Is Private Time Quality Time? A National Private–Public Comparison of Prison Quality

Abstract

The prison privatization debate has thus far formed around both normative and empirical frameworks. Although not denying the importance of the debate over whether private prisons should exist, the fact that privatization has occurred necessitates empirical examination. We, therefore, use census of federal and state correctional facilities data to compare the quality of confinement across federal, state, and private prisons. After controlling for important institutional differences, we find that across many of the domains of quality, public and private prisons are similar. Public and private prisons did differ on some measures. Private prisons experienced less crowding than either state or federal prisons. Conversely, federal prisons scored higher on measures of activity (work, treatment, education) than privately run institutions.

Keywords

private state and federal prisons quality of confinement

By any measure, private “for profit” correctional institutions have gained a substantial foothold in the prison landscape. In 1995, there were less than 30 adult confinement facilities operated by private contractors. Within 5 years, this number increased to more than 100 (Stephan & Karlberg, 2003). The Bureau of Justice Statistics reported that at year end 2006, almost 114,000 state and federal inmates—more than 7% of adult prisoners—were being held in private correctional facilities. Alaska, New Mexico, and Wyoming now house more than one-third of their inmates in private facilities, and seven additional states house at least 20% of inmates in private prisons (Sabol, Couture, and Harrison, 2007). Taken together, private facilities would follow Texas and California as the third largest “state-level” prison system (Bales, Bedard, Quinn, Ensley & Holley, 2005).

Given the scope of prison privatization, the discourse on this topic has shifted from normative debate to empirical examination. Although the issue of whether prisons should be privatized remains important and unsettled, (Durham, 1989; Reisig & Pratt, 2000; Sechrest & Shichor, 1996) the fact that a substantial proportion of inmates are already held in privately managed prisons demands empirical investigation. Prison quality has emerged as a central point of comparison between privately and publicly run prisons. Privatization skeptics suggest that the potential costs savings from private prison management might be realized by sacrificing prison quality (Shichor, 1995). In contrast, those who support privatization believe that private sector innovation and efficiency will allow private prisons to equal or outperform government-operated institutions (Logan, 1992). Empirical studies of differences in the quality of confinement between publicly and privately run prisons have thus far yielded mixed results (Armstrong & MacKenzie, 2003; Austin & Coventry, 2001; Camp & Gaes, 2002; Drowota, 1995; Logan, 1992, 1996; Thomas, 1997; Urban Institute, 1989).

Ambiguity in the literature stems in part from a reliance on research designs where a small number of private and public institutions are matched on key variables. In addition, scholars have conceptualized prison quality differently. For example, outcome measures range from postrelease recidivism to various markers of quality of life within prison walls. The present study seeks to contribute to the empirical debate surrounding quality comparisons of publicly and privately run prisons in two ways. First, we analyze data from more than 1,000 prisons by utilizing the census of federal and state correctional facilities. This sample size allows us to introduce important statistical controls. Second, our outcome measures are consistent with “domains of quality” framework first outlined by Logan (1992) and later refined by Perrone and Pratt (2003). This allows us to assess quality in a way that promotes replication in the future research.

The Goals of Private Prisons

Although the private sector has a long history of involvement in corrections, the rise of private for-profit prison management in the 1980s represented a qualitative shift in the relation between corrections and private business. Accordingly, the privatization of prisons created a heated normative debate. Critics of privatization question whether private entities should be allowed to exercise coercive control over citizens, or whether this right is exclusive to the state (Reisig & Pratt, 2000; Sechrest & Shichor, 1996). As Perrone and Pratt (2003) note, however, “[R]egardless of whether the most defensible legal-philosophical position dictates that policy makers should not have the authority to grant private agencies the power to punish, legislatures are already contracting correctional services to private companies at an increasing rate” (p. 303). Indeed, the spread of prison privatization highlights the importance of empirical evaluation of private facilities. Empirically, have private prisons lived up to their expectations?

The social context in which prison privatization emerged makes clear that private prisons are supposed to be more efficient and cost less than publicly operated prisons. In the 1980s, the combination of the Reagan administration’s emphasis on a smaller government role and the increased privatization in other spheres, prison crowding paved the way for the emergence of private prison management. Free market advocates argued that prison privatization would produce a cheaper, better quality, prison environment with less government bureaucracy (Logan & Rausch, 1985).

A recent survey of state correctional directors confirms these expectations. Directors reported turning to private prisons for two primary reasons, the desire for cost effectiveness and a need for rapid increases in capacity (MacDonald & Patten, 2005). On the surface, the determination of whether private prisons are cheaper than public prisons seems straightforward. The cost-effectiveness literature, however, has been plagued by methodological and cost-accounting problems. The per diem “cost” of a private facility is calculated based on a number of accounting assumptions that have a heavy influence on costs estimates. For example, should overhead costs for state bureaucracy be included in the cost estimate? Including overhead in the per diem cost, although reasonable on the surface, is based on the questionable assumption that “saving” one prison through privatization will “save” an equal share of overhead.

Such accounting issues make cost comparisons between private and public facilities hazardous. Methodologically, cost studies usually compare two or three prisons that are often not matched on important variables such as size, age, and security level. Given the accounting and methodological limitations associated with this literature, it is difficult to draw firm conclusions. Although, the majority of case studies find that per diem cost differences favor private prisons (Perrone & Pratt, 2003) a meta-analysis of 33 evaluations of the cost effectiveness of private prisons Pratt and Maahs (1999) found that ownership by a private entity was not a significant predictor of per diem inmate cost.

Linked directly with the issue of cost efficiency is the question of quality. Skeptics of privatization question the ability of the private companies to provide better (or comparable) quality of service in a private market that seeks to reduce costs and maximize profits (Ogle, 1999). Others believe that private sector innovation and efficiency will actually improve the quality of service (for an overview see Austin & Coventry, 2001). Private sector advocates argue that the competitive nature of the private market can produce a cheaper, better (or at least comparable) quality prison environment (Logan & Rausch, 1985).

As one might imagine, it is difficult to find consensus regarding the meaning of “quality.” Recent studies of postprison release recidivism illustrate one conceptualization of prison quality. Lanza-Kaduce, Parker, and Thomas (1999) found that a group of releasees from privately operated prisons in Florida had a lower recidivism rate than a matched group of releasees from state-operated facilities. They suggest lower recidivism results from factors such as higher quality programming and a better overall prison environment. Bales and associates (2005) also argue, based on the claims of some corrections companies, that postrelease recidivism is a fair measure of prison quality. Their research, however, found no differences in the recidivism of inmates released from private prisons compared to those released from public facilities (Bales et al., 2005).

Thomas (2005) contends that prison administrators should not be held responsible for what occurs outside of their prison’s walls and thus recidivism is not a fair measure of quality (also see Logan, 1992). Indeed, the great bulk of literature in this area examines differences in environmental quality within prison walls (Armstrong & MacKenzie, 2003; Austin & Coventry, 2001; Camp & Gaes, 2002; Logan, 1992). Such an approach is consistent with a “justice model” of corrections, where the essential purpose of a prison is to fairly and justly punish offenders (Cullen & Gilbert, 1982, Logan, 1992). Unfortunately, the research that has addressed quality from this perspective does not provide any firm conclusions (for a review, see Perrone & Pratt, 2003). In part, this reflects differences in the measurement of quality. In his widely read comparison of three prison systems, Dilulio (1987) measured quality in terms of amenity, service, and order. Although an important first step, DiIulio’s measures have been criticized for being too subjective and difficult to operationalize (see Perrone & Pratt, 2003). More recently, Logan (1992) and Perrone and Pratt (2003) have defined seven “domains of quality” which are more extensive and perhaps more empirically defensible. Although any typology is subjective, organizing quality along these domains does have the potential to create a more consistent body of empirical research.

Quality of Confinement and the “Domains of Quality”

Defined originally by Logan (1992) and refined by Perrone and Pratt (2003), the seven domains of quality are an attempt to bring clarity and reliability to the subjective “laundry list” of quality measures employed by researchers. Again, these seven domains are based on a confinement model of prison where quality is based on factors that occur inside prison walls. The seven domains of prison quality include condition, management, activity, care, security, safety, and order. Although no single study has examined all seven domains, almost all studies that have examined the quality of private prisons include measures of some of the domains.

Condition refers to the physical environment in which the inmates are held. Indications of a poorly kept prison such as crowding, noise, food, and sanitation have been used to measure condition. Studies that compared the conditions of public and private prisons have mixed results, with some finding results favorable to private prisons (Logan, 1992, 1996; Thomas, 1997).

Management refers to the ability of the prison administrators to effectively and efficiently run their institution. It is often measured by comparing staff turnover and stress rates. Logan (1992; 1996) found that private prisons have performed better than public prisons in the management domain, but more recently Camp and Gaes (2002) found private prisons to have higher staff turnover than federal prisons. Other research examining management has been either inconclusive (Archambeault & Deis, 1996) or found no significant differences between public and private correctional institutions (Urban Institute, 1989).

Activity refers to the ability of prison administrators to keep their inmate population involved and active in prison life. This domain is typically measured as the number of educational, treatment, and work programs available to and used by inmates. Several studies have shown private prisons to perform better in the activity domain (Archambeault & Deis, 1996; Austin & Coventry, 2001; OPPAGA, 1998; Sellers, 1989); yet one has found public prisons to be better at keeping their inmates active (Urban Institute, 1989).

As defined by Logan (1992), care specifically involves the extent and quality of medical care afforded to inmates. It is often evaluated by examining the availability of practitioners and different types of medical policies. Few studies have addressed the care domain, but those that have found that private prisons provide better medical attention than public ones (OPPAGA, 1998; 2000).

Security is a measure of how well the prison is able to keep its inmates separated from the outside world. Most studies examine this domain by measuring the number of escapes, but some have suggested that keeping contraband out of prisons is also an important measure of security. Studies that have examined differences in security have again found largely mixed results, with some indicating public institutions are more secure (Archambeault & Deis, 1996; Camp & Gaes, 2002; OPPAGA, 1998) and others finding that private prisons perform better than public ones (Drowota, 1995; Thomas, 1997; Urban Institute, 1989).

Safety refers to the danger to both inmates and prison staff of being assaulted or killed. Studies that have examined the safety of private and public prisons have been split, with some finding no difference (Camp & Gaes, 2002; Thomas, 1997) some finding private prisons to be more safe (Archambeault & Deis, 1996; OPPAGA, 1998), and some finding public prisons to be safer (Austin & Coventry, 2001; Drowota, 1995).

Order refers to the ability of an institution to control its population. It is usually measured by examining the number of disciplinary actions and disturbances that occur. For this domain, studies have typically found private prisons fare as well (Thomas, 1997; Urban Institute, 1989) if not better (Archambeault & Deis, 1996; Drowota, 1995; OPPAGA, 1998) than their public comparisons.

Although the domains of quality scheme may provide a framework to unify research on the quality of prisons, it is obvious that the empirical evidence across most of the domains remains mixed. One likely reason for this murkiness is the heterogeneity that exists among correctional facilities. There is a great deal of variation in the size, architecture, age, location, and administration of prisons both across and within state, federal, and private prison systems. To some extent then, one might expect some state or federal prisons to have higher or lower quality facilities than some private prisons. Unfortunately, the typical study design used by researchers in this area—the case study approach—exacerbates rather than alleviates this problem.

The majority of studies have compared only a small number of private institutions (usually one or two) to a public institution(s) matched on key variables. Because the small number of institutions examined, generalizing these findings to other institutions is problematic. Furthermore, matching on a couple of key variables may hide other important differences across prisons. In particular, there are strong reasons to believe that custody level, size, crowding, age, and architecture have a strong influence on measures of quality, independent of facility ownership. With only a few prisons, studies often fail to properly match prisons based on these important characteristics (Perrone & Pratt, 2003).

There has been, of late, an attempt to move beyond these small sample sizes and collect data on a larger number of private and public facilities (Armstrong & MacKenzie, 2003; Austin & Coventry, 2001; Camp & Gaes, 2002). Austin and Coventry (2001) examined differences between public and private prisons by comparing a national survey of 65 private prisons to the 1995 census of state and federal correctional facilities which includes more than 1,500 state and federal prisons. They examined two of the domains of quality and found that compared with public prisons, private prisons performed better in the activity domain but worse in the safety domain. Camp and Gaes (2002) examined responses to a national survey of 91 private prisons in 1999 and compared them to similar official data collected from federal prisons. They found that federal prisons fared better in the security and management domains and had inconclusive findings in regard to the safety domain.

Although these latter two studies have shed some light on the subject, they are not without their limitations. First, they fail to simultaneously assess differences across state, federal, and private prisons. Austin and Coventry (2001) lumped together state and federal prisons in a single “public” group whereas Camp and Gaes (2002) compared private prisons to only federal prisons. There has yet to be an analysis that compares private facilities to both state and federal institutions separately. Second, these studies have controlled for a limited number of theoretically important variables. For example, although Austin and Coventry (2001) examined differences in the rates of violations (effectively controlling for size of the inmate population), they failed to control for differences between minimum- and medium-security institutions. Finally, these studies examined a rather limited number (two or three) of quality domains.

Armstrong and MacKenzie (2003) addressed these shortcomings by using hierarchal linear modeling to compare all seven domains of quality between privately and publicly operated juvenile correctional facilities. They used self-reported perceptions of quality by juvenile offenders and controlled for differences in the composition of the juvenile population within each institution. They found that after adequately controlling for differences both within and across facilities, private and public institutions had similar levels of quality for all outcomes. Although important, it is unclear whether these self-report findings from juvenile facilities can be generalized to adult facilities or replicated with official data.

This article seeks to add to the growing “quality” literature that has moved beyond the case-study approach. Specifically, we use the census of federal and state correctional facilities 2000 to compare more than 100 private facilities with more than 900 state facilities and more than 80 federal facilities. The analysis includes multivariate (regression) models to properly control for differences between facilities while examining multiple measures of quality. In doing so, this research provides a national comparison of quality between private, state, and federal adult prisons.

Method

Data

The census of federal and state correctional facilities has been conducted approximately every 5 years since 1974 by the Census Bureau and the Bureau of Justice Statistics (U.S. Department of Justice, Bureau of Justice Statistics, 2004). In 2000, survey administrators contacted 1,668 institutions that housed either federal or state inmates and received a response rate of 100%. Data were collected through a mail-administered survey of prison management with mail and telephone follow-ups to ensure responses. The census is designed to collect information on a variety of topics ranging from the functions of the facility to detailed information regarding the demographics of its inmates, programming offered, and disciplinary infractions.

Because the census sends out surveys to a variety of institutions (halfway houses, drug-treatment facilities, and confinement facilities), it was necessary to first restrict the sample to adult confinement facilities that are run by either state, federal, or private entities. A small portion of facilities (1.3%) report being operated by joint or local entities and are excluded from the analysis. Also, as there are no private “super max” prisons, these facilities (1.6% of the total sample) were excluded from the analysis as well. These sample restrictions left us with 1,129 institutions, 105 which are private, 80 which are federal, and 944 which are operated by state governments. Preliminary analysis of the data revealed federal prison administrators systematically failed to answer questions regarding particular sets of questions.¹ Because of this, certain analyses exclude federal prisons.

Dependent Variables

The 2000 census was designed to examine numerous characteristics of institutions, including many key factors that measure the domains of quality identified in the literature. Accordingly, although the survey was not designed specifically to measure the institutional domains of quality, the census provides multiple measures of the seven domains. Table 1 provides descriptive statistics for all of the variables in the analysis. Specific items used to create indices are reported in the appendix.

Table 1.

Descriptive Statistics (N = 1,129)

Variable	$\bar{x}$	s
Prison ownership
State	0.07	0.25
Federal	0.84	0.37
Private	0.09	0.29
Activity
Work assignments	0.64	0.34
Educational scale	3.28	1.36
Treatment scale	5.43	2.10
Safety
Inmate assaults_a	2.17	1.61
Staff assaults_a	1.58	1.47
Order
Disciplinary reports_a	5.46	2.06
Disturbances	0.13	0.34
Care
Mental health scale	4.44	1.90
Suicide prevention scale	4.16	1.71
Disease prevention scale	3.47	0.90
Condition
Crowding	1.02	0.24
Court order	0.27	0.44
Management
Guard to inmate ratio	0.25	0.35
Professional to inmate ratio	0.05	0.05
Control variables
Average daily population	1,038.20	965.60
Maximum security	0.25	0.43
Medium security	0.43	0.50
Minimum security	0.32	0.47
Female facility	0.08	0.27
Mixed sex facility	0.08	0.28
Male facility	0.84	0.38
Facility age (years)	31.50	32.20

Natural logarithm.

Activity

Activity refers to the ability of prison administrators to keep their inmate populations involved and active in prison life. We include three measures of this domain. Consistent with prior research (Archambeault & Deis, 1996; Drowota, 1995; Logan, 1992; Urban Institute, 1989) work reflects the number of inmates who had some sort of work assignment relative to the total number of inmates at the facility during the study period. Also consistent with prior research (Austin & Coventry, 2001; OPPAGA, 1998; Sellers, 1989), several scales were created that measured the availability of different types of programming. The variable treatment was created by summing eight items that indicated whether or not (1 = yes) a prison had particular programs such as drug, alcohol, or sex offender treatment (α = .74). The availability of educational programs ranged from providing adult basic education to college courses. Accordingly, education was created by summing six items (1 = yes) that indicated whether the prison had a particular education program (α = .65).

Safety

Safety refers to the threat of assault, injury, or death to both inmates and prison staff. Accordingly, the number of official assaults reported that have been committed against both inmates and staff in the past calendar year were used to measure the domain of safety. In keeping with prior research, total assaults were disaggregated into assaults on inmates and assaults on staff (Austin & Coventry, 2001). Descriptive statistics indicated that counts of these measures are highly skewed. Therefore, the measures staff assaults and inmate assaults were transformed into the natural logarithm of the original count of assaults.

Order

As order refers to the ability of an institution to control its inmate population, disciplinary reports and disturbances are used. Disciplinary reports is a measure of the number of disciplinary reports filled out by staff in the past year.² Like inmate and staff assaults, the natural logarithm of the number of disturbances was used so the distribution of the variable approaches normality. Disturbances is dichotomous and refers to whether or not (1 = yes) the institution reported an incident involving five or more inmates that resulted in serious injury to anyone or significant property damage in the past year.

Care

As with educational and treatment programs, the Census gathered data on a variety of physical and mental health care policies/programs in place at each institution. Three indices were constructed to measure care. The variable mental health measures how many (out of six) policies or programs (i.e. assessment for mental disorders at intake) were in place to assist inmates with mental conditions (α = .82). Likewise, suicide prevention indicates the number (out of 6) of policies and programs (e.g., prevention teams, staff training) that were in place to prevent suicide (α = .70). Finally, disease prevention is an index of whether or not a prison had either (a) testing or (b) treatment available for three (HIV, Hepatitis B and C) communicable diseases (α = .61).

Condition

Condition refers to the physical environment of the prison. Logan (1992) originally operationalized condition as an indication of a poorly kept prison, using measures of crowding, sanitation, food, noise, and facilities maintenance. To address condition in our study, two measures were used. First, we focused on crowding, as the corrections literature consistently reveals that inmate crowding is a condition of confinement that influences many aspects of prison life such as violence, (Legger, 1988, Ruback & Carr, 1993), drug use (Gillespie, 2005), general misconduct (Wooldredge, Griffin, & Pratt, 2001) stress (Cox, Paulus, & McCain, 1984), and health (Paulus, Cox, & McCain, 1988). The variable crowding was created by dividing the average daily population of the facility by the rated capacity. We recognize that design capacity is a less subjective measure of prison space, but federal prisons report only rated capacity. The use of rated capacity results in a more conservative measure of crowding, yet still gives some indication as to the level of inmates relative to the capacity of the institution. Facilities also reported whether they were currently under a court order for conditions of confinement. The variable court order is a dichotomous measure indicating whether or not (1 = yes) the facility reported being under a court order for conditions of confinement.

Management

Management refers to the ability of prison administration to effectively run their institution and retain qualified staff. The census data have only limited measures related to management. Typically, past research has measured management through staff turnover rates and staff stress levels. Unfortunately, such measures are not available in the census. Instead, following Camp and Gaes (2002), this analysis uses measures of management that are based on staffing levels of correctional officers. We also provide a measure of levels of nonsecurity professionals to measure management. Thus, the ratios of both correctional officers and professionals to inmates are used to measure the management domain.

Security

Security deals with how well the prison is able to keep its inmates separated from the public. Most studies use escapes as a measure of the domain of security (Archambeault & Deis, 1996; Camp & Gaes, 2002; Drowota, 1995; Logan, 1992; OPPAGA, 1998; Thomas, 1997; Urban Institute, 1989). Thus, to measure security, we use a dichotomous variable that indicates whether the institution had any attempted or completed escapes.

Predictor Variables

The primary variable of interest in this article is prison management. Accordingly, a series of dummy variables are used to compare private, state, and federal prisons. Keeping private prisons as the reference group, the dummy variables state and federal reflect the difference between each group and privately operated prisons.

Perrone and Pratt (2003) identify three factors that are essential to control for when comparing prisons. Prison size is important because larger prisons are more likely to have more problems (e.g., assaults, disciplinary reports) than smaller institutions. The average daily population (ADP) is used to control for size. Security level is also important because inmates in higher custody facilities are assumed to be more dangerous and require more monitoring than those at lower custody levels. Security level is operationalized in a series of dummy variables, keeping minimum security as the reference group. The variables medium and maximum thus report differences in the outcome of medium and maximum security institutions relative to those that are minimum security.

The age of the facility can affect quality in a number of ways. Obviously, newer facilities will be in better physical condition and benefit from advancements in architecture (e.g., podular designs) that relate to security and management. Conversely, newer facilities may have less experienced administrators and have yet to fully implement all of the necessary policies to running a high quality prison. This is particularly important considering the youth of most private prisons. Age is taken by subtracting the year of original construction from 2000, the year the survey was administered.

Finally, in keeping with past research (Camp & Gaes, 2002), we control for the gender of the inmates housed in each facility. A series of dummy variables are used, keeping female prisons as the reference group. Institutions which housed only male and both male and female inmates (mixed) are compared to all female institutions.

Analysis Plan

To assess the effect of ownership (private, federal, or state) on prison quality, we computed a number of regression models that predict each measure of quality. For all models, we include the control variables (security level, average daily population, age of the institution, and gender of inmates) discussed above. Where the dependent variables are metric, we employ ordinary least square (OLS) regression.³ Where the variables are dichotomous, logistic regression is used.

Results

Tables 2 and 3 present regression coefficients predicting 15 outcomes of quality organized by domain. As noted, OLS regression models are used to predict continuous dependent variables. In these models, standardized beta coefficients (β) are reported. For dichotomous outcomes, logistic regression is used. Here, the logistic unstandardized coefficients (Exponential β) are reported. These indicate the change in the odds ratio for a one unit change in the independent variable.

Table 2.

Ordinary Least Square (OLS) and Logistic Regression Equations Predicting the Quality Domains of Activity, Safety, and Order

Domain	Activity			Safety		Security	Order
Variables	Education^a	Treatment^a	Work^a	Inmate assaults^a	Staff assaults^a	Escapes^b	Disciplinary reports^a	Disturbances^b
State	0.06	0.04	0.12**	−0.01	−0.08**	1.00	0.04	0.52
Federal	0.18**	0.19**	0.10*	0.09**	0.10**	—	−0.04	—
Maximum	0.16**	0.00	−0.16**	0.33**	0.49**	1.01	0.35**	4.51**
Medium	0.25**	0.20**	−0.08*	0.30**	0.29**	0.62**	0.24**	2.07
Male	−0.17**	−0.24**	−0.01	−0.01	−0.00	1.84*	0.11**	1.99
Mixed	−0.17**	−0.12**	−0.10*	−0.02	−0.07*	2.13*	0.03	0.39
ADP	0.24**	0.23**	0.01	0.40**	0.41**	1.00*	0.34**	1.00**
Age	0.04	0.07*	−0.06*	0.04	0.06*	1.01**	0.01	0.99
R ²	0.18**	0.16**	0.03**	0.36**	0.48**	0.046**	0.31**	0.33**
N	1,109	1,109	1,109	1,109	1,109	1,031	1,109	1,129

Note: ADP = average daily population.

Linear beta coefficients reported. ^bUnstandardized logistic regression coefficients reported; R² = Nagelkerke Pseudo R², (significance based on model χ²).

p < .05. **p < .01.

Table 3.

Ordinary Least Square (OLS) and Logistic Regression Equations Predicting the Quality Domains of Care, Condition, and Management

Domain	Care			Condition		Management
Variables	Mental health^a	Suicide^a	Disease^a	Crowding^a	Court order^b	Guard to inmate^a	Professional to inmate^a
State	0.04	0.04	0.22**	0.16**	4.68**	0.06	0.01
Federal	—	—	—	0.45**	—	—	—
Maximum	0.34**	0.37**	0.26	0.01	1.49	0.20**	0.21**
Medium	0.37**	0.38**	0.28	0.03	1.98**	0.13**	0.14**.
Male	−0.10**	−0.10**	−0.01	0.06	0.63	0.03	−0.16**
Mixed	−0.07	0.07	−0.11**	0.07	0.72	0.03	−0.06
ADP	0.22**	0.22**	0.13**	0.18**	1.00**	−0.18**	−0.19**
Age	0.08**	0.03	0.00	0.04	1.00	−0.02	0.02
R ^2c	0.24**	0.25**	0.19**	0.18**	0.17**	0.04**	0.07**
N	1,031	1,031	1,016	1,109	1,027	1,031	1,031

Note: ADP = average daily population.

Linear beta coefficients reported. ^bUnstandardized logistic regression coefficients reported; R² = Nagelkerke Pseudo R², (significance based on model χ²).

p < .05. **p < .01.

The first three columns from the left-hand side of Table 2 illustrate the regression results across the three activity domain variables. The results reveal several significant differences between private and public prisons. Federal prisons score significantly higher than private prisons on all three outcomes for activity. Furthermore, the beta coefficients reveal the differences are moderate for the educational programming and treatment (.18 and .19, respectively). The coefficients for the state dummy variable reveal no differences with private prisons in education and treatment. The beta coefficient of .12 reveals that compared with private prisons, state prisons have significantly higher proportions of inmates with work assignments, although the difference is modest.

The results for the domain of safety (fourth and fifth columns from left) reveal some significant differences between public and private prisons although these differences are not substantively large. Specifically, the coefficients representing federal prisons are positive and significant for both inmate (β = .09) and staff assaults (β = .10), suggesting federal prisons experience slightly more of both types of assault than private prisons. Compared to private prisons, state prisons report similar levels of inmate assaults and significantly fewer staff assaults. Again, the standardized regression coefficient (–.08) reveals that although significant, the difference between reported staff assaults in state prisons and private prisons is small.

There was no difference in our sole measure of security between state and private institutions. The logistic regression coefficients for the outcome escape reveals that after controlling for security level, average daily population, facility age, and gender of inmates, state and private facilities had similar odds of reporting an escape. Federal prisons were not included in the analysis because they failed to report the number of escapes.

The final two columns in Table 2 address the domain of order. Again, the results reveal no differences between private and public prisons. There are no differences in the natural log of disciplinary reports between private and either state or federal institutions. Also, the binary outcome of whether or not a disturbance was reported in the past year reveals that the odds are not significantly different between public and private prisons.

Table 3 illustrates the regression results for the domains of care, condition, and management. The first three columns from the left examine the three variables (mental health, suicide, and disease) under the domain of care. Federal prisons did not report on any of the outcomes for care. The results for state prisons reveal no differences in policies regarding treatment and screening of mental health and suicide. In contrast, state prisons have significantly more policies and treatments (β = .22) of communicable diseases than private prisons.

Condition domain variables are examined in the fifth and sixth columns in Table 3. In all available comparisons, private prisons outperformed public prisons. Specifically, state prisons experienced more crowding than private prisons. Furthermore, the logistic coefficients reveal a substantial difference between state and private prisons in terms of court orders. As an odds ratio, Exponential (β) is centered on 1, meaning that variables with odds ratios above 1 increase the odds of the outcome and variables below 1 are associated with a decrease in the odds. The coefficient for state prisons thus indicates that state prisons are more than three and a half times (β = 4.68) more likely to report being under a court order. While federal prisons failed to report on court orders, the results indicate that federal prisons also report significantly higher levels of crowding than private prisons. Furthermore, the beta coefficient reported (β = .45) is larger than that reported for any other outcome.

Outcomes for the management domain (the final two columns in Table 2) examine only state and private institutions. The regression models show that the state dummy coefficient is not a significant predictor. This indicates that private and state prisons report similar levels of ratios of guard and professional to inmate ratios.

Discussion and Conclusion

This research examined a large (national) sample of prisons to determine whether private institutions were superior in quality to either federal or state prisons. We conceptualized quality within the “confinement” or justice model of corrections, and categorized measures of quality according to the seven domains outlined in prior research. The findings yield several conclusions.

First, generally, this research shows a large degree of similarity between private and public prisons. Of the 15 outcomes examined, approximately half showed no difference between private prisons and either federal or state prisons. Across measures from the domains of security, order, and management, private and public prisons appear to be quite similar. Thus, consistent with Armstrong and MacKenzie (2003), this analysis indicates that, controlling for theoretically important variables in differences between private and public prisons, there is a general similarity in regard to quality.

Along the same line, the regression coefficients representing prison management (federal, state, or private) that did significantly predict quality outcomes were generally weak to moderate. Thus, a majority of the comparisons indicate that if there are differences, the differences are not that great. For example, in the domain of care, state prisons performed moderately better than private prisons on the scale for infectious diseases, but were similar on the other measures of this domain (the mental health and suicide-prevention scales).

Although there is a degree of similarity between private and public prisons, there are some key differences which were identified in this research. In the domain of condition, private prisons performed better than both state and federal prisons. After entering relevant controls, private prisons were shown to be less crowded than both state and federal prisons. Furthermore, although the difference in crowding between private and state prisons was moderate (β = .16), the standardized beta coefficients for the federal dummy variable was substantial (β = .45), revealing that private prisons are much less crowded than federal prisons.

Also of interest, state prisons were much more likely to report being under a court order for the conditions of confinement than private prisons. These findings are consistent with Austin and Coventry’s (2001) analysis which found that public facilities (both federal and state combined) were more likely to be under a court order for the conditions of their confinement. Furthermore, Austin and Coventry (2001) found that public prisons were also more likely to be under a court order specifically for crowding.

Clearly, then, private prisons are less crowded than publicly operated facilities. Given the adverse consequences of crowding, this is an important difference. Still, it is tough to attribute this finding to factors such as competition, innovation, and lack of government bureaucracy. A more likely explanation is that public facilities, like public education facilities, have much less control over the level and nature of new admissions.

In the domain of activity, federal prisons consistently performed better than private prisons. The standardized coefficients reveal that federal prisons were more likely to report higher scores on scales of treatment, education, and work. State prisons also performed better on the proportion of inmates with work assignments but were similar to private prisons on the treatment and education scales.

Consistent with this analysis, Austin and Coventry (2001) found that public prisons maintained higher percentages of inmates who were working. Of interest, Austin and Coventry (2001) found that private prisons fared somewhat better than public prisons in education and treatment measures. Conversely, our findings show that lumping federal and state prisons together masks differences between federal and state systems. Thus, when federal and state prisons are disaggregated, federal prisons are shown to perform better on education and treatment than both state and private prisons.

The measures of the domain of safety merit some discussion of the limitations of the present study. Although the differences were not substantively large, we found that federal prisons had more reports of assaults on both inmates and staff. It is important to note that these findings should be taken with a degree of caution, as notes in the codebook of census of state and federal correctional facilities specifically mention that, a large majority of assaults on both federal prison inmates and staff were classified as “nonserious.” Camp and Gaes (2002) also noted similar problems comparing federal and private prisons due to cross-jurisdictional differences in the reporting of assaults. As the majority of assaults are nonserious, the measure of assaults for federal prisons is likely to be inflated. Thus, it is probable that the slightly higher level of federal prison assaults is because of reporting differences.

Although this research offers an alternative to the typical case study methodology of past research, it is not without its shortcomings. Most significant is the failure to fully control for compositional differences in inmate population. Although we assume that classifying inmates to varying institutional security levels provides a partial control for the composition of the inmate population, nesting inmates within prisons would enable entering more comprehensive controls for inmate composition. For example, this would allow controls for inmate risk in examining differences between public and private prisons. Still, Armstrong and MacKenzie (2003) concluded that after controlling for compositional differences, private and public facilities were more alike than different, a finding consistent with the present research.

Another shortcoming of this research was the use of a survey not designed explicitly for the purpose of assessing the domains of quality. Clearly, we believe that we were able to capture relatively valid measures of the domains using the census data. Also, although the seven domains have been outlined, there is no gold standard for their operationalization. We recognize that an instrument designed for the task of comparing private and public prisons would yield more valid measures. Although future research would benefit by addressing these limitations, our analysis provides an important first step in the proper specification of models examining the quality of private prisons by controlling for average daily population, security level, gender composition, and facility age.

In sum, this research generally highlights a fair degree of similarity between private and public prisons. Consistent with prior research, (Armstrong & MacKenzie, 2003) we find that when controlling for proper characteristics, the differences between private and public prisons become relatively small. This highlights the importance of proper specification in research that seeks to compare private and public prisons. With respect to the privatization debate, our data suggest that private prisons are comparable, though not superior, to publicly operated facilities.

Footnotes

Appendix

Treatment. Whether or not any of the following educational programs are available (α = .76).

Drug dependency

Alcohol dependency

Psychological

HIV/AIDS

Sex offender

Employment

Life skills

Parenting/child rearing

Education. Whether or not any of the following educational programs are available (α = .67).

Adult basic education

Graduate equivalency diploma

Special education

Vocational training

College courses

Study release program

Mental Health. Whether or not the following mental health policies/programs exist (α = .88).

Intake policy for mental disorder

Psychiatric evaluations and assessments

24-hr mental health care

Therapy/counseling

Psychotropic medications

Assist-inmate to community health service

Suicide Prevention. Whether or not the following mental health policies/programs exist (α = .70).

Assessment at intake

Staff training

Inmate counseling or psychiatric services

Monitoring high risk

Watch cell or special location

Prevention teams

Disease Prevention. Whether or not the following testing or treatment was available (α = .91).

Communicable disease policies and treatments:

Test for hepatitis C virus (HCV)

Treatment for hepatitis C positive

Hepatitis B vaccine

Test for HIV

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Bios

Matthew D. Makarios is an assistant professor in the Department of Criminal Justice at the University of Wisconsin—Parkside. His research interests include institutional corrections, criminal justice policy, and criminological theory. His recent research appears in Crime and Delinquency and Criminal Justice and Behavior.

Jeff Maahs is an associate professor in the Department of Sociology and Anthropology at the University of Minnesota, Duluth. His teaching and research interests include corrections, criminological theory, and legal issues in criminal justice.

References

Archambeault

W. G.

Deis

D. R.

(1996). 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.

Armstrong

G. S.

MacKenzie

D. L.

(2003). Private versus public juvenile correctional facilities: Do differences in environmental quality exist? Crime & Delinquency, 49, 542-563.

Austin

Coventry

(2001). Emerging issues on privatized prisons (NCJ 181249). Washington, DC: Bureau of Justice Assistance: National Council on Crime and Delinquency.

Bales

W. D.

Bedard

L. E.

Quinn

S. T.

Ensley

D. T.

Holley

G. P.

(2005). Recidivism of public and private state prison inmates in Florida. Criminology & Public Policy, 4, 57-82.

Camp

Gaes

(2002). Growth and quality of U.S. private prisons: Evidence from a national survey. Criminology & Public Policy, 1, 427-450.

Cox

V. C.

Paulus

P. B.

McCain

(1984). Prison crowding research: The relevance for housing standards and a general approach for regarding crowding phenomena. American Psychologist, 39, 1146-1160.

Cullen

F. T.

Gilbert

(1982). Reaffirming rehabilitation. Cincinnati, OH: Anderson.

Dilulio

J. J.

Jr. (1987). Governing prisons: A comparative study of correctional management. New York, NY: Free Press.

Drowota

(1995). Comparative evaluation of privately-managed CCA prisons (South Central correctional center) and state managed prototypical prisons (Northeast correctional center, Northwest correctional center). Nashville, TN: Select Oversight Committee on Corrections, Tennessee Assembly.

10.

Durham

(1989). The privatization of punishment: Justification, expectations, and experience. Criminal Justice Policy Review, 3, 48-73.

11.

Gillespie

(2005). A multilevel model of drug abuse inside prison. The Prison Journal, 85, 223-246

12.

Lanza-Kaduce

Parker

F. F.

Thomas

C. W.

(1999). A comparative recidivism analysis of releasees from private and public prisons. Crime & Delinquency, 45, 28-47.

13.

Legger

R. G.

(1988). Perception of crowding, racial antagonism, and aggression in a custodial prison. Journal of Criminal Justice, 16, 167-181.

14.

Logan

(1992). Well kept: Comparing quality of confinement of private and public prisons. Journal of Criminal Law and Criminology, 83, 577-603.

15.

Logan

(1996). Public vs. private prison management: A case comparison. Criminal Justice Review, 21, 62-85.

16.

Logan

Rausch

(1985). Punish and profit: The emergence of private enterprise prisons. Justice Quarterly, 2, 303-318.

17.

McDonald

Patten

(2005). Governments’ management of private prisons (NCJ 203968). Washington, DC: National Institute of Justice.

18.

Office of Program Policy Analysis and Government Accountability (1998). Review of Bay correctional facility and Moore Haven correctional facility report (Report No. 97-68). Tallahassee, FL: Author.

19.

Office of Program Policy Analysis and Government Accountability (2000). South Bay correctional facility provides savings and success: Room for improvement (Private Prison Review, Report No. 99-39). Tallahassee, FL: Author.

20.

Ogle

(1999). Prison privatization: An environmental catch-22. Justice Quarterly, 16, 579-600.

21.

Osgood

D. W.

(2000). Poisson-based regression analysis of aggregate crime rates. Journal of Quantitative Criminology, 16, 21-43.

22.

Paulus

P. B.

Cox

V. C.

McCain

(1988). Prison crowding: A psychological perspective. New York, NY: Springer-Verlag.

23.

Perrone

Pratt

T. C.

(2003). Comparing the quality of confinement and cost-effectiveness of public versus private prisons: What we know, why we do not know more, and where to go from here. The Prison Journal, 83, 301-322

24.

Pratt

T. C.

Maahs

J. M.

(1999). Are private prisons more cost-effective than public prisons? A meta-analysis of evaluation research studies. Crime and Delinquency, 45, 358-371.

25.

Reisig

M. D.

Pratt

T. C.

(2000). The ethics of correctional privatization: A critical examination of the delegation of coercive authority. The Prison Journal, 80, 210-222.

26.

Ruback

R. B.

Carr

T. S.

(1993). Prison crowding over time: The relationship of density and changes in density to infraction rates. Criminal Justice & Behavior, 20(2), 130-148.

27.

Sabol

W. J.

Couture

Harrison

P. M.

(2007, December). Prisoners in 2006[Monograph]. Bureau of Justice Statistics Bulletin. Washington, DC: U.S. Department of Justice. Retrieved from http://bjs.ojp.usdoj.gov/content/pub/pdf/p06.pdf

28.

Sechrest

D. K.

Shichor

(1996). Comparing public and private correctional facilities in California: An exploratory study. In Mays

G. L.

Gray

(Eds.), Privatization and the provision of correctional services: Context and consequences (pp. 33-151). Cincinnati, OH: Anderson.

29.

Sellers

(1989). Private and public prisons: A comparison of costs, programs and facilities. International Journal of Offender Therapy and Comparative Criminology, 33, 241-256.

30.

Shichor

(1995). Punishment for profit: Private prisons/public concerns. Thousand Oaks, CA: SAGE.

31.

Stephan

J. J.

Karlberg

J. C.

(2003). Census of state and federal correctional facilities, 2000. Washington, DC: U.S. Department of Justice, Bureau of Justice Statistics.

32.

Thomas

(1997). Comparing the cost and performance of public and private prisons in Arizona [Monograph]. Phoenix: Arizona Department of Corrections.

33.

Thomas

(2005). Recidivism of public and private state prison inmates in Florida: Issues and unanswered questions. Criminology and Public Policy, 4, 89-100.

34.

Urban Institute. (1989). Comparison of privately and publicly operated corrections facilities in Kentucky and Massachusetts (Project Report). Washington, DC.: Author.

35.

U.S. Dept. of Justice, Bureau of Justice Statistics. (2004).Census of state and federal adult correctional facilities, 2000 (ICPSR ed.) [Computer file]. Conducted by U.S. Dept. of Commerce, Bureau of the Census[producer]; Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor].

36.

Wooldredge

Griffin

Pratt

T. C.

(2001). Considering hierarchical models for research on inmate behavior: Predicting misconduct with multilevel data. Justice Quarterly 18, 203-231.