Skip navigation

Department of Economics, Prison Health Care Contracting Study, 2007

Download original document:
Brief thumbnail
This text is machine-read, and may contain errors. Check the original document to verify accuracy.
Prison Health Care: Is Contracting Out Healthy?

Kelly Bedard
Department of Economics
University of California, Santa Barbara
kelly@econ.ucsb.edu

H.E. Frech III
Department of Economics
University of California, Santa Barbara
frech@econ.ucsb.edu

September 2007

Abstract
U.S. Prison health care has recently been in the news and in the courts. A particular issue
is whether prisons should contract out for health care. Contracting out has been growing
over the past few decades. The stated motivation for this change ranges from a desire to
improve the prison health care system, sometimes in response to a court mandate, to a
desire to reduce costs. This study is a first attempt to quantify the impact of this change
on inmate health. As morbidity measures are not readily obtainable, we focus on
mortality. More specifically, we use a panel of state prisons from 1979-1990 and a fixed
effects Poisson model to estimate the change in mortality associated with increases in the
percentage of medical personnel employed under contract. In contrast to the first stated
aim of contracting, we find that a 20 percent increase in percentage of medical personnel
employed under contract increases mortality by 2 percent.
Thanks are due to Mary Alice Conroy, Ph.D., Director of Graduate Studies in Forensic
Psychology, Sam Houston State University and Clifford Leonard, Ph.D. Staff
Psychologist, Pelican Bay State Prison, California, for helpful discussions and
background in prison health care. An earlier version of this paper was presented at the
International Health Economics Association meetings in Copenhagen, July 9, 2007.
Thanks are due to participants there, especially Avi Dor.
Keywords: prison health care, contracts, managed care, outcomes, mortality
JEL Numbers: I12, I18, K23, L14, L23

1. Introduction
There has been a large increase in contracting out for health care in U.S. prisons. As of
2004, 32 states contract for some or all prison health services (LaFaive 2006). Despite
the massive shift towards contracting out for prison health care, the popular press has
voiced concerns about the resulting quality of service for inmates. For example, a recent
series of New York Times articles by Paul von Zielbauer (2005A, B, C) blames
contracting out for poor health care in New York and Alabama, including inmate suicides
and prisoners dying after being denied treatment. He blames these poor outcomes on
Prison Health Services (PHS), a large health care company which has recently received
large contracts in the states he studies. In fact, PHS is the largest private healthcare
provider for penal institutions (both prisons and jails), providing care in 28 states for
237,000 inmates, about 10 percent of the penal population, grossing $690 million in 2004
(Zeilbaurer 2005A).1 While this is a big company, the market for outsourced medical
care is much larger still. The president of PHS estimates that $3 billion of the $7 billion
spent on penal medical care is contracted out (Business Week 2005).
While there are concerns that prison health care contracting out leads to
understaffing and under-treatment (Robbins 1999), it is also possible that outsourcing
produces efficiency gains by applying the principles of managed health care and thereby
reduces costs without reducing quality. In fact, contracting out has sometimes been
instituted in response to court orders as a means of improving prison health care quality
(McDonald 1999). The argument for outsourcing as a means to improve quality rests on

1

State and federal prisons house individuals who have been sentenced as punishment for crimes. Jails,
typically operated by local governments, largely house individuals who have been accused of crimes. Jail
stays, on average, are for shorter periods. While many economic and legal issues are the same for jails and
prisons, our data is limited to prisons.

2

the notion that independent organizations (often, but not necessarily profit-seeking firms)
are more flexible and efficient than governmentally operated prison health care staffs.
For one thing, contract health care providers are allowed to pay professionals wages that
exceed state-mandated pay schedules that are often too low for difficult work in prisons
in isolated areas (Gater 2005).2 Profit-seeking firms also have better incentives to
produce care more efficiently because managers are allowed to keep the residual earned
by reducing costs (Alchian and Demsetz 1972, Boardman and Vining 1989, Frech 1976,
Fizel and Nunnikhoven 1992).
While there is a substantial literature examining the relative efficiency of
government versus private firms in the context of goods produced directly for markets,
such as insurance or privatization of state-owned enterprises (Ehrlich, Gallais-Hamonno,
Lieu, 1994, Boardman and Vining 1989, Frech 1976, Shleifer 1998,), to the best of our
knowledge only Hart, Shleifer and Vishny (1997) formally models contracting for
services that are not bought on the open market. They set up a simple model where the
provider, either a government employee or a private contracting firm can invest in either
improving quality (which also tends to raise price) or reducing cost (which also tends to
reduce quality). They show that private contractors have stronger incentives to both
improve quality and reduce costs than government employees. The problem is that
private contractors may have incentives that are too strong to reduce costs, since they
ignore the adverse impact on quality. They apply the model to the question of privatizing
entire prisons. Like contracting out for prison health care, privatizing entire prisons is
growing in popularity in the U.S., though it is much less common than contracting out

2

All these issues were raised by Judge Thelton Henderson in appointing a receiver to take over the
California system (Plata v. Schwartznegger 2005).

3

health care alone.3 Their model suggests that prisons are not good candidates for
privatization because of the possibility for significant reductions in quality as a byproduct
of reducing costs. And, of course, inmates are not effective monitors of quality.
Consistent with this prediction, they present limited evidence that private prisons have
both lower costs and lower quality.
Given the theoretically ambiguous impact of prison health care contracting out on
the quality of inmate health care, the objective of this study is to quantify the impact.
Ideally, we would include measures of morbidity and mortality. Unfortunately,
morbidity data is not readily available. As such, we focus on mortality. In particular, we
use Census of Prison data from 1979-1990 and a fixed effects Poisson model to estimate
the impact of increases in health care outsourcing on inmate mortality. We find a
negative effect of contracting out on health (a positive effect on death). More
specifically, we find that a 20 percent increase in percentage of medical personnel
employed under contract increases mortality by 2 percent.

2. Prison Health Care Contracting
2.1. The History of Court Involvement
Before the federal courts began intervening, health care in American prisons was poor
and limited. Prison officials often considered medical care as a privilege, rather than a
right. It was sometimes withheld to discipline inmates. Care was often dispensed by
retired military corpsmen and untrained inmate nurses. The few physicians that existed
often had restricted institutional licenses (McDonald 1999, Anno 2004). The early
descriptive literature is harrowing. For example, Pennsylvania inmates who tried to hang
3

In 1995, private prisons had only about 3 percent of the market (Hart, Shleifer and Vishny 1998).

4

themselves were simply cut down, medicated and returned to their cells without
psychological evaluation (McDonald 1999, Anno 2004).
One of the complaints of the rioters in the infamous 1971 Attica New York prison
riot was inadequate health care. Although prisoners and prisoner advocates sued prisons
on the grounds that health care was inadequate, during this early period the courts
generally took a hands-off approach. The legal environment changed abruptly when the
federal courts began to view health care through the lens of the U.S. Constitution’s
Eighth Amendment prohibition of cruel and unusual punishment. An early landmark
case in this regard was the federal district court decision, affirmed by the Fifth Circuit of
Appeals, Newman v. State of Alabama (1974). Among the factual findings of the
decision was the story of a quadriplegic who was not given intravenous feeding in the
three days before his death. The court found the conditions barbarous and in violation of
the Eighth Amendment. In 1976, the Supreme Court addressed these issues in Estelle v.
Gamble. They declared that “deliberate indifference” to a prisoner’s serious medical
problem is a Constitutional violation. Subsequent rulings further established the right to
“reasonably adequate medical care” (McDonald 1999).
Partly due to the vagueness of these standards,4 these legal changes initiated an
endless stream of court cases and led to heavy involvement of the courts in forcing
improvements in prison health care. Medical care is the most litigated issue involving
prisons (Schlanger 2003). By 1996, 36 states were under federal court order to improve
prisons. The majority of these cases included health care (McDonald 1999). A survey of

4

Most observers cite under-treatment, especially of relatively sick prisoners. But, the concept of
“reasonably adequate medical care” is so vague that surprising outcomes of any kind can occur. For
example, in a controversial instance, the California prison system provided a heart transplant to twice
convicted of armed robber, at a cost of $1,000,000 (McKneally and Sade 2003).

5

prisons in 2003 showed that 56 percent were operating under court orders regarding
medical care and 65 percent had been under court orders that had been lifted by 2003.
In the new legal situation created by the federal decisions, various professional
organizations stepped in to set standards. Meeting these standards supports a legal
defense of following the usual practice. Free entry into standard setting led to as many as
four separate sets of standards. Today, there are still two, those of the American
Correctional Association and the National Commission on Correctional Health Care.
There are also generally higher standards that are not specific to prisons that some prisons
do meet.

2.2. Prison Health Care Costs and Health Status
Naturally, in this setting, costs for prison health care have risen substantially. Although
data are surprisingly difficult to come by, surveys indicate that spending in 1995 was
about $2,308 per prisoner per year. That was up from $880 in 1982, an increase of 160
percent (McDonald 1999). The variation in spending across state prison populations is
also striking. In 1998, costs ranged from a low of $1,001 in Alabama to a high of $4,365
in Massachusetts (Lamb-Mechanick and Nelson, undated). Note that per capita health
care spending in the U.S. as a whole in 1995 was $3,509 (U.S. Census 1997).
The seemingly high level of prison health care spending described above
underlies public policy towards cost control, including contracting out. But, one should
be careful making comparisons between inmate and non-inmate medical expenditures.
While prisoners are quite young, they are generally thought to be less healthy than the
population at large and therefore to need more medical services. However, the evidence

6

for this is mixed. For example, death rates in prisons are lower than for the general
population, after controlling for race, sex and age, though they are higher for infectious
disease and suicide (McDonald 1999). In a study of the Cook County (Chicago) Jail,
Kim et al (2006) found a 68 percent lower adjusted mortality rate for inmates than for the
general population. On the other hand, when prisoners are released, their death rates
jump and become much higher than those of the general population. In a study of former
Washington State inmates, the adjusted death rate for the former inmates was 3.5 times
the state’s overall death rate (Binswanger et al 2007).
We know of only one nationwide analysis of prison health care costs, done by
Lamb-Mechanick and Nelson (undated). They use state level data obtained from a
dedicated survey of state departments of corrections, plus the Federal Bureau of Prisons
(BOP) in 1998. Lamb-Mechanick and Nelson report per day health care costs per inmate
ranging from $2.74 in Alabama to $11.96 in Massachusetts, with a mean of $7.15. They
also study the determinants of costs using a simple OLS model and data from 38 states.
The regressors include several measures of medical professional inputs, and whether
juveniles are included in the budget. No state socio-economic variables are used. For
our purposes, the most interesting finding involves the dummy variable for whether the
state used capitated contracts (like many private sector HMOs) for ambulatory care. 18
states report using such contracts. Lamb-Mechanick and Nelson find that states with
capitated contracts have 31 percent lower costs per inmate. While this result is
interesting, it is reasonable to be quite concerned about omitted variables bias in this
context and one should therefore interpret these estimates with care.

7

One rationale for both contracting out and using private-sector managed care
techniques is to reduce costs. Another technique borrowed from the private sector (and
from Medicare) is prisoner copayments. This has been shown by Hyde and Brumfield
(2003) to be effective. They examine the initiation of small copayments ($3.00 for a sick
care visit and $2.00 for a prescription) in Idaho prisons in 1998.5 As one might expect,
the number of sick care requests declined by about 40 percent. This result is roughly
consistent with, though slightly larger than, the classic RAND study comparing free care
to small copayments (Newhouse et al 1981 and 1993). However, the comparison is not
perfect as there is likely to be significant non-price rationing in prisons, especially in the
absence of copayments.

2.3. Prison Health Care Quality and Outcomes
Our knowledge about the impact of contracting out on the quality of prison health care is
limited to case studies from Texas, Baltimore and Salt Lake City. As a result of
successful inmate lawsuits in the early 1990s, the U.S. District Court ordered Texas to
improve prison health care. In 1994, Texas responded by contracting out to a managed
care network established by and integrated with Texas state medical schools and their
affiliated teaching hospitals. The system uses a global capitation system, an HMO-like
system with strong incentives for cost control. Raimer and Stobo (2004) state that the
result has been improved health care by many process measures. For example, staff
vacancies declined greatly and compliance with practice guidelines improved. More
interestingly, several outcome measures also improved, including blood sugar levels in
diabetics, the proportion of inmates with high blood pressure, and death rates from AIDS
5

Indigent inmates (about 20 percent of inmates) do not have to pay these fees.

8

and asthma. At the same time, this contracting out strategy saved the state $215 million
over six years.
In a similar vein, as a result of prisoner protests in the Baltimore City Jail, health
care was contracted out to a newly created nonprofit organization in 1977. A comparison
of outcome measures pre and post contracting shows substantial effects. While the
number of sick visits fell from 62.9 to 27.4 patients per day per 1,000 inmates, the length
of time nurses spent with patients per visit rose from 2.8 to 10.9 minutes. At the same
time, clinical staffing at the jail increased by 60 percent while costs rose by only 13
percent, largely because hospital use declined. Overall, Freeman (1981) considers this to
be a substantial improvement in care.
Lastly, Szykula and Jackson (2005) detail a case study for managed mental health
techniques in a large Salt Lake City jail. They report lower costs and much lower levels
of psychotropic medication of the inmates after the initiation of manage care.

3. Inmate Mortality Data
We construct a balanced three year panel from the 1979 and 1984 Census of State Adult
Correction Facilities and the 1990 Census of State and Federal Adult Correction
Facilities. The sample is restricted to these three years because they are the only surveys
that include the necessary data. Because federal data are only reported in 1990, the panel
is also restricted to state prisons. All data are self reported at the institution level. As the
objective is to estimate the impact of medical contracting on inmate mortality, we restrict
the sample to facilities that are likely to offer at least some amount of medical care.
Operationally this means that the sample is restricted to state prisons housing adults with

9

a minimum capacity of 100 inmates and a positive number of professional staff in each of
the three years. This restriction reduces the sample from 1560 to 1095 prisons. Most of
the excluded prisons are very small and have no medical staff.
We use two dependent variables: deaths due to illness (including AIDS) and
deaths due to illness or suicide. Deaths due to violence are excluded. Our primary
independent variable is the percentage of professional staff under contract (not on regular
payroll). The number of professional staff is the number of medical doctors, dentists,
nurse, paramedics, psychiatrists, psychologists, educational and vocational counselors,
teachers, social workers, and so on. We are forced to amalgamate medical and other
professionals because the 1990 data does not separately identify them. Section 4.1
examines the possible biases introduced by this amalgamation. In particular, we use the
more detailed data available in the 1979 and 1984 data to approximately bound the
estimates. Table 1 reports the average percentage of professional staff employed under
contract at the prison level. Columns 1 and 2 show that the fraction of professional staff
employed under contract was essentially stable between 1979 and 1984, and then
increased substantially between 1984 and 1990.
The remaining control variables, other than the prison fixed effects, are listed in
Table 2. All models include the number of professional staff, the number of other staff,
prison population, prison capacity, prison security level, and the number of inmates killed
in the past year. The number of inmates killed is a proxy for social conditions in the
prison. The columns in Table 2 report these summary statistics for a variety of subsamples. Our primary sample includes all prisons with a minimum capacity of 100
inmates and a positive number of professional staff in each of the three years. Column 2

10

restricts the sample to prisons with a hospital, a shared hospital, or an infirmary and
column 3 restricts the sample to just prisons with a hospital. Columns 4 and 5 restrict the
sample to prisons with an average capacity of 500+ and 1000+, respectively. We use
these samples in Section 4 to check the sensitivity of the results to various sample
specifications.

4. Fixed Effects Poisson Model
The objective is to estimate the impact of medical contracting in prisons on inmate
mortality:
M it = α i + φt + βC it + γPit + X it θ + ε it

(1)

where i denotes prisons, t=1979, 1984, or 1990, M is the annual prison-level mortality
count, a is a vector of prison fixed effects, f is a vector of year indicators, C is the
proportion of professional staff employed on contract (ranges from 0 to 1), P is the
number of professional staff, X is a vector of time-varying prison characteristics, and ε is
the usual error term. The central feature of our prison mortality data is that it is a nonnegative count with a large number of zeros (see Table 3 and Figure 1). As such, we
estimate equation (1) using a fixed effects Poisson model. We also report OLS estimates
for comparison.
The estimates for equation (1) are reported in Table 4. Columns 1 and 2 report
the fixed effects Poisson estimates when mortality includes both illnesses and suicides
and excludes suicides, respectively. For comparative purposes, columns 3 and 4 report
the same estimates for a linear fixed effects model. All models are weighted by average
capacity. The sample sizes are smaller for the Poisson models compared to the OLS

11

models because prisons with constant mortality counts are dropped in the fixed effects
Poisson estimation. The first row reports the primary coefficient of interest, β , which is
the impact of contracting out on inmate mortality. Our preferred estimate includes deaths
due to suicide and uses a Poisson model (column 1). Under this specification, the
contracting coefficient is 0.352 with a standard error of 0.036.6 This impact is both
statistically significant and economically important. The coefficient estimate for
contracting out implies that a 20 percent increase in contracting (this is the mean for the
biggest change observed by prison ) increases mortality by 0.07 deaths or 2 percent
relative to a mean death count of 3.49. Alternatively, one can think about a complete
change from no contracting out to complete contracting out. In this case, the estimated
coefficient estimate implies that mortality will rise by 0.352, or 10 percent. However,
one should be careful with this interpretation since there are very few changes of such
magnitude in the data. The point estimates for the other three specifications are similar in
magnitude.
The other coefficients are reported in the remaining rows of Table 4. Focusing on
column 1, as one might expect, mortality falls as professional staff increases. On the
other hand, increases in non-professional staff and prison population are associated with
increases in inmate mortality. Also as expected, the trend in prison mortality is upward.
Table 5 replicates columns 1 through 4 in Table 4 under a variety of sample
specifications. For comparative purposes the first row reports the baseline estimates.
The second row restricts the sample to prisons with a hospital, shared hospital, or an
infirmary. Row 3 restricts the sample to prisons with a hospital. Rows 4 and 5 use prison
6

Two related notes. One, the standard errors are quite a bit smaller in the Poisson regressions than the
OLS regression. This makes sense, since this is count data. Two, the fixed effects are jointly significant in
all models, as one would expect with such heterogeneous institutions as prisons.

12

capacity, 500+ inmates and 1000+ inmates, as an alternative to direct measurement of
medical facilities, to focus on prisons that are more likely to provide a high proportion of
medical services in the prison itself. While the point estimates for large prisons and
prisons with a hospital are substantially larger than the baseline and the other two less
restrictive sub-samples, they are similar in percentage terms. A 20 percent increase in
contracting increases mortality by 0.13 deaths or 2 percent relative to a mean death count
of 5.75 for prisons with a hospital and by 0.13 deaths or 2 percent relative to a mean
death count of 5.23 for prisons with 1000+ inmates.

4.1. Medical Staff Measurement Problems
For our purposes, the primary flaw of available data is the fact that medical staff is not
separately identified from other professional staff in the 1990 survey. As a result, we are
forced to use all professional staff and the percentage of them employed under contract
instead of isolating medical contracting out. This lack of disaggregated data is
unfortunate since most of the substantial changes in contracting occur between 1984 and
1990. The 1979 and 1984 surveys do separate medical personnel from other professional
staff. Table 6 therefore replicates Table 4 with three differences. First, the sample only
includes the first two years, 1979 and 1984. Second, medical staff and other professional
staff enter all models separately (rows 1 and 3) as do the percentage of medical and other
professionals who are under contract (rows 2 and 4). Third, the sample is restricted to
prisons with at least some professional and medical staff in both 1979 and 1984.
Several features of Table 6 warrant comment. First, the point estimates are less
consistent across specifications. This is likely due to the limited number of prison

13

contracting changes between 1979 and 1984. Second, in this sample period, the prisons
not dropped from the Poisson model are generally large prisons; since these are the only
institutions with contracting and/or death count changes. As such, it may not be
surprising that the OLS and Poisson estimates are quite different, with the Poisson
estimates being similar to the large institution estimates reported in Table 5 and the OLS
estimates being more similar to the baseline estimates reported in Table 4. Third, while
the point estimates for the percentage of medical staff contracted out are estimated
reasonably precisely, the estimates for the other medical and professional measures are
very noisy. Again this likely reflects the fact there are very few changes between 1979
and 1984. Finally, although not reported in the table, the average percentage of
contracting is similar across medical and other professional categories, and changes very
little between years. In 1979 9.5 percent of medical staff are employed under contract
compared to 8.3 percent for other professionals. By 1988 these percentages had changed
slightly to 8.8 percent for medical staff and 8.0 for other professionals. Taken as a whole
these finding suggest that proxying medical contracting with professional contracting is
likely to be fairly reliable.

4.2. Endogenous Medical Contracting
The analysis of the impact of medical contracting on prisoner mortality raises the
question of endogeneity. More concretely, one may be concerned that the results partly
reflect the decision of prisons with high and rising death rates to switch towards medical

14

contracting to slow the rise in the death rate.7 As we have three years of data, we can
investigate this possibility by relating changes in mortality in the earlier period to
contracting out choices in the later period. We estimate the following simple model:

ΔC i90−84 = α + φ84 + πΔM i84−79 + γPi 84 + X i 84θ + ε i 84

(2)

where i denotes prisons, ΔC i90−84 is the change in the percentage of professional workers
employed under contract from 1984 to 1990, ΔM i84−79 is change in prison-level mortality
from 1979 to 1984, f is a 1984 year indicator, P is the number of professional staff, X is a
vector of prison characteristics as measured in 1984. Using equation (2), we ask whether
prisons that experienced increases in inmate mortality responded by changing their
professional staff contracting rate.
The results are reported in columns 1 and 2 in Table 7. Whether mortality
includes or excludes suicides, there is no relationship between the change in mortality
between 1979 and 1984 and the change in medical contracting between 1984 and 1990.
The point estimates are zero to three decimal places and the standard errors are small. In
order to check the sensitivity of this finding to alternative specifications, columns 3
through 8 replace the change in mortality with the level of mortality in 1984, 1979, both
and 1979 and 1984. The results for all specifications are similar: the data indicate that
prisons did not respond to mortality changes by changing the percentage of their medical
staff employed under contract, at least during the period of for which we have data.

7

To the extent that higher mortality rates deter crime, as shown Katz, Levitt and Shustorovich (2003), it is
also possible that the composition of prisoners is changing over time. While it is not obvious how this
would bias the reported estimates, we have no way to deal with possibility of such selection.

15

5. Conclusion
We find no evidence to support the positive rhetoric regarding the impact of prison health
care contracting out on inmate health, at least as measured by mortality. Our findings of
higher inmate mortality rates under contracting out are more consistent with recent
editorials raising concerns about this method of delivering health care to inmates. In fact,
the reported results lead one to wonder if Paul von Zielbauer (2005A, B, C) is indeed
correct asserting that contracting out may be as good as “death sentence,” for at least
some inmates?
It is, of course possible that the estimated declines in health care quality are offset
by gains in lower costs. The literature (Lamb-Mechanick and Nelson, undated) shows
that contracting out does reduce costs. These results, together with ours, suggest that the
theoretical analysis of Hart, Shleifer and Vishny (1997) may be right. Contracting out
may reduce both cost and quality.
Future work on the important issues of prison health care contracting would
benefit from better data on costs and on the details of the incentives created by different
types of contracts. Further, one might distinguish between contracting out to profitseeking versus nonprofit firms. And, it would be useful to repeat this study for more
recent data. All of these approaches require new datasets, going beyond the Census.

16

References
Alchian, Armen A. and Harold Demsetz, 1972. “Production, Information Costs, and
Economic Organization, American Economic Review 62(5) (Dec.): 777-795
Anno, B. Jaye, 2004. “Prison Health Services: An Overview, ” Journal of Correctional
Health Care, 10(3) (Fall): 287-301.
Boardman, Anthony E. and Aidan R. Vining, 1989. “Ownership and Performance in
Competitive Environments: A Comparison of the Performance of Private, Mixed, and
State-Owned Enterprises,”Journal of Law and Economics, 32(1) (Apr): 1-33.
Binswanger, Ingrid A., et al, 2007. “Release from Prison—A High Risk of Death for
Former Inmates,” New England Journal of Medicine, 356(2) (Jan. 11, 2007): 157-165.

Business Week, “Prison Health Services’ Hard Time,” (May 18, 2005).
http://www.businessweek.com/bwdaily/dnflash/may2005/nf20050518_2061_db008.htm
Ehrlich, Isaac, Georges Gallais-Hamonno, Zhiqiang Lieu, Randal Lutter, 1994.
“Productivity Growth and Firm Ownership: An Analysis and Empirical Investigation,”
Journal of Political Economy, 102(5) (Oct.): 1006-1038.

Estelle v. Gamble 429 U.S. 97; 97 S. Ct. 285; 50 L. Ed. 2d 251; 1976 U.S. LEXIS 175
Fizel, John L. and Thomas S. Nunnikhoven, 1992. “Technical Efficiency of For-Profit
and Non-Profit Nursing Homes,” Managerial and Decision Economics, 13(5) (Sept.):
429-439.
Frech, H. E. III, 1976. “The Property Rights Theory of the Firm: Empirical Results from
a Natural Experiment,” Journal of Political Economy, 84(1) (Feb.): 143-152.
Freeman, Richard W., et al, 1981. “Planning Health Services for a City Jail: Impact of
Contractual Services on Men’s Sick Call,” Medical Care 19 (4) (Apr.): 410-418.
Gater, Laura, 2005. “Outsourcing Healthcare Services,” Corrections Forum,
(July/August) 53-59
Hart, Oliver, Andrei Shleifer and Robert W. Vishny, “The Proper Scope of Government:
Theory and an Application to Prisons,” Quarterly Journal of Economics, 112(4) (Nov.):
1127-1161.
Hyde, Roberta and Beverly Brumfield, 2003. “Effect of Co-payment on the Use of
Medical Services by Male and Female Prisoners,” Journal of Correctional Health Care,
9(4) (Winter) 371-380.

17

Katz, Lawrence, Steven D. Levitt and Ellen Shustorovich, 2003. “Prison Conditions,
Capital Punishment and Deterrence,” American Law and Economcs Review, 5(2) (Fall):
318-343.
Kim, Seijeouung, et al. 2006. “Deaths in the Cook County Jail: 10-Year Report, 19952004,” Journal of Urban Health, 84(1): 70-84.
LaFaive, Michael D. 2006, Privatization for the Health of It, Mackinac Center for Public
Policy, http://www.mackinac.org/article.aspx?ID=6910, downloaded June 27, 2006.
Lamb-Mechanick, Deborah and Julianne Nelson, undated. Prison Health Care Survey:
An Analysis of Factors Influencing Per Capita Costs (undated), National Institute of
Corrections Website, http--www.nicic.org-pubs-2000-015999.pdf, downloaded May 16,
2007.
McDonald, Douglas C., 1999. “Medical Care in Prisons,” Crime and Justice, 26: 427478.
McKneally, Martin F. and Robert M. Sade, 2003. “The prisoner dilemma: Should
convicted felons have the same access to heart transplantation as ordinary citizens?
Opposing Views,” Journal of Thoracic and Cardiovascular Surgery, 125 (3) (Mar.): 451453.
Newhouse, Joseph P. and the Insurance Experiment Group, 1993. Free for All: Lessons
from the RAND Health Insurance Experiment, Cambridge: Harvard University Press.
Newhouse, Joseph P., et al, 1981. “Some Interim Results from a Controlled Experiment
in Health Insurance,” New England Journal of Medicine, 305 (25) (Dec. 17, 1981): 15011507.

N. H. Newman v. State of Alabama 503 F.2d 1320; 1974 U.S. App. LEXIS 6147.
(Marciano) Plata v. Arnold Schwartzenegger, 2005. Findings of Fact and Conclusions of
Law Re Appointment of Receiver, U.S. District Court for the Northern District of
California, Case No. C01-1351 TEH, Oct. 3, 2005.
Raimer, Ben G. and John D. Stobo, 2004. “Health Care Delivery in the Texas Prison
System,” Journal of the American Medical Association, 292 (4) (July 28, 2004): 1-9.
Robbins, Ira P., 1999. “Managed Health Care in Prisons as Cruel and Unusual
Punishment,” Journal of Criminal law and Criminology, 90 (1) (Autumn):195-238.
Schlanger, Margo, 2003. “Inmate Litigation: Results of a National Survey,” LJN
Exchange, Lonogmont, CO: National Institute of Corrections Information Center.

18

Shleifer, Andrei, 1998, “State versus Private Ownership,” Journal of Economic
Perspectives 12(4) (Autumn): 133-50.
Szykula, Steven A. and Dawn F. Jackson, 2005. “Managed Mental Health Care in Large
Jails: Empirical Outcomes on Cost and Quality,” Journal of Correctonal Health Care 11
(3): 223-240.
Von Zielbauer, Paul, 2005A “As Health Care in Jails Goes Private, 10 Days Can Be a
Death Sentence,” New York Times, (Feb. 27).
Von Zielbauer, Paul, 2005B. “In City’s Jails, Missed Signals Open Way to Season of
Suicides,” New York Times, (Feb. 28).
Von Zielbauer, Paul, 2005C. “A Company’s Troubled Answer for Prisoners With
H.I.V.” New York Times, (Aug. 1).
U.S. Bureau of the Census, 1997, Statistical Abstract of the United States: 1997 (117th
Ed.) Washington, D.C.

19

year==79

year==84

1
.75

Fraction of Prisons

.5
.25
.1
0
0

year==90

5

10

15

1
.75
.5
.25
.1
0
0

5

10

15

Number of Deaths

Figure 1. Annual Prison-Level Inmate Death Count

Table 1. Contracting Percentage (Measured at the Prison Level)

None
1-24 percent
25-49 percent
50-74 percent
75+ percent
Unweighted.

1979

1984

1990

53
27
15
4
1

59
24
11
5
2

31
41
15
7
7

Table 2. Prison-Level Summary Statistics
(1)
Deaths due to illness or suicide
Deaths due to illness

(2)

(3)

(4)

(5)

3.49

3.63

5.75

4.00

5.23

(6.06)

(6.16)

(8.88)

(6.39)

(7.23)

3.01

3.13

5.02

3.45

4.51

(5.63)

(5.73)

(8.35)

(5.96)

(6.79)

Percent contract professional

0.10

0.10

0.12

0.09

0.08

(0.19)

(0.18)

(0.21)

(0.17)

(0.16)

Professional staff

80.23

82.74

96.11

89.80

109.67

(69.36)

(69.83)

(89.64)

(70.15)

(74.45)

5.45

5.61

5.95

6.07

7.36

(4.64)

(4.66)

(4.65)

(4.66)

(4.89)

Other staff (in 100s)
Prison population (in 100s)
Prison capacity (in 100s)
Maximum security facility
Medium security facility
Minimum security facility
Inmates killed per 100 inmates

15.85

16.38

16.91

17.95

22.76

(13.04)

(13.09)

(13.37)

(12.91)

(12.89)

15.97

16.51

16.90

18.09

22.98

(13.20)

(13.26)

(13.30)

(13.08)

(13.05)

0.51

0.53

0.60

0.56

0.61

(0.50)

(0.50)

(0.49)

(0.50)

(0.49)

0.41

0.41

0.38

0.41

0.37

(0.49)

(0.49)

(0.49)

(0.49)

(0.48)

0.08

0.07

0.03

0.04

0.02

(0.28)

(0.25)

(0.16)

(0.19)

(0.13)

0.02

0.02

0.03

0.03

0.03

(0.06)

(0.06)

(0.07)

(0.06)

(0.05)

Due to illness or suicide (per 1000 Inmates)

2.2

2.2

3.4

2.2

2.3

Due to illness (per 1000 Inmates)

1.9

1.9

3.0

1.9

2.0

1095

966

255

615

276

No
No
No
No

Yes
No
No
No

No
Yes
No
No

No
No
Yes
No

No
No
No
Yes

Average Prison-Level Death Rate

Sample size
Sample restricted to prisons with:
A hospital, shared hospital, or infirmary
A hospital
Average Capacity 500+
Average Capacity 1000+
Weighted by average prison capacity.

Table 3. Prison-Level Death Counts by Year
Death Count

1979

1984

1990

0
1
2
3
4
5
6
7
8
9
10
11
13
14
15
16
18
19
20
22
26
29
32
38
42

251
52
20
8
17
8
2
1
1
1
2
1
0
0
0
0
0
1
0
0
0
0
0
0
0

202
69
37
13
11
7
5
4
4
5
1
2
1
1
0
1
0
1
0
0
0
1
0
0
0

163
67
36
24
20
9
10
7
3
5
0
2
0
3
2
0
3
1
2
2
1
1
1
2
1

Unweighted.

Table 4. The Impact of Contracting Out on Inmate Mortality
Poisson
(1)
Percent contract professional
Professional staff
Other staff (/100)
Prison population (/100)
Prison capacity (/100)

OLS: log(1+deaths)
(2)

(3)

(4)

0.352

0.284

0.307

0.243

(0.036)

(0.039)

(0.125)

(0.127)

-0.002

-0.001

-0.001

0.000

(0.000)

(0.000)

(0.001)

(0.001)

0.017

0.021

0.009

0.016

(0.002)

(0.003)

(0.010)

(0.011)

0.018

0.009

0.010

0.001

(0.002)

(0.002)

(0.008)

(0.008)

0.005

0.007

0.019

0.020

(0.002)

(0.002)

(0.009)

(0.009)

Inmates killed (per inmate)

0.484

0.117

0.070

-0.108

(0.094)

(0.108)

(0.355)

(0.359)

Medium security facility

-0.098

-0.117

-0.094

-0.102

(0.017)

(0.018)

(0.066)

(0.067)

Minimum security facility

-1.080

-0.993

-0.299

-0.292

(0.053)

(0.054)

(0.131)

(0.133)

0.484

0.488

0.198

0.194

(0.012)

(0.013)

(0.040)

(0.040)

1984
1990

0.969

1.041

0.436

0.428

(0.016)

(0.017)

(0.056)

(0.056)

Sample size

750

711

1095

1095

Deaths include suicides

Yes

No

Yes

No

All models also include prison indicators. Weighted by average prison capacity. Standard errors are in parentheses.
OLS standard errors are heteroskedastic consistent. Bold coeffcients are statistically significant at the 5% level and bold
italics are statistically significant at the 10% level.

Table 4. Robustness
Poisson

OLS: log(1+deaths)

(1)

(2)

(3)

(4)

Sample restricted to prisons with:
All prisons

Hospital, shared hospital, or infirmary

Hospital

Prison capacity 500+

Prison capacity 1000+

Deaths include suicides

0.352

0.284

0.307

0.243

(0.036)

(0.039)

(0.125)

(0.127)

[750]

[711]

[1095]

[1095]

0.373

0.298

0.345

0.269

(0.037)

(0.040)

(0.138)

(0.140)

[684]

[654]

[966]

[966]

0.659

0.521

0.677

0.472

(0.051)

(0.055)

(0.250)

(0.251)

[213]

[201]

[255]

[255]

0.369

0.294

0.414

0.331

(0.038)

(0.041)

(0.191)

(0.194)

[540]

[519]

[615]

[615]

0.652

0.522

0.728

0.558

(0.044)

(0.047)

(0.324)

(0.340)

[270]

[261]

[276]

[276]

Yes

No

Yes

No

All models include the variables listed in Table 3. Weighted by average prison capacity. Standard erros are in parentheses.
OLS standard erros are heteroskedastic consistent. Sample sizes are in square brackets. Bold coeffcients are statisticallylevel
significant at the 5% and bold italics are statistically significant at the 10% level.

Table 6. Separating Medical and Other Professionals
Poisson

OLS: log(1+deaths)

(1)

(2)

(3)

(4)

0.6754

1.0502

0.3303

0.3731

(0.0668)

(0.0838)

(0.1620)

(0.1567)

Medical staff

0.0022

0.0000

0.0024

0.0006

(0.0005)

(0.0005)

(0.0021)

(0.0021)

Percent contract non-medical professionals

-0.1924

-0.1579

-0.0028

-0.0200

(0.0613)

(0.0713)

(0.1843)

(0.1782)

Other professional staff

-0.0012

0.0008

-0.0014

0.0006

(0.0004)

(0.0005)

(0.0015)

(0.0014)

Other staff (/100)

0.0477

0.0781

0.0923

0.1352

(0.0095)

(0.0106)

(0.0381)

(0.0368)

Prison population (/100)

0.0352

0.0249

0.0134

-0.0025

(0.0031)

(0.0034)

(0.0122)

(0.0118)

Prison capacity (/100)

0.0241

0.0173

0.0173

0.0142

(0.0038)

(0.0042)

(0.0126)

(0.0121)

Inmates killed (per inmate)

-0.5864

-0.8387

-0.2255

-0.1596

(0.1427)

(0.1710)

(0.4392)

(0.4247)

Medium security facility

-0.3891

-0.4749

-0.1437

-0.1512

(0.0377)

(0.0432)

(0.0956)

(0.0925)

Minimum security facility

-0.9005

-1.0281

-0.2261

-0.2995

(0.0919)

(0.0945)

(0.1842)

(0.1781)

1984

0.2228

0.2627

0.0827

0.0922

(0.0179)

(0.0204)

(0.0495)

(0.0479)

Sample size

370

322

638

638

Deaths include suicides

Yes

No

Yes

No

Percent contract medical professionals

All models also include prison indicators. Weighted by average prison capacity. Standard errors are in parentheses.
OLS standard errors are heteroskedastic consistent. Bold coeffcients are statistically significant at the 5% level and bold
italics are statistically significant at the 10% level.

Table 7. Change in Professional Contracting between 1984 and 1990
(1)
Death Change (1984-1979)

(2)

0.000

0.000

(0.008)

(0.008)

Deaths in 1984

(3)

(4)

(5)

(6)

(7)

(8)

0.007

0.007

0.009

0.015

0.005

0.002

(0.007)

(0.007)

(0.010)

(0.011)

(0.008)

(0.008)

0.006

0.014

(0.011)

(0.012)

Deaths in 1979
Other controls measured in 1984:
Professional staff
Other staff (/100)
Prison population (/100)

-0.001

-0.001

-0.001

-0.001

-0.001

-0.001

-0.001

-0.001

(0.001)

(0.001)

(0.001)

(0.001)

(0.001)

(0.001)

(0.001)

(0.001)

0.004

0.004

0.002

0.002

0.004

0.004

0.003

0.003

(0.016)

(0.016)

(0.016)

(0.016)

(0.016)

(0.016)

(0.016)

(0.016)

0.012

0.012

0.012

0.012

0.012

0.012

0.012

0.012

(0.010)

(0.010)

(0.010)

(0.010)

(0.010)

(0.010)

(0.010)

(0.010)

-0.011

-0.011

-0.012

-0.012

-0.011

-0.011

-0.011

-0.011

(0.009)

(0.009)

(0.009)

(0.009)

(0.009)

(0.009)

(0.009)

(0.009)

Inmates killed (per inmate)

-0.336

-0.336

-0.346

-0.341

-0.381

-0.410

-0.372

-0.405

(0.284)

(0.284)

(0.282)

(0.282)

(0.287)

(0.286)

(0.287)

(0.287)

Medium security facility

-0.002

-0.002

0.003

0.002

0.005

0.006

0.006

0.007

(0.036)

(0.036)

(0.036)

(0.036)

(0.037)

(0.037)

(0.037)

(0.037)

0.030

0.030

0.032

0.031

0.034

0.034

0.034

0.034

(0.047)

(0.047)

(0.047)

(0.047)

(0.047)

(0.047)

(0.047)

(0.047)

0.117

0.117

0.118

0.118

0.115

0.116

0.116

0.117

(0.037)

(0.037)

(0.036)

(0.037)

(0.037)

(0.036)

(0.037)

(0.037)

Sample size

365

365

365

365

365

365

365

365

Deaths include suicides

Yes

No

Yes

No

Prison capacity (/100)

Minimum security facility
Constant

Bold coeffcients are statistically significant at the 5% level and bold italics are statistically significant at the 10% level. Unweighted.