The paucity of literature on whether the private sector is more successful than the public sector in rehabilitating prison inmates, as measured by reducing recidivism, is apparent (Austin and Coventry, 2001). To date, only three prior studies have compared the recidivism rates of private versus public prison inmates.4 While their results have not been consistent, prior research has indicated that at least some inmates released from private prisons recidivated somewhat less than inmates released from public prisons. These studies were all conducted using data from the FDOC.
Lanza-Kaduce, Parker and Thomas (1999) compared the recidivism of men released from private prisons and those released from public prisons in the state of Florida. The sample consisted of 396 males released from prison between June 1, 1996 and September 30, 1996; half were from public prisons, the other half were from private facilities. All of the subjects were classified by the FDOC as minimum or medium custody level at the time of release (close custody inmates were excluded). Inmates were defined as public or private according to the type of facility from which they were released, which included two private prisons and seven public prisons. Four recidivism measures were used: subsequent arrest, felony conviction, imprisonment for technical violation, and imprisonment for new offense. Recidivism data were collected for the 12 months following release.
Lanza-Kaduce et al. (1999) matched pairs of private and public inmates from the sample based on offense, race, prior record, and age. They achieved 149 pairs exactly matched on this important, but limited set of criteria, and 198 pairs matched by relaxing the age criterion. Analyzing these matched pairs in the aggregate, they found that releases from private prisons recidivated significantly less, using a sign test, than those released from public institutions on three of the four recidivism measures. Specifically, within 12 months following release, 10% of the private inmates were arrested compared with 19% of the public inmates; 6% of private inmates were convicted compared with 10% of public inmates; and 10% of private inmates were imprisoned for new offenses compared with 14% of public inmates. The authors found no significant difference between private and public inmate re-imprisonment for technical violations. In addition, using an aggregated measure for any indicator of recidivism, they found 17% of private inmates recidivated compared with 24% of public inmates.
In a subsequent study, Lanza-Kaduce and Maggard (2001) re-analyzed the same inmate pairs and recidivism measures from the 1999 study, extending the follow-up period through 48 months after release. In that study, the authors reported only one recidivism measure, imprisonment for either a technical violation or a new offense. Consistent with their earlier finding, they found that inmates released from private prisons recidivated at a lower rate than those released from public prisons over the longer follow-up period. Contrary to their earlier finding, however, this difference was at, best, marginally statistically significant (p<.10) based again on a sign test for matched pairs and only for a smaller, best matched subset (149 cases).
Farabee and Knight (2002) studied inmates released from FDOC between January 1, 1997 and December 31, 2000. The included inmates were released directly from public or private prisons. Inmates could not have been released to a detainer, but could be released at minimum, medium or close custody levels (n=8,848). The authors defined inmates as public or private by whether they had spent the last six months prior to release in a public or privately operated facility. For example, an inmate who was transferred to a private facility from a public facility three months prior to release was excluded. The basis for this exclusion criterion is the theoretical viewpoint, grounded in much program evaluation literature, that exposure to correctional programs for less than six months is unlikely to achieve a recidivism reduction benefit.
Farabee and Knight (2002) created a “matched” sub-sample using factors found to be significantly associated with recidivism in a FDOC (2001) study that analyzed the association between certain variables and recidivism. The authors matched public and private inmates on their commitment offense, custody level at release, race, age at release, education level, prior recidivism, the number of months served in prison, and the number of months since their release from prison. In presenting their results, they divided subjects into three groups; adult males, adult females, and youthful offender males.5 Inmates from public prisons included 4,912 adult men, 612 adult women, and 1,945 youthful offender males, and were compared to 2,341 adult men, 983 adult females, and 314 youthful offender males from private prisons.
Recidivism was defined in two ways: conviction for a new offense and incarceration for a new offense. The follow-up period for their study was three years after release. To compare the recidivism rates of the private inmates (treatment group) and public inmates (control group), the authors used a proportional hazard regression model, controlling for the same factors used to match the two groups. The data revealed that adult males released from public prisons and private prisons displayed rates of re-offense and re-imprisonment that were not statistically significantly different. In contrast, adult females released from private prisons had significantly lower rates of re-offense and re-incarceration than adult females released from public prisons over the three-year period (p<.05) (Farabee and Knight, 2002). They found women released from private facilities were 25% less likely to re-offend and 34% less likely to be re-incarcerated than female inmates released from public facilities. For youthful offender males, no significant differences were found in recidivism rates for public versus private inmates.
The current research makes several methodological improvements over the Lanza-Kaduce, et al. (1999) and Farabee and Knight (2002) studies. First, the method of measuring the critical private versus public prison experience is advanced. The first study compared inmates released directly from a private prison with those released back into the community from a public facility with no consideration given to the length of time spent in each type of facility (Lanza-Kaduce, Parker, and Thomas, 1999). This measure of the private prison effect has methodological problems. First, it results in a smaller case size for the treatment group, which may limit the reliability of recidivism rate estimates and the significance of their differences from a control group. Second, it does not quantify the amount of time in either, or both, private and public prisons. An analysis of the cases used in the Lanza-Kaduce, et al. study conducted by the FDOC,6 found that of the 198 inmates the researchers identified as private prison inmates, 69 (35%) had also been incarcerated in a public prison, other than a reception center. Of these 69“private” inmate cases that had both private and public prison time, 52 (75%) served more time in a public facility than in a private facility. This clearly shows that this early study did not adequately measure exposure to private prisons, by failing to exclude inmates exposed to both private and public prisons.
The second study was more refined in its measurement of the experimental effect (Farabee and Knight, 2002). This study, first, analyzed only inmates who had spent at least six months in the facility from which they were released to the community. Despite a theoretical basis for this method, from evaluations of correctional programs in the literature, it is not clear that this particular time limit rather than a longer one (e.g., 12 months) would be more appropriate when analyzing exposure to private prisons. Second, inmates were defined as having the private prison effect only if they were released from a private prison facility and met the first criterion. However, inmates who spent a substantial portion or length of their total prison time in a private prison but were transferred to a public prison just prior to release for medical reasons, to transfer closer to their release destination, etc., were excluded from the study. Whether excluding these cases from the final analysis resulted in any biased findings is unknown. Further, inmates who served significantly more than six months in a public facility but were released directly from a private facility after serving six months there would be defined as private inmates when in fact their public prison experience was more extensive. Though Farabee and Knight (2002) certainly improved on the method of identifying public versus private inmates over Lanza-Kaduce et al. (1999), there is simply not sufficient extant research to rely on only a single definition and measurement of exposure to private prisons.
The current study addresses the concerns above and improves on the methodological issues of the two previous studies. For this study, three concepts relating to the experimental effect of the private prison experience were used to guide the development of six different experimental measures. First, whether inmates were released directly from a private versus a public prison facility, regardless of the differentials in time served in each facility type. Second, the level of purity in the experimental effect measured by the time served, as actual time or percent of total time, in private prisons. Third, experimental measures that required specified periods of duration and percentages of total prison time in a private prison.
The purpose of employing differing measurements of the experimental group of inmates serving time in private prisons is threefold. First, there is minimal theoretical basis for preferring one method of quantifying a private prison experimental group. Using multiple definitions of the treatment group prevents concluding that an effect does or does not exist based simply on the choice of treatment group definition. Second, the consistent presence or absence of an effect derived from more than one definition of the treatment group will lend weight to the research conclusions. Third, if only one of several treatment group definitions yields an effect, the nature of the experimental group definition may explain what aspect of exposure to private prisons accounts for the effect. Such information, from an empirically-driven analysis in the present study, may then better guide future research on the effectiveness of private prisons.Notes: