THE EFFECTIVENESS OF JUVENILE
CORRECTIONAL FACILITIES: PUBLIC VERSUS
PRIVATE MANAGEMENT*
PATRICK BAYER and DAVID E. POZEN
Yale University

Abstract
This paper uses data on juvenile offenders released from correctional facilities in
Florida to explore the effects of facility management type (private for-profit, private
nonprofit, public state-operated, and public county-operated) on recidivism outcomes
and costs. The data provide detailed information on individual characteristics, criminal
and correctional histories, judge-assigned restrictiveness levels, and home zip codes—
allowing us to control for the nonrandom assignment of individuals to facilities far
better than any previous study. Relative to all other management types, for-profit
management leads to a statistically significant increase in recidivism, but relative to
nonprofit and state-operated facilities, for-profit facilities operate at a lower cost to
the government per comparable individual released. Cost-benefit analysis implies that
the short-run savings offered by for-profit over nonprofit management are negated
in the long run due to increased recidivism rates, even if one measures the benefits
of reducing criminal activity as only the avoided costs of additional confinement.

I.

Introduction

S

ince its beginnings in the mid-1980s, prison privatization in the United
States has provoked several rounds of congressional hearings and hundreds
of articles discussing its philosophical, organizational, economic, and legal
implications. At year-end 2001, privately operated facilities held over 6.5
percent of America’s total adult correctional facility population, which represents more than 90,000 adult offenders.1 And in late 1999, privately operated facilities held almost 30 percent of all juveniles in residential place* The authors wish to thank the Florida Department of Juvenile Justice (DJJ) and the Justice
Research Center (JRC), Inc., for providing the data used in this study. In particular, we wish
to thank Sherry Jackson, Steven Chapman, and Ted Tollett from the Florida DJJ and Julia
Blankenship and Kristin Winokur from the JRC for many helpful conversations concerning
the data and, more generally, concerning the operation of the DJJ. We also thank Joe Altonji,
Austan Goolsbee, Jonathan Gruber, John Simon, and an anonymous referee for valuable comments. We gratefully acknowledge the financial support of the Yale Law School and the Yale
Economics Department.
1
Paige M. Harrison & Allen J. Beck, Prisoners in 2001, at 8, table 9 (July 2002).
[Journal of Law and Economics, vol. XLVIII (October 2005)]
᭧ 2005 by The University of Chicago. All rights reserved. 0022-2186/2005/4802-0021$01.50

549

550

the journal of law and economics

ment, which represents more than 30,000 juvenile offenders.2 Following the
example of the United States, many other countries introduced private prisons
in the 1990s, and many more are considering the idea today.3
Setting off a national debate almost instantly, prison privatization emerged
in adult corrections when municipal and state governments—driven primarily
by concerns over excessive costs and crowding in public facilities—began
in 1985 to contract with private firms to run county jails and state prisons.4
Congress held hearings on prison privatization the next year, and almost
every criminal justice professional association took a stand on the issue.
Despite the protests of many, privatization has continued apace since then,
with the capacity of private secure adult correctional facilities increasing 856
percent between 1991 and 1998.5 By the end of 1999, 14 corporations were
operating more than 150 private correctional facilities for adults in the United
States,6 earning combined annual revenues in excess of a billion dollars.
Already a significant owner/operator of juvenile correctional facilities,
private-sector firms began to assume a substantially greater role in juvenile
corrections after Congress passed the Juvenile Justice and Delinquency Prevention Act in 1974.7 This act formally encouraged communities to develop
alternatives to traditional incarceration, and privatization emerged as the
primary mechanism for deinstitutionalization.8 In 1999, there were approx2
Melissa Sickmund, Juvenile Offenders in Residential Placement: 1997–1999, at 1 (2002).
Private organizations also provide court-mandated supervision for thousands of other juveniles
through nonresidential correctional programs such as mental health and substance abuse treatment programs.
3
Richard Harding, Private Prisons, in 28 Crime and Justice: A Review of Research 265,
268–69 (Michael H. Tonry ed. 2001).
4
Private contractors had, in fact, managed adult correctional facilities in a number of U.S.
states during the 1800s, but by the beginning of the twentieth century, all adult prisons were
government run. For a discussion of private prisons’ historic role in America, see generally
Blake McKelvey, American Prisons: A History of Good Intentions (1977); Shaneen Borna,
Free Enterprise Goes to Prison, 26 Brit. J. Criminology 321 (1986); and John G. DiPiano,
Private Prisons: Can They Work? Panopticon in the Twenty-first Century, 21 New Eng. J. on
Crim. & Civ. Confinement 171 (1995). For a discussion of how and why private prisons
returned to the United States, see David E. Pozen, Managing a Correctional Marketplace:
Prison Privatization in the United States and the United Kingdom, 19 J. L. & Pol. 253 (2003).
5
Gaylene Styve Armstrong, Private vs. Public Operation of Juvenile Correctional Facilities
2 (2001).
6
Charles W. Thomas, Private Adult Correctional Facility Census: A “Real-Time” Statistical
Profile, December 31, 1999 (1999) (on file with the authors).
7
Pub. L. No. 93-415, 88 Stat. 1109 (1974) (codified at 42 U.S.C. § 5601 (1994)). In contrast
with its episodic role in adult corrections, the private sector—largely in the form of nonprofit
charities and organizations—remained a consistently important player in the sphere of juvenile
corrections throughout the nineteenth and twentieth centuries. See, for example, Barry Krisberg,
The Legacy of Juvenile Corrections, 57 Corrections Today 122 (1995).
8
See Yitzhak Bakal & Harvey Lowell, The Private Sector in Juvenile Corrections, in Juvenile
Justice and Public Policy: Toward a National Agenda 196 (Ira M. Schwartz ed. 1992); Daniel
J. Curran, Destructuring, Privatization, and the Promise of Juvenile Diversion: Compromising
Community-Based Corrections, 34 Crime & Delinq. 363 (1988).

juvenile correctional facilities

551

imately 1,100 public and 1,800 private juvenile correctional facilities in
operation nationwide,9 and by 1990 almost 90 percent of states had at least
one contract with a nonprofit private corporation and 60 percent of states
had at least one contract with a for-profit corporation to operate a juvenile
correctional facility.10
A.

Prior Literature

For all the controversy engendered and for all the individuals affected by
prison privatization over the last 2 decades, empirical analysis has lagged
the public interest. Two leading surveys of research on prison privatization
explicitly lament the paucity of empirical work on the subject, especially
concerning recidivism outcomes.11 The empirical research has instead tended
to focus on cost and quality-of-confinement comparisons between public and
private facilities.12 Only one study comparing the recidivism rates of public
and private facilities, by Lonn Lanza-Kaduce and Karen Parker, has garnered
any significant attention, and it has been roundly criticized for its small sample
size and errors in design and methodology.13 Moreover, no comparative re9

Sickmund, supra note 2, at 1.
Robert B. Levinson & William J. Taylor, ACA Studies Privatization in Juvenile Corrections,
53 Corrections Today 242, 248 (1991).
11
Gerald G. Gaes, Scott D. Camp, & William G. Saylor, The Performance of Privately
Operated Prisons: A Review of Research, in Private Prisons in the United States: An Assessment
of Current Practice, app. 2, at 2, 31 (Douglas McDonald et al. eds. 1998); Kenneth L. Avio,
The Economics of Prisons, 6 Eur. J. L. & Econ. 143, 151 (1998).
12
Representative examples of public-versus-private studies in the United States include William G. Archambeault & Donald R. Deis, Jr., Cost Effectiveness Comparisons of Private versus
Public Prisons in Louisiana: A Comprehensive Analysis of Allen, Avoyelles, and Winn Correctional Centers (1996); Harry Hatry et al., The Urban Institute, Comparison of Privately and
Publicly Operated Correctional Facilities in Kentucky and Massachusetts (1989); Charles H.
Logan, Well Kept: Comparing Quality of Confinement in a Public and a Private Prison (1991);
Douglas and C. McDonald, The Costs of Operating Public and Private Correctional Facilities,
in Private Prisons and the Public Interest 86 (Douglas C. McDonald ed. 1990).
13
Lonn Lanza-Kaduce & Karen F. Parker, A Comparative Recidivism Analysis of Releasees
from Private and Public Prisons in Florida (1998) (finding that, relative to comparable inmates
released from public prisons, inmates released from private prisons had a lower recidivism
percentage on four of five indicators). For the key critiques of this study, see Florida Department
of Corrections, Bureau of Research and Data Analysis, Preliminary Assessment of a Study
Entitled: “A Comparative Recidivism Analysis of Releasees from Public and Private Prisons
in Florida” (1998); Gaes, Camp, & Saylor, supra note 11, at 28–30. These criticisms remain
apposite for the subsequent, published version of the study. Lonn Lanza-Kaduce, Karen F.
Parker, & Charles W. Thomas, A Comparative Recidivism Analysis of Releasees from Private
and Public Prisons, 45 Crime & Delinq. 28 (1999). Three additional studies have compared
recidivism rates of public and private releasees, although none of these attempted to account
for differences between individuals assigned to different correctional facilities. Peter W. Greenwood, Susan Turner, & Kathy Rosenblatt, Evaluation of Paint Creek Youth Center: Preliminary
Results (1989); Dale K. Sechrest & David Shichor, Quick Fixes in Corrections: Reconsidering
Private and Public For-Profit Facilities, 75 Prison J. 457 (1995); W. Clinton Terry, Lisa Stolzenberg, & Stewart J. D’Alessio, Private versus Public Placements: A Study of Recidivism
among Adjudicated Juvenile Offenders, 48 Juvenile & Family Court J. 33 (1997).
10

552

the journal of law and economics

cidivism analysis has ever considered the distinction between for-profit and
nonprofit management in the private juvenile corrections industry. Given the
large role played by both for-profit and nonprofit institutions in juvenile
corrections,14 the traditional public-versus-private dichotomy provides an
overly simplistic framework for evaluating privatization.
The Florida Department of Juvenile Justice (DJJ), our data source for this
study, rates the correctional facilities under its care on recidivism and cost
performance in its annual Program Accountability Measures (PAM) reports.15
These PAM reports provide the starting point for our analysis, but our study
differs from and expands on them in at least four important ways. First and
most fundamental, our study distinguishes facilities by management type—
private for-profit, private nonprofit, public state-operated, and public countyoperated—while the PAM reports do not. Second, our study uses multiple
definitions of recidivism, including a binary success/failure variable for both
criminal charges and adjudications,16 a survival-time measure, and variables
for 16 specific categories of crime. The PAM reports use only a binary
success/failure variable for adjudications. Third, our study incorporates many
more explanatory variables than the PAM reports do in order to control as
much as possible for differences in the populations served by different types
of correctional facilities. And fourth, our cost-benefit analysis is entirely
novel.
One other area of prior research merits a brief mention: studies of boot
camps’ effects on recidivism. Doris MacKenzie, David Wilson, and Suzanne
Kider’s recent meta-analysis of 29 studies “found no overall difference in
recidivism between boot camp participants (both juveniles and adults) and
comparison samples.”17 The U.S. Office of Juvenile Justice and Delinquency
Prevention reached a similar conclusion.18 It is unclear, however, how relevant
these results are for the case of Florida, as all of Florida’s juvenile boot
camps have the same distinctive management type—primary management
by county sheriff’s departments, with state-level oversight. Whether and how
14
Nonprofits play a much smaller role in the private adult corrections sector: for-profit
corporations manage the vast majority of private adult correctional facilities in the United
States, including 100 percent of the secure private facilities.
15
See, for example, Florida Department of Juvenile Justice, Justice Research Center, Inc.,
The 2003 PAM Report: A Two-Year Analysis (December 2002) (http://www.djj.state.fl.us/
Research/
statsnresearch/mr/2003_mr/2003-1programaccountabilitymeasurereport.pdf). The annual production of these Program Accountability Measures (PAM) reports is mandated by Florida Law.
Fla. Stat. Ann. § 985.412(4)(a)(b) (1997).
16
An adjudication, in the vernacular of the juvenile justice system, is analogous to a conviction in the adult justice system.
17
Doris Layton MacKenzie, David B. Wilson, & Suzanne B. Kider, Effects of Correctional
Boot Camps on Offending, 578 Annals Am. Acad. Pol. & Soc. Sci. 126, 126 (2001).
18
U.S. Department of Justice, National Criminal Justice Association, Juvenile Justice Reform
Initiatives in the States: 1994–1996, at 30–33 (1997).

juvenile correctional facilities

553

the Florida model of boot camp management affects recidivism remain open
questions.
B.

Basic Design of Our Study

This paper attempts to fill an empirical void in the debate over prison
privatization. Using a unique data set containing detailed information on over
5,000 juvenile offenders and 110 juvenile correctional facilities19 in Florida,
we investigate the effects of correctional facility management type—including public, private for-profit, and private nonprofit models—on releasees’
recidivism outcomes and on monetary costs to the state of Florida.20 By
investigating the effects of facility management type on recidivism outcomes,
this study achieves, necessarily, a second purpose: shedding light on the
impact of numerous personal and sociodemographic characteristics on recidivism risk.21
By using the extensive information on the criminal history, residential
locations, judge-assigned restrictiveness levels, and sociodemographic characteristics of the youths observed in the sample, we are able to control for
individual variation in the propensity to recidivate—and, consequently, for
the nonrandom assignment of individuals to facilities on the basis of these
characteristics—far better than any previous analysis. It is important to note,
however, that we are not able to control for the nonrandom assignment of
juveniles to facilities on the basis of any additional factors observable to
juvenile corrections officials but not to the researcher. Thus, the important
caveat remains that the estimated differences across facility management
types may be driven in part by nonrandom assignment according to unobserved factors.22 Many aspects of the analysis and results that follow, however,
19
The Florida DJJ, in its official reports, refers to correctional facilities under its supervision
as “programs” rather than “facilities.” The two terms can be thought of as interchangeable; all
correctional programs are located within a specific facility. We use “facilities” throughout this
paper for clarity and consistency.
20
As the Florida DJJ does in its PAM reports, we therefore assume that correctional facilities
have the potential to influence offenders’ postrelease criminality. This influence could result
from differentials in correctional facility features such as staff-inmate interactions, implementation of rehabilitative and other programmatic services, policies and procedures, and general
atmosphere and ideology. Michel Foucault famously argued that “[d]etention causes recidivism.” Michel Foucault, Discipline and Punish: The Birth of the Prison 265 (Alan Sheridan
trans. 1977) (emphasis added).
21
Currently, there is deep disagreement over what characteristics of an offender influence
his or her probability of recidivating, and to what extent. See Dean J. Champion, Measuring
Offender Risk: A Criminal Justice Sourcebook 92–93 (1994).
22
A second caveat inherent in attempts to study recidivism is that the theoretically relevant
dependent variables—time devoted to criminal activity and the intensity thereof—cannot be
directly measured. Traditionally, researchers have employed a binary “success/failure” measure
as a proxy recidivism variable. See Michael D. Maltz, Recidivism 23 (1984). Moreover, measured recidivism is the product of both offender behavior and enforcement activity. Lacking
information on the behavior of the various Florida police and prosecutorial bodies, this study
could not attempt to resolve the potential problems with simultaneity and enforcement activity

554

the journal of law and economics

limit the likelihood that nonrandom assignment according to unobserved
factors affects the qualitative nature of the main conclusions of the paper.
By exploring both recidivism and costs, we are able to examine the two
key variables of interest in the economic literature on prison privatization.
In a nutshell, economic theory predicts that private for-profit correctional
facilities should operate efficiently due to the profit motive, but in the absence
of explicit linkages between revenues and recidivism outcomes, they might
make decisions designed to increase profits at the expense of increased recidivism.23 Currently, the standard private-prison contract in the United States
remunerates the corporate operator based on the number of person-days of
confinement supplied, subject to some minimal level of amenities. A forprofit prison operator thus has almost no contractual incentive to provide
rehabilitation opportunities or educational or vocational training that might
benefit inmates after release, except insofar as these services act to decrease
the current cost of confinement.24 Decreasing recidivism likely has a bigger
role in the objective functions of publicly operated facilities and private
nonprofit facilities, if for no other reason than that the profit motive is not
as strong. Some commentators have argued that, of all the correctional facility
management types, private nonprofit operators ought to have the most success
at decreasing recidivism due to their organizational and programmatic flexibility, their mission focus, their use of volunteers, and their freedom from
political and profit constraints.25 Overall, then, without explicit linkages between revenues and recidivism outcomes, we might expect private nonprofit
facilities to have lower rates of recidivism, while we would generally expect
for-profit facilities to have higher rates of recidivism but to be able to operate
at a reduced per capita cost.
In Florida, the DJJ does in fact evaluate the correctional facilities it oversees on the basis of recidivism and costs.26 Because facilities’ revenues are
not directly linked to assessments of their recidivism performance except
through the possible elimination of particularly poorly performing facilities,
this linkage is not likely to be strong for the majority of facilities. Still, it
is important to bear in mind that our study analyzes the relative performance
effects by means of explicit controls. However, these problems are not a significant concern
to the extent that the inclusion of variables describing individual, criminal history, and neighborhood characteristics and, in particular, the 20 judicial circuit dummy variables controls for
the variation in the propensity of an individual, given a level of criminal behavior, to be charged
and adjudicated.
23
See Avio, supra note 11, at 150; Oliver Hart, Andrei Schleifer, & Robert W. Vishny, The
Proper Scope of Government: Theory and an Application to Prisons, 112 Q. J. Econ. 1127
(1997).
24
See Peter Schmidt & Ann D. Witte, An Economic Analysis of Crime and Justice: Theory,
Methods and Applications 345–46 (1984).
25
Daniel L. Low, Nonprofit Private Prisons: The Next Generation of Prison Management,
29 New Eng. J. on Crim. & Civ. Confinement 1, 4, 55–56 (2003); Richard Moran, A Third
Option: Nonprofit Prisons, N.Y. Times, August 23, 1997, at 23.
26
See, for example, Florida Department of Juvenile Justice, supra note 15.

juvenile correctional facilities

555

of facility management types in the presence of extensive evaluation and
monitoring of recidivism activity. In the absence of such evaluation and
monitoring, economic theory predicts that the performance of private forprofit facilities in terms of reducing recidivism is likely to be worse.
II.

Background on Florida Department of
Juvenile Justice and Data
A.

Florida Department of Juvenile Justice

The state of Florida has a number of features that make it suitable for a
study on the impact of correctional facility management type on recidivism.
Behind only California, Florida has the largest total number of juvenile
offenders under correctional supervision—on October 27, 1999, juvenile correctional facilities in Florida were maintaining 6,813 offenders in residential
placement27—and it has over 100 facilities to hold these offenders. Crucial
to the aims of this study, Florida is the only state with a significant sample
(n 1 10) of facilities managed each by public, private for-profit, and private
nonprofit entities. Just as crucial, Florida appears to be the only state that
comprehensively tracks and records the postrelease criminal behavior of all
juvenile offenders.28
Of all the U.S. states, Florida also operates the largest number of juvenile
boot camps29—the particular type of juvenile correctional facility that has
aroused the most interest and scrutiny over the past decade.30 Boot camps
are a distinct type of juvenile correctional facility, designed to shock their
youths (or, in boot camp jargon, their “recruits”) into compliance with
military-style discipline. All-male facilities in Florida except in one instance,
juvenile boot camps target youths ages 14 to 18. Procedures for selecting
youths for boot camps vary across counties, but all youths assigned to boot
camps must, in the opinion of DJJ commitment managers, “have medical
and psychological profiles conducive to successfully completing an intensive
work, educational, and disciplinary program.”31 After sanctioning the creation
27

Sickmund, supra note 2, at 2.
For its diligence in tracking and recording youth recidivism, the Florida DJJ has received
recognition from the National Center for Juvenile Justice under a National Institute of Justice
program that identifies best practices in the use of juvenile data. See Florida Department of
Juvenile Justice, Bureau of Data and Research, 2001 Program Accountability Measures
Report: A Two-Year Analysis 2 (March 2001) (on file with the authors).
29
Koch Crime Institute, Juvenile Boot Camps and Military Structured Youth Programs: 2000
Directory (2000).
30
See MacKenzie, Wilson, & Kider, supra note 16.
31
Fla. Stat. Ann. § 985.309(4) (1995). The primary considerations for boot camp selection,
according to DJJ officials, are that the youths do not have any physical problems that would
make the rigorous exercise program dangerous and that they do not suffer from any mental
illnesses requiring psychotropic medication.
28

556

the journal of law and economics

of juvenile boot camps in 1989, the Florida legislature authorized county
governments to implement and operate them, subject to a contractual agreement with the DJJ.32 As a result, the juvenile boot camps operating in Florida
are all run by county sheriff’s departments with oversight from the DJJ.33 In
this way, along with one youth development center that is also county operated, Florida’s boot camps are fundamentally distinct from other juvenile
correctional facilities not only because of their policies and procedures but
also because of sheriff’s departments’ role in managing them. Recognizing
the difference between county and state management, we treat the boot camps
and the youth development center run by sheriff’s departments as a fourth
category of juvenile correctional facility—separate from the public stateoperated, private for-profit, and private nonprofit facilities—in our study,
referring to them as “county” facilities.
B.

Data Overview

The primary data source for this study is the internal database that the
Florida Department of Juvenile Justice maintains for juvenile offenders under
its care. We were granted access to the DJJ’s records on all youths released
from a Florida-based juvenile correctional facility between July 1, 1997, and
June 30, 1999. These data provide complete histories of the experience of
each individual in the Florida juvenile justice system, including records of
all past arrests, adjudications, sentences, and facility assignments. The data
also provide some basic sociodemographic information such as date of birth,
race, and zip code of residence. A total of 16,164 youths are included in the
full sample.
The DJJ tracks the criminal activity of all juvenile offenders for 1 year
after their release so that if a releasee commits a recidivism offense while
still a juvenile (under 18 years of age), our data indicate the date and type(s)
of crime(s) for which he or she was charged and/or adjudicated.34 For our
analysis, we restrict attention to those individuals who were exactly 17 years
of age or younger at the time of release. This restricts the sample to 8,400

32

Fla. Stat. Ann. § 985.309(1) (1995).
Florida Department of Juvenile Justice, Bureau of Data and Research, A Profile of Recruits
Admitted to Department of Juvenile Justice Boot Camp Programs between Inception and March
15, 1997, 1 Res. Digest 1, 1 (May 1997) (http://www.djj.state.fl.us/Research/statsnresearch/
r_digest/issue1.pdf).
34
In the analyses that follow, we use both criminal charges and adjudications to measure
recidivism. Other possible definitions of recidivism include subsequent arrests and correctional
placements, but we lack data on these. Because they represent sentenced crimes,
(re)adjudications provide the most reliable indicator of (known) recidivism behavior; for this
reason, the Florida DJJ also uses adjudications to evaluate recidivism. See, for example, Florida
Department of Juvenile Justice, supra note 15, at 36.
33

juvenile correctional facilities

557

TABLE 1
Individuals and Facilities by Restrictiveness Level and Management Type
Level 1:
Minimum
Risk
State:
Releasees
Facilities
For-profit:
Releasees
Facilities
Nonprofit:
Releasees
Facilities
County:
Releasees
Facilities
Total:
Releasees
Facilities

Level 2:
Low
Risk

Level 3:
Moderate
Risk

Level 4:
High
Risk

Level 5:
Maximum
Risk

Total

≤ age 17

411
8

420
3

896
18

126
3

0
0

1,853
32

≤ age 17

0
0

74
1

821
9

207
8

13
2

1,115
20

≤ age 17

860
24

1,298
19

2,141
44

526
20

0
0

4,825
107

≤ age 17

0
0

0
0

422
8

183
1

2
1

607
10

≤ age 17

1,271
32

1,792
23

4,280
79

1,042
32

15
3

8,400
169

Source.—The data were provided by the Florida Department of Juvenile Justice.

individuals, about whose correctional placements and subsequent charges
and/or adjudications we have complete information.35
Table 1 presents the number of facilities and individuals released from
these facilities by management type and restrictiveness level. Florida law
determines five restrictiveness levels for juvenile correctional facilities, ranging from minimum risk to maximum risk.36 A restrictiveness level is assigned
to each adjudicated individual by his or her judge, reflecting the judge’s
evaluation of the appropriate detention environment for the individual. According to the DJJ, “Higher restrictiveness levels are characterized by tighter
physical security, closer supervision, . . . longer lengths of stay, [and] more
intensive treatment and overlay services such as mental health and drug
treatment.”37 As the table clearly demonstrates, only one of the facilities
below moderate restrictiveness level (level 3) is operated by for-profit or
county management, and none of the maximum restrictiveness level (level
35
For individuals observed multiple times in the sample (repeat recidivists), we include only
their first correctional placement as an observation. Including the subsequent correctional
placements of these youths would have had the effect of oversampling individuals with shorter
sentences.
36
Fla. Stat. Ann. § 985.03(45) (1997). Since the time of this study’s evaluation period, the
Florida DJJ has ceased using the “minimum risk” classification; the DJJ now classifies these
facilities as “nonresidential.” See Florida Department of Juvenile Justice, supra note 15, at
1–6.
37
Florida Department of Juvenile Justice, Bureau of Data and Research, 2000 Outcome
Evaluation Report, app. 1, at 8 (February 2000) (on file with the authors). This report also
contains descriptions of the five restrictiveness levels. Id. at 8–16.

558

the journal of law and economics

5) facilities are operated by nonprofit or state management.38 Moreover, the
minimum restrictiveness level (level 1) facilities are all nonresidential (the
juveniles committed to them are allowed to return home at night) and may
therefore be substantially different in nature from the other facilities. This
suggests that an analysis that compares the performance of facility management types across all restrictiveness levels may not be comparing like facilities. Consequently, in the analyses that follow, we restrict our attention
to the facilities in moderate and high restrictiveness levels (levels 3 and 4)
in order to ensure maximum comparability.39 This restriction decreases the
sample size to 5,322 youths.
Descriptive statistics are presented in Table 2 for the sample of individuals
17 years of age or younger at the time of release from a level 3 or a level
4 facility. Seventy percent of these individuals had a subsequent criminal
charge, and 51 percent had a subsequent adjudication within 1 year of release.
In most cases, the data provide the exact date of all recidivism offenses, so
for 2,660 individuals who had a subsequent adjudication within 1 year of
release, we were able to construct a time-until-recidivism variable (Survival
Time)40 to reflect how long each individual lasted before reoffending. The
mean survival time for the recidivists in the sample is 124 days. For the
3,703 individuals who were charged with a subsequent criminal offense, our
data also indicate the exact nature of the offense(s).41 The most common
offense for which individuals were subsequently charged is assault and battery (21 percent of the sample), followed by felony weapon offenses (15
percent) and burglary (14 percent). The means of these 16 Recidivism—
Charged criminal categories do not sum to .70 because most of the recidivist
youths were charged with multiple crimes.
The criminal history variables in Table 2 reflect the categories of crimes
for which individuals were formally charged within the Florida system prior
to placement in a correctional facility during the evaluation period. The
individual characteristics listed in Table 2 provide basic information on the
youths’ age, gender, race, and length of stay under their correctional facility’s
38
Whereas the Florida DJJ refers to minimum-risk facilities as level 2 facilities, low-risk
facilities as level 4, moderate-risk facilities as level 6, high-risk facilities as level 8, and
maximum-risk facilities as level 10, for simplicity’s sake, we refer to minimum-risk facilities
as level 1 facilities, low-risk facilities as level 2, and so on. Id.
39
In a previous version of this paper, we ran many of the same regressions without this
restriction. Patrick Bayer & David E. Pozen, The Effectiveness of Juvenile Correctional Facilities: Public versus Private Management (Discussion Paper No. 863, Yale Univ. Econ. Growth
Ctr., July 2003). The results without this restriction are largely similar, although restricting the
analysis to the more direct comparison of level 3 and level 4 facilities increases the magnitude
and statistical significance of the results concerning facility management type in all cases.
40
Throughout the paper, the names of all variables used in regressions appear with initial
capital letters.
41
Unfortunately, we were unable to adequately link adjudications after release with corresponding charges. Consequently, we were unable to create distinct variables that characterize
readjudication in specific crime categories.

juvenile correctional facilities

559

care. The typical length of stay in a facility is a little over half a year. The
neighborhood characteristics in this study all come from the 1990 Census of
Population and Housing except for Crime Rate in Zip, which comes directly
from DJJ records on delinquency referrals by zip code. The facility and peer
characteristics in Table 2 include information about facilities’ restrictiveness
levels as well as cost and compositional information. Note that in the discussion that follows, we treat restrictiveness levels as both facility and individual characteristics. This reflects the fact that each juvenile is explicitly
assigned a restrictiveness level by the judge and is subsequently placed in a
correctional facility that handles only individuals in that restrictiveness
category.
The last category of variables in Table 2, facility management type, characterizes facilities as state, for-profit, nonprofit, or county. The state facilities
are those directly controlled by the DJJ, while the county facilities are boot
camps (primarily) and youth development centers managed by county
sheriff’s departments, with oversight from the DJJ.42 The for-profit and nonprofit facilities are all privately operated. While privatization of correctional
facilities can take a variety of forms, for the purposes of this study a facility’s
management type is determined by its operational administration—whatever
entity has total operational administration of the facility, even if it does not
own the facility, is deemed its manager. The extent to which Florida’s juvenile
justice system is privatized is reflected by the fact that 50 percent of the
juvenile offenders in our sample were released from private nonprofit correctional facilities, and 19 percent were released from private for-profit facilities (19 percent and 12 percent were released from public DJJ facilities
and county-operated facilities, respectively). The small number of juveniles
who served sentences in adult correctional facilities does not appear in this
sample.
A final, important set of dummy variables—specifying which of Florida’s
20 geographically defined judicial circuit courts assigned an individual to
the facility under study—is not reported in Table 2.43 The inclusion of these
variables in the analysis below controls for regional variation in an individual’s propensity to recidivate, for variation in prosecutorial, police, and
sentencing practices across jurisdictions, and for any regional variation in
the cost of operating a correctional facility.

42
The state and county facilities are therefore both publicly operated, but we consider them
separately because of their very different management structures. See discussion in Section
IIA supra.
43
For information on Florida’s judicial circuit system for juvenile crime, see Florida Department of Juvenile Justice, Juvenile Justice Organizational Fact Sheet (http://www
.djj.state.fl.us/Research/statsnresearch/factsheets/organization.html); Florida Department of Juvenile Justice, Juvenile Justice Regional and Circuit Map (http://www.djj.state.fl.us/
AboutDJJ/agency/circ_region-key_map.pdf).

TABLE 2
Descriptive Statistics for Sample Used in Analysis
Variable

560

Recidivism:
Recidivism—Charged
Recidivism—Adjudicated
Survival Time—Adjudicated
Recidivism—Charged:
Assault Battery
Felony Weapon
Misd Weapon
Felony Drug
Misd Drug
Felony Sex
Misd Sex
Auto Theft
Burglary
Grand Larceny
Petty Larceny
Robbery
Vandalism
Disorderly Conduct
Escape
Trespassing

N

Mean

5,322
5,322
2,660

.70
.51
123.5

5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322

.21
.15
.01
.11
.09
.01
.00
.10
.14
.10
.11
.05
.05
.04
.02
.08

SD

Definition

.46 1 if client was criminally charged within 1 year of release
.50 1 if client was adjudicated within 1 year of release
87.5
Days elapsed from client’s release date to date of recidivism offense
.41
.36
.11
.31
.29
.12
.06
.31
.35
.29
.32
.22
.23
.20
.13
.28

1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

if
if
if
if
if
if
if
if
if
if
if
if
if
if
if
if

charged
charged
charged
charged
charged
charged
charged
charged
charged
charged
charged
charged
charged
charged
charged
charged

with
with
with
with
with
with
with
with
with
with
with
with
with
with
with
with

assault and battery within 1 year of release
felony weapon offense within 1 year of release
misdemeanor weapon offense within 1 year of release
felony drug offense within 1 year of release
misdemeanor drug offense within 1 year of release
felony sex offense within 1 year of release
misdemeanor sex offense within 1 year of release
auto theft within 1 year of release
burglary within 1 year of release
grand larceny (excluding auto theft) within 1 year of release
petty larceny (excluding auto theft) within 1 year of release
robbery (excluding auto theft) within 1 year of release
vandalism within 1 year of release
disorderly conduct within 1 year of release
unlawful escape (from aftercare) within 1 year of release
trespassing within 1 year of release

561

Criminal history:
Felonies
Felonies 0
Felonies 1
Felonies 2–3
Felonies 4–6
Felonies 7ϩ
Felony Weapon
Misd Weapon
Felony Drug
Misd Drug
Felony Sex
Misd Sex
Auto Theft
Burglary
Grand Larceny
Petty Larceny
Robbery
Vandalism
Disorderly Conduct
Escape
Trespassing
Individual characteristics:
Female
Black
Age at First Offense
Age at Exit
Length of Stay
Neighborhood characteristics:a
Crime Rate in Zip
Per-Cap Inc Race
Percent Own Race in Zip
Unemployment Rate
Incarcerated in Zip

5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322
5,322

5.63
.04
.11
.25
.29
.31
.42
.05
.16
.18
.08
.01
.31
.62
.39
.62
.15
.34
.10
.11
.35

5,322
5,322
5,322
5,322
5,322

.10
.50
12.55
15.79
194.4

5,216
4,621
4,621
4,621
4,621

360
10,488

Number of felony charges on client’s record
1 if zero prior felony charges
1 if one prior felony charge
1 if two or three prior felony charges
1 if four, five, or six prior felony charges
1 if 7 or more prior felony charges
1 if any felony weapon offense charges on client’s record
1 if any misdemeanor weapon offense charges on client’s record
1 if any felony drug offense charges on client’s record
1 if any misdemeanor drug offense charges on client’s record
1 if any felony sex offense charges on client’s record
1 if any misdemeanor sex offense charges on client’s record
1 if any auto theft charges on client’s record
1 if any burglary charges (excluding auto theft) on client’s record
1 if any grand larceny charges (excluding auto theft) on client’s record
1 if any petty larceny charges (excluding auto theft) on client’s record
1 if any robbery charges on client’s record
1 if any vandalism charges on client’s record
1 if any disorderly conduct charges on client’s record
1 if any unlawful escape charges on client’s record
1 if any trespassing charges on client’s record

.31 1 if client is female
.50 1 if client is black
1.98 Client’s age in years at first adjudicated criminal offense
.96 Client’s age in years at exit from facility
109.9
Total length of time (days) spent in facility
263
4,256

.59
.07
112

5.06
.20
.32
.44
.45
.46
.49
.21
.36
.38
.27
.10
.46
.48
.49
.48
.36
.47
.30
.31
.48

Total number of juvenile referrals in client’s home zip code, FY 2000-01
Median per-capita income of client’s racial group in home zip code (1990$)
.33 % of inhabitants in client’s home zip code of same racial group as client, 1990
.03 % unemployment rate in client’s home zip code, 1990
302
Number of people incarcerated in client’s home zip code, 1990

TABLE 2 (Continued)

Variable
Facility and peer characteristics:
Moderate Risk
High Risk
Cost per Release
Facility Size
Percent Black
Average Stay
Same County
Facility management type:
State
For-profit
Nonprofit
County

N

Mean

5,322
5,322
5,322
5,322
5,322
5,322
5,322

.80
.20
24,089
55.90
.52
195.50
.28

5,322
5,322
5,322
5,322

.19
.19
.50
.12

SD

Definition

.40 1 if facility is classified as Moderate Risk (level 3)
.40 1 if facility is classified as High Risk (level 4)
18,972
Average annual cost ($) to the Florida DJJ per client released from facility
82.01 Daily average number of individuals in facility from which individual was released
.50 % of clients released from facility identified as black
80.34 Average length of stay (days) in facility for released clients
.45 1 if facility is located in same county as client’s home county
.39 1 if facility is
.40 1 if facility is
.50 1 if facility is
.32 1 if facility is

operated
operated
operated
operated

directly by the Florida DJJ
by private for-profit management
by private nonprofit management
by a Florida County Sheriff’s Department

Source.—The data were provided by the Florida Department of Juvenile Justice (DJJ), except as otherwise indicated.
Note.—The sample used in the analysis includes individuals age 17 or younger at date of release from a level 3 or level 4 facility. All means are reported for individuals.
Sixty of the 2,720 individuals who were readjudicated within a year are missing information concerning the date of their recidivism offense(s). Neighborhood characteristics
are constructed for Florida zip codes only. Individuals with zip codes from other states are assigned a zero for all neighborhood characteristics, and a dummy variable
denoting that an individual has an out-of-state zip code is included in all regressions. This allows us to maintain the full sample for the regressions and controls for the
potential problem that out-of-state youths are less likely to recidivate in Florida. SD p standard deviation.
a
The neighborhood characteristics all come from the 1990 Census of Population and Housing except for Crime Rate in Zip, which comes directly from DJJ records
on delinquency referrals by zip code.

juvenile correctional facilities

563

TABLE 3
Profile by Facility Management Type
State
Number of facilities
Clients released ≤ age 17
% Recidivism—Charged in first year
% Recidivism—Adjudicated in first year
Mean Survival Time (for those who
readjudicate in first year)
Mean felonies per individual
Mean annual cost to DJJ per release ($)
Mean length of stay (days)
Mean age at exit (years)
Mean facility size (daily average number
of individuals)
Mean % black clients per facility
Mean % male clients per facility
Number of releases by programming type:
Wilderness and work program
Halfway house—male
Halfway house—female
Intensive halfway house—male
Intensive halfway house—female
Serious or habitual offender program
Youth development center
Boot camp
Therapeutic wilderness camp
Special needs program
Sexual offender program
Forestry youth academy
Out of state
Intensive residential treatment

21
996
69.0
51.0

For-Profit

Nonprofit

County

17
1,028
76.4
57.3

64
2,667
68.1
49.4

9
631
65.9
48.5

121.2
117.4
126.1
127.9
(85.0)
(86.4)
(89.9)
(82.7)
5.38
6.40
5.36
5.94
(4.65)
(5.21)
(5.07)
(5.29)
24,807
20,259
24,952
22,750
(23,013)
(10,292)
(20,853)
(5,181)
164
218
194
201
(101)
(106)
(110)
(117)
15.79
15.90
15.68
16.10
(.86)
(.84)
(1.07)
(.70)
19.9
(18.6)
.56
(.13)
.89
(.31)

37.4
(62.4)
.42
(.07)
.94
(.23)

58
697
106
25
0
0
101
0
0
0
0
9
0
0

0
199
24
64
34
49
631
0
0
0
0
0
27
0

19.4
(10.0)
.47
(.14)
.86
(.34)
368
1,238
280
221
23
62
157
0
72
193
37
0
0
16

39.9
(70.9)
.49
(.14)
.96
(.20)
0
0
0
0
0
0
183
448
0
0
0
0
0
0

Source.—The data were provided by the Florida Department of Juvenile Justice (DJJ).
Note.—The sample comprises individuals age 17 or younger at date of release from a level 3 or level
4 facility. Means are reported for individuals in each facility management type except for the mean facility
size, which is reported for all facilities. Standard deviations are in parentheses. Descriptions of each programming type are provided in the Appendix.

C.

Facility Characteristics by Management Type

Table 3 summarizes the key characteristics of the four facility management
types. With 64 facilities, nonprofit is by far the most common management
type, and county is the least common management type with nine facilities.
For-profit facilities have the highest recidivism rates as measured by both
charges and adjudications, and county facilities the lowest. For-profit facilities

564

the journal of law and economics

also have the shortest mean values for Survival Time for those individuals
who are readjudicated. Without any consideration of individual characteristics, therefore, these measures would suggest inferior recidivism performance by the for-profit facilities. Before drawing any conclusions about the
effects of facility management type on recidivism, however, it is important
to control for differences in the backgrounds and offending histories of individuals assigned to each facility management type, as these differences
likely influence their propensity to recidivate.
Indeed, the next set of statistics in Table 3 immediately suggests that the
for-profit facilities serve the most challenging clients, followed by county
facilities: compared to releasees from these two management types, state and
nonprofit releasees have, on average, fewer prior felony charges. The longer
average Length of Stay values and older average Age at Exit values of youths
in for-profit and county facilities also likely derive, at least in part, from
having more serious offenders assigned to these facilities.44 The overall Cost
per Release measure indicates that state and nonprofit facilities are more
expensive for the DJJ on a per capita basis than the county and for-profit
facilities. This cost differential does not seem to reflect the extra security
costs of maintaining a more dangerous inmate population, but it may derive
from economies of scale, as the cheaper county and for-profit facilities are
substantially larger on average than the other management types.45 Whether
facility size should be used as a control variable or viewed instead as the
endogenous choice of the facility manager is a question we explore in detail
below.46
The final set of statistics in Table 3 indicates the prevalence of different
facility programming types across the four management-type categories. The
Florida DJJ classifies all juvenile correctional facilities under its supervision
not only by their restrictiveness level but also by their programming type.
A facility’s programming type refers to the program of services and activities
it offers its inmates; correctional facilities with the same programming type
tend to have similar philosophies, guidelines, and strategies concerning their
44
The longer stays of youths in for-profit facilities may also derive, to an extent, from
deliberate attempts by the corporate operators to maintain occupancy by making it more difficult
for inmates to accumulate the “good behavior” points necessary for early release. See, for
example, Burton A. Weisbrod, The Nonprofit Economy 40 (1988). In Florida, juvenile correctional facility operators have significant discretion to lengthen or shorten their inmates’
length of stay. Judges recommend sentence lengths, but as the DJJ states on its website,
“Juvenile offenders are committed to [correctional] programs for an indeterminate length of
time. They must complete an individually designed treatment plan, based on their rehabilitative
needs, as one of the requirements for release.” Florida Department of Juvenile Justice, Juvenile
Justice Residential Services (http://www.djj.state.fl.us/DJJServices/Residential/residentialindex
.shtml).
45
Note that the mean facility size reported in Table 3 is averaged across facilities rather than
individuals, which explains why it does not match the overall mean reported in Table 2. All
other means in Table 3 are reported for individuals in facilities of the given management type.
46
See Section IIIG infra.

juvenile correctional facilities

565

TABLE 4
Effect of Facility Management Type on Recidivism—Adjudicated:
Ordinary Least Squares Estimation

State
Nonprofit
County
Controls:
Individual characteristics
Criminal history and
restrictiveness level
Neighborhood, facility, and
peer characteristics
Judicial circuit dummies
R2

(1)

(2)

(3)

(4)

(5)

Ϫ.063ϩ
(.038)
Ϫ.079*
(.032)
Ϫ.088*
(.043)

Ϫ.061*
(.027)
Ϫ.066**
(.021)
Ϫ.071*
(.033)

Ϫ.049ϩ
(.026)
Ϫ.056**
(.020)
Ϫ.069*
(.031)

Ϫ.052*
(.026)
Ϫ.064**
(.021)
Ϫ.075*
(.036)

Ϫ.052*
(.023)
Ϫ.060**
(.019)
Ϫ.071*
(.032)

No

Yes

Yes

Yes

Yes

No

No

Yes

Yes

Yes

No
No
.004

No
No
.046

No
No
.058

Yes
No
.061

Yes
Yes
.075

Note.—The sample comprises individuals age 17 or younger at date of release from a level 3 or level
4 facility. For-Profit is the omitted category. Standard errors, in parentheses, account for clustering of
observations at the facility level. The full list of controls, except for the judicial circuit dummies, is shown
in Table 6. N p 5,322.
ϩ
Statistically significant at the .10 level.
* Statistically significant at the .05 level.
** Statistically significant at the .01 level.

treatment of offenders. As a result, including controls for programming type
ought to go a long way toward isolating the specific effects of facility management type on releasees’ recidivism behavior. Table 3 reveals that most
programming types in our sample are run under one or two management
types, with nonprofit facilities host to the greatest number of programming
types and county facilities the least. Only youth development centers are
operated by all four of the management types. Full descriptions of each
programming type are provided in the Appendix.
III.
A.

Results

The Effect of Management Type on Recidivism

To predict recidivism, we use a linear probability model relating recidivism
to facility management type and other control variables, including variables
that characterize criminal history, individual attributes, neighborhood attributes, facility attributes, and judicial circuit assignments. In the analyses that
follow, we consider multiple definitions of recidivism, including (1) whether
a releasee was subsequently adjudicated, (2) whether a releasee was subsequently charged with any crime, and (3) whether a releasee was subsequently
charged in each of 16 specific categories of crime.
Table 4 reports the parameter estimates for the facility management-type

566

the journal of law and economics

variables for various specifications of a regression that uses Recidivism—
Adjudicated as the dependent variable. Standard errors are adjusted for clustering at the facility level throughout the entire analysis. The State, Nonprofit,
and County coefficients are interpreted relative to the omitted category: ForProfit. The specification shown in column 1 essentially restates information
from Table 3 in regression form. Without controlling for any other characteristics of the correctional facilities or their inmates, we find that youths
released from for-profit facilities have the highest probability of being readjudicated—57.3 percent of them are readjudicated within 1 year of release.
Youths released from county and nonprofit facilities are, respectively, 8.8
and 7.9 percentage points less likely to recidivate; these results are statistically
significant at the 5 percent level. Youths released from state facilities are 6.3
percentage points less likely to recidivate (significant at the 10 percent level).
Column 2 of Table 4 controls for observable individual characteristics
including sex, race, age at first offense, age at exit, and length of time in
the facility. Column 3 includes controls for the individual’s criminal history,
including the number of prior felonies, the nature of past crimes, and whether
the restrictiveness level of his or her facility is moderate versus high. This
restrictiveness level, as it reflects the judge’s evaluation of the appropriate
detention environment for the individual, likely picks up aspects of an individual’s criminal history and propensity to recidivate unobserved elsewhere
in the data. Column 4 of Table 4 adds controls for the individual’s home
neighborhood, correctional facility, and peers while in commitment. The
neighborhood controls include the unemployment rate, per capita income in
the neighborhood, racial composition of the neighborhood, and average youth
crime rate in the neighborhood. This last variable is particularly valuable
since it controls for any unobserved characteristics of the neighborhood that
increase the likelihood of all youths in the neighborhood to commit crime.
The facility and peer variables control for the racial composition of other
juveniles in the same facility and the average number of days juveniles spend
in the facility. Finally, column 5 of Table 4 includes additional controls for
the judicial circuit in which the individual was adjudicated prior to being
assigned to the facility under evaluation. The inclusion of this judicial circuit
information, which provides a measure of control for regional variation in
an individual’s propensity to recidivate as well as variation in prosecutorial
and police practices across jurisdictions, has little effect on the parameter
estimates.47
The estimates in column 5 of Table 4 imply that relative to youths released
from for-profit facilities, comparable youths released from state facilities are
47

The inclusion of the judicial circuit information does, however, significantly raise the
predictive power of the model, to an unadjusted R2 value of .075. In general, recidivism at the
individual level is difficult to predict, so these low R2 values are not surprising. See Champion,
supra note 20, at 83 (discussing the low predictive ability of existing models of recidivism
risk).

juvenile correctional facilities

567

TABLE 5
Effect of Facility Management Type on Recidivism—Charged:
Ordinary Least Squares Estimation

State
Nonprofit
County
Controls:
Individual characteristics
Criminal history and
restrictiveness level
Neighborhood, facility, and
peer characteristics
Judicial circuit dummies
R2

(1)

(2)

(3)

(4)

(5)

Ϫ.074*
(.032)
Ϫ.083**
(.023)
Ϫ.104*
(.041)

Ϫ.066**
(.022)
Ϫ.067**
(.015)
Ϫ.085*
(.033)

Ϫ.052**
(.019)
Ϫ.053**
(.014)
Ϫ.083**
(.026)

Ϫ.063**
(.020)
Ϫ.052**
(.016)
Ϫ.086**
(.033)

Ϫ.073**
(.017)
Ϫ.058**
(.015)
Ϫ.084**
(.030)

No

Yes

Yes

Yes

Yes

No

No

Yes

Yes

Yes

No
No
.055

No
No
.085

No
No
.107

Yes
No
.112

Yes
Yes
.124

Note.—The sample comprises individuals age 17 or younger at date of release from a level 3 or level
4 facility. For-Profit is the omitted category. Standard errors, in parentheses, account for clustering of
observations at the facility level. The full list of controls, except for the judicial circuit dummies, is shown
in Table 6. N p 5,322.
* Statistically significant at the .05 level.
** Statistically significant at the .01 level.

5.3 percentage points less likely to be readjudicated within a year (statistically
significant at the 5 percent level); youths released from nonprofit facilities
are 6.0 percentage points less likely to be readjudicated (statistically significant at the 1 percent level); and youths released from county facilities are
7.1 percentage points less likely to be readjudicated (statistically significant
at the 5 percent level). Confirming the intuition provided in Table 3, the
change in parameter estimates from column 1 to column 5 of Table 4 implies
that, relative to the other management types, for-profit facilities are indeed
assigned individuals more likely to recidivate on the basis of observable
characteristics. Yet since the estimates for the facility management type coefficients decline only 15 to 25 percent from column 1 to column 5 , it
appears that observable differences in individual, facility, neighborhood, and
regional variables across facilities operated by different management types
explain only a small portion of the raw differences in recidivism.
Table 5 replicates the analyses performed in Table 4 using Recidivism—
Charged as the dependent variable instead of Recidivism—Adjudicated. The
coefficients in this case are slightly larger than those reported in Table 4,
consistent with the additional likelihood that individuals are charged with a
subsequent crime within a year of release (70 percent) relative to being
readjudicated (51 percent). Again, the coefficients decline only slightly from
column 1 to column 5 as the full set of controls is included in the regression.
The final estimates for the facility-type coefficients in column 5 of Table 5

the journal of law and economics

568

imply that relative to youths released from for-profit facilities, comparable
youths released from state, nonprofit, and county facilities are, respectively,
7.3, 5.8, and 8.5 percentage points less likely to be charged for a criminal
offense within a year of release. All of these estimates are statistically significant at the 1 percent level. These results provide additional support for
the conclusion that state, nonprofit, and county facilities are more effective
at reducing recidivism than are for-profit facilities.
B.

More General Predictors of Recidivism

Table 6 provides estimates for the full specification used to generate column
5 of Table 4.48 A number of interesting results emerge. Compared to otherwise
identical males, female releasees are 16 percentage points less likely to be
readjudicated. Black juveniles are 14 percentage points more likely to recidivate, although it is important to point out that race may stand in for other
socioeconomic differences in this case. The two age variables have a small
but significant effect on youths’ recidivism risk: youths who are released at
a younger age and who committed their first criminal offense at a younger
age are more likely to be readjudicated (within a fixed time period). The
length of a youth’s stay in a correctional facility, however, does not seem to
have a discernible effect on his or her recidivism behavior, likely because a
youth’s restrictiveness level assignment and criminal history already adequately capture the effect of past offenses on his or her propensity to
recidivate.
Recidivism is, by and large, an increasing function of the number of prior
felonies, with youths with seven or more prior felonies 8.0 percentage points
more likely to recidivate than those without any prior felonies—and this is
over and above the effect of the particular prior felonies included in the
analysis. Of the specific crime categories included, a history of felony drug
offenses, auto theft, petty larceny, or prior escapes from correctional facilities
makes an individual especially more likely to recidivate (beyond having an
additional prior felony more generally). The neighborhood characteristics,
including the neighborhood youth crime rate, have very little predictive power
once the set of individual and criminal history variables are incorporated into
the analysis. These neighborhood results provide more evidence that the final
specification shown in column 5 of Table 4 is fairly robust to additional
selection bias, as one would certainly expect these variables to have predictive
48
In its PAM report using the same cohort of offenders as our study, the Florida DJJ identified
four factors as significant to a youth’s probability of recidivating—gender, age at release, age
at first adjudication, and number of prior adjudications—and used only these four risk factors
in its regressions. Florida Department of Juvenile Justice, 2001 Program Accountability Measures Report, supra note 28, at 30. We also found these four factors to have a significant impact
on recidivism risk, but we found the number of prior felonies to be a better predictor of
recidivism than the number of prior adjudications, and we found numerous other factors to
significantly predict recidivism as well.

TABLE 6
Predicting Recidivism—Adjudicated within 1 Year
Variable
Facility management type:
State
Nonprofit
County
Individual characteristics:
Female
Black
Age at First Offense
Age at Exit
Length of Stay (/100)
Criminal history:
Felonies 1
Felonies 2–3
Felonies 4–6
Felonies 7ϩ
Felony Weapon
Misd Weapon
Felony Drug
Misd Drug
Felony Sex
Misd Sex
Auto Theft
Burglary
Grand Larceny
Petty Larceny
Robbery
Vandalism
Disorderly Conduct
Escape
Trespassing
Facility restrictiveness level:
Moderate Risk
Neighborhood characteristics:
Crime Rate in Zip (/1,000)
Per-Cap Inc Race (/1,000)
Percent Own Race in Zip
Unemployment Rate
Incarcerated in Zip (/1,000)
Facility and peer characteristics:
Percent Black
Average Stay (/100)
Same County

Estimate

SD

Ϫ.053
Ϫ.060
Ϫ.071

.023*
.019**
.032*

Ϫ.164
.141
Ϫ.008
Ϫ.048
.004

.025**
.025**
.005*
.008**
.009

.040
.027
.052
.080
Ϫ.005
Ϫ.016
.033
.032
Ϫ.026
Ϫ.012
.053
.023
Ϫ.010
.037
.011
.021
.021
.049
.001

.040
.037
.038
.043ϩ
.016
.028
.019ϩ
.020
.023
.064
.016**
.021
.013
.013**
.021
.013
.023
.020*
.016

Ϫ.010

.013

.004
.007
.023
.043
Ϫ.001

.030
.003**
.028
.394
.023

.015
Ϫ.041
Ϫ.006

Ϫ.075
.015**
.015

Note.—The sample comprises individuals age 17 or younger at date of release from a
level 3 or level 4 facility. For-Profit, Felonies 0, and High Risk are the omitted categories.
Standard errors account for clustering of observations at the facility level. Judicial circuit
dummies are included in the regression. N p 5,322; SD p standard deviation.
ϩ
Statistically significant at the .10 level.
* Statistically significant at the .05 level.
** Statistically significant at the .01 level.

the journal of law and economics

570

power in the absence of sufficient controls for the individual’s own recidivism
risk.49 Of the facility and peer characteristics, exposure to peers who have
longer average commitments has a negative effect on recidivism.
C.

Recidivism in Specific Crime Categories

Using information on the exact offense(s) with which the recidivists in
the sample were charged, Table 7 reports results for 16 separate crime categories across the facility management types. This analysis augments those
of the previous tables, as it reveals the precise criminal areas in which
different management types have better or worse performance. Note that
these totals need not sum to the total effect shown in Table 5, for individuals
can be charged in multiple categories. Also note that full controls, including
the judicial circuit dummies, are included in all regressions.
Significant performance differentials arise in a number of important crime
categories. Relative to otherwise identical youths released from for-profit
facilities, youths released from state facilities are significantly less likely to
be charged with assault and battery, felony weapon offenses, felony sex
offenses, auto theft, petty larceny, or trespassing; youths released from nonprofit facilities are significantly less likely to be charged with felony sex
offenses, auto theft, robbery, or trespassing; and youths released from county
facilities are significantly less likely to be charged with felony drug offenses,
felony sex offenses, auto theft, burglary, petty larceny, robbery, or trespassing. These results suggest that certain management types may be particularly
well suited to dealing with (that is, decreasing the likelihood of ) certain
categories of recidivism crimes: county facilities, for example, seem especially successful at reducing felony drug offenses, burglary, and petty larceny
among their releasees. In no criminal area are releasees from for-profit facilities significantly less likely to recidivate than comparable releasees from
a facility of a different management type.50

49
Indeed, these variables are highly significant in regressions including only neighborhood
controls.
50
In a previous version of this paper, we also considered a specification of the basic recidivism
regression that allowed for interactions between facility management type and five key individual characteristics—gender, race, age at exit, age at first offense, and length of stay—in
order to examine whether certain types of individuals fare better or worse under specific
management types. See Bayer & Pozen, supra note 39, table 11. Our notable findings included
that males do relatively better in for-profit facilities, while females do relatively better in all
other types of facilities; black individuals do relatively better in state and (especially) nonprofit
facilities; older individuals do relatively better in state facilities; and, perhaps most interestingly,
otherwise identical individuals serving longer versus shorter sentences are less likely to recidivate when released from for-profit facilities, so that individuals serving especially long
sentences actually have better recidivism results on average when released from for-profit
facilities. Given the facilities’ ability to lengthen or shorten an offender’s sentenced commitment
period, this finding might mitigate concern over the possible financial incentives of for-profit
facilities to extend commitments artificially. See the discussion in note 44 supra.

juvenile correctional facilities

571

TABLE 7
Recidivism—Charged across Crime Categories

Assault Battery
Felony Weapon
Misd Weapon
Felony Drug
Misd Drug
Felony Sex
Misd Sex
Auto Theft
Burglary
Grand Larceny
Petty Larceny
Robbery
Vandalism
Disorderly Conduct
Escape
Trespassing

State

Nonprofit

County

Ϫ.025ϩ
(.015)
Ϫ.028ϩ
(.015)
.006
(.005)
Ϫ.019
(.013)
Ϫ.007
(.011)
Ϫ.013*
(.005)
Ϫ.001
(.003)
Ϫ.035**
(.011)
Ϫ.004
(.017)
.000
(.014)
Ϫ.017ϩ
(.009)
.000
(.010)
Ϫ.006
(.010)
Ϫ.006
(.009)
.008
(.007)
Ϫ.036**
(.010)

Ϫ.006
(.012)
Ϫ.002
(.013)
.001
(.003)
Ϫ.017
(.012)
Ϫ.004
(.011)
Ϫ.012**
(.004)
Ϫ.002
(.003)
Ϫ.036**
(.010)
Ϫ.020
(.016)
.003
(.013)
Ϫ.009
(.007)
Ϫ.012ϩ
(.007)
.000
(.010)
.001
(.008)
.005
(.005)
Ϫ.021ϩ
(.011)

.004
(.018)
Ϫ.017
(.016)
.002
(.004)
Ϫ.030*
(.014)
Ϫ.011
(.011)
Ϫ.015**
(.005)
.002
(.003)
Ϫ.032*
(.014)
Ϫ.054*
(.021)
Ϫ.004
(.016)
Ϫ.051**
(.014)
Ϫ.020ϩ
(.011)
Ϫ.005
(.012)
.008
(.010)
.030
(.005)
Ϫ.027ϩ
(.015)

Note.—The sample comprises individuals age 17 or younger at date of release from
a level 3 or level 4 facility. For-Profit is the omitted category. Standard errors, in
parentheses, account for clustering of observations at the facility level. The full set of
variables shown in Table 6 and the judicial circuit dummies are included in all regressions. N p 5,322.
ϩ
Statistically significant at the .10 level.
* Statistically significant at the .05 level.
** Statistically significant at the .01 level.

D.

Survival Analysis

Table 8 reports results for various specifications based on an analysis of
survival time. These specifications are able to glean additional information
from the precise timing of recidivism in the first year, which leads to more
precise estimates of the relative effectiveness of the different facility management types. The results are consistent across the three specifications: in
all cases, youths released from state, nonprofit, and county facilities have
significantly lower daily hazard rates than comparable releasees from for-

the journal of law and economics

572

TABLE 8
Survival Time Analysis and Costs to the State
Survival Time Analysis (Dependent Variable:
Recidivism—Adjudicated)

State
Nonprofit
County
R2
N

Proportional
Hazards
(1)

Weibull
Distribution
(2)

Exponential
Distribution
(3)

.875*
(.056)
.838**
(.045)
.810*
(.077)

.874*
(.059)
.834**
(.047)
.811*
(.081)

.872*
(.060)
.831**
(.049)
.809*
(.082)

5,322

5,322

5,322

Cost (Dependent Variable:
Cost per Release ($)):
Ordinary Least Squares
(4)
11,563**
(3,627)
6,123**
(2,169)
2,760
(6,677)
.535
111

Note.—The sample comprises individuals age 17 or younger at date of release from a level 3 or level
4 facility. For-Profit is the omitted category. Standard errors, in parentheses, account for clustering of
observations at the facility level. All specifications include full controls. Full controls are the complete set
of variables shown in Table 6 along with the judicial circuit dummies.
* Statistically significant at the .05 level.
** Statistically significant at the .01 level.

profit facilities. Releasees from state, nonprofit, and county facilities have
hazard rates approximately 13, 17, and 19 percent lower, respectively, than
otherwise identical individuals released from for-profit facilities. These estimates are significant at the 5, 1, and 5 percent levels, respectively, and
again imply that for-profit facilities have systematically worse recidivism
performance.
E.

Costs to the State

While For-Profit correctional facilities perform worse than state, nonprofit,
and county facilities with respect to recidivism, they may still be desirable
as a public policy tool if they come at a lower cost to the state.51 The Florida
DJJ’s data also provide information on each facility’s average cost to the
state per release. Cost in this instance is defined as the annual amount spent
by the Florida DJJ;52 it averages approximately $24,089 per release. With
51
In adult corrections, the Florida legislature requires that for-profit facilities operate at lower
cost to the state. By law, the Florida Correctional Privatization Commission “may not enter
into a contract [with a corporate operator] . . . unless [it] determines that the contract or series
of contracts in total for the facility will result in a cost savings to the state of at least seven
percent over the public provision of a similar facility.” Fla. Stat. Ann. § 957.07(1) (1993).
52
Data were not available on other possible sources of public funding for the correctional
facilities, such as federal grants or local school board allocations. Expenditures by the DJJ
represent the bulk of all the facilities’ public funding, however. Apart from public funding,
private for-profit and nonprofit correctional facilities may have other sources of income available
to them; for-profit facilities can potentially draw on investment income and budget allocations
from their parent corporations, while nonprofit facilities can potentially draw on donations,

juvenile correctional facilities

573

the average individual having spent 194 days in a facility, this works out to
$124 per day. Column 4 of Table 8 reports the results of a regression of the
average cost to the state per release on the full set of control variables. In
this case, the regression is run at the facility level, weighted by the number
of individuals released from each facility and controlling for facility averages
for the full set of controls included in the recidivism regressions.
The results imply that for-profit facilities do in fact generally require a
smaller outlay by the Florida DJJ per release, once the characteristics of the
facilities and the individuals assigned to these facilities are taken into account.
Specifically, the DJJ pays $11,563 more annually per release of a comparable
individual from state facilities and $6,123 more per release from nonprofit
facilities relative to for-profit facilities. These results are statistically significant at the 1 percent level. Comparable releases from county facilities cost
the DJJ $2,760 more annually on average than for-profit releases, although
this result is not statistically significant. The difference between these estimates and the raw differences in costs reported in Table 3 suggest that forprofit facilities are handling more costly juveniles in terms of observable
characteristics. The coefficients on the controls in the cost regression imply,
for example, that individuals with more felonies are more costly to house in
a correctional facility. Some of these additional costs may be related to the
increased difficulty of rehabilitating these individuals, while other costs may
be related to greater security precautions they require.
F.

Programming Type

In order to understand better why recidivism performance and costs differ
across management types, Table 9 shows results for our main specifications
after controlling for facility programming type—each facility’s selfdetermined set of philosophies and practices. Column 1 of Table 9 is comparable to column 5 of Table 4; column 2 is comparable to column 5 of
Table 5; column 3 is comparable to column 1 of Table 8; and column 4 is
comparable to column 4 of Table 8. Once again, For-Profit is the omitted
management-type category, and Boot Camp is the omitted programmingtype category. In all four of the specifications, the results for state and nonprofit facilities remain quite similar to those reported in the previous tables:
these management types continue to have significantly superior recidivism
performance to that of for-profit facilities using all measures, and they continue to cost more to the state per comparable individual released. In the
case of county facilities, on the other hand, adding the programming-type
controls changes the results substantially. The recidivism performance of
county facilities becomes statistically indistinguishable from that of for-profit
grants, and endowment income. Since we do not have information on these possible extragovernmental income flows, our results reflect only facilities’ cost effectiveness from the
perspective of the Florida DJJ.

TABLE 9
Robustness Analysis: Controlling for Programming Type
Recidivism—Adjudicated: Recidivism—Charged: Recidivism—Adjudicated: Cost per Release ($):
Ordinary Least Squares Ordinary Least Squares
Proportional Hazards
Ordinary Least Squares
(1)
(2)
(3)
(4)

574

Management type:
State
Nonprofit
County
Programming type (relative to boot camps (448)):
Halfway house—male (2,134)
Youth development center (1,072)
Wilderness and work program (426)
Halfway house—female (410)
Intensive halfway house—male (310)
Special needs program (193)

Ϫ.060*
(.030)
Ϫ.066*
(.028)
Ϫ.009
(.025)

Ϫ.082**
(.020)
Ϫ.066**
(.017)
Ϫ.007
(.020)

.877
(.071)
.843*
(.068)
.9930
(.088)

12,899**
(3,720)
8,048*
(3,544)
Ϫ25,923**
(6,154)

.101*
(.044)
.101*
(.048)
.091ϩ
(.047)
.130*
(.060)
.127*
(.055)
.052
(.065)

.129**
(.036)
.117**
(.040)
.136**
(.038)
.102
(.062)
.137**
(.047)
.058
(.051)

1.355*
(.189)
1.420*
(.216)
1.327ϩ
(.196)
1.476
(.354)
1.495*
(.292)
1.169
(.201)

Ϫ39,650**
(6,658)
Ϫ31,945**
(6,384)
Ϫ42,266**
(7,111)
Ϫ46,049**
(11,773)
Ϫ53.981**
(11,687)
Ϫ40,035**
(7,451)

Serious or habitual offender program (111)
Therapeutic wilderness camp (72)
Intensive halfway house - female (57)
Sexual offender program (37)
Out of state (27)
Intensive residential treatment (16)
Forestry youth academy (9)
N
R2

.071
(.108)
.145
(.089)
.112
(.120)
.079
(.072)
Ϫ.035
(.078)
.195*
(.078)
.017
(.051)
5,322
.079

.106ϩ
(.062)
.124
(.095)
.140*
(.068)
Ϫ.066
(.059)
.048
(.066)
.128ϩ
(.066)
Ϫ.043
(.048)
5,322
.128

1.291
(.461)
1.456
(.385)
1.373
(.721)
1.154
(.305)
.930
(.236)
1.672ϩ
(.441)
.944
(.150)
5,322

Ϫ44,820**
(11,896)
9,631
(16,284)
Ϫ63,485**
(16,695)
Ϫ18,641
(28,851)
Ϫ61,106**
(13,291)
6,173
(12,755)
3,787
(9,038)
111
.762

575

Note.—The sample comprises individuals age 17 or younger at date of release from a level 3 or level 4 facility. For-Profit and Boot Camp are the omitted categories.
Standard errors, in parentheses, account for clustering of observations at the facility level. Descriptions of each programming type are provided in the Appendix. The
number shown in parentheses following each programming type is the total number of individuals age 17 or younger released during the evaluation period. All specifications
include full controls. Full controls are complete set of variables shown in Table 6 along with the judicial circuit dummies.
ϩ
Statistically significant at the .10 level.
* Statistically significant at the .05 level.
** Statistically significant at the .01 level.

576

the journal of law and economics

facilities, but their cost per release becomes far lower—an annual average
of $26,047 less per comparable release relative to the cost for for-profit
facilities. It is important to point out, however, that since all county facilities
have only one of two programming types, youth development center or boot
camp, these results simply reflect how the single youth development center
operated by a county compares to the youth development centers operated
by the other management types.
Interpreted relative to the case of boot camps, the results in the lower part
of Table 9 indicate the relative effectiveness of the different programming
types. The recidivism performance of boot camps looks very good, as all of
the other major programming types have statistically inferior recidivism performance on at least one measure and in most cases on all three. These better
returns in terms of recidivism do come at an increased cost, though, as the
other programming types typically cost the state much less than boot camps.
More generally, it is important to keep in mind that all of the boot camps
operated in Florida are county facilities. Thus, to evaluate the performance
of Florida’s boot camps relative to other programming types run by a different
management type, one must weigh the combined effect of Boot Camp versus
another programming type and County versus another management type. In
most comparisons, Florida’s boot camps give rise to substantial decreases in
recidivism rates at an increased cost easily justifiable by a cost-benefit
analysis.
G.

Facility Size

In order to shed further light on why recidivism performance and costs
differ across management types, Table 10 explores the impact of including
facility size as a further control.53 As Table 3 makes clear, one of the key
differences between for-profit and county facilities versus state and nonprofit
facilities is the relatively large size of for-profit and county facilities. In
conducting our analyses, therefore, an interesting question arises concerning
how to treat facility size. On the one hand, facility size might be thought of
as an endogenous choice of the facility manager. If increased size allows the
facility manager to reduce per capita costs at the possible expense of greater
rates of recidivism, it would not be appropriate to include facility size as a
control variable. On the other hand, if one assumes that facility size is correlated with unobserved individual characteristics or is simply predetermined
53
As yet another way to explore the possible factors behind our recidivism results, we also
conducted an analysis of the “best and worst” facilities in our sample, where we compared
facilities’ actual Recidivism—Adjudicated rates to the rates predicted by the regression reported
in Table 6 without including the facility management variables, and then ranked the facilities
on the basis of that differential. These results were broadly consistent with the relative prevalence and overall recidivism performance of each management type, and they indicated that
our primary recidivism results are not driven by a few particularly excellent or substandard
facilities in one or more of the management-type categories.

TABLE 10
Robustness Analysis: Controlling for Programming Type and Facility Size

State
Nonprofit
County
Facility Size (/100)
N
R2

Recidivism—Adjudicated:
Ordinary Least Squares
(1)

Recidivism—Charged:
Ordinary Least Squares
(2)

Recidivism—Adjudicated:
Exponential Distribution
(3)

Cost per Release ($):
Ordinary Least Squares
(4)

Ϫ.049
(.030)
Ϫ.056ϩ
(.029)
Ϫ.054*
(.024)
.039**
(.012)
5,322
.079

Ϫ.078**
(.020)
Ϫ.062**
(.018)
Ϫ.024
(.027)
.015
(.014)
5,322
.128

.879
(.076)
.864ϩ
(.075)
.849ϩ
(.075)
1.125*
(.052)
5,322

9,378**
(3,361)
4,805
(3,086)
Ϫ12,476*
(5,459)
Ϫ11,760**
(4,087)
111
.793

Note.—The sample comprises individuals age 17 or younger at date of release from a level 3 or level 4 facility. For-Profit is the omitted category. Standard errors,
in parentheses, account for clustering of observations at the facility level. All specifications include full controls. Full controls are complete set of variables shown
in Table 6 along with the judicial circuit dummies, facility size, and the variables characterizing programming type.
ϩ
Statistically significant at the .10 level.
* Statistically significant at the .05 level.
** Statistically significant at the .01 level.

578

the journal of law and economics

by the DJJ, one might want to use facility size as an extra control variable.
Thus, in order to provide a clear picture of the extent to which this latter
assumption would affect our results, Table 10 repeats the analyses of Table
9 controlling for facility size.
As one might have expected, smaller facilities produce significantly better
results with respect to recidivism at a significantly higher cost. Yet while
part of the differences between the performance of for-profit facilities and
the performance of state and nonprofit facilities can therefore be explained
by size differences, size differences do not explain very much; when facility
size is added as a control, the results for state and nonprofit facilities are
broadly similar to before. The magnitudes and statistical significance of
their coefficients generally decline relative to their levels in Table 9, but
these two management types continue to outperform for-profit facilities by
a wide margin in terms of recidivism while costing more to the DJJ. In a
more considerable change from Table 9, county facilities’ Recidivism—
Adjudicated performance becomes significantly better than that of for-profit
facilities with the inclusion of the facility size control. The cost savings
offered to the DJJ by county facilities over for-profit facilities, meanwhile,
is halved but remains substantial at $12,476 annually per comparable individual released.
Overall, then, when facility size and programming type are both added to
the full set of controls, state and nonprofit facilities continue to have substantially better recidivism performance than for-profit facilities, but they
come at a higher cost to the state. County facilities also offer better recidivism
performance than for-profit facilities, and they come at a lower cost to the
state. County facilities therefore seem clearly preferable to for-profit facilities
from the perspective of the DJJ, while a more precise cost-benefit analysis
is necessary to determine whether the recidivism benefits of state and nonprofit facilities justify their additional costs relative to for-profit facilities.
IV.

Cost-Benefit Analysis

Taken together, the results presented above imply that for-profit management leads to a statistically significant increase in recidivism but at significantly lower costs when compared to nonprofit and especially to state facilities. This raises an interesting question for public policy: are the immediate
cost savings offered by for-profit facilities enough to justify the future costs
associated with increased recidivism? To provide a better sense of the magnitudes involved in this trade-off, Table 11 presents a comparison of the costs
and benefits of the different management types.
Our cost-benefit analysis is conducted for the mean individual (15.8 years
of age) in the sample used in the analyses presented above (17 years of age
or younger at the time of release from a moderate- or high-risk facility). We
calculate the expected number of future days that this individual, having been

juvenile correctional facilities

579

TABLE 11
Cost-Benefit Analysis by Management Type
For-Profit
Predicted daily survival rate
Predicted daily hazard rate
Expected cost difference per release ($)
Expected future number of days in
correctional facility:
1 year from release
At age 18 (2.2 years)
5 years from release—assumption 1
5 years from release—assumption 2
Expected additional days out of
confinement relative to for-profit:
1 year from release
At age 18 (2.2 years)
5 years from release—Assumption 1
5 years from release—Assumption 2
Value ($) of additional day out of
confinement to justify choice over
for-profit:
1 year from release
At age 18 (2.2 years)
5 years from release—assumption 1
5 years from release—assumption 2

State

.99775
.00225
11,563
90.8
223.0
432.5
533.6

Nonprofit

.99804
.00196

County

.99813
.00187
6,123

.99819
.00181
2,760

81.5
201.4
389.1
483.5

78.4
194.0
374.3
466.2

76.6
189.9
366.1
456.5

9.2
(4.4)
21.6
(10.4)
43.5
(20.9)
50.1
(24.3)

12.4
(3.6)
29.0
(8.5)
58.2
(17.2)
67.4
(19.3)

14.2
(6.6)
33.2
(15.6)
66.4
(30.3)
77.0
(35.4)

1,253
(526)
535
(225)
266
(112)
231
(98)

496
(163)
211
(70)
105
(30)
91
(29)

196
(302)
84
(129)
42
(63)
36
(55)

Note.—Standard errors (in parentheses) were bootstrapped using the standard errors reported for the
corresponding parameters. Assumption 1: recidivism rates fall by one-third between the ages of 18–21
relative to the ages of 16–18. Assumption 2: differences in recidivism rates remain as they are for the
juveniles in our sample.

released from a facility operated by each of the four management types,
would spend in a correctional facility. We then divide the estimated cost
differential between management types by this result to arrive at a cutoff
value of an expected future day in a correctional facility needed to justify
one type of facility over another. To determine which of these facility types
should be preferred, this cutoff value must be compared to the social benefit
of avoiding an additional day in confinement, including the avoided costs of
the crime and prosecution as well as the cost of confinement.54
54
For estimates of the costs of juvenile crime (and hence also of the benefits of reduced
juvenile recidivism rates), see, for example, Howard N. Snyder & Melissa Sickmund, Juvenile
Offenders and Victims: 1999 National Report 82–83 (1999); David A. Anderson, The Aggregate
Burden of Crime, 42 J. Law & Econ. 611 (1999).

580

the journal of law and economics

For this analysis, we seek to provide a lower bound regarding the benefits
of other facility management types versus for-profit management and therefore make assumptions that are either neutral to or favor for-profit management in comparison to other management types. First, the calculation of
future days in a correctional facility assumes a sentence length of 195 days,
which is equal to the mean sentence length in the current sample. This is
clearly a lower bound, given that sentences will subsequently be assigned to
older individuals with more criminal experience. Second, the calculation
assumes that once released from each type of facility, an individual, while
still a juvenile, recidivates according to the mean hazard rate associated with
that management type, conditional on the full set of controls used in the
analysis reported in Table 8. Once over 18 years of age, the mean individual
becomes an adult in Florida’s criminal justice system; we report results
following two assumptions regarding recidivism rates as an adult. In the first
instance (assumption 1), we assume that recidivism rates fall by one-third
between the ages of 18 to 21 relative to the rates of individuals ages 16 to
18. This assumption is consistent with the decline over time in recidivism
rates for all state prison releasees in Florida and serves to minimize differences in future recidivism rates across facility management types.55 In the
second instance (assumption 2), we assume that differences in recidivism
rates remain as they are for juveniles in our sample. Finally, we allow for
the possibility of multiple recidivistic commitments to a correctional facility.
In this case, we continue to apply the same hazard rate to an individual that
was associated with the original facility management type. This too is a
conservative assumption regarding differences between for-profit and other
management types, for we assume that the fact that an individual recidivates
once has no effect on his or her subsequent recidivism rate, future sentence
length, or future costs of confinement.
The results presented in Table 11 indicate that individuals released from
for-profit facilities are expected to spend an average of 223 additional days
in a correctional facility by their 18th birthday, 2.2 years in the future. This
figure reflects the fact that well over half of such individuals recidivate within
a year and, consequently, that a significant number will have recidivated
twice within this 2.2-year period. A comparable individual released from a
nonprofit facility is expected to spend only 194 future days in a correctional
facility. Assuming that recidivism rates decline by one-third between the ages
of 18 and 21 (assumption 1), which is our preferred assumption, we estimate
that individuals released from state, nonprofit, and county facilities will spend
an average of 44, 58, and 66 fewer future days in a correctional facility,
respectively, relative to comparable individuals released from for-profit fa55
Florida Department of Corrections, Recidivism Report: Inmates Released from Florida
Prisons July 1995 to June 2001, at 10–12, chart 1 (July 2003) (http://www.dc.state.fl.us/pub/
recidivism/2003/full.pdf).

juvenile correctional facilities

581

cilities. The corresponding numbers generated under assumption 2 are even
larger.
Table 11 also reports the cutoff value of an additional day spent out of a
correctional facility needed to justify the choice of state, nonprofit, and county
facilities over for-profit facilities. Following the more conservative assumption 1, the results imply that 5 years after release the cutoff value of an
additional day spent out of a correctional facility needs to be $266, $105,
and $42 to justify the choice of a state, nonprofit, and county facility, respectively, over a for-profit facility. To provide a benchmark from our data,
the average daily direct cost to the Florida DJJ for the juveniles in our sample
is $124 per day in a correctional facility. Thus, even without accounting for
any of the other possible benefits (both private and social) of avoiding recidivism as well as the direct costs of apprehension and prosecution, county
facilities emerge as clearly preferable for the DJJ to for-profit and state
facilities on the basis of cost savings related to future time spent in a correctional facility.
The results of our analysis for the nonprofit versus for-profit comparison
come within a standard deviation of the benchmark $124 cost to the state
of a future day spent in a correctional facility. This benchmark, however,
does not include any of the benefits of a reduction in crime except for cost
savings from avoided confinement. Moreover, as described above, our methodology is conservative in many other dimensions. Consequently, the analysis
clearly points to the conclusion that the nonprofit facilities are preferable to
the for-profit ones from the point of view of the state. it is not possible to
reach a strong conclusion regarding state management versus for-profit management since it would depend on what the full social benefits of avoiding
future recidivism are. The results of Table 11 do imply, however, that nonprofit and county facilities outperform state facilities, as they provide slightly
better recidivism performance at a substantially lower cost.56
V.

Conclusion

The results of our analysis indicate that for-profit management has a statistically significant impact on recidivism as measured by both 1-year recidivism
rates (approximately 5 to 8 percent higher than the other management types
in terms of adjudications and charges) and daily hazard rates (approximately
13 to 19 percent higher). However, for-profit management is also associated
with significantly lower costs when compared to nonprofit and especially to
state-operated (Department of Juvenile Justice–operated) facilities—about
56

A comparable cost-benefit analysis based on the results reported in Table 10, which control
for facility size and programming type, leads to the same pattern of cutoff values as those in
Table 11. This finding again indicates the clearly superior performance of county management
over for-profit management and provides a point estimate of the cutoff value between nonprofit
and for-profit management remarkably similar to that reported in Table 11.

582

the journal of law and economics

$6,000 and $11,500 per release less than the average cost for a comparable
release from a nonprofit and a state-operated facility, respectively.
Consistent with economic theory,57 we thus find a trade-off between the
benefits of reduced recidivism provided by nonprofit and state management
on the one hand and the cost savings associated with for-profit management
on the other; in order to determine the relative attractiveness of each correctional facility management type, this trade-off requires a more careful
analysis of the magnitudes of these effects.58 Using a series of conservative
assumptions concerning the future impact of the estimated differences in
recidivism rates across management types, our cost-benefit analysis implies
that the short-run savings offered by for-profit facilities over nonprofit facilities are reversed in the long run due to increased recidivism rates. This
conclusion holds even if one ignores all the other possible benefits, both
private and social, of reducing criminal activity and measures the benefits
of reduced recidivism as only the avoided costs of additional confinement.
County management outperforms for-profit management both in terms of
recidivism and in terms of (direct) cost and therefore appears clearly preferable for the state of Florida. While state management yields both worse
recidivism performance and higher costs than nonprofit and county management, the cost-benefit analysis is inconclusive regarding a comparison of
state and for-profit management.
While the possibility remains that unobserved differences in the populations served by each facility management type explain some of the differences
attributed to these management types, several components of our analysis
provide assurance that such unobserved differences would be unlikely to
change the qualitative nature of our results. First, this study includes far more
controls for individual, criminal history, judge-assigned restrictiveness level,
judicial circuit, facility programming type, neighborhood, and peer characteristics than any previous study of juvenile recidivism. Considering that the
inclusion of these numerous important controls reduces the raw differences
between management types by only 10 to 30 percent across multiple recidivism measures, it appears highly improbable that remaining unobserved
differences in commitment populations explain the estimated differences between facility management types.
These findings have immediate implications for public policy. Certainly,
the Florida Department of Juvenile Justice should continue to expand the
role of county-operated facilities (which are usually boot camps) and private
57

See discussion in Section IB supra.
Of course, in determining the relative attractiveness of any facility management type,
corrections officials might take into account many considerations other than recidivism and
cost, such as facilities’ quality of confinement and the quality of their educational, vocational,
rehabilitative, and health services. As the Florida DJJ recognizes, however, recidivism and cost
are fundamental considerations in shaping a corrections system. See, for example, Florida
Department of Juvenile Justice, supra note 15.
58

juvenile correctional facilities

583

nonprofit facilities in its portfolio of correctional facilities relative to the role
of private for-profit and state-operated facilities. Given the many legal, political, and ethical complications associated with profit-seeking correctional
facilities,59 it seems easy to recommend a movement away from for-profit
facilities in Florida’s juvenile justice system. Moreover, given that the Florida
DJJ explicitly evaluates facilities on the basis of recidivism and costs—
thereby providing some incentive for at least the worst-performing facilities
to reduce recidivism—the performance of for-profit facilities in jurisdictions
that do not collect data and evaluate facilities on the basis of recidivism is
likely to be even worse.
The results also suggest that certain facility management types may be
particularly well suited to decreasing the likelihood of certain categories of
recidivism crimes. County-operated facilities, for example, are especially
successful at reducing felony drug offenses, burglary, and petty larceny
among their releasees, while state-operated facilities are especially successful
at reducing future assault and battery and felony weapon offenses. For-profit
facilities do not exhibit superior recidivism performance in any of the 16
criminal categories, which suggests that their weaknesses in reducing recidivism are systematic.
APPENDIX
Programming Type Descriptions60
Wilderness and work programs provide services for youths committed by the juvenile
court as well as youths tried as adults and sentenced back to the juvenile system.
The programs maintain a population of approximately 20 to 25 males aged 15 to
18 years. They operate in an environmentally secure setting in a remote, isolated
rural location. They provide academic and vocational training with moderate overlay services such as mental health and drug abuse treatment. The programs also
emphasize outdoor activities, labor-intensive work projects, and behavior management. The designed length of stay is 12 months.
Halfway houses each serve a population of approximately 15 to 30 youths of the
same gender aged 14 to 18 years. These programs serve youths who have committed first-degree misdemeanors, felonies, or similar offenses and are classified
59
For representative scholarly critiques of private prisons, see David Shichor, Punishment
for Profit: Private Prisons/Public Concerns (1995); Eric Bates, Private Prisons, The Nation,
January 5, 1998, at 13; John J. DiIulio, Jr., What’s Wrong with Private Prisons, 92 Pub. Int.
66 (1988); Joseph E. Field, Making Prisons Private: An Improper Delegation of a Governmental
Power, 15 Hofstra L. Rev. 649 (1987); J. Robert Lilly & Paul Knepper, The CorrectionsCommercial Complex, 39 Crime & Delinq. 150 (1993); Robert G. Porter, The Privatisation of
Prisons in the United States: A Policy That Britain Should Not Emulate, 29 Howard J. Crim.
Just. 65 (1990); Mick Ryan & Tony Ward, Privatization and the Penal System: Britain Misinterprets the American Experience, 14 Crim. Just. Rev. 1 (1989).
60
Except for the descriptions of boot camps, forestry youth academies, and out-of-state
programs, the following descriptions of programming types all come from one Florida Department of Juvenile Justice report. Florida Department of Juvenile Justice, 2000 Outcome
Evaluation Report, supra note 37, app. 1, at 11–15 (2000). Some descriptions are adapted
slightly.

584

the journal of law and economics

as moderate risks to public safety. The programs provide an intentional therapeutic
environment based on control theory, structured learning, and behavior management techniques that emphasize social skills, academics, prevocational and vocational training, and life skills. The designed length of stay is 3 to 6 months.
Intensive halfway houses each serve a population of approximately 15 to 30 offenders
aged 14 to 18 years. These programs provide services at the high restrictiveness
level similar to those provided by a halfway house at the moderate restrictiveness
level. In addition to more intense physical, staff, and procedural security, there is
also increased structure and behavioral management to maximize protection of the
public. The designed length of stay ranges from 6 to 9 months.
Serious or habitual offender programs each provide services for a population of no
more than 25 males who are serious or habitual offenders. These programs are
designed for youths aged 14 to 19 years. These programs employ physical security
features and procedures to ensure protection of the public. The services provided
by these programs are statutorily mandated, and the designed length of stay is 9
to 12 months.
Youth development centers include programs formerly known as training schools.
These large institutions serve more than 100 youths of the same gender aged 13
to 18 years who have committed felonies of violent misdemeanors. Youths who
have committed sex offenses are not eligible for admission. These programs provide
a high degree of physical and staff security. Minimum security features include a
security perimeter fence at least 12 feet high with an inside overhang or razor
wire, external facility doors that are accessed electronically or through the use of
a key, passage doors that are hardware secure, and windows that must be secure
and constructed of break-resistant material. These programs provide services
through a multidisciplinary approach within an institutional setting. Services include behavior management, academics, vocational training, mental health and
substance abuse treatment, physical fitness activities, and health care. The designed
length of stay is 9 months.
Boot camps61 are designed as a series of phases. Each camp includes a high-intensity
intake on the first day. The next 10 to 14 days, called the forming stage, is an
orientation period during which the recruits are oriented to the basics of military
protocol and the rules and regulations of the facility. Once orientation has been
completed, educational, mental health, and other overlay services are added. Other
features of boot camps include a silence rule (recruits may not speak to one another
except under special circumstances); individual rooms; military bearing, discipline,
drill, ceremony, and physical training; long, structured days with little or no free
time or recreation; a minimum of 5 hours a day of education; use of different
colored hats to designate progress through the program; and transition programs
at the end of recruits’ stay in order to prepare them for return to “civilian life.”
Boot camps are typically all male.
Therapeutic wilderness camps each provide services for a population of approximately 30 to 50 emotionally disturbed all-male or all-female youths, generally
aged 11 to 16 years. The programs are designed to provide a camp environment
that emphasizes outdoor experiential learning, structured peer interaction, teamwork, and personal accountability. The designed length of stay is 12 to 18 months.
Special needs programs each serve a population of approximately 10 to 30 youths
of the same gender aged 13 to 18 years. These programs provide specialized clinical
treatments services in the areas of substance abuse, mental health, developmental
61
Florida Department of Juvenile Justice, Bureau of Data and Research, Determining Best
Practices in Florida’s Juvenile Boot Camps 12 (November 2000) (on file with the authors).

juvenile correctional facilities

585

disability, or sexual behavior dysfunction. The designed length of stay is 4 to 6
months.
Sexual offender programs provide services for approximately 20 to 30 male sexual
offenders aged 14 to 18 years. These programs provide a continuum of treatment
services specifically tailored to the needs of sexual offenders and focus on overcoming denial, treatment of maladaptive thought patterns, and alleviating or reducing dysfunctional sexual behaviors. The designed length of stay is 6 to 12
months.
The Forestry Youth Academy62 functions as a secondary, optional phase to boot camp
facilities and operates under the arm of the DJJ. Candidates for the 2-year program
are approximately 16 years old, have successfully graduated from a boot camp
facility, and have been prescreened by the DJJ. At this moderate-risk facility,
education and life skills are linked with discipline and teamwork. Participants
receive credit-bearing vocational training in practical forestry skills utilizing the
latest technology and nontraditional methodology to ensure the greatest chance of
employability upon graduation. Graduates leave the program with a G.E.D. or high
school diploma and a vocational certification.
Out-of-state programs refer to The Glen Mills Schools,63 a private, residential school
in Pennsylvania for court-adjudicated male delinquents aged 15 to 18 years. Students are referred to the school by state departments of juvenile justice throughout
the country, including the Florida DJJ. The school has two basic mandates for
students: to change behavior from antisocial to prosocial and to develop life skills
that will help sustain this change. Each student receives year-round instruction
designed to meet his educational needs.
Intensive residential treatment programs are for offenders aged 10 to 13 years. These
programs provide services for a population of approximately 25 young males who
have committed serious felony offenses, including capital or life felonies. Statutory
provisions allow the programs to retain youths until age 21 when necessary. The
programs provide intensive treatment services that address the areas of education,
behavior management, substance abuse, mental health, sexual behavior dysfunction, life skills, gang-related behavior, and family issues. The designed length of
stay is 9 to 12 months.

Bibliography
Anderson, David A. “The Aggregate Burden of Crime.” Journal of Law and
Economics 42 (1999): 611–42.
Archambeault, William G., and Deis, Donald R., Jr. Cost Effectiveness Comparisons of Private versus Public Prisons in Louisiana: A Comprehensive
Analysis of Allen, Avoyelles, and Winn Correctional Centers. Baton Rouge:
Louisiana State University, School of Social Work, 1996.
Armstrong, Gaylene Styve. Private vs. Public Operation of Juvenile Correctional Facilities. New York: LFB Scholarly Publishing, 2001.
Avio, Kenneth L. “The Economics of Prisons.” European Journal of Law
and Economics 6 (1998): 143–75.
Bakal, Yitzhak, and Lowell, Harvey. “The Private Sector in Juvenile Cor62

Florida Division of Forestry, Forestry Youth Academy, Helping Reclaim Florida’s Youth
(http://www.fl-dof.com/youth_academy.html).
63
The Glen Mills Schools (http://www.glenmillsschool.org).

586

the journal of law and economics

rections.” In Juvenile Justice and Public Policy: Toward a National
Agenda, edited by Ira M. Schwartz, pp. 196–213. Lexington, Mass.: Lexington Books, 1992.
Bates, Eric. “Private Prisons.” Nation, January 5, 1998, pp. 11–18.
Bayer, Patrick, and Pozen, David E. The Effectiveness of Juvenile Correctional Facilities: Public versus Private Management. Discussion Paper
No. 863. New Haven, Conn.: Yale University, Economic Growth Center,
2003.
Borna, Shaneen. “Free Enterprise Goes to Prison.” British Journal of Criminology 26 (1986): 321–34.
Champion, Dean J. Measuring Offender Risk: A Criminal Justice Sourcebook.
Westport, Conn.: Greenwood Press, 1994.
Curran, Daniel J. “Destructuring, Privatization, and the Promise of Juvenile
Diversion: Compromising Community-Based Corrections.” Crime and Delinquency 34 (1988): 363–84.
DiIulio, John J., Jr. “What’s Wrong with Private Prisons.” Public Interest 92
(1988): 66–83.
DiPiano, John G. “Private Prisons: Can They Work? Panopticon in the
Twenty-first Century.” New England Journal on Criminal and Civil Confinement 21 (1995): 171–202.
Field, Joseph E. “Making Prisons Private: An Improper Delegation of a
Government Power.” Hofstra Law Review 15 (1987): 649–75.
Florida Department of Corrections. Recidivism Report: Inmates Released
from Florida Prisons July 1995 to June 2001. Tallahassee: Florida Department of Corrections, 2003.
Florida Department of Corrections, Bureau of Research and Data Analysis.
Preliminary Assessment of a Study Entitled: “A Comparative Recidivism
Analysis of Releasees from Public and Private Prisons in Florida.” Tallahassee: Florida Department of Corrections, 1998.
Florida Department of Juvenile Justice. Bureau of Data and Research. “A
Profile of Recruits Admitted to Department of Juvenile Justice Boot Camp
Programs between Inception and March 15, 1997.” Research Digest 1
(1997): 1–2.
Florida Department of Juvenile Justice. Bureau of Data and Research. Determining Best Practices in Florida’s Juvenile Boot Camps. Management
Report No. 2000-11. Tallahassee: Florida Department of Juvenile Justice,
2000.
Florida Department of Juvenile Justice. Bureau of Data and Research. 2000
Outcome Evaluation Report. Management Report No. 2000-7. Tallahassee:
Florida Department of Juvenile Justice, 2000.
Florida Department of Juvenile Justice. Bureau of Data and Research. 2001
Program Accountability Measures Report: A Two-Year Analysis. Management Report No. 2001-4. Tallahassee: Florida Department of Juvenile
Justice, 2001.

juvenile correctional facilities

587

Florida Department of Juvenile Justice. Justice Research Center, Inc. The
2003 PAM Report: A Two-Year Analysis. Management Report No. 03-01.
Tallahassee: Florida Department of Juvenile Justice, 2002.
Foucault, Michel. Discipline and Punish: The Birth of the Prison, translated
by Alan Sheridan. New York: Pantheon Books, 1977.
Gaes, Gerald G.; Camp, Scott D.; and Saylor, William G. “The Performance
of Privately Operated Prisons: A Review of Research.” Appendix 2 in
Private Prisons in the United States: An Assessment of Current Practice,
edited by Douglas McDonald, Elizabeth Fournier, Malcolm Russell-Einhorn, and Stephen Crawford. Cambridge, Mass.: Abt Associates, Inc.,
1998.
Greenwood, Peter W.; Turner, Susan; and Rosenblatt, Kathy. Evaluation of
Paint Creek Youth Center: Preliminary Results. Santa Monica, Calif.:
RAND, 1989.
Harding, Richard. “Private Prisons.” In Crime and Justice: A Review of
Research, Vol. 28, edited by Michael Tonry, pp. 265–346. Chicago: University of Chicago Press, 2001.
Harrison, Paige M., and Beck, Allen J. Prisoners in 2001. Washington, D.C.:
U.S. Department of Justice, Bureau of Justice Statistics, 2002.
Hart, Oliver; Schleifer, Andrei; and Vishny, Robert W. “The Proper Scope
of Government: Theory and an Application to Prisons.” Quarterly Journal
of Economics 112 (1997): 1127–63.
Hatry, Harry P.; Brounstein, Paul J.; Levinson, Robert B.; Altschuler, David
M.; Chi, Keon; and Rosenberg, P. Philip. Comparison of Privately and
Publicly Operated Correctional Facilities in Kentucky and Massachusetts.
Washington, D.C.: Urban Institute, 1989.
Koch Crime Institute. Juvenile Boot Camps and Military Structured Youth
Programs: 2000 Directory. Topeka, Kans.: Koch Crime Institute, 2000.
Krisberg, Barry. “The Legacy of Juvenile Corrections.” Corrections Today
57 (1995): 122–54.
Lanza-Kaduce, Lonn, and Parker, Karen F. A Comparative Recidivism Analysis of Releasees from Private and Public Prisons in Florida. Gainesville:
Center for Studies in Criminology and Law, University of Florida, 1998.
Lanza-Kaduce, Lonn; Parker, Karen F.; and Thomas, Charles W. “A Comparative Recidivism Analysis of Releasees from Private and Public Prisons.” Crime and Delinquency 45 (1999): 28–47.
Levinson, Robert B., and Taylor, William J. “ACA Studies Privatization in
Juvenile Corrections.” Corrections Today 53 (1991): 242–48.
Lilly, J. Robert, and Knepper, Paul. “The Corrections-Commercial Complex.”
Crime and Delinquency 39 (1993): 150–66.
Logan, Charles H. Well Kept: Comparing Quality of Confinement in a Public
and a Private Prison. Washington, D.C.: National Institute of Justice, 1991.
Low, Daniel L. “Nonprofit Private Prisons: The Next Generation of Prison

588

the journal of law and economics

Management.” New England Journal on Criminal and Civil Confinement
29 (2003): 1–66.
MacKenzie, Doris Layton; Wilson, David B.; and Kider, Suzanne B. “Effects
of Correctional Boot Camps on Offending.” Annals of the American Academy of Political and Social Science 578 (2001): 126–43.
Maltz, Michael D. Recidivism. Orlando, Fla..: Academic Press, 1984.
McDonald, Douglas C. “The Costs of Operating Public and Private Correctional Facilities.” In Private Prisons and the Public Interest, edited by
Douglas C. McDonald, pp. 86–106. New Brunswick, N.J.: Rutgers University Press, 1990.
McKelvey, Blake. American Prisons: A History of Good Intentions. Montclair, N.J.: Patterson Smith, 1977.
Moran, Richard. “A Third Option: Nonprofit Prisons.” New York Times,
August 23, 1997, p. 23.
National Criminal Justice Association. Juvenile Justice Reform Initiatives in
the States: 1994–1996. Washington, D.C.: U.S. Department of Justice,
Office of Juvenile Justice and Delinquency Prevention, 1997.
Porter, Robert G. “The Privatisation of Prisons in the United States: A Policy
That Britain Should Not Emulate.” Howard Journal of Criminal Justice
29 (1990): 65–81.
Pozen, David E. “Managing a Correctional Marketplace: Prison Privatization
in the United States and the United Kingdom.” Journal of Law and Politics
19 (2003): 253–82.
Ryan, Mick, and Ward, Tony. “Privatization and the Penal System: Britain
Misinterprets the American Experience.” Criminal Justice Review 14
(1989): 1–12.
Schmidt, Peter, and Witte, Ann D. An Economic Analysis of Crime and
Justice: Theory, Methods, and Applications. New York: Academic Press,
1984.
Sechrest, Dale K., and Shichor, David. “Quick Fixes in Corrections: Reconsidering Private and Public For-Profit Facilities.” Prison Journal 74 (1995):
457–78.
Shichor, David. Punishment for Profit: Private Prisons/Public Concerns.
Thousand Oaks, Calif.: Sage Publications, 1995.
Sickmund, Melissa. Juvenile Offenders in Residential Placement: 1997–1999.
Fact sheet. Juvenile Offenders and Victims: National Report Series No. 7.
Washington, D.C.: U.S. Department of Justice, Office of Juvenile Justice
and Delinquency Prevention, 2002.
Snyder, Howard N., and Sickmund, Melissa. Juvenile Offenders and Victims:
1999 National Report. Washington, D.C.: U.S. Department of Justice, Office of Juvenile Justice and Delinquency Prevention, 1999.
Terry, W. Clinton; Stolzenberg, Lisa; and D’Alessio, Stewart J. “Private
versus Public Placements: A Study of Recidivism among Adjudicated
Juvenile Offenders.” Juvenile and Family Court Journal 48 (1997): 33–41.

juvenile correctional facilities

589

Thomas, Charles W. “Private Adult Correctional Facility Census: A ‘RealTime’ Statistical Profile, December 31, 1999.” On file with the authors.
Weisbrod, Burton A. The Nonprofit Economy. Cambridge, Mass.: Harvard
University Press, 1988.