to Albonetti, courtroom
actors rely on substantively rational criteria to make decisions in order to reduce uncertainty.
Prosecutorial outcomes, thus, result from an interaction between the formal considerations of laws
and the substantive considerations of prosecutors about individual offender and case characteristics.
Steffensmeier and colleagues (Steffensmeier & Demuth, 2000; Steffensmeier, Ulmer, &
Kramer, 1998; see also Johnson, 2003; Johnson, 2005; Kramer & Ulmer, 2002; Ulmer, Kurlycheck,
& Kramer, 2007) argue that these substantive considerations revolve around three primary “focal

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 3

concerns:” blameworthiness of the offender, dangerousness of the offender, and practical
constraints and consequences of sentences for offenders and organizations. Courtroom actors then
relate their interpretations of these focal concerns to particular offender and case characteristics.
Similar to Albonetti’s (1986; 1987; 1991) uncertainty/attribution theory, courtroom actors make
decisions by making subjective determinations of blameworthiness, dangerousness, and the
consequences of sentences based on particular case/defendant characteristics. The focal concerns
perspective recognizes that courtroom actors’ decision-making begins with legal factors (e.g.
offense severity, defendant criminal history) as “benchmarks” for decisions but then incorporates
“situational attributions” about defendants’ character and risk based on case characteristics and
defendant characteristics (e.g. race, gender) (Steffensmeier & Demuth, 2000; Steffensmeier et al.,
1998).
While these attributions of offender/case characteristics and interpretations of focal concerns
may be idiosyncratic to particular courtroom actors, scholars often combine the focal concerns
perspective with a court communities perspective (Eisenstein et al. 1988; Eisenstien & Jacob 1977;
Johnson 2003; 2006; Ulmer 1997). The court communities perspective argues that decision making
is also the product of courtroom social contexts (Eisenstein, Flemming, and Nardulli, 1988;
Eisenstien & Jacob 1977; Johnson, 2003; Johnson, 2006; Ulmer 1997). According to this
perspective, a community is formed between regular courtroom workgroup members – judges,
prosecutors, defense attorneys, law enforcement, courtroom personnel (Eisenstein and Jacob, 1977).
Through regular interactions over a long period of time, this workgroup forms a set of
interdependent relationships and produces a local legal culture characterized by shared traditions,
values, and norms (Eisenstein et al., 1988). For example, day-to-day interactions produce a set of

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 4

shared expectations about the value or prioritization of cases, the proper resolution of cases, and
how other courtroom actors will behave in future interactions (Ulmer, 1997).
In response to uncertainty, these workgroups establish “going rates” and norms that determine
decisions in most cases and that make the decision-making process more predictable (Eisenstein &
Jacob, 1977; Eisenstein et al., 1988; Ulmer, 1997). The focal concerns and court communities
perspectives acknowledge that these going rates and norms are often developed to ensure efficiency;
indeed, courtroom actors operate under the need for or goal of organizational efficiency (Dixon,
1995; Engen & Steen, 2000). The need to dispose of cases, avoid court and case backlogs, and
conserve resources are goals often shared among courtroom actors that contribute to outcomes.
Although efficiency is an “extralegal” goal, it is generally seen as separate from other substantively
rational concerns and one that may supersede both formally rational rules and substantively rational
concerns of courtroom actors (Engen & Steen, 2000; Kramer & Ulmer, 2002). Moreover, members
of the courtroom workgroup generally find it in their professional interests to abide by the values
and norms of the court community or face informal sanctioning by other members of the workgroup
(Eisenstein & Jacob, 1977). Prosecutors also appear to have a “downstream orientation” that leads
them to anticipate and consider how other actors not yet involved in the process, such as judges and
juries, will respond to a case (Frohmann, 1997; Spohn and Holleran, 2001). Thus, rather than
efficiency, some argue that decision making is often governed by a need to maintain good
relationships with other courtroom actors to achieve desired outcomes (Eisenstein & Jacob, 1977;
Ulmer, 1997).
1.2.2 Empirical Analyses of Prosecutorial Outcomes
Research on prosecutorial discretion conducted over the last two decades has attempted to
identify factors influencing prosecutorial decision making. The bulk of the research in this area has
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 5

focused on unraveling how prosecutors balance and blend “legal” and “extra-legal” factors in
estimating convictability. Much of this research supports the uncertainty avoidance thesis and focal
concerns/court communities perspectives.
Studies have shown that prosecutors rely heavily on legal factors, including the type of offense
(Albonetti, 1987; Hartley et al., 2007; Jacoby, Mellon, Ratlidge, & Turner, 1982a; Schmidt &
Steury, 1989), strength of the evidence (Albonetti, 1987; Jacoby et al., 1982a; Spohn & Holleran,
2001; Miller & Wright, 2008), and defendant culpability (Adams & Cutshall, 1987; Albonetti,
1987; Schmidt & Steury, 1989). Moreover, some have argued that in more serious cases the
outcome is determined primarily by legal factors but that discretion, and the role of extra-legal
factors, plays a larger part in less serious cases (Spohn and Holleran, 2001; Hartley et al., 2007; in
contrast, Adams and Cutshall, 1987, argue that extra-legal factors matter less in less-serious cases).
Defendant demographic characteristics have received a significant amount of research attention,
particularly race, ethnicity, and gender. There is mixed evidence that race plays a role in
prosecutorial decision making. In a review of 24 studies of prosecutorial charging decisions and 19
studies of decisions by prosecutors to seek the death penalty, Free (2002) found that race clearly
affected the decision to seek the death penalty. However, evidence on the role of race in charging
was less clear; 15 of the 24 studies found no effect of race on charging decisions. More recently,
Ulmer et al. (2007) found that prosecutors were almost twice as likely to seek mandatory sentences
against Hispanic defendants as white defendants. Chen (2008) found that black defendants were
more likely to be charged with and receive third-strike sentences than white defendants, particularly
for offenses known as “wobblers,” which can be prosecuted either as a felony or a misdemeanor.
The few studies that look at the impact of defendant sex on the decisions of prosecutors agree that,
after controlling for other factors, women tend to be treated differently than men at a variety of

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 6

decision points. In a study of pre-trial diversion of drug offenders, Alozie and Johnston (2000)
found that women were more likely to be diverted than men. Albonetti (1986) found that
prosecutors were generally more likely to file charges against men than women. Finally, research
has shown that prosecutors generally charge men with more serious offenses than women for
similar conduct (Miethe, 1987) and are more likely to seek mandatory sentences against men than
similarly eligible women (Bjerk, 2005; Ulmer et al., 2007).
Characteristics of the victim and the victim-defendant relationship have also been shown to play
a part (Albonetti, 1986; Kingsnorth & MacIntosh, 2004; Schmidt & Steury, 1989; Spears & Spohn,
1997; Spohn & Holleran, 2001; Stanko, 1981-82). Most studies, particularly of sexual assault, have
found that prosecutors rely a combination of legal and extra-legal factors to make decisions. Spohn
and Holleran (2001), for example, found that prosecutors were more likely to file charges when
there was corroborating physical evidence, the defendant had a prior felony conviction, the victim
did not engage in risk taking behavior, and there were no questions about the victim’s moral
character. They also found that the victim-defendant relationship impacts decision making; in cases
involving acquaintances and intimate partners, prosecutors were less likely to file charges if there
were questions about victims’ character or behavior at the time of the incident. In cases involving
strangers, however, prosecutors were more likely to file charges if the suspect used a weapon or the
victim was white. Studies of domestic violence and sexual assault cases have also demonstrated the
important role of prosecutors’ perceptions of victim credibility on decision making (see e.g.,
Kingsnorth & MacIntosh, 2004; Spears and Spohn, 1997; Stanko, 1981-82). Schmidt and Steury
(1989) also found that in domestic violence cases a defendant’s current and past behavior, in
particular use of alcohol or drugs, were better predictors of whether charges would be filed than any
legal factors.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 7

Although, as a group, the studies investigate the effects of a wide variety of factors on
prosecutors’ decisions, two areas long suspected of being important – but that have remained
unstudied – are the effects of prosecutor characteristics, such as demographics and experience
(Spears & Spohn, 1997) and organizational constraints, such as caseloads and inter-agency
relationships (Stanko, 1981-82). A growing body of literature has begun to explore the impact of
presiding judges and county contextual effects on sentencing outcomes (see, e.g., Johnson, 2005;
Ulmer et al., 2007). Similar aspects of prosecutors and contextual factors also may be expected to
impact case outcomes during the prosecutorial process. Indeed, idiosyncratic evaluations of focal
concerns or attributions of case characteristics by prosecutors may lead to variation in case
outcomes across prosecutors. Similarly, organizational differences across offices or within offices
over time may similarly lead to variation and change in prosecutorial decision making.
The uncertainty/attribution theory, focal concerns perspective, and court communities model
imply that prosecutors primarily are concerned with convictability and efficiency; generally
overlooked in such discussions, however, is the issue of justice. According to the American Bar
Association’s General Standards for the Prosecution Function (American Bar Association, 1993,
Standard 3- 1.2(c)), “the duty of the prosecutor is to seek justice, not merely to convict.” Thus, one
may expect prosecutors to be motivated not just by uncertainty avoidance, focal concerns, or
convictability, but by a desire to achieve or ensure justice. The ABA standards, however, do not
define justice nor do they instruct the prosecutor in what factors to use in ensuring justice in their
decision making.
In evaluating criminal justice decision making, legal philosophers and social scientists have
generally differentiated between distributive justice and procedural justice (Rawls, 1999; Tyler,
2002). Distributive justice is focused on outcomes and whether the outcome of legal decision

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 8

making is equitable. In the case of prosecutorial decision making, distributive justice is achieved if
outcomes are consistent across social groups or across prosecutors. In contrast, procedural justice is
focused on processes and whether the procedures used in legal decision making are fair. In the case
of prosecutorial decision making, procedural justice is achieved if decision making processes are
consistently applied across social groups or across prosecutors. Researchers have identified several
attributes of decision making that contribute to perceptions of procedural justice; these include:
whether individuals involved in the case have an opportunity to state their case (“voice); whether
decision makers are unbiased, honest, and principled (“neutrality”); whether decision makers were
benevolent, caring, and consider the needs of individuals (“trustworthy”); and whether others
involved in the case were treated with dignity and respect (“respect”) (Tyler, 2003;Tyler & Hua,
2002).
Research has consistently found that the extent to which decision making processes are
perceived as fair shapes perceptions of the legitimacy of the legal authorities responsible for the
decision; in other words, individuals are more likely to perceive prosecutors as legitimate if those
individuals feel a sense of procedural justice (for a review, see Tyler, 2002). A dearth of research
exists, however, examining how prosecutors define justice or whether their decision making is
oriented toward ensuring distributive justice or procedural justice. Although the
uncertainty/attribution theory, focal concerns perspective, and court communities provide necessary
insights into how legal and extra-legal factors may influence outcomes, it may also be necessary to
understand how prosecutors define and operationalize justice within these contexts and how
prosecutors orient decision making toward ensuring distributive and procedural justice.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 9

1.2.3 The Limitations of Prior Research
Despite the welcome growth in research on prosecutorial decision making, this work suffers
from a number of limitations relating to generalizability and research design and analysis. Results
from prior studies are not typically generalizable. To date, research overwhelmingly has examined
just one jurisdiction, just one decision point (e.g., whether to prosecute a case and what to charge),
or just one offense type. As a consequence, results cannot be applied more generally across
jurisdictions, decision points, or offenses. Spohn and Holleran (2001) are the only researchers of
whom we are aware to have focused on more than one jurisdiction in their analysis of the
prosecution of sexual assault cases. Similarly narrow in focus, prior studies have mainly examined
the initial decision whether to prosecute a case, and if so, what charges to file. No studies, of which
we are aware, have looked at whether factors affecting decision making differ at different stages or
decision points in the prosecutorial process (e.g., dismissal of charges, plea offers, sentence
recommendations) or whether there is a cumulative effect of different factors as cases advance
toward their conclusion. Furthermore, although there are several studies on domestic violence and
sexual assault (see, e.g., Kingsnorth & MacIntosh, 2004; Schmidt & Steury, 1989; Spears & Spohn,
1997), there are no recent studies that look at factors affecting decisions for multiple offense types
or severity levels.
Prior research has also been limited with respect to research designs and analyses. First, few
studies adopt a comparative design, whether across time or place. There are just two exceptions,
Miethe (1987) and Spohn and Holleran (2001). Miethe gauges the effect of sentencing guidelines on
prosecutorial discretion in Minnesota by comparing cases from 1978 (pre-guidelines) with cases
from 1980 and 1982 (post-guidelines). Spohn and Holleran’s examination of prosecutorial
discretion is one of the only studies to use data from more than one site (Kansas City, Missouri and

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 10

Philadelphia, Pennsylvania); however, surprisingly, they aggregate data from the two sites,
obviating cross site comparisons. Second, most prior research in this area is quantitative, using
administrative data to examine what factors affect decision making, but do not consider how and
when prosecutors weigh these factors. We are aware of just two qualitative studies. Stanko (198182) observed the felony-arrest screening process in New York County (Manhattan), New York, and
Frohmann (1997) conducted an ethnographic field study of a sexual assault unit in an unnamed
prosecutor’s office on the West Coast. Although methodologically groundbreaking, both studies
are limited, relying on anecdotal and non-rigorous analytical techniques (Stanko, 1981-82) or
focusing on just one unit that handles just one offense type (Frohmann, 1997).
Despite the gaps and weaknesses just discussed, the existing research has been instrumental in
showing that prosecutors take multiple factors into account, both legal and extra-legal, when
making case-processing decisions. The current study builds on this tradition by addressing some of
its deficiencies. The project (1) adopts a dual-site, comparative design, (2) incorporates multiple
offense types and decision points, (3) includes factors relating to prosecutor characteristics and
organizational constraints, and (4) collects and analyzes data from a variety of sources using an
integrated set of methodologically rigorous quantitative and qualitative techniques. These
components are specifically designed and integrated to bolster the study’s comprehensiveness,
validity, and, ultimately, utility to policymakers and practitioners.

1.3 Research Questions
The study is driven by several research questions. First, there is considerable evidence that both
legal and extralegal factors affect prosecutorial case outcomes. There is little research, however,
into the effect of those factors at different stages of the prosecutorial process or the effect of
individual prosecutors on case outcomes. The study examines the following research question: What
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 11

factors influence case outcomes? Specifically, using administrative data on case outcomes, the
study examines how defendant, victim, offense, case, and prosecutor characteristics affect the
decision to accept or reject a case for prosecution, the number and level of charges to file, the
amendment or dismissal of charges after filing, the number and level of charges offered during plea
negotiations, or the recommendation of an incarceration or non-incarceration sentence at trial.
Using a factorial survey design, the study further explores these questions: How does decision
making vary within a prosecutor’s office for a similar set of cases? How do factors such as strength
of the evidence and defendant criminal history affect case outcomes? Do evaluations of these case
factors vary by prosecutor characteristics?
Second, although prior research has examined the factors that predict case outcomes, little of
that research has explored how prosecutors evaluate these factors or weigh them in making
decisions. Indeed, the analyses of case outcomes provide only a partial glimpse of decision making;
specifically, it does not provide any information about why particular factors affect outcomes.
Understanding these issues requires a different methodological approach that considers how
prosecutors make decisions and how formal and informal mechanisms impinge or control
prosecutorial decisions. The study is motivated by the following research question: How do
prosecutors evaluate and weigh the different factors affecting a case? Specifically, using interviews
and surveys with prosecutors, the study examines how prosecutors interpret and use defendant,
victim, offense, and case characteristics in making decisions and how do they balance
organizational needs for efficiency and resource management with the maintenance of inter-agency
relationships and the just outcome of cases. Building on this analysis, the study further explores
these questions: How do prosecutors’ offices regulate prosecutorial decision making? How do

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 12

office structures, policies, and practices regulate decision making? What formal and informal,
internal and external controls are placed on prosecutors when making decisions?

1.4 Site Selection
The study relies on data from two county prosecutors’ offices – Northern County and Southern
County. The two offices provide ideal sites for examining issues of prosecutorial decision-making.
On the one hand, both sites are similar in several important ways that facilitate cross-site
comparisons: medium-sized offices serving urban/suburban populations and handling large numbers
and varied types of cases. On the other hand, the sites differ in two key ways that facilitate crosssite contrasts: organizational structures and operational approaches.
Both Northern County and Southern County have populations of just less than 1 million people,
with one large central urban center and several surrounding suburban municipalities (Table 1.4-1).
While Northern County witnessed a flat population growth over the last decade (increasing just 0.8
percent between 2000 and 2010), Southern County experienced rapid population growth, growing
32 percent over the last decade and making it one of the fastest growing urban regions in the
country. The two counties are very similar demographically; roughly 50 percent of the general
populations in both counties are white, 30 percent are African American, and 12 percent are
Hispanic. The defendants prosecuted in each county are also similar demographically – although
roughly 42 percent of the general population is non-white, approximately 66 percent of the
defendant population in each county is non-white.
The minority populations in both counties are largely concentrated in the central urban areas,
while the non-minority populations are largely concentrated in the suburban municipalities.
Between 2000 and 2010, both counties also saw increases in the proportion of African American
and Hispanic residents in the population and decreases in the proportion of white residents. These
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 13

fluctuations were relatively small in Northern County; in contrast, in Southern County the relative
proportion of white residents in the population decreased 10 percentage points (from 61 percent to
51 percent) as the proportion of Hispanic residents increased 5 percentage points (from 7 percent to
12 percent) and the proportion of African American residents increased 3 percentage points (from
28 percent to 31 percent).
Table 1.4-1 Characteristics of the population served, by research site

Selected Characteristics

Northern County

Southern County

2000

2010

2000

2010

900,000

950,000

700,000

900,000

% white, non-Hispanic

58%

54%

61%

51%

% black, non-Hispanic

26%

27%

28%

31%

% Hispanic, any race

11%

13%

12%

$40,500

$43,000

$64,000

$55,000

% of housing owner occupied

53%

51%

62%

61%

% of population below poverty

21%

19%

11%

13%

Approximate total population

Median household income (2010$)

The median household income is significantly higher in Southern County relative to Northern
County ($55,000 versus $43,000 in 2010). Northern County, however, experienced an increase in
median household income between 2000 and 2010, while Southern County experienced a decrease.
Moreover, while Northern County saw a 2 percentage point decrease in the poverty rate between
2000 and 2010, Southern County experienced a 2 percentage point increase.
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 14

The sentencing and corrections systems are similar in each research site as well. Both
jurisdictions operate within states that have abolished discretionary parole release from prison; yet
both states have maintained some form of mandatory supervision after release. Both jurisdictions
also operate within states with some form of sentencing guidelines. The guidelines under which
Northern County operates are advisory (meaning that judges are not required to follow the sentence
recommendations in the guidelines and neither the prosecutor or defense can appeal sentences that
do not adhere to the guidelines); guidelines sentences are determined according to seriousness of
offense and a subjective assessment of several factors related to future risk of re-offending. In
contrast, the guidelines under which Southern County operates are presumptive (meaning that
judges are required to follow the sentence recommendations in the guidelines and both the
prosecutor and defense can appeal sentences that do not adhere to the guidelines); guidelines
sentences are determined according to seriousness of the offense and prior criminal history of the
defendant.
The Northern County prosecutor’s office employs approximately 125 Assistant District
Attorneys (ADAs) who handle roughly 30,000 felony and misdemeanor cases per year (Table 1.42). The office is organized into a series of eighteen specialized units that handle specific offense
types (e.g. homicide, domestic violence, felony drug, guns) and five general crimes units that handle
all felony and misdemeanor cases not handled by specialized units. All new ADAs in Northern
County are assigned to one of the five general crimes unit comprised of both new and experienced
ADAs; ADAs may remain in a general crimes unit for their entire careers. All ADAs are
responsible for screening cases within their unit; cases accepted for prosecution are then assigned to
specific ADAs and prosecuted vertically (i.e. a single ADA handles the case throughout the entire
prosecutorial process). The office is structured along a three-tiered system of management, with

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 15

ADAs reporting to twenty-three unit managers who are supervised by five deputy prosecutors who,
in turn, report to the District Attorney. The District Attorney in Northern County was first elected
within the last ten years and has implemented innovative prosecution models, such as creating
community-prosecution units, organizing units around geographic areas, and instituting programs
based on restorative-justice models.
The Southern County prosecutor’s office employs roughly 75 ADAs who handle approximately
13,500 felony and misdemeanor cases per year (Table 1.4-2). The office is organized around seven
specialized felony units that handle broad categories of offense types (e.g. property, person, drugs)
and one misdemeanor unit that handles all misdemeanor and criminal traffic cases. All new ADAs
in Southern County are assigned to the misdemeanor unit which is comprised solely of new ADAs;
ADAs are then transferred to another unit, usually the drug unit, after nine to eighteen months.
Experienced ADAs are responsible for screening cases; cases accepted for prosecution are then
assigned to specific ADAs within units and prosecuted vertically, with the exception of felony drug
offenses which are prosecuted horizontally (i.e. cases are handled by multiple ADAs, each handling
the case at one stage of the prosecutorial process). The Southern County prosecutor’s office is a flat
system, with ADAs reporting to unit managers who report directly to the District Attorney; two
deputy prosecutors in Southern County function as office managers, but do not act as intermediaries
between unit managers and the District Attorney. At the time of the study, the District Attorney in
Southern County had retained the office for more than two decades and followed a fairly traditional
prosecution model.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 16

Table 1.4-2 Selected characteristics of participating prosecutors’ offices
Characteristics

Southern County

Northern County

Typical number of ADAs

125

Approximate number of
criminal cases per year

13,500

30,000

Office organization

 7 felony units, specialized
by crime type
 1 misdemeanor unit

 18 units that handle both
felonies and misdemeanors,
specialized by crime type
 5 general crimes units that
handle all other felony and
misdemeanor cases

Managerial structure

ADAs report to 8 unit heads
who report to the DA

ADAs report to 23 unit heads,
who are supervised by 5
deputies who report to the DA

Vertical or horizontal
prosecution

Horizontal for felony drug
cases; vertical for other cases
after initial screening

Vertical after initial screening

Strong orientation toward
diversion programs and
community prosecution?
Tenure of the DA (at start of
project)

Yes

30 years

2 years

We derived a limited amount of demographic information about prosecutors in each jurisdiction
from a general survey of prosecutors, in which prosecutors were asked their age, gender,
race/ethnicity, and level of experience; some descriptive statistics were also available in
administrative data maintained by the Northern County District Attorney’s office. Although we had
limited demographic information about ADAs, the few characteristics we were able to capture
indicate that the ADAs in the two research sites are quite different (Table 1.4-3). Although
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 17

prosecutors in both jurisdictions are similar in terms of gender (roughly 56 percent of ADAs in each
jurisdiction are male), prosecutors in Northern County tend to be older and have more experience
than prosecutors in Southern County. For example, roughly 46 percent of ADAs in Northern
County were 40 years of age or older, compared to just 19 percent of ADAs in Southern County. In
addition, while roughly 40 percent of ADAs in Northern County have 10 or more years experience
as a prosecutor, just 9 percent of ADAs in Southern County have a similar level of experience.
Table 1.4-3 Characteristics of prosecuting attorneys, by jurisdiction
Northern County
Southern
Characteristic
County Survey
Administrative
Survey Dataa
Dataa
Datab
Number of prosecutors
65
62
145
responding
Percentage male

56%

71%

56%

Percentage nonwhite or Hispanic

17%

10%

Age distribution
Less than 30 years old
30 – 39 years old
40 years old or older

26%
55%
19%

15%
39%
46%

Distribution of experience in
present DA’s office
Less than 1 year
1 – 9 years
10 or more years

11%
80%
9%

13%
49%
38%

14%
46%
40%

Notes:

From responses to the general survey for ADAs assigned to adult felony cases. The response rate for that
group was 95 percent in Southern County and 67% in Northern County.
b
From administrative data for ADAs who screened cases between January 2009 and June 2011. Comparison
with the survey data suggests that the survey responses were biased toward male respondents in Northern
County.

1.5 Methods Overview
This study used a multi-method approach, relying on both quantitative and qualitative methods
to examine prosecutorial decision making in Northern and Southern Counties (Figure 1.5-1). To
examine actual case outcomes and the factors associated with those outcomes, the study relied on
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 18

analyses of administrative data derived from the case management systems in each office and a
review of case files in Southern County. Using logistic regression and hierarchical linear modeling,
the analyses examined the impact of defendant, offense, victim, and prosecutor characteristics on
outcomes at several decision points – screening, charging, dismissal/amendment, and plea offer.
The study further explored case outcomes using a factorial survey containing a series of
hypothetical cases in which prosecutors were asked to make and justify screening, charging, and
plea offer decisions; the analyses considered the impact of evidence strength, offense severity,
defendant criminal history, and prosecutor characteristics on outcomes and provided an opportunity
to examine consistency in evaluations across prosecutors.
The analyses of actual and hypothetical case outcomes were complemented by an attitudinal
survey of prosecutors and series of interviews/focus groups with prosecutors. Prosecutors in both
jurisdictions were invited to respond to a structured survey, designed to elicit general attitudes about
definitions of individual and organizational success; the influence of relationships among
prosecutors, police, defense attorneys, and judges; resource and policy constraints; principles that
guide screening decisions and plea offers; the general goals and functions of the criminal justice
system; and internal training and oversight. Two waves of interviews with the District Attorney
(DA) and Deputy District Attorneys (Deputies) in each site and two waves of focus group sessions
with line prosecutors and unit managers also were conducted in each of the research sites. The first
wave of interviews/focus groups focused primarily on contextual conditions and circumstances that
influence decision making (e.g., guiding philosophies, policies, relationships with other system
actors and colleagues, resource constraints). The second wave of interviews/focus groups focused
primarily on case-specific factors that influence decision making (e.g., strength of evidence,

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 19

seriousness of the instant offense, defendant’s criminal history, and special aggravating/mitigating
circumstances).
Rather than provide a full methods section at the beginning of the report or separate detailed
methods sections within each chapter, we include relatively brief methods sections detailing the
approaches used in each chapter and detailed methods for the entire study in Appendix A.
Figure 1.5-1 Overview of Research Methods

The arrows in the diagram depict the influence of each project component on the design and
implementation of subsequent components. Administrative data analysis began before Wave 2
and continued afterward, so it both influenced and was influenced by the Wave 2 focus groups.

1.6 Report Overview
This study documents prosecutorial decision making in two county prosecutors’ offices,
examining legal and extra-legal factors influencing decision making. Our observations and
recommendations are based on the analysis of administrative data tracking case outcomes, focus
groups with prosecutors and supervisors, an attitudinal survey of prosecutors, and a factorial survey
using case vignettes. Drawing on these data, Part 2 first examines prosecutors’ perspectives of the

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 20

factors influencing decision making – prosecutorial philosophy, case-specific factors, contextual
constraints – and explores how prosecutors balance these varied influences at different stages of the
prosecutorial process. Part 3 then examines actual case outcomes and assesses the extent to which
different legal and extra-legal factors predict outcomes at different stages of the process. Part 4
discusses the policy implications of the findings. Parts 2 and 3 provide short descriptions of the
methods used in each section; detailed appendices provide information on the methods, additional
descriptive analyses of the data, and instruments used in the study.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 21

Part 2. The Prosecutor’s Perspective
2.1 Introduction
Much of the prior research on prosecutorial decision making has been devoted to examining the
influence of case characteristics on outcomes, exploring those legal and extra-legal case-specific
characteristics (e.g. defendant/victim demographics, offense severity, strength of the evidence) that
determine whether a case is declined or prosecuted, the level of charges ultimately filed, or the
sentence recommended upon conviction. Less research, however, has explored how prosecutors
weigh these case characteristics in making decisions or what contextual factors may influence how
prosecutors evaluate such characteristics.
This study employed a combination of qualitative and quantitative analyses to examine how
prosecutors evaluate and weigh the different factors affecting a case and how prosecutors’ offices
regulate prosecutorial decision making. Specifically, this chapter seeks to answer several questions:
What do prosecutors see as the primary goals of prosecution? How do prosecutors evaluate strength
of the evidence and how do these evaluations change throughout the life of a case? How do
prosecutors balance organizational needs for efficiency and resource management with the
maintenance of inter-agency relationships and the just outcome of cases? How do office structures,
policies, and practices regulate decision making? What formal and informal internal and external
controls guide prosecutors when making decisions about cases?

2.2 Summary of Research Methods: Focus Groups and General Survey
To answer these questions, the study relied on focus group interviews with prosecutors and
supervisors and a general survey of prosecutors in the participating jurisdictions (for a detailed
description of methods, see Appendix A).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 22

2.2.1 Focus Groups
Two waves of interviews and focus group sessions were conducted in each of the research sites.
Individual interviews were conducted with the District Attorney (DA) and Deputy District
Attorneys (Deputies) in each site; focus group interviews were conducted with ADAs and unit
managers. Focus group participants were divided by years of experience as a prosecutor (less than
one year experience, one to ten years experience, unit managers) and efforts were made to interview
the same prosecutors during each wave of focus groups. In Southern County, focus groups included
six prosecutors with less than one year experience, eight prosecutors with one to ten years
experience, and seven unit managers. In Northern County, there were not enough participants to
divide focus groups by years of experience; instead, focus groups included five prosecutors
responsible for different types of cases (general crimes, domestic violence, drugs, weapons) and
seven unit managers.
The first wave of interviews and focus groups focused primarily on contextual conditions and
circumstances that influence decision making (e.g., guiding philosophies, policies, relationships
with other system actors and colleagues, resource constraints). The second wave of interviews and
focus groups focused primarily on case-specific factors that influence decision making (e.g.,
strength of evidence, seriousness of the instant offense, defendant’s criminal history, and special
aggravating/mitigating circumstances).
The feedback received from prosecutors was recorded as field notes and analyzed across topics
and sites, noting the clustering of responses around specific issues or actors, as well as outliers and
other unique data. Since interviews/focus groups were not recorded, the discussion of interview and
focus group responses below is not able to produce exact quotes in all instances; thus, phrases
appearing in italics are partial or paraphrased quotes derived from interviewer notes.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 23

2.2.2 General Survey
Prosecutors in both jurisdictions were invited to respond to a structured survey, designed to
elicit general attitudes in eight substantive areas: 1) factors that define individual success; 2) factors
that define organizational success; 3) the influence of relationships among prosecutors, police,
defense attorneys, and judges; 4) resource and policy constraints; 5) principles that guide screening
decisions; 6) principles that guide the development of plea offers; 7) general goals and functions of
the criminal justice system; and 8) training and oversight. The survey instrument was accompanied
by a background questionnaire that captured respondent age, race, ethnicity, gender, and years of
experience as a defense attorney and prosecutor. Copies of the complete survey instrument, the
instructions to respondents, and the background questionnaire are included in Appendix D.
In Southern County, survey responses were received from 74 respondents from a pool of 78
prosecutors (95 percent response rate). Excluding the district attorney, the deputy district attorneys,
and juvenile court prosecutors, the sample analyzed for this report included 65 respondents from a
pool of 69 prosecutors (93 percent response rate). In Northern County, responses were received
from 81 prosecutors from a pool of 135 prosecutors (60 percent response rate). Excluding the
district attorney, the deputy district attorneys, and juvenile court prosecutors, the sample analyzed
for this report included 62 respondents from a pool of 93 prosecutors (67 percent response rate).
Preliminary analyses found that respondents tended to cluster their ratings at the upper or lower
end of the scale for most items which produced ratings with restricted variability. To examine the
consequences of these response biases, standardized responses (z-scores) were created for each
item, relative to the personal means and standard deviations of each respondent’s ratings for items
within each of the item categories listed above. The resulting z-scores were then grouped to create a
5-level standardized scale. For ease of interpretation, this report focuses primarily on the original

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 24

scaling of responses as structured in the survey instrument, but de-emphasizes items for which the
results of analyses differ according to which scaling is adopted. Finally, to reduce the amount of
detail in the presentation of results, principal components analyses were conducted to determine the
number of underlying dimensions of response for each category of items and then to identify the
one to three specific items that most strongly represented the underlying dimensions within each
category. The items selected for illustrative purposes are ones that (a) yield consistent patterns of
results using either the original scaling or the standardized scaling of responses and (b) are among
the items found to be most representative of a relevant underlying dimension within one of the eight
categories listed above. Table B34 in Appendix B identifies the underlying dimensions in each
category and lists the specific items most representative of each dimension.

2.3 The Goals of Prosecution: Justice, Consistency, and Efficiency
Prosecutors in both jurisdictions maintained that they were provided few specific guidelines or
rules for how to handle cases. Indeed, prosecutors at all levels recognized not only the power that
the office of the district attorney holds within the criminal justice system, but also the wide
discretion that they as individual prosecutors exercise in individual cases. Nonetheless, prosecutors
in both jurisdictions argued that their discretion was not unguided. Rather, prosecutors maintained
that the DAs in both sites articulated three primary goals of prosecution that governed decision
making in all cases: justice, consistency, and efficiency.
2.3.1 Justice
In both jurisdictions, the DA set out a simple philosophy for prosecuting cases that unit
managers and ADAs clearly understood – do justice.1 The DA in Southern County maintained that
he wanted people to do the right thing. This guided not only decisions on outcomes but also
1

Throughout the report, statements made by prosecutors in the individual interviews and focus group sessions are
presented in italics to indicate that the statements have been paraphrased rather than quoted verbatim.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 25

interactions with police officers, defense attorneys, judges, and defendants. The DA expected ADAs
to exercise complete candor when discussing cases with other system actors and instructed ADAs
that part of their job is to set the right tone. In the end, the DA argued that the critical part of the
prosecutor’s job was about ethics and how you treat people. For the DA, that meant that
prosecutors should try to do justice, be open-minded, and treat people fairly and with respect.
Indeed, as discussed in Section 2.5 below, many ADAs in Southern County said that this
philosophy of justice, openness, and respect governed their interactions and led to better outcomes
for the office.
The DA in Northern County communicated a similar philosophy, stating that prosecutors were
told to do the right thing, to do justice, to help the community. For the DA in Northern County,
doing the right thing consisted of two things: keeping the community safe and protecting the
constitutional rights of defendants. This articulation of philosophy as an overarching desire to do
justice shaped how the DA viewed success; success was not necessarily about winning cases and
getting convictions, but about protecting constitutional rights of citizens and safety. In some
instances, it also acted as a constraint, forcing prosecutors to do what is appropriate, not everything
that you can. As we discuss in Part 2.4, it also set up a fundamental question with which prosecutors
must contend: even if the defendant is guilty, is a conviction the right outcome? The DA in
Northern County recognized this tension and the difficulty in communicating how to balance
ensuring justice for the defendant with ensuring justice for the victim and the community.
Indeed, with a philosophy of doing justice as the primary guideline for handling cases, the
interpretation of “doing justice” in each case is left to individual ADAs, which, as the DA in
Southern County acknowledged, can vary from person to person. To gain some consistency in the
meaning of justice, each office relied heavily on unit managers and office peer pressure to

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 26

communicate and ensure adherence to the philosophy. As one unit manager in Northern County
noted, the office seeks to hire good people and teach them certain virtues. The unit approach in
each office also allowed for direct supervision of a small number of ADAs by one supervisor and
for the routine communication among unit members about the just or fair decision in individual
cases. This did not, however, always lead to the same evaluation of justice since, as one unit
manager acknowledged, a supervisor’s idea of “do the right thing” may be different than an ADAs.
In such instances, most prosecutors in both sites agreed that deference was often given to the
individual ADA’s definition of justice, noting that as long as ADAs can explain how they got to a
decision, that is acceptable. This response underscores the power of the individual ADA – with a
strong acceptance of office philosophy and close supervision, the individual ADA’s perception of
justice or decision on an individual case is allowed to prevail even when it conflicts with a
supervisor’s perception of justice or of the right decision in a particular case. In both offices,
respondents noted that the office only functions if the DA and supervisors trust ADAs to make the
right decisions.
This puts a lot of pressure on ADAs to make the “right” decision. Absent more explicit guidance
at the office level, prosecutors will necessarily pursue objectives that reflect their personal and
collective beliefs concerning the fundamental purposes of the criminal justice system and the
appropriate role of the prosecutor in serving those purposes. Often, the office philosophy conflicts
with the general perceptions of the role of the prosecutor, which some see as ensuring tough
sanctions for violations of the law. Several items from the general survey help to better understand
personal beliefs that may influence individual perceptions of justice. What is most striking about the
survey results is the amount of variation in responses; while focus group responses indicated a

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 27

consensus on the overarching philosophy of the office, survey responses indicated much more
divergence in opinions about the goals of prosecution.
Perceptions of the goals of the criminal justice system
Responses to items pertaining to the goals of the criminal justice system varied along three
underlying dimensions: the most important functions of the system, punishment orientation, and the
rights and rehabilitation of defendants. Overall, a majority of respondents (58 percent) agreed or
strongly agreed that the most important function of the criminal justice system is to prevent and
repress crime, but 22 percent disagreed or strongly disagreed that this was the most important
function (Figure 2.3.1-1). The tendency to emphasize a crime prevention function varied
systematically by jurisdiction and level of experience. The average level of agreement that crime
prevention is the most important function was marginally higher in Northern County than in
Southern County (p = .055, F = 3.760, df = 1). Controlling for differences between jurisdictions, 2
agreement that crime prevention is the most important function declined significantly with
increasing years of experience (p = .046, F = 2.823, df = 3), with more experienced ADAs
disagreeing that crime prevention was the most important function (Figure 2.3.1-2).

Because there were significant differences in response patterns between jurisdictions for some items, and there also
was a significant difference between jurisdictions in average level of experience, analyses of the differences in
responses across levels of experience controlled for differences in responses between jurisdictions.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 28

Figure 2.3.1-1 Responses to the question “The most important function of the criminal justice
system is to prevent and repress crime,” by research site
Percentage of Respondents within County

60.0
55.0
50.0
45.0

40.0
35.0
30.0
25.0
20.0
15.0
10.0
5.0
0.0
Strongly disagree

Disagree

Neutral

Northern County (n=62)

Agree

Strongly agree

Southern County (n=65)

Figure 2.3.1-2 Responses to question “The most important role of the criminal justice system
is to prevent and repress crime,” by research site and ADA experience

Mean Level of Agreement

5
4
3
2
1
0
Less than 1 year

At least 1 year but less At least 5 years but less
than 5 years
than 10 years

10 years or more

ADA years of experience in current DAs office
Northern County

Southern County

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 29

Overall, prosecutors were evenly divided as to whether the most important function of the
criminal justice system is protecting the rights of the accused: 32 percent agreed or strongly agreed
and 32 percent disagreed or strongly disagreed (Figure 2.3.1-3). Prosecutors in Northern County
were more likely than prosecutors in Southern County (39 percent vs. 26 percent) to agree or
strongly agree that the most important function is protecting the rights of the accused, and the
difference in average ratings of agreement was highly statistically significant (p = .009, F = 7.13, df
= 1). Controlling for differences between jurisdictions, the average ratings did not vary significantly
across levels of experience. Thus, prosecutors within the same offices hold very conflicting views
of the criminal justice system’s function. Many agree that the system functions primarily to prevent
crime, while a very large group also agrees that the system functions primarily to protect the rights
of the accused. Moreover, as prosecutors gain experience, they believe less in the crime prevention
function of the system.
Figure 2.3.1-3 Responses to the question “The most important function of the criminal justice
system is protecting the rights of the accused,” by research site
Percentage of Respondents within County

60.0
55.0
50.0
45.0
40.0
35.0
30.0
25.0
20.0
15.0
10.0
5.0

0.0
Strongly disagree

Disagree

Northern County (N=62)

Neutral

Agree

Strongly agree

Southern County (N=65)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 30

Prosecutors also expressed surprisingly non-punitive orientations. For both offices, only 18
percent of prosecutors agreed that sanctioning offenders should involve punishment rather than
rehabilitation, while 39 percent disagreed (Figure 2.3.1-4). The responses in Southern County
reflected a significantly stronger punishment orientation than the responses in Northern County. In
Southern County, 23 percent agreed that sanctioning should involve punishment rather than
rehabilitation and 25 percent disagreed; in Northern County, only 13 percent agreed that sanctioning
should involve punishment rather than rehabilitation and 54 percent disagreed (p = .001, F =
12.106, df = 1 for difference in average ratings).
Figure 2.3.1-4 Responses to the question “Sanctioning offenders should involve punishment
rather than rehabilitation,” by research site
Percentage of Respondents within County

60.0
50.0
40.0
30.0
20.0
10.0
0.0
Strongly disagree

Disagree

Northern County (N=61)

Neutral

Agree

Strongly agree

Southern County (N=65)

In addition, respondents in Southern County were significantly more likely than respondents in
Northern County (63 percent vs. 39 percent) to agree that many community-based programs do not
provide sufficient punishment for offenders (p=.002, F=10.129, df=1 for difference in average
ratings) (Figure 2.3.1-5). Finally, prosecutors’ opinions were divided as to whether many offenders
currently imprisoned could be adequately handled in non-prison sanctions; 26 percent agreed that
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 31

offenders could be handled in non-prison sanctions, but 44 percent disagreed (Figure 2.3.1-6).
Controlling for differences in responses between jurisdictions, the average level of agreement with
the statement “Offenders do not need to be punished in order to be rehabilitated,” did not differ
significantly across levels of experience; the average level of agreement initially decreased with
increasing experience, but then increased sharply for those with 10 years experience or more (p =
.024, F = 3.265, df = 3) (Figure 2.3.1-7).3
Figure 2.3.1-5 Responses to the question “Many community-based programs do not provide
sufficient punishment for offenders,” by research site
Pecentage of Respondents within County

60.0
50.0
40.0
30.0

20.0
10.0
0.0
Strongly disagree

Disagree

Northern County (N=61)

Neutral

Agree

Strongly agree

Southern County (N=67)

The group of respondents with 10 or more years of experience yielded results at odds with the patterns evident across
the other levels of experience for a number of the survey items. These inconsistencies are only partially accounted for
by differences in levels of experience between jurisdictions. It could not be determined from the limited data available
on prosecutor characteristics how this group might differ systematically from those with less experience with respect to
other characteristics. While it is reasonable to speculate that this group might include some or all of the unit managers,
that could not be confirmed from the data available for this study.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 32

Figure 2.3.1-6 Responses to the question “Many offenders currently imprisoned could be
adequately handled in non-prison sanctions,” by research site
Percentage of Respondents within County

60.0
50.0

40.0
30.0
20.0
10.0
0.0
Strongly disagree

Disagree

Northern County (N=61)

Neutral

Agree

Strongly agree

Southern County (N=65)

Figure 2.3.1-7 Responses to the question: “Offenders do not need to be punished in order to be
rehabilitated,” by research site and ADA experience

Mean Level of Agreement

5
4
3
2
1
0
Less than 1 year

At least 1 year but <5
years

At least 5 years but <10
years

10 years or more

ADA years of experience in current DAs office
Northern County

Southern County

The most striking pattern emerging from the analyses of items relating to criminal justice
system goals is the considerable divergence of opinion regarding the system’s most important
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 33

functions, the importance of punishment, and the extent to which the system responds appropriately
to the rights and rehabilitation potential of criminal defendants. It is reasonable to ask whether this
wide variation in a few basic beliefs might be accompanied by similar variation in beliefs and
attitudes that were not measured, and whether such variation translates into unwarranted differences
in the operational objectives pursued by different prosecutors and different prosecution teams. The
general survey explored convergence and divergence of these more concrete objectives by asking
prosecutors about the factors that define personal success and success for the office as a whole.
Perceptions of individual success
Respondents rated 15 potential outcomes as to their importance for defining personal success as
a prosecutor. Responses varied along four underlying dimensions: accuracy in charging and
diverting cases; respect and relationships (in working with colleagues, supervisors, police, defense
attorneys, and judges); obtaining convictions and guilty pleas; and fairness. Responses to items
reflecting respect and relationships are discussed later in a section focusing specifically on working
relationships. Items representing the remaining three dimensions are discussed below.
The general survey explored prosecutors’ perceptions of three potential performance indicators
that could be used to gauge accuracy in charging and diverting cases: dismissal rates,
deferral/diversion rates, and successful diversions. Low dismissal rates after charges are filed were
considered important or very important by 27 percent of respondents, but were considered of little
importance or unimportant by 44 percent of respondents. Ratings of the importance of low dismissal
rates, however, differed dramatically by jurisdiction (p = .000, F = 49.751, df = 1) (Figure 2.3.1-8).
Low dismissal rates were rated important or very important by 53 percent of respondents in
Northern County but only 1.5 percent (1 respondent) in Southern County. Controlling for
differences between jurisdictions, ratings of the importance of low dismissal rates also differed

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 34

significantly by level of experience (p = .039, F = 2.870, df = 3) (Figure 2.3.1-9). Average ratings
declined with increasing levels of experience, but increased for respondents with 10 or more years
of experience (see footnote 3 above). In other words, low dismissal rates were more important for
both the least experienced and most experienced prosecutors.

Percentage of Respondents within County

Figure 2.3.1-8 Responses to the question: “Low dismissal rates after charges are filed is
important for individual success,” by research site
60.0
55.0
50.0
45.0
40.0
35.0
30.0

25.0
20.0
15.0
10.0
5.0
0.0
Unimportant

Of little importance

Northern County (N=62)

Moderately
important

Important

Very important

Southern County (N= 64)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 35

Figure 2.3.1-9 Responses to the question: “Low dismissal rates after charges are filed is
important for individual success,” by research site and ADA experience

Mean Importance Rating

5
4
3
2
1
0
Less than 1 year

At least 1 year but <5
years

AT least 5 years but <10
years

10 years or more

ADA years of experience in current DAs office
Northern County

Southern County

Prosecutors had similar perceptions of deferral/diversion rates for eligible defendants. High
deferral/diversion rates were rated important or very important by 23 percent of respondents, but
were rated of little importance or unimportant by 37 percent of respondents. In addition, prosecutors
were evenly divided in their ratings of the importance of high success/completion rates for
defendants who were deferred or diverted, with 39 percent considering it important or very
important and 41 percent considering it of little importance or unimportant. There were significant
differences between jurisdictions in prosecutors’ average ratings of the importance of both high
deferral/diversion rates (p = .002, F = 10.122, df = 1) and high success/completion rates (p = .001, F
= 11.515, df = 1). High deferral/diversion rates were considered important by 33 percent of
respondents in Northern County but only 14 percent of respondents in Southern County (Figure
2.3.1-10). High success/completion rates were considered important by 48 percent of respondents in
Northern County but only 31 percent of respondents in Southern County (Figure 2.3.1-11).
Controlling for differences between jurisdictions, average ratings of the importance of high
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 36

deferral/diversion rates also differed significantly by levels of experience (p = .028, F = 3.142, df =
3), with the perceived importance declining substantially with increasing experience (Figure 2.3.112).

Percentage of Respondents within Counthy

Figure 2.3.1-10 Responses to the question: “High deferral/diversion rates for eligible
defendants are important for individual success,” by research site
60.0
55.0
50.0
45.0
40.0
35.0
30.0
25.0
20.0
15.0

10.0
5.0
0.0
Unimportant

Of little importance

Northern County (N=61)

Moderately
important

Important

Very important

Southern County (N=65)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 37

Percentage of Respondents within County

Figure 2.3.1-11 Responses to the question: “High success/completion rates for defendants
deferred/diverted are important for individual success,” by research site
60.0
55.0
50.0
45.0
40.0
35.0
30.0
25.0
20.0
15.0
10.0
5.0
0.0
Unimportant

Of little importance

Moderately
important

Northern County (N=61)

Important

Very important

Southern County (N=65)

Figure 2.3.1-12 Responses to the question: “High deferral/diversion rates for eligible
defendants is important for individual success,” by research site and ADA experience

Mean Importance Rating

5
4
3

2
1
0
Less than 1 year

At least 1 year but <5
years

At least 5 years but <10
years

10 years or more

ADA years of experience in current DAs office
Northern County

Southern County

While respondents did not maintain a highly punitive attitude, the survey clearly indicated that a
large majority of prosecutors assessed their success in terms of convictions, guilty pleas to the

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 38

highest charges filed, and incarceration for serious offenses. Forty-one percent of respondents
indicated that achieving high conviction rates was an important or very important criterion for
evaluating their own success (Figure 2.3.1-13). A majority (53 percent) responded that a high rate
of guilty pleas to the most serious charge(s) was important or highly important (Figure 2.3.1-14).
Finally, 57 percent also thought high imprisonment rates for serious crimes were important or very
important (Figure 2.3.1-15). The average rating of the importance of high rates of guilty pleas to
most serious charge(s) was slightly lower among prosecutors in Northern County than among
prosecutors in Southern County (p = .05, F = 3.833, df = 1); otherwise, the average orientation
toward these measures was similar across the two participating counties. The average rating of the
importance of high rates of guilty pleas to most serious charges was also strongly related to level of
experience (p = .016, weighted F for linear effect = 5.931, df = 1), with the perceived importance
declining substantially with increasing experience (Figure 2.3.1-16).

Percentage of Respondents within County

Figure 2.3.1-13 Responses to the question: “High conviction rates is important for individual
success,” by research site
60
55
50
45
40
35
30
25
20
15
10

5
0
Unimportant

Of little importance

Northern County (N=61)

Moderately
important

Important

Very important

Southern County (N=65)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 39

Percentage of Respondents within County

Figure 2.3.1-14 Responses to the question: “High rate of guilty pleas to most serious charge(s)
is important for individual success,” by research site
60
55
50
45
40
35
30
25
20
15
10
5
0
Unimportant

Of little importance

Northern County (N=62)

Moderately
important

Important

Very important

Southern County (N=64)

Percentage of Respondents within County

Figure 2.3.1-15 Responses to the question: “High rate of imprisonment for serious crimes is
important for individual success,” by research site
60
55
50
45
40
35
30
25
20
15
10
5

0
Unimportant

Of little importance

Northern County (N=62)

Moderately
important

Important

Very important

Southern County (N=64)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 40

Figure 2.3.1-16 Responses to the question: “High rate of guilty pleas to most serious charge(s)
is important for individual success,” by research site and ADA experience

Mean Importance Rating

5
4
3
2
1
0
Less than 1 year

At least 1 year but <5
years

At least 5 years but <10
years

10 years or more

ADA years of experience in current DAs office
Northern County

Southern County

Finally, 96 percent of prosecutors responded that fair treatment of defendants was important or
very important for evaluating their own success. The average rating of the importance of fair
treatment of defendants did not vary significantly by jurisdiction; ratings did increase with
increasing levels of experience, but the effect was small and not statistically significant (overall
effect, p=.446, F = .895, df = 3; linear effect, p = .127, weighted F = 2.366, df = 1) (Figure 2.3.117).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 41

Figure 2.3.1-17 Responses to the question: “Fair treatment of defendants is important for
individual success,” by research site and ADA experience

Mean Importance Rating

5
4
3
2
1
0
Less than 1 year

At least 1 year but <5
years

At least 5 years but <10
years

10 years or more

ADA years of experience at current DAs office
Northern County

Southern County

By a large margin, the outcome considered by prosecutors to be most important for defining
their own success was fair treatment of defendants, with an average rating of 4.75 on a 5-point scale
and very little variation across sites, levels of experience, or individual respondents. Items
indicating an orientation toward convictions, guilty pleas, and incarceration all elicited average
ratings slightly above the mid-point of the importance scale, with substantial variability of ratings
across individual respondents and levels of experience. Finally, items emphasizing an orientation
toward dismissals and deferrals all elicited average ratings slightly below the mid-point of the
importance scale, with substantial variation across individual respondents, jurisdictions, and levels
of experience.
Perceptions of organizational success
Prosecutors were also asked to differentiate evaluations of their own success from evaluations of
the success of the district attorney’s office. Several dimensions were important to both individual
success and organizational success and some new dimensions emerged (See Table B34 in Appendix
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 42

B). Respondents rated nineteen potential outcomes in their importance for defining success for the
district attorney’s office. Responses varied along five underlying dimensions: community
orientation, obtaining convictions and guilty pleas, crime control, consistency/fairness, and
relationships. The items addressing relationships and consistency are discussed in later sections of
the report. The presentation of importance ratings for organizational success focuses primarily on
dimensions that were not apparent in the ratings for individual success, or on items that yielded
different ratings for organizational success.
Respondents clearly saw the satisfaction of the community as a strong measure of the success of
the district attorney’s office. A high level of citizen satisfaction with the DA’s office was
considered to be important or very important by 68 percent of respondents, and was only considered
to be of little importance or unimportant by 6 percent (Figure 2.3.1-18). The average rating of the
importance of citizen satisfaction was significantly higher for Northern County than for Southern
County (p=.000, F=38.211, df=1); but, controlling for the difference in average ratings between
jurisdictions, the ratings did not differ significantly across levels of experience.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 43

Percentage of Respondents within Coumty

Figure 2.3.1-18 Responses to the question: “A high rate of citizen satisfaction with the DA’s
office is important for office success,” by research site
60
55
50
45
40
35
30
25
20
15
10
5
0
Unimportant

Of little importance

Northern County (N=62)

Moderately
important

Important

Very important

Southern County (N=65)

The orientation favoring convictions and guilty pleas as one of the defining characteristics of a
successful prosecutor’s office was best represented by responses to the same three items that
defined individual success: high conviction rates, high imprisonment rates for serious crimes, and
low dismissal rates after charges are filed. However, these dimensions were rated as even more
important for defining organizational success. Achieving high conviction rates was rated important
or very important for organizational success by 57 percent of respondents, compared to 41 percent
of respondents who rated this as important or very important for individual success. Similarly,
achieving high imprisonment rates for serious crimes was rated important or very important for
organizational success by 74 percent of respondents, compared to 57 percent who saw it as
important or very important for individual success. Achieving low dismissal rates was rated
important or very important for organizational success by 46 percent of respondents, while just 27
percent of respondents found it important or very important as a measure of individual success.
Finally, high deferral/diversion rates were much more frequently rated by prosecutors as important
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 44

or very important for organizational success (46 percent) than for their own individual success (23
percent). For organizational success, average ratings of the importance of high conviction rates were
significantly higher for Northern County than for Southern County (p=.007, F=7.607, df=1); there
were no significant differences across jurisdictions for other measures. Similarly, controlling for
differences between jurisdictions, ratings of the importance of these measures did not differ
significantly across levels of experience.
Similar to ratings of individual success, prosecutors saw the fair treatment of defendants as very
important to defining organizational success. The distribution of ratings of the importance of fair
treatment of defendants was nearly identical to the distribution of responses to the same item with
respect to defining individual success. These ratings did not differ significantly by jurisdiction, but
the ratings of the importance of fair treatment for defining organizational success did increase
significantly with increasing levels of experience (p=.044, F=2.775, df=1). As with the prosecutors’
ratings of criteria for defining individual success, the outcome considered by prosecutors to be most
important for defining organizational success was fair treatment of defendants, with an average
rating of 4.65 on a 5-point scale, with very little variation across sites or individual respondents.
Finally, prosecutors were asked to rate the importance of crime control as a measure of the
success of the district attorney’s office. This dimension was not included in the section of the survey
pertaining to individual success (the reason being that no individual prosecutor can or should be
seen as affecting crime rates). For organizational success, the dimension was best represented by
two items, pertaining to fewer defendants re-arrested after prosecution and lower crime rates.
Having fewer re-arrests among defendants after prosecution was considered important or very
important for defining organizational success by a majority of respondents (54 percent), while 17
percent judged lower re-arrest rates to be of little importance or unimportant (Figure 2.3.1-19).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 45

Similarly, nearly two-thirds (65 percent) of prosecutors considered lower crime rates to be
important or very important, while only 10 percent considered lower crime rates to be of little
importance or no importance (2.3.1-20). The importance of both re-arrest rates and crime rates for
defining organizational success was rated higher among respondents in Northern County than
among respondents in Southern County. The observed differences were not statistically significant
for ratings of the importance of re-arrest rates but were significant for crime rates (p=.027, F=5.034,
df=1); moreover, there was a significant interaction between jurisdiction and level of experience
(p=.045, F=2.764, df=3), such that the difference between jurisdictions was only evident among
respondents with 5 to 10 years of experience. For both re-arrest rates and crime rates, importance
ratings declined slightly with increasing levels of experience. However, the differences across levels
of experience were not statistically significant, after controlling for differences between
jurisdictions and the interaction between jurisdiction and experience.

Percentage of Respondents within County

Figure 2.3.1-19 Responses to the question: “Fewer defendants re-arrested after prosecution is
important for office success,” by research site
60
55
50
45
40
35
30
25
20
15
10

5
0
Unimportant

Of little importance

Northern County (N=62)

Moderately
important

Important

Very important

Southern County (N=65)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 46

Percentage of Respondents within County

Figure 2.3.1-20 Responses to the question: “Lower crime rates is important for office success,”
by research site
60
55
50
45
40
35
30
25
20
15
10
5
0
Unimportant

Of little importance

Northern County (N=62)

Moderately
important

Important

Very important

Southern County (N=65)

Individual perceptions and justice
Prosecutors generally defined their jobs in terms of justice. They routinely noted that their job
was to do justice, to ensure justice for defendants, victims, and the community, and to get a just
outcome in every case. Justice was not defined in terms of convictions or severe sentences. Rather
it was defined as doing the right thing. Sometimes, this meant ensuring a conviction or a severe
sentence. But often this meant declining to prosecute a case, reducing the charges in a case, or
offering a lower plea. In the end, it was often simply stated as getting the right outcome given the
circumstances.
2.3.2 Consistency and Flexibility
Although there was a clearly articulated philosophy of “doing justice” in each jurisdiction, this
was not necessarily seen as the same thing as consistency in outcomes. Doing justice was seen as
being fair or considering the implications of decisions for defendants, victims, and society.
Consistency was seen as uniformity in outcomes or, at the least, uniformity in decision-making.
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 47

Yet, in some cases, ADAs recognized that the difficulties in determining just outcomes often led
them to see consistency as justice. One ADA in Northern County observed that while ADAs were
told to do justice, this did not tell them how to handle a case – there is no way to do justice in an
absolute sense. Therefore, some ADAs focus on treating similarly situated defendants the same
since, according to some ADAs, this is the only approach that makes sense.
Indeed, in both offices, managers put mechanisms in place that were designed to ensure
consistency in outcomes and decision-making processes. As the deputy district attorney in Southern
County noted, supervisors wanted to ensure that personal opinions are not allowed to translate into
inconsistency in outcome. The goal was to discuss the general expectations and norms of the office
and mediate personal opinions. This was achieved through strategies like division of staff into small
units or teams, routine case review within units, supervisor approval of pleas, and review of overall
case statistics. In Southern County, most units relied on routine “roundtables” in which individual
cases were discussed and, in some instances, plea offers and dismissal decisions made. As one unit
manager in Southern County noted, these roundtables were where peoples’ opinions tend to stick
out and where such opinions are corrected. The offices also took steps to avoid hiring individuals
with opinions that differed widely from those of the DA. As the deputy in Southern County noted,
the office tried to avoid hiring zealots – or those not open to prioritization of cases and who do not
realize that they cannot go to the max on every case. In other words, the office tried to avoid hiring
prosecutors who would tend to be overly punitive or harsh, supporting the discussion above that the
norm in the office was an adherence to less punitive attitudes or, at least, mixed punitive attitudes.
In the end, inconsistency in Southern County was generally not a concern; the office was
divided into units by crime type with no two units prosecuting the same offenses. As the DA in
Southern County noted, within teams inconsistency is likely attenuated due to tight supervision and

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 48

roundtabling of cases. The use of roundtabling was discussed by all respondents in Southern
County. In this particular office, the practice was nearly universal and occurred on a very regular
basis, with the exception of the misdemeanor unit which was comprised of the least experienced
ADAs. Thus, ironically, the least experienced ADAs in Southern County – those likely with
opinions furthest from the norm – are not corrected through the use of roundtables. For other units,
however, the roundtable is often used to determine the charges to file and the plea offer to make,
resulting in a great deal of consistency in outcomes across cases. As discussed in Section 2.5, this
particular strategy often determines the outcomes of cases in other ways, particularly in the face of
limited resources and time constraints. When court space was limited, the group rather than the
individual prosecutor often made the decision to alter a plea offer or dismiss a case in order to make
space for more important cases.
Similar roundtables did not occur in Northern County. The domestic violence unit in Northern
County had formal monthly meetings to discuss cases and monthly trainings in which they went
over files; according to the unit manager, this was largely because many young ADAs were in the
domestic violence unit and needed that direction. In contrast, the sensitive crimes unit in Northern
County (which handles all sex offenses) never met as a group to discuss cases but encouraged
informal interactions among unit members to discuss cases. Northern County did have a similar
structure of small units which was intended to similarly attenuate inconsistency. However, this
attenuation was seen as dependant largely on peer pressure from colleagues. As a unit manager in
Northern County noted, people in units try to be consistent with each other. A perceived need for
greater consistency led to a change in the structure of the office into general crimes teams, which
prosecute both felony and misdemeanor offense types not handled by specialized units, and the
staffing of these general crimes teams with both experienced and inexperienced ADAs. The goal

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 49

was to have inexperienced ADAs working alongside experienced ADAs, who would act as a
sounding board for new ADAs and teach new ADAs office norms for handling cases. Moreover,
Northern County has several general crimes units which handle identical cases. As such, there is
more potential for inconsistency across the office in handling similar cases. This potential was
recognized by ADAs, who noted that there is a great deal of consistency within units, but across
units there may be some inconsistency. In fact, ADAs in Northern County indicated that consistency
in the office was likely impossible. Several ADAs expressed concern about inconsistency across
judges, noting that there is no need to worry about inconsistency in the office because judges are so
different that once you get to court, inconsistencies will be introduced anyway. In the end, in
Northern County the only strict level of supervision over decision making comes at the plea offer
stage in which any recommendation for prison must be run by a supervisor; no other decisions need
be discussed with colleagues or supervisors.
While consistency was seen as important, many respondents also maintained that there was a
need for flexibility. In fact, in both jurisdictions, there was a sense that the office could use the fact
that people have different opinions to make things better, by getting people to debate how to handle
cases. In some respects, consistency was not always seen as a clear goal of prosecution. In fact,
inconsistency was often seen as acceptable, particularly by new ADAs in Southern County. As
noted above, new ADAs work on the misdemeanor unit and do not roundtable cases like other units;
rather, they are supervised by one unit manager who monitors decisions after the fact. As several
new ADAs noted, each individual prosecutor has particular types of offenses that they care about
more and for which they are more opinionated and take a particular approach that may be [their]
own. As one ADA noted, he hated people who pass stopped school buses. While the ADA
acknowledged that he could offer something similar to his colleagues, he did not; other ADAs may

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 50

not care about passing a stopped school bus and will plead it down to something else, but he told
defendants that they can either plead guilty to that offense or go to trial. Other new ADAs in
Southern County saw consistency as consistency in approach rather than outcome, noting that
consistency lies in supporting each other on decisions, not necessarily in giving the same deals or
getting identical outcomes. Again, as noted above, this approach was supported by supervisors in
Southern County who maintained that individual prosecutors were allowed to seek outcomes
outside the norm as long as the prosecutor could justify the result. In Northern County, ADAs also
were generally willing to accept a certain level of inconsistency in outcomes. Since prosecutors are
dealing with facts that control what is done, prosecutors argued that inconsistencies are bound to
occur because no two cases have the same facts. According to some ADAs, they were comfortable
with other prosecutors coming to different conclusions than their supervisors, as long as unit
managers know what ADAs on their unit are doing. Thus, according to some ADAs in Northern
County, inconsistency in outcomes was acceptable as long as it was supervised.
As this discussion indicates, prosecutors who participated in the focus groups expressed a
mixture of concern about consistency, confidence in the processes in place to promote consistency,
and acceptance of a certain degree of inconsistency. Ratings by the broader sample of prosecutors
who responded to the general survey were less mixed. Large majorities identified consistency as an
important goal. Nearly three-quarters of respondents (74.8 percent) considered similar outcomes for
similar cases within units to be important or very important in defining organizational success,
while only 3 percent considered it to be of little importance or unimportant (Figure 2.3.2-1).
Consistent with the focus group responses, respondents were slightly less concerned about
consistency across units. Sixty-three percent of respondents considered similar outcomes across
units to be important or very important, whereas only 10 percent considered it to be of little

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 51

importance or unimportant (Figure 2.3.2-2). Roughly 87 percent respondents agreed or strongly
agreed with the statement: “For similar cases, there should be a great deal of consistency across
prosecutors in the factors that influence the decision in a case [emphasis added],” while only 4
percent disagreed, and none disagreed strongly (Figure 2.3.2-3). The average level of agreement did
not vary significantly across jurisdictions or level of experience for any of the three items related to
consistency.

Percentage of Respondents within County

Figure 2.3.2-1 Responses to the question: “Similar outcomes for similar cases within units is
important for office success,” by research site
60
55
50
45
40
35
30
25
20
15
10
5
0
Unimportant

Of little importance

Northern County (N=62)

Moderately
important

Important

Very important

Southern County (N=65)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 52

Percentage of Respondents within County

Figure 2.3.2-2 Responses to the question: “Similar outcomes for similar cases across units is
important for office success,” by research site
60
55
50
45
40
35
30
25
20
15
10
5
0
Unimportant

Of little importance

Northern County (N=62)

Moderately
important

Important

Very important

Southern County (N=64)

Percentage of Respondents within County

Figure 2.3.2-3 Responses to the question: “For similar cases there should be a great deal of
consistency across prosecutors in the factors that influence the decision on a case,” by
research site
70
65
60
55
50
45
40
35
30
25
20
15
10
5
0
Strongly disagree

Disagree

Northern County (N=61)

Neutral

Agree

Strongly agree

Southern County (N=65)

There is some question as to the appropriate interpretation of these findings. Some of the focus
group discussions explicitly distinguished between consistency of outcomes and consistency of
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 53

approach while others did not; as a result, it was not always clear which form of consistency
prosecutors were discussing, which may explain variation in responses. However, prosecutors’
responses to the general survey showed less variation in their support for consistency in both
decision criteria and outcomes. The statement that elicited the strongest agreement (87 percent)
referred explicitly to consistency in decision criteria, but substantial majorities of respondents also
rated consistent outcomes within units (75 percent) and consistent outcomes across units (63
percent) to be important objectives for defining organizational success. It is possible that the survey
responses reflected primarily abstract ideas, while the focus group discussions may have reflected
more pragmatic views about what it is possible to achieve.
2.3.3 Efficiency
Unlike justice and consistency--which focus on process and how case outcomes affect
defendants, victims, and society--efficiency was a goal of the prosecutor’s office that clearly
focused on how cases affected the office itself and the court system generally. Moreover, unlike
justice and consistency, efficiency as a goal was also difficult for supervisors to convey to new
ADAs. The deputy in Southern County noted that the toughest thing to get across to new ADAs is
that they cannot try all cases and, as ADAs are promoted from the misdemeanor unit to felony
units, supervisors must keep repeating the lesson. This was echoed by unit managers in Southern
County, who routinely instruct new ADAs that they have to dismiss, plead down, or otherwise get
rid of cases. And newer ADAs in Southern County clearly recognized the tension between doing
their job and ensuring efficiency; one new ADA in Southern County saw the job as to get rid of
every case.
In some instances, prosecutors saw their role more broadly as gatekeepers to the criminal justice
system. As such, they often saw themselves as ensuring efficiency not just in the district attorney’s
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 54

office but for the entire system as well. One ADA in Northern County argued that prosecutors see
themselves as stewards of criminal justice resources and seek to move minor cases to municipal
court so they could do better justice to more serious crimes. Often, the prosecutor’s office was seen
by ADAs as the only party interested in efficiency. This was particularly true in Southern County,
in which judges do not carry dockets, rather they rotate through courtrooms according to a set
schedule which lasts just a few weeks. As the deputy in Southern County noted, in most
jurisdictions, judges, the public defender, and the DA have a need for efficiency, but with judges not
having their own calendar, judges have no need for efficiency except to get out of court; in turn, the
deputy maintained that the public defender has some interest in delay which means it is only the
prosecutor that has an interest in efficiency.
The general survey revealed similarly mixed opinions regarding the importance of efficiency,
which may be a reflection of the difficulty in communicating the need for efficiency to new ADAs
and the clear tension that exists when decisions on individual cases must be made for efficiency
reasons. On the one hand, 88 percent of respondents agreed or strongly agreed that the quick
resolution of cases is a legitimate goal of the criminal justice system, with no significant differences
in average level of agreement by jurisdiction or level of experience. On the other hand, prosecutors
were divided in their ratings of the importance of expediting cases. A majority of respondents (53
percent) considered it to be of little importance or unimportant to examine cases at screening in
terms of their plea bargaining potential, their potential for early disposition, and their potential to
reduce the number of cases in the system, while 19 percent considered it important or very
important. Among the choices offered in the general survey for priority consideration at
screening—expediting cases, examining the appropriateness of diversion, assessing convictability,
and considering constitutional or evidentiary issues—expediting cases received the lowest ratings of

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 55

importance. The average rating for the importance of expediting cases was significantly higher
among prosecutors in Southern County than among prosecutors in Northern County (p=.050,
F=3.925, df=1) and the average ratings declined with increasing levels of experience for Southern
County (p=.069, F=2.428, df = 3, for the interaction between jurisdiction and level of experience)
(Figure 2.3.3-2).4

Percentage of Respondents within County

Figure 2.3.3-1 Responses to the question: “How important is it to examine cases at screening
in terms of their plea bargaining potential, their potential for early disposition, and their
potential to reduce the number of cases in the system,” by research site
60
55
50
45
40
35
30
25
20
15
10
5
0
Unimportant

Of little importance

Northern County (N=62)

Moderately
important

Important

Very important

Southern County (N=58)

The decline in the importance of expediting cases associated with increasing levels of experience may be an artifact of
confounding between experience and unit assignment in Southern County. All first-year ADAs and some second-year
ADAs are assigned to the misdemeanor unit, where the pressure to expedite cases is the greatest. Unfortunately, the
survey data available for these analyses did not include unit assignment of the respondents.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 56

Figure 2.3.3-2 Responses to question: “I examine cases at screening in terms of their plea
bargaining potential, their potential for early disposition, and their potential to reduce the
number of cases in the system,” by research site and ADA experience

Mean Importance Rating

5
4
3
2
1
0
Less than 1 year

At least 1 year but <5
years

At least 5 years but <10
years

10 years or more

ADA years of experience in current DAs office
Northern County

Southern County

Prosecutors were also divided in how frequently they said they were willing to adjust their
decisions in order to increase courtroom efficiency. Only 20 percent said they were willing to do so
frequently or very frequently, whereas 34 percent said they were rarely willing or never willing to
adjust their decisions for the sake of efficiency (Figure 2.3.3-3). The average frequency rating for
willingness to adjust decisions for the sake of efficiency was significantly higher among
respondents in Southern County than among respondents in Northern County (p=.000, F=13.946,
df=1). Controlling for the difference between counties, there were no significance differences in
average frequency rating across levels of experience.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 57

Percentage of Respondents within County

Figure 2.3.3-3 Responses to the question: “I am willing to adjust my decisions to increase
courtroom efficiency,” by research site
60
55
50
45
40
35
30
25
20
15
10
5
0
Unimportant

Of little importance

Northern County (N=61)

Moderately
important

Important

Very important

Southern County (N=65)

The goal of efficiency – and to some extent the need for efficiency – is discussed in greater
detail in Section 2.5, in which resource constraints and their impact on decision making are
discussed. The higher ratings of the need to expedite cases and adjust decisions in order to increase
courtroom efficiency in Southern County are likely the result of specific constraints on access to
courtrooms in that jurisdiction. As we discuss below, the limited court space for trying cases has led
prosecutors in Southern County to routinely re-evaluate cases to determine which are the most
important to try; as a result, more cases are considered for early pleas and dismissals in order to free
court space for more serious cases.
2.3.4 Prosecutors’ Perspective on the Goals of Prosecution
In both offices, the District Attorney had a simple philosophy to “do justice.” Individual
prosecutors and prosecution units were granted considerable discretion with little formal guidance
for translating that broad philosophy into operational objectives. As a result, there is the potential
for the decision making of individual prosecutors to reflect their personal beliefs and for the
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 58

definition of “justice” to take on several conflicting meanings. Taken together, the focus group
discussions and the survey responses revealed substantial variation in prosecutors’ beliefs and
opinions concerning the general goals and functions of the criminal justice system, objectives that
define success for individual prosecutors, and objectives that define success for the district
attorney’s office. The question is whether these variations translate into differences in how
prosecutors evaluate cases or in actual case outcomes.

2.4 Strength of the Evidence and Severity of the Case: Case-Specific Factors in Decision
Making
As expressed by many prosecutors in the focus groups, a clear philosophy does not provide
clear guidance on how to handle a case; rather, as one prosecutor noted, decisions are based on the
facts of the case. Thus, when examining how prosecutors make decisions, case specific factors are
paramount. Prior research has consistently shown that case-specific factors – primarily, strength of
the evidence, seriousness of the offense, and defendant criminal history – drive prosecutorial
decision making (Spohn & Holleran, 2001). According to prosecutors, these three factors come
together to determine how to process a case. Understanding how these factors are used in
processing a case, however, is less clear. Interviews with prosecutors revealed that these factors are
used to answer two fundamental questions. Strength of the evidence is used to answer the question:
Can I prove the case? In other words, prosecutors look to the strength of the evidence to determine
if a case can be prosecuted and result in a conviction. Seriousness of the offense and defendant
criminal history are then used to answer the question: Should I prove the case? In other words,
while strength of the evidence may indicate that a case can succeed at trial, prosecutors then look to
other factors to determine if a case should be prosecuted and result in a conviction. Answers to both
of these questions may evolve over time, as evidence deteriorates or new evidence comes to light,
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 59

as more complete information about defendant characteristics and circumstances becomes available,
and as external circumstances influence prosecution strategies or the prioritization of cases. The
following sections examine how prosecutors evaluate strength of the evidence and other casespecific factors and how these ultimately affect decision making.
2.4.1 Strength of the evidence
According to prosecutors, the most important factor considered in determining whether a case
will go forward is the strength of the evidence. In all of the focus groups, all prosecutors stated that
a case will not be accepted for prosecution unless it has strong evidence. As the DA in Southern
County noted if the prosecutor cannot prove the case, it does not matter if the crime is the worst
imaginable or the defendant is the worst you have ever seen – if you cannot prove the case then it
doesn’t matter. In both counties, the standard used to evaluate strength of the evidence was
likelihood of success at trial. Prosecutors examined the case on the merits, determining first, if the
facts are present and, second, if the evidence was there to support the facts.
This consensus extended to the broader sample of prosecutors who responded to the general
survey. Over 95 percent of respondents said that convictability and probability of success at trial
were important or very important considerations in screening cases. Asked to choose which of four
factors is the most important consideration in screening cases, two-thirds chose probability of
conviction and success at trial, three times the fraction that chose constitutional issues (22 percent),
and nearly seven times the fraction that chose deciding between diversion and vigorous prosecution
(10 percent) (Figure 2.4.1-1). (Only one respondent chose the fourth option—relating to aspects of
the potential for expediting the case—as the most important consideration at screening.) Average
responses did not vary significantly by jurisdiction or level of experience for either the importance

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 60

rating for convictability considered alone or for the choice of the most important factor to consider
at screening.

Percentage of Respondents within County

Figure 2.4.1-1 Responses to question: “The most important consideration when screening a
case for prosecution,” by research site
70
60
50
40
30

20
10
0
Potential for plea
Need for diversion or
bargaining, early
vigorous prosecution
disposition, and caseload
reduction

Northern County (N=61)

Probability of conviction Constitutionality and
and success at trial
evidentiary issues that
may affect case

Southern County (N=60)

While prosecutors agreed that strength of the evidence guided decisions, particularly at
screening, the study sought to explore how prosecutors defined strong evidence, what determined
strong evidence, and how the evaluation of the strength of evidence varied throughout the life of a
case. Three general themes emerged from the focus groups that begin to unpack the nuances of
strength of the evidence: the quality of information received, the types of information received, and
the variation in the importance of information by crime type.
Quality of information
Evaluating and determining strength of the evidence is not a straightforward decision. Access to
information and the timing of information delivery are the first obstacles that must be overcome in
the evaluation of evidence. While a case may have strong evidence, if it is not delivered to the
prosecutor in a timely fashion, the case may not proceed. The primary hindrance to properly
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 61

evaluating strength of the evidence is access to information or, at the least, access to information in
a timely manner. The goal is to get as much information as early in the process as possible. At the
earliest stages of the process, however, prosecutors are completely reliant on law enforcement to
supply the information, which, according to the DA in Northern County, is not contextualized
information – information that allows a prosecutor to see multiple angles of a case. In other words,
the information is necessarily incomplete. And, as one ADA in Southern County noted, there is
information that the prosecutor does not even know to ask for. As a result, prosecutors in both
jurisdictions push officers to get more information to prosecutors at screening. Often prosecutors
will pend cases at screening (i.e. not make a decision) and request officers to acquire additional
information before the prosecutor will make a decision to accept or decline a case. In Southern
County, prosecutors give law enforcement ten to fourteen days to produce the additional
information. As we discuss in the following chapter, this creates some tension between the
prosecutor and law enforcement.
When discussing strength of the evidence, prosecutors focused primarily on the initial screening
decision as the stage at which the evaluation of the strength of the evidence is paramount. And,
since all of the information at screening is derived from law enforcement, prosecutors agreed that
the evaluation of the quality of evidence is affected by who collects it. As we discuss in the
following chapter, the relationships with officers – knowing who the “good” officers are and who
the “bad” officers are – is important to determining how a case proceeds. To a certain extent, these
relationships also determine how evidence is evaluated. Moreover, as we discuss in the following
chapter, recent changes in the primary municipal police departments in each jurisdiction have
affected the quality of information delivered to the DA’s office. According to prosecutors in both
jurisdictions, the primary police department has started to devote fewer resources to investigation;

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 62

as a result, less information is delivered to prosecutors at screening and, according to most
prosecutors, the quality of information delivered has decreased. In other words, prosecutors
maintained that they now received less information and the information that they do receive is of a
lower quality. While an evaluation of the actual quality of information currently delivered to
prosecutors is beyond the scope of this study, the responses clearly indicated that the amount and
quality of information was vitally important to prosecutors’ feelings that they could carry out their
jobs and effectively prosecute cases.
The quality of the evidence also may change over time. Although evidence may be strong at the
beginning of a case, it may weaken over time. As one ADA in Northern County noted, as soon as
you put a case through intake, the expiration date starts. This is primarily because there are always
holes in the information. The declining strength of evidence is also particularly true for cases that
rely on victim and witness testimony. Prosecutors noted that witnesses are passionate right after an
event, but are less passionate three years later, particularly marginal witnesses. Prosecutors also
noted the problem of simple memory loss over time which makes witnesses look bad. This expiring
of information over time determined how prosecutors evaluated different types of information in
their determination of the strength of evidence.
Types of information
Obviously, not all evidence is equal. In determining whether evidence is strong or weak,
prosecutors look to several things. According to ADAs in Northern County, strong evidence
includes scientific evidence, confessions, witnesses; prosecutors were quick to point out that it is
generally not one piece of evidence that makes a case strong, but all of the pieces together
determine when evidence is strong. Moreover, strong evidence also involves having the right
evidence for the case. As a unit manager in Northern County noted, it comes down to what you

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 63

would expect the evidence to be, or whether the evidence is there when you expect it to be there.
What prosecutors expect is also driven by what juries will expect. As noted above, the primary
metric in evaluating strength of the evidence is probability of success at trial. As a result, evaluating
the types of information depends partially on how a jury may evaluate the information. As one unit
manager in Northern County noted, there are expectations among juries now of certain evidence in
certain cases. For example, prosecutors noted that videos are now used in all arrests for DUI and
people are expecting to see a person very drunk in the videos; however, as one ADA pointed out,
not everyone looks really drunk in the videos.
Although prosecutors must rely on witnesses in many cases, credibility issues with witnesses or
victims often require prosecutors to re-evaluate the strength of the evidence. Evidence that may
appear strong at the beginning of a case may, indeed, be weak once witnesses are interviewed.
Thus, physical evidence is weighed more heavily than testimonial evidence, which may change over
time and is open to cross examination. As one ADA in Northern County noted, the quality of
physical evidence is less likely to change over time and is less open to interpretation. Physical
evidence is generally seen by all parties as the same thing – everyone will agree that a gun is a gun;
in contrast, testimonial evidence will be interpreted differently by different people and the jury may
interpret it very differently than the prosecutor.
Testimonial evidence, particularly from victims, was discounted by many of the prosecutors in
the focus groups. As one unit manager in Northern County noted, when screening a case
prosecutors often think about if they can prove the case without the victim showing up and look for
evidence that allows the prosecutor to prove the case without a witness. In some instances,
prosecutors took hostile attitudes toward victims. As one ADA in Northern County noted, “If the
victim does not care about the case, why should I?” The evaluation of the victim is often subjective

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 64

as well. A new ADA in Southern County noted that it is very difficult to determine which witnesses
are credible; as a result, this particular ADA tried to read people and body language and evaluate
what people were telling her.
Evaluating victim or witness testimony is also problematic when a victim or witness was also
once a defendant. Prosecutors in both jurisdictions commented that most of the crime in their
jurisdictions occurs in fairly concentrated areas and that many of their victims are also defendants in
prior or pending cases. A unit manager in Southern County admitted that in felony cases it was very
difficult to work with victims who were once defendants, but maintained that if the evidence is there
and the victim is accessible, then the case will proceed. However, in one case screening observed in
Northern County, the victim’s prior criminal record and pending criminal cases were weighed
heavily in the screening decision. In this particular case the individual was the victim of a shooting,
but also had been prosecuted several times and had open cases. The credibility of the victim’s
potential testimony was weighed heavily in the decision to prosecute; while not unsympathetic to
the victim’s injuries, the screening ADA also saw the victim as a liability to the case and chose not
to accept it for prosecution.
While there was general agreement about what types of evidence were stronger than others, it
also was clear that this evaluation varied by offense type. Respondents were in general agreement
that evidence in drug and gun cases was generally strong relative to other offense types, largely
because such cases relied primarily on physical evidence. This may explain the generally lower
declination rates and higher conviction rates for drug offenses relative to other offense types (see
Part 3). Also, while physical evidence was generally seen as stronger than testimonial evidence for
all offense types, testimonial evidence was often seen as equally or more important in sex crimes
and domestic violence cases. As the unit manager of the sensitive crimes unit in Northern County

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 65

noted, “testimonial evidence is always important because often that may be all that you have.”
Indeed, prosecutors handling sex crimes and domestic violence pointed out that there is often very
little evidence to evaluate in sex offenses and domestic violence. As a result, the strength of the
evidence may depend on a single witness – the victim in the case. In turn, according to prosecutors,
the lowest common denominator is: will the jury believe the victim?
Since victims and witnesses are primarily involved in person offenses, cases involving these
types of offenses are more likely to be affected by the re-evaluation of evidence over time and, as a
result, more likely to be dismissed or amended as the case proceeds. Thus, domestic violence cases
are often subject to re-evaluation of evidence over time. As the unit manager of the domestic
violence unit in Northern County noted, the longer it takes a DV case to get to trial, the less chance
there is to get the victim to show up. Another ADA in Northern County expressed it this way,
“Actually, anything that involves people, after time passes, the reward just isn’t enough.” As
discussed in Section 2.5, the significant delays in court processing time, largely the result of fewer
courts in each jurisdiction, may be affecting person offenses more than other offenses, largely
because of this reliance on victims.
2.4.2 Severity of the offense, defendant criminal history, and other case-specific factors
Prosecutors clearly saw a distinction between the strength of the evidence and the merits of the
case. Strength of the evidence is a necessary but not sufficient condition for a case to proceed. It
merely answers the question, “Can I prove the case?” Once a prosecutor determines that there is
strong evidence and they can, indeed, prove the case, they answer a second question, “Should I
prove the case?” As one unit manager in Northern County noted, “the volume of evidence
determines if I can prove a case, but the specific circumstances determine whether I should prove a
case.” Basically, prosecutors determine if the case is serious enough to merit charges. In answering
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 66

this question, prosecutors consider the severity of the offense, the defendant’s criminal history, and
other defendant and case-specific factors. Prosecutors admitted that while a serious offense with
weak evidence may proceed, a non-serious offense with strong evidence may, in turn, not proceed.
As we discuss in Section 2.5, often this is due to resource constraints that prevent the district
attorney’s office from pursuing all viable cases; with limited staff and court time, units or individual
prosecutors may choose, for example, to pursue person offenses involving injury or weapons over
cases involving threats of injury or no weapons.
While statutory offense categories clearly delineate the severity of one offense relative to other
offenses, these statutory provisions often do not align with local evaluations of offense severity; as a
result, prosecutors’ evaluations of offense severity are often determined by practical considerations.
As one ADA in Northern County noted, prosecutors have to worry about the ‘I don’t care’
sentiment of the jury in bringing less serious cases. Even if there is strong evidence, prosecutors
may not bring a case if the jury just wouldn’t care. This was seen as particularly true for drug
offenses. As the unit manager of the drug unit in Northern County noted, drug cases have a higher
bar to prove to the jury; most people see all drug cases as non serious. As a result, prosecutors in
drug cases often look for some other factor to get the jury excited – a long criminal record, gang
affiliations. As discussed above, prosecutors also rely on their own evaluations of offense severity.
As one ADA in Southern County pointed out, every prosecutor has their own specific offense that
they care about; as a result, individual prosecutors may rank certain offenses as more severe than
their colleagues and, in turn, pursue these cases when their colleagues would not.
Defendant criminal history was the third axis nearly all prosecutors mentioned as determining
whether a case would go forward. Criminal history worked in two directions in determining whether
a case should be prosecuted: defendants with more serious criminal histories were more likely to be

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 67

prosecuted even when faced with weaker evidence and defendants with less serious criminal
histories were sometimes considered for no prosecution even faced with stronger evidence. Again,
prosecutors sought to answer the question “should I prove the case,” and often determined that this
case should proceed because this defendant is a ‘bad guy’ or this case should not proceed because
this defendant is not a bad guy. Indeed, prosecutors expressed some tension in declining cases that
had weak evidence but involved a “really bad guy.” But, this also poses challenges; as unit
managers in Southern County noted, it is easy to fall into the trap of taking cases when there is
serious injury or a ‘bad” defendant.
Although prosecutors agreed that defendant criminal history was a primary factor in
determining whether a case would proceed, prosecutors noted many defendant characteristics that
appeared to affect decisions as much as criminal history. In determining whether a case should be
proved, the answer often comes down to determining what is fair for the defendant. As one ADA in
Northern County noted, the ultimate resolution of a case should be based on what is fair, not about
what can be proved. As a result, several defendant characteristics are considered: the age of the
defendant, the potential impact of a conviction on the defendant, and the demeanor of the
defendant.
Age of the defendant was often a fairly straightforward consideration. As one ADA noted,
prosecutors often get cases involving kids getting in fights in response to broken video games or
some prior interaction; prosecutor often consider it enough to explain to the defendant, ‘you can’t
do that.’ More often, however, the age of the defendant interacts with considerations of the impact
of a conviction on the defendant. Prosecutors noted that they ask the questions: what is this going to
do to the defendant’s life? Do we want to have a 17 year old with a felony conviction? Often
prosecutors concluded that pursuing a case will do more harm than good and will choose not to

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 68

prosecute. But this is also influenced by the demeanor of the defendant. As one unit manager in
Northern County noted, the way the defendant handles himself matters. This can include taking
responsibility for their actions or showing disrespect to courtroom actors; in the former, if the
offense is not severe, the case may not be pursued because the defendant is remorseful; in the latter,
even if the offense is not severe, some ADAs (particularly newer ADAs) argued that the case may
be pursued to teach the defendant a lesson. Moreover, as discussed above, new ADAs in Southern
County working in misdemeanor court have very little information and very little time to devote to
assessing a case when it comes in. As a result, rather than relying on concrete information about a
case, these ADAs rely on a feeling about a person. Thus, in these particular cases, it appeared that
demeanor of the defendant mattered a great deal. As discussed in Section 2.5, prosecutors also
noted that much of the evaluation of defendant characteristics was dependant on information
supplied by the defense attorney. Thus, the evaluation of defendant characteristics becomes largely
dependent on the amount and quality of information supplied by the defense attorney.
While credibility and probable participation of victims may go into evaluating the strength of
the evidence, victim wishes also matter in determining whether a case should proceed. This is a
slightly different calculation than determining whether victims will actually show up for trial.
Victims may not show up for trial because they cannot be contacted or the prosecutor believes that
they will not show up for trial, which goes to the question of ability to prove a case or the
probability of success at trial. In other instances, however, the express wishes of the victim also
determine whether a case should proceed. As one unit manager in Northern County noted, the
system makes it worse for the victim – making them go through the process. As a result, some
prosecutors felt it was often unfair to make victims go through it if they do not want to. This is
particularly true for certain offenses, such as sex offenses. As the unit manager of the sensitive

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 69

crimes unit in Northern County noted, in sex crimes, prosecutors ask the victims if they want to
proceed and place a lot of weight on what the victim wants. However, this was markedly different
for domestic violence cases in Northern County. As the unit manager of the domestic violence unit
in Northern County argued, most of victims of domestic violence do not want to proceed, so
prosecutors do not factor in victim willingness to proceed or testify; rather in domestic violence
cases, prosecutors try to keep the victim out of the process. The approach of the domestic violence
unit was largely shared by other prosecutors (with the noted exception of the sex crimes unit):
uncooperative victims were considered when evaluating the strength of the evidence (i.e. whether
victims were central to the case and whether victims could be counted on the appear at trial); but
uncooperative victims or victims who did not want the case to proceed were not considered when
evaluating whether a case should proceed.
2.4.3 Improving the evaluation of evidence
Because evaluating strength of the evidence is often difficult, the DA in Southern County
reorganized the office so that more experienced ADAs were the only ones evaluating cases at
screening. Part 3 of this report presents empirical analyses that examine the relationships between
strength of evidence and both screening decisions and plea offers, as well as consistency of
screening and plea bargaining decisions among prosecutors for cases with similar evidence profiles.
Part 3 also examines the relationship between defendant, offense, and victim characteristics and
case outcomes for a series of offense types. Before turning to the analyses of case outcomes,
however, the following section examines contextual factors – factors outside the confines of a case
– that also impact prosecutorial decision making.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 70

2.5 Rules, Resources, and Relationships: Outside Influences on Decision Making
As discussed above, prosecutors determine whether a case will proceed based on case specific
factors – strength of the evidence, severity of the offense, and defendant criminal history. Recent
empirical research has shown that contextual factors also affect decision making in the criminal
justice system. Case loads, office policies, and relationships among courtroom actors can influence
charging decisions (Ulmer et al., 2007) as well as sentencing outcomes (Johnson, 2005). As
prosecutors noted, cases do not exist in a vacuum; they exist within a system that acts as a help or a
hindrance to decision making. For example, office policies may screen out certain cases even if
evidence is strong or may require a specific charge or sentence in a plea offer. A lack of available
support staff to track down witnesses or a lack of courtroom space may alter the ability to fully
pursue a case, requiring prosecutors to reassess whether a case will proceed. A close relationship
with law enforcement may lead prosecutors to accept less evidence when screening a case, while an
antagonistic relationship may lead them to discount officers’ version of events. The study explored
three primary contextual factors that may affect decision making – rules, resources, and
relationships.
2.5.1 Rules
While formal criminal procedural rules govern how a case may proceed, these rules do not
necessarily govern decision making. Policies within the district attorney’s office, however, may. For
example, both jurisdictions in the current study had rules pertaining to specific offenses that, for
example, required a deferred prosecution, or could not be pled down to a lesser offense, or required
a recommendation of a prison sentence following conviction. Nonetheless, in both jurisdictions
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 71

prosecutors noted a general absence of formal policies or rules that governed decision making
except in a very few select cases. Even when formal policies existed, the policies were often not
well-publicized or well-communicated to staff. As one ADA in Northern County noted, “Office
policies are urban legends.” This comment indicated a tension that existed, at least in Northern
County, between a desire for some policy or guidance on decision making and a desire for unguided
discretion. As the District Attorney in Northern County noted, “ADAs want 100% discretion with
100% guidance.”
While craving some guidance, few saw the absence of formal policies as a problem; rather, most
noted that crafting formal policies was impossible given the complexity and uniqueness of cases. In
place of formal written policies, a set of informal policies and practices pervaded each office. In
many instances, these were described as “cultural norms of practice” illustrative of the guiding
philosophies discussed above. As the District Attorney from Northern County noted, “You can only
give guidance on basic principles.” These basic principles were then used to guide more formal
policy-making within specialized units. It was the formal and informal policies within these units
that more closely guided or constrained decision making in individual cases.
Office-wide policies
In both jurisdictions a limited number of office-wide policies dictated how ADAs were to
handle a few specific cases. For example, in Southern County, ADAs were required to prosecute
DUI cases above a certain blood-alcohol level and could not reduce felony residential breaking and
entering to a misdemeanor. According to the deputy in Southern County, except for DUI and
residential breaking and entering, all other policies are advisory. While the deputy did not
articulate additional advisory policies, he did note the office response to those ADAs who did not
follow such policies; in such advisory instances, the ADA has to evaluate how they will feel sitting

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 72

across from the DA explaining why they did not follow the advisory policy. Thus, there was a sense
among prosecutors that, although additional advisory policies existed, they functioned much like
formal policies since ADAs generally feared a confrontation with the DA. New ADAs in Southern
County noted two additional formal policies: ADAs could not reduce a speeding ticket to driving
school and could not reduce or dismiss cases involving the possession of weapons on school
grounds; according to prosecutors, the latter offenses must result in plea of guilty or a trial – only
the DA can bargain these. More experienced ADAs detailed two additional policies: residential
break-ins involving defendants with no criminal history much receive at least a 30 day split
sentence and armed robbery cannot be pled down to an unarmed robbery. The fact that each focus
group detailed different formal policies or added formal policies not mentioned by previous groups
underscores the lack of communication of policies that many prosecutors expressed. The DA in
Northern County also noted that there was conscious decision not to constrain the discretion of
ADAs with written policies. Indeed, prosecutors in Northern County described just one formal
policy: residential burglary cases should be treated as violent crimes with the expectation that the
defendant will get prison.
While office-wide policies were largely absent, the district attorney in each jurisdiction
recognized the potential need for some formality. As the DA from Northern County noted, the job
of prosecution is as much about working with your heart as working with your head, and, as a
result, sometimes decision making gets a little loose. Thus, at times, it appeared that formal policies
would provide needed direction and consistency in decision making. Indeed, some ADAs in
Northern County recognized this need as well, particularly for ensuring consistency across teams
that handle the same types of crimes and when changes in management occur. At the same time,
ADAs and the DAs in both jurisdictions noted the problems inherent in creating formal policies. As

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 73

the DA in Southern County noted, it is difficult to write a policy for screening each case due to the
variation in the factors of cases. A written policy would be such a matter of judgment that it would
not be of much value and would create more problems than it solves. ADAs in Northern County
also noted that rigid formal policies potentially disrupt interactions with defense attorneys and
judges, forcing defense attorneys to more aggressively argue to have charges initially filed that are
not bound by the written policy and affecting prosecutors’ credibility in front of the judge by tying
prosecutors’ hands without clearly knowing the context. This was echoed by the DA from Northern
County who described the reaction following the creation of the office-wide policy on residential
burglary. Residential burglary in the jurisdiction involved a high degree of overlap between the
juvenile and adult system, with many offenses committed together by a 16 year old juvenile and an
18 year old adult offender; the formal policy created a disparity between juvenile and adult court
which created problems with defense attorneys who saw the disparity in treatment for defendants
who were nearly identical in age. As a result, many ADAs referred to the written policy when
negotiating cases with defense attorneys and, in turn, the public defender felt concerned that ADAs
were just covering their ass by following the guideline. As ADAs in Northern County noted,
policies should be flexible and evolve over time; written policies are not flexible. This was echoed
by ADAs in Southern County, who maintained that you could not build a rule book big enough for
all the different types of cases; rather, prosecutors have a lot of flexibility and that is the only way it
could work.
The general survey did not include items that addressed directly the prevalence or advisability
of office-wide policies. However, it did include items that support indirect inferences about beliefs
and opinions relating to the assertions summarized above. Most respondents did not believe that
their decisions were unduly constrained by office policies. Eighty percent disagreed that they felt

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 74

constrained by office policies and practices about when to accept or decline cases for prosecution
(Figure 2.5.1-1). Over 85 percent said that office policies never or rarely compelled them to decline
cases they would have preferred to prosecute (Figure 2.5.1-2). Less than 9 percent agreed that office
policies require outcomes with which they disagree (Figure 2.5.1-3). These response patterns were
consistent across jurisdictions and levels of experience. They could reflect an overall lack of
policies at the office level, or a high degree of congruence between office policies and the
individual judgments of most prosecutors. However, there was evidence presented in earlier
sections of this report suggesting considerable divergence of opinion among prosecutors with
respect to the fundamental purposes of the criminal justice system, the goals of prosecution, and the
most effective strategies for pursuing prosecutorial objectives. In addition, more than three-quarters
of respondents disagreed with the statement that “There should be more constraints on the
discretion of officials in the criminal justice system.” Thus, it seems most likely that these survey
responses simply reflect the general lack of formal office policies and general agreement that the
high degree of discretion granted ADAs is appropriate.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 75

Percentage of Respondents within County

Figure 2.5.1-1 Responses to question: “I feel constrained by office policies and practices about
when to accept or decline cases for prosecution,” by research site
70
65
60
55
50
45
40
35
30
25
20
15
10
5
0
Strongly disagree

Disagree

Northern County (N=61)

Neutral

Agree

Strongly agree

Southern County (N=62)

Percentage of Respondents within County

Figure 2.5.1-2 Responses to question: “Office policies compel me to decline cases that I would
prefer to prosecute,” by research site
70
65
60
55
50
45
40
35
30
25
20
15
10
5
0
Never

Very rarely

Northern County (N=61)

Occassionally

Frequently

Very frequently

Southern County (N=62)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 76

Figure 2.5.1-3 Responses to question: “Office priorities require case outcomes that I often
disagree with,” by research site
Percentage of Respondents within County

60
55
50
45

40
35
30
25
20
15
10
5
0
Strongly disagree

Disagree

Northern County (N=61)

Neutral

Agree

Strongly agree

Southern County (N=65)

On the other hand, prosecutors who responded to the general survey in Northern County
expressed a need for greater formal guidance for new prosecutors and prosecutors being transferred
to new units. Only 10 percent agreed that new prosecutors receive adequate training before they
start independently handling cases and only 10 percent agreed that prosecutors receive adequate
training when they are transferred to new units (Figure 2.5.1-4). The opposite pattern was observed
for Southern County, with 43 percent of respondents indicating that new prosecutors and transfers
received adequate training. In both jurisdictions, respondents were divided with respect to whether
office goals and priorities are clearly communicated to staff. For the two jurisdictions combined,
nearly 40 percent agreed or strongly agreed that there was clear communication of office goals and
priorities, but roughly 30 percent disagreed or strongly disagreed (Figure 2.5.1-5).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 77

Figure 2.5.1-4 Responses to question: “Prosecutors in my office receive adequate training
when they are transferred to a new unit,” by research site
Percentage of Respondents within County

60
55
50
45

40
35
30
25
20
15
10
5
0
Strongly disagree

Disagree

Northern County (N=60)

Neutral

Agree

Strongly agree

Southern County (N=61)

Figure 2.5.1-5 Responses to question: “Office goals and priorities are clearly communicated to
staff,” by research site
Percentage of Respondents within County

60
55

50
45
40
35
30
25
20
15
10
5
0
Strongly disagree

Disagree

Northern County (N=61)

Neutral

Agree

Strongly agree

Southern County (N=65)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 78

Unit-specific policies
Rather than formal office-wide policies, prosecutors argued that unit-specific policies often
governed decision making in both jurisdictions. Both offices are highly decentralized, with ADAs
assigned to small units of four to ten prosecutors; in most cases, these units subject-specialized (e.g.
drugs, weapons, person offenses, etc.), with the exception of the general crimes units in Northern
County and the misdemeanor unit in Southern County. The DA in both counties then delegated
supervisory and policy-making responsibilities to unit managers. As the DA in Southern County
stated, policy making was delegated to individual teams because the problems of each unit are
different based on the nature of the crimes each unit prosecutes. As the deputy in Southern County
noted, unit managers are free to make up new policies, but they must rise to some level of
justification – they have to be able to justify it to the DA. This was echoed by the DA in Northern
County who structured the office so that policy leaders are the unit managers – they are to be the
people to whom young ADAs go to for advice.
According to the DAs, this unit level approach allows for flexibility. However, it also allows for
potential inconsistency across teams that handle similar cases and across time, as new unit managers
are promoted. For example, in Southern County a unit captain had a policy of accepting all or nearly
all cases brought by the police department and dismissing problem cases later in the process; in
effect, the unit was not screening cases at the start of the process. As a result, declinations were very
low but dismissals were high. When a new unit manager was promoted, the unit began screening
cases more vigorously and declination rates rose; in turn, dismissal rates dropped since many
problematic cases were disposed of at screening.
While it was widely acknowledged that unit-specific policies existed, it was not clear if these
were written or unwritten policies. As such, it did not appear that there was a formal policy-making

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 79

process that occurred at the unit-level; rather, policy-making was more informal, with policies
evolving out of practice. Some ADAs described these policies as norms of practice that grew out
office philosophy. These policies were learned through interactions with colleagues and often
articulated clear outcomes for cases. This was echoed by the DA in Southern County who argued
that rather than policies we have norms – a regular way of handling cases and expectations about
outcomes. ADAs in Northern County noted a similar creation of policy through routine practice
stating that there were not necessarily policies in the office but longstanding practices about how to
handle cases; these are practices that have become routine.
ADAs of varying experience levels noted the presence of several such policies. For example,
new ADAs in Southern County stated that defendants arrested on five or more charges must plead
to the top charge, community service could not be given in exchange for a dismissal for any offense,
and defendants charged with possession of a gun may not receive deferred prosecution including
simple safety classes. New ADAs in Southern County also noted that many of these informal
policies come not from more experienced ADAs, but from colleagues at the same level. In Southern
County, new ADAs are assigned to a unit that deals only with misdemeanors, are paired with one
other new ADA, and are assigned to one courtroom in which to work. As a result, a norm of
practice develops among these two, relatively inexperienced ADAs. As one newer ADA noted,
“Everyone just seems to go with what your partner suggests.”
Whether written or unwritten, the unit-level policies were clearly seen as binding on
prosecutors. While there are too many units to describe all unit-specific policies, it was clear that
these policies governed decision making in certain circumstance, calling for prosecutors to decline
certain cases at screening, charge cases in a particular way, and offer specific criteria in plea offers.
Moreover, DAs in both offices were fairly comfortable with this approach; essentially, the unit

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 80

structure allows for a delegation of authority and supervision to very experienced ADAs who are
then responsible for articulating the philosophy into practice. The DA in Northern County also
recognized the danger of allowing informal policies to develop, particularly if those policies are
unorganized or lead to conflicting outcomes across units. Indeed, unit managers also recognized the
problem of allowing each unit to develop policies autonomously. As one unit manager noted, this
may be a negative or it may be a positive. Too many policies are a bad thing; but not enough
policies leads to a potential problematic culture. Indeed, several prosecutors noted that this could
lead to inconsistencies in priorities and outcomes across units.
The ability to achieve consistent results through heavy reliance on unit structure and informal
processes depends critically on various forms of internal communication. Responses to the general
survey clustered around four ways in which information about practices and norms was shared
within the offices: office policies, training, supervision, and consultation among colleagues. Office
policies and training were previously discussed in the subsection addressing formal policies. The
following examines other kinds of communication within units.
A slight majority of respondents (52 percent) agreed that supervisors let them know how well
they were doing in their jobs. A larger majority (66 percent) said that supervisors provided adequate
guidance on when to accept or decline a case for prosecution. Average responses to these items did
not differ significantly across jurisdictions or levels of experience. In contrast, clear majorities said
they often discussed how to handle cases with their colleagues (92 percent), that they often
compared their cases to those of their colleagues to ensure they were getting similar outcomes (52
percent) (Figure 2.5.1-6), and that it was important to routinely review cases as a group (60 percent)
(Figure 2.5.1-7). Nevertheless, 60 percent also agreed that there needed to be more communication
among staff to ensure consistency of outcomes (Figure 2.5.1-8). The average levels of agreement

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 81

were significantly higher among respondents in Southern County than among those in Northern
County for the statements about reviewing cases as a group (p=.000, F=21.572, df=1) and
comparing cases to those of colleagues (p=.039, F=4.376, df=1). This is likely due to the routine use
of roundtables to determine how to handle cases in Southern County and the near absence of this
approach in Northern County. Controlling for differences between counties, prosecutors with less
than one year experience or more than ten years’ experience were less likely than others to say they
often discussed how to handle cases with their colleagues (p=.017, F=3.536, df=3).
Figure 2.5.1-6 Responses to question: “I often compare my cases to those of my colleagues to
ensure that I am getting similar outcomes,” by research site
Percentage of Respondents within County

60
50
40
30
20
10
0
Strongly Disagree

Disagree

Northern County (N=61)

Neutral

Agree

Strongly Agree

Southern County (N=65)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 82

Percentage of Respondents within County

Figure 2.5.1-7 Responses to question: “It is important to routinely review cases as a group,”
by research site
60
50

40
30
20
10
0
Strongly disagree

Disagree

Northern County (N=59)

Neutral

Agree

Strongly agree

Southern County (N=65)

Figure 2.5.1-8 Responses to question: “There needs to be more communication among staff to
ensure consistency of outcomes,” by research site
Percentage of Respondents within County

50
40
30
20
10
0
Strongly disagree

Disagree

Northern County (N=60)

Neutral

Agree

Strongly agree

Southern County (N=65)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 83

Overall, the survey responses suggest that substantial percentages of prosecutors in both
jurisdictions were comfortable with a philosophy that relies heavily on unit-level supervision and
communication among colleagues to establish guidelines and norms of practice that foster
consistency of results. According to respondents’ claims about their own collaborative activity,
adherence to these principles appears to be even more prevalent among prosecutors in Southern
County than among prosecutors in Northern County. In addition, there was a perceived need for still
more communication among staff, which could also be interpreted as a commitment to the informal
norming process.
However, some of the more detailed patterns are difficult to interpret. In particular, the patterns
of association between average ratings of agreement and levels of experience are erratic and
sometimes counter-intuitive. If routine communication between supervisors and staff and
collaborative communication among staff are effective in promoting consensus on basic objectives
and establishing norms of practice, then we might expect to find increasing commitment to the
approach with increasing levels of experience in the office. However, the expected pattern was only
evident for Southern County, and only then if the view is limited to results for the first 10 years of
experience (See Figures 2.5.1-9 through 2.5.1-12). Some of the divergence from expected patterns
is undoubtedly due to the very small number of respondents with less than one year experience in
the offices (eight in Northern County and seven in Southern County for most analyses).5 In addition,
as with other analyses presented earlier in this report, the average level of responses from
prosecutors with more than 10 years’ experience often do not conform to the pattern (if there is one)
established across the three levels covering 10 years of experience or less.

We experimented with other ways of categorizing length of experience, but the responses of those with less than one
year’s experience were so often substantially different from the responses of those with more experience, it was decided
that it was best to keep that group separate to avoid inflating the variance or otherwise distorting the findings for the
more experienced groups.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 84

Figure 2.5.1-9 Responses to question: “I often discuss how to handle cases with my
colleagues,” by research site and ADA experience

Mean Level of Agreement

0
Less than 1 year

At least 1 year but <5
years

At least 5 years but <10
years

10 years or more

ADA years of experience in current DAs office
Northern County

Southern County

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 85

Figure 2.5.1-10 Responses to question: “I often compare my cases to those of my colleagues to
ensure that I am getting similar outcomes,” by research site and ADA experience

Mean Level of Agreement

5
4
3
2
1
0
Less than 1 year

At least 1 year but <5
years

At least 5 years but <10
years

10 years or more

ADA years of experience in current DAs office
Northern County

Southern County

Figure 2.5.1-11 Responses to question: “It is important to routinely review cases as a group,”
by research site and ADA experience

Mean Level of Agreement

5
4
3
2
1
0
Less than 1 year

At least 1 year but <5
years

At least 5 years but <10
years

10 years or more

ADA years of experience in current DAs office
Northern County

Southern County

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 86

Figure 2.5.1-12 Responses to question: “There needs to be more communication among staff
to ensure consistency of outcomes,” by research site and ADA experience

Mean Level of Agreement

5
4
3
2
1
0
Less than 1 year

At least 1 year but <5
years

At least 5 years but <10
years

10 years or more

ADA years of experience in current DAs office
Northern County

Southern County

These findings are only indicative of prosecutors’ beliefs, opinions, and priorities. How their
beliefs, opinions, priorities, and length of experience relate to consistency in case-level decisions is
explored in Part 3 of this report. Nonetheless, the preceding discussion reflects general agreement
on several key points. First, formal, office-wide policies were largely absent in both participating
jurisdictions, though there were a few policies governing decisions for specific types of cases under
narrowly defined circumstances. Second, prosecutors believed that office-wide policies should
generally emphasize ethical standards and guiding principles, rather than rules that specify
particular dispositions for particular types of cases. Third, despite this, in theory, more formal
policies might be desirable to promote consistency across prosecutors and across time; however, it
is nearly impossible to develop formal office-wide policies that can adequately accommodate the
wide variety of case characteristics and contextual circumstances that prosecutors encounter.
Fourth, prosecutors are comfortable with the creation of unit-specific policies that function like
formal policies governing decision-making. Moreover, fifth, in many instances these policies dictate
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 87

a clear outcome for a case, regardless of case-specific factors such as strength of the evidence,
severity of the offense, or defendant characteristics.
2.5.2 Resources
While rules influenced decision making in specific cases, prosecutors argued that a more
important influence on the outcomes of many cases was resources. The DA in Southern County felt
that rather than written policies prosecutors need more flexibility to survive the problem of resource
constraints. Concerns about resources dominated much of the discussion in both offices and
referred to resource constraints both inside and outside the prosecutor’s office. For example, in both
jurisdictions the restricted availability of courtrooms limited the ability of prosecutors to bring cases
to trial. The lack of internal resources for support staff and investigations limited the ability of
prosecutors to devote time to “legal” work. In Northern County, a shortage of ADAs and high
caseloads limited the capacity of ADAs to pursue all cases. In some instances, it resulted in the
dismissal of cases; in others, is resulted in the changing of plea offers to expedite a resolution. In the
end, the lack of resources often resulted in prosecutors having to dispose of cases in ways they
would not otherwise have to if adequate resources were available.
Court space and court resources
In both jurisdictions, a lack of court space (i.e. the lack of available court time to try all cases
prosecutors believe should be tried) was expressed as a persistent problem facing prosecutors.
According to prosecutors, in Southern County, the lack of court space was the product of a limited
number of superior courtrooms and not enough judges, clerks, and court reporters to staff all
courtrooms. As a result, there are more triable cases than available slots to hear those cases. In
Northern County, prosecutors maintained that there were not enough courtrooms and that judges
routinely opened court late in the day or closed court early, which significantly shortened the
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 88

amount of time available each day for hearing cases. According to the DA in Northern County, this
led to a decrease in jury trials, with a particular decrease in misdemeanor trials; according to the
DA, the message from the judiciary is that they do not have enough time to handle misdemeanor
jury trials.
In Southern County the lack of available court space was due not only to a lack of resources; it
was also due to a quirk in the system in which felony court judges do not carry inventories of cases.
Rather, cases are assigned to a courtroom and judges rotate through the courtroom. Thus, a judge
handles whatever cases are in the courtroom during their assigned week and then the judge rotates
out. If a case is unresolved when the judge is scheduled to rotate out, the judge is supposed to stay
in the courtroom until the case is resolved. However, if the case is continued, then the judge may
rotate out. According to prosecutors in Southern County, this results in judges either closing court
mid-week to avoid starting a trial that may extend beyond the time at which they are to rotate out or
a continuance which would move the trial to another week and, in turn, to another judge. According
to the DA in Southern County the office experiences roughly 100 judicially initiated continuances
per week. More importantly, perhaps, the lack of available court space and high rate of
continuances in Southern County significantly affected the ability of the office to try cases and
altered the way the office handled cases. Knowing that all triable cases could not be tried,
prosecutors had to change plea offers to ensure a high volume of cases were disposed of before
trial. As the DA in Southern County noted, “[The county] has the capacity to try about 450 trials
per year. Our office processes about 10,000 felonies per year. In almost every case we make a plea
offer because we do not have the capacity to try all cases…Therefore, we are trying to avoid trials
in most cases.” According to the DA, the prosecutors are forced to resolve cases in ways they would

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 89

not otherwise do. As one experienced ADA stated, “To deal with this, you give away the farm.” In
other words, prosecutors do everything they can to get a guilty plea.
The DA in Northern County argued that the lack of court resources did not affect how cases
were handled, but it did affect how resources were allocated within the office. According to the DA,
cases are evaluated the same and views about the right thing to do on a case remains unchanged;
however, the low trial numbers should free up ADAs to do more work on cases at the front end and,
as a result, the office can restructure the criminal justice process to get people out of the system
early. In other words, since the office knows that court resource constraints prevent it from trying
all triable cases, prosecutors work harder to evaluate cases for declination and deferral. Thus, while
the DA did not see resource constraints as affecting decision making, it appeared that resource
constraints did affect outcomes by funneling more cases out of the system earlier than would have
occurred in the absence of such a constraint. Indeed, the DA in Northern County recognized this,
noting that the resolution of cases early in the process comes with a political cost and pushes staff to
make different decisions.
The lack of courtroom space or the continuance of cases required prosecutors to re-evaluate
pleas and to try to come up with better pleas to get rid of cases. Staff have to pick and choose what
to do – some cases are continued and some cases are simply dismissed. In some instances,
prosecutors in Southern County work as a group to get rid of the worst cases the unit had on its
caseload. It was a process of deciding whether they wanted to waste a trial on a particular case.
Prosecutors described a process of prioritizing in which cases are ranked from strongest to weakest,
from the worst to least bad offense, from a high amount of victim cooperation to a low amount, from
newest to oldest, from in custody defendants to not in custody. All of this changed the threshold of
what prosecutors were willing to accept or dismiss and often cases were dismissed simply because

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 90

they sat at the bottom of the trial calendar. According to some, this resulted in prosecutors making
decisions that they may not be comfortable with because of court constraints and time.
In Southern County, the decision of what to do on a particular case is often outside of the
control of an individual prosecutor; rather, when resource constraints push the office to re-evaluate
the cases to be tried, roundtables are used. This generally means that all cases are evaluated together
and the unit decides which cases the entire unit will dispose of. As the unit manager of the drug unit
noted, roundtables were used to decide which cases to keep or re-evaluate and what to try and
when. Another unit manager in Southern County argued that prosecutors may consider our entire
inventory of cases when making a plea offer; for example, if there is not enough trial time for all
cases if they went to trial, prosecutors may make an offer that the defendant is more likely to take.
So, there is an incentive to make the best offer to get the case resolved. According to one
experienced ADA in Southern County, prosecutors compare the merits of one case to the merits of
other cases. Delays in cases present a similar problem. As one unit manager in Southern County
noted, continuances mean that some cases will have to be dismissed. When there are more cases
scheduled for trial than available court space, prosecutors roundtabled cases to decide what will be
dismissed or have a new plea offer made.
Staffing
Available court space limits the number of cases that can be tried and, as a result, affects the
way prosecutors evaluate cases and craft plea offers. But the availability of internal resources
affects the prosecutor’s ability to prepare a case in the first place. Even if courtroom space is
available, if the prosecutor does not have resources to track down witnesses or access information,
this may also change how cases are handled. The lack of internal resources affected both
jurisdictions and revolved primarily around the availability of support staff.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 91

In Southern County, the DA noted that internal resources were routinely a problem. As the DA
noted, the lack of support staff forces ADAs to do clerical work – track down files, set
appointments, keep track of victims and witnesses – which constrains their ability to make timely,
well informed decisions. Experienced ADAs in Southern County estimated that about 50 percent of
ADA’s time is spent doing clerical and support staff work. ADAs in Northern County noted similar
problems with a lack of support staff, arguing that a lack of support staff means that prosecutors
have to spend a lot of time searching for files, typing letters, doing paper work and then have no
time for legal work. The result, according to experienced ADAs, is that the quality of legal work is
poor.
ADAs also noted that the lack of support staff particularly affected cases that involved a lot of
preparatory or investigative work, such as person and property cases; drug cases, on the other hand,
were less affected since they did not involve witnesses or victims. Since there was little support
staff to assist in following up with victims and witnesses, prosecutors said they are more inclined to
not issue a case in such circumstances. Moreover, the lack of support staff, particularly for
assistance with follow-up after a case was charged, was seen as more problematic for general
crimes teams and less problematic of specialized units such as drugs or sensitive crimes. Specialized
police units continued investigations for specialized prosecution units after charging; however,
general crimes prosecution units that dealt only with general crimes investigators did not get the
same level of investigation from the police. As discussed below, this was primarily due to recent
changes in the police departments in both jurisdictions.
Other practical constraints affected how cases were handled as well. For example, ADAs in both
jurisdictions noted a lack of case management systems that would allow them to keep track of cases.
Even in Northern County, which has a sophisticated case management system, the system was not

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 92

available to ADAs in court; ADAs were still required to carry paper files to court, which was, again,
hampered by a lack of support staff to track and store files. Moreover, the lack of laptops, TVs, CD
burners, etc. prevented the effective presentation of electronic evidence in court, which, in turn,
affected the type or quality of evidence available to present.
Despite the myriad of problems potentially associated with lack of staff support, prosecutors
who responded to the general survey were far from unanimous in their evaluations of the
consequences of such problems. Only 18 percent of respondents said that a lack of investigators
frequently or very frequently affected the outcomes of their cases (Figure 2.5.2-1). Less than half of
respondents (45 percent) indicated that lack of support staff to coordinate contacts with victims,
witnesses, and defense counsel frequently or very frequently adversely affected their cases (Figure
2.5.2-2). Lack of support staff to coordinate contacts was seen as frequently or very frequently
problematic by a significantly higher percentage of Southern County prosecutors (60 percent) than
Northern County prosecutors (30 percent), and in Southern County, average ratings of the frequency
of these problems tended to increase with increasing levels of experience.
Prosecutors’ greatest concerns over lack of staff resources appeared to be centered around
paperwork, paper file maintenance, and the implications of the lack of technology resources for
clerical workload. Nearly two-thirds of respondents indicated that there was very rarely or never
enough clerical staff to assist with paperwork on all their cases, and only 18 percent indicated there
was frequently or very frequently enough clerical assistance. Forty-six percent of respondents
indicated that the lack of technology resources frequently or very frequently made work difficult,
whereas only 14 percent said it was very rarely or never a problem.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 93

Figure 2.5.2-1 Responses to question: “The lack of investigators in my office affects the
outcomes of my cases,” by research site
Percentage of Respondents within County

60
55
50
45

40
35
30
25
20
15
10
5
0
Never

Very rarely

Northern County (N=59)

Occasionally

Frequently

Very frequently

Southern County (N=60)

Percentage of Respondents within County

Figure 2.5.2-2 Responses to question: “The lack of support staff to coordinate contacts with
victims, witnesses, and defense counsel adversely affects my cases,” by research site
60
55
50
45
40
35
30
25

20
15
10
5
0
Never

Very rarely

Northern County (N=60)

Occasionally

Frequently

Very frequently

Southern County (N=63)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 94

Figure 2.5.2-3 Responses to question: “There is enough clerical staff to assist me with
paperwork on all my cases,” by research site
Percentage of Respondents within County

60
55
50
45

40
35
30
25
20
15
10
5
0
Never

Very rarely

Northern County (N=59)

Occasionally

Frequently

Very frequently

Southern County(N=63)

Figure 2.5.2-4 Responses to question: “The lack of technology in the office makes work
difficult,” by research site
Percentage of Respondents within County

60
55

50
45
40
35
30
25
20
15
10
5
0
Never

Very rarely

Northern County (N=61)

Occasionally

Frequently

Very frequently

Southern County(N=64)

A final problem of internal resources expressed in Northern County was the lack of ADAs and
resulting high caseloads. As one unit manager noted, high caseloads forced prosecutors to decide to
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 95

get rid of some cases, to not prosecute certain cases. ADAs recognized the problem as well, noting
that the high caseloads prevented them from devoting enough time to cases. Although a majority of
the prosecutors who responded to the general survey (60 percent) indicated that caseloads
frequently or very frequently prevented them from devoting enough time to all their cases, a
majority also indicated that they very rarely or never declined or dismissed cases when the amount
of time and effort needed to obtain a conviction exceeded the benefits of the potential sentence
(Figure 2.5.2-5; Figure 2.5.2-6). In both jurisdictions, prosecutors noted that a lack of resources did
not affect whether a case would be accepted. As the deputy in Southern County noted, “We work
with the idea that we do not spend money on cases we will not win. So we push police to exhaust all
resources on investigation before we make that decision. But to take a perfectly good winnable
felony and say that we will not accept it because we do not have resources – we do not do that.”
Figure 2.5.2-5 Responses to question: “Caseloads prevent me from devoting enough time to all
my cases,” by research site
Percentage of Respondents within County

60
55
50
45
40
35
30
25
20
15
10
5
0
Never

Very rarely

Northern County (N=62)

Occasionally

Frequently

Very frequently

Southern County (N=65)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 96

Figure 2.5.2-6 Responses to question: “I decline or dismiss cases when the amount of time and
effort need to obtain a conviction exceeds the benefits of the potential sentence,” by research
site
Percentage of Respondents within County

60
55
50
45
40
35
30
25
20

15
10
5
0
Never

Very rarely

Occasionally

Northern County (N=61)

Frequently

Very frequently

Southern County (N=64)

Again, these findings are only indicative of prosecutors’ opinions and may represent general
dissatisfaction with perceived resource constraints; in other words, while prosecutors maintain that
resource constraints influence decision making, in actuality they may have little real import. This
seems particularly true for internal resource constraints. While prosecutors may argue that a lack of
support staff or perceived inadequate staffing affects cases, on the general survey few admitted that
these constraints actually led them to handle cases differently. Rather, prosecutors described a lack
of internal resources generally leading to additional paper work and clerical work rather than an
actual need to dispose of cases. In contrast, external resource constraints clearly affected case
outcomes. In Southern County, prosecutors at all levels discussed several specific ways in which the
lack of available court space led directly to the re-evaluation of evidence and the eventual dismissal
or downgrading of plea offers in many cases.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 97

2.5.3 Relationships
Prosecuting cases relies on a web of relationships with other justice system actors – law
enforcement, defense attorneys, judges, other prosecutors. The quality of relationships with these
actors, like rules and resources, may affect how a prosecutor approaches a case. While a good or
bad relationship may not determine the outcome of a case, these relationships certainly affect how a
case may be handled or how certain information is evaluated. Moreover, routine interactions with
other system actors creates a set of expectations, for example, about how a specific judge will
sentence or what types of pleas a specific defense attorney will be receptive too. In this sense,
relationships with other actors may influence the outcome of a particular case.
Judiciary
Prosecutors in both jurisdictions noted that the judge on a particular case can affect how the case
is handled. In several instances, ADAs noted that prosecutors had to know their judge. This meant
that prosecutors had to know, for example, what the judge would view as an acceptable charge,
what motions would likely be looked on favorably, or what evidence the judge would expect to see.
As one newer ADA in Southern County noted, “There is a diverse array of what judges want and do
not want. A judge will let you know what they want. You need to learn the different styles of judge
and develop approaches to everything.” As another newer ADA noted, “The judges have different
stands on different issues. Certain statutes they do not necessarily agree with. There is certain
evidence that they will not accept and certain evidence that they expect to see.” However, the ADAs
were quick to maintain that the particular judge on a case did not determine decisions; however,
they also stated that they know what judges will do and tailor charges, pleas, etc. based on that.
Indeed, as one ADA in Northern County noted, “We must work within the confines of the system –
who the judge is will affect pleas and trials. Judges affect what ADAs CAN do.”

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 98

Surprisingly, only 6 percent of prosecutors who responded to the general survey indicated that
they frequently tailored their decisions to fit the expectations of judges, while 30 percent said they
very rarely or never did. Moreover, only three percent of respondents indicated that the specific
charges they filed in a case were affected by the judge that will hear the case, while 88 percent
indicated that the specific charges they filed were very rarely or never affected by the judge that will
hear the case. In spite of this apparent consensus among the prosecutors who responded to the
general survey, prosecutors who participated in the focus groups explained in some detail a variety
of ways that knowing what judge will hear a case and understanding that judge’s style and
preferences can affect how a prosecutor handles a case—apparently, according to the focus group
participants, including what charges to file.
The ADAs in Southern County did state that a case may be continued if assigned to a particular
judge. While newer ADAs claimed that they did not judge shop, they also admitted that
occasionally they would continue a case if they got a judge with whom they did not agree; in some
instances they would punt a case and continue it until they were out of the courtroom, thus leaving
it for another ADA to handle. This was true only for newer ADAs in Southern County assigned to
the misdemeanor unit; in misdemeanor court, ADAs appear before the same judge all the time. As
such, they have a good sense of the judge. Unit managers and more experienced ADAs who handle
felonies in Southern County did not see the judge affecting a case. According to this group of
prosecutors it was impossible to tailor decisions to judges because they did not know who the judge
will be on a case. However, they did acknowledge that the judge may affect which motions to file
and when a case is called; overall, it may affect the approach but not a decision (e.g. what to offer
in a plea) because we do not know who the judge will be. Rather than tailoring a decision to a
particular judge, the more experienced ADAs in Southern County tailored decisions to the typical

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 99

judge. Prosecutors know what the extremes are in judges’ sentencing decisions and shoot for the
middle. Basically, there is a sense of what the going rate is among the collective of judges in the
district and prosecutors tailor decisions to what the going rate or norm is.
In contrast, in Northern County, all ADAs appear before the same judge for roughly one year
before the judge is rotated out. Moreover, ADAs screen cases that will be assigned to their teams;
thus, when screening and charging a case, the screening ADA knows who the assigned judge will
be. As a result, in all cases – felony and misdemeanor – ADAs are very familiar with how the judge
runs their courtroom. Some unit managers argued that, as a result, ADAs alter screening, charging,
and plea offer decisions based on the judge. As one unit manager noted, “ADAs are cognizant of
what judges are willing to accept in terms of charges and plea offers and what kinds of evidence the
judge will admit. All of this determines what charges to file, what offer to make, etc. because the
ADA gets tired of being rejected and fighting with the judge.” Indeed, a deputy in Northern County
hinted that tailoring decisions to the judge was a matter of efficiency, arguing that prosecutors have
to get a resolution with the defense attorney and knowing what the judge will accept is important.
While some unit managers agreed with this statement, others disagreed, arguing that “for the most
part, we issue cases based on what we can prove and do not modify it for the judge.” Indeed, an
ADA noted that “The judge should not pull the ADA; rather the ADA must pull the judge to the
appropriate place. Our job is to decide what is just and worthwhile regardless of what the judge will
do. The DA has the right to have put on the record what they think is just and fair.”
Prosecutors in both jurisdictions also argued that cases revolved around assessments of who the
“good” judges were. In this sense, prosecutors scheduled cases for trial in order to trust the “good”
judges with the cases prosecutors really care about. For example, if a case turns on complicated
issues, prosecutors may want a certain judge and, therefore, may not fight a defense request for a

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 100

continuance. In Northern County, ADAs had a slightly different notion of a “good” judge. As one
ADA pointed out, some judges make doing the job difficult; he maintained that it is not about
ideology, but about inconsistency in how the judge will respond to facts. Indeed, several prosecutors
noted that predictability or consistency in judicial decision making was very important. As another
ADA noted, if the judge is predicable, it is easier to negotiate with the defense; if the judge is
inconsistent, this makes it harder to make a deal. To combat such inconsistency, prosecutors may
try to limit the judge’s discretion by limiting the number of charges that the defendant must plead
to, which then limits the judge’s sentencing options. However, if the judge is predictable, then
ADAs may alter practice and avoid a sentence recommendation in a plea offer. In other words, if
prosecutors can trust the judge and are sure that he will be fair, they are willing to offer a plea of
“sentence at discretion of the court.”
Defense
While judges may affect what the prosecutor can do, the relationship with the defense attorney
may affect the ease with which prosecutors do their job. In Southern County, prosecutors had very
different opinions about their relationship to defense attorneys, with opinions varying by experience
level. For example, some newer ADAs, who work exclusively in misdemeanor court, had very
antagonistic relationships with defense attorneys. As one ADA noted, “The PD [public defender]
needs a good relationship with the ADA to get good deals. But if there is a bad relationship with a
PD, the worst that can happen for an ADA is that they will have to try every case, which is what I
love to do.” As this ADA stated, the antagonistic relationship often resulted in “punishing” the
public defender in some way, for example, calling cases early or scheduling several cases for that
particular defense attorney in a row to make it tougher for the defense attorney to prepare. Other
newer ADAs in Southern County saw the defense as very important, primarily as conduits for

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 101

information about cases; moreover, these newer ADAs noted that if defense attorneys were
cooperative, then they would get more cooperation from the ADAs in the form of uncontested
continuances or a break without an objection. Experienced ADAs saw the importance of a good
relationship with defense attorneys as well, but noted that a good relationship did not affect decision
making. Specifically, like newer ADAs, they argued that they did not tailor decisions to defense
attorneys, but were willing give trusted defense attorneys more time to prepare a case. They noted
that if they trust and respect a defense attorney, they will listen to what he has to say about a client
and may reconsider a plea offer or outcome. In contrast, for those defense attorneys that prosecutors
do not respect, they will make them work for it. However, prosecutors at all levels were quick to
point out that they do not make it tough for the defendant or punish the defendant for the behavior
of the attorney; they simply try to make it tough for the defense attorney.
More experienced prosecutors also noted the need for a good relationship with defense attorneys
to ensure an efficient system, noting that the defense can penalize the ADA with a lot of motions and
continuances. This was echoed by ADAs in Northern County as well, who argued that courthouses
were small enough towns that all parties know each other so that no one benefits from having a bad
relationship; but a good relationship gets a more efficient outcome. Indeed, unit managers in both
jurisdictions claimed that it was imperative that prosecutors work with the public defender’s office
to ensure case flow. According to one unit manager in Northern County, public defenders were
willing to bend the rules to make the system work smoother, and, in turn, prosecutors were willing
to bend rules for them. This bending of the rules pertained to procedural rules, such as changing the
way cases were set for trial to accommodate the public defender’s schedule.
Prosecutors generally saw the largest benefit of a good relationship with defense attorneys as a
better flow of information and a more just resolution of a case. According to several ADAs

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 102

prosecutors rarely have enough information about a defendant or his social background and that is
what a good defense attorney will give you. Given the limited amount of information with which
prosecutors work, they felt that a good relationship with defense attorney brought a great deal more
information to the case; an antagonistic relationship made defense attorneys less inclined to come to
ADAs with additional information, waiting until trial to introduce it. A good relationship with the
defense was also seen as benefiting the prosecution. The judge sees a good relationship between the
prosecutor and defense and is more willing to listen to an ADA's arguments. It narrows the issues if
you trust someone and you can focus only on what the differences are about, for example, why a
party is seeking a continuance or why one party is filing a particular motion.
The idea expressed in focus groups that prosecutors value good working relationships with
defense attorneys but do not change their decisions based on relationships alone was reinforced in
the responses of prosecutors to the general survey. Forty-three percent of respondents considered
good relationships with defense attorneys to be an important or very important criterion for
evaluating their individual success as prosecutors, while only 10 percent considered it to be of little
importance or unimportant (Figure 2.5.3-1). As a criterion for organizational success, 51 percent of
respondents considered good relations with the defense bar to be important or very important, while
only 6 percent considered it to be of little importance or unimportant (Figure 2.5.3-2). Nearly threequarters of respondents (72 percent) also indicated that they would very rarely or never consider
altering their decisions for defense attorneys they respect; yet, this indicates that some proportion of
prosecutors do alter decisions for defense attorneys (Figure 2.5.3-3). Eighty-five percent said they
would very rarely or never tailor their decisions to gain or maintain the trust of defense attorneys;
again, indicating that some proportion would (Figure 2.5.3-4).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 103

Percentage of Respondents within Coumty

Figure 2.5.3-1 Responses to question: “Good relations with defense attorneys is important to
individual success,” by research site
60
55
50
45
40
35
30
25
20
15
10
5
0
Unimportant

Of little importance

Northern County (N=62)

Moderately
important

Important

Very important

Southern County (N=65)

Figure 2.5.3-2 Responses to question: “Good relations with defense bar is important to office
success,” by research site
Percentage of Respondents within County

60
55
50
45
40
35
30
25
20
15
10
5

0
Unimportant

Of little importance

Northern County (N=61)

Moderately
important

Important

Very important

Southern County (N=65)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 104

Figure 2.5.3-3 Responses to question: “I would consider altering my decisions for defense
attorneys who I respect,” by research site
Percentage of Respondents within County

60
55
50
45

40
35
30
25
20
15
10
5
0
Never

Very rarely

Northern County (N=60)

Occasionally

Frequently

Very frequently

Southern County (N=64)

Percentage of Respondents within County

Figure 2.5.3-4 Responses to question: “I tailor my decisions to gain or to maintain the trust of
defense attorneys,” by research site
70
65
60
55
50
45
40
35
30
25
20
15
10
5
0
Never

Very rarely

Northern County (N=61)

Occasionally

Frequently

Very frequently

Southern County (N=63)

Thus, prosecutors do not alter their decisions merely because of good relationships with defense
attorneys. Rather, good relationships are valued for other reasons—because good working
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 105

relationships promote efficiency in case processing, and because mutual trust fosters open and caserelevant communication. Good communication, in turn, can affect prosecutors’ decisions indirectly
by providing prosecutors with important information about defendants or the circumstances
surrounding a case that may not otherwise come to their attention.
Law enforcement
Of primary concern in both jurisdictions was the relationship between prosecutors and law
enforcement officers. In both jurisdictions, relationships with law enforcement have deteriorated
somewhat in recent years, primarily due to a change in leadership in each jurisdiction’s primary
urban police department. This has affected the political relationship between law enforcement and
prosecution that has then trickled down to the daily interactions between individual police officers
and ADAs. In turn, this has influenced relationships that are already characterized by tension and
has introduced a new level of acrimony between the two groups.
As the DA in Southern County noted, there is an institutional tension that always exists with
law enforcement, since prosecutors have an obligation to scrutinize cases brought by law
enforcement and must decline to prosecute some cases. Generally, ADAs saw this tension as good.
As unit managers in Southern County noted, police officers should not agree with everything
prosecutors are doing and prosecutors should not accept everything that law enforcement brings.
Some ADAs in Northern County welcomed pressure from law enforcement because it shows that
officers care about the case.
However, this tension often resulted in pressure being applied to ADAs to accept cases.
Generally, this pressure was felt most directly by newer ADAs who may be less confident in their
ability to confront law enforcement officers or decline their cases; in turn, they may be more likely
to accept questionable cases due to such pressure. As the deputy in Southern County noted, young

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 106

ADAs feel tension when they have face to face interactions with officers because the ADA may have
to tell the officer “no”, which people do not say to law enforcement; he pointed out that it is difficult
to instruct young ADAs how to overcome this tension because of the authority of law enforcement
that comes with the gun, badge, uniform. The mechanisms for dealing with such pressure generally
involved making arguments about the justification for not taking a case or deflecting responsibility
for decisions to other actors in the system. Other ADAs make the “efficiency argument” that it will
be easier if they just disposed of the case and some make the “fairness argument” that the
defendant really does not deserve to be punished. In contrast, more experienced ADAs appear to be
more comfortable discussing the legal aspects of the case with officers. Experienced ADAs in
Southern County stated that the tactic is to let them know about problems since this is usually what
the tension is about. In some instances, prosecutors succumb to the pressure from law enforcement
and accept cases they would not otherwise accept. This happens even with experienced ADAs, who
admitted that they had on several occasions accepted bad cases due to pressure during screening.
Often acceptance of weak cases derives from pressure applied by officers with whom prosecutors
have a good relationship. Nonetheless, ADAs were instructed to build good working relationships
with law enforcement. Much like defense attorneys, law enforcement was seen as a conduit for
quality information.
This contrast between operational tensions between police and ADAs and a philosophical
commitment to good working relationships was also evident in responses to the general survey.
Over half of respondents (56 percent) considered good relationships with law enforcement officers
to be an important or very important criterion for evaluating their individual success, while only 5
percent considered it to be of little importance or unimportant (Figure 2.5.3-5). As a criterion for
organizational success, nearly three-quarters of respondents (72 percent) considered good relations

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 107

with law enforcement agencies to be important or very important, while only 6 percent considered it
to be of little importance or unimportant (Figure 2.5.3-6). The survey results were somewhat mixed
with respect to case level interactions with the police officers who submit cases for prosecution.
Thirty-six percent of respondents indicated that they frequently or very frequently felt pressure from
law enforcement officers to accept cases for prosecution and 50 percent said it occurred
occasionally (Figure 2.5.3-7). Average ratings of the frequency of occurrence did vary significantly
across levels of experience (p=.021, F=3.364, df=3), but did not follow the expected pattern (Figure
2.5.3-8). Instead, the perceived frequency of pressure from police officers was lowest among ADAs
with less than one year experience,6 then peaked among ADAs with 1 to 5 years’ experience and
declined with increasing experience thereafter.
Figure 2.5.3-5 Responses to question: “Good relationships with law enforcement officers is
important to individual success,” by research site
60
55

Percentage of Respondents within County

45
40
35
30
25
20
15
10
5
0
Unimportant

Of little importance

Northern County (N=62)

Moderately
important

Important

Very important

Southern County (N=65)

The apparently anomalous results for the least experienced group could be due to the very small number of
respondents in that group. However, the same pattern was found in both jurisdictions.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 108

Figure 2.5.3-6 Responses to question: “Good relations with law enforcement agencies is
important to office success,” by research site
Percentage of Respondents within County

60
55
50
45
40
35
30
25
20
15
10
5
0
Unimportant

Of little importance

Northern County (N=61)

Moderately
important

Important

Very important

Southern County (N=65)

Figure 2.5.3-7 Responses to question: “I feel pressure from law enforcement officers to accept
cases for prosecution,” by research site
Percentage of Respondents within County

60
55

50
45
40
35
30
25
20
15
10
5
0
Never

Very rarely

Northern County (N=62)

Occassionally

Frequently

Very frequently

Southern County (N=63)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 109

Figure 2.5.3-8 Responses to question: “I feel pressure from law enforcement officers to accept
cases for prosecution,” by research site and ADA experience

Mean Frequency Rating

5
4
3
2
1
0
Less than 1 year

At least 1 year but <5
years

At least 5 years but <10
years

10 years or more

ADA years of experience in current DAs office
Northern County

Southern County

While prosecutors recognized the conflicts in working relationships with law enforcement, some
did acknowledge that the law enforcement officer assigned to a case does impact the way a case is
handled. For example, if an officer is not very good at testifying on the stand, ADAs will avoid
calling the officer to testify at trial. This may affect plea offers, as ADAs seek to dispose of a case
in order to avoid having an officer testify. As one experienced ADA in Southern County noted,
prosecutors learn quickly the difference between officers who want to do a good job and those who
do not and that is considered when prioritizing cases. This was echoed in Northern County where
prosecutors noted that if they have less confidence in the use of police at trial, they may work harder
to get a plea. Another ADA in Northern County argued that past experiences or knowledge of
particular cops affects a case from start to finish. According to this ADA, prosecutors were less
likely to charge a case if a “bad cop” is on the case and more likely to give a light offer in a plea
when a “bad cop” is involved.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 110

Both jurisdictions have also been affected by changes in the largest municipal police
departments in each county. According to all respondents in each jurisdiction, the municipal police
department focuses on arrests but not investigations of cases. As a result, according to nearly all
respondents, the quality of information and the quality of cases coming from the police department
has deteriorated. In turn, many cases are not prosecuted or are dismissed because of a lack of quality
information.
In Northern County the change in focus from investigations to arrests accompanied a
reorganization within the police department, which affected the quality of information delivered to
prosecutors. According to the DA in Northern County, the police chief disbanded the specialized
units and put the emphasis on response to crime rather than resolution of cases; as a result, the
quality of cases is going down dramatically because the most experienced law enforcement officers
are not working on investigating cases. According to the DA, all of this leads to high no process
and high dismissal rates in the office. The reorganization in the police department has also affected
the types of cases the DA’s office gets. As one deputy in Northern County commented, the DA’s
office can only deal with the cases brought to them by the police department. And according to
many, the police department has started to deemphasize certain cases such as officer-initiated
offenses and undercover police investigations. For the drug unit, specifically, this meant a
significant decrease in large-scale drug offenses and an increase in low-level drug cases.
The deputy in Southern County pointed out a similar pattern of law enforcement focusing on
arrest and not investigation. As the deputy noted, this had practical implications for the DA’s office:
the police officer’s goal is to get probable cause to make an arrest, but they do not finish the
investigation necessary to move from probable cause to beyond a reasonable doubt. In both
jurisdictions, prosecutors described increasing conflict with officers on a regular basis. This was

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 111

echoed by experienced ADAs in the office as well, who maintained that officers get increasingly
focused on the arrest but not the investigation and that this has gotten worse with the new chief. In
Southern County, the DA responded by putting experienced ADAs at the screening desk who know
what to look for and can get more respect from the officer. As a result of more stringent screening
processes, the primary police department in Southern County has also stopped seeking pre-warrant
approval by the district attorney’s office; in the past, police would get cases pre-screened by the
DA’s office before seeking a warrant to arrest a suspect. Given the change in focus of the police to
make arrests and the increased scrutiny in screening by the DA’s office, police are now avoiding the
pre-warrant screening. As a result, many more cases are declined at the screening stage. The result
of changes in law enforcement in both jurisdictions has been an increase in cases declined for
prosecution, pended for additional information, or dismissed due to poor follow-up investigation.
Colleagues and supervisors
Relationships to colleagues also impact decision making. ADAs often look to colleagues to
gauge performance and to ensure that their decisions are within office norms. At one level,
colleagues communicate the over-riding philosophy of the department. As one experienced ADA in
Southern County noted, “There are a lot of different ideas of justice – to the victim, to the
community, to the defendant. It takes a lot of experience to balance these. A person who only
pushes for one type of justice or with one view of justice will not survive because their offers and
decisions will be outside the norm.” The office generally relies on colleagues to train new ADAs.
As the deputy in Southern County noted, ADAs learn from their peers and from supervisors about
how this office wants them to make decisions. For example, colleagues and supervisors let new
ADAs know the range of acceptable pleas or what offers should be accepted. Moreover, the reliance

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 112

on colleagues is by design; supervisors want ADAs to look to their colleagues about what to do and
what is acceptable.
But colleagues affect decisions not just through training, but as barometers for what is
acceptable or the right decision in a case. As one new ADA noted, new ADAs are often concerned
about doing something wrong or something that someone doesn’t like. Unit managers expressed this
as a desire to make sure decisions were consistent with those of colleagues. As one unit manager
noted, colleagues definitely affect decisions – you want your decisions to be consistent. It shapes
what we do even if it does not dictate actual decisions…We do not want to make decisions that you
will have to explain later.
The importance of relationships with colleagues and supervisors was also reflected in
prosecutors’ responses to the general survey. There was considerable consensus among respondents
that having the respect of their colleagues and supervisors were important or very important criteria
for evaluating their own personal success. Surprisingly, though, a majority of respondents indicated
that they very rarely or never make decisions based on how their colleagues perceive their
performance (85 percent) and that the decisions they make are very rarely or never affected by how
they think they will be perceived by their supervisors (52 percent) (Figure 2.5.3-9 and Figure 2.5.310). Nevertheless, non-negligible percentages of respondents did indicate that their decisions were
occasionally affected by the perceptions of colleagues and supervisors, and the average frequency
ratings declined significantly with increasing levels of experience for both the perceptions of
colleagues (p=.001, F=5.741, df=3) and the perceptions of supervisors (p=.046, F=2.745, df=3)
(Figures 2.5.3-11 and 2.5.3-12). This latter finding suggests that the least experienced prosecutors
rely somewhat on the judgments of their colleagues and supervisors, while the more experienced
prosecutors are perhaps more comfortable with their own judgments.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 113

Figure 2.5.3-9 Responses to question: “I make decisions based on how my colleagues will
perceive my performance,” by research site
Percentage of Respondents within County

60
55
50
45

40
35
30
25
20
15
10
5
0
Never

Very rarely

Northern County (N=61)

Occassionally

Frequently

Very frequently

Southern County (N=65)

Figure 2.5.3-10 Responses to question: “The decisions I make are affected by how I think they
will be perceived by my supervisor,” by research site
Percentage of Respondents within County

60
55
50
45
40
35
30
25
20
15
10
5
0
Never

Very rarely

Northern County(N=61)

Occasionally

Frequently

Very frequently

Southern County (N=65)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 114

Figure 2.5.3-11 Responses to question: “I make decisions based on how my colleagues perceive
my performance,” by research site and ADA experience

Mean Frequency Rating

5
4
3
2
1
0
Less than 1 year

At least 1 year but <5
years

At least 5 years but <10
years

10 years or more

ADA years of experience in current DAs office
Northern County

Southern County

Figure 2.5.3-12 Responses to question: “The decisions I make are affected by how I think they
will be perceived by my supervisor,” by research site and ADA experience

Mean Frequency Rating

5
4
3
2
1
0
Less than 1 year

At least 1 year but <5
years

At least 5 years but <10
years

10 years or more

ADA years of experience in current DAs office
Northern County

Southern County

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 115

2.6 Understanding the Prosecutor’s Perspective
According to the prosecutors in Northern and Southern Counties, prosecution involves
continually considering two fundamental questions: ’Can I prove the case?’ and ‘Should I prove the
case?’ Answering the first question requires the seemingly objective evaluation of the evidence—a
case can proceed if the evidence is strong enough to obtain a conviction. Prosecutors must answer
the first question in the affirmative before the second question can even be asked; but the fact that a
case can proceed does not imply that it should proceed. Answering the second question requires the
subjective evaluation of the merits of the case. To determine whether a case should be prosecuted,
prosecutors said they consider multiple factors, including the seriousness of the offense, the
defendant’s criminal history, and other factors internal to a case.
But prosecutors also noted that cases do not exist in a vacuum; rather, individual cases exist
within a system that often constrains prosecutorial decision-making. Internal rules or policies within
the prosecutor’s office sometimes determine whether a case is accepted for prosecution or how to
craft an appropriate plea. The lack of resources of the prosecutor’s office and the local court system
may require prosecutors to reject, dismiss, or amend charges in order to work within available
resource limits. Relationships with law enforcement officers, judges, and defense attorneys may
alter how a case will be handled. These constraints may trump evaluations of strength of the
evidence, severity of the offense, and defendant criminal history, forcing prosecutors to make
decisions that they may not make in the absence of such constraints.
In Part 3 we explore the effect of case-specific factors on case outcomes. Although we do not
examine empirically the impact of rules, resources, and relationships on case outcomes, we begin to
explore how prosecutors view other case factors at different stages of the prosecutorial process and
how prosecutor characteristics may influence variation in case outcomes.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 116

Part 3. The Stages of the Prosecutorial Process: Analyzing Case Outcomes
3.1. Introduction
Prosecutors’ perspectives provide the necessary context for understanding how prosecutorial
decisions are made and for illuminating the factors that prosecutors report they reference when
making decisions. A separate question remains regarding how much these factors affect the
outcomes of cases. A growing body of research has empirically examined the extent to which legal
and extra-legal factors affect prosecutorial outcomes. Studies have shown that variation in outcomes
is associated primarily with legal factors, including the type of offense (Albonetti, 1987; Hartley et
al., 2007; Jacoby et al., 1982a; Schmidt & Steury, 1989), strength of the evidence (Albonetti, 1987;
Jacoby et al., 1982a; Spohn & Holleran, 2001), and defendant culpability (Adams & Cutshall, 1987;
Albonetti, 1987; Schmidt & Steury, 1989). Several studies have also found that variation in
outcomes is associated with extra-legal factors, including defendant race (Chen, 2008; Free, 2002;
Frohmann, 1997; Hartley et al., 2007; Ulmer et al., 2007), defendant gender (Albonetti, 1986;
Alozie and Johnston, 2000; Bjerk, 2005; Farrell, 2003; Frohmann, 1997; Miethe, 1987; Hartley et
al., 2007; Ulmer et al., 2007), and characteristics of the victim (Albonetti, 1986; Kingsnorth &
MacIntosh, 2004; Schmidt & Steury, 1989; Spears & Spohn, 1997; Spohn & Holleran, 2001;
Stanko, 1981-82).
The following sections explore the impact of legal and extra-legal factors on case outcomes in
Southern and Northern Counties. The analyses explore several outcomes throughout the
prosecutorial process: screening, charging, plea offers (including sentence recommendations), and
dismissals. This part of the report is organized into four sections, each addressing one of these four
decision points. Within each section, we present prosecutors’ perspectives on the respective

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 117

decision point; we then turn to analyses of case outcomes to examine the influence of legal and
extra-legal factors.

3.2. Summary of Research Methods: Administrative Data, Factorial Survey, Focus
Groups, and Surveys
The analyses of case outcomes rely primarily on two strategies: an analysis of administrative
data documenting outcomes in actual cases and an analysis of a factorial survey documenting
prosecutors’ responses to hypothetical cases. These analyses were supplemented with additional
information gleaned from focus groups and the general survey of prosecutors. (For a detailed
description of methods, see Appendix A.).
3.2.1 Examining Case Outcomes in Southern County
The analyses of case outcomes in Southern County relied on individual-level administrative data
collected from two sources. Data pertaining to the processing of felony cases assigned to the drug
crimes unit were extracted from an automated case management system maintained by the District
Attorney’s office. Data pertaining to felony cases assigned to the person crime unit and property
crime unit were coded manually from paper case files using a structured data entry routine with preestablished response lists for the coded data fields. Data from both sources identified unique
individuals, individual complaints, individual charges within complaints, and unique “cases”
(defined as a complaint-defendant combination).
The drug unit data tracked the status of individual charges from intake to initial screening,
through acceptance or rejection at screening, lower court processing, grand jury preparation, grand
jury decision, plea offer, felony court processing, and final disposition. The person unit and
property unit data collected from paper case files were less finely articulated, capturing the point at
which a charge first appeared in the file and the latest point at which it appeared, for initial
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 118

screening, at some undefined point after acceptance but prior to entry to upper (felony) court, at the
plea offer, and at some undefined point in upper court processing following the plea offer. Data
from both sources included basic defendant characteristics (age, race, gender), charge details
(offense type and statutory classification), reasons for dismissal, case-level disposition, statutorily
defined prior record level as reported in the plea offer, prosecutors’ sentence recommendations, and
the identities of the assigned prosecutors. While the data coded from paper case files were less
precise than the drug unit data with respect to processing stages, the paper case files yielded
additional information not available for the drug cases, most notably characteristics of the victims,
victim-offender relationships, victims’ willingness to testify, and physical evidence cited in the
police report. However, neither source yielded information about the characteristics of the
prosecutors.
The drug unit data included information on all closed felony drug cases screened by the office
from May 1, 2007 through July 31, 2009. These included a total of 12,225 unique charges in 4,890
unique cases. The person unit and property unit data included all closed felony cases initially
forwarded from the largest urban police agency in Southern County and screened between January
1, 2007 and June 30, 2007. Homicide cases were excluded during data collection and cases in which
the top charge was for kidnapping or any sex offense were excluded later during data file
preparation. The final person unit sample included 1,118 unique charges in 508 unique cases; and
the property unit sample included 1,316 unique charges in 658 unique cases. Since cases often
involved multiple charges, a procedure was devised to categorize and track cases according to the
“controlling offense” at each stage of the prosecutorial process based on offense severity (using a
11-point severity scale, ranging from 1=most serious felony to 11=any misdemeanor).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 119

Analyses were conducted for the following outcome measures: whether a case was accepted or
rejected at screening (0=rejected, 1=issued); the number of charges filed for each case accepted for
screening (continuous); the statutory rank of the top filing charge (ordinal, 1=most serious, 11=least
serious); the number of charges requiring guilty pleas in the formal plea offer (continuous); the
statutory rank of the top plea offer charge (ordinal, 1=most serious, 11=least serious); whether the
plea offer recommended a period of incarceration (0=no incarceration recommended,
1=incarceration recommended); whether all charges in a case were disposed in favor of the
defendant—that is, no conviction in the case (0=convicted on one or more charges, 1=all charges
disposed in favor of the defendant); potential exposure to incarceration if sentenced to incarceration
for top charge (in months); and aggregate exposure to incarceration if sentenced to incarceration for
all charges (in months).
Potential explanatory variables incorporated in the analyses included: four measures of
defendant characteristics (age, race, gender, and a statutorily defined prior record level); five
measures of case seriousness (number of intake charges, statutory rank of the top intake charge,
whether the top intake charge was robbery, whether the top intake charge was burglary, and whether
there were codefendants in the case); over a dozen measures of type and amount of physical
evidence (e.g., total number of items of evidence, number of items of evidence per charge, presence
of evidence with the potential for forensic identification, and items of evidence relating to weapons,
drugs, and several other categories); and over two dozen measures of victim characteristics and
victim-offender relationships (e.g., victim age, race, and gender; total number of victims; number of
victims willing to prosecute; and victims to whom the defendant was a stranger).
The influence of the above factors on case outcomes was analyzed using standard multivariate
regression techniques. For outcomes measured with ordinal or interval measures, we estimated

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 120

standard ordinary least squares (OLS) linear regression models. For the three outcomes measured in
dichotomous measures—the probability of acceptance at screening, the probability that a plea offer
recommended a period of incarceration, and the probability of non-conviction—we estimated
binary logistic models (Hosmer & Lemeshow, 2000). Regression analyses were conducted
separately for person unit cases, property unit cases, and drug unit cases. Each of these analytic
strategies is explained in more detail in sections reporting the associated analyses.
3.2.2 Examining Case Outcomes in Northern County
The analyses of case outcomes in Northern County relied on individual-level administrative data
collected in the case management system of the District Attorney’s office. These data identified
unique individuals, charges, and cases and featured a number of substantive data fields: offender
demographics (age, race, gender), victim demographics (age, race, gender), charge characteristics
(arresting agency, arrest offense, charged offense, offense disposition, charge enhancers), and
prosecutor information (arbitrary prosecutor ID, prosecutor unit). The data included information on
all misdemeanor and felony arrest charges screened by the office between January 1, 2009 and June
16, 2011.
The initial sample included 111,704 unique arrest charges. Each charge carried a unique
identifier which linked all charges corresponding to a single case/defendant. The sample included a
total of 77,987 unique cases/defendant records (hereafter, cases). This study sample was further
reduced once covariates were introduced into analytic models and cases were deleted due to missing
data for some covariates (listwise deletion was employed). The final sample included 110,437
unique arrest charges and 76,721 unique cases. Since cases often involved multiple charges, a
procedure was devised to categorize and track cases according to the “controlling offense” at each

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 121

stage of the prosecutorial process based on offense severity (using a 13 point severity scale, ranging
from ordinance violation=0 to Class A Felony=13).
The analyses of case outcomes focused on measures at three stages of the prosecutorial process:
screening, charging, and dismissal. At screening, a single dichotomous variable measured whether a
case was accepted for prosecution (0=rejected, 1=issued). A case was considered accepted for
prosecution if any single arrest charge was prosecuted. In other words, all arrest charges had to be
rejected for a case to be considered rejected for prosecution, but any single charge accepted was
sufficient for a case to be considered accepted for prosecution. At charging, two outcome measures
were used to measure the severity and scope of charges issued: most serious arrest charge issued
and number of charges reduced. A single dichotomous variable measured whether the most serious
arrest charge was one of the charges issued at initial charging (0=most serious arrest charge
rejected, 1=most serious arrest charge issued). For cases involving multiple arrest charges, a
measure of the reduction in the number of charges issued was created; this compared the number of
charges issued to the number of arrest charges screened. If a case has fewer issued charges than
arrest charges, it was considered to have a charge reduction. The measure was operationalized as a
dichotomous variable (0=number of charges not reduced, 1=number of charges reduced). These
measures were used to measure outcomes only for cases that were issued; in other words, cases
rejected for prosecution at screening were not included in the charging analyses. Finally, a single
dichotomous variable measured whether a case was dismissed after charging (0=not dismissed,
1=dismissed). A case was considered dismissed if all issued charges within the case were dismissed.
In other words, all issued charges had to be dismissed or otherwise disposed of without a guilty
verdict for a case to be considered dismissed.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 122

Several variables measuring defendant characteristics, victim characteristics, offense
characteristics, and case characteristics were derived from the administrative data and were used a
as predictor variables in outcome analyses. The influence of these factors on case outcomes was
analyzed using hierarchical linear modeling (HLM) procedures designed to account for the nested
nature of multilevel prosecution data. A two-level hierarchy represents the current data, with
individual cases nested within prosecutors. All variables were centered on their grand means and
results reported are based on offense-specific models using robust standard errors (Raudenbush &
Bryk, 2002, pp. 276–280). Seven separate models were run, one for each offense type – person,
property, drugs, public order, domestic violence, weapons, and DUI. This allowed us to examine the
impact of defendant, victim, offense, and case characteristics on case outcomes varied by offense
type.
Several tables of descriptive statistics of cases at each stage of the prosecution process in
Northern County are presented in Appendix C.
3.2.3 Factorial Survey
The term “factorial survey” refers to an approach in which respondents are asked to make
judgments about a structured set of hypothetical cases or “vignettes” (Jasso, 2006; Rossi &
Anderson, 1982). In this study, each participant was asked to respond to 10 vignettes, each of which
described the circumstances surrounding an arrest (summary of incident, offenses charged by the
police, evidence presented to the prosecutor by the police, suspect and victim information). These
base scenarios were selected from a standard case set developed and studied extensively by Joan
Jacoby and colleagues (Jacoby, Mellon, Ratlidge, & Turner, 1982b). A list of 30 vignettes from this
standard case set were ranked by offense seriousness and strength of evidence, classifying them as
low, medium, or high on each dimension; one vignette was then selected from each of the nine
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 123

combinations of offense seriousness and evidence levels and an additional vignette was added to
ensure inclusion of a drug offense. The vignettes presented to prosecutors in the factorial survey
represented several different offense types (person, property, drugs). All respondents received the
same base scenarios, but two factors were systematically varied across vignettes – defendant
criminal history and defendant race. As a result, each respondent received a packet of 10 vignettes
that was unique with respect to combinations of the basic scenarios, defendant prior criminal
history, and defendant race. The resulting classification of vignettes is displayed in Table 3.2.3-1.
Table 3.2.3-1 Arrest charges in factorial survey vignettes, classified by strength of evidence
and seriousness of top arrest charge
Seriousness of Top
Arrest Charge

High-level felony

Low
Vignette #1

Strength of Evidence (Design Level)
Medium
High
Vignette #4
Vignette #6

-Burglary
(Residential)
-Assault

-Burglary
(commercial)
-Theft
-Possession of a Stolen
Vehicle

Vignette #2

-Aggravated Assault
with a Deadly
Weapon (a pistol)
-Carrying Concealed
Weapon
-Assault in the Third
Degree (2 counts)
Vignette #5

-Robbery 2
-Conspiracy

-Robbery 1
-Conspiracy

-Forgery (4 counts)
-Theft: under $300 (4
counts)

Low-level felony

Vignette #9

Vignette #7
-Possession of a
Controlled Substance
(Heroin)

Misdemeanor

Vignette # 8

Vignette #3

Vignette #10

-Failure to move on

-Criminal Mischief
-Criminal Trespass

-Possession of a stolen
credit card
-Attempted Illegal Use
of a Credit Card

The resulting vignettes were reformatted to resemble a standardized arrest report. Each
hypothetical report was accompanied by a questionnaire, asking prosecutors to indicate whether

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 124

they would accept or reject the case for prosecution, give their reasons for rejection if applicable,
specify the charges they would file initially, specify the charges to which the defendant must plead
guilty in connection with a plea bargain, and indicate what sentence recommendation they would
make, if any. The packet also included a questionnaire asking the respondents about some of their
characteristics: age, race, gender, and professional experience in various capacities as attorneys
handling criminal cases. The instructions to respondents, the survey questionnaires, and arrest report
templates for each vignette are included in Appendix E. We were also able to link participants’
responses to the factorial survey with their responses to the general survey. This made it possible to
include information about prosecutors’ beliefs and attitudes in the analyses of their decisions in the
hypothetical cases presented in the factorial survey.
In Southern County, participation in the factorial survey was solicited from 67 prosecutors who
had previously responded to the general survey. Sixty-two prosecutors completed the survey for a
response rate of 93 percent. Each respondent answered questions about 10 different case vignettes,
yielding up to 620 observations for each question. Unfortunately, the number of responses to the
factorial survey received from Northern County was not adequate to reflect the variation of factors
built into the survey design and was not sufficient to support valid statistical analyses of the
influences of those factors. Only 18 responses were received initially from Northern County, and a
concerted follow-up effort yielded only three additional responses. Consequently, only the factorial
survey data from Southern County were analyzed.
Six dependent variables were defined for statistical modeling purposes: whether the respondent
would reject the case at screening (0=accept, 1=reject); the number of filing charges the respondent
listed (continuous); the rank of the statutory class of the top filing charge (ranging from 1=most
serious to 11=least serious); the number of charges to which the plea offer would require a guilty

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 125

plea (continuous); the rank of the statutory class of the top plea offer charge (from 1=most serious
to 11=least serious); and whether or not the sentence recommendation would include a period of
incarceration (0=no incarceration recommended; 1=incarceration recommended).
The independent variables retained for analysis included case seriousness (misdemeanor arrest,
low level felony arrest, or high level felony arrest), prosecutor’s rating of the strength of evidence,
prosecutor’s rating of the seriousness of defendant criminal history, the respondents’ years of
experience in the current prosecutor’s office, and several measures and items pertaining to
respondents’ beliefs and attitudes from the general survey (see Appendix A for a description of
these measures). Defendant age was confounded with defendant criminal history, and defendant
gender was determined by the facts of the case in a given scenario, so neither of those factors was
incorporated in analyses as independent variables. In addition, preliminary analyses found no
associations of any of the outcome variables with defendant race, respondent race, respondent age,
or respondent gender, so those variables were dropped early in the modeling process to conserve
degrees of freedom in the face of a small sample size.
In the factorial survey design adopted for this study, unique vignettes are nested within
respondents, and variables are measured at both the case level and the respondent level. The
appropriate approach for this design is a multi-level analysis using hierarchical modeling techniques
(Raudenbush & Bryk, 2002). To conduct these analyses, we used the Hierarchical Linear and
Nonlinear Modeling statistical package (HLM6; Raudenbush, Bryk, Cheong, & Congdon, 2004).
By keeping the models relatively simple, we were able to estimate multi-level binary logistic
models for the rejection decision and incarceration recommendation, but we were forced to settle
for estimating continuous dependent variable models for the other four outcome variables.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 126

3.2.4 Focus Groups and General Survey
Interpretation of our quantitative analyses of case outcomes were guided, in part, by discussions
in the same focus group sessions and general survey responses previously presented in Part 2. The
following sections draw on the participants’ more concrete discussions and responses concerning
the detailed stages in case processing – screening and charging cases, voluntarily dismissing or
adjusting charges, recommending sentences – as well as the procedures, processes, and strategies
that are employed at each stage. It also examines charging and plea bargaining strategies employed
by ADAs, variation in approach across prosecutors, the factors that govern the choice of strategy,
and prosecutors’ evaluations and use of case specific factors. The discussion of interviews and focus
group responses is not able to produce exact quotes in all instances; thus, phrases appearing in
italics are partial quotes or paraphrases derived from interviewer notes.

3.3 Case Processing Overview
Our quantitative analysis of the factors that affect case outcomes begins by examining selected
measures of those outcomes across key stages in case processing. The results are presented
separately for Southern County and Northern County because the available data captured outcomes
at different stages and for different crime types in the two counties.
3.3.1 Case Processing Overview for Southern County
In Southern County, the data described the processing of cases through several stages of the
prosecutorial process: initial screening (either as a result of arrests or in connection with pre-arrest
warrants), filing (charges either accepted at screening or introduced by the prosecutor shortly after
screening), indictment by the grand jury, plea offers, and final disposition. Table 3.3.1-1 presents
descriptive statistics on the number of cases processed through each of these stages. A “case” was
defined as a complaint-defendant; that is, if more than one defendant was named on a single
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 127

complaint, they were treated as separate cases in our analyses, regardless of whether they were
prosecuted together or separately. The drug unit cases were divided into two cohorts – the Early
Drug Unit Sample (cases entering May 2007 through March 2008) and the Late Drug Unit Sample
(cases entering April 2008 through July 2009) – to reflect a change in the unit manager supervising
the unit and a resulting change in policy (from a “loose” approach to screening to a “strict”
approach to screening.)
The percentage of cases accepted for prosecution varied substantially across crime types,
ranging from nearly 90 percent for the early drug unit sample to slightly less than two-thirds for the
property unit sample. The acceptance rate was slightly lower in the more recent drug unit sample.
Both the person and property unit samples were limited to earlier cases (screened from January
2007 through June 2007), so it cannot be determined whether acceptance rates have also declined
for cases screened by those two units. Despite the differences in case-level acceptance rates,
however, the average numbers of charges filed were quite similar across crime types.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 128

Table 3.3.1-1 Selected outcome measures by crime type and selected processing stages in
Southern County
Accepted/
Plea Offer:
b
Selected Outcome Measures
Screened
Filed
Indicted
Must Plead
Early Drug Unit Sample (Cases entering May 2007 through March 2008)
N of cases with final disposition
2475
2185
1527
1359
% of screened cases
100.0%
88.3%
61.7%
54.9%
Average N of charges
2.5
2.6
2.4
1.5
a
Avg. statutory rank of top charge
9.1
9.2
8.5
8.6
Late Drug Unit Sample (Cases entering April 2008 through July 2009)
N of cases with final disposition
2415
1981
1106
873
% of screened cases
100%
82.0%
45.8%
36.1%
Average N of charges
2.5
2.5
1.1
1.3
a
Avg. statutory rank of top charge
9.0
9.4
7.9
8.0
Person Unit Sample
N of cases with final disposition
508
372
317
253
% of screened cases
100%
73.2%
62.4%
49.8%
Average N of charges
2.2
2.4
2.5
1.6
Avg. statutory rank of top chargea
7.3
7.4
7.2
7.9
Property Unit Sample
N of cases with final disposition
658
426
289
249
% of screened cases
100%
64.7%
43.9%
37.8%
Average N of charges
2.0
2.1
2.3
1.7
Avg. statutory rank of top chargea
9.2
9.2
9.0
9.3

Convicted

1133
45.8%
1.4
8.9
842
34.0%
1.3
8.4
226
44.5%
1.5
8.2
224
34.0%
1.6
9.2

Highest rank = 1. Larger values indicate less serious statutory classification.

Initial charging includes a few charges active in lower court that were added after initial screening.

For all four samples examined, the average number of charges screened and the average number
of charges filed were nearly identical, indicating that when prosecutors accepted a case for
prosecution they tended to file all charges requested by the police. For the early drug unit, the
person unit, and the property unit samples, there also was little change in the average number of
active charges between initial filing and indictment, but in all three cases there were substantial
reductions in the average number of indicted charges and the average number of charges for which
the plea offer required guilty pleas; thus, prosecutors tend to retain all charges until the plea offer
and then reduce charges at that stage. The pattern was somewhat different for the late drug unit
sample, for which it appears that reductions in the average number of charges occurred before
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 129

indictment. For the person and property unit samples, the average statutory rank of the top charge
remained virtually unchanged throughout the process. However, the pattern was quite different for
drug unit cases, the average statutory rank of the top charge increased between initial filing and
indictment and then maintained through plea offer and ultimate conviction.
Overall, it appears that the decisions with the greatest consequences for defendants occurred at
the initial decision to accept or reject a case and in the preparation of plea offers. Consequently, our
analyses of factors affecting case processing decisions in Southern County focused most heavily on
the initial screening decision and the changes between initial filing and the plea offer.
3.3.2 Case Processing Overview for Northern County
The analyses of case outcomes in Northern County focused primarily on three stages of the
prosecutorial process: initial screening, filing, and final disposition (whether dismissed or
convicted). Table 3.3.2-1 presents descriptive statistics on the number of cases processed through
each of these stages. As above, a “case” was defined as a charge or set of charges for a single
defendant prosecuted as a result of a single complaint. Because of the detail available at the charge
level, it was possible to distinguish between cases in which all charges were dismissed prior to
adjudication and cases in which at least one charge was not dismissed and remained active to be
adjudicated by guilty plea or verdict at trial. Cases in which at least one charge remained active are
counted in the column labeled “Adjudicated” in Table 3.3.2-1, followed by a column that presents
results for cases in which at least one charge resulted in a guilty plea or a guilty verdict. Analyses
focused on cases in which the top charge was for one of seven crime types: person, property, drugs,
public order, domestic violence, weapons, or DUI.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 130

Table 3.3.2-1 Selected outcome measures by crime type and selected processing stages in
Northern County
Selected Outcome Measures
N of cases with final disposition
% of screened cases
Average N of charges
Avg. statutory rank of top chargea
N of cases with final disposition
% of screened cases
Average N of charges
Avg. statutory rank of top chargea
N of cases with final disposition
% of screened cases
Average N of charges
Avg. statutory rank of top chargea
N of cases with final disposition
% of screened cases
Average N of charges
Avg. statutory rank of top chargea
N of cases with final disposition
% of screened cases
Average N of charges
Avg. statutory rank of top chargea
N of cases with final disposition
% of screened cases
Average N of charges
Avg. statutory rank of top chargea
N of cases with final disposition
% of screened cases
Average N of charges
Avg. statutory rank of top chargea
a

Accepted/
Filed

Screened
Drugs
4,214
3,057
100%
72.5%
1.5
1.5
8.2
8.3
Person
2,761
1,082
100%
39.1%
1.3
2.1
6.9
6.6
Property
5,301
2,970
100%
56.0%
1.5
1.7
7.6
7.9
Public Order
9,221
6,358
100%
68.9%
1.2
1.4
9.6
9.7
Domestic Violence
9,582
2,433
100%
25.3%
1.1
1.5
9.8
9.7
Weapons
1,267
729
100%
57.5%
1.8
1.7
8.3
8.3
DUI
2,498
2,173
100%
87.0%
1.8
2.1
9.8
9.8

Adjudicatedb

Convicted

2,363
56.0%
1.6
8.2

2,320
55.0%
1.1
8.4

890
32.2%
2.3
6.3

849
30.7%
1.3
6.7

2,518
47.5%
1.8
7.8

2,490
46.9%
1.2
7.7

4,633
50.2%
1.5
9.7

4,262
46.2%
1.0
9.7

1,526
15.9%
1.8
9.7

1,485
15.4%
1.2
9.7

606
47.8%
1.8
8.4

596
47.0%
1.1
8.4

2,126
85.1%
2.1
9.9

2,073
82.9%
1.0
9.8

Highest rank = 1. Larger values indicate less serious statutory classification.

Refers to cases in which at least 1 charge remained active to be adjudicated through guilty pleas or trials

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 131

Similar to Southern County, acceptance rates in Northern County varied widely across crime
types, from a low of roughly 25 percent for domestic violence cases to nearly 87 percent for DUI
cases. The average statutory rank of the top filing charge varied considerably across crime types, in
a pattern generally consistent with expectations. Despite large differences in acceptance rates across
crime types and some differences in the average seriousness of filed charges, there were two
patterns that were extremely consistent across crime types. First, for every crime type, there was a
substantial drop in average number of active charges between initial filing and eventual conviction.
Second, for every crime type, there were virtually no changes in average statutory rank of the top
charge across any of the stages captured in this analysis. Later in this report, we discuss the relative
consequences to defendants of reductions in the number of active charges versus reductions (or lack
of reductions) in the seriousness of top charge as they related to the potential threat of incarceration.
3.3.3 Comparison and Summary
Both counties experienced substantial variation across crime types in the percentage of screened
cases accepted for prosecution. However, for comparable crime types, the acceptance rates were
consistently higher in Southern County than in Northern County. Neither county experienced any
significant change in average number of charges or average seriousness of the top charge between
intake to screening and initial filing of accepted cases. Neither county experienced reductions in
average seriousness of the top charge between initial filing and final disposition, and there were
noticeable increases in average seriousness for drug unit cases in Southern County. In contrast,
both counties experienced considerable reductions in the average numbers of active charges
between initial filing and final disposition. In Southern County, those reductions took place largely
between indictment and plea offer, except for the late drug sample, where it appears the decision to
drop charges may have occurred somewhat earlier. In Northern County, the lack of specific data on
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 132

indictment charges and plea offers makes it difficult to determine exactly when and how the
reductions in number of charges took place. However, it is quite plausible, given what is understood
from common knowledge and reinforced by the findings of the focus group discussions, that the
reductions in Northern County were also a product of plea negotiations.

3.4 Screening
The prosecutor is often seen as the gatekeeper for the criminal justice system. After the police
respond to a criminal incident, the prosecutor must review the facts of the case and determine
whether the state will formally charge an individual with a crime. This is generally referred to as
screening. In each jurisdiction, the timing of when a prosecutor screens a case depends on whether
an officer seeks a warrant for arrest or the defendant is arrested at the scene of an incident. Where
there has been an investigation and an officer wants to make an arrest, the case may be screened by
the district attorney’s office prior to obtaining a warrant from a magistrate. In neither jurisdiction is
this pre-warrant screening mandatory; rather, the district attorney’s office sees it as an efficient and
just process of ensuring that evidence is strong enough prior to the arrest and possible detention of
individuals. According to prosecutors, the notion is that if the evidence is not strong enough prior to
arrest, then the case will be rejected for prosecution after the arrest anyway; this just prevents
unnecessary or premature arrests. If the defendant is already arrested, the process works essentially
the same way. In both situations, law enforcement delivers a full case file – including all relevant
evidence, arrest charges, and circumstances of arrest – to the district attorney’s office for a face-toface meeting between a law enforcement officer and an ADA. At this stage the ADA determines
whether a person should be charged with a crime and what the charges should be.
In felony cases, an ADA will speak with the arresting/investigating officer and review all
reports and records, including witness statements and the suspect’s prior record. In Northern
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 133

County, witnesses may also be available for the ADA to interview; in Southern County witnesses
are not available and ADAs rely solely on the report filed by law enforcement. The screening ADA
has the option to reject the case for prosecution, send the case back to the police for follow-up
investigation, or accept the case for prosecution and issue a complaint. The ADA will put the initial
file together at this time, which includes all police reports, the initial charges filed, and generally a
recommendation for final disposition. In both jurisdictions, screening takes place seven days a week
and teams rotate through the screening area generally on a weekly basis, with each unit taking
several days or a week to screen cases for their unit. In Southern County, only experienced ADAs
screen felony cases; in Northern County all ADAs screen cases.
Misdemeanors in Northern County follow the same process. As with felonies, misdemeanor
cases are screened by ADAs of all experience levels and are generally handled within a general
crimes unit which prosecutes both misdemeanor and felony offenses. The primary difference in the
screening process for misdemeanors is that ADAs may speak with liaison officers from the relevant
precinct/agency rather than the arresting officer and may review files in bulk rather than one at a
time with the officer. Misdemeanors in Southern County, however, follow a completely different
process with no formal screening conference with law enforcement; rather, ADAs screen, charge,
and plea cases as part of the same process. Moreover, unlike felony cases in Southern County which
are screened by experienced ADAs, all misdemeanor cases are handled by the least experienced
ADAs in the office. Misdemeanor cases are processed in a lower court which includes both
misdemeanor offenses and criminal traffic offenses; as such, the courtroom handles an extremely
high volume of cases (approximately 100,000 cases per year). Individuals arrested for misdemeanor
offenses are generally given a citation with a date to come to court for a first appearance (for those
in custody, this is done via video conference). At this appearance, defendants will be told of the

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 134

charges against them and asked if they would like counsel to be assigned or if they are interested in
taking a plea and waiving counsel. All plea offers are made on the spot and are not subject to any
review by a supervisor; pleas are based only on the information contained on the charge sheet and
the defendant’s record. If a defendant does not express interest in a plea, or the case is not
appropriate to plead at this point, then counsel will be assigned (if necessary) and the case will be
adjourned approximately six weeks for trial. The trial date will depend on who the arresting officer
is; officers are assigned court dates two days per month.
The screening decision is generally seen as the most important decision in the processing of a
case. The DA in Southern County described screening as a mean task – it is stressful, requires
experience and patience, and requires training of the police. Moreover, at no other stage of the
process is the prosecutor’s decision more obscured and less reviewed than at screening; indeed, one
deputy in Northern County described the screening decision as unfettered discretion. While the
screening decision may be the most important, it is also the decision which prosecutors have the
least time to make. Prosecutors described having just enough time to make a decision of whether the
facts are present. As unit managers in Northern County noted, the minimal amount of time
available for screening meant that many cases may be dismissed immediately after screening
because the assigned ADA looks closer.
The screening decision is also much more complex than simply rejecting or accepting a case to
prosecute. Many cases—if not most—involve multiple arrest charges and, potentially, multiple filed
charges. Determining whether a case can go forward and whether a case should go forward involve
decisions at the individual charge level and decisions at the overall case level. Evaluating the
probability of conviction is first a charge-by-charge determination concerning whether there is
sufficient evidence to support each potential charge. The evaluation that a case (the collection of all

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 135

the charges) go forward implies a judgment that there is sufficiently strong evidence to support one
or more of the potential charges and that there are no known case-level characteristics (e.g., a
sympathetic defendant or contextual factors) that would make a jury likely to discredit or ignore the
evidence, lead a judge to exclude key evidence, or lead the prosecutor to otherwise reject a case.
Thus, the screening decision involves decisions about individual charges and the overall case:
individual charges may be rejected while the case itself is accepted. To account for this, we consider
the screening decision as it affects the entire case; a case may be rejected for prosecution or
accepted for prosecution. We consider the charging decision (in the following section) as it affects
the individual charges within a case; if the case is accepted for prosecution, for example, the most
serious arrest charge may be accepted or rejected.
Overall, the outcomes of screening decisions differed substantially between counties and across
crime types (Figure 3.4-1). Acceptance rates in Southern County were significantly higher than
those in Northern County for all three comparable offense types – felony drug, felony person, and
felony property offenses. The difference may be a reflection of different approaches to the screening
process. In Northern County, the District Attorney relies heavily on a strict screening decision to
divert cases from prosecution and to reject weak cases. Only recently has the District Attorney in
Southern County placed a greater emphasis on a strict screening, historically relying on dismissals
as a mechanism for rejecting weak cases after acceptance.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 136

Percentage of cases accepted

Figure 3.4-1 Percentage of cases accepted for prosecution in Southern and Northern Counties,
by crime type and offense level
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%

Northern County

Southern County

Person

Property

Drugs

Person

Felony

Property

Drugs

Public
Order

Domestic Weapons
Violence

DUI

Misdemeanor

We examined case-level and prosecutor-level factors influencing screening decisions using two
methods. First, we examined administrative data capturing the initial decision to accept or reject
actual cases. Second, we examined factorial survey data capturing prosecutors’ decisions to reject
or accept hypothetical cases for prosecution.
3.4.1 Screening Decisions in Southern County
Descriptive Statistics
The analyses of screening decisions in Southern County were based on 508 person unit cases,
653 property unit cases, 2,403 early drug sample cases, and 2,316 late drug sample cases. The
acceptance rates at screening in Southern County ranged from a low of 65.5 percent for property
crimes to a high of 80.1 percent for the more recent drug crime sample (Figure 3.4.1-1).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 137

Figure 3.4.1-1 Acceptance rates for cases screened in Southern County
Percentage of cases accepted

100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Person Team

Property Team

Early Drug
Sample

Late Drug Sample

Table 3.4.1-1 displays simple descriptive statistics showing how acceptance rates varied across
several case-level factors. The case characteristics presented in Table 3.4.1-1 are limited to ones that
proved to be significant factors in one or more of the final logistic regression models. A much larger
set of case factors and their interactions was tested for possible inclusion in the models. A complete
list of variables tested in the logistic regression analyses is given in Appendix A.
Table 3.4.1-1 Percentage of screened cases accepted for prosecution in Southern County, by
selected case characteristics
Case Characteristics
Number of charges presented for screening
1
2
3
4
5
Rank of top charge presented for screening
4
5
6
7
8
9
10
Codefendants
No
Yes

Person
Unit

Property
Unit

Early Drug
Sample

Late Drug
Sample

52%
84%
88%
87%
89%

49%
74%
84%
78%
87%

77%
86%
95%
94%
96%

67%
78%
87%
91%
92%

91%
72%
70%
73%
80%
78%
78%

66%
69%

100%
98%
90%
97%
86%
85%

97%
82%
94%
80%
76%

76%
72%

63%
70%

91%
79%

83%
70%

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 138

Table 3.4.1-1 (cont’d) Percentage of screened cases accepted for prosecution in Southern
County, by selected case characteristics
Case Characteristics
Black defendant
No
Yes
Male defendant
No
Yes
Defendant age at screening
<21
21-25
26-30
31-35
36-40
41+
Number of items of evidence per charge
<1
1-2
2+
Total number of items of evidence listed
0
1
2
3
4
5
Evidence with potential for forensic ID
No
Yes
Number of weapon-related items of evidence
0
1
2+
Number of victims
1
2+
Number of victims willing to prosecute
0
1
2+
Any non-person victims
No
Yes
Number of victim strangers
0
1
2+
Any victims need medical attention
No
Yes

Person
Unit

Property
Unit

Early Drug
Sample

Late Drug
Sample

62%
75%

63%
67%

90%
87%

81%
80%

71%
74%

53%
68%

89%
87%

81%
80%

72%
80%
77%
73%
53%
67%

73%
58%
64%
55%
70%
61%

84%
87%
(21-30)

74%

69%
82%
80%

65%
62%
78%

54%
84%
76%
84%
95%
96%

56%
61%
75%
91%
94%
92%

73%
72%

65%
83%

66%
86%
85%

66%
53%
80%

71%
77%

65%
75%

42%
80%
86%

46%
70%
77%

69%
81%

61%
72%

68%
74%
79%

53%
70%
76%

69%
81%

66%
-

84%
(31+)

82%
74%

-N < 10 cases

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 139

As Table 3.4.1-1 indicates, perceived seriousness of the offense appears to influence the
screening decision. Higher statutory class of the top charge presented for screening was associated
with higher acceptance rates, though the number of cases was small for some specific statutory
classes, and the resulting pattern was somewhat erratic. Among the measures of offense
characteristics, the most consistent pattern was found for the total number of charges presented for
screening. For all four samples, a greater number of charges was associated with substantially
higher acceptance rates. The meaning of this latter pattern is unclear. This may reflect an effect of
case seriousness, or it may be an evidence-related effect in which more charges provide a greater
opportunity that some charge or charges will be provable.7
In the two samples for which evidence and victim information was available, the descriptive
statistics show highly consistent relationships between acceptance rates and direct evidence
measures and between acceptance rates and evidence-related victim characteristics. For both the
person unit cases and the property unit cases, acceptance rates were substantially higher for cases
with a greater total number of items of physical evidence listed in the police report, more items of
evidence per charge, more weapons-related items of evidence, a greater number of victims
(especially the number of victims willing to prosecute), and any non-person victims (such as a
business or “the state,” for which testimonial evidence is more likely to be reliable or pertinent
records may be supplied). Cases in which the defendant was a stranger to the victim were more
likely than others to be accepted for prosecution, as were cases involving crimes against persons in
which the victims required medical attention.

On the other hand, cases involving more charges may be inherently more complex and more vulnerable to dismissal
later in the process. We present evidence later in this report suggesting that cases with more charges and more serious
top charges are likely to experience greater reductions between initial charging and plea offers but are less likely to be
dismissed completely.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 140

Other measures displayed in Table 3.4.1-1 exhibited less consistent results. The presence of
codefendants was associated with lower acceptance rates for drug crimes but somewhat higher
acceptance rates for property crimes. Cases involving black defendants had noticeably higher
acceptance rates for person crimes but showed no clear pattern for property crimes or drug crimes,
whereas cases involving male defendants had noticeably higher acceptance rates for property crimes
but no clear pattern for person crimes or drug crimes. There also was a slight but inconsistent
tendency for older defendants to experience lower acceptance rates for person and property crimes,
but no clear pattern for drug crimes.8 Finally, the presence of forensic evidence with the potential
for identifying the defendant (DNA, fingerprints, blood, etc.) was associated with substantially
higher acceptance rates for property crimes but not for person crimes. However, it should be noted
that sex offenses were excluded from all of these analyses, and forensic evidence may be less
commonly collected in higher volume crimes against persons such as assaults and robberies.
Logistic Regression Analyses
Separate logistic regression models for the probability of acceptance were estimated for person
unit cases and property unit cases. The early drug unit sample and the late drug sample were
combined in a single analysis to permit explicit tests of the effect of historical period and the
interactions between historical period and other factors. We employed binary logistic regression
modeling to determine which among the potential explanatory variables had the strongest unique
effects on the probability of acceptance at screening. However, as we have mentioned elsewhere in
this report, the relatively small samples available for some of the analyses, combined with the
relatively large number of candidate variables, means the selection of a particular variable among a
8

Prior criminal history scores were available for the defendants included in the samples from each unit. However, prior
criminal history was not included among the potential explanatory variables in analyses of screening outcomes, because
prior record was generally not available to prosecutors at the time of initial screening for felony cases. A statutorily
defined prior record score becomes available to prosecutors somewhat later in the processing of a case; analyses of plea
offers and dismissals presented later in this report examine the potential influence of prior record at those stages.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 141

related set of variables may often be a matter of chance. Thus, we caution readers not to attach too
much importance to which among related variables happens to show significant effects in a single
given model. Rather, we are attending to the overall pattern of findings across samples, measures,
and conditions.
Table 3.4.1-2 presents the final models for the effects of selected factors on the probability of
acceptance for cases screened by the person unit, property unit, and drug unit in Southern County.
The values in the body of the table are odds ratios. A ratio greater than one indicates that an
increase in the value for the variable is associated with an increase in the odds of acceptance. A
ratio less than one indicates that an increase in the value for the variable is associated with a
decrease in the odds of acceptance. Thus, for example, an odds ratio of 1.50 indicates an increase of
50 percent in the odds of acceptance, and an odds ratio of .50 indicates a decrease of 50 percent in
the odds of acceptance.9

Odds are not probabilities. Odds are related to probabilities by the function odds = p/(1-p). For example, if the
probability of acceptance is .80, the odds of acceptance are 4-to-1 (.80/.20 = 4). If the probability is .50, the odds are 1to-1 (.5/.5 = 1). The magnitude of an effect in the probability metric depends on the reference level. For example, if the
probability of acceptance for some reference group is .20 (odds of .25-to-1) and the probability for some target group is
.50 (odds of 1.0-to-1), then the odds ratio would be 4.0, and the ratio of probabilities would be 2.5. However, if the
probability for the reference group is .50 (odds of 1.0-to-1), an odds ratio of 4.0 would imply a probability of .80 for the
target group, and the ratio of probabilities would only be 1.6.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 142

Table 3.4.1-1 Final logistic regression models for effects of selected factors on the odds of
acceptance at screening in Southern County
Drug
Unit

Odds Ratios
Person
Unit

Offense Characteristics
Late drug sample
N of charges screened (up to 5)
Statutory class of top charge at screeninga
Codefendants
Burglary

.544***
1.682***
.900***
ns

1.595***
.902
.470***

1.496***
ns
ns
2.712***

Defendant Characteristics
Race (black)
Gender (= male for drugs; female otherwise)
Defendant age (3 levels for drug; 6 otherwise)

.773**
ns
1.238***

ns
ns
ns

ns
.485
.851***

.632
ns
ns
ns

.475*
2.188***
1.397*
2.210***

Evidence
Total N of items of evidence listed (up to 5)
N of items of evidence per intake charge (3 levels)
N of weapons-related items of evidence

ns
3.803***
ns

1.606***
ns
.329***

Interactions
Robbery BY N of victims willing to prosecute
Multiple charges BY race
Multiple charges BY N of vics willing to prosecute
Multiple charges BY evidence for forensic ID

1.911***
2.508**
2.446***
.418**

ns
ns
ns
ns

479
.36

616
.28

Independent Variables

Victim Characteristics
N of victims
N of victims willing to prosecute
N of victims to whom defendant was a stranger
Any nonperson victims

*
a

Ns
Nagelkerke R2

4383
.11

Property
Unit

p<=.10; **p<=.05; ***p<=.01; for coefficients without asterisks p<=.15
Statutory class was coded as 1 = highest rank; larger values correspond to less serious charges. Thus, odds ratios
greater than one denote inverse relationships with seriousness.

The direct and indirect influences of evidence are apparent in the results for both the person unit
cases and the property unit cases, but they show up differently in the two samples. For person unit
cases, there was a highly significant effect of the number of items of evidence per charge, as well as
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 143

effects for the number of victims willing to prosecute in robbery cases and cases with multiple
intake charges. For property unit cases, there were highly significant effects for the total number of
items of evidence, for the involvement of nonperson victims, and for the number of victims willing
to prosecute. However, for both samples, the effects of the number of victims willing to prosecute
were offset somewhat by a reduction in the odds of acceptance for greater overall numbers of
victims; the combination of these two variables perhaps suggest that greater numbers of victims
only enhance a case if the victims are willing to prosecute and may be detrimental otherwise.
Similarly, for person unit cases, there was a tendency for the positive effect of the number of items
of evidence per charge to be offset in cases with multiple charges by a negative effect of evidence
with the potential for forensic identification. There were only 29 person unit cases with such
evidence, so it is not clear whether this finding is merely a random anomaly, or whether it
represents a substantive influence, such as, for example, if cases dependent on forensic evidence
were especially prone to some damaging flaw.
A few of the significant effects were confined to one of the samples or were inconsistent across
samples. Cases in the late drug sample were much less likely to be accepted than earlier cases,
perhaps reflecting general efforts in Southern County to improve case screening or a specific
change in policy associated with the change in drug unit supervisor. Person unit cases involving
codefendants were less likely than other cases to be accepted; this could be a result of inducing
defendants to cooperate with the prosecution, but it is not clear why that would not also apply to
property cases and especially drug cases. Drug cases involving black defendants were less likely
than drug cases involving nonblack defendants to be accepted for prosecution, but the opposite was
true for person unit cases with multiple charges. Finally, cases involving female defendants were
less likely to be accepted than those involving male defendants for property crime cases but not for

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 144

the other two samples, and older defendants were more likely than younger ones to have their cases
accepted for prosecution in drug cases but less likely in property cases. Consistent with the caution
expressed earlier, we are inclined to view most of these inconsistencies as non-findings, rather than
attaching much substantive import to the pattern of inconsistencies.
Perhaps the most surprising result of these analyses was the consistent influence of case
seriousness across measures and samples. More serious top charge at intake was associated with
greater odds of acceptance for both the drug unit cases and the person unit cases. Greater numbers
of charges at intake were strongly associated with greater odds of acceptance in all three samples. In
addition, for property crimes, acceptance rates were higher for cases in which the defendant was a
stranger to the victim, and for burglaries (reflecting an explicit office policy in Southern County).
As noted in the discussion of the descriptive summaries, the number of charges could be influential
in more than one way—as an indication of case seriousness or as an influence on the ability to
obtain a conviction. Either way, it appears from these findings that case seriousness begins to
influence case decisions from the outset.
3.4.2 Screening Decisions in Northern County
Unlike Southern County, data in Northern County was available for all offense types and all
offense levels and was not limited to samples of data; rather, we were able to examine screening
decisions for all cases reviewed by the office over the entire study period. Therefore, we provide
additional descriptive statistics on the distribution of cases across offense types and offense levels in
Northern County.
Descriptive Statistics
The analyses of screening decisions in Northern County were based on 46,358 misdemeanor
cases and 28,274 felony cases. The largest portion of cases screened involved domestic violence
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 145

offenses (28.2 percent of all cases), followed by public order offenses (22.9 percent), property
offenses (15.8 percent), and drug offenses (13.9 percent) (Figure 3.4.2-1). While roughly 38 percent
of all cases involved a felony offense, this distribution was not true for all offense types. Roughly
76 percent of person cases, 64 percent of property cases, 84 percent of drug cases, and 57 percent of
weapons cases involved a felony as the top arrest charge. In contrast, just 19 percent of public order
cases, 9 percent of domestic violence cases, and 6 percent of DUI cases involved a felony.
Figure 3.4.2-1 Number of cases screened in Northern County, by offense type and offense
severity level
25,000

Number of Cases

20,000
15,000
10,000

5,000
0
Person

Property

Drugs

Public Order Domestic
Violence

Misdemeanor

Weapons

DUI

Felony

Figure 3.4.2-1 also indicates that a large portion of cases were misdemeanor domestic violence
cases (25.6 percent of all cases) and misdemeanor public order cases (18.7 percent of all cases);
these two offense type/offense severity categories accounted for roughly 44 percent of all cases and
71 percent of all misdemeanor cases screened in the office. As Figure 3.4.2-2 shows, domestic
violence battery cases were the most frequent cases screened in the office, accounting for nearly 16
percent of all cases screened, followed by driving on a suspended license (8 percent of all cases)
and possession with intent to distribute a controlled substance (6 percent of all cases).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 146

Overall, 30 percent of cases screened involved multiple arrest charges. This distribution was not
true for all offense types. As Figure 3.4.2-3 indicates, roughly 84 percent of DUI cases and 50
percent of weapons cases involved multiple arrest charges. In contrast, just 16 percent of domestic
violence cases, 23 percent of person offense cases, and 23 percent of public order offenses cases
involved multiple arrest charges.
Figure 3.4.2-2 Ten most frequently screened cases in Northern County, by offense type and
offense severity level
14,000
Number of Cases

12,000

10,000
8,000
6,000
4,000
2,000
0

Misdemeanor

Felony

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 147

Figure 3.4.2-3 Percent of cases screened in Northern County involving single or multiple
arrest charges, by offense type
100%
90%
Percent of Cases

80%
70%
60%
50%

40%
30%
20%
10%
0%
Person

Property

Drugs

Single arrest charge

Public Order Domestic
Violence

Weapons

DUI

Multiple arrest charges

As noted in Part 2, Northern County contains one primary municipal police department and
several suburban municipal police departments. Roughly 69 percent of all cases screened involved
arrests made by the primary municipal police department. This distribution was fairly stable across
offense types, with the exception of public order and DUI cases; for both of these offense types a
markedly lower percentage of cases involved arrest made by the primary agency (Figure 3.4.2-4).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 148

Figure 3.4.2-4 Percent of cases screened in Northern County involving primary municipal
police department or other law enforcement agency, by offense type
100%
90%
Percent of Cases

80%
70%

60%
50%
40%
30%
20%
10%
0%
Person

Property

Drugs

Public Order Domestic
Violence

Primary Municipal

Weapons

DUI

Other Agency

Roughly 82 percent of the cases screened in Northern County involved male defendants. Over
62 percent of defendants were black, 31 percent were white, 6 percent were Hispanic, and 1 percent
were of other racial backgrounds. When gender and racial characteristics of defendants are
combined, black males clearly represented the largest racial/gender group of defendants with nearly
52 percent of all cases screened in the office involving black male defendants (Figure 3.4.2-5);
however, felony cases were more likely to involve black men than misdemeanor cases, with over 62
percent of felony cases involving black male defendants compared to 46 percent of misdemeanor
cases. The distribution of racial/gender groups was fairly constant across offense types, with the
exception of weapons offenses and DUI offenses. Roughly 74 percent of weapons offense cases
involved a black male defendant, while over 55 percent of DUI offense cases involved a white male
defendant; in fact, over 80 percent of weapons offense cases involved black defendants and over 75
percent of DUI offense cases involved white defendants.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 149

Figure 3.4.2-5 Percent of cases screened in Northern County, by defendant race/gender

All Cases
5% 1%
10%

24%

53%
White Male

White Female

Black Male

Black Female

Hispanic Male

Hispanic Female

Misdemeanor Cases

Felony Cases

6% 1%
10%

4% 0%
10%

19%

27%
5%

9%
47%
62%

Roughly 55 percent of defendants were in custody at the time their cases were screened. Again,
the custody status of defendants varied widely for each offense type (Figure 3.4.2-6). As Figure
3.4.2-6 indicates, 84 percent of defendants in drug cases, 80 percent of defendants in weapons
cases, and 68 percent of defendants in property cases were in custody at the time of screening. In
contrast, just 21 percent of defendants in DUI cases and 38 percent of defendants in public order
cases were in custody at the time of screening.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 150

Figure 3.4.2-6 Percent of cases screened in Northern County involving defendants in custody
or not in custody, by offense type
100%
90%
Percent of Cases

80%
70%

60%
50%
40%
30%
20%
10%
0%
Person

Property

Drugs
In custody

Public Order Domestic
Violence

Weapons

DUI

Not in custody

Overall, roughly 50 percent of the cases screened in Northern County involved an identifiable
victim; victim information was available in 99 percent of person offenses, 95 percent of domestic
violence cases, and 64 percent of property cases.10 The demographic characteristics of victims were
quite different than those of defendants (Figure 3.4.2-7). While the racial composition of victims
was similar to that of defendants – roughly 57 percent of victims were black compared to 62 percent
of defendants – the gender composition was quite different – nearly 60 percent of victims were
female compared to 18 percent of defendants. When gender and racial characteristics of victims
were combined, black females clearly represented the largest racial/gender group of victims
screened in the office (over 41 percent of all cases screened) followed by white females (over 22
percent of all cases screened). The high percentages of female victims were due largely to domestic
violence cases. As Figure 3.4.2-7 shows, over 54 percent of domestic violence cases involved black

Naturally, some types of offenses – drug offenses, public order offenses, and DUI offenses specifically and some
property and weapons offenses – do not involve victims. In addition, some cases categorized as, for example, drug
cases in our analyses have victims listed; however, these victims are not involved in the controlling offense (i.e. the
most severe charge). These cases are rare, however, and victim information is not included in such cases. Thus, victim
information is included only when it pertains to the most serious arrest offense.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 151

female victims and nearly 25 percent involved white female victims. Person cases showed similar
high rates of black female victims (nearly 33 percent of all person cases) and black victims
generally (roughly 56 percent of all person cases); property cases displayed higher rates of white
male victims (over 31 percent of all property cases) and white victims generally (roughly 52 percent
of all property cases).
Figure 3.4.2-7 Percent of cases screened in Northern County, by victim race/gender

All Cases
1%

15%

43%

23%
16%

Person Cases
2%

White Male

White Female

Black Male

Black Female

Hispanic Male

Hispanic Female

Property Cases
2%

16%

24%

DV Cases
1%

32%

6%
25%

33%
20%
54%
20%
26%

21%

10%

Given the division of the district attorney’s office into specialized units, a high percentage of
cases – over 55 percent – were screened by specialized units. Naturally, this was higher for those
offenses targeted by such specialized units (Figure 3.4.2-8). For example, nearly 100 percent of
domestic violence cases, 77 percent of drug cases, and 62 percent of weapons cases were screened
by specialized units dedicated to these offense types, respectively; just 7 percent of property cases,
29 percent of public order cases, and 30 percent of DUI cases were screened by a specialized unit.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 152

Figure 3.4.2-8 Percent of cases screened in Northern County by a specialized or general
crimes unit, by offense type
100%
90%
Percent of Cases

80%
70%

60%
50%
40%
30%
20%
10%
0%
Person

Property

Drugs

Specialized prosecution unit

Public Order Domestic
Violence

Weapons

DUI

General crimes prosecution unit

A total of 145 unique ADAs screened cases during the study period. While the case
management system requires screening ADAs to identify themselves, this information was missing
in 9.7 percent of cases. For those cases with screening ADA information, just over 65 percent were
screened by male ADAs; this was fairly stable across offense types, with the exception of weapons
cases and DUI cases, in which over 80 percent of cases were screened by male ADAs. Overall,
approximately 38 percent of cases were screened by ADAs with less than two years of experience,
23 percent were screened by ADAs with two to ten years of experience, and 39 percent were
screened by ADAs with more than ten years of experience (Figure 3.4.2-9). Again, this varied
dramatically across offense types. For example, while just 18 percent of weapons offenses were
screened by ADAs with less than two years of experience, over 70 percent of domestic violence
cases were screened by these ADAs. For most offense types, ADAs with less than two years of
experience screen between 20 percent and 35 percent of cases, consistent with their representation
in the office.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 153

Figure 3.4.2-9 Percent of cases screened in Northern County by inexperienced, mid-career,
and experienced ADAs, by offense type
100%
90%
Percent of Cases

80%
70%

60%
50%
40%
30%
20%
10%
0%
Person

Property

ADA <2 years experience

Drugs

Public Order Domestic Weapons
DUI
Violence
ADA 2-10 years experience
ADA >10 years experience

Overall, roughly 57 percent of cases were accepted for prosecution during the study period; in
other words, of the 74,632 cases screened in Northern County, 42,420 cases (56.8 percent) were
accepted for prosecution and 32,212 cases (43.2 percent) were rejected (Table 3.4.2-1). Cases
involving felonies were accepted at a slightly higher rate (63.0 percent) than cases involving
misdemeanors (54.7 percent). Moreover, acceptance rates differed markedly by offense type. Over
94 percent of DUI cases, 75 percent of drug cases, and 71 percent of public order cases were
accepted for prosecution, compared to just 34 percent of domestic violence cases and 44 percent of
person cases. This is fairly consistent with prosecutors’ impressions as described in Part 2; those
cases that rely heavily on victim participation or testimony, such as person offenses and domestic
violence offenses, tend to have lower acceptance rates than other types of cases. Conversely, cases
that rely primarily on physical evidence, such as drug offenses and weapons offenses, tend to have
higher acceptance rates. Table 3.4.2-1 displays simple descriptive statistics showing how
acceptance rates varied across several case-level factors.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 154

Table 3.4.2-1 Percentage of screened cases accepted for prosecution in Northern County, by
selected case characteristics
Overall

Person
44%

Property
60%

Drugs
75%

Public
Order
71%

DV
34%

Weapons
59%

DUI
94%

Offense Characteristics
Charge Severity
Misdemeanor
Felony

41%
46%

66%
57%

56%
78%

71%
69%

33%
34%

56%
62%

94%
97%

Number of charges
Single charge
Multiple charges

37%
67%

54%
72%

71%
80%

69%
76%

30%
56%

47%
70%

73%
98%

Enhancement Offense
No
Yes

44%
53%

60%
75%

74%
83%

71%
57%

30%
43%

58%
71%

94%
100%

Arresting Agency
Primary municipal
Other

47%
38%

55%
73%

75%
72%

65%
81%

35%
34%

60%
56%

88%
95%

Drug Type
Cocaine
Heroin
Marijuana
Other

-----

80%
87%
75%
65%

-----

45%
44%
49%

64%
58%
66%

71%
76%
77%

77%
66%
74%

32%
34%
42%

54%
60%
61%

95%
91%
93%

Gender
Female
Male

43%
45%

60%
60%

65%
76%

67%
71%

22%
37%

44%
60%

84%
95%

Custody Status
Not in custody
In Custody

32%
54%

53%
63%

61%
77%

75%
66%

30%
39%

39%
64%

95%
92%

Defendant Characteristics
Race
White
Black
Hispanic

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 155

Table 3.4.2-1 (cont’d) Percentage of screened cases accepted for prosecution in Northern
County, by selected case characteristics
Victim Characteristics
Race
White
Black
Hispanic
Gender
Male
Female
Prosecutor Characteristics
Gender
Female
Male
Experience
<2 years
2-10 years
>10 years
Unit
General crimes unit
Specialized unit

Person

Property

Drugs

Public
Order

Weapons

DUI

46%
41%
46%

59%
50%
61%

----

34%
32%
40%

----

41%
46%

55%
56%

---

22%
35%

---

45%
44%

56%
63%

77%
73%

68%
71%

31%
37%

60%
59%

93%
95%

45%
44%
45%

63%
56%
60%

78%
75%
68%

73%
60%
73%

32%
33%
46%

65%
57%
57%

94%
92%
95%

36%
50%

59%
63%

70%
76%

72%
67%

-34%

53%
61%

95%
83%

As Table 3.4.2-1 indicates, acceptance rates varied considerably along several dimensions of
case characteristics, defendant characteristics, and prosecutor characteristics. Across all offense
types, cases involving multiple arrest charges were more likely to be accepted for prosecution than
cases involving single arrest charges. Cases involving felonies generally were more likely to be
accepted, with the exception of property and public order offenses. Acceptance rates also varied by
arresting law enforcement agency; however, the patterns were not consistent. For some offenses –
drugs, weapons, domestic violence, and DUI – acceptance rates were roughly the same for both the
primary municipal police department and other agencies. For property and public order cases,
acceptance rates for case involving the primary municipal police department were roughly 20
percentage points higher than other agencies and for person offenses they were roughly 10
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 156

percentage points lower. These differences may be a reflection of differences in the quality of cases
presented by different agencies, described by prosecutors in Part 2.
Overall acceptance rates were nearly identical for white and Hispanic defendants (roughly 64
percent), which were slightly higher than acceptance rates for black defendants (approximately 55
percent); acceptance rates for men (58 percent) were also higher than rates for women (51 percent).
When racial/gender groups were compared, cases involving Hispanic male defendants and white
male defendants tended to be accepted at higher rates than cases involving other racial/gender
groups, with a few exceptions. Acceptance rates based on the victim racial/gender categories were
similar to those based on defendant racial/gender categories, with cases involving black victims
accepted at lower rates than cases involving white victims. This is likely an artifact of intra-racial
offending patterns for person, property, and domestic violence offenses. Generally, cases involving
white female victims tend to be accepted at rates higher than all other racial gender groups. Cases
involving defendants in custody at the time of screening also had slightly higher acceptance rates
than cases in which the defendant was not in custody (60.1 percent versus 56.1 percent,
respectively). This was true for all offense types, except public order and DUI offenses.
Acceptance rates also varied slightly by prosecutor gender and level of experience. Overall,
male prosecutors accepted cases at a higher rate than female prosecutors (59 percent versus 51
percent). Less experienced ADAs also tended to accept cases at a higher rate than more experienced
ADAs, with the exception of domestic violence cases. Finally, there was no clear pattern to
acceptance rates based on the type of prosecution unit screening the case.11 Specialized prosecution
units accepted cases at a higher rate than general crimes units for cases involving person offenses,
property offenses, and public order offenses; but specialized units accepted cases at a lower rate
11

Since less than 100 domestic violence cases were screened outside the specialized domestic violence unit over the
entire study period, the screening decisions for these cases are not included.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 157

than general crimes units for cases involving drug offenses and DUI offenses. The differences in
acceptance rates are likely a result of the severity of cases that are referred to specialized units. For
example, all felony drug offenses are screened and prosecuted by the specialized drug unit, while
most misdemeanor drug offenses are screened by general crimes units; thus, the differences in
acceptance rates may be a reflection of differences in the underlying severity of offenses screened
by each unit. The same may be true for person offenses. Northern County has a specialized unit
which screens and prosecutes all sex offenses and crimes involving children; all other person
offenses are screened and prosecuted by general crimes units.
HLM Analyses
Table 3.4.2-2 presents results from the two-level unconditional models of screening for each
offense type. Results suggest that across all models approximately 10 percent of the total variation
in the likelihood of a case being accepted can be attributed to differences between prosecutors; the
variance attributed to prosecutors ranged from 4 percent in DUI cases to 11 percent in drug cases.12

Since the screening outcome is a dichotomous variable, it lacks a meaningful individual-level variance component,
σ2. The Level 1 model , however, can be conceived of in terms of a latent variable (see Snijders and Bosker, 1999, cited
in Raudenbush and Bryk, 2002: 334) and the Level 1 random effect can be assumed to have a standard logistic
distribution with a mean of 0 and σ2 = π2/3. The intraclass correlation can then be estimated using the between variance,
τ2, and this estimation of the within variance, σ2, as follows: τ 2/( τ2+ σ2). This tells us the portion of the total variance
that occurs between prosecutors. Others, however, have cautioned that the meaningfulness of this estimation depends
on the validity of the underlying distributional assumptions.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 158

Table 3.4.2-2 HLM unconditional models predicting cases accepted at screening in Northern County

Fixed Effects
Intercept

Person
b
SE
-.28 .08***

Property
b
SE
.57 .07***

b
.93

SE
.08***

Random Effects
Level 1
Level 2

Var.
-.38

Var.
-.36

Var.
-.41

SD
-.64***

Between-prosecutor
proportion of
variance

SD
-.62***

.10

SD
-.60***

.10

Drugs

.11

Public Order
b
SE
.77 .06***
Var.
-.36

SD
-.60***

.09

Domestic
Violence
b
SE
-.43 .64***

Weapons
b
SE
.37 .09***

b
2.61

SE
.10***

Var.
-.41

Var.
-.24

Var.
-.13

SD
-.37***

SD
-.64***

.11

SD
-.49***

.06

DUI

.04

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 159

Table 3.4.2-3 presents results for the individual-level influences on screening. Two models
examining property offenses are included. Since not all property offenses included an identifiable
victim, separate models were run for property crimes with a victim and all property crimes. Table
3.4.2-4 then presents the full two-level models. Coefficients presented in the models represent the
increases in the likelihood of a case being accepted generated by each predictor variable. These
coefficients are presented as odds ratios – coefficients greater than 1 imply an increased probability
of failure and coefficients below 1 indicate a reduction in such probability. In all models, “white” is
taken as the reference category when comparing outcomes across defendant and victim racial
groups; the coefficients listed for other races are the effects of these races on case outcomes
compared to whites. For drug offenses, “cocaine” was used as the reference category when
comparing across drug offenses; the coefficients listed for other drug types are the effects of those
drug types on case outcomes relative to cocaine.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 160

Table 3.4.2-3 Individual-level factors predicting cases accepted at screening in Northern County
Model 1
Person

Model 2a
Property

Model 2b
Property

Coefficients (Odds Ratios)
Model 3
Model 4
Drugs
Public Order

Model 5
DV

Model 6
Weapons

Model 7
DUI

1.067***
1.938***
1.345
1.676***

1.037***
1.484***
1.277
.412***

1.082***
1.539***
1.086
.541***

1.154***
1.328***
1.328**
.672***

.920***
1.600***
.635***
.396***

1.142***
2.200***
1.219***
1.377***

.964
1.823***
1.148
.964

1.327***
2.438***
--

-----

-1.097
1.143
.584***

-----

Defendant Characteristics
Race
White (reference)
Black
Hispanic
Gender (Male)
Age (Years)
Custody Status (In Custody)

-.902
.905
1.006
.993*
2.36***

-.985
1.120
.975
1.007***
2.070***

-.972
1.010
1.232**
.997
2.017***

-.847*
.974
1.416***
1.009***
2.075***

-.920
.991
1.232***
1.011***
1.137*

-1.275***
1.265**
1.680***
.997
1.616***

-1.022
1.218
1.676**
1.000
2.739***

-.639*
.664
1.328
.998
1.480

Victim Characteristics
Race
White (reference)
Black
Hispanic
Gender (Male)
Age (Years)

-.670***
.790
.692***
.999

------

-.817**
1.312
.797***
1.007***

------

-.692***
1.124
.636***
1.005**

------

Offense Characteristics
Charge Severity
Number of charges (>1)
Enhancement Offense (Yes)
Arresting Agency (Primary)
Drug Type
Cocaine (reference)
Heroin
Marijuana
Other

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 161

As Models 1 through 7 show, there are no consistent predictors of case outcomes across offense
types, with the exception of variables measuring offense characteristics. For all offense types, cases
involving multiple arrest charges were more likely to be accepted for prosecution than cases
involving a single arrest charge. Having more than one arrest charge increased the likelihood that a
case would be accepted by anywhere from 30 percent (for drug offenses) to nearly 140 percent (for
DUI offenses) after controlling for other factors. This may be an indication of the seriousness of the
case; however, the effect of the number of arrest charges was independent of charge severity (i.e.
the severity of the controlling arrest charge) and charge enhancements. In six of the eight models
(person, property, drugs, domestic violence, and DUI), cases with more serious arrest charges were
more likely to be accepted for prosecution, although the effect size was small for some offenses. For
example, each increase in the severity of the controlling arrest offense (e.g. from Class A
misdemeanor to Class I felony) increased the likelihood of accepting a person offense by 6 percent;
in contrast, the same one level increase in the severity of an offense increased the likelihood of
accepting a drug offense by 15 percent. Surprisingly, more serious public order offenses were less
likely to be accepted for prosecution; similarly, the presence of a charge enhancer decreased the
likelihood that a public order offense was accepted. Moreover, this was contrary to the impact of
charge enhancers on the outcomes for drug and domestic violence offenses. Finally, for drug
offenses, the type of drug had little impact on screening decisions; however, cases involving drugs
such as prescription drugs, LSD, or other Schedule IV and V drugs were less likely to be accepted
for prosecution relative to cocaine.
Arresting agency influenced the screening decision in five of the eight models (person, property,
drugs, public order, and domestic violence). For property, drugs, and public order offenses, arrests
made by the primary municipal police department were less likely to be accepted for prosecution. In

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 162

the case of property and drug offenses, this may be an indication of changes in the investigation
units within the police department, as discussed in Part 2. Property offenses, in particular, may be
more affected by decreases in the quality of investigations, which translates into lower acceptance
rates for such offenses, consistent with prosecutors’ perceptions. The lower acceptance rates for
public order offenses may be due to similar circumstances. As discussed by prosecutors in focus
groups, the primary police department is increasingly focused on making arrests and getting their
numbers up; this may result in an increase in weak arrests for low level offenses and, according to
prosecutors, little follow through by police officers after the arrest. Conversely, for person offenses,
arrests made by the primary municipal police department were more likely to be accepted for
prosecution. This may be due to the inclusion of sex offenses within the person offense category.
In focus groups, prosecutors noted that specialized investigation units for sex offenses remained in
the primary municipal police department; as such, the quality of arrests may remain high and, as a
result, such cases may be more likely to be accepted.
Defendant characteristics – particularly gender and age – were fairly strong predictors of
screening outcomes. The presence of a male defendant increased the likelihood that a case was
accepted in five of the eight models (property, drugs, public order, domestic violence, and
weapons), consistent with the descriptive statistics presented above. For most offense types, this
may not be surprising; the large majority of defendants in all cases were male and the very large
sample sizes may create a problem in which trivial differences are exaggerated and found to be
significant. In contrast, it may be an indication that prosecutors view cases involving female
defendants differently, and reject cases at higher rates for such defendants. Defendant age was also
significant in four of the eight models (person, property, drugs, and public order). For property,
drug, and public order offenses, cases involving older defendants were more likely to be accepted

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 163

for prosecution. The effect, however, was rather small; overall, each one year increase in a
defendant’s age increased the likelihood that a case would be accepted by roughly one percent.
Nonetheless, this lends some support to prosecutors’ focus group statements that defendant
characteristics such as age matter when determining whether a case should be prosecuted; in these
cases, it appears that prosecutors view cases involving younger defendants as cases that, perhaps,
should not be prosecuted. These are also offenses that do not involve violence or potential physical
harm to others. In contrast, person offenses, cases involving older defendants were less likely to be
accepted for prosecution; again, the effect was rather small with each one year increase in a
defendant’s age decreasing the likelihood that a case would be accepted by roughly one percent.
Defendant race/ethnicity was significant in just two of the eight models (drugs and domestic
violence). For drug offenses, cases involving black defendants were less likely to be accepted for
prosecution than cases involving white defendants; the presence of a white defendant increased the
likelihood that a drug case was accepted for prosecution by 18 percent relative to cases involving
black defendants. Conversely, for domestic violence cases, cases involving black defendants or
Hispanic defendants were more likely to be accepted for prosecution; specifically, the presence of a
black defendant or a Hispanic defendant increased the likelihood that a domestic violence case was
accepted for prosecution by roughly 27 percent relative to cases involving white defendants.
Defendant custody status was a significant predictor in all eight models. In each model, a case
with a defendant in custody was more likely to be prosecuted than a case with a defendant not in
custody; in each model, the presence of a defendant in custody increased the likelihood that the case
was accepted by over 110 percent.
While defendant race/ethnicity was not a consistent predictor of case outcomes, victim
race/ethnicity was. For the three offense types for which victim information was available (person,

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 164

property, domestic violence), cases involving black victims were less likely to be accepted for
prosecution than cases involving white victims; compared to cases involving black victims, the
presence of a white victim increased the likelihood that a case was accepted by 23 percent for
property offenses to 49 percent for person offenses. Cases involving male victims were also less
likely to be accepted for prosecution in all three models, with the presence of a female victim
increasing the likelihood that a case was accepted by 25 percent (property offenses) to 56 percent
(domestic violence offenses). Finally, victim age was significant in two models (property, domestic
violence). In both models, cases involving older victims were more likely to be accepted for
prosecution; like the effect of defendant age, the overall effect was small with each one year
increase in the victim’s age increasing the likelihood that the case would be accepted by less than 1
percent.
Overall, offense characteristics were strong predictors of screening decisions. Taken together,
the effects of offense characteristics on screening outcomes appear to indicate that more serious
cases (i.e. with more severe charges, more charges, and charge enhancers) generally are more likely
to be accepted for prosecution. This is consistent with the findings in Southern County and with
prior research that has found case characteristics generally to be the strongest predictors of decisionmaking. Moreover, this supports the prosecutors’ perceptions as discussed in Part 2. Prosecutors
stated that after strength of the evidence, the seriousness of the case (which was partially
determined by the severity of the charges) was one of the determining factors in whether a case
should be prosecuted.
What is surprising – although also perhaps reassuring – is that defendant characteristics were
very inconsistent predictors of screening decisions. For DUI, weapons, and person offenses, for
example, defendant characteristics mattered very little. In DUI cases, only the race of the defendant

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 165

mattered; defendant gender, age, and custody status had no impact on the screening decision. As
noted in Part 2, DUI cases may be driven primarily by physical evidence against defendants (e.g.
the results of a breathalyzer) which may be less open to interpretation; as a result, such cases may
generally have very strong evidence. Moreover, such cases are increasingly a target for tougher
enforcement and, as such, defendant characteristics may not influence prosecutors’ decisions to
decline cases. This also may be true for weapons offenses – offense types that are increasingly
targeted for tougher enforcement and, as such, less open to extra-legal influences. In fact, for
weapons and DUI offenses, very few variables were significant predictors of screening decisions;
for both offense types, offense characteristics appeared to be more consistent predictors. For other
offense types, particularly drug offenses, defendant characteristics appeared to matter more; for
drug offenses, defendant race, gender, age, and custody status were all associated with the screening
decision. As prosecutors described in focus groups, they rely on defendant characteristics to
determine whether a case should be prosecuted. This may be occurring for drug offenses in which
black defendants, female defendants, and younger defendants are less likely to be prosecuted.
Table 3.4.2-4 presents the results of the two-stage HLM models with prosecutor-level variables
included. As Table 3.4.2-4 indicates, the characteristics of prosecutors available for these analyses
mattered very little in the screening decision.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 166

Table 3.4.2-4 Two-stage HLM models predicting cases accepted at screening in Northern County, prosecutor-level predictors
Model 1
Person

Model 2a
Property

Model 2b
Property

Coefficients (Odds Ratios)
Model 3
Model 4
Drugs
Public Order

Model 5
DV

Model 6
Weapons

Model 7
DUI

1.062***
1.939***
1.363
1.672***

1.037***
1.484***
1.277
.414***

1.083***
1.539***
1.081
.543***

1.147***
1.330***
1.322**
.672***

.923**
1.603***
.636***
.396**

1.143***
2.200***
1.220*
1.378***

.956
1.829***
1.146
.962

1.327***
2.438***
--

----

1.100
1.127
.579***

----

Defendant Characteristics
Race
Black
Hispanic
Gender (Male)
Age (Years)
Custody Status (In Custody)

.992
.909
1.008
.992*
2.372***

.984
1.120
.975
1.007***
2.070***

.969
1.009
1.232**
.997
2.024***

.849**
.978
1.419***
1.009***
2.062***

.920
.989
1.232***
1.011**
1.138

1.275***
1.266*
1.679***
..997
1.611***

1.022
1.240
1.692**
1.000
2.730***

.639*
.664
1.328
.998
1.480

Victim Characteristics
Race
Black
Hispanic
Gender (Male)
Age (Years)

.671***
.792
.695***
1.000

-----

.819**
1.318
.797**
1.007***

-----

.691***
1.124
.635***
1.005*

-----

Prosecutor Characteristics
Gender (Male)
Experience (years)
Supervisor
Specialized unit
Caseload
Percentage offense-specific

1.039
1.000
.858
1.374
1.000
1.000

1.248
.991
.970
1.220
.999
1.007

1.201
.984*
1.082
1.117
.999
1.008

.913
.997
.997
.904
1.001
1.006*

.922
1.003
.958
.975
1.000
1.005

.887
1.027***
1.125
1.355
1.003*
.995

.986
.994
.768
1.747*
.997
.986*

-------

Offense Characteristics
Charge Severity
Number of charges (>1)
Enhancement Offense (Yes)
Arresting Agency (Primary)
Drug Type
Heroin
Marijuana
Other

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 167

While prosecutor-level characteristics explained roughly 10 percent of the variance in outcomes
(see Table 3.4.2-2 above), very few prosecutor-level predictors were significant in the models.
Surprisingly, prosecutor tenure, supervisory responsibilities, and familiarity with a particular type of
offense had little impact on screening outcomes for most offenses. For property offenses, however,
more experienced ADAs were less likely to accept cases for prosecution – each one year increase in
experience decreased the likelihood of acceptance by just roughly 1.6 percent. For domestic
violence cases, more experienced ADAs were more likely to accept cases for prosecution, with each
one year increase in experience increasing the likelihood of acceptance by roughly 2.7 percent. The
percentage of offense-specific cases screened by the prosecutor was significant in only two models.
For drug offenses, prosecutors who had screened more cases involving drug offenses were more
likely to accept drug cases for prosecution, although the effect size was very small; in contrast, for
weapons offenses, prosecutors who had screened more cases involving weapons offenses were less
likely to accept such cases for prosecution. The fact that the prosecutor was a supervisor was not
significant in any of the models and the fact that a prosecutor was part of a specialized unit was
significant in just one model, with prosecutors in specialized prosecution units being more likely to
accept cases involving weapons offenses. Finally, contrary to prosecutors’ perceptions of the impact
of caseloads on case outcomes, prosecutor caseloads did not impact screening decisions, although
our ability to accurately capture caseloads at particular points in time was limited. The measure
used here captures average weekly caseload during the study period; a more accurate measure
would be weekly or monthly caseloads tied to the week or month in which the screening decision
was made.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 168

3.4.3 Factors Influencing the Screening Decision: Hypothetical Cases
While outcomes in actual cases provide insights into the aggregate effect of case specific factors
on outcomes, the study also sought to examine variation in outcomes across prosecutors for
identical cases using the factorial survey design and the set of ten case vignettes.
Descriptive Statistics
The analyses of hypothetical screening decisions relied on one dependent variable: whether the
respondent would reject the case at screening (0=accept, 1=reject). As noted above, sixty-two
prosecutors responded to the factorial survey. Respondents were remarkably consistent in their
screening judgments (Figure 3.4.3-1). For six of the ten vignettes, nearly all of the respondents
(between 95 and 100 percent) indicated they would accept the case for prosecution. For vignette 2
(medium offense seriousness, low evidence strength), 87 percent indicated they would reject the
case. Two vignettes (vignette 2 and vignette 8 – low offense seriousness, low evidence strength)
accounted for 75 percent of the rejections. Three vignettes—vignette 6 (high offense seriousness,
high evidence strength), vignette 7 (medium offense seriousness, medium evidence strength), and
vignette 8 (low offense seriousness, low evidence strength)—accounted for most of the variance
among prosecutors in screening decisions.
This pattern provides a preliminary suggestion that prosecutors were primarily influenced in
their screening decisions either by case seriousness or strength of evidence, since those were the
fixed attributes of the vignettes. It suggests they were less likely influenced primarily by criminal
history, since criminal history patterns were assigned randomly to vignettes within respondent
packets. Thus, if prosecutors were responding primarily to criminal history, more variation in
decisions would be expected within vignettes. The statistical modeling addresses these distinctions
more directly.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 169

Figure 3.4.3-1 Percentage of hypothetical cases rejected at screening, by vignette
Percentage of hypotehtcial cases rejected at
screening, by vignette

100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
1

Vignette

Three case-level independent variables were emphasized in the statistical modeling: researchers’
classification of the seriousness of the top arrest charges (offense seriousness), prosecutors’ ratings
of the strength of evidence (evidence rating), and the prosecutors’ ratings of the seriousness of
defendant criminal histories (criminal history rating). Twelve prosecutor-level independent
variables were also included in the analyses: a measure of prosecutors’ experience and eleven
measures of prosecutors’ values and preferred prosecution strategies derived from their responses to
the general survey. (See Section A.3.4 in Appendix A for detailed explanations and descriptive
analyses of the independent variables.)
HLM Analyses
Multilevel modeling of the factorial survey data was necessary to account for the dependencies
among observations due to the hierarchical structure of the data—i.e., the fact that unique cases are
nested within respondents. It also makes it possible to examine the potential influence of prosecutor
characteristics on case-level decisions.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 170

Sixty-two prosecutors responded to 10 vignettes each, yielding a maximum of 620 observations
for each case level decision addressed in the factorial survey. There were 570 cases with complete
data across the four case-level variables incorporated in the modeling: the accept/reject decision,
offense seriousness level, evidence rating, and criminal history rating. Prosecutor-level data were
matched to complete case-level data for 60 of the 62 respondents. For case-level data and
prosecutor-level data combined, a total of 560 observations were available for multi-level modeling
of the screening decision.13
Two-level models of the probability of rejection at screening were estimated using HLM 6
(Raudenbush et al., 2004) with the Bernoulli sampling model and a logit link function. Because the
number of observations was somewhat limited, given the number of parameters to be estimated and
the sample requirements for stable estimation of non-linear models, modeling proceeded in a
cautious, step-wise fashion. First, the case-level parameters (intercept and regression weights) were
constrained to be the same across all prosecutors, and a simple model was estimated for the three
case-level predictors with no interaction terms. Second, if convergence was achieved on a plausible
model, the constraints were removed and the case-level regression parameters were allowed to vary
across prosecutors. Third, if convergence was achieved on a plausible unconstrained model, and the
model did not imply substantively different conclusions than the constrained model, the three twoway interactions among the case-level predictors were introduced together. Fourth, non-significant
interactions were removed one at a time, beginning with the least significant, until only statistically

With race and gender excluded, 12 of the 14 prosecutor-level measures described above were considered for inclusion
in the multi-level modeling. These 12 measures proved to be relatively independent of one another. Of the 66 unique
rank order correlations among the 12 measures, only 13 were statistically significant, and all but two of those had
absolute values in the range from .27 to .36. These correlations were not expected to pose multi-collinearity problems in
estimating model parameters, because the models tested never included more than two or three of the prosecutor-level
measures at a time. One measure did exhibit strong correlations with two others: the all charges dichotomy had a
correlation of -.60 with both the top charges dichotomy and the should plead dichotomy. However, that was simply
because the three dichotomies together constituted a nearly exhaustive set (see Table 3.4.1-3); no more than one of
those was ever included in a single model.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 171

significant interactions remained. Main effects of case-level predictors that participated in
significant interactions were retained, whether or not they were statistically significant. All models
tested survived these four steps. Only the final models are presented in this section.
The final model for probability of rejection at screening is summarized in Table 3.4.3-1.
Neither offense seriousness, nor criminal history rating, nor any of the two-way interactions among
the case-level measures had a significant effect on the probability of rejecting a case. Consistent
with the claims of prosecutors who participated in the focus group sessions, the only case-level
measure with a statistically significant influence on the screening decision for the hypothetical cases
was prosecutors’ rating of the strength of evidence. The effect of evidence was dramatic; on
average, each step down on the evidence rating scale was associated with approximately a 257
percent increase in the odds of rejecting the case.
Table 3.4.3-1 Hierarchical logistic regression model for the decision to reject at screening
Effect

Case level
Intercept
Offense seriousness
Evidence rating
Criminal history rating
Prosecutor level
Item Q4c
(Low priority if effort
exceeds benefits)a
Item Q5a
(evaluate early disposition
potential)a
Consistency scalea
a

Coefficient

Odds
Ratio

T Ratio

P value

-1.72
0.11
-1.27
-0.08

0.18
1.12
0.28
0.92

-21.50
1.004
-15.644
-1.262

58
61
61
61

.000
.320
.000
.212

-0.23

0.80

-3.07

.004

0.18

1.19

2.64

.011

-0.07

0.93

-2.02

.048

See Table A.3-2 for explanations of Q4c, Q5a, and the consistency scale.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 172

Figure 3.4.3-2 displays the observed pattern of screening outcomes across levels of offense
seriousness and evidence ratings. The modeling results provided an estimate of the average effect of
strength of evidence for each step up in the prosecutors’ evidence ratings. Figure 3.4.3-2 suggests
the effect was non-linear, with a dramatic drop in the probability of rejecting a case occurring
between the lowest evidence rating and the average evidence rating, and relatively little difference
in the outcomes across average, mid-high, and highest evidence ratings.

Percentage of cases rejected at screening

Figure 3.4.3-2 Percentage of hypothetical cases rejected at screening, by offense seriousness
and prosecutor’s rating of strength of evidence
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
1 Lowest

2 Mid-Low

3 Average

4 Mid-High

5 Highest

Strength of evidence (prosecutor's rating)
Misdemeanor

Low level felony

High level felony

Three prosecutor-level measures had statistically significant associations with variations in the
case-level intercepts (viewed here as main effects of prosecutor-level measures on the odds of
rejecting a case for prosecution). Higher ratings of the importance of consistency in approach and
outcomes (consistency scale) were associated with lower odds of rejection at screening. Given
prosecutors’ comments indicating that there is little time to evaluate cases thoroughly at screening,
and that cases accepted at screening are often quickly dismissed upon further evaluation, this result
may reflect a tendency to continue ambiguous cases, pending more definitive decisions based on
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 173

further information. Higher ratings of the importance of examining cases at screening in terms of
their potential for plea bargaining, early disposition, and caseload reduction (Item Q5a) were
associated with higher odds of rejection at screening, perhaps reflecting a desire to make valid
distinctions among cases at screening with respect to fact patterns and associated strength of
evidence. Thus, the effect of the consistency scale and the effect of Item Q5a appear to be generally
consistent with each other and with the themes that emerged in the focus group discussions.
Ratings of the frequency of declining or dismissing cases “when the amount of time and effort
needed to obtain a conviction exceeds the benefits of the potential sentence” (Item 4c) had an
unexpected association with the odds of rejection. The higher a prosecutors’ rating of that item, the
lower the odds of rejecting a case. This appears to contradict the finding for Item Q5a and much of
the discussion in the focus groups. We have no clear explanation for this apparent contradiction.
The rank-order correlation between responses to Q5a and Q4c in the general survey was, in fact,
quite low (rho = .11), indicating respondents did not view those two items as addressing a common
issue. However, there are many ways the items could have elicited differing interpretations. For
example, respondents could have focused on dismissals rather than declinations in Item Q4c,
whereas Item Q5a is more explicit in drawing attention to the screening decision. Or, at a more
abstract level, respondents could have viewed Q4c as a more simplistic criterion and item Q5a as
reflecting a more reasoned approach, more reflective of their decision-making process.14
There were no significant interactions between the effects of prosecutor-level measures and
case-level measures on the odds of rejecting a case at screening.

In a later section of the report concerning reasons for dismissals, we discuss the related finding
that the majority of reasons prosecutors cited for rejecting cases at screening were related to
deficiencies in evidence, whereas few related to waste of time and effort relative to the potential
outcome.
14

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 174

The descriptive analysis of rejection rates presented at the beginning of this section showed that
prosecutors who responded to the factorial survey were remarkably consistent in their screening
decisions. This conclusion is further supported by an analysis of the variation in case-level model
parameters across prosecutors, from an intermediate model (not shown) which included only the
three case-level measures and no prosecutor-level predictors other than the average level of
responding across prosecutors. The HLM 6 estimates for the corresponding random effects show no
statistically significant deviations from the average effects for the case-level intercept (χ2=50.03,
df=57, p>.500), the offense seriousness coefficient (χ2=43.68, df=57, p>.500), the evidence rating
coefficient (χ2=48.38, df=57, p>.500), or the criminal history rating coefficient (χ2=32.31, df=57,
p>.500). Although introduction of prosecutor-level variables accounted for still more of the
variation among prosecutors and reduced the residual χ2 statistics still further, the magnitudes of
those effects were necessarily small, since so much of the variation was already accounted for by
case characteristics.
3.4.4 Factors Influencing the Screening Decision: The Primacy of Evidence and the Severity of the
Case
Findings from the administrative data analyses and factorial survey analyses clearly support the
prosecutors’ identification of strength of evidence and severity of the case as the primary
considerations at initial screening. In the analyses of factorial survey data, prosecutors’ ratings of
the strength of evidence was the only case-level variable with a statistically significant influence on
the screening decision, and the effect was dramatic. For actual person and property cases screened
in Southern County, direct measures of evidence, evidence-related victim characteristics, the
number of arrest charges, the type and seriousness of arrest charges, defendant race, and the

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 175

presence of codefendants all influenced screening outcomes, but the combined effects of evidencerelated measures outweighed the combined effects of the other factors.
Measures of the strength of evidence were not available for the analyses of actual felony drug
cases screened in Southern County or for any cases screened in Northern County. However, the
results of the statistical analyses indirectly suggested the importance of strength of evidence.
Whereas the models that incorporated measures of evidentiary strength for predicting acceptance of
person offense and property offense cases in Southern County were moderately strong (Nagelkerke
R2 = .36 and .28, respectively), the models for predicting acceptance for drug cases in Southern
County and all cases in Northern County, which lacked such measures, were much weaker
(Nagelkerke R2 equaled .11 for felony drug cases in Southern County and was similar for all crime
types in Northern County). Absent measures of evidence, there were significant effects of other
factors, many of which were related to the seriousness of the offense, including multiple arrest
charges and more serious arrest charges.

3.5 Charging
If a case is accepted for prosecution, an ADA determines what charges to file. This occurs
nearly simultaneously with the decision to accept or reject a case at screening. Although screening
decisions and charging decisions are conceptually distinct, in practice the screening decision is
necessarily linked to the initial charging decision. As discussed above, prosecutors must make
decisions about multiple arrest charges during the screening process. While a case may be accepted
for prosecution, individual charges may be rejected. Moreover, a prosecutor may add additional
charges to the case that were not part of the original arrest charges. The result may be an increase or
decrease in the severity of charges filed relative to the original arrest charges or an increase or
decrease in the aggregate number of charges filed relative to the original arrest charges.
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 176

In focus groups, several prosecutors also argued that the screening and charging decisions are
based on different information. As noted in Part 2, ADAs decide whether they can prove a case
using the strength of the evidence; they decide whether they should prove a case using primarily
offense severity and defendant criminal history. According to prosecutors, these latter factors
determine the charging decision. The DA in Southern County argued that evidence determines
whether a case gets in the door. Offense severity and defendant’s record determine the charge.
Indeed, prosecutors noted that strong cases do not necessarily result in more severe charges; in
other words, if two cases involve the same underlying conduct, one will not get a higher charge
simply because it has stronger evidence. As one unit manager in Northern County noted, “More
evidence does not make the defendant more culpable or the offense more serious.” Rather, a
separate calculation occurs in which prosecutors determine what should be charged based on similar
evaluations of convictability and fairness to the defendant, victim, and society.
The result of the charging decision determines the amount of exposure that defendants will face
– the number of charges that they may be convicted of and the most severe charge that they may be
convicted of. There were noticeable differences of opinion among the prosecutors regarding the
optimum approaches for deciding how many charges and which charges to file at initial screening.
In the general survey, prosecutors were asked to choose which of four options best describes their
general approach to charging. A plurality of respondents (47 percent) said they file only the charges
they believe the defendant should plead guilty to, while 23 percent said they file only the charges
they believe the defendant will plead guilty to, and 29 percent said they file all charges possible.
The pattern of responses differed significantly by jurisdiction (p=.000, χ2=41.059, df=1), with
respondents in Northern County more oriented toward filing only charges the defendant should
plead to and respondents in Southern County more oriented toward filing all charges possible. A

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 177

substantial majority (63 percent) also agreed or strongly agreed that the charging decision should
include the highest charges that could be proven at trial with the realization that these may be
reduced later through a plea bargain, but a non-negligible minority (17 percent) disagreed or
strongly disagreed. The average level of agreement was significantly higher among Southern
County prosecutors than among Northern County prosecutors (p=.000, F=39.000, df=1).
Figure 3.5-1 Responses to question: “Realizing that each case is unique, which of the following
best describes your general approach to charging?” by research site
60

Number of Prosecutors

50
40
30
20
10
0
I file all possible charges

I file only the most
I file only the charges I
serious charges possible believe the accused will
plead guility to

Northern County

I file only the charges I
believe the accused
should plead guilty to

Southern County

The charging decision involves a balancing act, particularly when multiple charges are possible.
Prosecutors have to balance being tough with efficiency. Some prosecutors want to charge the most
serious charges available and then negotiate down from those charges; others choose to charge what
they would like the defendant to plead guilty to and negotiate up if the defendant chooses not to
plead guilty. But there are dangers in both approaches. As one unit manager in Northern County
noted, if a prosecutor wants a case to be resolved, it is easier to resolve a case if you have “throw
away” charges; but eventually the defense attorney will know that this is what will happen if you
always include throw away charges and you will never get a top charge to stick. Others noted
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 178

dangers in charging all possible offenses, particularly when there are multiple counts of the same
offense. One ADA argued that it is good to avoid charging a lot of counts because a lot of counts
confuses the jury. For example, you would not want 300 counts of child pornography – one for
every photo on someone’s computer – because neither the judge, jury, nor ADA could keep track of
each count.
Some of the observed variability in opinions about charging strategies may be explained by
differences in the contingencies faced by prosecutors handling different types of cases. For
example, for misdemeanor cases in Southern County, initial charging is controlled by the
magistrate, and the ADA’s decisions about charges enter at the point of making a plea offer. Review
of the magistrates’ charges, the plea offer, the defendant’s response, and scheduling a trial if
necessary all take place in the courtroom within a span of a few minutes. Similarly, the initial
charging decision may tend to be closer to the final outcome for crime types for which cases are
frequently expedited (e.g., drug cases), and one unit manager in Southern County noted that the
initial charging decisions tend to stick. Conversely, unit managers in both jurisdictions noted that
charging decisions are affected by the likelihood of going to trial. A unit manager in Southern
County noted that if a guilty plea is likely, a prosecutor may get rid of extra charges early, whereas
if a case appears headed for trial, retaining multiple charges may help obtain a guilty plea or a
conviction at trial. Similarly, a unit manager in Northern County explained that if a case is likely to
go to trial, the prosecutor will charge enough crimes to ensure that all of the key facts of a case will
be presented to the jury, while trying to avoid introducing too much for the jury to handle.
Furthermore, prosecutors assigned to units that handle cases that often go to trial (e.g., domestic
violence and sex crime cases) tend to charge all cases as if they were going to trial. ADAs in the
domestic violence unit in Northern County described their charging style as “the boy scout way – be

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 179

prepared. When we charge, we are thinking about the trial. So, we charge enough so it represents
the incident and reflects all of the facts of the case.” This does not necessarily mean that ADAs
charge everything possible; only that the charges represent all conduct in the incident.
The starkest difference in opinion was seen between new ADAs and more experienced ADAs,
particularly in Southern County. As one new ADA in Southern County argued: it is better to be the
toughest person in the office rather than the easiest person. If you are the easiest person in the
office, then you should be a defense attorney. More experienced ADAs argued that a style of always
charging high and dismissing charges later undermined a prosecutor’s credibility when the case
actually warranted a very high charge; the defense and the judge will assume it is like other cases in
which the ADA simply reduced charges. This was echoed by an ADA in Northern County who
noted that charging based on desired exposure shows the prosecutor’s credibility to the defense
attorney and judge. We do not have to go through BS arguments about charges. Although many
prosecutors noted potential problems with charging high, they also noted that it was not something
that should necessarily be avoided. Unit managers in Southern County argued that loading up does
not mean that you do not charge what is appropriate. This may mean that someone simply takes the
time to look at all possible charges. Conversely, filing a limited number of charges or reducing
charges does not necessarily mean that a prosecutor is simply charging what they think the
defendant will plead guilty to. One experienced ADA in Southern County saw the charging decision
as an opportunity to correct police work, arguing that most of the time, the police overcharge and
the prosecutor can seek to correct this.
Indeed, while all prosecutors agreed that charging style was completely dependent on the ADA,
most also agreed that there was an office philosophy to not charge everything possible. As one unit
manager in Northern County noted, the office philosophy is “Charge what you want but do not

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 180

overcharge. The goal is to charge with the amount of exposure necessary for the defendant.
Therefore, charge enough to get the sentence that is appropriate.” Most of the justifications for
charging styles were consistent with prior research (Frohmann, 1997) finding that many
prosecutorial decisions are forward looking; in this case, several prosecutors argued that they think
about the harm caused and what the response to that harm should be. In this sense, most
prosecutors had an idea of either the sentence or the ultimate conviction charge that they wanted in
a case (this is discussed in more detail in the following section on plea offers).
We examined the charging decision, again, using two methods. First, we examined
administrative data capturing the number and severity of charges filed relative to the initial arrest
charges. Second, we examined factorial survey data capturing prosecutors’ charging decisions in
hypothetical cases.
3.5.1 Factors Influencing the Charging Decision: Administrative Data from Actual Cases
Analyses of administrative data in both sites revealed that, among cases accepted for
prosecution, there were few differences between the arrest charges and the charges initially filed by
the screening prosecutor. There were only slight differences between the average number of arrest
charges per case and the average number of filed charges per case; and, there were only slight
differences between the average severity of the most serious arrest charge and the average severity
of the most serious filed charge. Thus, overall, it appears that prosecutors in both jurisdictions
usually tended to either accept all of the charges presented for screening or reject them all.
For example, the most serious arrest charge was filed in roughly 90 percent of cases for all
offense types and offense levels for which data were available (Figure 3.5.1-1). In Northern
County, prosecutors were more likely to issue the top charge in misdemeanor cases than in felony
cases. In Southern County, prosecutors were slightly less likely to file the top arrest charge in drug
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 181

cases than in cases involving person or property crimes. The largest reduction in charges occurred
in felony domestic violence cases in Northern County, for which prosecutors issued the most
serious arrest charge in roughly 80 percent of cases.
Figure 3.5.1-1 Percent of cases in which most serious arrest charge was issued

Percentage of cases

100%

Southern County

Northern County

80%
60%
40%
20%
0%
Person Property

Drugs

Person Property

Felony

Drugs

Public
Order

Weapons

DUI

Misdemeanor

Since, for most crime types, the charging decision did not appear to be functionally separate
from the screening decision, we did not conduct a separate statistical analysis of factors affecting
the charging decision for actual cases accepted for prosecution in either site. Instead, we emphasize
analyses of the factors associated with post-filing dismissals and changes in the number and
seriousness of charges between filing and final case disposition (these analyses are described in
subsequent sections). Before turning to those analyses, however, we first present prosecutors’
charging decisions in the factorial survey.
3.5.2 Factors Influencing the Charging Decision: Hypothetical Cases
In addition to asking prosecutors whether they would accept a hypothetical case at screening,
the factorial survey also asked prosecutors to specify the number and severity of charge or charges

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 182

they would file. Similar to the analyses of screening decisions, we examined case-level and
prosecutor-level factors associated with variation in these charging decisions.
Descriptive Statistics
Among the 570 hypothetical cases with complete data (see section 3.4.3), prosecutors indicated
that 470 (82.5 percent) would be accepted for prosecution. However, in fourteen of those cases, the
respondents indicated that there was insufficient information provided in the vignette to determine
which specific charges should be filed, leaving 456 accepted cases with complete data for the caselevel measures. These represented responses from 60 of the 62 prosecutors who completed the
survey.
The charges listed by the respondents as those they would charge were coded to indicate the
number of charges and the statutory class of the top filing charge. For each case, the statutory class
of the top charge was assigned a numerical rank, with a value of “1” corresponding to the most
serious felony class and a value of “11” corresponding to any class of misdemeanor. Thus, in all of
the analyses reported in this section, high values for the rank variable correspond to low levels of
charge seriousness, and positive indices of association reflect inverse relationships with charge
seriousness.
The average number of charges prosecutors indicated they would file varied substantially by
vignette, from a low of 1.07 charges for vignette 8 (low offense seriousness, low evidence strength)
to a high of 5.02 charges for vignette 9 (medium offense seriousness, high evidence strength)
(Figure 3.5.2-1). The variation in the number of filing charges per case also differed among
vignettes, from a very narrow range for vignette 8 (range = 1; s.d. = .3) to a much wider range for
vignette 9 (range = 11; s.d. = 2.6).15
15

However, the ratios of standard deviations to vignette means were fairly similar across most vignettes—ranging from
33 percent of the mean to 42 percent of the mean for 8 of the 10 vignettes. The two vignettes with the most extreme

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 183

Figure 3.5.2-1 Mean number of charges to be filed per hypothetical case, by vignette
Mean number of charges to be filed

6
5
4
3
2
1
0
1

Vignette number

The average statutory ranks of top filing charges varied across vignettes, and were fairly evenly
distributed across the range from a mean of 6.70 for vignette 4 (high offense seriousness, medium
evidence strength) to a mean of 11.00 for vignette 8 (low offense seriousness, low evidence strength
(Figure 3.5.2-2). Variation in statutory rank of top filing charge within vignette was greatest for
vignette 4 (range = 8; s.d. = 2.27) and least for vignette 8 (range = 0, s.d. = 0, all misdemeanors).16

ratios of standard deviation to the mean were vignette 9 (2.581/5.020 => 51%) involving alleged forgery and theft and
vignette 8 (.258/1.070 => 24%) involving alleged failure to move on. The distribution for vignette 8 was also strongly
positively skewed (skewness = 3.59). The restricted relative variability and positive skew for vignette 8 were
consequences of the high rejection rate for that case.
16
As with the number of charges to be filed, the restricted range in statutory classification of top filing charge for
vignette 8 was a function of the high rejection rate and the downgrading of charges by prosecutors who reported they
would have accepted the case.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 184

Figure 3.5.2-1 Mean statutory rank of top filing charge, by vignette

Mean statutory rank of top filing charge

12
10
8
6
4
2
0
1

Vignette number
NOTE: A value of 1 = felony with highest statutory rank; 11 = misdemeanor.

The statistical models exploring factors influencing the charging decision incorporated the same
independent variables as the models examining the screening decision (see Section A.3.4 in
Appendix A for explanations and descriptive analyses of independent variables). There were only
slight differences in the distributions of the independent variables at the charging stage relative to
the screening stage, due to the rejection of some vignettes and offender criminal history scores. For
example, vignette 2 (medium offense seriousness, low evidence strength) resulted in extremely high
rejection rates at screening (87 percent) which resulted in obvious underrepresentation of responses
at charging. Whereas the previous analysis showed that the even distribution across categories
established by the factorial survey design was maintained among completed surveys, that balance
was not maintained among accepted cases. The same is true with respect to representation of
prosecutors’ ratings of strength of evidence across levels of offense seriousness. Whereas the
analysis of all completed surveys found significant numbers of “lowest” and “mid-low” evidence

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 185

ratings for low-level felonies, the analysis of accepted cases found relatively few ratings in those
categories. Nevertheless, average prosecutors’ ratings of strength of evidence and the variability of
those ratings were quite well balanced across categories of offense seriousness.17 Finally, among
cases prosecutors indicated they would accept for prosecution, the distribution of prosecutors’
ratings of the seriousness of defendants’ criminal histories across levels of offense seriousness
remained stable relative to screening.
The multi-level modeling of charging decisions incorporated the same prosecutor-level
variables as the analyses of screening decisions. Except for minor changes from one analysis to
another in the patterns of missing data, the two analyses were also based on responses from the
same set of prosecutors. Therefore, descriptive analyses of prosecutor-level independent variables
are not repeated here. Descriptions and analyses of the 12 prosecutor-level measures tested for
inclusion in multi-level models of charging decisions are provided in Section A.3.4 in Appendix A.
HLM Analyses
Two different aspects of the initial charging decision were modeled: the number of charges
prosecutors indicated they would file and the statutory class of the top filing charge. As with the
screening analyses, preliminary analyses found no independent relationship between defendant race
and either of the dependent variables, so the race variable was dropped from consideration in order
to conserve degrees of freedom given the relatively small sample available for analysis.
Nevertheless, the balance of race across seriousness categories could be indirectly relevant, if for
example, prosecutors’ ratings of defendant criminal histories were subtly related to defendant race.

Mean evidence ratings were 3.09, 3.35, and 3.63 for misdemeanors, low-level felonies, and high-level felonies,
respectively. The corresponding standard deviations were .89, .85, and .91—representing 29.8 percent, 25.5 percent,
and 25.1 percent of their respective means. The distributions of ratings within levels of offense seriousness were close
to normal, with skewness values of -.12, -.38, and -.027.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 186

As with the modeling of screening decisions, the statistical modeling of charging decisions
focused on three case-level independent variables: researchers’ classification of the seriousness of
the top arrest charges (offense seriousness), prosecutors’ ratings of the strength of evidence
(evidence rating), and the prosecutors’ ratings of the seriousness of defendant criminal histories
(criminal history rating). The analyses show adequate variability of responding for each of the
three dimensions, though some caution is warranted due to underrepresentation of low evidence
ratings for low-level felonies.18
Table 3.5.2-1 summarizes the results of the final two-level model for the number of charges
prosecutors indicated they would file.

The selection bias due to the exclusion of rejected cases does not appear to have resulted in much change in the
relationships among case-level independent variables. As was true for the broader sample of cases included in the
screening analyses, the inter-correlations among case-level independent measures for accepted cases exhibit substantial
independence. The rank order correlations of criminal history rating with offense seriousness and evidence rating were
non-significant and negligible (rho = -.002 and .003, respectively). Also, as in the analysis of screening decisions, the
rank order correlation between evidence rating and offense seriousness was statistically significant but small (rho =
.227, p=.000, n=463). These inter-correlations were not considered strong enough to threaten the ability to estimate
independent effects.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 187

Table 3.5.2-1 Hierarchical regression model for the number of charges to file
Effect
Case level
Intercept
Offense seriousness
Evidence rating
Criminal history rating
Seriousness by Evidence
Prosecutor level
Should plead (file only charges
to which the defendant
should plead guilty)a
Item Q8da (file highest
provable charges)
Prosecutor by case level
Evidence by Item Q8aa
(defendant should plead to all
charges filed)
a

Coefficient

T Ratio

P value

2.77
-0.20
-0.21
-0.16
0.31

37.15
-0.90
-1.56
-2.33
4.66

57
59
58
59
59

.000
.375
.124
.023
.000

-0.38

-2.72

.009

0.34

3.30

.002

-0.19

-3.67

.001

See Table A.3-2 for explanation of items Q8a, Q8d, and Should Plead.

As Table 3.5.2-1 indicates, at the case-level, there was a highly significant interaction between
the influence of offense seriousness and the influence of evidence rating. Though they were not
statistically significant, the main effects for seriousness and evidence were retained in the model to
aid in the interpretation of the interaction effect. The combined effect of the interaction and its
component main effects were such that the expected number of charges filed was approximately the
same across levels of offense seriousness when strength of evidence was rated at the lowest level.
The expected number of charges filed increased as the evidence rating increased, and the rate of
increase was much greater for more serious offenses than for less serious offenses. That is,

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 188

according to the model, offense seriousness mattered more when the evidence was perceived to be
strong, and, equivalently, strength of evidence mattered more for more serious cases.
The expected numbers of charges resulting from the combined evidence and seriousness effects
depend on choices of reference levels for the other variables in the model. However, a sense of the
magnitudes of the effects can be gained from Figure 3.5.2-3, which displays the pattern of the raw
numbers of charges specified by the respondents. In this case, the model and the raw data yield
nearly identical conclusions, except that the graph of the raw data suggests that the evidence effect
may be nonlinear for low-level felonies, and that the effect of evidence for low-level felonies may
only diverge from the effect for high-level felonies at the highest levels of evidentiary strength.
However, for low-level felonies, there were only 5 observations at the lowest evidence level and 11
observations at the highest level. It may be that the only reliable distinction is the greater effect of
strength of evidence for felonies, relative to the effect for misdemeanors.
Figure 3.5.2-3 Mean number of charges to be filed per hypothetical case by offense seriousness
and prosecutor’s rating of strength of evidence (among cases accepted for prosecution)
Mean number of charges to be filed

6
5
4
3
2
1
0
Lowest

Mid-Low

Average

Mid-High

Highest

Strength of evidence (prosecutor's rating)
Misdemeanor

Low level felony

High level felony

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 189

The model also suggests a counter-intuitive effect of criminal history rating—with higher
ratings of the seriousness of criminal history being associated with slightly fewer filing charges. The
graph of the raw data in Figure 3.5.2-4 also hints at complex interactions, in which the average
number of filing charges increases in the range from lowest to average criminal history ratings for
misdemeanors but decreases for felonies, and then the numbers of charges diverge again in the
range from average to highest criminal history ratings. Even if reliable, these would be interactions
among nonlinear effects, and our linear modeling found no statistically significant interaction
effects for criminal history ratings. While the exact nature of the relationship is difficult to discern
and interpret, it is clear that, overall, ratings of the seriousness of criminal history were not
consistently associated with greater numbers of filing charges. Thus, at least for these hypothetical
cases, the role of criminal history in the initial charging decisions appears to be more complex than
might be inferred from the focus group discussions.
Figure 3.5.2-4 Mean number of charges to be filed per hypothetical case by offense seriousness
and prosecutor’s rating of defendant’s criminal history (among cases accepted for
prosecution)

Mean number of charges to be filed

5
4
3
2
1
0
Lowest

Mid-Low

Average

Mid-High

Highest

Defendant criminal history (prosecutor's rating)
Misdemeanor

Low level felony

High level felony

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 190

Three of the prosecutor-level measures incorporated in the analyses were significantly related to
the average number of filing charges. The measures that proved significant were prosecutors’
responses to three individual items on the general survey. The more strongly prosecutors agreed that
“the charging decision should include the highest charges that could be proven at trial, with the
realization that these may be reduced later through a plea bargain” (item Q8d), the greater the
average number of charges the prosecutors indicated they would file. Similarly, prosecutors who
indicated that their general approach to charging was to “. . . file only the charges I believe the
accused should plead guilty to” (item Q7, response option 4) identified fewer charges for initial
filing than prosecutors who chose other responses. In addition, there was a significant prosecutorlevel by case-level interaction, such that the modeled influence of strength of evidence was
attenuated among prosecutors who agreed more strongly that “a plea offer should include all of the
charges filed, with an offer to forego additional charges if the offer is accepted” (item Q8a). This
latter result could simply reflect the influence of a general “seek-everything” orientation; if a
prosecutor aims to file as many charges as possible, the strength of the evidence may become less
relevant. On the other hand, it is possible that prosecutors do tend to file fewer charges initially if
they believe that defendants should ultimately plead guilty to all charges filed.
The variation in case-level model parameters across prosecutors was analyzed by examining the
results for an intermediate model (not shown) which included only the three case-level measures
and the case-level seriousness-by-evidence interaction, as well as prosecutor-level intercepts
reflecting average estimates of the random case-level parameters. The HLM 6 estimates for the
corresponding random effects show no statistically significant deviations from the average offense
seriousness coefficient (χ2=26.75, df=46, p > .500), the average evidence rating coefficient
(χ2=23.22, df=46, p > .500), or the coefficient for seriousness-by-evidence interaction (χ2=28.74,

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 191

df=46, p > ,500). However, there was significant variation among prosecutors in the influence of
criminal history ratings (χ2 for slope coefficient = 78.03, df=46, p = .002) and marginally significant
variation among prosecutors in the average number of charges they indicated they would file (χ2 for
intercept = 60.05, df=46, p = .080). In the final model, some of the variation among prosecutors in
average response was accounted for by their responses to selected general survey items (resulting χ2
for intercept = 48.80, df=44, p = .286), demonstrating that at least some of the variation in decisions
among prosecutors was attributable to differences among prosecutors in their beliefs about
appropriate prosecutorial strategies. However, inclusion of the general survey items did not reduce
the residual variance for the criminal history coefficient, leaving unexplained the differences among
prosecutors in the influence of criminal history on charging decisions.
Table 3.5.2-2 summarizes the results of the final two-level model for the seriousness of top
filing charge. It suggests a highly significant effect of prosecutors’ ratings of strength of evidence
in the counter-intuitive direction—higher statutory rank number (lower seriousness) appears
associated with stronger evidence. However, the combination of effects for evidence rating,
seriousness of the top arrest charge, and a highly significant interaction between seriousness and
evidence together yielded a different overall pattern. None of the prosecutor-level measures we
tested exhibited significant relationships with either the average level of response (case-level
intercepts) or the average influence of case-level predictors (case-level regression coefficients)
(thus, we do not display them in Table 3.5.2-2).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 192

Table 3.5.2-2 Hierarchical regression model for statutory rank of top filing charge
Effect

Coefficient

T Ratio

P value

Case level
Intercept
Offense seriousness
Evidence rating
Criminal history rating
Seriousness by evidence
Seriousness by history

8.99
-0.49
0.47
-0.27
-0.36
0.12

115.99
-1.24
2.74
-2.08
-3.70
1.45

59
59
59
59
59
59

.000
.221
.009
.042
.001
.153

Prosecutor level
[No significant effects]

Note: The dependent variable, rank of top filing charge, is scaled with smaller numerical values corresponding to more
serious charges. Consequently, effects with positive signs in the above table reflect inverse relationships with
the seriousness of top filing charge, and vice versa.

The most serious top arrest charges (high-level felonies) were associated with the most serious
top filing charges, the least serious top arrest charges (misdemeanors) were associated with the least
serious top filing charges, and low-level felony arrests were associated with intermediate-level top
filing charges. For high level felony arrests, higher evidence ratings were strongly associated with
more serious top filing charges. For misdemeanor arrests, seriousness of top filing charge either
decreased slightly across increasing levels of evidentiary strength (according to the model) or
increased slightly across increasing levels of evidentiary strength (compare Figure 3.5.2-5, which
displays the raw data patterns). For low-level felonies, the results were intermediate with respect to
both average level of top filing charge and the slope of the relationship between statutory level of
the top filing charge and prosecutors’ ratings of the strength of evidence.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 193

Figure 3.5.2-5 Mean statutory rank of top filing charge by offense seriousness and
prosecutor’s rating of strength of evidence (among cases accepted for prosecution)

Mean statutory rank for top filing charge

12
10
8
6
4
2
0
Lowest

Mid-Low

Average

Mid-High

Highest

Strength of evidence (prosecutor's rating)
Misdemeanor

Low level felony

High level felony

The model shows no significant relationship between prosecutors’ ratings of the seriousness of
defendants’ criminal histories and the seriousness of the top filing charge, and a graph of the raw
data patterns suggests the same interpretation (Figure 3.5.2-6).

Mean for statutory rank of top filing charge

Figure 3.5.2-6 Mean statutory rank of top filing charge by offense seriousness and
prosecutor’s rating of criminal history (among cases accepted for prosecution)
12
10
8
6
4
2
0
Lowest

Mid-Low

Average

Mid-High

Highest

Defendent criminal history (prosecutor's rating)
Misdemeanor

Low level felony

High level felony

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 194

The HLM 6 estimates for the random variation of case-level regression parameters across
prosecutors show statistically significant deviations from the average coefficients for all of the caselevel effects included in the final model: case-level intercept (χ2 = 71.89, df = 45, p = .007),
offense seriousness (χ2 = 71.89, df = 45, p = .002), strength of evidence (χ2 = 71.89, df = 45, p =
.051), criminal history rating (χ2 = 71.89, df = 45, p = .000), and the interaction between offense
seriousness and strength of evidence (χ2 = 71.89, df = 45, p = .002). None of this residual variation
(after controlling for case seriousness, strength of evidence ratings, and criminal history ratings)
could be reliably attributed to any of the prosecutor-level measures tested in these analyses. With
respect to the seriousness level of the top charges prosecutors indicated they would file in these
hypothetical cases, the significant variation among prosecutors in their average levels of
responding, and the variation in how they weigh offense seriousness, strength of evidence, and
criminal history in their initial charging decisions, remain unexplained by these analyses.

3.6 Plea Offers
While the charging and plea offer decisions can occur far apart in time, they are not entirely
separate. Prosecutors often approach charging in anticipation of the plea offer and outcome. This
was evident in prosecutors’ discussions of what guided their plea offers. Prosecutors noted that plea
offers and changes to plea offers were often guided by a desire to either get a certain sentence or get
a conviction on a certain charge. One prosecutor noted that if we know the amount of time we want
the defendant to receive, then we can be creative in what we offer as a plea. It is often about getting
the right sentence. According to the DA in Southern County, the plea offer is based on the likely
sentence that would be imposed at trial. The ADA has to read the file and judge what a jury is
likely to do and, if convicted, what the likely sentence would be. Then the ADA says, ‘This is the
worst thing that will happen to the defendant; therefore, what plea offer would get the defendant to
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 195

plead guilty rather than be exposed to the likely sentence.’ Another ADA noted that in other
instances, it is about getting the right conviction offense, such as sex offender notification or a drug
offense with a gun. Sometimes we would rather the defendant’s record reflects what the defendant
did, to better reflect the crime committed, not just what is the appropriate sentence. This was
echoed by the unit manager of the sensitive crimes unit in Northern County who argued that in
sexual assault cases, if the victim does not want to testify, then we make sure to have enough
charges to ensure a felony conviction. Prosecutors also argued that the decision was not always
entirely up to the prosecutor. As one ADA in Northern County noted, “This is also dependant on
the judge. If the judge changes sentences from those in plea offers, then sometimes all that sticks
around is the charge.”
These sentiments were echoed in the general survey. There was no clear consensus in the
responses to the general survey as to what aspect of charging was most important to consider in
preparing plea offers. Half of the respondents (50 percent) agreed or strongly agreed that plea
bargaining should focus primarily on the expected sentence the defendant is facing, while 19
percent disagreed or strongly disagreed (Figure 3.6-1). Fifty-five percent of respondents agreed or
strongly agreed that plea bargaining should focus primarily on the severity of the most serious
charge, while 18 percent disagreed or strongly disagreed (Figure 3.6-2). A slight majority (54
percent) disagreed or strongly disagreed that plea bargaining should focus primarily on the number
of charges the defendant is facing, while only 10 percent agreed.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 196

Percentage of Respondents
within County

Figure 3.6-1 Responses to question: “Plea offer should focus primarily on length of sentence,”
by research site
60
55
50
45
40
35
30
25
20
15
10
5
0
Disagree

Neutral

Northern County (N=60)

Agree
Southern County (N=65)

Percentage of Respondents
within County

Figure 3.6-2 Responses to question: “Plea offer should focus primarily on offense severity,” by
research site
60
55
50
45
40
35
30
25
20
15
10
5
0
Disagree
Northern County (N=61)

Neutral

Agree
Southern County (N=64)

While prosecutors agreed that plea offers often reflected a desire to ensure a certain sentence,
they disagreed about whether the plea offer should include a recommendation for a particular
sentence or whether the sentence was something that could be negotiated. For example, some ADAs
maintained that they do not make specific sentence recommendations, but rather leave it up to the
judge. Others disagreed and maintained that it is the prosecutor’s responsibility to declare what the
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 197

state believes the proper sentence would be. As one ADA in Northern County noted, the sentence
recommendation also is about the credibility of the ADA. If the ADA is seen as fair to the defendant,
then extraordinary recommendations – for example, sentence recommendations that are very high
or very low – will be taken seriously by everyone involved. Others argued that the sentence
recommendation in the plea offer did not mean much in the end. One ADA in Northern County
maintained that the sentence recommendation is symbolic – the judge will do whatever he or she
wants to do.
Given this conflicted view of the sentence recommendation, prosecutors had conflicted views
about whether they would negotiate the sentence with the defense. In most cases, prosecutors stated
that they did not negotiate the sentence. However, some argued that it was about balancing
efficiency concerns. As one ADA in Northern County noted, “If the sentencing recommendation is
holding up the plea, then I may bend on the sentencing recommendation.” Prosecutors also varied in
their opinion of whether the plea stage was simply an offer or a negotiation. Generally, prosecutors
felt the plea was an offer that could be accepted or rejected, but not negotiated. However, the DA in
Southern County noted that if the defendant can provide additional information, the ADA should
listen and re-evaluate the plea offer.
Prosecutors who responded to the general survey had differing opinions about the best strategies
for achieving the desired end results. They were equally divided as to whether a plea offer should
include all of the charges filed with an offer to forego additional charges if the plea is accepted (31
percent agreed, 33 percent disagreed, and 36 percent were neutral) (Figure 3.6-3). A nearly opposite
approach, suggesting that a plea offer should include only the charges the defendant should plead
guilty to with a threat to add additional charges if the plea is not accepted, also elicited divergent
opinions (27 percent agreed, 46 percent disagreed, and 37 percent were neutral) (Figure 3.6-4). Our

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 198

analyses of hypothetical cases in the factorial survey suggest that strategic preferences such as these
account for some of the inconsistency among ADAs in critical case processing decisions.
Figure 3.6-3 Responses to question: “Plea offer should include all charges filed with an offer
to forego additional charges if the plea offer is accepted,” by research site
50
Percentage of Respondents
within County

45
40

35
30
25
20
15
10
5
0
Disagree

Neutral

Northern County (N=61)

Agree

Southern County (N=64)

Figure 3.6-4 Responses to question: “A plea offer should include only the charges the
defendant should plead guilty to with a threat to add additional charges if the plea offer is not
accepted,” by research site
50
45
Percentage of Respondents
within County

40
35
30
25
20
15
10
5
0
Strongly disagree

Disagree

Northern County (N=55)

Neutral

Agree

Southern County (N=62)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 199

One concern among prosecutors was the need for consistency across ADAs in plea offers. In
Southern County, a common practice across the units is to do a roundtable involving all members of
the unit to discuss plea offers. This promotes consistency and uniformity in plea offers. As the unit
manager of the drug unit in Southern County noted, the unit roundtables all cases before a plea
offer; whatever the majority vote is, that is the offer. The drug unit prosecutes cases horizontally,
meaning that no single attorney handles an entire case; rather, different ADAs may handle different
aspects of the case. Therefore, the unit is concerned with consistency because everyone on the unit
will see the work. The roundtable approach, however, can lead to unintended consequences; since
the person offering the plea is not the person taking the case to trial, there is an incentive to be
tough at plea offer. In contrast, the person crimes unit in Southern County prosecutes cases
vertically, so that each ADA has his or her own inventory; as a result, according to the unit manager
of the person crimes unit, unit members are less cognizant about consistency across ADAs. There is
less peer pressure, but there is still a concern for consistency. In Northern County, prosecutors were
similarly conflicted about the need for consistency in plea offers. As one ADA in Northern County
argued, “What we decide a person should get often depends on the judge. So, we tailor offers to the
judge.” Others disagreed and argued that the offer for a specific offense should be the same every
time. Similar to the charging decision, the plea offer was based primarily on defendant
characteristics rather than strength of the evidence.
Prosecutors maintained that the strength of the evidence will bring a recommendation down but
it will not bring it up. In other words, we negotiate downward later if the case is weak, but we do
not raise the charges upward just because it is strong. The focus on the characteristics and
circumstances of defendants in the preparation of plea offers is a reflection of prosecutors’ efforts to
do the right thing with respect to appropriate consequences for the defendants. Potential

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 200

consequences include a wide range of possibilities, such as expected sentence type and sentence
length, the length and seriousness of the resulting official criminal record, payment of fines and
restitution, obligations such as community service or participation in treatment, and restrictions on
personal liberty – all of which can depend wholly or in part on the number and severity of
conviction charges.
We examined the plea offer decision using two methods. First, we examined administrative data
capturing the plea offers made in Southern County. Data on plea offers was unavailable in Northern
County. Therefore, the following section details plea offers in Southern County only. Second, we
examined factorial survey data capturing prosecutors’ plea offer decisions in hypothetical cases.
3.6.1 Factors Influencing the Plea Offer Decision: Administrative Data from Actual Cases
The analyses of plea offers in Southern County examined several aspects of the plea offer
decision. First, we examined changes in the number of charges facing defendants, by comparing
the filed charges to the charges listed in the plea offer. Second, we examined changes in the
seriousness of charges a defendant was facing, by comparing the seriousness of the top charge at
filing to the top charge to which the plea offer required a guilty plea. Finally we examined the
associated sentence recommendations, looking at three variables related to sentencing: 1) top charge
incarceration exposure – an estimate of the number of months of incarceration a defendant could
face if convicted of the top charge; 2) aggregate incarceration exposure – an estimate of the number
of months of incarceration a defendant could face if convicted of all the active charges; and 3),
whether the plea offer recommended a period of incarceration.19

The actual threat of incarceration is also strongly determined by a statutorily defined prior record scale under
Southern State’s sentencing guidelines; prior record level, however, is not subject to prosecutorial discretion. Thus,
these exposure measures reflect only the potential consequences of charge seriousness, but array those consequences on
interval scales.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 201

The following sections detail separate analyses for each of these outcome variables. The
analyses in each section rely on the same samples of cases that advanced beyond indictment/grand
jury in Southern County. These included 253 person unit cases, 249 property unit cases, and 2,232
drug unit cases. Rather than summarize descriptive statistics here, brief summary statistics for the
outcome variables are presented at the beginning of each section.
Explanatory variables included in the analyses differed somewhat across the three data sets.
Measures of defendants’ personal characteristics, prior record, and case seriousness at intake were
available from all three sources. Victim characteristics and strength of evidence were available only
for the person unit and property unit cases. Finally, a measure of concurrent experience was
constructed by counting the number of cases handled by each prosecutor within each of the three
teams during the time period covered by the associated data set.
The potential influences of available case-level factors on changes from initial filing to plea
offer were evaluated using linear regression modeling. Analyses of the changes in number of
charges were restricted to cases in which more than one charge was filed initially. For the analyses
of actual drug cases, the early drug unit sample and the late drug unit sample were combined to
permit explicit tests of the differences between time periods in outcomes and difference between
time periods in the relative influence of case characteristics. Analyses of factors influencing the
probability of an incarceration recommendation were based on logistic regression models. In
addition, logistic regression modeling and a series of one-way ANOVAs were conducted to explore
the variability in average outcomes across ADAs.
Number of charges requiring guilty pleas.
Overall, the average number of charges for which the plea offer would require guilty pleas was
similar across the broad crime categories, ranging from an average of 1.3 charges requiring a guilty

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 202

plea for the late drug unit cases to 1.7 charges requiring a guilty plea for the property unit cases (See
Table 3.6.1-1).20 For the drug unit samples, this represented substantial reductions from the average
number of charges initially filed. Late drug unit cases involved 1.5 fewer charges in the plea offer
than initially filed; early drug unit cases involved 1.2 fewer charges in the plea offer than initially
filed. Property unit cases saw the smallest reduction, with these cases involving just 0.6 fewer
charges in the plea offer than initially filed.
Table 3.6.1-1 Number of charges filed and requiring a guilty plea in plea offers

Selected Outcome Measures
Total N of cases with plea offer
% of initially screened cases
Avg. N of charges filed in cases with plea offersa
Avg. N of charges requiring guilty plea in offer
Avg. reduction in N of charges
a

Person
Unit
253
49.8%
2.5
1.6
0.9
(251)b

Property
Unit
249
37.8%
2.3
1.7
0.6
(248)

Early
Drug
Unit
1359
54.9%
2.7
1.5
1.2
(1335)

Late
Drug
Unit
873
36.1%
2.8
1.3
1.5
(857)

Calculations for N of filing charges and reduction in N of charges are limited to cases with plea offers.
Numbers in parentheses indicate N of cases with non-missing values for both N of filing charges and N of plea
offer charges.

Table 3.6.1-2 presents the results of the linear regression models examining changes in the
number of charges per case from initial filing to plea offer. Changes in the number of drug charges
per case were reliably associated with several case-level characteristics (N = 1653, R2 = .56).
However, the average effects were relatively small in absolute magnitude. Factors that tended to
increase the average number of charges were black defendant (+.11 charges) and prior record (+.05
charges per increase in level). Factors that tended to reduce the average number of charges were
class of the top charge at screening (-.03 charges per increase in statutory class), number of charges
at screening (-.95 plea offer charges per charge screened), and the time period during which the case

The somewhat lower average number of charges for the late drug unit sample appears to have been more than offset
by retention of more serious charges. See the analyses of incarceration exposure.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 203

was processed (-.27 charges for cases processed in the more recent time period). Clearly, more
serious cases with more room for plea bargaining did, in fact, experience greater reductions in the
number of charges between initial filing and plea offers. The same effect was observed for both the
early drug unit sample and the late drug unit sample, but the reductions between initial charging and
plea offer were somewhat greater overall in the later sample.
Table 3.6.1-2 Final linear regression models for effects of selected factors on change in
number of charges from initial filing to plea offer in Southern County
Independent Variables
Model Constant
Case Characteristics
Late drug sample
N of charges screened (up to 5)
Statutory class of top charge at screeninga
Top charge screened was burglary

ns
ns
+.797***

Defendant Characteristics
Race (black)
Prior record level

ns
ns

Victim Characteristics
(none significant)

-.534***
+.667**

-.342**
+1.236***

Interactions
Robbery BY N of charges screened
Robbery BY gender (male)
Robbery BY N of evidence items
Burglary BY female victim

Coefficients
Model 2
Property
Unit
-1.035

ns
ns

Evidence
Total N of items of evidence listed (up to 5)
N of items of evidence per intake charge (3 levels)

Model 1
Person
Unit
-.534

Pairwise Ns
Model R2

-.690***
+1.021***
+.443***
159-192
.32

Model 3
Drug
Unit
+.823
-.271***
-.951***
+.031**

+.112**
+.045***

-.681*

161-171
.11

1653 - 1776
.56

p<=.10; **p<=.05; ***p<=.01; for coefficients without asterisks p<=.15

Statutory class was coded as 1 = highest rank; larger values correspond to less serious charges. Thus, coefficients with
positive signs denote inverse relationships with seriousness.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 204

Prediction models for changes in the number of charges were much weaker and somewhat more
complicated for the person unit sample (N = 159-192, R2 = .32) and the property unit sample (N =
161-171, R2 = .11).21 For the person unit sample, several effects were limited to the subset of cases
arrested for robbery. For the property unit sample, several effects were limited to the subset of cases
arrested for burglary. That is, they were governed by significant interactions with robbery and
burglary, respectively. In the person unit sample, the net result was that, for robberies, a greater
number of items of evidence per charge and whether the alleged robber was female were associated
with increases or smaller reductions in the number of plea offer charges, and a greater number of
charges at intake was associated with greater decreases in the number of plea offer charges. For
other crimes against persons, none of the factors tested explained the observed reduction in the
number of charges. In particular, it was somewhat surprising that neither defendant characteristics,
nor prior record level, nor any of the measures of victim characteristics were significantly related to
the reduction in number of charges for crimes against persons, and that strength of evidence
remained as the dominant consideration.
Strength of evidence also emerged as the most influential factor in analyses of the number of
plea offer charges in the property unit sample. A three-level grouping of the number of items of
evidence per charge (<1, 1 to 2, 2 or more) was associated with an average increase of 1.2 charges
per level between the number filed and the number for which the plea offer would require guilty
pleas. However, this was partially offset by the total number of items of evidence forwarded to the
prosecution from the police (-.34 charges per item, up to a maximum of 5 items). This combination
of effects and the similar combination observed for the person unit sample suggest that more
21

Sample sizes are reported as ranges for regressions performed with pairwise deletion for missing values. For those
analyses, the larger value for N is the number of cases that had non-missing values for the dependent variable, and the
smaller value for N is the smallest number of cases involved in estimating the covariance between any pair of
variables—among only the variables included in the final model. Though listwise deletion is usually preferred, it was
decided for these analyses that the potential selection bias associated with listwise deletion could be more damaging
than the problems associated with pairwise deletion.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 205

complex cases may be more vulnerable to reductions in the number of charges. In addition,
burglaries were associated with a significant average increase in the number of charges (+.80
charges), except when one or more of the victims was female, perhaps reflecting prosecutors’
greater confidence in the testimony and other evidence associated with institutional victims
(businesses, “the State”, etc.).22
Statutory rank of the most serious charge requiring a guilty plea.
Overall, there were only slight changes in average statutory rank of the top charge between
initial filing and plea offer (Table 3.6.1-3). Statutory rank of charges is coded so that less serious
charges have a higher value in the coding (i.e. highest level felony =1 and misdemeanor =11); thus,
a positive change in the statutory rank of charges would imply a reduction in the seriousness of
charges and a negative change would imply an increase in seriousness of charges. As Table 3.6.1-3
indicates, person unit and property unit cases experienced a decrease in charge seriousness from
initial filing to plea offer, with late drug unit cases experiencing a slight increase.
Table 3.6.1-3 Average statutory rank of top filed charge and top charges requiring a guilty
plea in plea offers

Selected Outcome Measures
Total N of cases with plea offer
% of initially screened cases
Avg. statutory rank of top filed charge for cases
with plea offersa
Avg. statutory rank of top plea offer chargec
Avg. change in statutory rank of top charge
a

Person
Unit
253
49.8%

Property
Unit
249
37.8%

Early
Drug
Unit
1359
54.9%

Late
Drug
Unit
873
36.1%

7.2

9.1

8.7

8.1

7.9
+0.7
(251)b

9.3
+0.2
(248)

8.6
0.0
(1335)

8.0
-0.1
(857)

Calculations for statutory rank and change in statutory rank are limited to cases with plea offers.
Numbers in parentheses indicate N of cases with non-missing values for both rank of top filed charge and rank of
top plea offer charge.
c
Highest rank = 1. Positive change in rank value indicates a reduction in seriousness.
b

Presence of a female victim in a burglary case may serve here as a proxy for household burglary.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 206

Table 3.6.1-4 presents the results of the linear regression models examining changes in the
statutory rank of the top charge from initial filing to plea offer. As Table 3.6.1-4 indicates, the
relatively small changes in statutory rank of the top charge resulted in a relative inability to identify
significant influences. This was particularly true for the person unit sample due to the small
available sample size (N = 176 – 251, R2 = .05). There was, however, a statistically significant
effect of the number of charges presented at intake, such that each increase in the number of charges
(up to a maximum of 5) was associated with an average reduction of .33 levels in statutory class
(recall, the relationship is inverse). This is consistent with other findings suggesting that higher
initial starting points tend to be associated with greater reductions later.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 207

Table 3.6.1-4 Final linear regression models for effects of selected factors on change in
statutory class of top charge from initial filing to plea offer in Southern County
Independent Variables
Model Constant
Case Characteristics
Late drug sample
N of charges screened (up to 5)
Statutory class of top charge at screeninga
Top charge screened was robbery
Top charge screened was burglary

Model 4
Person
Unit
-.010

+.331***
ns
ns

Coefficients
Model 5
Property
Unit
4.194

Model 6
Drug
Unit
+2.417

+.131*
-.497***

ns
ns
-.171***

-.625***

Defendant Characteristics
Race (black)
Gender (= male for drugs; female otherwise)
Prior record level
Defendant age at screening (4 levels)

ns
ns
ns
ns

ns
+.364*
ns
ns

Victim Characteristics
(none significant)

Evidence
Total N of items of evidence listed (up to 5)
N of items of evidence per intake charge (3 levels)

ns
ns

Interactions
Late drug sample BY prior record level
Pairwise Ns
Model R2

-.212***
-.206***
-.144***
-.079**

-.059**
176-251
.05

198-248
.26

.08

p<=.10; **p<=.05; ***p<=.01; for coefficients without asterisks p<=.15
a
Statutory class was coded as 1 = highest rank; larger values correspond to less serious charges. When the independent
variable is statutory class of top charge screened, the sign of the coefficient represents a “double negative”; a positive
sign means more serious intake charge is associated with increase in seriousness between filing and plea offer. For all
other independent variables, the inverse is true; a positive sign denotes a decrease in seriousness of the top charge
between initial filing and plea offer.

Analysis of the property unit sample yielded somewhat stronger prediction (N = 198 – 248, R2 =
.26). Greater reductions in seriousness of top charges between initial filing and the plea offer were
associated with greater numbers of charges at intake, higher statutory class of the top charge at

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 208

intake, and female defendants. Cases in which the top charge at intake was for burglary experienced
less reduction, on average, than other cases.
For drug cases, the much larger sample yielded statistically significant effects for several
predictors. Consistent with the findings for property cases, reductions in statutory class between
initial filing and plea offer were greater in cases with higher statutory class at intake. However,
defendants with more serious prior records, black defendants, male defendants, and older
defendants experienced less reduction, on average, than other defendants. In addition, the
enhancing effect of prior record was stronger for the more recent drug unit sample than for the
earlier cases. Yet, overall, these measures did not account for much of the slight variation in
changes in class rank (N = 2045 – 2192, R2 = .08).
Change in top charge exposure
The potential minimum number of months of incarceration implied by the most serious active
charge increased only slightly between initial filing and plea offer for drug cases and declined only
slightly for person and property cases (Table 3.6.1-5).
Table 3.6.1-5 Average months of exposure for top filed charge and for top charges requiring a
guilty plea in plea offers

Selected Outcome Measures
Total N of cases with plea offer
% of initially screened cases
Avg. months of exposure for top filed charge
among cases with plea offersa
Avg. months of exposure for top plea offer charge
Avg. change in months of exposure
a

Person
Unit
253
49.8%

Property
Unit
249
37.8%

Early
Drug
Unit
1359
54.9%

Late
Drug
Unit
873
36.1%

41.4

9.2

12.2

18.9

36.6
-7.9
(223)b

8.2
-1.0
(239)

14.1
+1.9
(1331)

21.0
+2.1
(854)

Calculations for exposure for top filed charge and change in exposure are limited to cases with plea offers.
Numbers in parentheses indicate N of cases with non-missing values for top charge exposure at both filing and
plea offer.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 209

Table 3.6.1-6 presents the results of the linear regression models examining changes in months
of exposure from initial filing to plea offer. As Table 3.6.1-6 shows, several case-level factors were
significantly associated with these changes. For cases in the person unit sample (N = 153 – 223, R2
= .26), greater decreases in top charge exposure were associated with more charges at intake, more
victims willing to prosecute, and cases in which the top intake charge was for robbery. These
seemingly counterintuitive results are consistent with other findings in this study suggesting that
cases that start at a higher level, or are more complex, or possibly are more prone to overcharging
are more vulnerable to reductions later in case processing. Increases (or lesser decreases) in top
charge exposure were associated with a greater number of victims with physical injuries (up to a
maximum of 2), but this was offset by an average negative effect if any of the victims’ injuries were
categorized as “serious.” This too is counterintuitive, but may reflect a determination by
prosecutors, once victims have been interviewed and medical evidence has been evaluated, that a
significant number of such cases were initially overcharged.
For cases in the property unit sample (N = 192 – 239, R2 = .49), greater reductions in top charge
exposure were associated with older defendants, black defendants (especially in cases with multiple
charges), and cases involving codefendants. Lesser reductions23 in top charge exposure were
associated with defendants with more serious prior record, lower statutory rank of the top charge at
intake, a greater number of black or Hispanic victims (up to a maximum of 2), cases in which the
top charge at intake was burglary, and the number of victims with physical injuries (especially for
cases in which the top charge was burglary).

The model constant was -71.5 months, reflecting some degree of reduction for almost all cases.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 210

Table 3.6.1-6 Final linear regression models for effects of selected factors on change in top
charge incarceration exposure from initial filing to plea offer in Southern County
Independent Variables
Model Constant
Case Characteristics
Late drug sample
N of charges screened (up to 5)
Statutory class of top charge at screeninga
Top charge screened was robbery
Top charge screened was burglary
codefendants
Defendant Characteristics
Race (black)
Gender (= male for drugs; female otherwise)
Prior record level
Defendant age at screening (4 levels)
Victim Characteristics
Any victim with serious injury
Number of injured victims (up to 2)
N of victims willing to prosecute
N of black or Hispanic victims (up to 2)
Evidence
Total N of items of evidence listed (up to 5)
N of items of evidence per intake charge (3 levels)
Interactions
Late drug sample BY prior record level
Multiple charges BY class of top charge at
screening
Burglary BY N of injured victims
Multiple charges BY black defendant
Pairwise Ns
Model R2

Model 7
Person
Unit
+1.666

-5.250***
ns
-10.081***

Coefficients
Model 8
Property
Unit
-71.501

Model 9
Drug
Unit
-28.269

+7.550***

ns
ns
+2.001***

+6.581***
-2.632**

ns
ns
ns
ns

-12.797***
+5.575***
-5.043***
ns

ns
ns
ns

ns
ns
+2.674***
-1.718***

+2.698***
+2.355***
+1.688***
+1.105**

ns
ns
ns
+1.843

ns
ns
ns
+0.769***

+1.120**
ns
ns
153-223
.26

ns
+17.916***
-3.455**
192-239
.49

ns
ns
ns
2045-2192
.08

*p<=.10; **p<=.05; ***p<=.01; for coefficients without asterisks p<=.15
a
Statutory class was coded as 1 = highest rank; larger values correspond to less serious charges. Thus, coefficients with
positive signs denote inverse relationships with seriousness.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 211

For drug cases, the ability to explain changes in top charge exposure between initial filing and
plea offer was substantially more limited (N = 2045 – 2192, R2 = .08). The factors found to
influence the change in top charge exposure were similar to those identified as influential for the
person unit and property unit samples, but not necessarily in the same direction. As with the person
and property unit samples, higher statutory classification of the top charge at intake was associated
with greater reduction in top charge exposure. Similar to the results for the property unit sample,
more serious prior record classification was associated with less reduction in top charge exposure,
and the effect was stronger for the more recent drug unit sample than for earlier cases. However,
unlike the results for the property unit sample, lesser reductions in top charge exposure were
associated with older defendants, black defendants, and male defendants.
Most notable is that fact that none of the analyses of changes in top charge exposure found any
direct effects of the evidence initially cited by police when the cases were forwarded for screening,
though other factors that could later affect the evaluation of evidence did emerge. In general, the
findings support statements by the prosecutors who participated in the focus groups that factors
such as case seriousness, prior record, and victim characteristics, as well as factors that may affect
the evaluation of evidence, become relatively more important as a case progresses.
Change in aggregate exposure.
The number of months of incarceration a defendant could face if convicted of all of the active
charges was a function of both the number of active charges and the statutory classification of each
active charge. The resulting aggregate exposure declined between initial charging and plea offer in
all of the samples (Table 3.6.1-7). The average decline was greatest for the person unit sample (21.2
month reduction), but smaller or negligible for the other samples (2.8 month reduction for early
drug cases, 0.3 month reduction for property cases, and 5.5 month reduction for late drug cases).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 212

However there were substantial differences in the amount of reduction among individual cases in all
samples and analyses identified a number of factors associated with those differences.
Table 3.6.1-7 Average months of aggregate exposure at filing and for aggregate exposure at
plea offer

Selected Outcome Measures
Total N of cases with plea offer
% of initially screened cases
Avg. months of aggregate exposure at filing
among cases with plea offersa
Avg. months of aggregate exposure at plea offer
Avg. change in months of aggregate exposure
a

Person
Unit
253
49.8%

Property
Unit
249
37.8%

Early
Drug
Unit
1359
54.9%

Late
Drug
Unit
873
36.1%

69.1

12.4

17.1

29.4

47.9
-21.2
(223)b

12.1
-0.3
(239)

17.1
-2.8
(1331)

23.9
-5.5
(854)

Calculations for aggregate exposure at filing and change in exposure are limited to cases with plea offers.
Numbers in parentheses indicate N of cases with non-missing values for aggregate exposure at both filing and plea
offer.

Table 3.6.1-8 presents the results of the linear regression models examining changes in
aggregate months of exposure. For the person unit sample, analyses identified ten factors

associated with change in aggregate exposure (N = 153-223, R2 = .59). Factors associated with
greater reductions in aggregate exposure included the number of charges at intake, cases in which
the top charge at intake was robbery, cases in which any victims needed medical attention, the
number of victims willing to prosecute, and whether there were codefendants in the case. For
example, each additional charge filed at intake (up to a maximum of five charges) resulted in a 30
month reduction of aggregate exposure in the plea offer. Similarly, the presence of any victims
needing medical attention resulted in a 26 month reduction in aggregate exposure. For robbery
cases only, the number of black or Hispanic victims also reduced the aggregate exposure at plea
offer, resulting in a 12 month reduction for each additional black or Hispanic victim (up to 2).
Factors associated with lesser reductions or increases in aggregate exposure in person unit cases
included the total number of items of evidence cited by the police (+7 months for each increase, up
to a maximum of 5 items), the number of victims to whom the defendant was a stranger (+11
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 213

months for each increase, up to a maximum of 2 victims), prior record level (+6 months for each
increase in level, up through level 4), and, for cases with multiple charges only, the statutory class
of the top charge at screening (+5 months for each decrease in class).
Table 3.6.1-8 Final linear regression models for effects of selected factors on change in
aggregate incarceration exposure from initial filing to plea offer in Southern County
Independent Variables
Model Constant
Case Characteristics
Late drug sample
N of charges screened (up to 5)
Statutory class of top charge at screeninga
Top charge screened was robbery
Top charge screened was burglary
Codefendants

Model 10
Person
Unit
+29.712
-29.989***
na
-11.597
-10.944*

Coefficients
Model 11
Property
Unit
-66.714

Model 12
Drug
Unit
-51.492

-4.244***
+7.273***

-2.364***
-3.629***
+4.241***

+8.994***

Defendant Characteristics
Race (black)
Gender (= male for drugs; female otherwise)
Prior record level
Defendant age at screening (4 levels)

na
na
+6.062**
na

+2.617***
-2.052***

Victim Characteristics
Any victims need medical attention
N of injured victims (up to 2)
N of victims willing to prosecute
N of victims to whom defendant was a stranger

-25.697***
na
-15.108***
+11.442***

na
+10.122**
na
na

Evidence
Total N of items of evidence listed (up to 5)
N of items of evidence per intake charge (3 levels)

+7.152***
na

+1.379*
na

Interactions
Multiple charges BY class of top charge at screening +4.578***
Robbery BY N of black or Hispanic victims
-11.782**

Pairwise Ns
Model R2

153-223
.59

192-239
.38

+5.286***
+2.390**
+3.331***
+1.628***

2039
.21

*p<=.10; **p<=.05; ***p<=.01; for coefficients without asterisks p<=.15
Statutory class was coded as 1 = highest rank; larger values correspond to less serious charges. Thus, coefficients with
positive signs denote inverse relationships with seriousness.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 214

For the property unit sample, nearly all cases experienced some decline in aggregate
incarceration exposure. Starting from an average reduction of 67 months,24 analyses identified three
factors associated with relatively greater reductions in aggregate exposure and four factors
associated with relatively lesser reductions (N = 192 – 239, R2 = .38). Factors associated with
relatively greater reductions included statutory rank of the top charge at intake (-7 months for each
increase in statutory class), the number of charges at intake (-4 months for each additional charge,
up to a maximum of 5 charges), and defendant age at screening (-2 months for each step up in a 4
category grouping of ages). Factors associated with relatively lesser reductions in aggregate
exposure included cases in which the top charge was burglary (+9 months), prior record level (+3
months for each step up in level, up through level 4), the number of injured victims (+10 months for
each increase in the number of injured victims, up to a maximum of 2 victims), and the total number
of items of evidence cited by the police (1 month for each increase in the number of items, up to a
maximum of 5 items).
For drug cases, analyses identified three factors associated with relatively greater reductions in
aggregate incarceration exposure and four factors associated with relatively lesser reductions (N =
2,039, R2 = .21). Starting from an average reduction of 51 months,25 factors associated with
relatively greater reductions included statutory class of the top intake charge (-4 months for each
increase in class), number of charges at intake (-4 months for each increase in the number of
charges, up to a maximum of 5 charges), and historical time period (-2 months for cases processed
in the more recent period). Factors associated with relatively lesser reduction in aggregate exposure
included prior record level (+3 months for each increase in level, up through level 4), defendant
race (+5 months if the defendant was black), defendant age at screening (+2 months for each step up
24
25

The model constant was -66.714 months.
The model constant was -51.492.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 215

in a 4-category grouping of ages), and gender (+2 months if the defendant was male). Separate tests
of interactions with historical period (early drug sample vs. late drug sample) suggested that the
prior record effect was stronger in the later sample and that the race and gender effects were weaker
in the later sample. However, the interactions could not all be tested jointly in the same model
without introducing multicollinearity problems, so the interaction effects were excluded from the
final model reported.
Sentence recommendations
While sentencing is ultimately the responsibility of the judge, prosecutors may employ different
strategies to achieve what they believe are fair outcomes given the offense, the defendant’s prior
record, the interests of victims, the consequences to the defendant, and other circumstances
surrounding the case. They may try to limit a judge’s discretion by limiting the number of charges
to which the defendant must plead guilty, which then limits the judge’s sentencing options. If a
judge is predictable, ADAs may avoid making an explicit sentence recommendation and offer a plea
of “sentence at the discretion of the court.” However, in the majority of cases, plea offers include
explicit sentence recommendations. We examined one key aspect of such recommendations—
whether or not a plea offer included a recommendation for a period of incarceration.
A series of logistic regression analyses were conducted to identify factors most strongly
associated with the probability that a plea offer would recommend a sentence of incarceration.
Separate analyses were conducted for the person unit sample, the property unit sample, and the drug
unit sample. The resulting statistical models achieved moderately strong predictive power for all
three samples (see Table 3.6.1-9).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 216

Table 3.6.1-9 Final logistic regression models for effects of selected factors on the probability
of an incarceration recommendation in Southern County
Independent Variables
Case Characteristics
Late drug sample
N of charges screened (up to 5)
Statutory class of top charge at screeninga
Top charge screened was robbery
Top charge screened was burglary
Defendant Characteristics
Race (black)
Gender (= male for drugs; female otherwise)
Prior record level
Defendant age at screening (4 levels)
Victim Characteristics
N of victims
Any victims need medical attention
N of black or Hispanic victims (up to 2)
Any person victims
Evidence
Total N of items of evidence listed (up to 5)
N of items of evidence per intake charge (3 levels)
N of weapons-related items of evidence

Model 13
Person
Unit

Coefficients
Model 14
Property
Unit

.742*
.520***
2.300

.423***

.199**
.264
3.131***
.696**

ns
2.583***
2.013***
1.520***

3.656***
.763**

.482
.176
.422**
2.030

2.487***
3.448*

Interactions
Late drug sample BY defendant age at screening
Late drug sample BY class of top charge at
screening
Late drug sample BY race (black)
Ns
Nagelkerke R2

Model 15
Drug
Unit

.754*
1.285***
.358***
157
.47

182
.35

2056
.47

*p<=.10; **p<=.05; ***p<=.01; for coefficients without asterisks p<=.15
Statutory class was coded as 1 = highest rank; larger values correspond to less serious charges. Thus, odds ratios
greater than one denote inverse relationships with seriousness.

For the person unit sample, the analysis identified four variables associated with a higher
probability of an incarceration recommendation and five variables associated with a lower
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 217

probability (N = 157, Nagelkerke R2 = .47). Variables associated with a higher probability of an
incarceration recommendation included robbery as the top charge at intake, higher statutory class of
the top charge at intake, a greater number of items of evidence per charge, and a higher prior record
level. Variables associated with a lower probability of an incarceration recommendation included
older defendant, black defendant, female defendant, higher number of victims, and the number of
victims requiring medical attention. As previously noted, the counterintuitive effects for numbers of
victims may be related to situations in which victim witnesses may be unreliable or police are
especially prone to overcharging.
For the property unit sample, the analysis identified three variables associated with a higher
probability of an incarceration recommendation and three variables associated with a lower
probability (N = 182, Nagelkerke R2 = .35). Variables associated with a higher probability of an
incarceration recommendation included higher prior record level, any person victim, and weapons
included among the items of evidence cited by the police. Variables associated with a lower
probability of an incarceration recommendation included older defendant, more black or Hispanic
victims, and more charges at intake.
For drug cases, the analysis identified five variables associated with a higher probability of an
incarceration recommendation, but three of those were significantly less influential in the more
recent sample than in earlier cases (N = 2,056, Nagelkerke R2 = .46). In both time periods, males
and defendants with higher prior record classification were more likely than others to face a
recommendation for incarceration in the plea offer. Older defendants, black defendants, and cases in
which the statutory classification of the top charge was higher were also more likely to face
incarceration recommendations during the early sample. However, the effects of age and class were
attenuated in the more recent sample, and the effect of race was reversed, with black defendants

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 218

becoming less likely than others to face incarceration recommendations in the more recent time
period.
Concurrent vs. consecutive sentences
The observed changes in top charge incarceration exposure and observed changes in aggregate
incarceration exposure lead to different conclusions about the impact of plea bargaining on the
likely consequences for defendants. For drug cases, aggregate exposure declined slightly between
initial charging and plea offer, but average top charge exposure actually increased. For the person
unit sample, average aggregate exposure declined substantially between initial charging and plea
offer, but average top charge exposure declined only slightly.26 Whether top charge exposure or
aggregate exposure is the more relevant measure depends critically on whether judges tend to
impose concurrent or consecutive sentences in cases with convictions for multiple offenses. If they
tend mostly to impose concurrent sentences, then top charge exposure is the more relevant measure
of the potential consequences to the defendant. If they tend mostly to impose consecutive sentences,
then aggregate exposure is the more relevant measure.
The data available for the present study did not include direct information about the mode of
sentencing. However, some indirect inferences are possible by comparing the top charge exposure
measure and the aggregate exposure measure for cases actually sentenced to incarceration to the
actual sentence lengths for those cases. The results suggest different conclusions depending on
crime type. For crimes against persons in which the defendant was actually sentenced to
incarceration, even the top charge exposure measure overestimated the potential length of
incarceration for crimes against persons. However, the opposite was true for property crimes and
drug crimes; for those cases, actual incarceration sentences were much closer to the aggregate
exposure measure than to the top charge exposure measure.
26

There was little average change in either measure for the property unit sample.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 219

The aggregate exposure measure is driven largely by the number of active charges. The fact that
actual incarceration sentences were closer to the aggregate exposure measure than to the top charge
exposure measure for drug crimes and property crimes suggests that, for those crime types,
prosecutors’ offers to drop charges in exchange for guilty pleas provided a clear benefit to
defendants. On the other hand, it appears that in reducing the number charges in cases involving
crimes against persons, prosecutors gave away relatively little in terms of the ultimate consequences
to defendants.
Consistency among ADAs
A series of analyses were conducted to explore the variation among the ADAs responsible for
preparation of plea offers; these analyses examined the reduction in the number of the charges and
the reduction in the statutory class of the top charge between initial charging and plea offer.
Individual ADAs were ranked according to their “concurrent experience”—the number of cases
they handled for the relevant unit during the period covered by the study. ADAs who handled
fewer than 10 cases each were grouped into a single analytic unit, as were ADAs who handled
between 10 and 19 cases each. ADAs who handled 20 or more cases were each treated as a separate
analytic unit, ordered by the number of cases they handled. Then each ADA or ADA grouping was
treated as a separate group in a one-way analysis of variance, with either the change in number of
charges or the change in statutory rank of the top charge as the dependent variable.
For the early drug unit sample, there was less average reduction in the number of charges among
the more experienced ADAs. Both the overall differences among ADAs and the linear component
of the differences were highly statistically significant. The average change in number of charges
per ADA ranged from -.74 charges to -2.57 charges. There was also significantly less reduction in
statutory class among more experienced ADAs, with actual average increases among the five

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 220

ADAs with the most concurrent experience and average changes in rank across all ADAs that
ranged from -.33 (an increase in class) to +.52 (a decrease in class) .
For the late drug unit sample, ADAs with more than approximately 132 cases in the study
period were quite consistent with one another, but there was substantial variation among ADAs with
less concurrent experience, with average changes in the number of charges that ranged from -.79
charges to -2.31 charges. Regarding changes in statutory class of the top charge, there was no clear
association with level of experience, but there was a clear separation between a high group and a
low group, for which the low group consisted of the two most inexperienced ADA groups and two
much more experienced ADAs.
For the person unit sample, there was much greater variation among ADAs who had handled
fewer than 50 cases than among those who had handled more than 50 cases. For the less
experienced group, average change in the number of charges ranged from -1.15 to -.42, whereas the
average change ranged only from -1.14 to -.75 for the more experienced ADAs. More striking for
this sample were the highly significant differences in within-ADA variation, with standard
deviations ranging from .67 to 2.04 for ADAs who had handled fewer than 50 cases but only from
1.00 to 1.56 for the more experienced ADAs. Changes in statutory class were also highly variable
for this sample, with average changes in rank number per ADA ranging from +.08 to +1.31 (both of
which represent decreases in the reverse coded class), as well as within-ADA standard deviations
that ranged from .28 to 2.59, but with no clear pattern of association with concurrent experience.
Similar results were found for the property unit sample, with significant differences among ADAs
in both the changes in number of charges and the changes in statutory class, but with no clear
pattern of association with concurrent experience.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 221

3.6.2. Factors Influencing Plea Offers: Hypothetical Cases
Building on the factorial survey analyses from previous sections, for each hypothetical case that
prosecutors indicated they would accept for prosecution, prosecutors were asked to specify the
charge(s) and sentence that they would suggest in the plea offer. The cases included in the plea
offer analyses were the same as those included in the charging decision analyses discussed in
Section 3.5.2.
Descriptive Statistics
The charges that the prosecutors indicated would require guilty pleas in their plea offers were
coded to indicate the number of charges and the statutory class of the top plea offer charge. For each
case, the statutory class of the top charge was assigned a numerical rank, with a value of “1”
corresponding to the most serious felony class and a value of “11” corresponding to any class of
misdemeanor. Thus, in all of the analyses reported in this section, high values for the rank variable
correspond to low levels of charge seriousness, and positive indices of association reflect inverse
relationships with charge seriousness. The sentence recommendations the prosecutors offered were
coded to indicate simply whether the recommendations included a period of incarceration (0=no
incarceration; 1=incarceration).
Among cases accepted for prosecution, the average number of charges for which prosecutors
would have required guilty pleas varied from a low of .71 to a high of 2.69, with the averages for 6
of the 10 vignettes clustered between .96 and 1.28 (Figure 3.6.2-1). The variation in numbers of
filing charges per case also differed among vignettes, from a very narrow range for vignette 7
(medium offense seriousness, medium evidence strength) (range = 1; s.d. = .2) to a much wider

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 222

range for vignette 9 (medium offense seriousness, high evidence strength) (range = 7; s.d. = 1.7)
(Figure 3.6.2-2).27
Figure 3.6.2-1 Mean number of charges requiring guilty plea per plea offer, by vignette
Mean number of charges requiring guilty plea

0
1

Vignette number

However, the ratios of standard deviations to vignette means were fairly similar across a majority of the vignettes—
ranging from 33 percent of the mean to 45 percent of the mean for 7 of the 10 vignettes. The three vignettes with the
most extreme ratios of standard deviation to the mean were vignette 7 (.208/.96 => 22%) involving an arrest for
possession of cocaine, vignette 9 (1.722/2.69 => 64%) involving an arrest for forgery and theft, and vignette 2 (.488/.71
=> 69%) involving an arrest for robbery and conspiracy. The within-vignette distributions of numbers of charges
requiring guilty pleas were highly skewed for several of the vignettes, reflecting narrow ranges and unequal frequencies
across 2 or 3 values. The most highly skewed was the distribution for vignette 7, for which 96 percent of the plea offers
specified only one charge, and the remaining 4 percent recommended diversion or deferred prosecution and did not
require guilty pleas to any charges. Across all vignettes, the overall mean number of plea offer charges requiring guilty
pleas was 1.56, with a range from 0 to 8, a standard deviation of 1.031, and a skewness index of 2.320.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 223

Figure 3.6.2-2 Range of number of charges requiring guilty plea per plea offer, by vignette,
across prosecutors
Number of charges requiring guilty plea per
plea offer

9
8
7

6
5
4
3
2
1

0
1

Vignette

The average statutory ranks of top plea offer charges varied across vignettes, and were fairly
evenly distributed across the range from a mean of 7.36 for vignette 4 (high offense seriousness,
medium evidence strength) to a mean of 11.00 (all misdemeanors) for vignette 8 (low offense
seriousness, low evidence strength) (Figure 3.6.2-3). Variation in statutory rank of top plea offer
charge within vignette was greatest for vignette 4 (range = 7; s.d. = 2.58) and least for vignette 8
(range = 0; s.d. = 0, all misdemeanors) (Figure 3.6.2-4).28

The ratios of standard deviations to vignette means were fairly similar for most of the vignettes, ranging from 0
percent of the mean to 10 percent of the mean for 8 of the 10 vignettes. The two vignettes with the most extreme ratios
were vignette 1 (1.794/8.23 => 22%) involving burglary and assault and vignette 4 (2.576/7.36 => 35%) involving
aggravated assault with a pistol. The distributions of statutory rank within vignettes were highly skewed for several of
the vignettes, with the most extreme skew (skewness = -4.27) observed for vignette 9. Across all vignettes, the overall
mean statutory rank of top plea offer charge was 9.27, with a range from 4 to 11, a standard deviation of 1.731 and a
skewness index of 1.731.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 224

Mean statutory rank for top plea offer charge

Figure 3.6.2-3 Mean statutory rank of top plea offer charge, by vignette
12
10
8
6
4
2
0
1

Vignette number

Figure 3.6.2-4 Range of statutory rank of top plea offer charge, by vignette, across
prosecutors
Statutory class ranking code for top charge
requiring guilty plea

12
10
8
6
4
2
0
1

Vignette number

The wide ranges for vignettes 1 (high offense seriousness, low evidence strength), 4 (high
offense seriousness, medium evidence strength), and 9 (medium offense seriousness, high evidence
strength) were due to the fact that some prosecutors indicated they would require guilty pleas to a
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 225

habitual felon charge (statutory rank = 4), based on their interpretation of the prior record profiles
with which they were presented. Prior record profiles were paired randomly with vignettes; for a
given vignette, different respondents were presented with different prior record profiles. Therefore,
a valid analysis of variation in responding requires controls for seriousness of prior record, which
are incorporated later in this section.
Among hypothetical cases that respondents indicated they would accept for prosecution, the
proportion for which they indicated they would recommend a period of incarceration varied
substantially across vignettes, from a low of 0 percent for vignette 8 (low offense seriousness, low
evidence strength) to a high of 76 percent for vignette 4 (high offense seriousness, medium
evidence strength) (Figure 3.6.2-5). For six of the ten vignettes, the proportions of responses
recommending incarceration were in the range from 20 percent to 50 percent, with two falling
below that range and two falling above that range.29 The overall proportion of cases with
incarceration recommendations, across all vignettes, was 38 percent.

Since incarceration recommendation was measured as a dichotomy, the variance is a fixed function of the proportion,
p(1-p), and is not analyzed separately here.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 226

Figure 3.6.2-5 Proportion of accepted hypothetical cases for which plea offer would
recommend a period of incarceration

Proportion of plea offers with incarceration
recommended

100%

80%

60%

40%

20%

0%
1

Vignette number

The case-level independent variables tested in statistical models for number of plea offer
charges, statutory rank of plea offer charges, and incarceration recommendation were the same as
those previously introduced in Sections 3.4.3 and 3.5.2 for modeling screening and charging
decisions. The primary variables included researchers’ design classification for offense seriousness
(based on seriousness of the top arrest charge), prosecutors’ ratings of strength of evidence, and
prosecutors’ ratings of the seriousness of defendants’ criminal histories. In addition, the analyses of
plea offers presented in this section were based on the same subset of cases used in the analyses of
charging decisions—namely, cases that prosecutors indicated they would accept for prosecution at
screening (Refer to Section A.3.4 in Appendix A for explanations and descriptive analyses of the
case-level independent variables in cases accepted for prosecution).
The prosecutor-level independent variables tested in statistical models for number of plea offer
charges, statutory rank of plea offer charges, and incarceration recommendation were also the same

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 227

as those previously introduced in Sections 3.4.3 and 3.5.2. They included prosecutors’ years of
experience in the Southeast County prosecutor’s office, three short scales constructed from
responses to items on the general survey, and responses to several other individual items on the
general survey. (Refer to Section A.3.4 in Appendix A for explanations and descriptive analyses of
the prosecutor-level independent variables).
HLM Analyses
Three different aspects of plea offers were modeled: the number of charges requiring guilty
pleas, the statutory class of the top plea offer charge requiring a guilty plea, and whether or not a
period of incarceration was recommended.
The number of charges requiring guilty pleas. Table 3.6.2-1 summarizes the results of the final
two-level model for the number of plea offer charges requiring guilty pleas.
Table 3.6.2-1 Hierarchical regression model for the number of plea offer charges requiring
guilty plea
Effect
Case level
Intercept
Offense seriousness
Evidence rating
Criminal history rating
Seriousness by Evidence

Coefficient

T Ratio

P value

1.54
-0.15
-0.10
-0.01
0.15

30.42
-1.12
-1.09
-0.19
3.79

59
59
59
59
58

.000
.270
.281
.847
.001

0.02

2.37

.021

Prosecutor level
(none)
Prosecutor by case level
Experience by Seriousness
by Evidence

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 228

The main effects for offense seriousness and evidence rating were non-significant, but they
participated in a highly significant interaction effect. The combined result of the two main effects
and their interaction was that the expected number of charges requiring guilty pleas increased across
increasing levels of evidentiary strength, and that the amount of the increase was greater for more
serious offenses. These modeled expectations for the interaction between case seriousness and
strength of evidence were generally consistent with the observed patterns in the raw data graphed in
Figure 3.6.2-6. The model shows no statistically significant effect of criminal history, a result that
is consistent with the raw data patterns depicted in Figure 3.6.2-7. The interactions between
criminal history rating and offense seriousness and between criminal history rating and evidence
rating were also both non-significant and were removed from the model.

Mean N of charges requiring guilty plea

Figure 3.6.2-6 Mean number of plea offer charges requiring guilty plea by offense seriousness
and prosecutor’s rating of strength of evidence (among cases accepted for prosecution)
5
4
3
2
1
0
Lowest

Mid-Low

Average

Mid-High

Highest

Strength of evidence (prosecutor's rating)
Misdemeanor

Low level felony

High level felony

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 229

Mean N of charges requiring guilty please

Figure 3.6.2-7 Mean number of plea offer charges requiring guilty plea by offense seriousness
and prosecutor’s rating of defendant’s criminal history (among cases accepted for
prosecution)
5
4
3
2
1
0
Lowest

Mid-Low

Average

Mid-High

Highest

Defendant criminal history (prosecutor's rating)
Misdemeanor

Low level felony

High level felony

None of the prosecutor-level measures had significant effects on the average case-level intercept
or main effect coefficients. However, years of prosecutors’ experience in the Southern County
prosecutor’s office did affect the strength of the case-level interaction between offense seriousness
and evidence rating. The more experienced the respondent, the stronger the interaction—that is, the
more rapidly the number of plea offer charges for more serious offenses diverged from the number
of plea offer charges for less serious offenses across increasing levels of evidentiary strength
(Compare Figure 3.6.2-8 with Figure 3.6.2-9). The cumulative effect could be substantial, since the
number of years of experience for the respondents ranged from 0 years (less than 6 months) to 40
years, with a majority falling in the range from 0 to 10 years.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 230

Mean N of charges requiring guilty plea

Figure 3.6.2-8 Mean number of plea offer charges requiring guilty plea by offense seriousness
and prosecutor’s rating of strength of evidence (for prosecutors with 2 to 5 years’ experience
in the Southern County DA’s office)
5
4
3
2
1
0
Lowest

Mid-Low

Average

Mid-High

Highest

Strength of evidence (prosecutor's rating)
Misdemeanor

Low level felony

High level felony

Mean N of charges requiring guilty plea

Figure 3.6.2-9 Mean number of plea offer charges requiring guilty plea by offense seriousness
and prosecutor’s rating of strength of evidence (for prosecutors with more than 5 years’
experience in the Southern County DA’s office)
5
4
3
2
1
0
Lowest

Mid-Low

Average

Mid-High

Highest

Strength of evidence prosecutor's rating
Misdemeanor

Low level felony

High level felony

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 231

In the focus group discussions, prosecutors indicated that it is often important to charge enough
offenses to cover the essential facts of a case, especially if a case is likely to go to trial. More
serious cases may often have more complicated fact patterns or be more dependent on forensic
evidence and therefore be more sensitive to the strength of the evidence—which could account for
the case-level interaction between offense seriousness and evidence rating in the model for number
of filing charges (see Section 3.5.2). The model presented in this section suggests that the effect
carried over to the number of charges requiring guilty pleas in the plea offers for these hypothetical
cases, and that the effect was somewhat stronger among more experienced prosecutors. The raw
data patterns depicted in Figures 3.6.2-8 and 3.6.2-9 suggest that the effect of experience was to
decrease the number of charges requiring guilty pleas in less serious cases with strong evidence
(rather that to increase it for more serious cases). This may suggest that experienced prosecutors are
more likely than less experienced prosecutors to distinguish between a need for retaining complete
charging in more serious (more complex) cases and a lesser need for complete charging in less
serious (less complex) cases.
The descriptive analyses of the number of charges for which the plea offer would require guilty
pleas found considerable variation across prosecutors for several of the vignettes (Figure 3.6.2-2).
The focus on individual vignettes effectively controlled for case seriousness and strength of
evidence. However, caution is advised due to the fact that different respondents were presented
with different pairings of vignettes with criminal history profiles. Thus, a valid analysis of variation
in responding among prosecutors requires controlling simultaneously for all of the case level
factors, including criminal history, and the interactions among case-level factors.
The HLM analyses provided estimates of the residual variation among prosecutors, controlling
for the case level factors. Variability in the average number of plea offer charges was assessed by

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 232

examining the variance of case-level intercepts across prosecutors, from an intermediate model (not
shown) which included the three case-level measures and the case-level seriousness-by-evidence
interaction, but no prosecutor-level predictors. The analysis found significant variation among
prosecutors in the case-level intercept (χ2=59.07, df=42, p = .042). This significant variation among
prosecutors in the average number of plea offer charges was not accounted for by any of the
prosecutor-level variables tested in development of the final model and remains unexplained.
The HLM analyses also examined variability among prosecutors in the influence of offense
seriousness, evidence rating, and criminal history rating on the number of charges for which the
plea offer would require guilty pleas. The HLM 6 estimates for the corresponding random effects
showed no statistically significant deviations from the average offense seriousness coefficient (χ2=
24.61, df=42, p > .500), the average evidence rating coefficient (χ2=26.78, df=42, p > .500), the
average criminal history rating coefficient (χ2= 48.59, df=42, p = .224), or the average coefficient
for seriousness-by-evidence interaction (χ2=24.73, df=42, p > .500). That is, there were no
significant differences among prosecutors in the weight they assigned to any of the case-level
factors included in these analyses.
The statutory class of the top plea offer charge requiring a guilty plea. Table 3.6.2-2
summarizes the results of the final two-level model for the seriousness of top plea offer charge.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 233

Table 3.6.2-2 Hierarchical regression model for statutory rank of top plea offer charge
Effect
Coefficient T Ratio
df
P value
Case level
Intercept
Offense seriousness
Evidence rating
Criminal history rating
Seriousness by evidence
Seriousness by history

9.27
-0.28
0.52
-0.31
-0.41
0.16

144.75
-0.68
2.88
-2.85
-3.84
2.53

59
59
59
58
59
58

.000
.498
.006
.006
.000
.015

-0.03

-2.08

.042

Prosecutor level
(none)
Prosecutor by case level
Item Q8a by seriousness
by history

Note: The dependent variable, rank of top plea offer charge, is scaled with smaller numerical values corresponding to
more serious charges. Consequently, effects with positive signs in the above table reflect inverse relationships
with the seriousness of top plea offer charge, and vice versa.

As Table 3.6.2-2 shows, there were significant effects for evidence rating and the interaction
between evidence rating and offense seriousness. Although the main effect of offense seriousness
was non-significant, it was retained in the model because it participates in the significant evidenceby-seriousness interaction. Although the main effect for evidence rating was in the counterintuitive
direction, the combined result of evidence rating, offense seriousness, and their interaction was that
cases with more serious top arrest charges were associated with more serious (lower numbered
rank) top plea offer charges, and the differences in top plea offer charges among levels of offense
seriousness increased across increasing levels of evidentiary strength.
Figure 3.6.2-10 depicts the patterns of relationships among offense seriousness, strength of
evidence, and seriousness of the top plea offer charge, as reflected in the raw data. The case-level
patterns identified in the statistical model are mostly consistent with the raw data, although the

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 234

model calculations imply a slightly positive association between top charge seriousness and strength
of evidence for low-level felonies that is not evident in the graph of the raw data.

Mean statutory rank for top charge requiring
guilty plea

Figure 3.6.2-10 Mean statutory rank of top plea offer charge per hypothetical case by offense
seriousness and prosecutor’s rating of strength of evidence (among cases accepted for
prosecution)
12
10
8
6
4
2
0
Lowest

Mid-Low

Average

Mid-High

Highest

Strength of evidence (prosecutor's rating)
Misdemeanor

Low level felony

High level felony

The model also suggests a significant influence of prosecutors’ ratings of the seriousness of
defendants’ criminal histories on the seriousness of top plea offer charge, as well as a significant
interaction between offense seriousness and criminal history rating. Calculations based on the
model coefficients suggest a combined effect of offense seriousness, criminal history rating, and
their interaction in which seriousness of top plea offer charge increases with increasing seriousness
of criminal history to a greater extent for misdemeanors than for low-level or high-level felonies.
However, this is not very consistent with the overall pattern in the raw data depicted in Figure 3.6.211, which suggests a greater effect of criminal history for low-level felonies. This discrepancy
could be due to the fact that the statistical model controls for a significant effect of a prosecutorlevel influence on the interaction in questions, whereas the raw data patterns naturally do not.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 235

There was also a discontinuity in the measure of top plea offer charge seriousness which could lead
to counter-intuitive results. When prior criminal history warrants it, prosecutors may include a
habitual felon charge in the plea offer. Sometimes that was the highest charge requiring a guilty
plea. The habitual felon charge is at a fixed statutory class that does not depend on the
classification of the underlying offense. Thus, habitual offender charges could be either more
serious or less serious than the underlying offenses, and could alter the relationship between
seriousness of the top arrest charge (offense seriousness in these analyses) and the seriousness of the
top plea offer charge.

Mean statutory rank for top charge requiring
guilty plea

Figure 3.6.2-11 Mean statutory rank of top plea offer charge per hypothetical case by offense
seriousness and prosecutor’s rating of criminal history (among cases accepted for
prosecution)
12
10
8
6
4
2
0
Lowest

Mid-Low

Average

Mid-High

Highest

Defendant criminal history (prosecutor's rating)
Misdemeanor

Low level felony

High level felony

There was a significant effect of prosecutors’ responses to general survey item Q8a on the caselevel interaction between seriousness of the top arrest charge and prosecutors’ ratings of defendant
criminal history. Item Q8a asked prosecutors how much they agreed or disagreed that “a plea offer
should include all of the charges filed, with an offer to forego additional charges if the offer is

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 236

accepted.” The more strongly prosecutors agreed with that item, the weaker the case-level
interaction between offense seriousness and criminal history rating. Thus, for prosecutors who
disagreed with the all-charges orientation, offense seriousness mattered more for defendants with
low criminal history ratings than for defendants with high criminal history ratings (Figure 3.6.2-12).
However, for prosecutors who agreed with the all-charges orientation, there was no consistent
relationship between criminal history rating and the effect of offense seriousness on seriousness of
the top plea offer charge (Figure 3.6.2-13). As suggested in an earlier section, it may be that a
strong orientation to pursuing all charges may render other considerations less relevant.

Mean statutory class rank code for top charge
requiring guilty plea

Figure 3.6.2-12 Mean statutory rank of top plea offer charge by offense seriousness and
prosecutor’s rating of criminal history (for prosecutors who disagreed with General Survey
Item Q8a)
12
10
8
6
4
2
0
Lowest

Mid-Low

Average

Mid-High

Highest

Defendant criminal history (prosecutor's rating)
Misdemeanor

Low level felony

High level felony

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 237

Mean statutory rank for top charge requiring
guilty plea

Figure 3.6.2-13 Mean statutory rank of top plea offer charge by offense seriousness and
prosecutor’s rating of criminal history (for prosecutors who agreed with General Survey Item
Q8a)
12

10
8

6
4
2
0
Lowest

Mid-Low

Average

Mid-High

Highest

Defendant criminal history (prosecutor's rating)
Misdemeanor

Low level felony

High level felony

The descriptive analyses of the statutory rank of the top plea offer charge found considerable
variation across prosecutors for several of the vignettes (see Figure 3.6.2-4). However, caution is
advised due to the fact that different respondents were presented with different pairings of vignettes
with criminal history profiles. A valid analysis of variation in responding among prosecutors
requires controlling simultaneously for all of the case level factors. The HLM analyses produced
estimates of the residual variation among prosecutors, controlling for the case level factors.
Variability in the average statutory rank of top plea offer charges was assessed by examining the
variance of case-level intercepts across prosecutors from an intermediate model (not shown), which
included the three case-level measures and the case-level seriousness-by-evidence and seriousnessby-history interactions, but no prosecutor-level predictors.
The HLM 6 estimates for the random variation of case-level regression parameters across
prosecutors found statistically significant deviations from the average coefficients for all of the

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 238

case-level effects included in the final model: case-level intercept (χ2 = , df = 45, p = .015), offense
seriousness (χ2 = 71.87, df = 35, p = .000), strength of evidence (χ2 = 49.25, df = 35, p = .055),
criminal history rating (χ2 = 50.54, df = 35, p = .043), the interaction between offense seriousness
and strength of evidence (χ2 = 59.97, df = 35, p = .006), and the interaction between offense
seriousness and criminal history rating (χ2 = 56.47, df = 35, p = .012).30 The significant residual
variation for the case-level intercepts means that the observed differences among prosecutors in the
average statutory class rank of their top plea offer charges were not accounted for by the case-level
factors included in the analysis. The significant residual variation among prosecutors relative to the
average case-level regression weights also means that different prosecutors tended to weigh case
characteristics differently in making plea offer decisions for this set of hypothetical cases.
The residual variation in model intercepts was reduced to a non-significant level when
prosecutors’ orientation toward pleas for all charges was incorporated in the model. The significant
variation among prosecutors in how they weigh offense seriousness, strength of evidence, and
criminal history in their plea offer decisions remain unexplained by these analyses. Both the fact
that some of the variation in plea offer decisions was explained by differences among prosecutors in
their choice of prosecutorial strategies and the fact that there remains unexplained random variation
in decision making criteria for these hypothetical cases are cause to question how consistent the
decision making might be among prosecutors deciding real cases.
The probability of an incarceration recommendation. Table 3.6.2-3 summarizes the results of
the final two-level model for probability that the plea offer would recommend a period of
incarceration.

There was sufficient information to calculate the chi-square statistics for only 36 of the 60 respondents included in the
analysis. However, the finding of significant unexplained variance for all of the random parameters in the model for
those 36 is sufficient to raise questions about possible lack of consistency in decision criteria across prosecutors.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 239

Table 3.6.2-3 Hierarchical logistic regression model for the probability of an incarceration
recommendation
Odds
Effect
Coefficient
T Ratio
df
P value
Ratio
Case level
Intercept
Offense seriousness
Evidence rating
Criminal history rating

-0.68
1.16
0.49
0.67

0.51
3.19
1.64
1.96

-5.15
10.26
4.80
10.56

60
59
59
57

.000
.000
.000
.000

0.20
-0.17
0.17
0.26
0.28

1.22
0.83
1.19
1.30
1.33

2.51
-2.54
2.14
4.45
1.77

59
59
57
57
57

.015
.014
.037
.000
.082

Prosecutor level
(none)
Prosecutor by case level
Item Q5a by seriousness
Fairness by evidence
Item Q8a by history
Item Q8d by history
Top charges by history

As Table 3.6.2-3 shows, there were highly significant positive effects for all three of the caselevel independent variables. Each step up in the 3-level classification of seriousness of the top
arrest charge was associated with a 3-fold increase in the odds of an incarceration recommendation.
The odds also nearly doubled for each increase in criminal history rating and increased by nearly
two-thirds for each increase in evidence rating. Large effects of offense seriousness and criminal
history rating might be expected, since sentencing options in Southern State are tightly constrained
by statute on the basis of conviction offense and conviction history. However, in these analyses, the
variable labeled offense seriousness is a broad classification based on top arrest charge. Prosecutors
exercise considerable discretion in deciding what charges to file and what charges require guilty
pleas, so the seriousness of top arrest charge does not necessarily determine the seriousness of either

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 240

the top plea offer charge or the top conviction charge. Thus, ultimate case outcome given offense
seriousness as alleged by an arresting officer constitutes one kind of summary of the exercise of
discretion across the stages of prosecution, and it is interesting that the statutory level of top arrest
charge so strongly influences plea offers and sentence recommendations despite the intervening
exercise of prosecutorial discretion.
What is perhaps more surprising is the significant effect of strength of evidence on plea offers
and the associated sentence recommendations. Prosecutors who participated in the focus group
sessions asserted the opposite—that strong cases do not necessarily result in more severe charges;
and if two cases involve the same underlying conduct, one will not get a higher charge simply
because it has stronger evidence. However, the outcome still could reflect the cumulative effects of
charging decisions at initial filing and during the preparation of plea offers relating to constraints
imposed by deficiencies in the evidence—or the lack of such constraints.
Graphs of the raw data patterns for offense seriousness by strength of evidence (Figure 3.6.2-14)
and offense seriousness by defendant’s criminal history (Figure 3.6.2-15) suggest possible
interaction effects. Figure 3.6.2-14 suggests that the effect of strength of evidence might be
stronger for high-level felonies than for low-level felonies or misdemeanors. Figure 3.6.2-15
suggests that the effect of criminal history rating might be stronger for felonies than for
misdemeanors and might operate only at the highest level of criminal history for misdemeanors.
However, both interactions were tested in the statistical modeling, and neither was statistically
significant given the size of the sample available for analysis.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 241

Figure 3.6.2-14 Proportion of accepted hypothetical cases with incarceration
recommendations by offense seriousness and prosecutor’s rating of strength of evidence
100%
Percentage of accepted cases with
incarceration recommendations

90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Lowest

Mid-Low

Average

Mid-High

Highest

Strength of evidence (prosecutor's rating)
Misdemeanor

Low level felony

High level felony

Figure 3.6.2-15 Proportion of accepted hypothetical cases with incarceration
recommendations by offense seriousness and prosecutor’s rating of defendant’s criminal
history
Percentage of accepted cases with
incarceration recommendation

100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Lowest

Mid-Low

Average

Mid-High

Highest

Defendant criminal history (prosecutor's rating)
Misdemeanor

Low level felony

High level felony

Five prosecutor-level measures had statistically significant effects on case-level regression
coefficients. The more importance prosecutors attached to the potential for plea bargaining, early
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 242

disposition, and caseload reduction (general survey item Q5a), the more strongly their decisions
were influenced by offense seriousness. The more importance prosecutors attached to fair treatment
of defendants (fairness scale), the less strongly their decisions were influenced by strength of
evidence.
Three measures of prosecutors’ preferences for particular prosecutorial strategies affected the
model estimates of the influence of prosecutors’ ratings of defendants’ criminal histories. The
positive effect of criminal history rating on the estimated odds of an incarceration recommendation
increased further as a function of increasing agreement that a plea offer should include all of the
charges filed (general survey item Q8a), that the charging decision should include the highest
charges that could be proved at trial (item Q8d), and that the respondent prefers to file only the most
serious charges possible (item Q7, response option 2). Each of these items reflects a strategy
oriented toward obtaining severe outcomes. In deciding whether or not to recommend incarceration
in the plea offer, prosecutors who indicated agreement with that orientation gave greater weight to
the criminal history rating than those who disagreed.
The HLM 6 estimates for the variation in case-level model parameters across prosecutors found
no statistically significant deviations from the average offense seriousness coefficient (χ2 = 52.52, df
= 56, p > .500), the average evidence rating coefficient (χ2 = 61.59, df = 56, p = .283), or the
average criminal history rating coefficient (χ2 = 54.32, df = 56, p > .500 ). However, the residual
variation in regression intercepts across prosecutors was highly statistically significant (χ2 = 92.76,
df = 56, p = .002). None of this residual variation could be reliably attributed to any of the
prosecutor-level measures tested in these analyses. The significant variation among prosecutors in
their average rates of recommending incarceration remains unexplained by these analyses.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 243

3.7 Dismissals
There are many reasons a case or selected charges within a case might be dismissed after initial
filing. Throughout the processing of a case, prosecutors must continually re-evaluate the core
questions “Can I prove the case?” and “Should I prove the case?” Whole cases or selected charges
may be dismissed because prosecutors become less certain of their ability to prove the charges:
because information about the victim, the defendant, or the circumstances surrounding the case
causes prosecutors to reconsider what would be a fair outcome; because prosecutors plan from the
outset to dismiss some charges as part of their prosecution strategy; or simply because resource
constraints force prosecutors to dismiss lower priority cases or craft more attractive plea offers.
The most direct reason for dismissals relates to the availability and quality of evidence.
Evidence thought to be available and credible at the outset may later be unavailable or found to be
flawed in some way. Prosecutors noted that the minimal amount of time available for screening
meant that many cases may be dismissed immediately after screening because the assigned ADA
looks closer. Prosecutors also indicated that changes in law enforcement in both jurisdictions have
led to an increase in cases declined for prosecution, pended for additional information, or dismissed
due to poor follow-up investigation. New information may cast doubt on the credibility of witness
testimony, and victims and other potential witnesses may decide not to cooperate. For a variety of
such reasons, evidence may be weaker than initially thought or may weaken as time passes; rarely,
according to prosecutors, does the evidence prove to be stronger than initially thought or become
strengthened with the passage of time.
Other case characteristics can affect the probability of dismissal either directly or indirectly
through their relationship to the strength of evidence. A salient example is the influence of case
seriousness. On the one hand, cases involving more charges and more serious charges leave room

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 244

for making attractive plea offers if necessary to avoid dismissing the entire case, and we noted in
the previous section that such cases do indeed experience greater reductions than less serious, less
complicated cases. In addition, more serious cases may be considered higher priority than other
cases, perhaps making it less likely that the entire case would be dismissed. On the other hand, such
cases may have complicated evidentiary requirements and be more vulnerable than less serious
cases due to lack of expected forensic evidence, perceived flaws in the evidence, and poor
credibility or declining interest of victims and other witnesses.
In the face of resource constraints such as limited availability of court time, the likelihood of
dismissing an entire case or dismissing selected charges to make a plea offer more attractive can
also be affected by the priority of the case, relative to the priority of other pending cases. Relative
priority, in turn, might be influenced by a variety of factors: the (updated) strength of the evidence;
the seriousness of the offense; the prior record and perceived dangerousness of the defendant;
whether the defendant is in custody; whether the prosecutor considers time already spent in custody
as sufficient consequence given the seriousness of the offense; the wishes of the victim; the age of
the case; and so on.
In the previous section, we examined case-level and prosecutor-level factors that influence
reductions between initial filing and presentation of a plea offer in the number of charges and the
seriousness of the top charge. However, we only had plea offer information for Southern County. In
this section, we examine case characteristics associated with the probability of pre-adjudication
dismissal of all charges in Northern County and Southern County, and then examine the reasons
cited by prosecutors for dismissal of individual charges in Southern County.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 245

3.7.1 Dismissal Decisions in Southern County
For Southern County, we were unable to distinguish between cases dismissed pre-adjudication
in trial court and cases disposed by acquittal at trial. However, since only a small fraction of cases
were disposed in trial court, and only some fraction of those were disposed by acquittal at trial, we
adopted non-conviction as a proxy for dismissal.
Descriptive Statistics
Our analyses of non-conviction rates were based on 372 Person Unit cases, 428 Property Unit
cases, 2,105 Early Drug Sample cases, and 1,854 Late Drug Sample cases filed following
acceptance at initial screening. Figure 3.7.1-1 displays the acceptance rates at screening and the
non-conviction rates among accepted cases for each of the four samples. The non-conviction rates
for Southern County were 33.1 percent for felony person crimes, 38.6 percent for felony property
crimes, 37.7 percent for the early felony drug crime sample, and 48.4 percent for the late felony
drug crime sample. These compare to dismissal rates of approximately 20 percent for felony drug
crimes and less than 15 percent for felony person and felony property crimes in Northern County
(see below).31
The bivariate relationships between non-conviction rates for filed cases and selected case
characteristics are displayed in Table 3.7.1-1. The case characteristics presented in Table 3.7.1-1 are
limited to ones that proved to be significant factors in one or more of the final statistical models.

These differences are much larger than could be attributable to the substitution of non-conviction as a proxy for
dismissal.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 246

Figure 3.7.1-1 Acceptance rates and non-conviction rates among accepted cases by prosecutor
unit in Southern County
100%
90%

Percentage of cases

80%
70%
60%

50%
40%
30%
20%
10%
0%
Person Team

Property Team

% Accepted

Early Drug Sample

Late Drug Sample

% NC Among Accepted

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 247

Table 3.7.1-1 Percentage of filed cases with non-conviction dispositions in Southern County,
by selected case characteristics
Case Characteristics
N of charges presented for screening

Rank of top charge presented for screening

Codefendants
Black defendant
Male defendant
Prior record level

N of items of evidence per charge
Total N of items of evidence listed

N of victims
N of victims willing to prosecute
N of victims of strangers
N of black or Hispanic victims
Any victims need medical attention?
*N less than 10

1
2
3
4
5
4
5
6
7
8
9
10
N
Y
N
Y
N
Y
1
2
3
4
5
6
<1
1-2
2+
0
1
2
3
4
5
0
1
2+
0
1
2+
0
1
2+
0
1
2+
N
Y

Person
Unit
45.0
33.8
23.9
24.4
22.6
19.0
34.7
43.8
12.5
28.1
26.7
40.7
33.9
32.2
16.0
35.7
41.7
32.1
26.7
40.3
16.7
28.6
35.7
26.5
33.3
53.0
27.4
26.9
22.0
20.0
25.0
35.4
28.6
63.0
29.8
26.7
44.0
29.8
26.1
28.2
39.1
31.9
35.7
23.1

Property
Unit
41.9
37.7
39.5
39.3
19.0
*
*
*
*
*
39.0
36.0
39.6
35.2
43.5
36.2
35.3
39.0
28.9
29.2
18.2
32.1
39.6
36.0
39.6
47.9
33.7
39.8
36.6
40.0
33.3
38.9
35.6
75.9
32.6
33.3
46.3
35.6
38.7
32.9
50.4
57.1
-

Early Drug
Sample
48.5
39.4
34.5
31.4
23.0
15.8
8.2
30.7
*
14.8
37.2
45.3
36.8
40.9
45.2
34.5
39.5
37.4
25.6
22.9
20.7
21.5
19.0
15.4

Late Drug
Sample
60.7
53.4
41.1
39.6
37.4
*
16.7
22.0
*
18.5
46.2
61.8
51.0
37.7
58.2
46.0
55.6
47.1
22.0
23.2
13.6
10.5
35.5
29.0

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 248

As Table 3.7.1-1 indicates, the number of charges initially forwarded to the prosecutor for
screening exhibited the most consistent relationships with the probability of non-conviction. In all
four samples, the greater the number of charges screened (regardless of the number filed, which
tended to be similar), the lower the ultimate non-conviction rate. The magnitude of the observed
differences was minimal for property crimes but substantial for person crimes and drug crimes
(reductions in the order of 20 percentage points across increasing numbers of charges). This is
consistent with the finding for Northern County that cases with multiple charges filed were less
likely to be dismissed than cases in which only a single charge was filed (see below).
The observed relationship of non-conviction rates with race was inconsistent. For person crimes,
non-conviction rates were twice as high for black defendants as for non-black defendants. In
contrast, for property and drug crimes, non-conviction rates were approximately 10 percentage
points lower for black defendants. Table 3.7.1-1 also shows that cases involving black or Hispanic
victims had higher non-conviction rates than others—for both person and property crimes, but
especially for property crimes.
Other offense-related factors and defendant characteristics had less consistent relationships with
non-conviction rates across crime types. The presence of codefendants was associated with lower
non-conviction rates for person crimes, property crimes, and the late drug crime sample, but not for
the earlier drug crime sample. Cases involving male defendants had lower non-conviction rates for
person crimes and both drug crime samples, but not for property crimes. Prior record level was not
consistently related to non-conviction rates for person or property crimes. For drug crimes,
increasing prior record level was generally associated with decreasing non-conviction rates. The
two highest prior record levels did not fit this pattern for the late drug crime sample, but the
apparent deviations may have been due to small numbers of cases at the two highest levels. The

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 249

statutory class of the top charge presented at screening showed no consistent pattern of relationships
with non-conviction rates for person or drug crimes, and, for property crimes, cases were spread too
thinly across higher classes to support reliable estimates for individual classes.
Information about available evidence and victim characteristics was only available for person
and property cases. For person crimes, the greater the number of items of physical evidence, the
lower the non-conviction rate. For property crimes, the relationship was somewhat less consistent,
where the main distinction seemed to be between having some physical evidence and none. Among
victim characteristics, more victims, more victims willing to prosecute, more victims to whom the
defendant was a stranger, and any victims needing medical attention were all associated with lower
non-conviction rates.
Logistic Regression Analyses
Separate logistic regression models for the probability of non-conviction were estimated for
person unit cases, property unit cases, and drug unit cases. The early drug sample and the late drug
sample were combined in a single analysis to permit explicit tests of the effect of historical period
and the interactions between historical period and other factors. Relatively small samples were
available for these analyses, limiting the number of parameters that could be estimated reliably.32 In
addition, the set of potential predictors included subsets consisting of alternative measures of the
same or similar constructs, thus presenting a high risk of multicollinearity problems. Consequently,
model development proceeded in a stepwise fashion.33 First, only potential main effects were

A larger set of potentially predictive factors and their interactions was tested for possible inclusion in the models.
Additional information about the variables tested in the logistic regression analyses is given in Appendix A.
33
It is well known that stepwise selection and elimination procedures are sensitive to small changes in the sample data
when candidate variables are substantially intercorrelated; slight differences between samples can lead to differences in
the order of selection among correlated variables. In this study that means, for example, that which among the set of
measures of victim characteristics or which among the set of evidence measures is included in a final model might have
been different as a result of sampling variation. Thus, for example, it is safest to interpret a significant coefficient for a
given victim characteristic as an indication that something about victim characteristics or circumstances influenced the

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 250

entered and subjected to backward elimination processing. At each step, model parameters were
examined for signs of problems in model fitting. If problems were detected, variables likely to be
causing the problems were removed and the analysis was repeated. Once reliable results were
achieved, variables with significance values greater than p = .15 were discarded.34 Interactions
between robbery and other factors (for the person unit sample), burglary and other factors (for the
property unit sample), and historical period and other factors (for the combined drug unit sample)
were tested in forward stepwise fashion.
The three final logistic regression models for predicting non-conviction from selected case
characteristics, defendant characteristics, victim characteristics, and two measures of physical
evidence are summarized in Table 3.7.1-2. The entries in the table are odds ratios, indicating the
increase or decrease in the odds of non-conviction.

probability of non-conviction, but not to place too much emphasis on the particular measure that proved most
significant in a given sample.
34
This liberal inclusion criterion follows a suggestion by Hosmer and Lemeshow (2000, p. 118), who cite evidence that
more stringent criteria often result in misspecification (excluding variables from the model that are necessary to avoid
spurious effect estimates for other included variables).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 251

Table 3.7.1-2 Logistic regression models for predicting non-conviction given filing in Southern
County
Independent Variables

Model 16
Person Unit

Case Characteristics
N of charges presented for screening
Rank of top charge presented for screeninga
Codefendants

Odds Ratios
Model 17
Property Unit

Model 18
Drug Unit

ns
ns
ns

0.78
1.38
0.55**

0.82***
ns
ns

Defendant Characteristics
Black defendant
Male defendant
Prior record level

5.15**
ns
ns

ns
ns
ns

ns
ns
0.93*

Victim Characteristics
N of victims
N of victims willing to prosecute
N of victims of strangers
N of black or Hispanic victims
Any victims need medical attention

3.29**
0.40***
0.59*
ns
0.44*

2.98
0.17***
ns
2.42***
ns

Evidence
N of items of evidence per charge
Total N of items of evidence listed

1.74*
0.70**

ns
ns

220
.19

248
.16

N
Nagelkerke R2

2036
.02

*p<=.10; **p<=.05; ***p<=.10; an odds ratio displayed without an asterisk indicates p<=.15
a
Highest statutory class rank = 1; larger values correspond to less serious offenses; thus, odds ratios greater than 1
indicate inverse effects.

For both person unit cases and property unit cases, the odds of non-conviction were strongly
associated with various combinations of victim characteristics. Whereas there were significant
effects of evidence for person crimes, the odds of non-conviction for property crimes were more
strongly associated with the number and seriousness of charges and the presence or absence of
codefendants. While it is somewhat surprising that the evidence measures did not exhibit significant
effects for property crimes, it may be that the effect of evidence is felt indirectly through victim
characteristics and the willingness of victims to prosecute.
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 252

The effects of case characteristics on the odds of non-conviction among property unit cases
were fairly straightforward. For each increase in the number of charges screened at intake the odds
of non-conviction decreased by approximately 20 percent. For each decrease in statutory class35 of
the top charge screened at intake, there was nearly a 40 percent increase in the odds of nonconviction. As previously noted, both a greater number of initial charges and more serious charges
provide some leeway in negotiating pleas and may also be associated with cases considered high
priority by prosecutors. The presence of codefendants reduced the odds of non-conviction by almost
half, perhaps reflecting an ability to induce codefendants to cooperate with the prosecution.
It is somewhat surprising that no significant effects of case characteristics were found for person
crimes, but it may be that the effects of case seriousness are captured indirectly through the
significant effects of victim characteristics and evidence. For example, cases with a greater number
of victims and/or a greater number of items of physical evidence are likely also to be ones in which
a greater number of offenses were charged.
The size of the effect of race may also seem surprising; it indicates that, controlling for other
variables in the model, the odds of non-conviction were five times greater in person crime cases
involving black defendants than in person crime cases involving non-black defendants. However,
that may not translate into a large difference in the probability of non-conviction. As a hypothetical
illustration, suppose the probability of non-conviction for a non-black defendant with average
values on the other predictors in the model were .10; that translates to odds of .1/.9 = .11. Five
times that, odds of .55, would translate to a probability of non-conviction for black defendants of
approximately .35 (controlling for other factors), not necessarily an unbelievable value.
Nevertheless, the finding that person crime cases involving black defendants were significantly

Statutory class rank is reverse coded; higher values correspond to less serious statutory classes.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 253

more likely than other cases to result in dismissal or acquittal warrants further investigation into the
possibility of systematic overcharging at arrest in cases involving black defendants.
The combined effects of victim characteristics and the combined effects of evidence measures
are also somewhat complicated to interpret. Both the modeled effect of number of victims and the
modeled effect of number of items of physical evidence are in the counterintuitive direction, relative
to the patterns displayed in the raw descriptive statistics. In both cases, the descriptive statistics
suggest greater numbers of victims or items of evidence were associated with lower non-conviction
rates. However, in the logistic regression models, both variables exhibit the opposite effects—
greater numbers appear to be associated with higher non-conviction rates. These are called
suppressor effects; they have to be interpreted in combination with other variables. In the case of
victim characteristics, the combination of number of victims with number of victims willing to
prosecute suggests that more victims yield lower non-conviction rates only if those victims are
willing to prosecute. For property unit cases, for example, if there were two victims (an increase of
one over the minimum) but neither victim was willing to prosecute (an increase of zero over the
minimum), then the odds of non-conviction would be increased almost three-fold. On the other
hand, if both victims were willing to prosecute (an increase of two over the minimum), then the
combined effect would be (2.98)(.17)(.17) = .09, more than a 90 percent decrease in the odds of
non-conviction.
Similar reasoning applies to the combined effects of number of items of physical evidence and
number of items per charge (classified into three levels as less than one item per charge, 1 or more
items but less than 2, and 2 or more items per charge). However, the result is still somewhat
counterintuitive. The model suggests that each increase in the number of items of evidence was
associated with a 30 percent decrease in the odds of non-conviction, as long as there were fewer

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 254

items of evidence than there were charges. If the number of items of evidence was roughly equal to
the number of charges (odds ratio = 1.74), then there would have to be 2 or more items and charges
to yield a combined reduction in the odds of non-conviction. And if the number of items of
evidence was double or more than double the number of charges (1 charge and 2 or more items, 2
charges and 4 or more items, etc.), there would be a net increase in the odds of non-conviction for 1
charge and only 2 or 3 items of evidence, but a net decrease for any combination involving 4 or
more items of evidence. We have no substantive explanation to offer for this pattern of results.
However, we suspect that combinations of evidence measures are indirectly reflecting some
important but unmeasured characteristic of the cases.36 The extremely weak model for predicting
non-conviction in drug cases may be an indirect reflection of the importance of evidence, since
neither direct measures of evidence nor the possible indirect effects of evidence through victim
characteristics could be incorporated in the analyses.
3.7.2 Dismissal Decisions in Northern County
The examination of dismissal decisions in Northern County relies on cases filed between
January 1, 2009 and December 31, 2009; we rely on this subsample of cases to ensure time to final
disposition before the end of the data collection period. The subsample included a total of 18,802
cases.
Descriptive Statistics
Overall, dismissal rates in Northern County were substantially lower than non-conviction rates
in Southern County for comparable offense types (see Figure 3.7.1-1 above). Roughly 22 percent of
cases were dismissed during the study period; in other words, of the 18,802 cases filed in Northern
County between January 1, 2009 and December 31, 2009, 4,140 cases (22.0 percent) were
36

We tested more than two dozen measures of types and amount of evidence, but only the two cited here had consistent
associations with the modeled outcomes. However, we did not have measures of quality of evidence or the match
between available evidence and the evidentiary requirements of a case.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 255

dismissed before adjudication and 14,662 cases (78 percent) were adjudicated through guilty pleas
or trials.
Cases involving felonies were generally dismissed at a lower rate than cases involving
misdemeanors. Moreover, dismissal rates differed markedly by offense type. For example, nearly
40 percent of misdemeanor domestic violence cases were dismissed during the study period,
compared to just 3 percent of misdemeanor DUI cases. Dismissal rates for felony drug, public
order, domestic violence, and weapons offenses were remarkably similar at roughly 20 percent,
significantly higher than the rates for felony person, property, and DUI offenses. These patterns
may be reflections of differences in screening practices and offense characteristics across the
different offense types. Recall that person and property offenses generally had lower acceptance
rates than other offenses, implying either a stricter screening process or, as indicated by prosecutors,
the influence of perceived victim/witness unreliability. Consequently, stronger cases generally made
it through the screening process, resulting in lower dismissal rates. Conversely, drug and weapons
offenses tended to have higher acceptance rates, implying either a weaker screening process or, as
indicated by prosecutors, the influence of perceived evidence reliability. Higher dismissal rates for
these offenses may, in turn, be a reflection of this weaker screening pattern or the re-evaluation of
evidence later in the process.
By comparison, the non-conviction rates for Southern County were 33.1 percent for felony
person crimes, 38.6 percent for felony property crimes, 37.7 percent for the early felony drug crime
sample, and 48.4 percent for the late felony drug crime sample. These compare to dismissal rates of
approximately 20 percent for felony drug crimes and less than 15 percent for felony person and
felony property crimes in Northern County.37
37

These differences are much larger than could be attributable to the substitution of non-conviction as a proxy for
dismissal.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 256

The bivariate relationships between dismissal rates for filed cases and selected case
characteristics in Northern County are displayed in Table 3.7.2-1.
Table 3.7.2-1 Percentage of accepted cases dismissed in Northern County, by selected case
characteristics
Overall
Offense Characteristics
Charge Severity
Misdemeanor
Felony
Number of charges
Single charge
Multiple charges
Enhancement Offense
No
Yes
Arresting Agency
Primary municipal
Other
Drug Type
Cocaine (reference)
Heroin
Marijuana
Other
Defendant Characteristics
Race
White
Black
Hispanic
Gender
Male
Female
Victim Characteristics
Race
White
Black
Hispanic
Gender
Male
Female

Person
18%

Property
15%

Drugs
23%

Public
Order
27%

DV
37%

Weapons
17%

DUI
2%

25%
14%

19%
11%

26%
21%

28%
21%

39%
23%

9%
23%

2%
8%

31%
8%

20%
8%

29%
11%

38%
4%

51%
16%

19%
15%

6%
2%

19%
10%

16%
8%

24%
15%

27%
19%

36%
42%

15%
26%

2%
5%

18%
16%

15%
15%

21%
28%

28%
26%

41%
25%

18%
14%

2%
2%

-----

23%
20%
19%
34%

-----

18%
18%
14%

17%
14%
16%

32%
19%
18%

26%
29%
20%

26%
42%
36%

15%
17%
16%

2%
1%
3%

17%
21%

13%
22%

21%
41%

27%
26%

38%
33%

17%
16%

2%
2%

15%
20%
19%

13%
16%
25%

----

30%
44%
40%

----

18%
18%

16%
14%

---

32%
40%

---

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 257

Dismissal rates varied considerably along several dimensions of case characteristics and
defendant characteristics. Across all offense types, cases involving more serious charges, multiple
charges, and enhancement charges were less likely to be dismissed. The only exceptions were for
weapons offenses and domestic violence offenses; for both of these, cases involving enhancement
charges were more likely to be dismissed than cases without enhancements, and, for weapons
offenses, cases involving felonies were more likely to be dismissed than misdemeanors. Combined
with the screening outcomes – which found that cases involving multiple arrest charges were more
likely to be accepted for prosecution – this indicates that more serious, complex cases are more
likely to continue in the prosecutorial process.
Dismissal rates varied only slightly by arresting law enforcement agency. For most offenses –
person, property, public order, weapons, and DUI – dismissal rates were roughly the same for both
the primary municipal police department and other agencies. For drug cases, however, dismissal
rates for cases involving the primary municipal police department were roughly 10 percentage
points lower than other agencies. For domestic violence cases, however, dismissal rates for cases
involving the primary municipal police department were roughly 15 percentage points higher than
other agencies.
Overall dismissal rates were nearly identical across racial groups with a few exceptions. For
drug offenses, dismissal rates for white defendants were significantly higher than the rates for black
and Hispanic defendants (32 percent versus 19 percent). For public order offenses, dismissal rates
for black defendants (29 percent) were considerably higher than those for both white defendants (26
percent) and Hispanic defendants (20 percent). Finally, dismissal rates for domestic violence
offenses displayed a similar pattern; for these offenses, dismissal rates for black were roughly 6
percentage points higher than the rates for Hispanic defendants and 16 percentage points higher

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 258

than the rates for white defendants. Dismissal rates for men were consistently lower than the rates
for women, with the exception of public order offenses and DUI offenses (in which they were
nearly equal) and domestic violence cases (in which the rates for men were roughly 5 percentage
points higher).38
Dismissal rates based on the victim racial categories showed markedly different trends than
those for defendant racial categories. For person, property, and domestic violence offenses (the
only offense types with available victim information), dismissal rates were consistently lower for
cases involving white victims. In addition, for property offenses, dismissal rates for cases involving
Hispanic victims were markedly higher than rates for cases involving white or black victims
(roughly 12 percentage points higher relative to white victims and 9 percentage points higher
relative to black victims).
Logistic Regression Analyses
The following analyses rely on logistic regression models to examine the decision to dismiss a
case in Northern County. The subsample of 18,803 cases was used to generate a series of logistic
models predicting the likelihood that all charges in a case would be dismissed, controlling for
individual-level factors. Models 1 through 7 in Table 3.7.2-2 examine case outcomes for each
offense type. Two models examining property offenses are included; since not all property offenses
included an identifiable victim, separate models were run for property crimes with a victim and all
property crimes. Coefficients presented in Models 1 through 7 represent the increases in the
likelihood of a case being dismissed generated by each predictor variable. These coefficients are
presented as odds ratios – coefficients greater than 1 imply an increased likelihood of case dismissal

Unfortunately, defendant custody information was not consistently reported after initial charge filing; thus, the
defendant custody variable was not included in the dismissal analyses. Prosecutor information was also missing for
more than 50 percent of the cases. As a result, the study was unable to examine the influence of prosecutor
characteristics on dismissal decisions.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 259

and coefficients below 1 indicate a reduction in such likelihood. In all models, “white” is taken as
the reference category when comparing outcomes across defendant and victim racial groups; the
coefficients listed for other races are the effects of these races on case outcomes compared to
whites. For drug offenses, “cocaine” was used as the reference category when comparing across
drug offenses; the coefficients listed for other drug types are the effects of those drug types on case
outcomes relative to cocaine.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 260

Table 3.7.2-2 Logistic regression predicting case dismissal in Northern County
Model 1
Person

Model 2a
Property

Model 2b
Property

Model 3
Drugs

Model 4
Public Order

Model 5
DV

Model 6 Model 7
Weapons DUI

.897**
.522***
.595
1.026

.819***
.436***
.565
1.153

.894*
.583***
.572
1.129

.939*
.493***
.822
1.022

.905***
.107***
.715
1.074

1.190***
.265***
.961
1.539**

1.700***
.699***
1.343
1.405

1.297**
.670
1.035
.631

-----

-.996
.569***
1.052

-----

Defendant Characteristics
Race
White (reference)
Black
Hispanic
Gender (Male)
Age (Years)

-.648
.939
1.011
1.023**

-.822
.729
.589
1.002

-.849
1.193
.597***
.981***

-.589***
.506**
.478***
.995

-1.196*
.725*
1.137
.993**

-1.154
1.056
.665
1.003

-1.066
1.082
.807
1.014

-.321
1.713
1.099
1.036**

Victim Characteristics
Race
White (reference)
Black
Hispanic
Gender (Male)
Age (Years)

-2.270*
1.447
.862
.992

-1.198
2.532*
1.308
.997

------

-1.322
1.171
1.905
.988*

------

.689
876
733.08
.166
93.76***

2.200
1,407
1,044.80
.153
127.04***

1.609
1,529
1,262.07
.081
73.64***

3.645***
3,053
3,027.14
.117
243.30***

6.988***
2,351
6,472.32
.202
952.54***

1.010
1,864
2,150.76
.225
337.83***

.021***
728
611.53
.104
46.62***

.010***
2,166
423.16
.073
29.87***

Offense Characteristics
Charge Severity
Number of charges
Enhancement Offense (Yes)
Arresting Agency (Primary)
Drug Type
Cocaine (reference)
Heroin
Marijuana
Other

Constant
Total cases
-2 log likelihood
Nagelkerke R2
Chi2

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 261

As Models 1 through 7 show, there were no consistent predictors of case outcomes across
offense types, with the exception of variables measuring offense characteristics. For person,
property, drug, and public order offenses, the strongest predictor of case dismissal was number of
charges issued – for all four offense types, cases involving more serious charges and a greater
number of charges were less likely to be dismissed. In other words, it appears that prosecutors were
less likely to dismiss cases that were more serious and more complex. For example, each increase in
the severity of the controlling filed offense (e.g. from Class A misdemeanor to Class I felony)
decreased the likelihood of dismissal of a drug offense by six percent, while the same one level
increase in the severity of a public order offense decreased the likelihood of dismissal by 14
percent. Having more than one filed charge decreased the likelihood that a case would be dismissed
by anywhere from 50 percent (for drug offenses) to nearly 90 percent (for public order offenses)
after controlling for other factors. Moreover, this was consistent with findings for screening, in
which cases involving more serious charges and a greater number of charges were less likely to be
declined for prosecution for nearly all offense types. In other words, more serious and more
complex cases are generally more likely to move forward in the prosecutorial process – more likely
to be accepted for prosecution and more likely to avoid dismissal. For domestic violence, weapons,
and DUI offenses, this was only partially true – cases involving a greater number of charges were
also less likely to be dismissed, but, for all three offense types, cases with more serious charges
were more likely to be dismissed. The presence of a charge enhancer had no effect in any of the
models. Finally, for drug offenses, the type of drug had little impact on screening decisions;
however, cases involving marijuana were less likely to be dismissed relative to cocaine.
Unlike the screening decision, arresting agency had very little influence on the decision to
dismiss, reaching significance in only two models. For public order and domestic violence offenses,

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 262

arrests made by the primary municipal police department were more likely to be dismissed; while
not significant, the coefficient for arresting agency in all other models indicated a similar trend.
Again, this may be an indication of changes in the investigation units within the police department,
as discussed in Part 2. Prosecutors maintain that there has been a decrease in the quality of
investigations, which translates into less follow-up and cooperation on cases by the policy
department after charges are filed, increasing the likelihood of dismissal of such cases. The higher
dismissal rates, particularly for public order offenses, may be due to what prosecutors argued was
the police department’s increased focus on making arrests.
Defendant characteristics had almost no impact on dismissal decisions, with the exception of
drug offenses, public order offenses, and property offenses. For drug offenses, the presence of a
black or Hispanic defendant decreased the likelihood that a case was dismissed relative to cases
involving white defendants; the effect was quite large, decreasing the likelihood of dismissal by
roughly 50 percent. Moreover, this was contrary to the impact of race on the screening decision for
drug offenses, in which cases involving black defendants were more likely to be declined for
prosecution than cases involving white defendants. This could be an indication that cases involving
black defendants are screened more carefully (resulting in lower dismissal rates later) due to threats
of racial bias or over-charging by law enforcement. Conversely, for public order offenses cases
involving black defendants were more likely to be dismissed, while cases involving Hispanic
defendants were less likely to be dismissed. For both property and drug offenses, cases involving
male defendants were less likely to be dismissed than cases involving female defendants. Finally,
defendant age was significant in four of the eight models (Person, Property, Public Order, and DUI).
For person and DUI offenses, cases involving older defendants were more likely to be dismissed.
The effect appears small; overall, each one year increase in a defendant’s age increased the

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 263

likelihood that a case would be dismissed by less than 2 percent. This, however, could be a nonnegligible effect. The cumulative effect of the difference between age 18 and age 28, for example,
would be about a 22 percent increase in the odds of dismissal for these crime types. This again lends
some support to prosecutors’ focus group statements that defendant characteristics such as age
matter when determining whether a case should be prosecuted or continued.
Victim characteristics also were not consistent predictors of case outcomes. For person offenses,
cases involving black victims were more likely to be dismissed than cases involving white victims;
the presence of a black victim increased the likelihood that a case was dismissed by 127 percent.
Moreover, this was consistent with screening outcomes, in which the presence of a black victim
increased the likelihood that a case was declined for prosecution. Together, these findings imply
that cases involving black victims are funneled out of the system at much higher rates than cases
involving white victims. For property offenses, cases involving Hispanic defendants were more
likely to be dismissed. While not significant, coefficients for race/ethnicity variables hinted similar
trends for all three offense types with victim information (person, property, domestic violence).
Contrary to analyses of screening decisions, victim gender had no impact on dismissal decisions.
Finally, for domestic violence offenses, cases involving older victims were less likely to be
dismissed (while not significant, coefficients in other models were in the same direction); like the
effect of defendant age, the overall effect was small with each one year increase in the victim’s age
decreasing the likelihood that the case would be dismissed by just 1 percent; again, this still could
be a large cumulative effect, if for example the comparison is between young adults and older adults
(e.g. there would be a 40 percent decrease from age 28 to age 48, assuming a linear effect in the logodds metric).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 264

Overall, as with screening decisions, offense characteristics were strong predictors of dismissal
decisions. Taken together, the effects of offense characteristics on dismissal outcomes appear to
indicate that more serious cases (i.e. with more severe charges, more charges) generally are more
likely to continue in the prosecutorial process. This is consistent with prior research that has found
case characteristics generally to be the strongest predictors of decision-making and the perceptions
of prosecutors.
3.7.3 Reasons for Dismissal in Southern County
The case management system used in drug unit cases in Southern County included an
opportunity for prosecutors to select coded reasons for dismissals at each of several stages in case
processing. We incorporated that same coding structure in the instrument used to code data from
paper case files for person unit and property unit cases. From both sources, the reasons for dismissal
were attached to specific charges and the analyses presented in this section were performed at the
charge level rather than the case level. Reasons were not indicated for all of the individual charges
that were dismissed; but reasons were found for a large enough number of dismissed charges to
support an informative analysis. The distributions of reasons for dismissal for person unit and
property unit cases are presented in Table 3.7.3-1 and the distributions of reasons for dismissal for
drug unit cases are presented in Table 3.7.3-2. The analyses of drug unit cases are presented
separately from those of the person and property units due to differences in the detail of data
capturing the stage of the process where the dismissal occurred; the data from the drug unit was
much more precise regarding the timing of dismissal.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 265

Table 3.7.3-1 Distribution of reasons for charge dismissal by processing stage for the Person
and Property Teams in Southern County

Reason for Rejection/Dismissal

Primary
Reason for
Rejection at
Screening

N of charges with dismissal reasons
Contradictory/inclusive lab results
Evidence only supports misdemeanor charge
Evidence was destroyed or missing
Incomplete/missing witness statements
Insufficient evidence for prosecution
Insufficient nexus
No corroboration of evidence
Physical evidence insufficient
Other evidence problem
Total for evidence-related reasons
Total treated as misdemeanor
Pled guilty to other charge in other complaint
Pled guilty to other charge in this complaint
Prosecuting other charge
Total for pursuit of other charges
Interest of justice
No probable cause for arrest
Defendant found incompetent
Other due process problems
No papering
Feds took the case
Deferred prosecution
Death of defendant
Total for all other reasons

760
0.9%
0.7%
0.4%
12.9%
37.4%
6.2%
11.8%
0.1%
0.3%
70.7%
9.6%
0.4%
0.7%
5.8%
6.8%
5.4%
1.4%
0.7%
3.6%
0.8%
-1.1%
-12.9%

Reason for
Voluntary
Dismissal
Before Upper
Court
289
-1.4%
0.7%
8.0%
16.6%
1.4%
3.8%
-1.4%
33.2%
22.8%
2.8%
15.2%
4.8%
22.8%
4.8%
-0.7%
5.2%
2.4%
3.1%
4.8%
-21.1

Reason for
Voluntary
Dismissal in
Upper Court
860
---7.8%
10.8%
0.7%
1.0%
0.1%
1.0%
21.5%
1.4%
13.7%
47.3%
1.3%
62.3%
1.0%
-0.7%
6.5%
-5.6%
0.1%
0.8%
14.8%

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 266

Table 3.7.3-2 Distribution of reasons for charge dismissal by processing stage for the Drug
Unit in Southern County
Reason for Rejection/Dismissal

Screening

N of charges with dismissal reasons
Contradictory/inclusive lab results
Evidence only supports misdemeanor charge
Evidence was destroyed or missing
Incomplete/missing witness statements
Insufficient evidence for prosecution
Insufficient nexus
No corroboration of evidence
Victim delay in reporting
Victim unlocatable
Witness credibility/bias
Witness refuses to cooperate
Analytical results insufficient
Physical evidence insufficient
Other evidence problem
Total for evidence-related reasons
Total treated as misdemeanor
Pled guilty to other charge in other complaint
Pled guilty to other charge in this complaint
Prosecuting other charge
Total for pursuit of other charges
Interest of justice
No probable cause for arrest
Unlawful search, no warrant
Other due process problems
Affirmative defense
Defendant found incompetent
No papering
Feds took the case
Deferred prosecution
Low priority for unspecified reasons
Resource limitations
Death of defendant
Total for all other reasons

1171
-1.6%
-0.1%
87.5%
3.8%
-0.2%
0.1%
0.1%
0.3%
-0.4%
2.0%
96.1%
0.3%
--1.6%
1.6%
-0.5%
-1.1%
0.1%
-0.2%
-----1.9%

District
Court
532
1.3%
---13.3%
----0.9%
0.8%
0.8%
0.2%
1.1%
18.4%
-6.0%
38.5%
-44.5%
1.5%
--0.8%
--33.1%
1.7%
----37.1%

GJ and
GJ Prep
964
3.9%
.4%
-1.0%
13.9%
-.1%
-0.1%
3.0%
-3.1%
1.5%
2.1%
29.1%
-3.4%
28.4%
-31.8%
0.6%
-0.7%
0.3%
0.2%
--0.5%
-36.6%
--38.9%

Admin
Court
2844
2.0%
0.1%
--5.9%
----1.4%
-1.1%
0.1%
0.9%
11.5%
-13.9%
66.9%
-80.8%
0.6%
--0.1%
-0.4%
0.3%
6.3%
----7.7%

Trial
Court
224
0.9%
0.4%
-1.3%
29.0%
------0.4%
2.7%
6.3%
41.0%
-16.5%
30.8%
-47.3%
7.1%
------4.0%
--0.4%
-11.5%

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 267

As these tables indicate, prosecutors cited evidence-related reasons for rejecting charges at
screening for 70.7 percent of the charges rejected in person and property cases and 96.1 percent of
charges rejected in drug cases. For person and property cases, the percentage attributed to
evidentiary problems declined to 33.2 percent of charges dismissed prior to filing in upper court and
21.5 percent of the charges dismissed after reaching upper court. For drug cases, the percentage
attributed to evidentiary reasons declined dramatically to 18.4 percent of charge dismissed in lower
court, 29.1 percent of charges dismissed during grand jury preparation or presentation, and 11.5
percent of charges dismissed in felony administrative court. However, evidence-related reasons
were cited for 41.0 percent of the charges dismissed in felony trial court.
As the percentage of dismissals attributable to evidentiary problems declined, other reasons
necessarily accounted for increasing percentages of dismissals. The most striking was the increase
in the percentage of dismissals attributable to the fact that prosecutors were pursuing other charges
or considered guilty pleas to other charges to constitute acceptable outcomes. For person and
property cases, the percentage of dismissals attributable to a focus on other charges grew from 6.8
percent of charges rejected at screening, to 22.8 percent of charges dismissed prior to reaching
upper court, and to 62.3 percent of charges dismissed in felony administrative court or felony trial
court. For drug cases, the percentage grew from 1.6 percent of charges rejected at screening to 44.5
percent of cases dismissed in lower court, 31.8 percent of charges dismissed during grand jury
preparation or presentation, 80.8 percent of charges dismissed in felony administrative court, before
dropping back somewhat to 47.3 percent of charges dismissed in felony trial court. These patterns
almost certainly involve dismissing some charges in exchange for guilty pleas to the charges
prosecutors consider most appropriate given case characteristics and the existing circumstances. It is
not clear from these data how often prosecutors obtain guilty pleas to the charges they consider

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 268

most appropriate, and how often they end up dismissing charges they do not believe should be
dismissed due to contextual circumstances that render a case difficult to prove or require lowering
its priority relative to other pending cases.
Three other specific items are worth highlighting. For person and property cases, a significant
number of charges (22.8 percent of those dismissed prior to reaching upper court) were reduced to
misdemeanors and pursued in lower court. For drug cases, 33.1 percent of the cases dismissed in
lower court were dismissed for lack of “papering” (the case preparation associated with filing a
charge in court). Also for drug cases, 36.6 percent of the cases dismissed during grand jury
preparation were assigned an undocumented reason code. However, that code only occurs in
connection with a particular type of plea offer. In recording plea offers in the case management
system, prosecutors in the drug unit assign a priority code to each charge: “1” for “must plead
guilty” and “2” for “will dismiss.” Some charges are assigned a priority code of “3,” which means
that the charge has already been dismissed and the fact that it was dismissed is considered a part of
the plea bargain. Grand jury preparation and plea offer development occur more or less
concurrently, and charges coded as already dismissed in the plea offer usually have the
undocumented code mentioned above recorded as the reason for dismissal. Thus, the true reason for
dismissal of those charges may be similar to those associated with the category labeled “pursuit of
other charges,” which may then account for as many as 65 percent of the charges dismissed during
grand jury preparation.
The results of the analyses are highly consistent with our findings from other analyses presented
in this report. Prosecutors who participated in the focus group discussions indicated that the strength
of evidence was the dominant consideration at initial screening, and that while continual
reevaluation of evidence remained important throughout the life of a case, it was balanced by the

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 269

increasing importance of other considerations in making charging and plea offer decisions and
setting priorities among cases. Our analyses of screening, charging, plea offers, and sentence
recommendations across several data sets generally support this conception.

3.8 Understanding Case Outcomes across the Prosecutorial Process
The analyses of case outcomes clearly support prosecutors’ identification of strength of
evidence as the primary consideration at initial screening. In the analyses of hypothetical cases in
the factorial survey, prosecutors’ rating of the strength of evidence was the only case-level variable
with a statistically significant influence on the screening decision and higher ratings of the strength
of evidence were associated with greater numbers of charges requiring guilty pleas in a plea offer.
For actual person and property cases screened in Southern County, the combined effects of
evidence-related measures far outweighed the combined effects of other factors in predicting the
screening outcome. In the analyses of reasons for dismissal in Southern County, evidence related
reasons were the dominant reasons for dismissal at screening.
Yet, throughout the prosecutorial process, prosecutors continue to ask the question, ”Should the
case proceed?” as additional information becomes available or prior information changes. In
general, the findings support statements by prosecutors that evidence related factors decrease in
importance as a case progresses. Indeed, in analyses of reasons for dismissal in Southern County,
the percentage of dismissals attributable to evidentiary problems declined dramatically after initial
screening, which would be expected if screening is successful in removing weak cases. According
to prosecutors, non-evidence related factors – particularly offense seriousness – become relatively
more important after initial screening. When making charging decisions in hypothetical cases, for
example, prosecutors were more likely to issue additional charges at charging in cases involving
felonies than misdemeanors. When making plea offers in hypothetical cases, more serious offenses
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 270

were associated with greater numbers of charges requiring guilty pleas. In Northern County, cases
involving felonies were less likely to be dismissed than cases involving misdemeanors.
Analyses of case outcomes in both counties indicate that additional factors were also associated
with case outcomes. In Northern County, for example, cases involving younger defendants were
more likely to be rejected for prosecution and were more likely to be dismissed for several offense
types. Overall, race had little impact on case outcomes in either jurisdiction. But, the presence of a
black defendant decreased the likelihood that a drug case was accepted for prosecution in both
Northern and Southern Counties; conversely, the presence of a black or Hispanic defendant
increased the likelihood that a domestic violence case accepted for prosecution in Northern County.
For a few offense types in Northern County, the presence of a male defendant increased the
likelihood that a case was accepted for prosecution and decreased the likelihood that a case was
dismissed. Although defendant race/ethnicity was not a consistent predictor of case outcomes,
victim race/ethnicity was. For person, property, and domestic violence offenses in Northern
County, cases involving black victims were less likely to be accepted for prosecution than cases
involving white victims; and cases involving male victims were also less likely to be accepted for
prosecution in all three models. Victim characteristics had little impact on decisions after the initial
screening decision in either jurisdiction.
Although prosecutors maintained that unique case factors affected case outcomes, they also
expressed a mixture of concern about consistency. A majority of prosecutors considered achieving
consistency in outcomes—both within and across units—to be an important objective for defining
office-level success and nearly all agreed that there should be a great deal of consistency across
prosecutors in the case-specific factors that influence decision-making. Yet, there remained
significant variation in screening decisions across prosecutors. In Southern County, for example, the

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 271

range of acceptance rates among prosecutors screening person crimes varied from a low of 70
percent to a high of 84 percent; for property crimes, prosecutors’ acceptance rates ranged from 59
percent to 72 percent and for drug crimes they ranged from 76 percent to 87 percent. Analyses of
screening outcomes in Northern County found much wider variations in acceptance rates across
prosecutors: for felony person offenses, acceptance rates varied from a low of 32 percent to a high
of 61 percent and for felony property offenses ranged from a low of 36 percent to a high of 87
percent. Other offense types showed similarly wide variations, with the exception of weapons
offenses and DUI offenses.
Although there was wide variation in acceptance rates across prosecutors, there was little
variation in charging decisions for actual cases. Analyses of administrative data in both sites
revealed that prosecutors tended to either accept all of the arrest charges presented for screening or
reject them all and, overall, few differences existed between the number or severity of arrest charges
and the number or severity of charges initially filed by the screening prosecutor. Among the cases
accepted for prosecution in Northern County, for example, the most serious arrest charge was filed
in roughly 90 percent of cases for all offense types and offense levels. Thus, it appears that the
screening decision is the more important decision. It may also indicate that the screening decision
and charging decision are even separable; prosecutors may determine what charges they could
prove given the case presented at screening and then make the decision to proceed on that basis.
The factorial survey results, however, indicated that these decisions may be separable in some
instances. In the factorial surveys analyses, there was significant variation among prosecutors in
average seriousness of top filing charge (controlling for crime seriousness, strength of evidence, and
criminal history), as well as significant variation in how prosecutors weighed crime seriousness,
strength of evidence, and criminal history in deciding what to charge.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 272

Given the wide variation in acceptance rates across prosecutors, even for similar cases, some
analyses examined the effects of prosecutors’ characteristics on decision outcomes. In Northern
County, for example, analyses of screening decisions showed that prosecutor-level characteristics
explained roughly 10 percent of the variance in case outcomes; however, very few prosecutor-level
predictors were significant in the models. Moreover, the few significant results were mixed. For
property offenses, more experienced prosecutors were less likely to accept cases for prosecution, but
for domestic violence offenses more experienced prosecutors were more likely to accept cases for
prosecution. For drug offenses, prosecutors who had screened more drug cases were more likely to
accept drug cases for prosecution; in contrast, for weapons offenses, prosecutors who had screened
more cases involving weapons offenses were less likely to accept such cases for prosecution. The
fact that the prosecutor was a supervisor was not significant in any of the models, and the fact that a
prosecutor was part of a specialized unit was significant in just one model, with prosecutors in
specialized prosecution units more likely to accept cases involving weapons offenses.
Although screening decisions in the factorial survey were very consistent across prosecutors,
differences in prosecutors’ attitudes did account for some variation in screening decisions. For
example, higher prosecutor ratings of the importance of consistency (from responses to the general
survey) were associated with lower odds of rejection at screening; higher ratings of the importance
of examining cases at screening in terms of their potential for plea bargaining, early disposition, and
caseload reduction were associated with higher odds of rejection at screening. Prosecutors were less
consistent in their choices of the most serious charge to file in the factorial survey; the number of
charges prosecutors indicated they would file was, again, influenced by their general attitudes
toward the most appropriate prosecution strategies. The more strongly prosecutors agreed that the
charging decision should include the highest charges that could be proven at trial, the greater the

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 273

average number of charges the prosecutors indicated they would file. Similarly, prosecutors who
indicated that their general approach to charging was to file only the charges that the accused should
plead guilty to identified fewer charges for initial filing.
Overall, the analyses of case outcomes in Northern and Southern Counties supports much of the
research on the factors associated with case outcomes. However, the analyses also highlight several
new findings – the relative decreasing importance of evidence-related factors over the life of a case,
the often conflicting influence of case-specific factors across offense types, the lack of consistency
in outcomes across prosecutors and the impact of prosecutor characteristics on case outcomes.
These findings point to potential new areas of research that can both deepen our understanding of
prosecutorial outcomes and broaden our approach to exploring them.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 274

Part 4. Conclusions
Understanding the anatomy of prosecutorial discretion is no easy task. In any individual case,
prosecutors make innumerable subtle decisions that propel a case forward or hold it back or that
increase the exposure a defendant may face upon conviction or decrease that exposure. In any
individual case, these decisions are based on a complex evaluation of case-level factors and a
calculation of expected and desired outcomes. Although it may be clear which case-level factors are
most important in making these decisions and how prosecutors weigh those case-level factors, these
subtle decisions about cases are not made in a vacuum. Although prosecutors discuss their
discretion in individual cases as being nearly unbounded, external forces significantly shape and
constrain that discretion. Case-level decisions are often determined by, or at least limited by,
resource constraints, office policies, and a balancing of practical needs for efficiency, consistency,
and fairness. Thus, in the end, individual prosecutorial discretion is not nearly as unbounded as
traditionally thought.
In this chapter we briefly summarize what we learned about prosecutorial decision-making in
Northern and Southern Counties. Rather than compiling findings from each of the substantive
chapters included in this report, we sought to highlight several key findings and primary challenges
facing prosecutors.

4.1 Key Findings
4.1.1 Prosecutors Are Guided by Two Questions: ‘Can I prove the case?’ and ‘Should I prove the
case?’
According to prosecutors, decisions in any individual case are based on the strength of the
evidence, the seriousness of the offense, and the defendant’s criminal history. Interviews with
prosecutors revealed that strength of the evidence is used to answer the question ‘Can I prove the
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 275

case?’ Prosecutors also argued that seriousness of the offense and defendant criminal history are the
primary factors then used to answer the question ‘Should I prove the case?’ In other words, while
strength of the evidence is used to determine if a case can likely result in a conviction, prosecutors
look to other factors to determine if a case should be prosecuted at all, regardless of the likelihood
of conviction. In this sense, prosecutors clearly see a distinction between the strength of the
evidence and the merits of the case and see strength of the evidence as a necessary but not sufficient
condition for a case to proceed.
Analyses of actual case outcomes and hypothetical cases clearly supported prosecutors’
perceptions of the impact of these factors. Moreover, as a case proceeds through the prosecutorial
process, the relative importance of each factor changes. Strength of the evidence is very important
at initial screening, but its relative importance declines as a case proceeds; in turn, seriousness of
the offense and defendant criminal history increase in importance as a case proceeds. Several
defendant characteristics also were considered by prosecutors in addition to criminal history when
determining whether or how a case should proceed, including defendant age, the potential impact of
a conviction on the defendant, and the demeanor of the defendant. Moreover, analyses of actual
case outcomes revealed that defendant and victim characteristics such as race and gender were also
associated with case outcomes.
The influence of factors beyond strength of the evidence, seriousness of the offense, and
defendant criminal history on case outcomes supports the argument made by most prosecutors that
prosecutorial decision making is incredibly complex. Decisions are a balance of multiple case-level
factors. In the end, prosecutors argued that determining the right outcome for a case or answering
the question of whether or how a case should proceed revolved around determining what was fair
for the particular defendant, a reflection of the guiding philosophy within each office of “doing

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 276

justice.” While prosecutors were given discretion to translate that philosophy into practice and
could, conceivably, translate it into a focus on crime control, most prosecutors saw their job as
ensuring justice for defendants and evaluated their own success and office success in terms of the
fair treatment of defendants.
4.1.2 Prosecutorial Decision Making Is Often Affected by External Constraints
Individual cases exist within a system that often constrains prosecutorial decision-making.
Three external constraints operated in both Northern and Southern Counties: rules, resources, and
relationships. In some instances, these constraints determined case outcomes. In other instances,
they placed limits on available outcomes. And, in some instances, they simply shaped the decisions
in a case.
In both jurisdictions, a very limited number of office-wide rules dictated how prosecutors were
to handle cases. These rules generally involved specific offenses targeted by the District Attorney
and called for specific outcomes at particular stages of the prosecutorial process (e.g. DUI cases
above a certain blood-alcohol level cannot be rejected for prosecution, felony residential breaking
and entering cannot be reduced to a misdemeanor, residential burglary cases must have a sentence
recommendation of prison). A complex system of unit-specific rules, however, dictated outcomes in
a much wider array of offenses and decision points. The district attorney in each jurisdiction relied
on experienced unit supervisors to develop rules within units. The rules were unorganized, in that
rules were not coordinated across units or always expressed as formal practices; rather, rules
evolved out of norms of practice and could shift with a change in unit supervisor. Rules did not
cover all offenses or decisions; but the rules that did exist were seen as largely binding by
prosecutors and called for prosecutors to decline certain cases at screening, charge cases in a
particular way, and offer specific charges or sentences in plea offers. Prosecutors did agree that they
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 277

could disregard these rules if they could justify their reasoning to supervisors and colleagues;
however, prosecutors also noted that peer pressure within each unit tended to normalize most
responses.
While rules influenced decision making in specific cases, prosecutors argued that a more
important influence on case outcomes was the lack of external resources. In both jurisdictions, a
lack of court resources (i.e. a limited number of courtrooms; not enough judges, clerks, and court
reporters to staff all courtrooms; closing of courtrooms mid-day or mid-week) was expressed as a
persistent problem facing prosecutors. In Southern County, the lack of courtroom space or the
continuance of cases due to a lack of space required prosecutors to charge cases differently, to
continually re-evaluate and change plea offers, and in some instances, to simply dismiss cases to
ensure a high volume of cases were disposed of before trial. Prosecutors described a process of
prioritizing in which cases were ranked from strongest to weakest (based on evidence, offense
severity, victim cooperation, time since initial filing, etc.), which changed the threshold of what
prosecutors were willing to accept or dismiss. According to some, this resulted in prosecutors
making decisions that they would not otherwise make. Moreover, the decision of what to do on a
particular case was often outside of the control of an individual prosecutor; rather, when resource
constraints required a re-evaluation of cases, all cases were evaluated by the unit as a whole and the
unit decided which cases to dispose of. In Northern County, the lack of court resources appeared to
affect only the screening and charging decisions; since the office knew that court resource
constraints prevented it from as many cases as they would otherwise think appropriate, prosecutors
worked harder to evaluate cases for declination and deferral at screening.
Prosecuting cases relies on a web of relationships with other justice system actors – law
enforcement, defense attorneys, judges. Of primary concern in both jurisdictions was the

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 278

relationship between prosecutors and law enforcement officers and the pressure law enforcement
officers often applied to prosecutors to accept cases. Prosecutors at all experience levels admitted
that, in some instances, they succumb to the pressure from law enforcement and accept weak cases
they would not otherwise accept. Prosecutors also acknowledged that the law enforcement officer
assigned to a case impacted the way a case was handled in very practical ways; for example, if an
officer was not very good at testifying on the stand, prosecutors may seek to dispose of a case early
through a generous plea offer in order to avoid having an officer testify. Both jurisdictions have also
been affected by changes in the largest municipal police departments in each county, with
relationships deteriorating somewhat due to a change in law enforcement leadership. According to
all respondents, the quality of information and the quality of cases coming from the police
department has deteriorated due to a decrease in law enforcement emphasis on investigations. The
result of changes has been an increase in cases declined for prosecution, pended for additional
information, or dismissed due to poor follow-up investigation.
The particular judge assigned to a case had limited impact on prosecutorial decision- making. In
Southern County, prosecutors argued that the judge did not affect outcomes primarily due to a
particular feature of the court structure, in which judges rotated through the county every month and
cases were scheduled for trial before the judge was assigned. In contrast, in Northern County, each
unit was assigned to a specific courtroom and appeared before the same judge for roughly one year
before the judge was rotated out. Some prosecutors acknowledged that, as a result, prosecutors
tailored screening, charging, and plea offer decisions based on the judge – e.g. rejecting cases that
the specific judge would likely not find serious, increasing or decreasing charges to ensure the judge
would impose a particular sentence, or offering a plea based on what the judge was likely to find
acceptable.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 279

4.1.3 Inconsistency and Variation in Approach Are Common
Prosecutors who participated in the focus groups expressed a mixture of concern about
consistency. Prosecutors were generally confident that the combination of the unit organization and
the supervisory and communication practices within units served to adequately attenuate
inconsistency, but they were also accepting of the idea that a certain degree of inconsistency is both
inevitable and desirable. Ratings by the broader sample of prosecutors who responded to the general
survey were less mixed. Large majorities considered achieving consistency in outcomes—both
within and across units—to be an important objective for defining office-level success and nearly
all agreed that there should be a great deal of consistency across prosecutors in the case-specific
factors that influence decision-making.
There was, however, a great deal of variability in opinions and outcomes across prosecutors. In
the general survey, prosecutors displayed widely divergent views about the goals of the criminal
justice system, charging philosophies, and plea bargaining strategies. These differences in views
accounted for some of the variation in screening, charging, and plea offer decisions in hypothetical
cases. Greater variation among prosecutors was observed for screening decisions in actual cases,
with the ranges of acceptance rates for individual prosecutors varying widely for nearly all offense
types. In analyses of screening decisions in Northern County, prosecutor-level characteristics
explained roughly 10 percent of the variance in case outcomes. In some instances, prosecutor
experience – either tenure as prosecutor or familiarity with a particular offense type – affected
opinions and outcomes; in other instances, such variation was not explained by any measured
prosecutor characteristic.
The difference between the high degree of consistency in the screening judgments in responses
to the factorial survey and the greater variability in outcomes among prosecutors screening actual

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 280

cases is interesting. It suggests that the factorial survey responses may be closer to what the
prosecutors believed would be the ideal responses, unconstrained by contextual influences. If so, it
could be that a primary source of inconsistency in prosecutors’ decision making in actual cases may
be found in how prosecutors respond to the pressures of contextual circumstances, rather than how
they evaluate case characteristics.

4.2 Continuing Challenges
4.2.1 External Resources
A significant challenge facing prosecutors’ offices is the lack of resources. Prosecutors often
struggle with securing adequate resources within their own offices for maintaining staffing levels,
updating technology, and investigating cases. Often overlooked, however, is the impact that
external resources – primarily court resources – may have on the prosecution process. While not
entirely constrained by resources, prosecutors clearly see the lack of adequate court resources as
limiting their ability to fully prosecute cases. Moreover, in several instances, resource constraints
force prosecutors to dismiss cases or reduce plea offers although they feel such decisions are not the
appropriate or right decisions.
Addressing this challenge may be beyond the ability of any individual district attorney. In
Northern County, the solution has been to shift internal resources to the screening decision and the
diversion of cases out of the system. Stricter screening procedures result in lower acceptance rates
which alleviate some of the pressure in courtrooms; working with outside agencies to develop and
implement programs can also help divert cases away from courtrooms. In Southern County, the
solution has been to ‘roundtable’ cases. When court resource constraints begin to affect case flow,
individual units evaluate cases to determine which cases are the best candidates for dismissal,
decreases in plea offers, or continuance for trial.
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 281

Both solutions work to ensure that the most important cases move forward and both solutions
have benefits. In Northern County, the stricter screening process may produce more efficient
outcomes as the office expends fewer resources on cases that may ultimately necessitate a dismissal.
In Southern County, the use of roundtables may produce more consistent outcomes as the office
considers the totality of cases in making decisions and weighs the relative merits of cases against
each other.
4.2.2 Training
District Attorneys also struggle with training new prosecutors who must learn everything from
practical skills, such as filling out paper work or properly addressing the court, to more nuanced
skills, such as evaluating evidence and crafting the appropriate plea offer. At the same time, new
prosecutors must be ‘normalized’ to ensure that their attitudes conform to office philosophy and the
outcomes in their cases are consistent with those of their colleagues.
District Attorneys can address this challenge in several ways – by having new prosecutors
‘shadow’ more senior prosecutors, by having supervisors review the decisions of new prosecutors,
by creating training and orientation periods for new prosecutors. In Southern County, the solution
has been to create a unit comprised solely of new prosecutors who handle only misdemeanor cases
for six to eighteen months and are supervised closely by an experienced prosecutor. The idea is to
teach new prosecutors how to work independently, to give them extensive courtroom experience,
and to allow them to explore the use of discretion; after new prosecutors are transferred to felony
units, they are then normalized through unit-specific rules, greater interaction with colleagues, and
roundtables in which they must justify and debate their decisions with colleagues. In Northern
County, the solution has been to create general crimes units comprised of both new and experienced
prosecutors who handle both misdemeanor and felony cases. The idea is to normalize new
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 282

prosecutors to the guiding philosophy of the office and to teach them the unit-specific rules and
expectations right away; new prosecutors still learn to work independently, gain extensive
courtroom experience, and explore the use of discretion, but do this largely with more experienced
mentors.
Again, both solutions work to ensure that prosecutors learn to balance independence with office
norms. In Southern County, the approach is to allow independence to grow and then teach
prosecutors to fit decisions to office policies and priorities. In Northern County, the approach is to
teach prosecutors to fit decisions to office policies and priorities and then allow prosecutors to
explore their independence within those constraints.
4.2.3 Balancing Consistency and Flexibility
District Attorneys also face the challenge of ensuring consistency in approach and outcomes
within an environment imbued with significant discretion. Training can help to attenuate differences
in philosophy that could translate into differences in approach and outcome. Specific rules for
screening cases, charging cases, dismissing cases, and plea offers can ensure uniform outcomes
across cases and prosecutors. Yet, the complexity of cases requires flexibility. Moreover, within an
environment of constrained resources, such flexibility may not only be desirable but necessary.
In Southern County, the solution has been to rely largely on roundtabling as a mechanism to
ensure consistency in both approach and outcomes. In some instances, decisions in cases are made
by the entire unit during roundtabling sessions; all members of the unit discuss the specific charges
to file, the plea offer to make, or whether a particular case should be dismissed. The office also
relies on the most experienced prosecutors within any unit to do all screening; thus, a small number
of prosecutors do all screening, and prosecutors perceived a noticeable increase in consistency in
screening decisions. Finally, the horizontal prosecution of cases within the drug unit in Southern
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 283

County ensures a certain level of consistency since prosecutors see the decisions of their colleagues
throughout the processing of all cases. In Northern County, the solution has been to rely largely on
training, peer pressure from colleagues through informal discussions, and unit supervisors to
communicate a philosophy and goals of prosecution. The district attorney also uses the office’s case
management system to routinely examine case outcomes with unit managers.
While both solutions work to ensure consistency, it appears that the approach in Southern
County produces a greater level of consistency in decisions across prosecutors. The routine
discussion of cases and requirement that prosecutors justify their decisions to colleagues tends to
create similar outcomes. The lack of formalized discussions about cases in Northern County appears
to create greater variation in outcomes across prosecutors.

4.3 Limitations
While these analyses provide a necessary understanding of prosecutorial decision-making, there
remain several limitations to the research. First, the analyses rely on two fairly large prosecutors’
offices. The majority of prosecutors’ offices in the United States are not nearly as large as those in
Southern and Northern Counties. Thus, the decision-making processes, the constraints on decisionmaking, and the attitudes prosecutors express about their roles in the criminal justice system may
vary greatly from those of smaller (or larger) prosecutors’ offices. Future research should explore
such issues in different sized offices.
Second, the analyses of actual case outcomes rely on a very limited set of covariates and
outcomes. The analyses in Northern County, for example, lack any measures of evidence strength,
victim/defendant relationship, or other factors potentially associated with outcomes (e.g. weapons
use, defendant criminal history, etc.); moreover, they are limited to an examination of screening,
charging, and dismissals, lacking any ability to examine plea offers or sentence recommendations.
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 284

In Southern County, the analyses of administrative data are limited to just three types of felonies
and contain no information on prosecutors. As such, the study is limited in its ability to determine
what case-level, prosecutor-level, and contextual variables affected case outcomes. Future
evaluations of prosecutorial case outcomes should seek to include additional covariates, particularly
measures of evidence strength and prosecutor-level factors, that may be associated with variation in
outcomes across several stages of the prosecutorial process.
Third, the study relies on a very small subsample of prosecutors in Northern County. Response
rates to focus group requests, factorial surveys, and the general survey were very low in Northern
County. Consequently, it was not possible to compare results across jurisdictions or to compare
responses on the general survey to factorial survey responses. Future research should seek to ensure
comparative analyses are possible in order to better explore the influence of structural differences or
office policies on prosecutors’ attitudes and decision-making processes.
Finally, influential contextual factors were identified in this study through qualitative analysis of
a limited number of individual interviews and focus group discussions in two jurisdictions.39 The
insights gained from these analyses provide a rich source of hypotheses for further study, but it is
not clear how well these findings might generalize across offices and circumstances. It remains to
develop objective measures of these factors, so that future research can begin to quantify and assess
the prevalence of their influence.

A total of seven individual interviews and 10 focus group sessions.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 285

4.4 Conclusion
Prosecutors make a multitude of inter-related decisions throughout the course of a case. How
they make these decisions is not straightforward; rather, it is influenced by case-level factors,
several internal and external constraints, and a balancing of several practical goals of prosecution.
Understanding the black box in which prosecutorial decision-making exists requires further
evaluation of this multitude of factors and the contexts in which prosecutors make decisions.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 286

References
Adams, K. & Cutshall, C. (1987). Refusing to prosecute minor offenses: the relative influence of
legal and extralegal factors. Justice Quarterly, 4, 595-609.
Albonetti, C. (1986). Criminality, prosecutorial screening, and uncertainty: toward a theory of
discretionary decision making in felony case processing. Criminology, 24, 623-44.
Albonetti, C. (1987). Prosecutorial discretion: the effects of uncertainty. Law and Society Review,
21, 291-313.
Albonetti, C. A. (1991). An integration of theories to explain judicial discretion. Social Problems,
38, 247-266.
Alozie, N. O. & Johnston, C.W. (1999-2000). Probing the limits of female advantage in criminal
processing: pretrial diversion of drug offenders in an urban county. Justice System Journal,
21, 239-59.
American Bar Association (1993). ABA Standards for Criminal Justice: Prosecution and Defense
Function, 3rd ed. Retrieved from http://www.americanbar.org/publications/criminal_justice_
section_archive/crimjust_standards_pfunc_tocold.htm.
Applebaum, L. D., Lennon, M.C., & Aber, J.L. (2006). “When effort is threatening: the influence of
the belief in a just world on Americans’ attitudes toward antipoverty policy. Political
Psychology, 27, 387-402.
Applegate, B. K. (1997). Specifying Public Support for Rehabilitation: A Factorial Survey
Approach. Washington DC: U.S. Department of Justice; prepared under grant number 96IJ-CX-0007.
Bjerk, D. (2005). Making the crime fit the penalty: the role of prosecutorial discretion under
mandatory minimum sentencing. Journal of Law and Economics, 48, 591-625.
Chen, E. Y. (2008). The liberation hypothesis and racial and ethnic disparities in the application of
California's Three Strikes Law. Journal of Ethnicity in Criminal Justice, 6, 83-102.
Davis, A. J. (2008). Racial fairness in the criminal justice system: the role of the prosecutor.
Columbia Human Rights Law Review, 39, 202.
Dixon, J. (1995). The organizational context of criminal sentencing. American Journal of Sociology,
100(5), 1157-1198
Engen, R. L. & Steen, S. (2000). The power to punish: discretion and sentencing reform in the war
on drugs. American Journal of Sociology, 105, 1357-1395.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 287

Eisenstein, J., Flemming, R. B., & Nardulli, P. F. (1988). Contours of justice: communities and
their courts. (National Criminal Justice Reference Service Document No. NCJ 107140).
Boston, MA: Little, Brown and Co.
Eisenstein, J. & Jacob, H. (1977). Felony Justice: An Organizational Analysis of Criminal Courts.
Boston: Little, Brown and Company.
Farrell, J. (2003). Mandatory minimum firearm penalties: a source of sentencing disparity? Justice
Research and Policy, 5, 95-115.
Foley, M.O. (2000). Police Perjury: A Factorial Survey. Ph.D. dissertation, The City University of
New York.
Franklin, T.W. (2010). Community influence on prosecutorial dismissals: a multilevel analysis of
case- and county-level factors. Journal of Criminal Justice, 38, 693-701.
Free, M. (2002). Race and presentencing decisions in the united states: a summary and critique of
the research. Criminal Justice Review, 27, 203-32.
Frohmann, L. (1997). Convictability and discordant locales: reproducing race, class, and gender
ideologies in prosecutorial decisionmaking. Law and Society Review, 31, 531-55.
Hartley, R.D., Maddan, S. & Spohn, C.C. (2007). Prosecutorial discretion: an examination of
substantial assistance departures in federal crack-cocaine and powder- cocaine cases. Justice
Quarterly, 24, 382-407.
Hosmer, D.W. & Lemeshow, S. (2000). Applied Logistic Regression. Second Edition. John Wiley
& Sons, Inc. New York.
Hunter, C. & McClelland, K. (1991). Honoring accounts for sexual harassment: a factorial survey
analysis. Sex Roles: A Journal of Research, 24, 725-750.
Jacoby, J.E., Mellon, L.R., Ratlidge, E.C., & Turner, S. (1982a). Prosecutorial Decision Making: A
National Study. Washington, DC: U.S. Department of Justice, National Institute of Justice.
Jacoby, J.E., Mellon, L.R., Ratlidge, E.C., & Turner, S. (1982b). The Standard Case Set: A Tool for
Criminal Justice Decisionmakers. Washington, DC: U. S. Department of Justice, National
Institute of Justice.
Jasso, G. (1998). Exploring the justice of punishment: framing, expressiveness, and the just prison
sentence. Social Justice Research, 11, 397.
Jasso, G. (2006). Factorial survey methods for studying beliefs and judgments. Sociological
Methods & Research, 34, 334-423.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 288

Johnson, B.D. (2003). Racial and ethnic disparities in sentencing departures across modes of
conviction. Criminology, 41, 501-542.
Johnson, B. D. (2005). Contextual disparities in guidelines departures: Courtroom social contexts,
guidelines compliance, and extralegal disparities in sentencing. Criminology, 43(3), 761796.
Johnson, B.D. (2006). The multilevel context of criminal sentencing: integrating judge and county
level influences in the study of courtroom decision making. Criminology, 44, 259-298.
Kingsnorth, R.F. & Macintosh, R.C. (2004). Domestic violence: predictors of victim support for
official action. Justice Quarterly, 21, 301-28.
Kramer, J.H. & Ulmer, J. (2002). Downward departures for serious violent offenders: local court
‘corrections’ to Pennsylvania’s Sentencing Guidelines. Criminology, 40, 601-36.
Mears, D.P. (1998). The sociology of sentencing: reconceptualizing decisionmaking processes and
outcomes. Law and Society Review, 32, 667-724.
Mellon, L.R. (1980). Concept for measuring the legal evidentiary strength of criminal cases.” In
J.E. Jacoby (Ed.), Prosecutorial Decisionmaking – Selected Readings. United States: Bureau
of Social Science Research, Inc., NCJ 79210.
Miethe, T. (1987). Charging and plea bargaining practices under determinate sentencing: an
investigation of the hydraulic displacement of discretion. Journal of Criminal Law and
Criminology, 78, 155-76.
Miller, J.L., Rossi, P.H., & Simpson, J.E. (1991). Felony punishments: a factorial survey of
perceived justice in criminal sentencing. Journal of Criminal Law and Criminology, 82,
396-422.
Miller, M.L. & Wright, R.F. (2008). The black box. Iowa Law Review, 94, 125-96.
O’Toole, R., O’Toole, A.W., Webster, S., & Lucal, B. (1997). Influence of perpetrators’ account on
nurses’ recognition and reporting of child abuse. Journal of Family Violence, 12, 49-62.
Raudenbush, S.W. & Bryk. A.S. (2002). Hierarchical Linear Models: Applications and Data
Analysis Methods (2nd Ed.). California: SAGE Publications.
Raudenbush, S.W., A.S. Bryk, Y.F. Cheong, & Congdon, R.T. (2004). HLM 6: Linear and
Nonlinear Modeling. Lincolnwood, IL: Scientific Software International, Inc.
Rawls, J. (1999) A theory of justice (revised edition). Cambridge, MA: Belknap/Harvard.
Rossi, P.H. & Nock, S.S. (eds.) 1982. Measuring Social Judgments: The Factorial Survey
Approach. Beverly Hills, CA: Sage.
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 289

Rossi, P. & Anderson, A.B. (1982). The factorial survey approach: An introduction. In Rossi, Peter
H. and Steven S. Nock (eds.) 1982. Measuring Social Judgments: The Factorial Survey
Approach. Beverly Hills, CA: Sage.
Schmidt, J. & Hochstedler Steury, E. (1989). Prosecutorial discretion in filing charges in domestic
violence cases. Criminology, 27, 487-509.
Spears, J.W. & Spohn, C.C. (1997). The effect of evidence factors and victim characteristics on
prosecutors’ charging decisions in sexual assault cases. Justice Quarterly, 14, 501-524
Spohn, C.C. & Holleran, D. (2001). Prosecuting sexual assault: a comparison of charging decisions
in sexual assault cases involving strangers, acquaintances, and intimate partners. Justice
Quarterly, 18, 651-88.
Stanko, E.A. (1981-1982). The impact of victim assessment on prosecutors’ screening decisions:
the case of the New York County district attorney’s office. Law and Society Review, 16,
225-39.
Steffensmeier, D. & Demuth, D. (2000). “Ethnicity and sentencing outcomes in U.S. Federal courts:
who is punished more harshly? American Sociological Review, 65, 705-729.
Steffensmeier, D., Ulmer, J., & Kramer, J. (1998). The interaction of race, gender, and age in
criminal sentencing: The punishment cost of being young, Black, and male. Criminology,
36, 763-797.
Taylor, B. J. (2006). Factorial surveys: Using vignettes to study professional judgment. British
Journal of Social Work, 36, 1187–1207.
Thurman, Q.C. (1986). Estimating social-psychological effects in decisions to drink and drive: a
factorial survey approach. Journal of Studies on Alcohol, 47, 447-454.
Thurman, Q.C. (1987). Modeling Judgments of Taxpayer Noncompliance. Ph.D. dissertation, The
University of Massachusetts, Amherst.
Thurman, Q.C., Jackson, S., & Zhao, J. (1993). drunk driving research and innovation: a factorial
survey study of decisions to drink and drive. Social Science Research, 22, 245-264.
Tyler, T.R. (2003). Procedural justice, legitimacy, and the effective rule of law. Crime and Justice,
30, 283-357.
Tyler, T.R. & Huo, Y.J. (2002). Trust in the Law: Encouraging Public Cooperation with the Police
and Courts. New York: Russell Sage Foundation.
Ulmer, J. T. (1997). Social Worlds of Sentencing: Court Communities under Sentencing Guidelines.
Albany: State University of New York Press.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 290

Ulmer, J.T., Kurlycheck, M.C., & Kramer, J.H. (2007). Prosecutorial discretion and the imposition
of mandatory minimum sentences. Journal of Research in Crime and Delinquency, 44, 427458.
Wallander, L. & Blomqvist, J. (2009). Modeling ideal treatment recommendations: a factorial
survey of Swedish social workers' ideal recommendations of inpatient or outpatient
treatment for problem substance users. Journal of Social Science Research, 35, 47-64.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 291

Dissemination of Research Findings
Presentations
“Prosecutorial Decision-Making: A Multi-Site Study of the Factors Affecting Prosecutors’
Decisions,” Don Stemen, American Society of Criminology Annual Meeting. San Francisco, CA,
November 2010.
“Balancing Strength of the Evidence and Contextual Constraints: A Multi-Method Study of Factors
Affecting Prosecutors’ Decisions,” Bruce Frederick and Don Stemen, American Society of
Criminology Annual Meeting. Washington, DC, November 2011.
“Contextual Constraints on Prosecutorial Discretion: The Effects of Policies, Resources, and
Relationships on Prosecutors’ Decisions,” Don Stemen and Bruce Frederick, American Society of
Criminology Annual Meeting. Washington, DC, November 2011
“Variation in the Screening, Charging, and Plea Bargaining Strategies of Prosecutors,” Don Stemen,
American Society of Criminology Annual Meeting. Chicago, IL, November 2012.
Vera Institute Web Site
This technical report, a summary report, and four podcasts are available via the Vera Institute of
Justice web site at www.vera.org.
Reports
Bruce Frederick and Don Stemen, The Anatomy of Discretion: An Analysis of Prosecutorial Decision
Making. New York: Vera Institute of Justice, 2012. Available at http://www.vera.org.
Bruce Frederick and Don Stemen, The Anatomy of Discretion: An Analysis of Prosecutorial Decision
Making—Technical Report. New York: Vera Institute of Justice, 2012. Available at
http://www.vera.org.
Podcasts

“The Anatomy of Discretion, Part I of IV, An Interview with Don Stemen, PhD, Assistant
Professor, Department of Criminal Justice, Loyola University Chicago.” Vera Voices [Video
podcast series]. Available at http://www.vera.org.
“The Anatomy of Discretion, Part II of IV, An Interview with Anne J. Swern, First Assistant
District Attorney, Kings County, NY, District Attorney’s Office.” Vera Voices [Video podcast
series]. Available at http://www.vera.org.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 292

“The Anatomy of Discretion, Part III of IV, An Interview with Judge Theodore A. McKee, U.S.
Court of Appeals for the Third Circuit.” Vera Voices [Video podcast series]. Available at
http://www.vera.org.
“The Anatomy of Discretion, Part IV of IV, An Interview with Anthony Thompson, Professor of
Clinical Law, NYU School of Law.” Vera Voices [Video podcast series]. Available at
http://www.vera.org.
Journal Articles
Don Stemen and Bruce Frederick. Rule, Resources, and Relationships: Contextual Constraints on
Prosecutorial Decision Making. Scheduled for publication in the Quinnepiac Law Review, November,
2012.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 293

Appendices

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 294

Appendix A: Methods
Much of the material presented in this appendix also appears in various locations throughout the
body of the text as introductory explanations for the associated analyses. It is reproduced here to
provide a methods overview in a single location, with certain supplemental material added for
additional explanation.
As noted throughout the report, this study used a multi-method approach, relying on both
quantitative and qualitative methods to examine prosecutorial decision making in Northern and
Southern Counties. These included: two waves of interviews and focus groups, a general survey, a
factorial survey, and an analyses of administrative data (Figure 1-1).
.
Figure A-1 Overview of Research Methods

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 295

A.1 Interviews and Focus Groups
Two waves of interviews and focus group sessions were conducted in each of the research sites.
Individual interviews were conducted with the district attorney and deputy district attorneys
(deputies) in each site; focus group interviews were conducted with ADAs and unit managers. The
first wave of interviews and focus groups focused primarily on contextual conditions and
circumstances that influence decision making: goals of prosecution and guiding philosophies;
formal and informal policies and practices; relationships with police, defense attorneys, and judges;
relationships with colleagues within the prosecutor’s office; resource constraints and the need for
efficiency; and processes that promote adherence to policy and consistency in decision making,
such as training, supervision, mentoring, and informal communication. The second wave of
interviews and focus groups focused primarily on case-specific factors that influence decision
making: strength of evidence, seriousness of the instant offense, defendant’s criminal history, and
special aggravating or mitigating circumstances. The same broad questioning guides were used for
individual interviews and focus groups in both counties; however, the order and emphasis of
specific topics were allowed to vary from one session to another, according to the dynamics of
group discussion and the priorities of individual interviewees.
In Southern County, individual interviews were conducted with the DA and the deputy in both
the first and second waves. Focus group participants in Southern County were divided by length of
experience in order to reduce the likelihood that junior level prosecutors might be unduly reticent in
the presence of their supervisors or other more experienced prosecutors. These groups included
those prosecutors with one year or less experience as a prosecutor, those with more than one year
but less than 10 years experience as a prosecutor, and unit managers. Efforts were made to include
the same prosecutors in both the first and second waves of focus groups. Among prosecutors with

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 296

one year or less of experience, the same six individuals participated in both the first and second
waves. Among those with more than a year and up to ten years of experience, eight individuals
participated in the first wave and five participated in the second wave. Seven unit managers
participated in the first wave focus group and five of these seven participated in the second wave
focus group.
In Northern County, the first wave included individual interviews with the DA and a deputy in
charge of administration, as well as a combined interview with two deputies. There were not
enough participants in Northern County to divide focus groups by length of experience. Instead,
first wave focus groups were conducted with five unit managers and five ADAs responsible for
prosecuting a variety of different types of cases (general crimes, domestic violence, drugs,
weapons). The second wave focus groups were conducted with seven unit managers and five
ADAs, again responsible for prosecuting a variety of different types of cases. No individual
interviews were conducted in the second wave in Northern County.
The feedback received from prosecutors was recorded as field notes. To analyze these data we
classified field notes by type of respondent, topic, and site. Notes were analyzed across topics and
sites, noting the clustering of responses around specific issues or actors, as well as outliers and other
unique data. Through iteration, we developed a number of substantive themes, some of which
coincided with those highlighted by our qualitative instruments others which emerged from the
interviews themselves.

A.2 General Survey
Prosecutors in both participating jurisdictions were invited to respond to a forced-choice, paper
and pencil survey. Whereas the focus group sessions allowed for clarification and in-depth
exploration of beliefs and opinions volunteered by the participants, the general survey permitted
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 297

researchers to elicit structured responses to a broader array of questions and statements, and to
examine the prevalence of ideas expressed in the focus groups among a broader sampling of
prosecutors.
The general survey incorporated a total of 76 items, organized in eight substantive categories: 1)
factors that define professional success for individual prosecutors (15 items); 2) factors that define
success for the district attorney’s office (19 items); 3) the influence of relationships among
prosecutors, police, defense attorneys, and judges on decision making (10 items); 4) resource and
policy constraints (13 items); 5) principles that guide screening decisions (6 items); 6) principles
that guide the development of plea offers (8 items); 7) general goals and functions of the criminal
justice system (9 items); and 8) training and oversight (16 items). Responses to all but two of the
items were ratings on five-point scales that reflected either perceived frequency of occurrence,
degree of agreement, or importance, as appropriate. Responses to the remaining two items involved
categorical choices rather than quantitative ratings. The survey instrument was accompanied by a
background questionnaire that captured age, race, ethnicity, gender, and length of respondents’
experience as defense attorneys and prosecutors. Copies of the complete survey instrument, the
instructions to respondents, and the background questionnaire are included in Appendix D (General
Survey Instrument).
In Southern County, responses were received from 74 out of a total of 78 prosecutors for an
overall response rate of 95 percent. Excluding the district attorney, the deputy district attorneys,
and juvenile court prosecutors, the sample analyzed for this report included 65 responses from a
pool of 69 prosecutors for a response rate of 93 percent. In Northern County, responses were
received from 81 out of a total of 135 prosecutors for an overall response rate of 60 percent.
Excluding the district attorney, the deputy district attorneys, and juvenile court prosecutors, the

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 298

sample analyzed for this report included 62 responses from a pool of 93 prosecutors, for a response
rate of 67 percent.
Preliminary analyses found substantial variation among respondents in the average levels and
variability of their ratings. Some respondents tended to select ratings toward the upper end of the
scale for most items (“yea-sayers”). Some others tended to select ratings toward the lower end of
the scale for most items (“nay-sayers”). As a result, for example, a rating of 3 on a 5-item scale
could be considered a low rating for some respondents but a high rating for other respondents. In
addition, due to floor and ceiling effects, respondents who consistently selected high ratings or low
ratings also produced ratings with restricted variability.
To examine the consequences of these response biases, standardized responses (z-scores) were
created for each item, relative to the personal means and standard deviations of each respondent’s
ratings for items within each of the item categories listed above. The resulting z-scores were then
grouped to create a 5-level standardized scale.40 A series of analyses examined the relationships of
each item with other items and with respondent age, race/ethnicity, gender, and length of
prosecutorial experience, comparing results obtained using the original scaling with results obtained
using the standardized scaling. For most items, the substantive interpretation of the findings would
be the same, regardless of which scaling was adopted. Consequently, for ease of interpretation, this
report focuses primarily on the original scaling of responses as structured in the survey instrument,
but de-emphasizes items for which the results of analyses differ according to which scaling is
adopted.

While respondent-level biases can distort the univariate distribution of responses and can attenuate relationships with
other variables, the standardized scaling also has drawbacks. The adjusted responses are purely relativistic, such that a
high rating on the standardized scale simply means that the respondent rated that item higher than that same respondent
tended to rate other items in the same group. Thus, the standardized scaling makes it possible to more reliably “rank”
items and produce more valid estimates of relationships between item responses and other variables, but it does not
yield any information about the absolute level of agreement, perceived frequency of occurrence, or rating of importance.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 299

Finally, principal components analyses were conducted to determine the number of underlying
dimensions of response for each category of items and then identify the one to three specific items
that most strongly represented the underlying dimensions within each category. Table B34 in
Appendix B (Supplemental Tables for General Survey Responses) identifies the underlying
dimensions in each category and lists the specific items most representative of each dimension. In
the presentation of findings, selected items are used to illustrate broader summaries of the findings
from analyses of the general survey responses. The items selected for illustrative purposes are ones
which (a) yield consistent patterns of results regardless of whether analyses are based on the
original scaling or the standardized scaling of responses, and (b) are among the items found to be
most representative of a relevant underlying dimension of variation within one of the eight
categories of items included in the general survey.

A.3 Factorial Survey
A.3.1 Survey Design
The term “factorial survey” refers to an approach in which respondents are asked to make
judgments about a structured set of hypothetical cases or “vignettes.” Using hypothetical cases
makes it possible to systematically vary the characteristics of cases, among cases for each
respondent and across respondents. Thus, structured experimental designs and balanced sampling
plans of varying complexity can be introduced, permitting economical survey administration and
identification of effects that might be difficult to disentangle relying on real-world observations.
This technique has been widely used to research decision making and judgment formation on a
variety of issues, including professional judgment, crime seriousness, sexual harassment,
pregnancy, fear of victimization, ideal substance use treatment recommendations, child abuse and
reporting of child abuse, drinking and driving, social welfare policy, the justice of punishment, and
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 300

police perjury (Applebaum, Lennon, & Aber, 2006; Applegate, 1997; Foley, 2000; Hunter &
McClelland, 1991; Jasso, 1998; Miller, Rossi, & Simpson, 1991; R. O’Toole, A. O’Toole, Webster,
& Lucal, 1997; Rossi & Nock, 1982; Taylor, 2006; Thurman, 1986; Thurman, Jackson, & Zhao,
1993; Wallander & Blomqvist, 2009).
The basic approach involves several steps leading to the construction of a packet or “deck” of
vignettes for each respondent (Jasso, 2006; Rossi & Anderson, 1982; Thurman, 1987). Researchers
first identify the factors hypothesized to influence respondents’ judgments and decide on the
specific values of measures to be used to represent each factor (5 categories of age, for example).
Then a “factorial object universe” is defined, which consists of all of the hypothetical cases formed
by all possible combinations of the characteristics to be studied. The factorial object universe can
be very large, ranging from a few hundred to millions of unique combinations. For example, if a
survey of perceived danger posed by offenders returning from prison involved 5 conviction crimes,
6 levels of statutory class, 10 age categories, 2 categories of gender, 3 categories of race, 2
categories of ethnicity, and a 10-point risk assessment scale, then the number of unique cases in the
factorial object universe would reach 36,000. Then, random samples are drawn from the vignette
population. Vignettes may be sampled so as to obtain a unique sample to present to each
respondent, or a smaller, fixed number of vignettes may be selected so multiple responses can be
obtained to each vignette. For example, Jasso (2006, p. 343) cites a study in which 10 unique sets
of vignettes were developed, each set contained 60 randomly sampled vignettes, and each set was
presented to 20 different respondents. As a result of this process, each respondent receives a set of
vignettes with different combinations of characteristics, but if the number of respondents is large
enough, all of the characteristics of interest can be adequately represented with balanced

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 301

frequencies across combinations of conditions. Standard statistical techniques can then be applied,
though both the sampling design and the analysis can be quite complicated.
Factorial survey methods produce the best coverage of the dimensions of interest when the
number of potential survey respondents is large, typically from several hundred to a few thousand.
For this study, though, the maximum number of potential respondents was limited to the number of
prosecutors in each of the two participating jurisdictions—roughly 125 in Northern County and
roughly 75 in Southern County. In addition, given prosecutor workloads, it was determined that it
would not be feasible to obtain judgments for more than a few vignettes from each prosecutor. This
combination of small samples of respondents and severe limits on the number of vignettes per
respondent resulted in significant adaptation of traditional factorial survey methods, while still
retaining some of its key features.
In this study, each participant was asked to read and respond to 10 vignettes, each of which
described an arrest and the circumstances surrounding the arrest, listed the offenses charged by the
police, provided a list of the evidence presented to the prosecutor by the police, and specified the
age, race, gender, and criminal history of the suspect. Each vignette consisted of a base scenario
(offense and evidence list) and two elements that varied systematically across vignettes – defendant
criminal history and defendant race. All respondents received the same 10 base scenarios (i.e. all
respondents received the same arrest offenses and evidence lists). However, each respondent’s
packet of vignettes was unique with respect to combinations of those base scenarios with the
variable elements of defendant criminal history and defendant race.41
The base scenarios and evidence lists were selected from a standard case set developed and
studied extensively by Joan Jacoby and her colleagues in the 1970s and 1980s (Jacoby et al.,
41

In order to conserve degrees of freedom in the face of a relatively small sample, the race dimension was limited to
black vs. white, and ethnicity was ignored in the construction of vignettes.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 302

1982b). The standard case set includes 30 scenarios. Reducing the number of vignettes from 30 to
10 placed some restrictions on the way case characteristics could be varied. Highest priority was
given to independently varying offense seriousness and strength of evidence. This was
accomplished by ranking the 30 original vignettes by offense seriousness and strength of evidence,
classifying them as low, medium, or high on each dimension, and then selecting a vignette
representing each the 9 combinations of offense and evidence levels. In a few cases, items were
deleted from the existing evidence lists to adjust strength of evidence to the desired level. There
was only one vignette among the original 30 for which the top charge was a drug offense, and that
one was selected as the 10th vignette in the final design.
Strength of evidence was rated as low, medium or high using a scoring procedure based on
some of the elements considered in the research underlying the development of the standard case set
from which the vignettes were drawn (Jacoby et al., 1982b; Mellon, 1980). It was not possible to
fully replicate the scoring procedure applied in the standard case set development, and actual
strength of evidence scores are not reported in the standard case set documentation (Jacoby et. al.,
1982b). An approximation based on a subset of the original items was used in this study.
Neither the initial classification of case seriousness nor the initial classification of strength of
evidence could be objectively definitive. The vignettes in the standard case set were deliberately
developed to be general enough to be applicable across jurisdictions. Therefore, some of the arrest
offenses cited in the vignettes have no specific equivalent in state law in the state in which Southern
County or Northern County is located. In those instances, classification was somewhat subjective,
based on researchers’ judgments as to which statutes were most like the offenses named in the
vignettes. For example, an offense labeled “Felony Criminal Mischief” in the vignettes most likely
corresponds to an offense classified as a misdemeanor in Southern State statutes. In addition,

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 303

because the scoring procedure used in the standard case set development could not be fully
replicated, vignettes may be assigned strength of evidence levels for this study that are different
from those intended in their original development. However, the purpose of this initial
classification was merely to ensure relatively independent variation across offense seriousness and
evidence strength, and data presented later in this section show that was accomplished.
Table A.3-1 Factorial survey vignette descriptions
Strength of Evidence (Design Level)

Seriousness of Top
Arrest Charge

High-level felony

Low
Vignette #1

Medium
Vignette #4

High
Vignette #6

-Burglary
(Residential)
-Assault

-Burglary
(commercial)
-Theft
-Possession of a Stolen
Vehicle

Vignette #2

-Aggravated Assault
with a Deadly
Weapon (a pistol)
-Carrying Concealed
Weapon
-Assault in the Third
Degree (2 counts)
Vignette #5

-Robbery 2
-Conspiracy

-Robbery 1
-Conspiracy

-Forgery (4 counts)
-Theft: under $300 (4
counts)

Low-level felony

Vignette #9

Vignette #7
-Possession of a
Controlled Substance
(Heroin)

Misdemeanor

Vignette # 8

Vignette #3

Vignette #10

-Failure to move on

-Criminal Mischief
-Criminal Trespass

-Possession of a stolen
credit card
-Attempted Illegal Use
of a Credit Card

Only two elements of the vignettes were systematically varied: 10 criminal history scenarios,
and 2 race designations (black vs. white). Suspect and victim genders were not varied, because they
were integral features of some of the scenarios, and varying them without rewriting the scenarios
would produce nonsensical results. Similarly, age and criminal history were confounded, because a
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 304

given list of arrest and conviction dates is only possible for persons of a certain minimum age. All
respondents received the same 10 criminal history scenarios, but the pairing of criminal history lists
with base scenarios was randomized across packets. Then, half of the defendants in each packet
were designated as black and the other half were designated as white, but the assignment of race
categories to vignettes within packets was counterbalanced across vignettes. As a result, each
respondent received a packet that was unique with respect to combinations of the basic scenarios
with age, race, and prior criminal history.
The resulting vignettes were reformatted to resemble a standardized arrest report. Each
hypothetical report was accompanied by a questionnaire, asking prosecutors to indicate whether
they would accept or reject the case for prosecution, give their reasons for rejection if applicable,
specify the charges they would file initially, specify the charges to which the defendant must plead
guilty in connection with a plea bargain, and indicate what sentence recommendation they would
make, if any. The packet also included a questionnaire asking the respondents about some of their
characteristics: age, race, gender, and professional experience in various capacities as attorneys
handling criminal cases. The instructions to respondents, the survey questionnaires, and arrest report
templates for each vignette are included in Appendix E.
A.3.2 Sample
In Southern County, participation in the factorial survey was solicited from 67 prosecutors who
had previously responded to the general survey. Sixty-two prosecutors completed the survey for a
response rate of 93 percent. Each respondent answered questions about 10 different case vignettes,
yielding up to 620 observations for each question. Unfortunately, the number of responses to the
factorial survey received from Northern County was not adequate to reflect the variation of factors
built into the survey design and was not sufficient to support valid statistical analyses of the
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 305

influences of those factors. Only 18 responses were received initially from Northern County, and a
concerted follow-up effort yielded only two additional responses. Consequently, only the factorial
survey data from Southern County were analyzed.
A.3.3 Dependant Variables
Six dependent variables were defined for statistical modeling purposes: whether the respondent
would reject the case at screening (0=accept, 1=reject), the number of filing charges the respondent
listed (continuous), the rank of the statutory class of the top filing charge (ranging from 1=most
serious to 11=least serious); the number of charges to which the plea offer would require a guilty
plea (continuous); the rank of the statutory class of the top plea offer charge (from 1=most serious
to 11=least serious); and whether or not the sentence recommendation would include a period of
incarceration (0=no incarceration recommended, 1=incarceration recommended). These were
designed to capture respondents’ decisions across the prosecutorial process: screening, charging,
plea offer, and sentence recommendation.
A.3.4 Independent Variables
The three case-level independent variables emphasized in the statistical modeling were
researchers’ classification of the seriousness of the top arrest charges (offense seriousness),
prosecutors’ ratings of the strength of evidence (evidence rating), and the prosecutors’ ratings of the
seriousness of defendant criminal histories (criminal history rating). The design used in the
construction of the factorial survey was intended to elicit variability of responding along each of
these three dimensions and relative independence of responding across dimensions. The intercorrelations among the measures exhibit substantial independence across dimensions. The rank
order correlations of criminal history rating with offense seriousness and evidence rating were nonsignificant and negligible (Spearman’s rho = .007 and .035, respectively). The rank order
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 306

correlation between evidence rating and offense seriousness was statistically significant but small
(rho = .193, p=.000, n=580), not large enough to threaten the ability to estimate independent effects.
Seriousness of the top arrest charges. As noted above, vignettes were selected to reflect three
levels of offense seriousness: low, medium, and high. Specific offense descriptions were translated
into specific offenses within the state codes in which Southern and Northern Counties reside.
Offense seriousness was then translated into a three-part categorization of offense seriousness
corresponding to general offense classifications: misdemeanor, low-level felony, and high-level
felony. As indicated in Table A.3-1, three offenses were coded as misdemeanor offenses, four
offenses were coded as low-level felony offenses, and three offenses were coded as high-level
felony offenses.
Strength of evidence. The strength of the evidence presented in each vignette was measured
two different ways. In designing the survey packets, vignettes were chosen to represent levels of
strength of evidence based on some of the evidentiary characteristics used in the original
development of the standard case set (Mellon, 1980). In addition, respondents were asked to rate
the strength of evidence for each of the vignettes they reviewed.
The results of the researchers’ initial design classification are displayed in Figure A.3-1. By
design, each level of strength of evidence is equally represented across levels of case seriousness,
with one exception: relative to the other conditions, twice as many vignette-packet combinations
reflected medium strength of evidence in low level felony cases. This was because the inclusion of
a drug case as the tenth vignette resulted in two vignettes being assigned to that combination (see
Table A.3-2).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 307

Figure A.3-1 Hypothetical cases, by arrest offense seriousness and researchers’ classification
of strength of evidence
25%
20%

Percent

15%
10%
5%
0%
Misdemeanor

Low level felony

High level felony

Offense seriousness
Low evidence strength

Medium evidence strength

High evidence strength

Figure A.3-2 Hypothetical cases, by arrest seriousness and prosecutors’ ratings of strength of
evidence
25%

20%

Percent

15%
10%
5%
0%
Misdemeanor

Low level felony

High level felony

Offense seriousness
Lowest evidence strength
Average evidence strength
Highest evidence strength

Mid-low evidence strength
Mid-high evidence strength

The more relevant measure for explaining prosecutors’ decisions is the strength of evidence as
rated by the prosecutors themselves. The correlation between the researchers’ original design

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 308

classification of strength of evidence and the prosecutors’ ratings was low (Spearman’s rho = .28, p
= .000). Nevertheless, prosecutors’ ratings were also fairly well-balanced across levels of case
seriousness. Mean ratings on a 5-point scale were 2.91 for misdemeanors, 2.79 for low-level
felonies, and 3.56 for high-level felonies, and the standard deviations within seriousness levels were
1.00, 1.18, and 1.11, respectively—slightly less than a third of their respective means and roughly a
fifth of the range. For all three levels, the distributions of responses were very slightly negatively
skewed, with skewness values of -.15 for misdemeanors, -.14 for low-level felonies, and -.13 for
high-level felonies. Only the prosecutors’ ratings of strength of evidence were incorporated in the
statistical modeling.
Defendant criminal history. Prosecutors also provided ratings of the seriousness of defendants’
criminal histories, as reflected in the randomly assigned lists of prior arrests and convictions
attached to each of the vignettes they reviewed. As intended in the design, this resulted in
considerable variability in criminal history ratings within each case seriousness level (Figure A.33). The means and standard deviations of responses on a 5-point rating scale were quite similar
across levels of offense seriousness: mean = 2.98 and standard deviation = 1.28 for misdemeanors;
2.86 and 1.31 for low-level felonies; and 2.94 and 1.30 for high-level felonies. Skewness differed
slightly across levels of offense seriousness, with slightly negative skew for misdemeanors (-.20)
and high-level felonies (-.11), and slightly positive skew for low-level felonies (.06). At all three
levels of seriousness, the standard deviations of criminal history ratings were slightly greater than
40 percent of their respective means and approximately one-fourth of the range of possible ratings.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 309

Figure A.3-3 Percent of hypothetical cases by offense seriousness and prosecutors’ rating of
defendant criminal history
20%

Percent

15%

10%

0%
Misdemeanor

Low level felony

Lowest history rating
Average history rating
Highest evidence strength

High level felony

Mid-low history rating
Mid-high history rating

Defendant race. By construction, defendant’s race (black vs. white) was counterbalanced across
combinations of case scenarios and criminal history scenarios, so that the numbers of white and
black defendants were equal within each level of case seriousness. A total of 78 respondent packets
were produced, 67 were distributed to the prosecutors in Southeast County who had responded to
the general survey, and 62 prosecutors completed their factorial survey packets. Figure A.3-4
shows that the balance of defendant’s race across levels of case seriousness was maintained among
the completed surveys available for analysis.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 310

Figure A.3-4 Percent of hypothetical cases, by defendant race and offense seriousness
25%

20%

Percent

15%

10%

0%
Misdemeanor

Low level felony

White

High level felony

Black

Preliminary analyses found no relationship between defendant race and any of the dependent
variables examined in analyses of the factorial survey results. Consequently, in order to conserve
degrees of freedom given the relatively small sample of observations available for this study, the
race variable was dropped from consideration and is not included in the statistical models presented
in this report.
Defendant age. Defendant age was a fixed characteristic of each criminal history scenario. Two
defendant ages (27 and 30) occurred in two scenarios each, so only eight different ages (ranging
from 27 to 45) were represented across the 10 criminal history scenarios (Figure A.3-5). It is likely
that defendant age influenced prosecutors’ ratings of the seriousness of defendants’ criminal
histories, and age/criminal history is necessarily treated as a single dimension in the analyses of
factorial survey responses for this study. Along with the associated criminal history scenarios,
defendant ages were adequately represented across the three levels of case seriousness in the 62
survey packets completed by Southeast County prosecutors. However, in the statistical analyses,
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 311

defendant age is represented only indirectly through prosecutors’ ratings of the seriousness of
criminal history (discussed below).
Figure A.3-5 Hypothetical cases, by offense seriousness and defendant’s approximate age at
arrest
10%
9%
8%
7%

Percent

6%
5%
4%
3%
2%
1%
0%
27

Age
Misdemeanor

Low level felony

High level felony

Prosecutor race and gender. Multilevel modeling of the factorial survey data is necessary to
account for the dependencies among observations due to the hierarchical structure of the data—i.e.,
the fact that unique cases are nested within respondents. It also makes it possible to examine the
potential influence of prosecutor characteristics on case-level decisions. Potentially relevant
information about the prosecutors was obtained from two sources: a brief questionnaire
accompanying the factorial survey and their responses to the general survey. The brief questionnaire
provided data on the prosecutors’ age, race, ethnicity, gender, and prosecutorial experience. Their
responses to the general survey provided information about their attitudes and beliefs concerning
the fundamental purposes of the criminal justice system, the goals of prosecution, and the optimum
strategies for achieving prosecutorial objectives. The distribution of race and gender among the 62
prosecutors who responded to the factorial survey is depicted in Figure A.3-6. Respondents were
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 312

about equally divided between males (55 percent) and females (45 percent). Overall, 18 percent
reported nonwhite race or Hispanic ethnicity, and the percentage minority was nearly identical for
males and females considered separately.
Figure A.3-6 Prosecutor race and gender
60

Number of Prosecutors

50
40

30
20
10
0
Male

Female

White/NonHispanic

Minority

Prosecutor experience. A questionnaire that accompanied the factorial survey asked prosecutors
to indicate how long they had worked in the current prosecutor’s office. The distribution of years of
experience in the prosecutor’s office is displayed in Figure A.3-7. Values are rounded to the nearest
whole year, so that, for example, a value of zero corresponds to less than 6 months’ experience, a
value of 1 corresponds to the range of experience from 6 months up to but not including 18 months,
a value of 2 corresponds to the range of experience from 18 months up to but not including 30
months, etc. In Figure 3.4.1-11, experience of 9.5 years or more is displayed as “10 or more,” but
the analyses were actually conducted on the full distribution of years of experience. That
distribution was highly skewed, with 23 percent of the respondents reporting years of experience
approximately evenly distributed across the range from 10 years to 40 years. Because of the

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 313

extreme skew, prosecutors’ length of experience was modeled as log10(years of experience). The
mean of the logarithms was .67, with a standard deviation of .45 and a slight positive skew of .15.
Figure A.3-7 Distribution of prosecutors’ years of experience in the Southern County
prosecutor’s office

Percent of prosecutors

40%

30%

20%

10%

0%
0

10 or
more

Years of experience in current DAs office

Prosecutor attitudes. We were also able to link participants’ responses to the factorial survey with

their responses to the general survey. This made it possible to include information about
prosecutors’ beliefs and attitudes in the analyses of their decisions in the hypothetical cases
presented in the factorial survey. Three brief scales were developed from responses to the general
survey: SEVERITY (sum of 4 items, Cronbach’s alpha = .74); FAIRNESS (sum of 3 items,
Cronbach’s alpha = .64; and CONSISTENCY (sum of 4 items, Cronbach’s alpha = .68). In
addition, responses to several individual general survey items were incorporated in the analyses.
Details of the scales and items are presented in the table below.

Prosecutors’ orientation toward obtaining severe case outcomes was measured using a brief
four-item scale (SEVERITY) derived from their responses to the general survey (see Table A.3-2
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 314

for descriptions of the component items). The scale scores were simply the unweighted sums of
ratings on a 5-level scale of importance across the four component items (Cronbach’s coefficient
alpha=.74). Thus, possible scores ranged from a low of 4 through a high of 20. The scale reflects
the importance prosecutors attached to achieving high rates of guilty pleas to the most serious
charges and high imprisonment rates for serious crimes. Observed scores on the severity scale
ranged from 8 through 20 (Figure A.3-8). Scores were concentrated between 13 and 16, with 10
percent in the range from 8 through 12, and 16 percent in the range from 17 through 20. The mean
score was 15.02 and the standard deviation was 2.33, slightly less than one-fifth of the observed
range. However, there was a noticeable negative skew (skewness = -.39) and a strong clustering of
values at a score of 16.
Figure A.3-8 Distribution of prosecutors’ scores on a 4-item severity scale incorporated in the
factorial survey analyses

Number of Prosecutors

0
8

Prosecutors orientation toward severe outcome for serious crimes

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 315

Table A.3-2 Descriptions of the component items for severity, fairness, and consistency scales
Scales/Items

Description
Q1b: How important for defining your personal success is it to obtain high
imprisonment rates for serious crimes?
Q1g: How important for defining your personal success is it to obtain a high rate of
SEVERITY
guilty pleas to most serious charges?
(sum of 4 items,
Q2b:
How
important for defining office success is it to obtain high imprisonment
alpha = .74)
rates for serious crimes?
Q2g: How important for defining office success is it to obtain a high rate of guilty
pleas to most serious charges?
Q1o: How important for defining your personal success is fair treatment of
defendants?
FAIRNESS
(sum of 3 items, Q2s: How important for defining office success is fair treatment of defendants?
alpha = .64)
Q9e: How much do you agree or disagree that the most important function of the
criminal justice system is protecting the rights of the accused?
Q2k: How important for defining office success is it to obtain similar outcomes for
similar cases within units?
Q2j: How important for defining office success is it to obtain similar outcomes for
CONSISTENCY
similar cases across units?
(sum of 4 items, Q10c: How much do you agree or disagree that, for similar cases, there should be a
alpha = .68)
great deal of consistency across prosecutors in the factors that influence the
decision in a case?
Q10g: How much do you agree or disagree that there needs to be more
communication among staff to ensure consistency of outcomes?
Q3m: How frequently are you willing to adjust your decisions in order to increase
courtroom efficiency?
Q4c: How frequently do you decline or dismiss cases when the amount of time and
effort needed to obtain a conviction exceeds the benefits of the potential
sentence?
Q5a: How important is it for you to examine cases at screening in terms of their plea
bargaining potential, their potential for early disposition, and their potential to
reduce the number of cases in the system?
Q8a: How much do you agree or disagree that a plea offer should include all of the
INDIVIDUAL
charges filed, with an offer to forego additional charges if the offer is
ITEMS
accepted?
Q8d: How much do you agree or disagree that the charging decision should include
the highest charges that could be proven at trial, with the realization that these
may be reduced later through a plea bargain?
Q7-1: Is your general approach to charging best described as “I file all possible
charges”?
Q7-2: Is your general approach to charging best described as “I file only the most
serious charges possible”?
Q7-4: Is your general approach to charging best described as “I file only the charges
I believe the accused should plead guilty to”?

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 316

The fairness scale (FAIRNESS) reflects the importance that prosecutors attached to protecting
the rights of the accused in general and fair treatment of defendants in the prosecution of individual
cases in particular. It was constructed as the unweighted sum of ratings on a 5-level scale of
importance across the three component items (Cronbach’s coefficient alpha = .64; see Table A.3-2
for descriptions of the component items). Thus, possible scores ranged from a low of 3 through a
high of 15. Figure A.3-9 shows the distribution of the observed scale scores for the 62 prosecutors
who completed the factorial survey. The distribution was approximately normal in the range from
10 through 15, but with a noticeable negative skew overall (skewness = -.70). The mean score for
fairness orientation was 12.18, with a standard deviation of 1.63—slightly more than one-fifth the
observed range.
Figure A.3-9 Distribution of prosecutors’ scores on a 3-item fairness scale incorporated in the
factorial survey analyses

Number of prosecutors

0
8

Prosecutors orientation towards fairness

The consistency scale (CONSISTENCY) reflects the importance prosecutors attached to
consistency across prosecutors in the criteria that influence their decisions, as well as consistency in

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 317

decision outcomes, within and across units. It was constructed as the unweighted sum of responses
to four of the general survey items (Cronbach’s coefficient alpha = .68; see Table A.3-2 for
descriptions of the component items). Possible responses ranged from a low of 4 through a high of
20. Figure A.3-10 shows the distribution of the observed consistency scale scores for the 62
prosecutors who completed the factorial survey. Scores were concentrated in the range from 13
through 18, but with 9.7 percent of the prosecutors’ scores falling in the range from 9 through 12
and 6.5 percent with scores of 19 or 20. Overall, the distribution was negatively skewed (skewness
= -.41) with a mean of 15.18 and a standard deviation of 2.39—approximately one-fifth the range of
the observed scores.
Figure A.3-10 Distribution of prosecutors’ scores on a 4 item consistency scale incorporated in
the factorial survey analyses

Percent of prosecturos

0
9

Prosecutors orientation toward consistency

In addition to the multi-item scales described above, several individual items from the general
survey were incorporated in the factorial survey analyses. They reflect respondents’ opinions about
various prosecutorial strategies. The items chosen were ones for which there was sufficient
variability in responses to warrant testing whether these differences in prosecutors’ expressed
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 318

opinions about appropriate strategies might be associated with differences in their decisions in the
set of hypothetical cases. The distributions of responses to five of the items are displayed in Table
A.3-3. The items are paraphrased in Table A.3-3; for the precise wording of the questions and
response options, see the copy of the complete survey instrument in Appendix D.
All of the items yielded ratings on a 5-point Likert scale, but the meanings of the ratings
differed for different items. For Q3m and Q4c, the ratings reflected prosecutors’ judgments as to
frequency of occurrence of the stated decision outcomes. For Q5a, the responses reflected
prosecutors’ ratings of the importance of the stated approach to plea bargaining. For Q8a and Q8d,
the responses reflect the extent of agreement or disagreement with statements about plea bargaining
and charging. Four of the five items in Table A.3-3 yielded mean responses near the center of the
range of possible response options (2.52 to 3.11) and standard deviations approximately one-fifth of
the possible range (2.89 to 1.13). However, item Q8d yielded a mean rating of 4.16, a standard
deviation of .606, and skewness of -.54, representing considerably restricted variability relative to
the other items.
One item from the general survey yielded categorical choices rather than quantitative ratings.
That item, Q7, asked prosecutors which of four statements best described their approach to charging
(see Table A.3-4). Three of the four response choices were dummy coded to create dichotomous
variables (0 vs. 1), labeled all charges (Q7-1), top charges (Q7-2), and should plead (Q7-4) in this
report. Because the resulting variables are dichotomies, their means are simply the proportions
equivalent to the percentages displayed in Table A.3-4, and their standard deviations are fixed
functions of the proportions. Thus, the all charges dichotomy represents slightly greater variability
in responding (s.d. = .50) than the top charges dichotomy (s.d. = .40) or the should plead dichotomy
(s.d. = .40).

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 319

Table A.3-3 Distributions of prosecutors’ responses to general survey items incorporated in
the factorial survey analyses
General Survey Item
Q3m: How frequently are you willing to adjust your decisions in
order to increase courtroom efficiency?
Q4c: How frequently do you decline or dismiss cases when the
amount of time and effort needed to obtain a conviction
exceeds the benefits of the potential sentence?
Q5a: How important is it for you to examine cases at screening in
terms of their plea bargaining potential, their potential for
early disposition, and their potential to reduce the number of
cases in the system?
Q8a: How much do you agree or disagree that a plea offer should
include all of the charges filed, with an offer to forego
additional charges if the offer is accepted?
Q8d: How much do you agree or disagree that the charging
decision should include the highest charges that could be
proven at trial, with the realization that these may be reduced
later through a plea bargain?

Prosecutor’s
Rating
2
3
4

4.8

16.1 45.2 30.6

3.2

%
(N)

%
N
%

17.7 32.3 32.3 16.1
11

3.2

1.6

6.5

1.6

14.5 21.0 38.7 17.7

1
8.1

35.5 33.9 24.2

5
3.2

66.1 25.8
41

Note: Items are paraphrased above. For the precise wording of questions and responses, see
Appendix D for a copy of the general survey instrument.
Table A.3-4 Distribution of prosecutors’ responses to general survey question 7 – “Realizing
that each case is unique, which of the following best describes your general approach to
charging?:”

Item Response
Q7-1: I file all possible charges that encompass the offense or offenses
alleged to have been committed by the accused.
Q7-2: I file only the most serious charges possible given the offense or
offenses alleged to have been committed by the accused.
Q7-3: I file only the charges I believe the accused will plead guilty to
given the offense or offenses alleged to have been committed by the
accused.
Q7-4: I file only the charges I believe the accused should plead guilty to
given the offense or offenses alleged to have been committed by the
accused.

Frequency

Percentage

59.7%

19.4%

1.6%

19.4%

Correlations among prosecutor-level independent variables. With race and gender excluded, 12
of the 14 prosecutor-level measures described above were considered for inclusion in the multi-

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 320

level modeling. These 12 measures proved to be relatively independent of one another. Of the 66
unique rank order correlations among the 12 measures, only 13 were statistically significant, and all
but two of those had absolute values in the range from .27 to .36. These correlations were not
expected to pose multi-collinearity problems in estimating model parameters, because the models
tested never included more than two or three of the prosecutor-level measures at a time. One
measure did exhibit strong correlations with two others: the all charges dichotomy had a correlation
of -.60 with both the top charges dichotomy and the should plead dichotomy. However, that was
simply because the three dichotomies together constituted a nearly exhaustive set (see Table A.3-3);
no more than one of those was ever included in a single model.
A.3.5 Analytic Approach
In the factorial survey design adopted for this study, unique vignettes are nested within
respondents, and variables are measured at both the case level and the respondent level. The
appropriate approach for this design is a multi-level analysis using hierarchical modeling techniques
(Raudenbush & Bryk, 2002). To conduct these analyses, we used the Hierarchical Linear and
Nonlinear Modeling statistical package (HLM6; Raudenbush, Bryk, Cheong, and Congdon, 2004).
As with the initial design, the limited sample size imposed certain limitations on the analysis.
HLM produces its results through iterative estimation of multivariate likelihood functions, and the
estimation algorithms often will not converge on stable results for small samples, unless the models
being tested are relatively simple and the underlying structure of the data is not too complex. This
was exacerbated in the present study by the scaling of the dependent variables. The screening
decision and the incarceration recommendation are dichotomies. The number of filing charges and
number of plea offer charges ideally should be modeled as count variables. The statutory ranks of
top filing charge and top plea offer charge ideally should be modeled as ordered categories.
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 321

However, these all involve nonlinear modeling techniques that require even larger samples for
stable estimation. By keeping our models relatively simple, we were able to estimate multi-level
binary logistic models for the rejection decision and incarceration recommendation, but we were
forced to settle for estimating continuous dependent variable models for the other four outcome
variables. In order to obtain dependable findings with these methods, we undertook a conservative,
step by step approach to the model fitting and interpretation. We are quite confident that the effects
we are reporting are reliable, although it is likely that we may have failed to discern some
potentially informative interactions among case level factors and between case level and respondent
level factors.
Preliminary analyses found no significant relationships between defendant race, prosecutor race,
or prosecutor gender and any of the hypothetical case outcomes examined in the factorial survey
analyses. Therefore, these variables were ultimately excluded from the multi-level modeling to
simplify the models and conserve degrees of freedom in the face of a relatively small sample. The
independent variables retained for analysis were researchers’ classification of nominal case
seriousness (misdemeanor arrest, low level felony arrest, or high level felony arrest), prosecutor’s
rating of the strength of evidence (low, medium, high), prosecutor’s rating of the seriousness of
defendant criminal history, the respondents years of experience in the current prosecutor’s office,
and the measures and items pertaining to respondents’ beliefs and attitudes. As noted above,
defendant age was confounded with defendant criminal history, and defendant gender was
determined by the facts of the case in a given scenario, so neither of those factors was incorporated
in analyses.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 322

A.4 Examining Case Outcomes in Southern County
A.4.1 Data
The analyses of case outcomes in Southern County relied on individual-level administrative data
collected from two sources. Data pertaining to the processing of felony cases assigned to the drug
crimes unit were extracted from an automated case management system maintained by the district
attorney’s office. Data pertaining to felony cases assigned to the person crime unit and property
crime unit were coded manually from paper case files using a structured data entry routine with preestablished response lists for the coded data fields.
Data from both sources identified unique individuals, individual complaints, individual charges
within complaints, and unique “cases”—where a case was defined as a complaint-defendant
combination. The drug unit data tracked the status of individual charges from intake to initial
screening, through acceptance or rejection at screening, lower court processing, grand jury
preparation, grand jury decision, plea offer, felony court processing, and final disposition. The
person unit and property unit data collected from paper case files were less finely articulated,
capturing the point at which a charge first appeared in the file and the latest point at which it
appeared, for initial screening, at some undefined point after acceptance but prior to entry to upper
(felony) court, at the plea offer, and at some undefined point in upper court processing following the
plea offer.
Data from both sources included basic defendant characteristics (age, race, ethnicity, gender),
charge details (offense type and statutory classification), reasons for dismissal, case-level
disposition, statutorily defined prior record level as reported in the plea offer, prosecutors’ sentence
recommendations, and the (coded) identities of the assigned prosecutors. However, neither source
yielded information about the characteristics of the prosecutors.
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 323

While the data coded from paper case files were less precise than the drug unit data with respect
to processing stages, the paper case files yielded additional information not available for the drug
cases, most notably characteristics of the victims, victim-offender relationships, victims’
willingness to testify, and physical evidence cited in the police report.
The drug unit data included information on all closed felony drug cases screened by the office
from May 1, 2007 through July 31, 2009. The person unit and property unit data included all closed
felony cases initially forwarded from the largest urban police agency in Southern County and
screened between January 1, 2007 and June 30, 2007. Homicide cases were excluded during data
collection and cases in which the top charge was for kidnapping or any sex offense were excluded
later during data file preparation.
A.4.2 Sample
Following several steps of data file preparation, the drug unit sample used in analyses included a
total of 4,890 unique complaint-defendant level records (hereafter, cases) and 12,225 unique
charge-level records. The final person unit sample included 508 unique cases and 1,118 unique
charges; and the property unit sample included 658 unique cases and 1,316 unique charges. In
addition, the person unit sample included 963 unique victim records and the property unit sample
included 746 unique victim records.
Most analyses were conducted at the case level. In order to represent charge-level information
and victim-level information at the case level, several different aggregation strategies were adopted.
A charge ranking scheme was adopted, and the top charge overall and the top charge for each of
several specific crime types were identified for each case at each identifiable stage in case
processing. Also computed were the number of charges per case at each processing stage, and two
derived measures: top charge incarceration exposure (the number of months of potential
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 324

incarceration if convicted of the top charge) and aggregate incarceration exposure (the number of
months of potential incarceration if convicted of all active charges in the case). Similar strategies
were adopted to determine the number of victims in each case with each of several victim
characteristics.
A.4.3 Measures
Analyses were conducted for the following outcome measures at the case level: whether a case
was accepted or rejected at screening (0=rejected, 1=accepted); the number of charges filed for each
case accepted for screening (continuous); the statutory rank of the top filing charge (1=most serious,
11=least serious); the number of charges requiring guilty pleas in the formal plea offer (continuous);
the statutory rank of the top plea offer charge (1=most serious, 11=least serious); whether the plea
offer recommended a period of incarceration (0=no incarceration recommendation, 1=incarceration
recommendation); and whether all charges in a case were disposed in favor of the defendant—i.e.,
no conviction in the case (0=not disposed of in favor of defendant, 1=disposed of in favor of
defendant). Analyses were conducted examining factors associated with the changes between
initial filing and plea offer in the number of active charges, the statutory rank of the top charge, the
top charge incarceration exposure, and the aggregate incarceration exposure. In addition, an
analysis of reasons for charge dismissal was conducted at the individual charge level.
Potential explanatory variables incorporated in the analyses of case outcomes typically included:
four measures of defendant characteristics (age, race, gender, and a statutorily defined prior record
level), five measures of case seriousness (number of intake charges, statutory rank of the top intake
charge, whether the top intake charge was robbery, whether the top intake charge was burglary, and
whether there were codefendants in the case); over a dozen measures of type and amount of
physical evidence (for example, total number of items of evidence, number of item of evidence per
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 325

charge, presence of evidence with the potential for forensic identification, and items of evidence
relating to weapons, drugs, and several other categories); and over two dozen measures of victim
characteristics and victim-offender relationships (for example, victim age, race, and gender; total
number of victims, number of victims willing to prosecute, and victims to whom the defendant was
a stranger). These variables were culled from a much larger set of candidate variables in a series of
preliminary analyses. The full set of variables potentially available for these analyses is
documented in the data file codebooks archived with the data files in the National Archive of
Criminal Justice Data.
A.4.4 Analytical Strategy
The influence of the above factors on case outcomes was analyzed using standard multivariate
regression techniques. For most of the outcomes, we estimated standard ordinary least squares
(OLS) linear regression models. For two outcomes reflected in dichotomous measures—the
probability of acceptance at screening and the probability that a plea offer recommended a period of
incarceration—we estimated binary logistic models. Regression analyses were conducted
separately for person unit case, property unit cases, and drug unit cases.
The case characteristics reported in final regression models a small subset of the variables
initially tested. A much larger set of potentially predictive factors and their interactions was tested
for possible inclusion in the models. A complete list of variables tested in the both the linear and
logistic regression analyses is given below. Relatively small samples were available for these
analyses, limiting the number of parameters that could be estimated reliably. In addition, the set of
potential predictors included subsets consisting of alternative measures of the same or similar
constructs, thus presenting a high risk of multicollinearity problems. Consequently, model
development proceeded in a stepwise fashion.
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 326

First, only potential main effects were entered and subjected to backward elimination
processing. At each step, model parameters were examined for signs of problems in model fitting.
If problems were detected, variables likely to be causing the problems were removed and the
analysis was repeated. Once reliable results were achieved, variables with significance values
greater than p = .15 were discarded.42 Then, interactions between robbery and other factors (for the
Person Unit Sample), burglary and other factors (for the Property Unit Sample), and historical
period and other factors (for the combined Drug Unit Sample) were tested in forward stepwise
fashion.
It is well known that stepwise selection and elimination procedures are sensitive to small
changes in the sample data when candidate variables are substantially intercorrelated; slight
differences between samples can lead to differences in the order of selection among correlated
variables. In this study that means, for example, that which among the set of measures of victim
characteristics or which among the set of evidence measures is included in a final model might have
been different as a result of sampling variation. Thus, for example, it is safest to interpret a
significant coefficient for a given victim characteristic as an indication that something about victim
characteristics or circumstances influenced the probability of non-conviction, but not to place too
much emphasis on the particular measure that proved most significant in a given sample.
The values reported for the results of the logistic regression analyses are odds ratios, indicating
the increase or decrease in the odds of non-conviction.43 An odds ratio greater than 1 indicates an

This liberal inclusion criterion follows a suggestion by Hosmer and Lemeshow (2000, p. 118), who cite evidence that
more stringent criteria often exclude important variables from the model.
43
Odds are not probabilities. Odds are related to probabilities by the function odds = p/(1-p). For example, if the
probability of acceptance is .80, the odds of acceptance are 4-to-1 (.80/.20 = 4). If the probability is .50, the odds are 1to-1 (.5/.5 = 1). The magnitude of an effect in the probability metric depends on the reference level. For example, if the
probability of acceptance for some reference group is .20 (odds of .25-to-1) and the probability for some target group is
.50 (odds of 1.0-to-1), then the odds ratio would be 4.0, and the ratio of probabilities would be 2.5. However, if the
probability for the reference group is .50 (odds of 1.0-to-1), an odds ratio of 4.0 would imply a probability of .80 for the
target group, and the ratio of probabilities would only be 1.6.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 327

increase in the odds associate with an increase in the value of the associated predictor, and an odds
ratio less than one indicates a decrease in the odds with an increase in the value of the associated
predictor. For example, an odds ratio of 1.5 would reflect a 50 percent increase in the odds (not the
probability) of non-conviction. An odds ratio of .50 would reflect a 50 percent decrease in the odds
of non-conviction. Unlike linear regression weights, odds ratios combine multiplicatively rather
than additively. For example, if the odds ratio for one predictor were 2.0 and the odds ratio for
another predictor were 3.0, the combined effect of one unit increases in both predictors would be to
increase the odds of non-conviction 6-fold.

A.5 Examining Case Outcomes in Northern County
A.5.1 Data
The analyses of case outcomes in Northern County relied on individual-level administrative data
collected in the case management system of the district attorney’s office. These data identified
unique individuals, charges, and cases and featured a number of substantive data fields: offender
demographics (age, race, gender), victim demographics (age, race, gender), charge characteristics
(arresting agency, arrest offense, charged offense, offense disposition, charge enhancers), and
prosecutor information (prosecutor ID, prosecutor unit). The data included information on all
misdemeanor and felony arrest charges screened by the office between January 1, 2010 and June 20,
2011. These data were employed to generate our study sample and to track charge- and case-level
outcomes.
A.5.2 Sample
The initial sample included 111,704 unique arrest charges. Each charge carried a district
attorney case identifier which linked all charges corresponding to a single case/defendant; in other

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 328

words, all charges with the same district attorney case identifier corresponded to the same unique
case/defendant record. The sample included a total of 77,987 unique cases/defendant records
(hereafter, cases). This study samples was further reduced once covariates were introduced into
analytic models due to missing data for some covariates (listwise deletion was employed). The
final sample included 107,374 unique arrest charges and 74,632 unique cases.
All analyses relied on the examination of cases. Since cases often involved multiple charges, a
procedure was devised to categorize and track cases according to the “controlling offense” at each
stage of the prosecutorial process.
To determine the controlling offense at screening, arrest charges were first ranked by offense
severity according to the state’s misdemeanor and felony classes. Arrest charges were then
classified into six distinct offense types (person, property, drugs, public order/public administration,
weapons, driving under the influence) according to state statutory designations; cases designated by
the district attorney’s office as domestic violence cases were used to create a seventh category of
domestic violence cases. The arrest charge with the highest offense severity in a case was
designated as the controlling offense at screening for analysis purposes. When a case contained two
arrest offenses with the same offense severity but different offense types, offenses were ranked
according to offense type in the following way: person (most severe), domestic violence, weapons,
property, drugs, DUI, public order (least severe); when a case contained two arrest offenses with the
same offense severity and the same offense type, offenses were allowed to randomly select as the
controlling offense.
To determine the controlling offense after screening, a similar strategy was used. All issued
charges (i.e. all charges not rejected at screening) were ranked according to offense severity and
offense type as above; the issued charge with the highest offense severity in a case was designated

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 329

as the controlling offense for analysis purposes. The controlling offense after screening was then
used to examine the dismissal and amendment of cases after filing.

A.5.3 Measures
The analyses of case outcomes focused on measures at three stages of the prosecutorial process:
screening, charging, and trial. At screening, a single outcome measure was used: case accepted for
prosecution. A case was considered accepted for prosecution if any single arrest charge within the
case had a status of “Issued,” “Amended,” “Re-Issued,” or “Revised” in the case management
system. In other words, all arrest charges had to be rejected for a case to be considered rejected for
prosecution, but any single charge accepted was sufficient for a case to be considered accepted for
prosecution. The final screening decision was operationalized as a dichotomous variable
(0=rejected, 1=issued).
At charging, two outcome measures were used to measure the severity and scope of charges
issued: most serious arrest charge issued and number of charges reduced. These measures were
used to measure outcomes only for cases that were issued; in other words, cases rejected for
prosecution at screening were not included in the charging analyses (see Analytical Strategy below).
A case was considered to have the most serious arrest charge issued if the controlling offense at
screening (i.e. the most serious arrest charge screened) had a status of “Issued,” “Amended,” “ReIssued,” or “Revised” in the case management system. This charging decision was operationalized
as a dichotomous variable (0=most serious arrest charge rejected, 1=most serious arrest charge
issued). The study also measured the number of charges issued. For cases involving multiple arrest
charges, a measure of the reduction in the number of charges issued was created; this compared the
number of charges issued to the number of arrest charges screened. If a case has fewer issued
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 330

charges than arrest charges, it was considered to have a charge reduction. The measure was
operationalized as a dichotomous variable (0=number of charges not reduced, 1=number of charges
reduced).
Finally, at trial, two outcome measures were used: case dismissed and most serious issued
charge dismissed. A case was considered dismissed if all issued charges within the case had a
status of “Charge Consolidated Into Another Case,” “Charge Dismissed but Read In,” “Discharged
After Being Found Incompetent,” “Dismissed Before Initial Appearance,” “Dismissed on Courts
own Motion,” “Dismissed on Defendants Motion,” “Dismissed on Prosecutors Motion,” or
“Dismissed Read In” in the case management system. In other words, all issued charges had to be
dismissed or otherwise disposed of without a guilty verdict for a case to be considered dismissed.
The dismissed decision was operationalized as a dichotomous variable (0=not dismissed,
1=dismissed). For cases not dismissed in their entirety (i.e. not all charges were dismissed), the
analyses also examined the processing of the most serious issued charge. A case was considered to
have the most serious issued charge dismissed if the controlling offense after screening (i.e. the
most serious charge issued) had a status of “Charge Consolidated Into Another Case,” “Charge
Dismissed but Read In,” “Discharged After Being Found Incompetent,” “Dismissed Before Initial
Appearance,” “Dismissed on Courts own Motion,” “Dismissed on Defendants Motion,” “Dismissed
on Prosecutors Motion,” or “Dismissed Read In” in the case management system. This decision
was operationalized as a dichotomous variable (0=most serious issued charge not dismissed,
1=most serious issued charge dismissed).
Several variables measuring defendant characteristics, victim characteristics, offense
characteristics, and case characteristics were derived from the administrative data and were used a
predictor variables in outcome analyses.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 331

Defendant characteristics included information on demographic attributes of defendants,
namely, age (in years), race (0= White, 1=Black, 2=Hispanic, 3=Other), gender (0=Female,
1=Male), and custody status (0=Not in custody, 1=In custody). Victim characteristics included
similar information on demographic attributes of victims, namely, age (in years), race (0= White,
1=Black, 2=Hispanic, 3=Other), and gender (0=Female, 1=Male).
Offense characteristics included offense severity, which was coded as a categorical variable
with thirteen categories (0=Ordinance violation (least serious), 13=Class A Felony (most serious))
and number of arrest offenses (0=single arrest offense, 1= multiple arrest offenses). We also
measured the presence of offense enhancers – specific factors that may increase the underlying
sentence for the arrest offense, such as habitual offender status, second or subsequent conviction for
the same offense, presence of a weapon, or repeat domestic abuser – which were flagged by ADAs
at screening; a dichotomous variable was created to operationalize the presence of any one of these
enhancers (0=No enhancer present; 1=Enhancer present). Finally, for drug offenses, the type of
drug was also captured as a categorical variable (0=Cocaine, 1=Heroin, 2=Marijuana, 3=Other).
Case characteristics included arresting agency. Fifty-four separate arresting agencies referred
arrest charges to the district attorney’s office in Northern County; 66 percent of these arrests were
made by the primary municipal police department in the county. Thus, we recoded arresting agency
into a dichotomous variable based on whether the arrest was made by the primary municipal police
department or another arresting agency (0=Other agency; 1=Primary municipal police department).
The data also captured the name of the screening ADA. This was used to determine the ADA’s
gender (0=Female, 1=Male), level of experience (in months) in the district attorney’s office, and
role as a supervisor in the office (0=Not a supervisor; 1=Supervisor); an ADA was coded as a
supervisor if they were either a unit leader, a deputy district attorney, or the district attorney.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 332

Finally, the data captured the prosecution unit within the district attorney’s office that screened the
case; a total of nineteen separate units screened cases – fourteen specialized units that process
specific types of offenses (e.g. felony drug offenses, sex crimes, gun crimes, domestic violence
offenses, etc.) and five general crimes units that process all cases not handled by specialized units.
The fourteen specialized units screened roughly 55 percent of cases during the study period; the five
general crimes teams screened the remaining 45 percent of cases. These units were categorized into
two groups based on whether the unit was specialized or a general crimes unit (0=General crimes
unit; 1=Specialized unit.
Our ability to examine additional covariates of case outcomes was limited by the type and
quality of available administrative data.
A.5.4 Analytical Strategy
The impact of defendant, victim, offense, and case characteristics on case outcomes was
analyzed using standard statistical procedures to examine categorical data in multivariate settings.
Specifically, our baseline estimations rely on a series of logistic regression models to estimate the
effect of these factors on the case outcome measures described. Because specific offense types are
evaluated differently (see Part 2 above), seven separate models were run, one for each offense type
– person, property, drugs, public order, domestic violence, weapons, and DUI. This allowed us to
examine the impact of defendant, victim, offense, and case characteristics on case outcomes varied
by offense type. We further disaggregated the data to examine the impact of these factors on the
three most specific offenses within each offense type. Several tables of descriptive statistics of
cases at each stage of the prosecution process in Northern County are presented in Appendix C.
Because of the nested nature of the data (cases nested within prosecutors) the analyses rely on
hierarchical linear modeling (HLM) procedures designed to account for the nested nature of
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 333

multilevel prosecution data. Cases screened by the same prosecutor are likely to have certain
similarities; as a result, residual errors are likely to be correlated within prosecutors, violating
fundamental error assumptions of OLS regression and resulting in misestimated standard errors.
HLM resolves this problem by incorporating into the statistical model a unique random effect for
each prosecutor (Raudenbush & Bryk, 2002, p. 100). HLM also allows the modeling of the
heterogeneity of regression coefficients that can occur when relationships between individual
characteristics and outcomes vary across aggregate units. For example, the effect of being a
minority defendant may differ across prosecutors. HLM allows for modeling this variation by
allowing both slopes and intercepts to vary across prosecutors.
A two-level hierarchy represents the current data, with individual cases nested within
prosecutors. All variables were centered on their grand means and results reported are based on
unit-specific models using robust standard errors (Raudenbush & Bryk, 2002: pp. 276–280). All
models estimate the probability of a case being accepted for prosecution at screening using HLM 6
(Raudenbush et al., 2004) with the Bernoulli sampling model and a logit link function.
The analysis begins by estimating unconditional models for each offense type – person,
property, drug, public order, domestic violence, weapons, and DUI offenses. These models produce
estimates of the relative amount of screening variation that occurs at the case- and prosecutor-levels
of analysis, providing insights into the relative importance of the prosecutor in screening decisions.
Case-level predictors are then added to the models to assess the degree to which prosecutor-level
variations are accounted for by compositional differences in cases. This provides information on the
extent to which case-level factors vary significantly across prosecutors. Two-level hierarchical
models are then estimated to investigate the direct effects of specific prosecutor-level influences on
screening decisions.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 334

A.6 Summary of Administrative Data Obtained from Northern and Southern Counties
Table A.6-1 Administrative data available for analysis, by participating jurisdiction
Northern
County
Data available for analysis

Felony
person and
property
crimes

Felony
drug
crimes

X
X
X
X
X
X

X
X
X
X

X
X

(All felonies and
misdemeanors)
Decision outcomes
Screening
Charging
Dismissal (charge- and case-level)
Plea offer
Sentence recommendation
Final case disposition
Case characteristics
Arresting agency
Prosecuting attorney
Number of charges
Types of offenses
Seriousness of offenses

Types and amounts of evidence
Codefendants
Reasons for dismissal
Defendant criminal history
Pretrial custody status
Defendant age
Defendant race
Defendant gender
Victim age
Victim race
Victim gender
Victim-offender relationships
a
b

Southern County

X
X
X

X
X
X
X
X
X
X

X
X
X

Multiple measures of seriousness
Multiple measures of evidence, but no information concerning quality or relevance of evidence

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 335

Appendix B: Supplemental Tables for General Survey Responses
Table B1: Response distributions for General Survey Question 1 – “Realizing that each case is unique, please indicate how
important the following outcomes are in how you evaluate your own success as a prosecutor:”
Item

4
43

5
9

1
6.3%

Percentage
2
3
4
16.7%
35.7%
34.1%

5
7.1%

Mean

3.19

1.01

High conviction rates
High imprisonment rates for
serious crime
Low dismissal rates after
charges are filed
Low declination rates
High deferral/diversion rates
for eligible defendants
High success/completion
rates for defendants
deferred/diverted
High rate of guilty pleas to
most serious charge(s)
Quick resolution of easy or
straightforward cases
Low rate of defendant
success on appeal
Good relationships with law
enforcement officers
Victim satisfaction with the
outcome of the case

126

1
8

126

4.8%

10.3%

27.8%

39.7%

17.5%

3.55

1.048

127

10.2%

33.9%

29.1%

18.1%

8.7%

2.81

1.118

118

27.1%

44.9%

24.6%

1.7%

2.06

.86

126

11.9%

25.4%

39.7%

19.0%

4.0%

2.78

1.019

126

16.7%

23.8%

20.6%

25.4%

13.5%

2.95

1.308

126

5.6%

11.9%

30.2%

41.3%

11.1%

3.4

1.021

126

5.6%

12.7%

24.6%

39.7%

17.5%

3.51

1.094

127

8.7%

15.7%

26.8%

30.7%

18.1%

3.34

1.197

127

.8%

3.9%

39.4%

37%

18.9%

3.69

.85

125

0.0%

4.0%

37.6%

20.8%

3.75

.829

Respect of your colleagues

127

.8%

15.7%

39.4%

43.3%

4.24

.801

Respect of your supervisors

126

.8%

1.6%

16.7%

34.1%

46.8%

4.25

.846

127

1.6%

8.7%

46.5%

33.1%

10.2%

3.42

.849

127

100

0.0%

3.9%

17.3%

78.7%

4.75

.519

Good relationships with
defense attorneys
Fair treatment of defendants

2
21

Frequency
3
45

1 = Unimportant; 2 = Of little importance; 3 = Moderately important; 4 = Important; 5 = Very important

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 336

Table B2: Response distributions for General Survey Question 2 – “Please indicate how important you believe the following
outcomes are to the District Attorney in evaluating the overall success of the office in which you currently work:”
Item

2
15

Frequency
3
37

4
61

5
12

1
1.6%

2
11.8%

Percentage
3
4
29.1%
48%

3.52

.881

127
126

1.6%

4.8%

19.8%

51.6%

22.2%

3.88

.864

127

5.5%

12.6%

36.2%

30.7%

15.0%

3.37

1.06

120

13.3%

35.8%

37.5%

11.7%

1.7%

2.53

.925

127

3.1%

11.8%

39.4%

34.6%

11.0%

3.39

.943

125

3.2%

10.4%

34.4%

17.6%

3.53

1.005

125

2.4%

8.0%

28.8%

43.2%

17.6%

3.66

.943

125

3.2%

4.0%

22.4%

33.6%

36.8%

3.97

1.023

Quick resolution of cases

126

1.6%

7.9%

29.4%

38.9%

22.2%

3.72

.952

126

2.4%

7.9%

27.0%

37.3%

25.4%

3.75

1.001

127

.8%

2.4%

22.0%

43.3%

31.5%

4.02

.84

127

2.4%

7.9%

24.4%

33.9%

31.5%

3.84

1.035

127

5.5%

11.8%

28.3%

33.1%

21.3%

3.53

1.119

127

4.7%

1.6%

26.0%

44.9%

22.8%

3.8

.971

125

4.0%

14.4%

24.0%

47.2%

10.4%

3.46

.996

126

.8%

6.3%

20.6%

50.8%

21.4%

3.86

.855

126

.8%

5.6%

42.9%

35.7%

15.1%

3.59

.842

127

.8%

1.6%

23.6%

48.8%

25.2%

3.96

.791

127

0.0%

.8%

7.1%

18.1%

74.0%

4.65

.647

Similar outcomes for similar cases
across units
Similar outcomes for similar cases
within units
Lower crime rates
Fewer defendants re-arrested after
prosecution
High rate of citizen satisfaction with
the DA’s office
Low rate of defendant success on
appeal
Good relations with law
enforcement agencies
Good relations with defense bar
Victim satisfaction with the
handling of cases
Fair treatment of defendants

5
9.4%

Mean

1
2

1 = Unimportant; 2 = Of little importance; 3 = Moderately important; 4 = Important; 5 = Very important

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 337

Table B3: Response distributions for General Survey Question 3 – “Please indicate how frequently you experience the
following situations/feelings in your work:”
Item
I feel pressure from law
enforcement officers to
accept cases for prosecution.
I reluctantly accept cases for
prosecution because of
pressure from law
enforcement officers.
I tailor my decisions to fit the
expectations of judges.
I return cases to law
enforcement for additional
investigation.
I would consider altering my
decisions for defense
attorneys who I respect.
The specific charges I file in a
case are affected by the
judge that will hear the case.
I am willing to adjust my
decisions in order to avoid
injuring relationships with
other agencies.
I lean toward declining cases
from historically “problematic”
units of law enforcement
agencies.
I tailor my decisions to gain
or to maintain the trust of
defense attorneys.
I am willing to adjust my
decisions in order to increase
courtroom efficiency.

Percentage
3

.8%

12.8%

50.4%

28.8%

29.5%

55.7%

14.8%

31.2%

30.4%

120

1.7%

124

126

118

124

126

Frequency
3

Mean

7.2%

3.29

.811

0.0%

1.85

.651

32.0%

5.6%

.8%

2.14

.956

1.7%

28.3%

45.0%

23.3%

3.87

.849

35.5%

36.3%

26.6%

1.6%

0.0%

1.94

.829

70.2%

17.7%

8.9%

1.6%

1.47

.85

31.7%

55.6%

12.7%

0.0%

1.81

.641

27.1%

39.0%

33.1%

.8%

0.0%

2.08

.797

59.7%

25.0%

12.9%

1.6%

.8%

1.59

.836

11.9%

22.2%

46.0%

17.5%

2.4%

2.76

.959

125

122

125

1 = Never; 2 = Very rarely; 3 = Occasionally; 4 = Frequently; 5 = Very frequently

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 338

Table B4: Response distributions for General Survey Question 4 – “Realizing that each case is unique, please indicate how
frequently you find yourself encountering the following circumstances:”
5

Percentage
3

2.4%

9.5%

30.2%

31.7%

46.3%

39.0%

9.8%

26.4%

33.6%

50.4%

119

126

124

127

120

125

Jury consultants are available in high profile cases.

Item

Frequency
3
4

Mean

26.2%

3.7

1.037

2.4%

1.76

.908

28.0%

11.2%

.8%

2.26

1.001

33.9%

14.2%

.8%

0.0%

1.65

.752

21.0%

16.0%

44.5%

14.3%

4.2%

2.65

1.094

20.6%

31.0%

34.9%

9.5%

4.0%

2.45

1.048

58.1%

33.1%

7.3%

1.6%

0.0%

1.52

.704

2.4%

10.2%

27.6%

30.7%

29.1%

3.74

1.063

5.8%

19.2%

38.3%

29.2%

7.5%

3.13

1.004

3.2%

11.2%

36.8%

23.2%

25.6%

3.57

1.088

102

81.2%

10.8%

6.9%

0.0%

1.0%

1.28

.68

The lack of support staff to coordinate contacts with
victims, witnesses, and defense counsel adversely
affects my cases.

123

7.3%

16.3%

30.9%

17.1%

28.5%

3.43

1.262

There is enough clerical staff to assist me with
paperwork on all my cases.

122

36.1%

29.5%

16.4%

15.6%

2.5%

2.19

1.159

High case loads force me to handle cases differently
than I would if I had fewer cases.
Office policies compel me to decline cases that I would
prefer to prosecute.
I decline or dismiss cases when the amount of time and
effort needed to obtain a conviction exceeds the benefits
of the potential sentence.
I make decisions based on how my colleagues perceive
my performance.
A lack of investigators in my office adversely affects the
outcomes of my cases.
The decisions I make are affected by how I think they
will be perceived by my supervisor.
I intentionally make decisions that are inconsistent with
office policies that I disagree with.
Caseloads prevent me from devoting enough time to all
my cases.
I have access to expert witnesses when needed on a
case.
The lack of technology resources in the office makes
work difficult.

126

123

125

126

1 = Never; 2 = Very rarely; 3 = Occasionally; 4 = Frequently; 5 = Very frequently

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 339

Table B5: Response distributions for General Survey Question 5 – “In addition to examining cases in terms of the legal
elements of the alleged offense, please circle the number that best corresponds to how important the following
considerations are in screening a case for prosecution:”
Item
I examine cases at
screening in terms of their
plea bargaining potential,
their potential for early
disposition, and their
potential to reduce the
number of cases in the
system.
I examine cases at
screening to determine the
need for the diversion of
the defendant or the need
for the vigorous
prosecution of the
defendant.
I examine cases at
screening in terms of
convictability and
probability of success at
trial.
I examine cases at
screening to determine the
constitutionality or
evidentiary issues that
may affect the cases later.

Percentage
3

23.3%

30.0%

27.5%

12.5%

4.2%

9.3%

17.8%

.8%

0.0%

.8%

120

118

120

Frequency
3

Mean

6.7%

2.49

1.174

43.2%

25.4%

3.76

1.068

3.3%

27.5%

68.3%

4.63

.636

7.5%

28.3%

63.3%

4.54

.672

1 = Unimportant; 2 = Of little importance; 3 = Moderately important; 4 = Important; 5 = Very important

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 340

Table B6: Response distributions for General Survey Question 6 – “Of the considerations listed below, which is the most
important when screening a case for prosecution?:”
Item
Potential for plea bargaining, early
disposition, and caseload reduction.
Need for diversion or vigorous
prosecution.
Probability of conviction and success at
trial.
Constitutionality and evidentiary issues
that may affect case.

121

Frequency

Percentage

.8%

9.9%

66.9%

22.3%

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 341

Table B7: Response distributions for General Survey Question 7 – “Realizing that each case is unique, which of the following
best describes your general approach to charging?:”
Item
I file all possible charges that
encompass the offense or offenses
alleged to have been committed by the
accused.
I file only the most serious charges
possible given the offense or offenses
alleged to have been committed by the
accused.
I file only the charges I believe the
accused will plead guilty to given the
offense or offenses alleged to have
been committed by the accused.
I file only the charges I believe the
accused should plead guilty to given
the offense or offenses alleged to have
been committed by the accused.

Frequency

Percentage

29.3%

23.3%

.9%

46.6%

116

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 342

Table B8: Response distributions for General Survey Question 8– “Please circle the number that best corresponds to your
level of agreement with the following statements:”
Item
A plea offer should include all of
the charges filed with an officer
to forego additional charges if
the plea is accepted.
A plea offer should include only
the charges the defendant
should plead guilty to with a
threat to add additional charges
if the plea is not accepted.
The first plea offer should be the
only plea offer available with no
negotiations.
The charging decision should
include the highest charges that
could be proven at trial with the
realization that these may be
reduced later through a plea
bargain.
The charges in the plea offer
should be higher than I want the
defendant to ultimately plead to
so I have some room to bargain
down.
Plea bargaining should focus
primarily on the severity of the
most serious charge.
Plea bargaining should focus
primarily on the number of
charges the defendant is facing.
Plea bargaining should focus
primarily on the expected
sentence the defendant is facing.

Frequency
3

Percentage
3

125

4.8%

28.8%

35.2%

24.8%

125

14.4%

32.0%

26.4%

127

18.9%

48.8%

125

4.0%

126

125

126

125

Mean

6.4%

2.99

.996

22.4%

4.8%

2.71

1.113

19.7%

12.6%

0.0%

2.26

.91

12.8%

20.0%

50.4%

12.8%

3.55

1.004

22.2%

44.4%

26.2%

7.1%

0.0%

2.18

.862

.8%

17.6%

26.4%

44.8%

10.4%

3.46

.929

9.5%

44.4%

35.7%

10.3%

0.0%

2.47

.807

4.0%

15.2%

30.4%

42.4%

8.0%

3.35

.969

1=Strongly disagree; 2=Disagree; 3=Neutral; 4=Agree; 5=Strongly agree

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 343

Table B9: Response distributions for General Survey Question 9– “Please indicate how much you generally agree or disagree
with each statement:”
Item
The quick resolution of cases
is a legitimate goal of the
criminal justice system.
The most important function
of the criminal justice system
is to prevent and repress
crime.
Offenders do not need to be
punished in order to be
rehabilitated.
Many offenders currently
imprisoned could be
adequately handled in nonprison sanctions.
The most important function
of the criminal justice system
is protecting the rights of the
accused.
There should be more
constraints on the discretion
of officials in the criminal
justice system.
Sanctioning offenders should
involve punishment rather
than rehabilitation.
The due process rights of
defendants are insufficiently
protected.
Many community-based
programs do not provide
sufficient punishment for
offenders.

Percentage
3

1.6%

4.8%

5.6%

75.4%

4.0%

14.3%

23.8%

15.2%

44.0%

126

5.6%

127

126

127

126

125

Frequency
3

Mean

12.7%

3.93

.718

42.1%

15.9%

3.52

1.049

20.8%

17.6%

2.4%

2.48

1.029

38.9%

29.4%

21.4%

4.8%

2.81

.994

5.5%

26.0%

36.2%

26.8%

5.5%

3.01

.988

14.3%

61.9%

17.5%

6.3%

0.0%

2.16

.742

5.6%

33.3%

42.9%

16.7%

1.6%

2.75

.855

25.2%

63.8%

7.1%

3.1%

.8%

1.91

.718

.8%

18.3%

29.4%

45.2%

6.3%

3.38

.884

1=Strongly disagree; 2=Disagree; 3=Neutral; 4=Agree; 5=Strongly agree

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 344

Table B10: Response distributions for General Survey Question 10– “Please circle the number that best corresponds to your
level of agreement with the following statements:”
Item

Percentage
3

Frequency
3
4
59

6.3%

18.9%

22.8%

46.5%

1.6%

10.4%

22.4%

0.0%

4.0%

0.0%

127

126

125

126

122

126

Mean

5.5%

3.26

1.033

55.2%

10.4%

3.62

.868

9.5%

66.7%

19.8%

4.02

.675

1.6%

6.3%

40.2%

52.0%

4.43

.685

0.0%

7.1%

14.2%

43.3%

35.4%

4.07

.884

3.2%

27.8%

34.1%

32.5%

2.4%

3.03

.912

0.0%

12.0%

28.0%

48.8%

11.2%

3.59

.843

1.6%

16.7%

13.5%

51.6%

16.7%

3.65

.999

19.5%

60.2%

8.1%

11.4%

.8%

2.14

.984

23.0%

30.2%

20.6%

19.8%

6.3%

2.56

1.223

126

4.0%

19.8%

24.6%

44.4%

7.1%

3.31

126

6.3%

26.2%

27.8%

33.3%

6.3%

3.07

1.052

121

14.9%

33.1%

25.6%

22.3%

4.1%

2.68

1.105

126

19.8%

48.4%

23.0%

7.9%

.8%

2.21

.882

It is important to routinely review cases as a group.

124

2.4%

16.1%

21.8%

46.8%

12.9%

3.52

.992

Office policies about how to handle cases are not
consistent with my own beliefs.

126

19.8%

51.6%

20.6%

7.9%

0.0%

2.17

.837

My supervisors let me know how well I am doing on the
job.
My supervisors provide adequate guidance on when to
accept or decline a case for prosecution.
For similar cases, there should be a great deal of
consistency across prosecutors in the factors that
influence the decision on a case.
I often discuss how to handle cases with my colleagues.
Experienced prosecutors are best qualified to screen
cases for prosecution.
Office policies provide clear guidance on how to handle
cases.
There needs to be more communication among staff to
ensure consistency of outcomes.
My supervisors rarely review the plea offers that I make
in cases.
I feel constrained by office policies and practices about
when to accept or decline cases for prosecution.
New prosecutors in my office receive adequate training
before they start independently handling cases.
I often compare my cases to those of my colleagues to
ensure that I am getting similar outcomes.
Office goals and priorities are clearly communicated to
staff.
Prosecutors in my office receive adequate training when
they are transferred to a new unit.
Office priorities require case outcomes that I often
disagree with.

127

125

126

127

1=Strongly disagree; 2=Disagree; 3=Neutral; 4=Agree; 5=Strongly agree

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 345

Table B11: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 1 – “Realizing that each case is unique,
please indicate how important the following outcomes are in how you evaluate your own success as a
prosecutor:”
Item
DF
High conviction rates

Site
F

Sig

Level of experience
F
Sig

Interaction between site and
experience
DF
F
Sig

.282

.596

.925

.431

.525

.167

.918

.075

.785

.683

.564

.066

.978

31.846

.000

2.870

.039

1.087

.358

1.932

.167

8.152

.000

1.505

.217

7.837

.006

3.142

.028

.157

.925

10.120

.002

1.205

.311

.179

.910

.746

.389

1.911

.125

.079

.971

5.601

.020

1.689

.173

1.025

.384

2.395

.124

1.253

.294

.268

.848

3.258

.074

1.933

.128

.402

.752

9.453

.003

.388

.762

.820

.485

Respect of your colleagues

3.830

.053

1.913

.131

.150

.930

Respect of your supervisors

3.841

.052

1.681

.175

.342

.795

Good relationships with defense
attorneys

1.164

.283

.178

.911

.177

.912

Fair treatment of defendants

.353

.553

.952

.418

.744

.528

High imprisonment rates for
serious crime
Low dismissal rates after charges
are filed
Low declination rates
High deferral/diversion rates for
eligible defendants
High success/completion rates for
defendants deferred/diverted
High rate of guilty pleas to most
serious charges
Quick resolution of easy or
straightforward cases
Low rate of defendant success on
appeal
Good relationships with law
enforcement officers
Victim satisfaction with the
outcome of the case

1 = Unimportant; 2 = Of little importance; 3 = Moderately important; 4 = Important; 5 = Very important

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 346

Table B12: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 2 – “Please indicate how important you
believe the following outcomes are to the District Attorney in evaluating the overall success of the office in
which you currently work:”
Item
DF

Site
F

Sig

Level of experience
DF
F
Sig

Interaction between site and
experience
DF
F
Sig

High conviction rates

.078

.780

.716

.544

1.029

.383

High imprisonment rates for serious crime

.091

.763

.203

.894

.497

.685

Low dismissal rates after charges are filed

3.601

.060

1.395

.248

2.261

.085

Low declination rates

.054

.816

4.991

.003

.622

.602

High deferral/diversion rates for eligible defendants

14.350

.000

1.635

.185

.610

High success/completion rates for defendants
deferred/diverted

14.235

.000

1.463

.228

.443

.723

High rate of guilty pleas to most serious charges

5.980

.016

.881

.453

.345

.793

Convictions in high profile cases

.450

.504

.269

.848

.846

.471

Quick resolution of cases

10.539

.002

1.833

.145

1.231

.301

Similar outcomes for similar cases across units

.643

.424

.403

.751

1.024

.384

Similar outcomes for similar cases within units

.387

.535

.500

.683

1.641

.184

Lower crime rates

5.034

.027

.890

.448

2.764

.045

Fewer defendants re-arrested after prosecution

2.551

.113

.457

.713

1.250

.295

High rate of citizen satisfaction with the DA’s office

21.302

.000

.983

.403

.071

.976

Low rate of defendant success on appeal

.994

.321

1.998

.118

.792

.501

Good relations with law enforcement agencies

.052

.820

.911

.438

.600

.616

Good relations with defense bar

.461

.498

.798

.497

1.594

.194

Victim satisfaction with the handling of cases

5.390

.022

.255

.857

2.960

.035

Fair treatment of defendants

.839

.362

2.775

.044

1.013

.389

1 = Unimportant; 2 = Of little importance; 3 = Moderately important; 4 = Important; 5 = Very important

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 347

Table B13: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 3– “Please indicate how frequently you
experience the following situations/feelings in your work:”
Item
DF
I feel pressure from law
enforcement officers to accept
cases for prosecution.
I reluctantly accept cases for
prosecution because of pressure
from law enforcement officers.
I tailor my decisions to fit the
expectations of judges.
I return cases to law enforcement
for additional investigation.
I would consider altering my
decisions for defense attorneys
who I respect.
The specific charges I file in a case
are affected by the judge that will
hear the case.
I am willing to adjust my decisions
in order to avoid injuring
relationships with other agencies.
I lean toward declining cases from
historically “problematic” units of
law enforcement agencies.
I tailor my decisions to gain or to
maintain the trust of defense
attorneys.
I am willing to adjust my decisions
in order to increase courtroom
efficiency.

Site
F

Sig

Level of experience
F
Sig

Interaction between site and
experience
DF
F
Sig

.558

.457

3.364

.021

.210

.890

.078

.781

.524

.667

.221

.882

.000

.997

2.586

.056

.114

.952

.070

.792

4.835

.003

.069

.976

.178

.674

1.232

.301

.136

.938

3.873

.051

.918

.435

1.028

.383

1.621

.205

.886

.451

.817

.487

.989

.322

.448

.719

1.580

.198

.193

.661

2.040

.112

.053

.984

13.946

.000

1.056

.371

.334

.801

1 = Never; 2 = Very rarely; 3 = Occasionally; 4 = Frequently; 5 = Very frequently

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 348

Table B14: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 4– “Realizing that each case is unique,
please indicate how frequently you find yourself encountering the following circumstances:”
Item
DF
High case loads force me to handle cases differently
than I would if I had fewer cases.
Office policies compel me to decline cases that I
would prefer to prosecute.
I decline or dismiss cases when the amount of time
and effort needed to obtain a conviction exceeds the
benefits of the potential sentence.
I make decisions based on how my colleagues
perceive my performance.
A lack of investigators in my office adversely affects
the outcomes of my cases.
The decisions I make are affected by how I think
they will be perceived by my supervisor.
I intentionally make decisions that are inconsistent
with office policies that I disagree with.
Caseloads prevent me from devoting enough time to
all my cases.
I have access to expert witnesses when needed on
a case.
The lack of technology resources in the office makes
work difficult.
Jury consultants are available in high profile cases.
The lack of support staff to coordinate contacts with
victims, witnesses, and defense counsel adversely
affects my cases.
There is enough clerical staff to assist me with
paperwork on all my cases.

Site
F

Sig

Level of experience
DF
F
Sig

Interaction between site and
experience
DF
F
Sig

2.766

.099

.904

.442

.080

.971

3.741

.056

1.477

.225

1.179

.321

5.907

.017

1.026

.384

1.303

.277

.753

.387

5.741

.001

.781

.507

1.607

.208

1.182

.320

2.220

.090

16.946

.000

2.745

.046

2.125

.101

2.849

.094

.335

.800

.228

.876

.282

.597

2.077

.107

.420

.739

.043

.836

1.180

.321

.207

.892

.546

.461

1.951

.125

2.754

.046

.001

.971

3.848

.012

1.945

.128

4.415

.038

1.913

.131

.397

.755

2.354

.128

2.504

.063

.299

.826

1 = Never; 2 = Very rarely; 3 = Occasionally; 4 = Frequently; 5 = Very frequently

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 349

Table B15: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 5– “In addition to examining cases in terms
of the legal elements of the alleged offense, please circle the number that best corresponds to how important
the following considerations are in screening a case for prosecution:”
Sig

3.925

.050

1.959

.124

2.428

.069

1.696

.196

.728

.537

.549

.650

.723

.397

.575

.633

1.217

.307

1.037

0.311

0.024

0.995

0.476

0.699

DF
I examine cases at screening in
terms of their plea bargaining
potential, their potential for early
disposition, and their potential to
reduce the number of cases in the
system.
I examine cases at screening to
determine the need for the
diversion of the defendant or the
need for the vigorous prosecution
of the defendant.
I examine cases at screening in
terms of convictability and
probability of success at trial.
I examine cases at screening to
determine the constitutionality or
evidentiary issues that may affect
the cases later.

Level of experience
F
Sig

Interaction between site and
experience
DF
F
Sig

Site
F

Item

1 = Unimportant; 2 = Of little importance; 3 = Moderately important; 4 = Important; 5 = Very important

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 350

Table B16: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 8– “ Please circle the number that best
corresponds to your level of agreement with the following statements:”
Site
F

Sig

2.254

.136

1.484

.222

.061

.980

8.737

.004

.687

.562

.977

.406

4.924

.028

.477

.699

.596

.619

39.000

.000

.247

.863

.195

.900

.027

.869

.894

.447

.723

.540

1.557

.215

1.394

.248

1.011

.390

.529

.469

2.496

.063

.256

.857

.157

.693

.101

.959

1.466

.227

DF
A plea offer should include all of the
charges filed with an officer to forego
additional charges if the plea is
accepted.
A plea offer should include only the
charges the defendant should plead
guilty to with a threat to add
additional charges if the plea is not
accepted.
The first plea offer should be the only
plea offer available with no
negotiations.
The charging decision should include
the highest charges that could be
proven at trial with the realization that
these may be reduced later through a
plea bargain.
The charges in the plea offer should
be higher than I want the defendant
to ultimately plead to so I have some
room to bargain down.
Plea bargaining should focus
primarily on the severity of the most
serious charge.
Plea bargaining should focus
primarily on the number of charges
the defendant is facing.
Plea bargaining should focus
primarily on the expected sentence
the defendant is facing.

Interaction between site and
experience
DF
F
Sig

Level of experience
F

Item

1=Strongly disagree; 2=Disagree; 3=Neutral; 4=Agree; 5=Strongly agree

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 351

Table B17: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 9– “Please indicate how much you generally
agree or disagree with each statement:”
Item

Sig

3.936

.050

1.708

.169

.110

.954

3.760

.055

2.823

.042

1.144

.335

.469

.495

3.265

.024

.436

.727

1.139

.288

.162

.921

.435

.728

4.980

.028

.616

.606

1.552

.205

.334

.565

4.895

.003

.819

.486

6.534

.012

.432

.730

.671

.571

.704

.403

1.356

.260

.518

.671

6.131

.015

1.902

.133

1.779

.155

DF
The quick resolution of cases is a
legitimate goal of the criminal
justice system.
The most important function of the
criminal justice system is to
prevent and repress crime.
Offenders do not need to be
punished in order to be
rehabilitated.
Many offenders currently
imprisoned could be adequately
handled in non-prison sanctions.
The most important function of the
criminal justice system is
protecting the rights of the
accused.
There should be more constraints
on the discretion of officials in the
criminal justice system.
Sanctioning offenders should
involve punishment rather than
rehabilitation.
The due process rights of
defendants are insufficiently
protected.
Many community-based programs
do not provide sufficient
punishment for offenders.

Level of experience
F
Sig

Interaction between site and
experience
DF
F
Sig

Site
F

1=Strongly disagree; 2=Disagree; 3=Neutral; 4=Agree; 5=Strongly agree

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 352

Table B18: ANOVA Results for Jurisdiction by Prosecutors’ Experience Question 10– “Please indicate how much you
generally agree or disagree with each statement:”
Item
DF

Sig

.046

.830

.339

.797

.997

.397

.067

.797

.698

.555

.928

.430

1.935

.167

.232

.874

2.550

.059

I often discuss how to handle cases with my colleagues.

.462

.498

3.536

.017

1.321

.271

Experienced prosecutors are best qualified to screen cases for
prosecution.

4.057

.046

3.358

.021

1.115

.346

Office policies provide clear guidance on how to handle cases.

6.296

.013

1.640

.184

1.359

.259

There needs to be more communication among staff to ensure
consistency of outcomes.

1.539

.217

.637

.593

1.112

.347

My supervisors rarely review the plea offers that I make in cases.

5.930

.016

1.891

.135

1.707

.169

.119

.730

.440

.725

.409

.747

29.223

.000

2.656

.052

1.316

.272

4.376

.039

.652

.583

1.423

.240

Office goals and priorities are clearly communicated to staff.

2.644

.107

1.063

.368

.232

.874

Prosecutors in my office receive adequate training when they are
transferred to a new unit.

15.953

.000

1.303

.277

1.330

.268

Office priorities require case outcomes that I often disagree with.

.053

.818

.515

.672

.324

.808

It is important to routinely review cases as a group.

21.572

.000

.200

.896

2.476

.065

Office policies about how to handle cases are not consistent with my
own beliefs.

2.211

.140

.866

.461

.410

.746

I feel constrained by office policies and practices about when to accept
or decline cases for prosecution.
New prosecutors in my office receive adequate training before they start
independently handling cases.
I often compare my cases to those of my colleagues to ensure that I am
getting similar outcomes.

Level of experience
DF
F
Sig

Interaction between site
and experience
DF
F
Sig

Site
F

1=Strongly disagree; 2=Disagree; 3=Neutral; 4=Agree; 5=Strongly agree

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 353

Table B19: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 1 – “Realizing that each case is unique,
please indicate how important the following outcomes are in how you evaluate your own success as a
prosecutor:”
< 1 year
Item
High conviction rates
High imprisonment rates for
serious crime
Low dismissal rates after charges
are filed
Low declination rates
High deferral/diversion rates for
eligible defendants
High success/completion rates for
defendants deferred/diverted
High rate of guilty pleas to most
serious charges
Quick resolution of easy or
straightforward cases
Low rate of defendant success on
appeal
Good relationships with law
enforcement officers
Victim satisfaction with the
outcome of the case
Respect of your colleagues
Respect of your supervisors
Good relationships with defense
attorneys
Fair treatment of defendants

Site
Northern
Southern
Northern
Southern
Northern
Southern
Northern
Southern
Northern
Southern
Northern
Southern
Northern
Southern
Northern
Southern
Northern
Southern
Northern
Southern
Northern
Southern
Northern
Southern
Northern
Southern
Northern
Southern
Northern
Southern

Mean
2.86
3.14
3.50
3.57
3.75
2.57
2.75
2.67
3.62
3.14
3.75
2.57
3.38
3.43
3.38
4.00
3.38
3.71
3.88
4.43
4.25
3.43
4.25
4.71
4.29
4.86
3.25
3.71
4.75
4.43

SD
.690
.690
.926
.976
1.035
.535
.707
.516
.916
.690
.886
.976
1.061
.787
1.188
.816
.744
.951
0.835
0.535
.707
.535
.886
.488
.951
.378
1.389
.756
.707
.976

N
7
7
8
7
8
7
8
6
8
7
8
7
8
7
8
7
8
7
8
7
8
7
8
7
7
7
8
7
8
7

At least 1 year but <5

At least 5 years but <10

Mean

3.32
3.38
3.73
3.67
3.64
2.24
2.70
2.00
3.05
2.59
3.50
2.76
3.59
3.72
3.23
4.00
2.95
3.65
3.55
3.78
4.00
3.54
4.18
4.41
4.27
4.41
3.36
3.54
4.64
4.76

SD
1.129
.828
1.077
.956
1.093
.723
1.031
.803
.950
1.013
1.406
1.300
1.221
.741
1.232
.816
1.253
1.184
1.101
0.750
.775
.730
.795
.725
.935
.725
.848
.730
.581
.435

N
22
37
22
36
22
37
20
32
22
37
22
37
22
36
22
37
22
37
22
37
21
37
22
37
22
37
22
37
22
37

2.88
3.13
3.50
3.67
2.63
2.07
1.88
1.67
2.75
2.13
2.87
2.27
3.13
3.47
2.88
3.73
2.75
3.07
3.38
3.87
3.63
3.53
4.13
4.53
4.13
4.40
3.38
3.47
4.88
4.80

SD
1.126
.990
.926
.900
.916
.704
.835
.617
.707
.834
.991
1.163
1.246
.915
1.126
.884
1.165
1.163
0.744
0.990
.744
.990
.835
.640
.835
.737
1.188
.743
.354
.561

N
8
15
8
15
8
15
8
15
8
15
8
15
8
15
8
15
8
15
8
15
8
15
8
15
8
15
8
15
8
15

10 years or more
Mean
3.13
3.00
3.26
3.33
3.48
2.17
1.61
1.67
2.95
2.17
3.27
2.33
2.96
3.17
3.14
3.00
3.43
3.67
3.48
3.50
4.09
3.33
3.78
4.00
3.74
4.17
3.26
3.33
4.83
4.83

1.290
1.095
1.287
1.366
1.163
.753
.499
.816
1.133
.983
1.386
1.211
1.022
1.169
1.167
1.265
1.199
1.633
0.665
0.837
.949
.516
.951
.632
.964
.753
.864
.816
.388
.408

23
6
23
6
23
6
23
6
22
6
22
6
23
6
22
6
23
6
23
6
23
6
23
6
23
6
23
6
23
6

1 = Unimportant; 2 = Of little importance; 3 = Moderately important; 4 = Important; 5 = Very important

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 354

Table B20: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 2 – “Please indicate how important you believe the
following outcomes are to the DA in evaluating the overall success of the office in which you currently work:”
< 1 year
At least 1 year but <5
Mean
SD
N
Mean
SD
N
Northern
3.38
.744
8
3.77
.869
22
High conviction rates
Southern
3.29
.756
7
3.51
.837
37
Northern
3.88
.835
8
4.09
.811
22
High imprisonment rates for serious crime
Southern
4.14
.378
7
3.81
.845
37
Northern
4.00
.926
8
3.86
.941
22
Low dismissal rates after charges are filed
Southern
3.29
.488
7
3.05
1.079
37
Northern
3.38
.916
8
2.80
.951
20
Low declination rates
Southern
3.00
.632
6
2.62
.922
34
Northern
3.50
.756
8
3.95
.899
22
High deferral/diversion rates for eligible defendants
Southern
3.14
.378
7
3.16
.866
37
Northern
4.00
.756
8
4.05
1.090
22
High success/completion rates for defendants
deferred/diverted
Southern
3.14
.900
7
3.38
.982
37
Northern
3.57
1.397
7
3.59
1.098
22
High rate of guilty pleas to most serious charges
Southern
3.86
.690
7
3.95
.848
37
Northern
3.75
1.035
8
4.14
1.283
22
Convictions in high profile cases
Southern
4.29
.756
7
3.92
1.038
37
Northern
4.00
1.069
8
3.05
.999
22
Quick resolution of cases
Southern
4.29
.756
7
4.14
.751
37
Northern
3.75
1.035
8
3.86
1.167
22
Similar outcomes for similar cases across units
Southern
4.14
.378
7
3.59
1.092
37
Northern
4.13
.835
8
4.09
.868
22
Similar outcomes for similar cases within units
Southern
4.14
.378
7
3.84
1.014
37
Northern
4.00
.926
8
4.00
.873
22
Lower crime rates
Southern
4.14
.690
7
3.86
1.032
37
Northern
3.75
1.165
8
3.59
1.260
22
Fewer defendants re-arrested after prosecution
Southern
3.43
.787
7
3.65
1.184
37
Northern
4.00
.535
8
4.14
.834
22
High rate of citizen satisfaction with the DA’s office
Southern
3.29
.488
7
3.24
1.234
37
Northern
3.00
.756
8
3.64
.902
22
Low rate of defendant success on appeal
Southern
3.71
.756
7
3.50
1.108
36
Northern
3.38
1.061
8
4.00
.926
22
Good relations with law enforcement agencies
Southern
3.71
.951
7
3.70
.968
37
Northern
3.25
1.035
8
3.32
1.041
22
Good relations with defense bar
Southern
3.43
.787
7
3.92
.795
37
Northern
3.88
.991
8
3.86
.990
22
Victim satisfaction with the handling of cases
Southern
3.57
.787
7
3.92
.795
37
Northern
4.00
1.069
8
4.50
.802
22
Fair treatment of defendants
Southern
4.43
.787
7
4.78
.479
37
1 = Unimportant; 2 = Of little importance; 3 = Moderately important; 4 = Important; 5 = Very important
Item

Site

At least 5 years but <10
Mean
SD
N
3.13
.835
8
3.67
.976
15
4.00
.756
8
3.73
.884
15
2.75
.707
8
3.27
1.223
15
2.13
.641
8
2.13
.915
15
3.63
.916
8
2.53
.834
15
3.75
1.035
8
2.67
.900
15
3.00
1.195
8
3.80
.775
15
4.00
.926
8
3.60
.910
15
3.25
1.035
8
3.80
.941
15
3.75
.707
8
3.71
1.069
14
3.88
.641
8
4.33
.724
15
4.62
.518
8
3.13
.990
15
4.13
.835
8
3.07
.961
15
4.38
.518
8
3.53
.640
15
2.88
1.246
8
3.13
1.187
15
4.00
.756
8
4.00
.655
15
3.75
.707
8
3.60
.507
15
4.00
.535
8
3.87
.640
15
4.88
.354
8
4.73
.594
15

10 years or more
Mean
SD
N
3.48
1.082 23
3.50
.548
6
3.78
1.126 23
3.83
.753
6
3.65
1.071 23
3.00
.894
6
2.13
.869
23
2.50
.548
6
3.74
.964
23
3.17
.753
6
3.77
.813
22
3.33
.516
6
3.30
.765
23
3.83
.753
6
4.18
.958
22
3.67
.816
6
3.48
.846
23
4.00
.632
6
3.96
.928
23
3.17
.753
6
4.17
.778
23
3.50
.548
6
3.78
1.278 23
3.33
.816
6
3.39
1.234 23
3.17
.408
6
4.48
.511
23
3.50
.548
6
3.65
.714
23
3.67
1.033
6
4.04
.706
23
3.83
.408
6
3.48
.846
23
3.33
.516
6
4.48
.511
23
3.33
.516
6
4.74
.541
23
4.67
.516
6

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 355

Table B21: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 3 – “Please indicate how frequently you
experience the following situations/feelings in your work:”
Item
I feel pressure from law enforcement
officers to accept cases for
prosecution.
I reluctantly accept cases for
prosecution because of pressure from
law enforcement officers.
I tailor my decisions to fit the
expectations of judges.
I return cases to law enforcement for
additional investigation.
I would consider altering my decisions
for defense attorneys who I respect.
The specific charges I file in a case
are affected by the judge that will hear
the case.
I am willing to adjust my decisions in
order to avoid injuring relationships
with other agencies.
I lean toward declining cases from
historically “problematic” units of law
enforcement agencies.
I tailor my decisions to gain or to
maintain the trust of defense
attorneys.
I am willing to adjust my decisions in
order to increase courtroom efficiency.

Site

Mean

< 1 year
SD

At least 1 year but <5
Mean
SD
N

At least 5 years but <10
Mean
SD
N

10 years or more
Mean
SD
N

Northern

2.62

.744

3.50

.859

3.25

.886

3.13

.626

Southern

3.00

.577

3.51

.781

3.33

.900

3.17

.983

Northern

1.88

.354

2.00

.756

1.75

.707

1.68

.568

Southern

1.86

.900

1.91

.678

1.87

.640

1.83

.408

Northern

2.43

.535

2.36

1.002

2.00

.756

1.68

.780

Southern

2.29

.951

2.32

1.107

2.20

.862

1.67

.816

Northern

3.14

.378

4.00

.756

4.13

.641

3.91

.684

Southern

3.00

1.265

3.91

.996

4.07

.730

4.00

.632

Northern

1.88

.641

1.77

.752

2.25

.886

1.91

.868

Southern

1.86

.900

1.89

.854

2.20

.775

2.17

1.169

Northern

1.71

.756

1.86

1.320

1.38

.518

1.65

.775

Southern

1.71

1.254

1.20

.473

1.27

.704

1.00

.000

Northern

2.00

.535

1.64

.658

2.00

.535

1.52

.511

Southern

2.00

.816

1.92

.604

1.93

.704

2.00

.894

Northern

2.00

.756

2.23

.813

2.00

.894

1.76

.889

Southern

1.86

.378

2.03

.770

2.47

.743

2.33

.816

Northern

1.29

.488

1.73

.827

1.75

.707

1.26

.541

Southern

1.43

.535

1.71

.957

1.87

1.187

1.33

.516

Northern

2.25

.707

2.45

.800

2.50

1.195

2.26

.915

Southern

3.14

1.069

3.11

.809

3.40

.910

2.67

1.033

1 = Never; 2 = Very rarely; 3 = Occasionally; 4 = Frequently; 5 = Very frequently

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 356

Table B22: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 4 – “Realizing that each case is unique,
please indicate how frequently you find yourself encountering the following circumstances:”
< 1 year
SD
1.309
1.113
.535
1.512

N
8
7
8
7

1.43

.535

2.18

.907

2.25

1.035

1.91

.848

Southern

2.71

1.113

2.67

.986

2.27

1.100

2.17

1.169

Northern
Southern
Northern
Southern
Northern

1.88
2.43
2.57
1.83
2.00

.991
.976
1.272
.753
.756

8
7
7
6
8

1.68
1.81
2.27
3.00
2.36

.646
.776
.935
1.225
1.002

22
37
22
33
22

1.75
1.60
2.75
2.93
2.00

.463
.828
.886
.799
.756

8
15
8
15
8

1.17
1.17
2.32
3.33
1.78

.388
.408
1.129
.816
.902

23
6
22
6
23

Southern

3.71

1.113

2.76

1.011

2.87

.915

2.17

.753

Northern

1.63

0.518

1.67

0.730

1.63

0.744

1.59

0.908

Southern

1.14

0.378

1.46

0.650

1.53

0.640

1.33

0.816

Northern
Southern
Northern
Southern
Northern
Southern
Northern
Southern
Northern

3.38
3.14
3.00
3.00
3.63
2.57
1.67
2.33
2.88

1.302
1.215
1.195
1.414
.744
1.512
.816
1.155
1.356

8
7
8
6
8
7
6
3
8

3.86
4.05
3.05
2.91
3.32
3.94
1.19
1.28
3.41

.889
.941
.899
1.055
1.041
1.040
.512
.614
1.054

22
37
22
34
22
36
21
25
22

3.88
3.87
3.00
3.27
3.63
4.13
1.63
1.08
3.38

.641
1.246
.756
.961
.744
1.125
1.408
.289
.518

8
15
8
15
8
15
8
12
8

3.30
3.83
3.45
3.50
3.09
3.67
1.19
1.00
2.68

1.185
.983
.945
.837
.921
1.033
.512
.000
1.086

23
6
20
6
22
6
21
6
22

Southern

3.00

1.291

3.86

1.287

4.00

1.512

3.67

1.033

Northern
Southern

2.57
2.17

.976
1.169

7
6

1.95
1.83

1.046
1.056

22
36

2.25
1.93

1.165
1.100

8
15

3.00
2.33

1.195
1.366

22
6

Item

Site

High case loads force me to handle cases
differently than I would if I had fewer cases.
Office policies compel me to decline cases
that I would prefer to prosecute.
I decline or dismiss cases when the amount
of time and effort needed to obtain a
conviction exceeds the benefits of the
potential sentence.
I make decisions based on how my
colleagues perceive my performance.
A lack of investigators in my office adversely
affects the outcomes of my cases.
The decisions I make are affected by how I
think they will be perceived by my
supervisor.
I intentionally make decisions that are
inconsistent with office policies that I
disagree with.
Caseloads prevent me from devoting enough
time to all my cases.
I have access to expert witnesses when
needed on a case.
The lack of technology resources in the
office makes work difficult.
Jury consultants are available in high profile
cases.
The lack of support staff to coordinate
contacts with victims, witnesses, and
defense counsel adversely affects my cases.
There is enough clerical staff to assist me
with paperwork on all my cases

Northern
Southern
Northern
Southern

Mean
3.50
3.71
1.50
2.57

Northern

At least 1 year but <5
Mean
SD
N
3.64
1.002
22
4.11
.854
36
1.73
1.120
22
2.00
.935
33

At least 5 years but <10
Mean
SD
N
3.38
.916
8
3.80
1.146
15
1.50
.535
8
1.67
.617
15

10 years or more
Mean
SD
N
3.35
.982
23
3.67
1.033
6
1.52
.730
23
1.50
.548
6

1 = Never; 2 = Very rarely; 3 = Occasionally; 4 = Frequently; 5 = Very frequently

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 357

Table B23: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 5 – “In addition to examining cases in
terms of the legal elements of the alleged offense, please circle the number that best corresponds to how
important the following considerations are in screening a case for prosecution:”
Item

Site
Mean

I examine cases at screening in
terms of their plea bargaining
potential, their potential for early
disposition, and their potential to
reduce the number of cases in the
system.
I examine cases at screening to
determine the need for the
diversion of the defendant or the
need for the vigorous prosecution
of the defendant.
I examine cases at screening in
terms of convictability and
probability of success at trial.
I examine cases at screening to
determine the constitutionality or
evidentiary issues that may affect
the cases later.

< 1 year
SD

At least 1 year but <5
Mean
SD
N

At least 5 years but <10
Mean
SD
N

10 years or more
Mean
SD
N

Northern

2.00

1.309

2.32

1.041

2.38

1.061

2.17

1.029

Southern

3.50

1.517

3.10

1.221

2.20

.862

2.00

.894

Northern

4.13

.835

3.86

1.037

3.50

1.069

4.14

.640

Southern

4.00

.632

3.53

1.252

3.53

1.302

3.33

1.366

Northern

4.75

.463

4.64

.658

4.38

.744

4.70

.470

Southern

4.33

.516

4.71

.461

4.60

.632

4.33

1.633

Northern

4.50

.756

4.59

.666

4.38

.744

4.43

.662

Southern

4.67

.516

4.52

.769

4.67

.617

4.67

.516

1 = Unimportant; 2 = Of little importance; 3 = Moderately important; 4 = Important; 5 = Very important

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 358

Table B24: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 8 – “ Please circle the number that best
corresponds to your level of agreement with the following statements:”
< 1 year
SD

3.00

1.069

3.32

.945

3.00

1.195

2.90

1.044

Southern

2.57

.535

3.05

1.026

2.80

1.082

2.50

.548

Northern

3.13

1.246

3.09

1.192

3.63

.744

2.86

1.153

Southern

2.71

1.113

2.22

.917

2.40

.910

2.67

1.366

Northern

2.00

.756

2.14

.774

2.25

.463

1.78

.850

Southern

2.43

.787

2.65

1.033

2.27

.884

2.50

.837

Northern

3.13

.641

3.05

1.133

3.13

.641

2.86

.990

Southern

4.14

.378

4.03

.654

4.27

.458

4.17

1.169

Northern

2.13

.835

2.36

.848

2.50

.756

1.77

.813

Southern

2.29

1.380

2.30

.968

2.13

.352

2.17

.408

Northern

3.38

.744

3.50

.913

3.75

1.035

3.55

1.057

Southern

3.43

1.134

3.41

.927

3.67

.617

2.67

1.033

Northern

2.50

.756

2.59

.734

2.63

.916

2.22

.951

Southern

2.43

.535

2.57

.765

2.60

.828

1.83

.753

Northern

3.88

.835

3.41

.959

3.13

1.126

3.18

1.097

Southern

3.00

.816

3.32

1.002

3.60

.828

3.33

.816

Item

Site

A plea offer should include all of the
charges filed with an officer to forego
additional charges if the plea is accepted

Northern

A plea offer should include only the
charges the defendant should plead guilty
to with a threat to add additional charges
if the plea is not accepted
The first plea offer should be the only
plea offer available with no negotiations
The charging decision should include the
highest charges that could be proven at
trial with the realization that these may be
reduced later through a plea bargain
The charges in the plea offer should be
higher than I want the defendant to
ultimately plead to so I have some room
to bargain down
Plea bargaining should focus primarily on
the severity of the most serious charge
Plea bargaining should focus primarily on
the number of charges the defendant is
facing
Pleas bargaining should focus primarily
on the expected sentence the defendant
is facing

Mean

At least 1 year but <5
Mean
SD
N

At least 5 years but <10
Mean
SD
N

10 years or more
Mean
SD
N

1=Strongly disagree; 2=Disagree; 3=Neutral; 4=Agree; 5=Strongly agree

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 359

Table B25: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 9 – “Please indicate how much you
generally agree or disagree with each statement:”
< 1 year
SD

3.38

1.506

3.73

.827

3.75

.886

4.00

.426

Southern

3.71

.756

4.11

.458

4.07

.704

4.17

.408

Northern

3.88

.991

3.64

.848

3.75

.886

3.30

1.329

Southern

3.43

1.134

3.72

.944

3.20

.941

2.50

1.049

Northern

2.75

1.035

2.23

.869

2.25

.886

3.09

1.151

Southern

2.29

.756

2.38

1.114

2.07

.799

3.00

.632

Northern

3.25

.707

2.86

1.125

2.88

1.126

2.82

1.006

Southern

2.57

.976

2.62

.924

2.87

1.060

2.83

.753

Northern

3.25

.707

3.27

.935

3.00

.926

3.26

1.010

Southern

2.86

.900

2.70

.996

3.20

1.082

2.17

.753

Northern

3.00

.926

2.00

.535

2.13

.835

1.96

.562

Southern

2.57

1.134

2.19

.776

2.13

.516

1.83

.753

Northern

2.25

.707

2.50

.740

2.75

1.165

2.50

.913

Southern

3.14

.690

3.03

.763

3.00

.845

2.67

1.033

Northern

2.25

.886

2.14

.990

1.63

.518

1.83

.717

Southern

1.86

.378

1.89

.658

1.73

.458

1.83

.753

Northern

3.00

.926

3.23

1.020

3.88

.991

2.83

.717

Southern

4.00

.577

3.54

.767

3.67

.900

3.50

.548

Item

Site

The quick resolution of cases is a
legitimate goal of the criminal justice
system

Northern

The most important function of the
criminal justice system is to prevent
and repress crime
Offenders do not need to be
punished in order to be rehabilitated
Many offenders currently imprisoned
could be adequately handled in nonprison sanctions
The most important function of the
criminal justice system is protecting
the rights of the accused
There should be more constraints on
the discretion of officials in the
criminal justice system
Sanctioning offenders should involve
punishment rather than rehabilitation
The due process rights of
defendants are insufficiently
protected
Many community-based programs
do not provide sufficient punishment
for offenders

Mean

At least 1 year but <5
Mean
SD
N

At least 5 years but <10
Mean
SD
N

10 years or more
Mean
SD
N

1=Strongly disagree; 2=Disagree; 3=Neutral; 4=Agree; 5=Strongly agree

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 360

Table B26: Mean Responses by Jurisdiction and Prosecutors’ Experience Question 10 – “Please circle the number that best
corresponds to your level of agreement with the following statements:”
< 1 year
SD
0.916
1.633
.641
1.254
.756

N
8
7
8
7
8

4.14

.378

3.97

.726

4.47

.516

3.83

.983

Northern
Southern
Northern
Southern
Northern
Southern

4.38
4.29
3.25
4.00
3.13
3.57

.744
.488
1.035
.816
.641
.535

8
7
8
7
8
7

4.73
4.46
3.64
4.22
2.50
3.32

.550
.605
1.093
.787
.802
.852

22
37
22
37
22
37

4.25
4.60
4.38
4.27
2.38
3.07

1.035
.507
.518
.594
.916
.884

8
15
8
15
8
15

4.22
3.83
4.26
4.50
3.09
3.00

.795
.753
.864
.548
1.041
.894

23
6
23
6
23
6

Item

Site

Northern
Southern
Northern
Southern
Northern

Mean
3.38
3.00
4.13
3.71
4.00

Southern

Experienced prosecutors are best qualified to screen
cases for prosecution.
Office policies provide clear guidance on how to handle
cases.
There needs to be more communication among staff to
ensure consistency of outcomes.
My supervisors rarely review the plea offers that I make
in cases.
I feel constrained by office policies and practices about
when to accept or decline cases for prosecution.
New prosecutors in my office receive adequate training
before they start independently handling cases.
I often compare my cases to those of my colleagues to
ensure that I am getting similar outcomes.
Office goals and priorities are clearly communicated to
staff.
Prosecutors in my office receive adequate training
when they are transferred to a new unit.
Office priorities require case outcomes that I often
disagree with.
It is important to routinely review cases as a group.
Office policies about how to handle cases are not
consistent with my own beliefs.

At least 1 year but <5
Mean
SD
N
3.32
1.086
22
3.32
0.915
37
3.50
1.102
22
3.66
.802
35
3.91
.610
22

At least 5 years but <10
Mean
SD
N
3.25
1.282
8
3.07
1.1
15
3.75
.707
8
3.67
.816
15
3.62
.916
8

10 years or more
Mean
SD
N
3.09
1.041 23
3.83
0.408
6
3.30
.765
23
3.83
.408
6
4.09
.515
23

Northern

3.50

.756

3.45

.912

3.63

.744

4.00

.756

Southern
Northern
Southern

3.14
4.00
4.14

.690
.926
.690

7
8
7

3.54
3.82
3.59

.767
.853
1.013

37
22
37

3.67
3.88
2.87

1.113
.991
1.125

15
8
15

3.33
3.91
3.00

.816
.949
.632

6
23
6

Northern

2.13

.835

2.32

1.041

2.00

1.069

1.91

.684

Southern
Northern
Southern
Northern
Southern
Northern
Southern

2.14
2.38
4.00
2.63
3.43
3.38
3.57

1.464
1.188
1.414
.916
.787
.744
1.272

7
8
7
8
7
8
7

2.15
1.59
3.08
3.23
3.49
2.86
3.19

.784
.734
1.010
1.110
.901
1.125
1.050

34
22
37
22
37
22
37

2.33
1.75
3.07
2.88
3.8
2.50
3.20

.976
1.035
1.100
1.126
0.941
.926
1.082

15
8
15
8
15
8
15

2.00
2.22
2.67
3.22
3
2.91
3.17

.632
1.085
1.366
.998
1.095
.996
1.169

6
23
6
23
6
23
6

Northern

2.71

.951

2.00

.873

1.75

1.035

2.43

.896

Southern
Northern
Southern
Northern
Southern
Northern
Southern

3.50
2.25
2.43
3.25
3.57
2.50
2.00

1.517
.707
.976
.886
.787
.926
1.000

6
8
7
8
7
8
7

3.15
2.18
2.27
3.09
3.95
2.45
2.11

.925
.853
.962
1.231
.705
1.057
.809

34
22
37
22
37
22
37

3.13
2.13
2.33
2.50
4.27
2.00
2.07

1.125
.991
.724
1.069
.594
.756
.594

15
8
15
8
15
8
15

2.67
2.13
1.83
3.10
3.67
2.13
1.83

1.366
.968
.753
.889
.516
.757
.753

6
23
6
21
6
23
6

1=Strongly disagree; 2=Disagree; 3=Neutral; 4=Agree; 5=Strongly agree
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 361

Table B27: Dimensions of Variation in Survey Responses
Question Group

Dimension of Variation

Q1: ADA’s own
success

Differentiated response

(15 items)

Respect and relationships

(4 components
account for 60%
of the variance in
item responses)

Q2: DA office
success
(19 items)
(5 components
account for 61%
of the variance in
item responses)

Severity of outcome
Fairness
Community orientation
Relationships
Severity of outcomes
Crime control
Consistency and Fairness

Q3: Courtroom
workgroup

General workgroup
considerations

(10 items)

Relationships with defense
attorneys
Relationships with judges

(4 components
account for 63%
of the variance in
item responses)

Relationships with law
enforcement

Most Representative Item(s)
(Values in parentheses are component loadings)
Q1e: High deferral/diversion rates for eligible defendants (.80)
Q1f: High success/completion rates for defendants deferred/diverted (.78)
Q1c: Low dismissal rates after charges are filed (.72)
Q1j: Good relationships with law enforcement officers (.75)
Q1n: Good relationships with defense attorneys (.80)
Q1l: Respect of your colleagues (.82)
Q1m: Respect of your supervisors (.82)
Q1a: High conviction rates (.73)
Q1b: High imprisonment rates for serious crimes (.78)
Q1g: High rate of guilty pleas to most serious charge (.73)
Q1o: Fair treatment of defendants (.77)
Q2n: High rate of citizen satisfaction with the DA’s office (.75)
Q2e: High deferral/diversion rates for eligible defendants (.74)
Q2q: Good relations with defense bar (.79)
Q2p: Good relations with law enforcement agencies (.64)
Q2a: High conviction rates (.84)
Q2b: High imprisonment rates for serious crimes (.81)
Q2c: Low dismissal rates after charges are filed (.60)
Q2m: Fewer defendants re-arrested after prosecution (.74)
Q2l: Lower crime rates (.71)
Q2k: Similar outcomes for similar cases within units (.86)
Q2j: Similar outcomes for similar cases across units (.82)
Q2s: Fair treatment of defendants (.53)
Q3m: I am willing to adjust my decisions in order to increase courtroom efficiency (.75)
Q3i: I am willing to adjust my decisions in order to avoid injuring relationships with other
agencies (.70)
Q3f: I would consider altering my decisions for defense attorneys who I respect
Q3l: I tailor my decisions to gain or to maintain the trust of defense attorneys (.46)
Q3h: The specific charges I file in a case are affected by the judge that will hear the
case(.87)
Q3d: I return cases to law enforcement for additional investigation (.76)
Q3a: I feel pressure from law enforcement officers to accept cases for prosecution (.70)
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 362

Question Group

Dimension of Variation

Q4: Resource
and policy
constraints

Case loads

(13 items)

Colleagues and supervisors

(4 components
account for 61%
of the variation in
item responses)

Support staff

Jury consultants

Q5: Sceening
(6 items)
(Not amenable to
component
analysis)

Q8: Plea offers
(8 items)
(3 components
account for 54%
of the variance in
item responses)

Most Representative Item(s)
(Values in parentheses are component loadings)
Q4h: Caseloads prevent me from devoting enough time to all of my cases (.63)
Q4c: I decline or dismiss cases when the amount of time and effort needed to obtain a
conviction exceeds the benefits of the potential sentence (.64)
Q4b: Office policies compel me to decline cases I would prefer to prosecute (.69)
Q4d: I make decisions based on how my colleagues perceive my performance (.83)
Q4f: The decisions I make are affected by how I think they will be perceived by my
supervisor (.82)
Q4l: The lack of support staff to coordinate contacts with victims, witnesses, and defense
counsel adversely affects my cases (.82)
Q4j: The lack of technology resources in the office makes work difficult (.71)
Q4e: A lack of investigators in my office adversely affects the outcomes of my cases (.69)
Q4k: Jury consultants are available in high profile cases (.92)

(weak component: inconsistent relationships with other items)
Q5a: I examine cases at screening in terms of their plea bargaining potential, their potential for early disposition, and their
potential to reduce the number of cases in the system
Q5b: I examine cases at screening to determine the need for the diversion of the defendant or the need for the vigorous
prosecution of the defendant.
Q5c: I examine cases at screening in terms of convictability and probability of success at trial.
Q5d: I examine cases at screening to determine the constitutionality or evidentiary issues that may affect the cases later .
Q6: Of the considerations listed above, which is the most important when screening a case for prosecution?
Q7: Realizing that each case is unique, which of the following best describes your general approach to charging? (See copy of
survey instrument for response options.)
All charges
Q8a: A plea offer should include all of the charges filed with an offer to forego
additional charges if the plea is accepted (.82)
Provable charges
Q8d: The charging decision should include the highest charges that could be
proven at trial with the realization that these may be reduced later through a
plea bargain (.71)
Sentence exposure
Q8h: Plea bargaining should focus primarily on the expected sentence the
defendant is facing (.76)
Q8g: Plea bargaining should focus primarily on the number of charges the
defendant is facing (.66)
Q8f: Plea bargaining should focus primarily on the severity of the most serious
charge (.57)
The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 363

Question Group
Q9: CJ system
goals
(9 items)
(3 components
account for 51%
of the variance in
item responses)
Q10: Training
and oversight
(16 items)
(5 components
account for 63%
of the variance in
item responses)

Dimension of Variation
Punishment orientation
Most important functions

Rights and rehabilitation of defendants
Adequate training

Constrained by policy

Consult with colleagues

Supervision
Consistency

Most Representative Item(s)
(Values in parentheses are component loadings)
Q9g: Sanctioning offenders should involve punishment rather than rehabilitation
(.80)
Q9b: The most important function of the criminal justice system is to prevent and
repress crime (.65)
Q9e: The most important function of the criminal justice system is protecting the
rights of the accused (.52)
Q9h: The due process rights of defendants are insufficiently protected (.76)
Q9c: Offenders do not need to be punished in order to be rehabilitated (.64)
Q10j: New prosecutors in my office receive adequate training before they start
independently handling cases (.91)
Q10m: Prosecutors in my office receive adequate training when they are
transferred to a new unit (.90)
Q10p: Office policies about how to handle cases are not consistent with my own
beliefs (.86)
Q10i: I feel constrained by office policies and practices about when to accept or
decline cases for prosecution (.75)
Q10n: Office priorities require case outcomes that I often disagree with (.74)
Q10o: It is important to routinely review cases as a group (.73)
Q10k: I often compare my cases to those of my colleagues to ensure that I am
getting similar outcomes (.67)
Q10d: I often discuss how to handle cases with my colleagues (.64)
Q10b: My supervisors provide adequate guidance on when to accept or decline a
case for prosecution (.80)
Q10a: My supervisors let me know how well I am doing on the job (.77)
Q10g: There needs to be more communication among staff to ensure consistency
of outcomes (.78)
Q10c: For similar cases, there should be a great deal of consistency across
prosecutors in the factors that influence the decision in a case (.66)

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 364

Appendix C: Supplemental Tables for Case Outcomes in Northern
County
Table C1: Descriptive Statistics for Individual- and Prosecutor-Level Predictors, Person
Cases
Variable
Mean
SD
Min
Max
Dependant variable
N=5,150
Accepted for prosecution
.43
.49
0.00
1.00
Case-level variables
N=5,150
Offense severity
5.91
3.22
0
13
Number of arrest charges
1.34
.84
1
12
Enhancement eligible
.03
.17
0
1
Arrested by primary LEA
.73
.44
0
1
Defendant age (years)
30.74
11.37
13
90
White defendant (reference)
.27
.45
0
1
Black defendant
.66
.47
0
1
Hispanic defendant
.05
.22
0
1
Male defendant
.77
,42
0
1
Defendant in custody
.55
.50
0
1
Victim age (years)
23.78
15.08
1
89
White victim (reference)
.34
.47
0
1
Black victim
.59
.49
0
1
Hispanic victim
.05
.21
0
1
Male victim
.45
.50
0
1
Prosecutor-level variables
N=66
Prosecutor experience (years)
10.07
10.06
.43
36.9
Male prosecutor
.56
.50
0
1
Supervising prosecutor
.17
.38
0
1
Special unit
.55
.50
0
1
Avg. monthly cases screened
33.28
22.42
2.46
145.79
Percent person cases
.21
.24
.02
.90

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 365

Table C2: Descriptive Statistics for Individual- and Prosecutor-Level Predictors, Property
Cases without Victim Characteristics
Variable
Mean
SD
Min
Max
Dependant variable
N=10,457
Accepted for prosecution
.60
.49
0.00
1.00
Case-level variables
N=10,457
Offense severity
5.54
2.47
0
11
Number of arrest charges
1.60
1.40
1
37
Enhancement eligible
.02
.13
0
1
Arrested by primary LEA
.73
.45
0
1
Defendant age (years)
28.78
11.47
13
77
White defendant (reference)
.26
.44
0
1
Black defendant
.69
.46
0
1
Hispanic defendant
.04
.20
0
1
Male defendant
.80
,40
0
1
Defendant in custody
.68
.47
0
1
Prosecutor-level variables
N=77
Prosecutor experience (years)
10.29
10.10
.43
36.9
Male prosecutor
.61
.49
0
1
Supervising prosecutor
.16
.37
0
1
Special unit
.58
.50
0
1
Avg. monthly cases screened
32.88
21.61
3.17
145.79
Percent person cases
.28
.17
.01
.87

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 366

Table C3: Descriptive Statistics for Individual- and Prosecutor-Level Predictors, Property
Cases with Victim Characteristics
Variable
Mean
SD
Min
Max
Dependant variable
N=6,762
Accepted for prosecution
.55
.50
0.00
1.00
Case-level variables
N=6,762
Offense severity
6.17
2.44
0
11
Number of arrest charges
1.62
1.49
1
37
Enhancement eligible
.02
.13
0
1
Arrested by primary LEA
.81
.39
0
1
Defendant age (years)
26.65
11.47
13
75
White defendant (reference)
.23
.42
0
1
Black defendant
.72
.45
0
1
Hispanic defendant
.04
.20
0
1
Male defendant
.84
,37
0
1
Defendant in custody
.71
.45
0
1
Victim age (years)
38.78
15.93
13
92
White victim (reference)
.51
.50
0
1
Black victim
.43
.49
0
1
Hispanic victim
.03
.18
0
1
Male victim
.55
.50
0
1
Prosecutor-level variables
N=77
Prosecutor experience (years)
10.29
10.10
.43
36.9
Male prosecutor
.61
.49
0
1
Supervising prosecutor
.16
.37
0
1
Special unit
.58
.50
0
1
Avg. monthly cases screened
32.88
21.61
3.17
145.79
Percent person cases
.28
.17
.01
.87

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 367

Table C4: Descriptive Statistics for Individual- and Prosecutor-Level Predictors, Drug Cases
Variable
Mean
SD
Min
Max
Dependant variable
N=9,970
Accepted for prosecution
.75
.43
0.00
1.00
Case-level variables
N=9,970
Offense severity
4.46
2.04
0
10
Number of arrest charges
1.58
0.95
1
13
Enhancement eligible
.10
.31
0
1
Arrested by primary LEA
.81
.39
0
1
Defendant age (years)
26.62
9.67
13
74
White defendant (reference)
.26
.42
0
1
Black defendant
.69
.46
0
1
Hispanic defendant
.04
.20
0
1
Male defendant
.88
,32
0
1
Defendant in custody
.84
.37
0
1
Prosecutor-level variables
N=74
Prosecutor experience (years)
8.75
9.26
.39
33.17
Male prosecutor
.61
.49
0
1
Supervising prosecutor
.16
.37
0
1
Special unit
.58
.50
0
1
Avg. monthly cases screened
39.16
40.02
3.13
313.71
Percent person cases
.30
.29
.01
.91

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 368

Table C5: Descriptive Statistics for Individual- and Prosecutor-Level Predictors, Public
Order Cases
Variable
Mean
SD
Min
Max
Dependant variable
N=14,193
Accepted for prosecution
.71
.46
0.00
1.00
Case-level variables
N=14,193
Offense severity
2.29
1.70
0
10
Number of arrest charges
1.38
0.79
1
15
Enhancement eligible
.03
.16
0
1
Arrested by primary LEA
.62
.48
0
1
Defendant age (years)
30.95
10.97
13
89
White defendant (reference)
.34
.42
0
1
Black defendant
.55
.50
0
1
Hispanic defendant
.10
.29
0
1
Male defendant
.80
,40
0
1
Defendant in custody
.41
.49
0
1
Prosecutor-level variables
N=99
Prosecutor experience (years)
10.00
10.14
.39
36.90
Male prosecutor
.57
.50
0
1
Supervising prosecutor
.17
.38
0
1
Special unit
.64
.48
0
1
Avg. monthly cases screened
34.57
35.92
1.62
313.71
Percent person cases
.23
.12
.01
.99

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 369

Table C6: Descriptive Statistics for Individual- and Prosecutor-Level Predictors, Domestic
Violence Cases
Variable
Mean
SD
Min
Max
Dependant variable
N=18,860
Accepted for prosecution
.34
.47
0.00
1.00
Case-level variables
N=18,860
Offense severity
2.87
1.06
0
11
Number of arrest charges
1.21
0.55
1
21
Enhancement eligible
.34
.47
0
1
Arrested by primary LEA
.79
.41
0
1
Defendant age (years)
31.94
10.62
13
89
White defendant (reference)
.26
.42
0
1
Black defendant
.68
.47
0
1
Hispanic defendant
.05
.23
0
1
Male defendant
.84
,36
0
1
Defendant in custody
.50
.50
0
1
Victim age (years)
32.14
11.54
1
91
White victim (reference)
.34
.36
0
1
Black victim
.60
.49
0
1
Hispanic victim
.05
.23
0
1
Male victim
.17
.37
0
1
Prosecutor-level variables
N=62
Prosecutor experience (years)
9.04
10.16
.19
33.17
Male prosecutor
.55
.50
0
1
Supervising prosecutor
.11
.32
0
1
Special unit
.77
.42
0
1
Avg. monthly cases screened
47.35
47.72
3.17
313.71
Percent DV cases
.35
.28
.01
.99

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 370

Table C7: Descriptive Statistics for Individual- and Prosecutor-Level Predictors, Weapons
Cases
Variable
Mean
SD
Min
Max
Dependant variable
N=2,380
Accepted for prosecution
.59
.49
0.00
1.00
Case-level variables
N=2,380
Offense severity
4.76
1.48
0
8
Number of arrest charges
1.82
1.04
1
8
Enhancement eligible
.06
.23
0
1
Arrested by primary LEA
.82
.38
0
1
Defendant age (years)
28.23
10.07
14
74
White defendant (reference)
.16
.23
0
1
Black defendant
.80
.40
0
1
Hispanic defendant
.03
.18
0
1
Male defendant
.91
,28
0
1
Defendant in custody
.80
.40
0
1
Prosecutor-level variables
Prosecutor experience (years)
N=50
8.71
8.65
.43
27.34
Male prosecutor
.68
.47
0
1
Supervising prosecutor
.20
.40
0
1
Special unit
.44
.50
0
1
Avg. monthly cases screened
43.23
44.61
8.18
313.71
Percent Weapons cases
.08
.12
.01
.52

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 371

Appendix D: General Survey Instrument

The Anatomy of Discretion:
Prosecutors’ Survey
Perspectives on Decision Making
(2010)

GENERAL INSTRUCTIONS
The Vera Institute of Justice is a private non-profit organization with a 50-year history of commitment to
improving the justice system through exploratory research and demonstration programming. Vera is
currently conducting a study of prosecutorial decision-making in two county prosecutors’ offices in the
United States. Funding for this project is provided through the National Institute of Justice.
We are interested in understanding how prosecutors make decisions – the information they use to evaluate a
case and the weight they give to different factors in a case. We are also interested in what factors outside of a
case, such as office policies, office practices and procedures such as supervisory review or group-decisionmaking, training, interactions with other justice system actors, and workloads may affect prosecutors’
decisions.
To appreciate how prosecutors think about these influences, we would like you to answer the questions in the
attached survey. It is important that you rely on your own experiences as a prosecutor and respond openly
and honestly.
Your participation is voluntary and is strictly confidential. Your rights as a participant are described in the
informed consent form that we ask you to sign before answering the questionnaire (see attached). While the
District Attorney has been informed about this survey, no one in the District Attorney’s Office will have
access to your answers. Only summary statistics and aggregate responses will be reported. If you decide to
participate, please return both the signed copy of the informed consent form and the completed survey to the
proctors after completion.
Please read each question carefully and circle the number that best corresponds with your answer to the
question. Please circle only one answer per question.
We would like to thank you in advance for your contribution and cooperation with this important project.
Please feel free to contact the Principal Investigators for any additional information on the scope of the
project, our methodology or any other area of interest related to our project:
Don Stemen, Loyola University Chicago, dstemen@luc.edu, (312) 915-7570
Bruce Frederick, Vera Institute of Justice, bfrederick@vera.org, (518) 391-5799

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 372

Section 1: Measuring Success in Prosecution
1. Measuring Individual Success. Below are some general statements about possible measures for
evaluating a prosecutor’s individual success. Realizing that each case is unique, please indicate how
important the following outcomes are in how you evaluate your own success as a prosecutor.
(Circle one number for each item)

Unimportant

Of Little
Importance

Moderately
Important

Important

Very
Important

m Respect of your supervisors

Good relationships with
defense attorneys

Fair treatment of defendants

a
b
c
d
e
f
g
h
i
j
k
l

High conviction rates
High imprisonment rates for
serious crimes
Low dismissal rates after
charges are filed
Low declination rates
High deferral/diversion rates
for eligible defendants
High success/completion rates
for defendants
deferred/diverted
High rate of guilty pleas to
most serious charge(s)
Quick resolution of easy or
straightforward cases
Low rate of defendant success
on appeal
Good relationships with law
enforcement officers
Victim satisfaction with the
outcome of the case
Respect of your colleagues

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 373

2. Measuring Organizational Success. Below are some possible measures for evaluating the
success of an entire prosecutor’s office. For these questions, please indicate how important you
believe the following outcomes are to the District Attorney in evaluating the overall success of the
office in which you currently work. (Circle one number for each item)
Of Little
Unimportant Importance
a
b
c
d
e
f
g

Moderately
Important

Important

Very
Important

Convictions in high profile cases

Quick resolution of cases

j
k
l
m
n
o
p

Similar outcomes for similar cases
across units
Similar outcomes for similar cases
within units
Lower crime rates
Fewer defendants re-arrested after
prosecution
High rate of citizen satisfaction
with the DA’s office
Low rate of defendant success on
appeal
Good relations with law
enforcement agencies

Good relations with defense bar

Victim satisfaction with the
handling of cases

Fair treatment of defendants

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 374

Section 2: Prosecutors and the Courtroom Workgroup
3. Below are some general statements about your experiences working with other actors in the
criminal justice system. Please indicate how frequently you experience the following
situations/feelings in your work. (Circle one number for each item)

a
b
c

I feel pressure from law enforcement
officers to accept cases for
prosecution.
I reluctantly accept cases for
prosecution because of pressure from
law enforcement officers.
I tailor my decisions to fit the
expectations of judges.

I return cases to law enforcement for
additional investigation.
I would consider altering my
f decisions for defense attorneys who I
respect.
The specific charges I file in a case
h are affected by the judge that will
hear the case.
I am willing to adjust my decisions in
i order to avoid injuring relationships
with other agencies.
I lean toward declining cases from
j historically “problematic” units of
law enforcement agencies.
I tailor my decisions to gain or to
l maintain the trust of defense
attorneys.
I am willing to adjust my decisions in
m order to increase courtroom
efficiency.
d

Never

Very
Rarely

Occasionally

Frequently

Very
Frequently

NOTE: Items e, g, and k were deleted in the final instrument. However, the earlier item designations were retained for
the remaining items, and all data files and documentation use the above designations.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 375

Section 3: Prosecutors and Office Resources/Policies
4. Below are some general statements about the availability of resources and the influence of office
policies on your work. Realizing that each case is unique, please indicate how frequently you find
yourself encountering the following circumstances. (Circle one number for each item)

a
b
c
d
e
f
g
h

Never

Very
Rarely

Occasionally

Frequently

Very
Frequently

I have access to expert witnesses when
needed on a case.

The lack of technology resources in the
office makes work difficult.

k
l
m

Jury consultants are available in high
profile cases.
The lack of support staff to coordinate
contacts with victims, witnesses, and
defense counsel adversely affects my
cases.
There is enough clerical staff to assist me
with paperwork on all my cases.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 376

Section 4: The Screening, Charging, and Plea Offer Decisions
5. The following questions pertain to how you screen cases. In addition to examining cases in terms
of the legal elements of the alleged offense, please circle the number that best corresponds to how
important the following considerations are in screening a case for prosecution. (Circle one number
for each item)
I examine cases at screening in
terms of their plea bargaining
potential, their potential for early
a
disposition, and their potential to
reduce the number of cases in
the system.
I examine cases at screening to
determine the need for the diversion of
b
the defendant or the need for the
vigorous prosecution of the defendant.
I examine cases at screening in terms of
c convictability and probability of
success at trial.
I examine cases at screening to
determine the constitutionality or
d
evidentiary issues that may affect the
cases later.

Of Little
Moderately
Very
Unimportant Importance Important Important Important

6. Of the considerations listed above, which is the most important when screening a case for
prosecution? (Circle one number)
1. Potential for plea bargaining, early disposition, and caseload reduction.
2. Need for diversion or vigorous prosecution.
3. Probability of conviction and success at trial.
4. Constitutionality and evidentiary issues that may affect case.
7. Realizing that each case is unique, which of the following best describes your general approach
to charging? (Circle one number)
1. I file all possible charges that encompass the offense or offenses alleged to have
been committed by the accused.
2. I file only the most serious charges possible given the offense or offenses alleged to
have been committed by the accused.
3. I file only the charges I believe the accused will plead guilty to given the offense or
offenses alleged to have been committed by the accused.
4. I file only the charges I believe the accused should plead guilty to given the
offense or offenses alleged to have been committed by the accused.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 377

8. The following questions pertain to the creation of plea offers. Please circle the number that best
corresponds to your level of agreement with the following statements. (Circle one number for each
item)

a
b
c
d

e
f

A plea offer should include all of the charges filed
with an offer to forego additional charges if the plea
is accepted.
A plea offer should include only the charges the
defendant should plead guilty to with a threat to add
additional charges if the plea is not accepted.
The first plea offer should be the only plea offer
available with no negotiations.
The charging decision should include the highest
charges that could be proven at trial with the
realization that these may be reduced later through a
plea bargain.
The charges in the plea offer should be higher than I
want the defendant to ultimately plead to so I have
some room to bargain down.
Plea bargaining should focus primarily on the
severity of the most serious charge.

Strongly
Disagree

Disagree

Neutral

Agree

Strongly
Agree

Plea bargaining should focus primarily on the
number of charges the defendant is facing.

Plea bargaining should focus primarily on the
expected sentence the defendant is facing.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 378

Section 5: Goals of the Criminal Justice System
9. Below are some general statements about criminal justice and the types of state and community
interventions which might be required in response to crime. Please indicate how much you
generally agree or disagree with each statement. (Circle one number for each item).
Strongly
Disagree

Disagree

Neutral

Agree

Strongly
Agree

The quick resolution of cases is a legitimate goal of
the criminal justice system.

The most important function of the criminal justice
system is to prevent and repress crime.

Offenders do not need to be punished in order to be
rehabilitated.

d
e
f

Many offenders currently imprisoned could be
adequately handled in non-prison sanctions.
The most important function of the criminal justice
system is protecting the rights of the accused.
There should be more constraints on the discretion
of officials in the criminal justice system.

Sanctioning offenders should involve punishment
rather than rehabilitation.

The due process rights of defendants are
insufficiently protected.

Many community-based programs do not provide
sufficient punishment for offenders.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 379

Section 6: Training and Oversight
10. The following questions pertain to your job and the District Attorney’s Office in which you
work. Please circle the number that best corresponds to your level of agreement with the following
statements. (Circle one number for each item)

a
b
c
d

Strongly
Disagree

Disagree

Neutral

Agree

Strongly
Agree

Experienced prosecutors are best qualified to screen
cases for prosecution.

Office policies provide clear guidance on how to
handle cases.

There needs to be more communication among staff
to ensure consistency of outcomes.
My supervisors rarely review the plea offers that I
h
make in cases.
I feel constrained by office policies and practices
i
about when to accept or decline cases for
prosecution.
New prosecutors in my office receive adequate
j
training before they start independently handling
cases.
I often compare my cases to those of my colleagues
k
to ensure that I am getting similar outcomes.
Office goals and priorities are clearly
L
communicated to staff.
Prosecutors in my office receive adequate training
M
when they are transferred to a new unit.
Office priorities require case outcomes that I often
N
disagree with.
O It is important to routinely review cases as a group.
Office policies about how to handle cases are not
P
consistent with my own beliefs.
g

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 380

Respondent ID: ___________________

Section 7: Background Information [To be separated]

The following questions are about you and your background. Please circle the number that best corresponds
to each item. (Circle one number for each item)
11. What is your gender?
1. Male
2. Female
12. What is your racial background?
1. Black or African-American
2. White
3. Asian
4. Other or more than one race (please specify) ____________
13. What is your ethnic background?
1. Hispanic
2. Non-Hispanic
14. What is your age as of your last birthday? ______ Years
15. Do you work in a specialized unit or a general crimes unit?
1. Specialized unit (e.g. drugs, homicide, sex crimes, etc.)
2. General crimes unit (e.g. misdemeanor unit, general felonies unit, etc.)
16. How long have you been in your current unit assignment?
If less than 2 years, please specify number of months

______ years
______ months

17. How long have you been with the DA’s Office?
If less than 2 years, please specify number of months

_____ years
_____ months

19. Before coming to the DA’s Office, were you ever a defense attorney?
1. Yes
2. No
20. Is your job in the prosecutor’s office your first professional experience?
1. Yes
2. No
21. Have you or anyone in your family ever worked in law enforcement?
1. Yes
2. No
That concludes our survey. On the back of this page, or anywhere else on this document, you are welcome to
write any comments you have about specific questions in this questionnaire, potential uses for the
information obtained, and any additional issues you feel may be beneficial to this research.
Thank you very much for your participation.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 381

Appendix E: Factorial Survey Instrument
In this study, each participant was asked to read and respond to ten case vignettes, each of which
described the circumstances surrounding an arrest, the offenses charged by the police, the evidence presented
to the prosecutor by the police, and the age, race, gender, and criminal history of the suspect. Each vignette
consisted of a base scenario that was the same for all respondents and two elements that were systematically
varied: defendant criminal history and race.
Ten criminal history profiles and two race designations (black vs. white) were created and allowed to
vary across vignettes. Suspect and victim genders were not varied because they were integral features of
some of the scenarios. Similarly, defendant age and criminal history were confounded in the final vignettes,
because a given list of arrest and conviction dates was only possible for persons of a certain minimum age.
All respondents received the same ten age/criminal history profiles, but the pairing of criminal history
profiles with base scenarios was randomized across respondents. For each respondent, half of the defendants
in the ten vignettes were designated as black and the other half were designated as white, but the assignment
of race categories to vignettes was counterbalanced across vignettes. As a result, each respondent received a
set of ten vignettes that was unique with respect to combinations of the basic scenarios with age/prior
criminal history and race.
The resulting vignettes were reformatted to resemble a standardized arrest report. Each hypothetical
report was accompanied by a questionnaire, asking prosecutors to indicate whether they would accept or
reject the case for prosecution, give their reasons for rejection if applicable, specify the charges they would
file initially, specify the charges to which the defendant must plead guilty in connection with a plea bargain,
and indicate what sentence recommendation they would make, if any. The packet also included a
questionnaire asking the respondents about some of their personal characteristics: age, race, gender, and
professional experience in various capacities as attorneys handling criminal cases. This appendix provides
the instructions to respondents, the survey questionnaires, the ten arrest report templates for each vignette,
and the ten age/prior record profiles.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 382

Respondent ID: __________________

FACTORIAL SURVEY QUESTIONNAIRE
The Anatomy of Discretion

GENERAL INSTRUCTIONS
The Vera Institute of Justice is a private non-profit organization with a 50-year history of
commitment to social justice through exploratory research and demonstration programming. Vera
is currently conducting a study of prosecutorial decision-making in two county prosecutors’ offices
in the United States. Funding for this project is provided through the National Institute of Justice.
We are interested in understanding how prosecutors make decisions in cases – the information they
use to evaluate a case and the weight they give to different factors in a case. We are also interested
in what factors outside of a case, such as office policies, interactions with other justice system
actors, or workloads, may ultimately affect prosecutors’ decisions.
To account for the experiences of prosecutors, you have been given a consent form, a questionnaire
sheet and a booklet containing 10 unique vignettes – short multidimensional descriptions of
hypothetical cases. The questionnaire will ask you to state your decisions for each hypothetical
case, explain your decisions, and rate your perceptions of seriousness of the offense, strength of the
evidence, and defendant criminal history. Please fill out a questionnaire for each vignette and
respond openly and honestly. Please do not put your name on the surveys.
Your participation is voluntary and is strictly confidential. Your rights as a participant are
described in the informed consent form that we ask you to sign before answering the questionnaire.
While the District Attorney has been informed about this survey, no one in the District Attorney’s
Office will have access to your answers. Only summary statistics and aggregate responses will be
reported. If you decide to participate, please fill out and sign the consent form upon receipt of your
packet and return it to our designated project staff on site.
Please complete the survey and return the entire set of completed questionnaires in the provided
envelope to our project staff on site. We will be collecting your questionnaires on
__________________________.
We would like to thank you in advance for your contribution and cooperation with this important
project. Please feel free to contact the Principal Investigators for any additional information on the
scope of the project, our methodology or any other area of interest related to our project:
Don Stemen, Loyola University Chicago, dstemen@luc.edu, (312) 915-7570
Bruce Frederick, Vera Institute of Justice, bfrederick@vera.org, (518) 391-5799

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 383

Respondent ID: __________________

FACTORIAL SURVEY QUESTIONNAIRE
The Anatomy of Discretion
Case Number: ___________
1
1. Circle the number that best represents the priority you, as a prosecutor, feel that this case should have for
prosecution:
Lowest Priority
1

Average Priority
3

Highest Priority
5

2. Circle the number that best represents how strong you believe the evidence in this case is.
Weakest
1

Average
3

Strongest
5

3. Circle the number that best represents how serious you believe the defendant’s criminal history is.
Not Serious
1

Average
3

Very serious
5

4. After reviewing the case, would you accept or reject this case for prosecution?
(1) Accept (proceed to Question 5)
(2) Reject
a. If you rejected the case, what is the primary reason for rejecting it:
______________________________________________________________
b. Are there any other reasons for rejecting it?
______________________________________________________________
5. If you accepted the case for prosecution, what charges would you recommend? (Include as many as you
believe are applicable).
___________________________________________________________________________
___________________________________________________________________________
___________________________________________________________________________
6. Given the charges that you recommended in Question 5, when drafting a plea offer, which charges would
you require the defendant to plead guilty to? (Include as many as you believe are applicable)
___________________________________________________________________________
___________________________________________________________________________
___________________________________________________________________________
7. What sentence would you recommend for this defendant? (Circle one)
(1) Fine/restitution
(3) Probation
(5) Prison
(2) Community Service
(4) Jail
(6) Other (explain)____________________

8. Are there any potential problems you think may arise as the case progresses

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 384

Respondent ID: __________________

FACTORIAL SURVEY QUESTIONNAIRE
The Anatomy of Discretion
Background Information
1. What is your gender?

Male (1)

Female (2)

2. What is your racial background?
5. Black or African-American (1)
6. White (2)
7. Asian (3)
8. Other or more than one race (4) ____________
3. What is your ethnic background?____________________________________
4. What is your current marital status?
_ Married (1)
_ Unmarried, but living with a partner (2)
_ Married, but living alone (3)
_ Unmarried, and living alone (4)
5. What is your age as of your last birthday? ______
6. What unit or unit are you currently assigned to? ________
7. How long have you been in this assignment?

_____

8. How long have you been with the DA’s Office?

_____

Weeks

_____

Months Years
_____

_____

Months Years
_____

9. How old were you when you joined the Office? ______
10. Before coming to the DA’s Office, were you ever a defense attorney?
_ Yes (1)
_ No (2)
11. Is your job in the prosecutor’s office your first professional experience?
_ No (2)
_ Yes (1)
12. That concludes our survey. On the back of this page, or anywhere else on this document, you are welcome to
write any comments you have about specific questions in this questionnaire, potential uses for the information
obtained, and any additional issues you feel may be beneficial to this research.
Thank you very much for your participation.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 385

Mailmerge Template for Vignette #1

ANYWHERE U.S.A. ARREST PROCESSING
CENTER
ARRESTEE’S INFORMATION
Complaint #:
«respid»«newcase»«casenum1»«packnum»
Last Name:
SSN:

Drivers License:

Date of Birth:
«dob»
Home block #

Place of Birth:

City:

Dir.

Arrest #:
20101111-«newcase»

PID #:

First Name:

Middle:

Street:

State:
NC
ARRESTEE’S PRIOR CRIMINAL HISTORY
Arrest Date
Charge
«priordate1»
«priorchg1»
«priordate2»
«priorchg2»
«priordate3»
«priorchg3»
«priordate4»
«priorchg4»
«priordate5»
«priorchg5»
«priordate6»
«priorchg6»
«priordate7»
«priorchg7»
«priordate8»
«priorchg8»
«priordate9»
«priorchg9»
«priordate10»
«priorchg10»
«priordate11»
«priorchg11»
«priordate12»
«priorchg12»
«priordate13»
«priorchg13»
«priordate14»
«priorchg14»
«priordate15»
«priorchg15»
«priordate16»
«priorchg16»
«priordate17»
«priorchg17»
«priordate18»
«priorchg18»

State:
NC
Height:
66 in
Zip:

Race:
«race»
Weight:
160 lbs

Arrest
Report

Suf:
Sex:
Male
Hair:
BLK
Type:

Eyes:
BLK
Apt:

Phone:
Disposition
«priordispo1»
«priordispo2»
«priordispo3»
«priordispo4»
«priordispo5»
«priordispo6»
«priordispo7»
«priordispo8»
«priordispo9»
«priordispo10»
«priordispo11»
«priordispo12»
«priordispo13»
«priordispo14»
«priordispo15»
«priordispo16»
«priordispo17»
«priordispo18»

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 386

ARREST CHARGE INFORMATION
Charge:
Burglary (Residential)
Charge:
Assault
Charge:
Charge:
ARREST REPORT
On November 11, 2010, the defendant was arrested for Burglary (Residential) and Assault.
The arresting officer was dispatched on a burglary in progress call. On his arrival at the victim’s dwelling, the
victim indicated that the suspect had fled and gave the arresting officer a description which was broadcast. As
the arresting officer drove in the immediate area adjacent to the victim’s dwelling, he saw a person who fitted
the description given by the victim. The officer stopped the subject and thereafter the victim within minutes
arrived on the scene and confirmed that it was the subject who had been in the victim’s basement.
Further investigation revealed that the subject (the defendant) had entered the dwelling through the basement
door and was rummaging through the basement when the victim surprised him. When the victim attempted to
detain him, the defendant shoved the victim and fled.

EVIDENCE
Testimony of the victim as to the presence of defendant in basement and as to method of entry; testimony of
victim that premises had been secured prior to entry and that defendant was there without permission.
Testimony of arresting officer as to circumstances surrounding description of defendant given him by victim
and arresting officer’s subsequent finding of defendant in the neighborhood and his arrest thereafter.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 387

Mailmerge Template for Vignette #2

ANYWHERE U.S.A. ARREST PROCESSING
CENTER
ARRESTEE’S INFORMATION
Complaint #:
«respid»«newcase»«casenum1»«packnum»
Last Name:
SSN:

Drivers License:

Date of Birth:
«dob»
Home block #

Place of Birth:

City:

Dir.

Arrest #:
20101102-«newcase»

PID #:

First Name:

Middle:

Street:

State:
NC
Height:
72 in
Zip:

Race:
«race»
Weight:
230 lbs

Arrest
Report

Suf:
Sex:
Male
Hair:
BRWN
Type:

Eyes:
BRWN
Apt:

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 388

ARREST CHARGE INFORMATION
Charge:
Robbery in the Second Degree
Charge:
Conspiracy
Charge:
Charge:
ARREST REPORT
On November 2, 2010, at 8:25 pm, the defendant was arrested for Robbery in the Second Degree and
Conspiracy.
On November 2, 2010, the victim was walking on the street in the city when she was approached by two males
who were jogging at the same time she first saw them. As they came up to her, one of the males grabbed the
small purse she had in her hand and sprinted up the street. The other individual, the defendant, continued
jogging at the same speed and upon hearing the victim scream stopped and came back to talk to her. At this
time the victim indicated that her purse had been taken and the defendant responded, “I know the man, he’s a
friend of time. I’ll get you stuff back for you.” The victim said that the defendant did not run simultaneously
with the other man who was never apprehended. There was no indication or evidence that the defendant and
the unknown other suspect had consulted prior to the other subject’s taking the victim’s purse. The defendant
was arrested by Witness #2 who while on routine patrol came on the scene and was told by the victim that her
purse had been taken.

EVIDENCE
Victim’s testimony as to the taking of the purse
Testimony of arresting officer as to defendant who was fleeing at the time he was apprehended.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 389

Mailmerge Template for Vignette #3

ANYWHERE U.S.A. ARREST PROCESSING
CENTER
ARRESTEE’S INFORMATION
Complaint #:
«respid»«newcase»«casenum1»«packnum»
Last Name:
SSN:

Drivers License:

Date of Birth:
«dob»
Home block #

Place of Birth:

City:

Dir.

Arrest #:
20101105-«newcase»

PID #:

First Name:

Middle:

Street:

State:
NC
Height:
73 in
Zip:

Race:
«race»
Weight:
205 lbs

Arrest
Report

Suf:
Sex:
Male
Hair:
BRWN
Type:

Eyes:
BRWN
Apt:

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 390

ARREST CHARGE INFORMATION
Charge:
Felony Criminal Mischief
Charge:
Criminal Trespass
Charge:
Charge:
ARREST REPORT
On November 5, 2010, defendant was arrested for Felony Criminal Mischief and Criminal Trespass.
At approximately 11:10 pm, on November 5, the police were contacted by Witness #1 who reported acts of
vandalism being committed at a local high school. A police officer (Witness #2) responded to the scene where
he was told by Witness #1 that four males had emerged from a white vehicle, climbed a short fence onto the
property and proceeded to push over three light poles along the driveway of the high school. Witness #1 had
copied a license plate number which he gave to the police. The license number was reported to central
communications and shortly thereafter the vehicle in question was stopped by arresting officer. The
defendant, who was the only person in the vehicle, was arrested and taken back to the scene. Witness #1
positively identified him as one of the four persons he had seen pushing over the light poles. It was
subsequently determined that the damage done at the crime scene amounted to $625.

EVIDENCE
Testimony of eyewitness as to acts of criminal mischief and trespass
Testimony of Witness #2 as to having given description of vehicle and license plate number
Testimony of arresting officer as to defendant’s presence in vehicle at time of arrest
Testimony of school official as to extent of damage.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 391

Mailmerge Template for Vignette #4

ANYWHERE U.S.A. ARREST PROCESSING
CENTER
ARRESTEE’S INFORMATION
Complaint #:
«respid»«newcase»«casenum1»«packnum»
Last Name:
SSN:

Drivers License:

Date of Birth:
«dob»
Home block #

Place of Birth:

City:

Dir.

Arrest #:
20101112-«newcase»

PID #:

First Name:

Middle:

Street:

State:
NC
Height:
71 in
Zip:

Race:
«race»
Weight:
190 lbs

Arrest
Report

Suf:
Sex:
Male
Hair:
BLK
Type:

Eyes:
BLK
Apt:

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 392

ARREST CHARGE INFORMATION
Charge:
Aggravated Assault with a Deadly Weapon (a pistol)
Charge:
Assault in the Third Degree – two counts
Charge:
Carrying a Concealed Dangerous Weapon
Charge:
Possession of a Weapon during Commission of a Felony
ARREST REPORT
On November 12, 2010, the defendant, a male, was arrested for Aggravated Assault with a Deadly Weapon (a
pistol), two counts of Assault in the Third Degree, Carrying a Concealed Dangerous Weapon, and Possession
of a Weapon during Commission of a Felony.
On November 12, 2010, the defendant knocked on the door of Witness #1’s apartment. She states that she did
not answer immediately and the defendant began pounding on the door. When Witness #1 opened the door,
the defendant ran upstairs and confronted Witness #2 who was visiting Witness #1. The defendant accused
Witness #2 of “messing around with his girlfriend.” (Witness #1). He began punching Witness #2 in the face,
eventually shoving him down the stairs of the apartment. The defendant then threatened to use a .38 caliber
Smith and Wesson revolver which he was carrying in a holster on his right hip, stating that he would kill
Witness #1 and Witness #2. After pushing Witness #2 down the steps, he followed him outside and fired the
revolver into the ground a few feet away from Witness #2. He then allowed Witness #2 to leave, went back
upstairs and assaulted Witness #1, causing her a black eye as well as many contusions and abrasions.
During the incident Witness #3 who was in the apartment viewed the assault by the defendant on Witnesses #1
and #2. Witness #4 also in the apartment observed the defendant fire the revolver. Witness #4 was in the
bathroom and did not come out while the commotion was taking place. Witness #5 responded to the telephone
call from the apartment made by Witness #3 who stated that the defendant was still in the apartment holding
Witness #1 as a hostage. At this point the arresting officer telephoned the apartment and advised the defendant
that he should come out of the apartment with his hands up. The defendant then walked out of the apartment
as he was instructed and he was taken into custody. The defendant was transported to jail by Witness #6.
Witness #5 attempted to recover the spent round from the weapon, however he was not able to find it. The
weapon which was taken from the defendant by Witness #5 is in evidence. When the defendant arrived at the
jail he was advised by Witness #6 of his rights and thereafter gave a full statement admitting to discharging the
weapon, assaulting Witnesses #1 and #2 and ripping the phone out of the wall at the apartment.
EVIDENCE
Weapon used by defendant
Testimony of victims (Witness #1 and #2) as to assault by the defendant and firing of pistol
Corroborative testimony of witnesses #3 and #4
Testimony of Witness #6 as to confession by defendant
Confession

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 393

Mailmerge Template for Vignette #5

ANYWHERE U.S.A. ARREST PROCESSING
CENTER
ARRESTEE’S INFORMATION
Complaint #:
«respid»«newcase»«casenum1»«packnum»
Last Name:
SSN:

Drivers License:

Date of Birth:
«dob»
Home block #

Place of Birth:

City:

Dir.

Arrest #:
20101112-«newcase»

PID #:

First Name:

Middle:

Street:

State:
NC
Height:
67 in
Zip:

Race:
«race»
Weight:
170 lbs

Arrest
Report

Suf:
Sex:
Male
Hair:
BLK
Type:

Eyes:
BLK
Apt:

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 394

ARREST CHARGE INFORMATION
Charge:
Robbery 1
Charge:
Conspiracy
Charge:
Charge:
ARREST REPORT
On November 12, 2010, at 10:40 PM, the defendant, a male, was arrested for Robbery 1 and Conspiracy.
On November 12, 2010, at approximately 7:30 PM, the victim accompanied by her son, 19 years of age, was
travelling in her automobile on a city street and noticed that the fire hydrant was spraying across the entire
street ahead of her. As she slowed her vehicle a male ran up to the driver’s side of the car and threw a bucket
of water into her vehicle. Two other suspects ran up to the vehicle, one on each side. The suspect on the
passenger’s side then reached inside the car window and grabbed the victim’s purse which was on the floor of
the car. At the same time he also struck the victim’s son on the face. The door on the driver’s side was
opened and the defendant attempted to pull the victim out of the vehicle. The victim slammed the door and
rolled up her window. The victim then struggled with the suspect who had taken her purse. As she grabbed
the purse and started to pull, the purse ripped and the defendant then took it out of her hand. In the purse was a
wallet containing $125, jewelry worth $200, and a variety of charge plates and credit cards.
The victim gave a detailed description of the subject who had taken the purse.
Upon the report of the crime a police unit responded to the vicinity where the crime had occurred and shortly
thereafter arrested the defendant. The victim, who was not seriously injured, and her son, at the crime scene
positively identified the defendant as the person who had taken the purse.

EVIDENCE
Testimony of the victim
Testimony of the arresting officer
Testimony of victim’s son.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 395

Mailmerge Template for Vignette #6

ANYWHERE U.S.A. ARREST PROCESSING
CENTER
ARRESTEE’S INFORMATION
Complaint #:
«respid»«newcase»«casenum1»«packnum»
Last Name:
SSN:

Drivers License:

Date of Birth:
«dob»
Home block #

Place of Birth:

City:

Dir.

Arrest #:
20101101-«newcase»

PID #:

First Name:

Middle:

Street:

State:
NC
Height:
74 in
Zip:

Race:
«race»
Weight:
215 lbs

Arrest
Report

Suf:
Sex:
Male
Hair:
BRWN
Type:

Eyes:
HZL
Apt:

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 396

ARREST CHARGE INFORMATION
Charge:
Burglary
Charge:
Theft
Charge:
Possession of a Stolen Vehicle
Charge:
ARREST REPORT
On November 1, 2010, the defendant a male, was arrested for Burglary, Theft, and Possession of a Stolen
Vehicle.
During the early morning hours of November 1, 2010, Witness #1 was awakened by a loud noise which she
described as sounding like “a car losing an engine.” When she arose and looked out of her window she
noticed a dark colored utility bed pick-up truck with clearance lights on top of the cab in the parking lot of the
bank across the street from her apartment. As she watched, a male got in the passenger side of the vehicle and
the vehicle thereafter departed.
A short while later she heard the truck pull back into the parking lot. Again she looked out the window and
called the police after seeing the passenger get out of the truck and go to the side of the bank. She thought that
the person she observed was tampering with either the 24-hour depository or the side door.
Witness #1 remained at her window and thereafter observed the truck return a third time. She again called the
police and informed the communications officer that the vehicle was back. The officer who received her call
held her on the phone until patrol cars arrived and apprehended the defendants. Accordingly Witness #1 was
able to observe all that transpired at the scene.
She indicated to the investigating officers that as the truck, occupied by the defendant, left the third time, a
brief period elapsed until it reappeared for the fourth time. It was at the s time that the vehicle was stopped by
a patrol car which had responded to the burglary call. At the time the vehicle was stopped by the investigating
officer (Witness #2), the defendant jumped out of the vehicle and came back towards the police car.
Defendant identified himself using a fictitious name but later gave his real name at the police station. The
investigating officer (Witness #2) placed defendant under two hour detention, gave him his rights and placed
him in a patrol car. Investigating officer (Witness #3) arrived in a second car on the scene. Witness #3
directed Witness #2 to search the inside of the truck for fruits or instruments of the crime which was believed
to have occurred at the bank. In the cab of the truck Witness #2 found a brown Trust Company deposit bag
#26A with the name “Church of the Holy Child.”
The officers returned the defendant to the scene where defendant told Witness #2 that he was on the way to
work when he saw the night deposit open.
Investigating officers (Witnesses #2 and #3) determined that entry into the night deposit vault was made by
using a key to open the outside deposit plate to the hopper. Thereafter the defendant inserted a tow chain hook
under the banner which was then attached to the trailer hitch on the vehicle in which the defendant was
apprehended. Apparently the original noise that woke up Witness #1 occurred when the hopper assembly was
pulled out to expose the night vault below. The investigating officer (Witness #2) surmised that the return

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 397

trips that followed by the defendant were attempts to remove money bags which were in the vault. The motor
vehicle in which the defendant was found, had been stolen earlier in the evening from Witness #4.

EVIDENCE
Brown deposit bag marked “Trust Company – Church of the Holy Child” containing @286 in bills and $251
in coins along with $883 in miscellaneous checks.
Testimony of Witness #2 as to defendant 1 giving false identification to the arresting officer.
Testimony of Witness #1 as to the defendant’s presence at the bank.
Testimony of owner of the pick-up truck as to ownership and theft.
Testimony of Pastor of Church of the Holy Child as to having made night deposit earlier in the evening on day
crime occurred.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 398

Mailmerge Template for Vignette #7

ANYWHERE U.S.A. ARREST PROCESSING
CENTER
ARRESTEE’S INFORMATION
Complaint #:
«respid»«newcase»«casenum1»«packnum»
Last Name:
SSN:

Drivers License:

Date of Birth:
«dob»
Home block #

Place of Birth:

City:

Dir.

Arrest #:
20101031-«newcase»

PID #:

First Name:

Middle:

Street:

State:
NC
Height:
68 in
Zip:

Race:
«race»
Weight:
200 lbs

Arrest
Report

Suf:
Sex:
Male
Hair:
BLK
Type:

Eyes:
BLK
Apt:

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 399

ARREST CHARGE INFORMATION
Charge:
Possession of a controlled substance (heroin)
Charge:
Charge:
Charge:
ARREST REPORT
On October 31, 2010 the defendant was arrested for possession of a controlled substance (heroin).
The arresting officer on the above date stopped the defendant for driving carelessly. The defendant was asked
for his operator’s license. As the defendant sat in his car seat and removed the license, the arresting officer
observed a number of foil packets in his wallet. Based on his past experience as a police officer familiar with
controlled substances, the arresting officer concluded that the packets contained heroin and arrested the
defendant at this time. Laboratory analysis of the eighteen packets seized from the defendant proved positive
for heroin.

EVIDENCE
Testimony of arresting officer as to circumstances surrounding defendant’s arrest.
Eighteen foil packets seized from person of defendant.
Lab technician’s testimony that foil packets contained heroin.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 400

Mailmerge Template for Vignette #8

ANYWHERE U.S.A. ARREST PROCESSING
CENTER
ARRESTEE’S INFORMATION
Complaint #:
«respid»«newcase»«casenum1»«packnum»
Last Name:
SSN:

Drivers License:

Date of Birth:
«dob»
Home block #

Place of Birth:

City:

Dir.

Arrest #:
20101104-«newcase»

PID #:

First Name:

Middle:

Street:

State:
NC
Height:
75 in
Zip:

Race:
«race»
Weight:
175 lbs

Arrest
Report

Suf:
Sex:
Male
Hair:
BRWN
Type:

Eyes:
BRWN
Apt:

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 401

ARREST CHARGE INFORMATION
Charge:
Failure to move on
Charge:
Charge:
Charge:
ARREST REPORT
On November 4, 2010, the defendant, a male, was arrested for failure to move on.
On the above date, the arresting officer was in the process of writing a traffic ticket for a person whom he had
stopped for driving at an excessive rate of speed. The defendant, a friend of the person stopped by the officer,
came on the scene and began to harangue the officer as to the unfairness of giving his friend a ticket “while all
of those criminals are permitted to run free.” The officer repeatedly told the defendant to move on and finally
arrested him for his failure to do so.

EVIDENCE
Testimony of arresting officer as to conduct of defendant which led to his arrest.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 402

Mailmerge Template for Vignette #9

ANYWHERE U.S.A. ARREST PROCESSING
CENTER
ARRESTEE’S INFORMATION
Complaint #:
«respid»«newcase»«casenum1»«packnum»
Last Name:
SSN:

Drivers License:

Date of Birth:
«dob»
Home block #

Place of Birth:

City:

Dir.

Arrest #:
20101109-«newcase»

PID #:

First Name:

Middle:

Street:

State:
NC
Height:
62 in
Zip:

Race:
«race»
Weight:
110 lbs

Arrest
Report

Suf:
Sex:
Female
Hair:
Eyes:
BLND
BLUE
Type:
Apt:
Phone:
Disposition
«priordispo1»
«priordispo2»
«priordispo3»
«priordispo4»
«priordispo5»
«priordispo6»
«priordispo7»
«priordispo8»
«priordispo9»
«priordispo10»
«priordispo11»
«priordispo12»
«priordispo13»
«priordispo14»
«priordispo15»
«priordispo16»
«priordispo17»
«priordispo18»

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 403

ARREST CHARGE INFORMATION
Charge:
Forgery – 4 counts
Charge:
Theft (under $300) – 4 counts
Charge:
Charge:
ARREST REPORT
On November 9, 2010, the defendant, a female, was arrested for four counts of Forgery and four counts of
Theft (under $300).
On July 19, 2010 a breaking and entering occurred in a grocery store in the city. Among other things, a
number of blank checks drawn from Central Bank were stolen. The serial numbers of the stolen checks were
provided to the police by the owner of the grocery store. The Central Bank was immediately notified by the
merchant that the theft had occurred.
On July 21, 2010, one of the stolen checks in the amount of $175 was forged and passed to Witness #1. On
the same date another one was forged in the amount of $154.30 and passed to the manager of another liquor
store in the city (Witness #2). On July 22, 2010, a third check was forged in the amount of $145.65 and passed
to Witness #3, the clerk in another liquor store. On July 22, 2010, a fourth stolen check in the amount of
$195.43 was forged and passed to the Peoples’ Bank in the city.
A video camera in the final liquor store in which Witness #3 received the check took a picture of the
defendant. The positive identification by Witness #3 resulted in the arrest of the defendant. The remaining
witnesses, including the clerk at the People’s Bank (Witness #4), made positive identification of the defendant.
EVIDENCE
Witness #1 and forged check in the amount of $175
Witness #2 and forged check in the amount of$154.30
Witness #3 and forged check in the amount of $145.65
Witness #4 and forged check in the amount of $195.43
Witness #5, representative of Central Bank, as to ownership of check and fact that check were reported stolen.
Witness #6, owner of grocery store, as to burglary and theft of checks.

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 404

Mailmerge Template for Vignette #10

ANYWHERE U.S.A. ARREST PROCESSING
CENTER
ARRESTEE’S INFORMATION
Complaint #:
«respid»«newcase»«casenum1»«packnum»
Last Name:

Arrest #:
20101105-«race»

PID #:

First Name:

Middle:

SSN:

Drivers License:

Date of Birth:
«dob»
Home block #

Place of Birth:
Dir.

Street:

City:
ARRESTEE’S PRIOR CRIMINAL HISTORY
Arrest Date
Charge
«priordate1»
«priorchg1»
«priordate2»
«priorchg2»
«priordate3»
«priorchg3»
«priordate4»
«priorchg4»
«priordate5»
«priorchg5»
«priordate6»
«priorchg6»
«priordate7»
«priorchg7»
«priordate8»
«priorchg8»
«priordate9»
«priorchg9»
«priordate10»
«priorchg10»
«priordate11»
«priorchg11»
«priordate12»
«priorchg12»
«priordate13»
«priorchg13»
«priordate14»
«priorchg14»
«priordate15»
«priorchg15»
«priordate16»
«priorchg16»
«priordate17»
«priorchg17»
«priordate18»
«priorchg18»

State:
NC

State:
NC
Height:
68 in
Zip:

Race:
«race»
Weight:
155 lbs

Arrest
Report

Suf:
Sex:
Female
Hair:
Eyes:
BRWN
HZL
Type:
Apt:
Phone:
Disposition
«priordispo1»
«priordispo2»
«priordispo3»
«priordispo4»
«priordispo5»
«priordispo6»
«priordispo7»
«priordispo8»
«priordispo9»
«priordispo10»
«priordispo11»
«priordispo12»
«priordispo13»
«priordispo14»
«priordispo15»
«priordispo16»
«priordispo17»
«priordispo18»

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 405

ARREST CHARGE INFORMATION
Charge:
Possession of a stolen credit card
Charge:
Attempted illegal use of a credit card
Charge:
Charge:
ARREST REPORT
On November 5, 2010, the defendant, a female was arrested for possession of a stolen credit card and
attempted illegal use of a credit card.
On the above date, the defendant attempted to purchase a number of sundry items at a drug store which she
sought to pay for by using a stolen credit card. When the sales clerk made a phone call to verify the card, he
was informed that it had been stolen and that the police had been called to come to the drug store. The
defendant was arrested by the arresting officer just as the defendant attempted to regain possession of the card
and leave the store.

EVIDENCE
Stolen credit card
Testimony of card owner that card had been in her purse at time it had been stolen.
Clerk’s testimony as to attempt by defendant
Arresting officer’s testimony as to apprehension of defendant

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 406

Defendant Age/Criminal History Profiles
Profile

DOB

Profile #1

01/02/65

Profile #2

9/04/77

Profile #3

9/01/83

Profile #4

05/01/76

Profile #5

10/23/80

Arrest Date

Charge

Disposition

7/14/87
12/20/05
6/01/01
8/08/03
12/03/05
5/18/06
11/17/09
02/10/04
05/01/05
11/16/95
12/30/95
03/19/96
11/02/98
01/16/99
12/15/99
07/30/00
04/22/02
04/29/02
03/09/04
09/27/06
10/18/08
11/23/00
10/01/01
11/10/01
11/10/01
12/01/01
12/31/01
12/31/01
12/31/01
01/28/02
10/10/02
04/14/05
09/22/07
06/04/08
06/04/08
11/25/08
04/21/09
10/04/09
10/04/09

Rape
Murder
Receiving Stolen Property
Misdemeanor Assault
Misdemeanor Assault
Misdemeanor Assault
Misdemeanor Assault
Arson
Misdemeanor Assault
Drunkenness
DWI
Drunkenness
Misdemeanor Assault
Receiving Stolen Property
Misdemeanor Assault
Misdemeanor Assault
Misdemeanor Assault
Misdemeanor Assault
Drunkenness
Possession of Heroin
Misdemeanor Larceny
Felony Larceny
Burglary, 1st degree
Felony Larceny
Possession of Cocaine
Contributing to Delinquency of a Minor
Possession of Cocaine
Possession of Mescaline
Sale of Heroin
Possession of Narcotic Equipment
Felony Larceny
Felony Larceny
Misdemeanor Assault
Intimidation
Disorderly Conduct
Carrying a Prohibited Weapon
Possession of Explosives
Intimidation
Possession of Marijuana

Convicted
Convicted
Convicted
Convicted
Convicted
Dismissed
Dismissed
Convicted
Convicted
Convicted
Convicted
Dismissed
Convicted
Dismissed
Dismissed
Convicted
Convicted
Convicted
Convicted
Convicted
Dismissed
Acquittal
Convicted
Convicted
Convicted
Dismissed
Convicted
Convicted
Convicted
Convicted
Convicted
Convicted
Acquittal
Acquittal
Convicted
Dismissed
Dismissed
Acquittal
Acquittal

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 407

Profile

DOB

Arrest Date

Charge

Conviction

Profile #6

03/28/83

No Prior Record

Profile #7

1/20/81

08/02/01
06/03/04
06/14/05
02/22/06
05/22/08

Felony Larceny
Forgery
Disorderly Conduct
Carrying Concealed Weapon
Felony Larceny

Convicted
Dismissed
Convicted
Convicted
Convicted

Profile #8

05/15/80

No Prior Record

01/16/79

04/05/94
11/24/96
05/09/98
12/15/99
05/30/01
07/05/02
10/05/03
04/04/04
08/13/04
03/16/05
01/01/06
06/01/07
12/24/08
12/06/09

Loitering
Burglary, 2nd Degree
Burglary, 2nd degree
Possession of Marijuana
Traffic Offense
Burglary, 2nd degree
Probation Violation
Possession of Heroin
Felony Larceny
Possession of Marijuana
Possession of Cocaine
Possession of Cocaine
Receiving Stolen Property
Possession of Heroin

Convicted
Acquittal
Convicted
Convicted
Acquittal
Acquittal
Acquittal
Convicted
Convicted
Convicted
Acquittal
Convicted
Dismissed
Dismissed

03/17/74

07/15/94
11/18/96
09/20/97
07/03/99
07/24/01
09/21/03
10/11/04
10/01/07

Misdemeanor Assault
Misdemeanor Assault
Misdemeanor Assault
Misdemeanor Assault
Misdemeanor Assault
Sale of Heroin
Burglary, 1st degree
Assault, Serious Bodily Injury

Convicted
Dismissed
Convicted
Convicted
Convicted
Dismissed
Convicted
Convicted

Profile #9

Profile #10

The Anatomy of Discretion: An Analysis of Prosecutorial Decision Making 408