Вы находитесь на странице: 1из 7

Journal of Criminal Justice 36 (2008) 389395

Contents lists available at ScienceDirect

Journal of Criminal Justice

The effect of faith program participation on prison misconduct: The Life Connections Program
Scott D. Camp , Dawn M. Daggett, Okyun Kwon, Jody Klein-Saffran
Ofce of Research and Evaluation, Federal Bureau of Prisons, Washington, DC 20534

a r t i c l e

i n f o

a b s t r a c t
Faith-based programming is one of many potential tools for preparing inmates for successful reentry into society. The current study used ofcial records of inmate misconduct and attitudinal survey data to investigate whether participation in a faith-based program reduced the likelihood of prison misconduct. The results indicated that program participation did lower the probability of engaging in serious forms of misconduct. No effect was discovered, though, for less serious forms of misconduct or for both types of misconduct considered simultaneously. Published by Elsevier Ltd.

Introduction Religion has been an important component of corrections in the United States dating back to the rst prison (Clear & Myhre, 1995; DixRichardson & Close, 2002; Rothman, 1995). The Eastern State Prison in Philadelphia was designed to hold inmates in silence and solitude so that they could reect upon their sins and be reformed. The inmates were required to perform penance for previous indiscretions. Currently, all federal prisons employ chaplains who provide religious services and spiritual guidance. There are also many volunteers with specic faith afliations who lead study of sacred scripture, counsel inmates, and serve as mentors. What is new in federal prisons is the development of a formal faith-based program, with curricula, targeted to decrease the propensity of criminal behavior and attitudes of inmates while incarcerated and upon release to the community, i.e., the Life Connections Program (LCP). This program is unique because it is a multi-faith approach which explicitly accommodates any faith choice, including atheism. The program is an eighteen-month residential program, meaning that the inmates live together in the same housing unit while in the program. Formal prison programs based on faith principles only recently appeared, therefore there are a limited number of studies that have evaluated this form of correctional programming. Johnson, Tompkins, and Webb (2002) point out, though, that the paucity of studies characterizes most areas where faith-based organizations provide services, not just in corrections. The few evaluations that have been conducted in corrections are restricted by methodological limitations, such as the reliance upon self-reports of outcome measures (for

examples see Clear & Sumter, 2002; Kerley, Matthews, & Blanchard, 2005) and/or limits to model specication created by unmeasured differences between the treatment group receiving the faith program and the comparison subjects (Camp, Klein-Saffran, Kwon, Daggett, & Joseph, 2006). The current study addressed some of the methodological shortcomings of previous evaluations of faith-based programs. In addition, a faith-based program with the explicit intention of changing inmate behavior, not just changing their spirituality, was the target of the study. The study addressed whether participation in the LCP was associated with a decline in prison misconduct while in the program. Only a few inmates completed the LCP by the time of this analysis, therefore it was not possible to examine their prison behavior after leaving the program. The current study used a quasi-experimental research design to compare ofcial misconduct data with the propensity-score approach pioneered by Rosenbaum and Rubin (Rosenbaum, 1995; Rosenbaum & Rubin, 1983, 1985). Prison misconduct is not typically used in evaluations of programs, though French and Gendreau (2006) point out that this is an opportunity missed. Prison misconduct is more readily available and timely than recidivism measures, and it captures antisocial behavior that cuts across social settings. Others have also found prison misconduct to be a reasonable proxy for criminal behavior after release (French & Gendreau, 2006; Gendreau, Goggin, & Law, 1997; Homant & Witkowski, 2003; Zamble & Poporino, 1988). Literature review There is a general lack of theorizing about the impact of religious programs upon change in inmate behavior that is matched by a lack of theorizing about the impact of religion upon crime. This is a surprising omission given the importance of religion in other areas of social

Corresponding author. Tel.: +1 202 616 7248; fax: +1 202 307 5888. E-mail address: scamp@bop.gov (S.D. Camp). 0047-2352/$ see front matter. Published by Elsevier Ltd. doi:10.1016/j.jcrimjus.2008.07.004

390

S.D. Camp et al. / Journal of Criminal Justice 36 (2008) 389395

theory and the role that religion assumed in traditional societies in imposing and enforcing a moral code. With the industrial revolution in Western nations and the rise of more secular societies, there is no doubt that the established church lost its position of supremacy in dictating the moral code of society and dening the nature of crime (Sutherland & Cressey, 1947), but religion remained a powerful agent of socialization. Differential association and social learning theories, for example, failed to capitalize on the role of religion, even though religious practice could provide a cornerstone for the social networks and social denitions that the theories respectively postulate as important for a crime-free lifestyle (Akers, 1977; Sutherland & Cressey, 1947). One of the remaining unknowns for each theory is the impact that social learning and association with peers have for adults. Most research within the respective traditions has focused upon adolescents, usually adolescent males. Likewise, social control theory, reecting the emphasis upon secular institutions found in criminological theory, emphasizes the development of social bonds and the commitment to and involvement with conventional activities as discouraging crime. While this would seem to point toward the importance of religion, religion is given practically no attention as an agent for developing bonds and commitment to conventional values. Instead, the emphasis has historically included the institutions of the family and education (Hirschi, 1969; Krohn & Massey, 1980), again reecting the emphasis upon adolescents. More recently, the age-graded theory of informal social control put forward by Sampson and Laub (1993) has included social bonds in adulthood such as marriage and employment. The signicance of religion for criminological theory has been mentioned recently in several contexts, although actual development has yet to follow from these pronouncements. Myers (2000), for example, criticized Hagan and McCarthy (1998) for not including bonds to conventional activities such as religion in their integration of differential association, social control, and conict theories. Laub and Sampson (2003) also have noted that the importance of forms of informal social control, such as religion, is largely unknown with respect to their effects upon persistence and desistance of crime. Religion has been used for many purposes and as justication for many practices within the specic context of prison. In addition to the often-stated purpose of bringing change to inmates, religion was also used as a means of institutional control over those incarcerated there. The tables have turned somewhat though concerning religion in prisons. There is currently concern that religion can be used to recruit inmates to radical political ideologies. While there are serious analyses of these changes (Pisciotta, 1994; Rothman, 1980), the omission of these concerns from mainstream accounts of criminality is somewhat surprising.1 Despite the lack of theoretical and empirical work, or perhaps because of it, religious programs and activities survived the downsizing of prison programs in the 1970s and continue to maintain a strong presence in the correctional setting (Kerley et al., 2005; Sumter & Clear, 2002), in part because they are relatively inexpensive (O'Connor & Perreyclear, 2002). Many correctional facilities offer a variety of religious services and programs which include large prison ministries throughout federal prisons and the fty states. Yet only a few studies examined the effect of religion or religious programs on either institutional misconduct or post-release behavior (Clear et al., 1992; Clear & Sumter, 2002; Johnson, 1984, 2002, 2004; Johnson, Larson, & Pitts, 1997; Kerley et al., 2005; O'Connor & Perreyclear, 2002; Young, Gartner, O'Connor, Larson, & Wright, 1995). Religious programs in prison, unlike most other programs, historically did not include a formal curriculum of standard and required program steps. For the most part, religious programs were assumed to work by bringing salvation or transformation to the inmates, and this new outlook would encourage inmates to lead a life without crime. Since prison programs were traditionally Christian, this focus upon salvation followed from the evangelical orientation of

most Christian denominations. Studies of the impact of religion in prison, for the most part, were designed to capture the effect of religiosity upon behavior, not the impact of a religious program. One of the original studies to examine the relationship between religion and prisoner adjustment was conducted by Johnson (1984). While his study failed to nd any identiable impact of religious activity on prison adjustment, Johnson paved the way for other researchers to begin to investigate the impact of religion on prison misconduct (Clear et al., 1992). One of the ways that religion was believed to help inmates while in prison was that it helped inmates adjust psychologically and emotionally to a foreign and uncertain environment, a fact that may be underplayed if attention is focused upon reducing recidivism (Clear, Hardyman, Stout, Lucken, & Dammer, 2000; Clear & Sumter, 2002). Clear et al. (1992) found that younger inmates in their study used religion as an aid in their psychological adjustment to imprisonment and that older inmates had fewer infractions of misconduct if they were involved in religion. A later study by Johnson et al. (1997), however, found no overall effect for participation in Prison Fellowship on measures of institutional adjustment and recidivism unless the analysis controlled for the level of participation in Prison Fellowship. Inmates with high levels of participation in Prison Fellowship were less likely to commit infractions than their counterparts who participated at a medium or lower rate of participation. Inmates with high levels of participation were also less likely to commit serious infractions than the nonreligious inmates. High participation in Prison Fellowship also reduced the likelihood of an arrest during the follow-up period. According to Johnson et al. (1997), this relationship appeared to be strongest for White Prison Fellowship inmates and was nonsignicant for African Americans in the sample. Sumter (1999) added ofcial criminal history data to the data collected by Clear et al. (1992) to determine if an inmate's religiousness inuenced post-release adjustment. The results did not indicate any difference in the recidivism rates of those involved in Prison Fellowship and those who were not. What was signicant in her research was that inmates who had a greater orientation in terms of their values at the time of release as well as those who attended some religious programming were less likely to recidivate (Sumter, 1999). Moreover, Sumter demonstrated that self-reports of participation in religious programming were associated with lower levels of rearrest as measured by ofcial FBI criminal history reports. The results of religion upon misconduct while in prison and arrest after release are mixed in the literature. Most results are from studies where an explicit evaluation of a program was not the intent of the study. Instead, given the focus upon salvation or religiosity, the studies were generally more concerned with demonstrating that high levels of religiosity, usually measured with the proxy measure of level of participation, were associated with favorable outcomes. There was less concern given to assessing whether specic programs were successful in bringing about the higher levels of religiosity or whether the religiosity existed prior to program participation. Prior evaluations of faith-based programming as opposed to the effect of religiosity have been few. The best planned of these studies was conducted by Johnson and Larson (2003). The original research design called for random assignment of program volunteers into participation and the control group. There were not enough volunteers to make this strategy feasible, however, and all volunteers were allowed to enter the program. With the original design subverted by practical limitations, Johnson and Larson then created matches for the program volunteers from inmates who did not volunteer for the program on a few variables: race, age, offense type, and salient factor score. When Johnson and Larson compared the postrelease behavior of all inmates who entered treatment and the control group, they did not nd an effect. An effect for reducing recidivism was found only if the treatment group was further limited to only

S.D. Camp et al. / Journal of Criminal Justice 36 (2008) 389395

391

those inmates who completed all phases of the program, including the post-release requirements. The results of the Johnson and Larson study suffer from at least two problems. First, it is highly unlikely, though, that the limited number of variables they used to create matched groups sufciently captured pertinent differences between the two groups. Camp et al. (2006) found that several factors not used in the Johnson and Larson study were related to participation in the LCP. The factors included several dimensions of religiosity, motivation for change, and related factors. The second related problem with the Johnson and Larson (2003) study is potentially more problematic. In comparing the subset of inmates who completed all phases of the faith program to the matched comparisons, Johnson and Larson violated the logic of their amended research design, a fact they themselves acknowledged but many proponents of faith programming ignored. In short, by the time the nal comparison was made by Johnson and Larson (2003), the unmeasured differences between the two groups may have outweighed the program effects of participation in the InnerChange Freedom Initiative. There is simply no method to make the determination given the limitations of their data. The Johnson and Larson (2003) study remains the most complete to date, and the analysis provided here built upon the approach taken by Johnson and Larson to identify more appropriate comparison groups. LCP The LCP at the Bureau of Prisons (BOP) provides support for encouraging desistance from crime in three ways identied as important in the literature. First, there is a moral component to the program where pro-social values are encouraged (Cullen & Gendreau, 2000), and for most inmates this is within an established religious framework. For agnostics and atheists, the values are formed from a more secular perspective. Second, in a related fashion, the program focuses upon having inmates recognize the cognitive decits that brought them to prison and to accept responsibility for harming victims with their actions (Samenow, 1998, 2004; Yochelson & Samenow, 1976). Many BOP programs are based upon a cognitivebehavioral framework, and the LCP is no exception. Not only are the inmates forced to identify their criminal thinking in structured exercises, they practice appropriate behaviors within the residential setting. Finally, the LCP recognizes the importance of social support and networks during the transition back to the community. Mentors from the local faith community are identied by Volunteers of America, and the volunteers work with the inmates to establish a lifestyle different from the one that brought them to prison. The LCP has a formal curriculum that all inmates follow. The program is residential and lasts for eighteen months. The same curriculum is used by all inmates, although the material is taught by spiritual guides (religious leaders contracted from outside the prison). The spiritual guides are hired by the LCP chaplain who serves as the administrator for the program and leads common activities such as sharing the word of the day. The spiritual guides help the participants to place the curriculum material within the different faith traditions represented by the inmates. Each program has up to ve spiritual guides at one time, each from a different faith group. Inmates also work with in-prison mentors drawn from faith organizations in the local communities surrounding the prisons. Upon release, inmates work with different mentors in the communities to which they are released. These mentors are identied by Volunteers of America. Inmates who are releasable in the U.S., speak English or fullled ESL obligations, have completed or are completing GED obligations, are meeting nancial responsibilities, and are within 24 to 60 months of release are eligible for the program. The release requirements are different for high-security male inmates and female inmates. For highsecurity male inmates, there is no requirement, and for females they must be within 24 and 120 months of release. A typical starting LCP

cohort has between thirty and fty inmates, and cohorts are spread out by six months so that three cohorts operate at an LCP prison at any one time. There is explicit focus upon personalizing the program to the religious values of the participants while at the same time employing techniques used in secular prison programs, such as encouraging inmates to accept responsibility for their own actions and practicing pro-social behaviors in a therapeutic environment (Sundt, Dammer, & Cullen, 2002). Without abandoning a core emphasis upon the religious values of the inmates, the LCP is more in the mainstream of typical prison programs than religious programs that lack the same level of structure and curriculum. Data and methods used The data for the current study were a combination of ofcial records and self-reported survey data. Ofcial records were used to obtain information on misconduct, program participation, and basic socio-demographic information. The self-reported data asked study participants about their religious practices, attitudes and knowledge of religion, motivation to change, and other relevant factors. The identication of new study subjects is ongoing, but the study participants identied here were surveyed between August 2004 and October 2005. The LCP participants were identied at the ve program facilities in operation during this period. The comparison subjects were identied at thirteen different prisons that matched the LCP site prisons on sex of the inmates and security level of the prison. Surveys were administered by members from the Ofce of Research and Evaluation in the Central Ofce of the BOP. The response rate for the survey was 68.2 percent. For the LCP participants, the response rate was 65.8 percent, and for the comparison group it was 69.4 percent. Ofcial records and survey data were collected for 1,611 inmates. Twenty-one survey respondents were dropped from the analysis because of improper coding of register number and/or name on the surveys. Of the study participants, 443 were enrolled in the LCP, and 1,147 inmates were comparison subjects. The survey data used here were provided and described by Camp et al. (2006). LCP participants were surveyed at a point near the time they started the program. For comparison subjects, the LCP criteria (including time left to serve) were used to screen inmates at non-LCP prisons, and random samples of these inmates were drawn. Non-LCP prisons were selected for identifying potential comparison subjects with the presumption that more inmates who are otherwise like LCP inmates would exist at these prisons without an LCP than at prisons with an LCP. The survey data were fairly complete in terms of item response, although there were some missing data. The question with the largest number of missing responses was the current religious afliation of the respondent. Data were missing for 77 of the 1,590 respondents. To retain these observations, values for the missing data were imputed with the techniques described by Schafer (1997) and incorporated into the SAS procedure MI. Five complete data bases were created and analyzed separately. Methods The issue of selection bias is endemic to observational studies of voluntary programs that attempt to replicate the experimental design of comparing treated (treatment) subjects and untreated (control) subjects. One of the more common methods for dealing with the bias in observational studies created by nonrandom assignment to groups is propensity score adjustments (Rosenbaum, 1995; Wainer, 1986). While there are several ways in which propensity scores can be used (Parsons, 2001), this study used the most common technique. The scores were used to create post hoc treatment-control matches among inmates. The matched groups were then compared on the outcome of interest, prison misconduct. In essence, this approach mimics the

392

S.D. Camp et al. / Journal of Criminal Justice 36 (2008) 389395

assignment of groups in a classical experiment. Where a welldesigned experiment produces comparable distributions on observed and unobserved covariates across the treatment and control groups, propensity score matches produce comparable distributions on observed covariates. To the extent that the observed covariates encompass the process by which individuals were selected into the treatment group, the results approximate those from an experimental design. Propensity models predicting LCP participation included the major variables identied in the literature regarding both participation and misconduct as well as variables not previously used, especially a scale of motivation for change. Variables were included for frequency of spiritual experiences, whether the respondent practiced religion before coming to prison, frequency of current religious practice, current religious afliation, feelings of self-worth, desire to integrate in the community upon release, a scale measuring internal motivation for change, the custody risk of the inmate (BOP measure summarizing criminal history), the number of previous incarcerations, age, Hispanic ethnicity, race, sex, education, marital status, and months of current incarceration before entering the study. The variables used in the propensity score models calculated here match the variables used by Camp et al. (2006) with one major exception. Camp et al. used history of misconduct before entering the program. Misconduct was not used here because it was a dependent variable in later models. Fit for the propensity score models on the ve complete data bases was good as the ROC value ranged between .800 and .804. A value of 1.00 is the maximum possible ROC score, and a value of .50 means that the model did no better in predicting LCP participation than a coin ip. Values above .70 are considered acceptable. Likewise, the Hosmer-Lemeshow t statistic also suggested a good t between predicted and observed LCP participation. For the ve complete data bases, the HosmerLemeshow value was smaller than the degrees of freedom, which meant that the probability values ranged between .6143 and .8808. The small values of the Hosmer-Lemeshow statistics show that the predicted values closely match the observed values. Complete results for the propensity score models are available from the authors. Treatment-control matches were made using the propensity scores with the PSMATCH2 module for Stata developed by Leuven and Sianesi (2003), the greedy matching technique implemented in SAS and described by Parsons (2001), and the NNMATCH module for Stata developed by Abadie, Drukker, Herr, and Imbens (2004). In all of the matching strategies, unique one-to-one matches were created since the number of potential control cases was not sufcient for multiple matches per treated inmate. Different matching strategies were used after a preliminary examination of the frequency distributions for the propensity scores revealed that the potential comparisons were distributed toward the lower propensity values of participation than was true for the LCP participants. Histograms of the respective propensity score distributions are available upon request. Five stratifying cells were created by cross-classifying sex (male-female) and misconduct (none, one, two or more). The cells for females with one incident and multiple incidents were combined because of limited observations, thus producing the ve cells instead of six. For the PSMATCH2 routine, a match based upon the propensity score was made for each cell with a caliper width of .25 times the standard deviation of the propensity scores. The matches were made with the nearest neighbor without replacement options. Matches were made for 87.8 percent of the LCP participants. For the greedy match algorithm, matches were made for 81.7 percent of the LCP participants. For the NNMATCH module, the matches are done internally to the module, and the number of matches is not reported. Exploratory analysis of the inmates without matches revealed that they tended to have higher propensity scores for program participation, were more likely to come from a prison with a higher security level, and were no more likely to be involved in misconduct during program participation. This latter nding is particularly relevant as it

suggests that the matching algorithms did not produce a subset of the LCP participants with articially low rates of misconduct for the analysis presented below. The dependent variable examined here was prison misconduct, but prison misconduct covers a whole range of behaviors from the very serious, such as a homicide or an escape, to the less serious, such as rules about where to smoke or when to move between buildings. The BOP codes misconduct to reect the seriousness of the offense. The most serious offenses in the BOP are those that are in the range of 100 to 199. The next most serious types of misconduct are those in the 200 to 299 range. The less serious forms of misconduct in the BOP are those in the 300 and 400 levels. A complete list of the BOP misconduct codes can be found in the Appendix to Camp, Gaes, Langan, and Saylor (2003). There were so few instances of any one specic type of misconduct, thus, it was necessary to group types of misconduct together. All forms of serious misconduct (codes 100-299) were classied together, and the less serious misconduct (300-499) was likewise grouped. The groups were compared for the three types of misconductall, serious, and less serioususing 2 tests for crosstabulation tables for each matching scheme. An important methodological note is that LCP participants were followed for misconduct for eighteen months, even if they withdrew from the LCP. The reason for this was to insure comparability of the two groups. The matches used in this study were based upon how the groups looked at entry, not completion of the program. Not counting misconduct by LCP participants after they left the program but prior to the completion of eighteen months would have provided an undeserved advantage to the LCP participant group. Descriptive statistics As noted previously, not all LCP participants were used in the analysis, but their exclusion did not seem to threaten the external validity of the ndings reported below. The inmates without matches were more likely to volunteer for treatment, so their exclusion did not produce a subset of star participants. As noted, the LCP participants without matches were just as likely to be involved in the outcome of interest for the present study, prison misconduct, as the matched LCP participants. Table 1 presents results of t-tests comparing LCP participants and the constructed comparison group on the variables used to generate the matches. Table 1 is a composite of the information produced for the ve multiple imputed data bases. The effect is that Table 1 gives the average characteristics used in the ve separate runs. The variables used in Table 1 (and in generating the matched comparisons) are comprehensive of those used in other evaluations of faith-based programs, and it includes measures of motivation, religiosity, desire for community integration, and self-worth not found in other evaluations of faith-based programs. As can be seen there, the only ways in which the two groups signicantly differ is that the LCP participants were at risk for misconduct, especially serious misconduct, longer than comparison inmates. If anything, this difference makes it harder to show an effect for LCP participation reducing misconduct. Table 1 also presents information on the number of prior instances of misconduct for the matched groups on average. As can be seen there, the two groups did not differ on the number of prior misconduct incidents for the three classications of misconduct. This is interesting because prior history of misconduct was not included in the models predicting propensity for LCP participation. It provides further evidence that the LCP group was not somehow different in terms of being at less risk for misconduct. Findings of the impact of LCP participation on prison misconduct As noted, the data were examined by using three different matching strategies to create post hoc groups for comparison with

S.D. Camp et al. / Journal of Criminal Justice 36 (2008) 389395 Table 1 Descriptive statistics of variables used in creating matches Variable name Wk34 Wk12 wkall monprior godly RLGN_BEF Rlgn_acx crel_c crel_i crel_o selfwrth Commsprt Prochtot P1BASE INCR_FRQ Age Hispanic Black Female Bed_hs Bed_byhs cmar_mar cmar_div Pcntall pcnt12 pcnt34 LCP 54.016 Comparison Difference t-value p-value Variable label 52.201 1.816 4.026 3.412 - 2.745 0.066 0.004 0.099 - 0.009 - 0.025 0.025 - 0.005 0.039 0.102 - 0.065 - 0.004 0.025 - 0.004 - 0.039 0.000 - 0.009 0.045 0.006 0.013 - 0.025 - 0.006 - 0.019 - 1.006 - 2.304 - 1.848 0.894 - 0.740 - 0.124 - 0.728 0.392 0.926 - 0.864 0.092 - 0.760 - 0.652 0.192 0.054 - 0.036 0.206 1.090 0.000 0.332 - 1.366 - 0.192 - 0.414 0.278 0.338 0.180 0.329 0.029 0.075 0.423 0.472 0.723 0.475 0.507 0.366 0.441 0.841 0.492 0.566 0.677 0.746 0.880 0.723 0.319 1.000 0.730 0.190 0.777 0.694 0.782 0.746 0.850 Weeks at risk less serious misconduct Weeks at risk serious misconduct Weeks at risk all misconduct Months before entering study Spiritual experience scale Practiced religion Frequency of service, current Currently Catholic Currently Islamic Currently other religion Scale of self-worth Community integration scale Prochaska motivation scale Custody classication score Number of incarcerations Actual age in years Hispanic, 1 = yes Black, 1 = yes Female, 1 = yes High school equivalent More than high school Currently married Currently divorced/ separated Count of prior all misconduct Count of prior serious misconduct Count of prior less serious misconduct

393

Table 2 Matched group comparisons for any type of misconduct results for the ve imputations across the three matching methods N PSMATCH2 method Imputation 1 Imputation 2 Imputation 3 Imputation 4 Imputation 5 Greedy match method Imputation 1 Imputation 2 Imputation 3 Imputation 4 Imputation 5 NNMATCH method Imputation 1 Imputation 2 Imputation 3 Imputation 4 Imputation 5 No matching performed 782 776 778 774 778 LCP 19.44% 21.39% 20.82% 20.41% 19.79% Comparison 23.02% 23.71% 22.62% 22.48% 22.11% Diff. (LCP-comp.) - 3.58% - 2.32% - 1.80% - 2.07% - 2.32% p-value 0.221 0.439 0.543 0.484 0.428

58.044 54.018 52.675 40.507 4.880 0.693 6.514 0.095 0.157 0.205 4.217 3.938 9.395 10.424 1.685 38.702 0.080 0.564 0.204 0.152 0.332 0.284 0.287 1.905 0.687 1.219 49.264 43.252 4.814 0.689 6.415 0.103 0.182 0.180 4.222 3.899 9.293 10.489 1.689 38.677 0.084 0.603 0.204 0.161 0.287 0.278 0.274 1.930 0.693 1.238

581 586 580 579 592

19.35% 21.18% 21.53% 19.71% 20.29%

23.67% 23.58% 23.24% 22.59% 23.02%

- 4.32% - 2.40% - 1.71% - 2.88% - 2.73%

0.207 0.490 0.627 0.400 0.425

1,590 1,590 1,590 1,590 1,590 1,590

NA NA NA NA NA 19.86%

NA NA NA NA NA 19.97%

2.89% 2.77% 2.52% 2.64% 2.77% - 0.11%

0.290 0.306 0.350 0.325 0.302 0.9641

Note: NA = not available. The NNMATCH procedure does not provide base rates for the two groups.

the comparison subjects in terms of having any type of misconduct during the study period. The results produced with the matches made by PSMATCH2 and Parsons' greedy algorithm seemed to nd that the LCP group had the lowest incidence of occurrence, whereas the comparisons made with data produced by the NNMATCH procedure suggested the opposite. Since none of the results was signicant, however, there is no real signicance to this observation. The results for serious misconduct are presented in Table 3. Again, all of the matching methods suggest the same conclusion. Inmates involved in the LCP were less likely to have an instance of serious misconduct than the comparison group. In fact, according to the matches produced by the PSMATCH2 and greedy algorithms, inmates in the LCP had about half of the amount of serious misconduct, as did

Simple average values are presented for the ve complete data bases created by multiple imputations. The sex of the inmates was the only variable that required an exact match. As a result, the proportion of females in the LCP and comparison groups was identical by denition, producing the otherwise strange result of a t-value of exactly 0.

Table 3 Serious misconduct N LCP 5.37% 5.67% 5.66% 5.43% 5.40% Comparison 11.51% 10.82% 11.05% 9.82% 10.54% Diff. (LCP-comp.) - 6.14% - 5.15% - 5.39% - 4.39% - 5.14% p-value 0.002 0.009 0.007 0.021 0.008

simple cross-tabulation tables. Since the outcome variables were binary, misconduct or not, the more commonly employed t-tests of means were not appropriate. Instead, the comparisons were between proportions of individuals falling into the respective categories and were based upon Pearson chi square statistics. These results are presented in Tables 24 below. Matched group comparisons The results for an inmate having any type of misconduct during the study period are presented in Table 2. As can be seen there, the results are presented for each of the three matching methods and for an unmatched representation of the data. The results are also inclusive of the multiple imputation method used to create the ve data bases. Since it was possible to present all of the results for this type of analysis, they were included in Table 2, as well as in Tables 3 and 4 where misconduct is decomposed into serious and less serious misconduct. All methods for examining the data suggest that the LCP participants did not differ in a statistically signicant fashion from

PSMATCH2 method Imputation 1 Imputation 2 Imputation 3 Imputation 4 Imputation 5 Greedy match method Imputation 1 Imputation 2 Imputation 3 Imputation 4 Imputation 5 NNMATCH method Imputation 1 Imputation 2 Imputation 3 Imputation 4 Imputation 5 No matching performed

782 776 778 774 778

581 586 580 579 592

5.36% 5.29% 6.19% 5.00% 5.29%

11.02% 11.38% 11.62% 10.46% 11.90%

- 5.66% - 6.09% - 5.43% - 5.46% - 6.61%

0.012 0.007 0.021 0.013 0.004

1,590 1,590 1,590 1,590 1,590 1,590

NA NA NA NA NA 5.64%

NA NA NA NA NA 8.63%

- 3.71% - 3.77% - 3.84% - 3.58% - 3.90% - 2.99%

0.029 0.030 0.025 0.037 0.022 0.0464

Note: NA = not available. The NNMATCH procedure does not provide base rates for the two groups.

394 Table 4 Less serious misconduct N PSMATCH2 method Imputation 1 Imputation 2 Imputation 3 Imputation 4 Imputation 5 Greedy match method Imputation 1 Imputation 2 Imputation 3 Imputation 4 Imputation 5 NNMATCH method Imputation 1 Imputation 2 Imputation 3 Imputation 4 Imputation 5 No matching performed 782 776 778 774 778 LCP 16.11% 17.78% 17.48% 17.05% 16.45% Comparison 15.35% 16.49% 15.17% 15.50% 14.91%

S.D. Camp et al. / Journal of Criminal Justice 36 (2008) 389395

Diff. (LCP-comp.) 0.76% 1.29% 2.31% 1.55% 1.54%

p-value 0.768 0.634 0.383 0.559 0.554

determine whether differences existed between LCP participants and comparison group members in terms of their involvement in misconduct while subjects in the study. The answer was yes, but only for serious misconduct. The results also demonstrate the importance of doing statistical adjustments before comparing the LCP participants and comparisons. Contrary to what might be assumed, the unadjusted comparisons produced a result demonstrating a difference between LCP participants and comparison subjects, but the effect was smaller than the effect found in the analyses on the matched groups. Additionally, the statistical power for the unmatched groups was lower. Conclusion The different matching methods used to create comparison groups and analyze the occurrence of misconduct during the study period produced consistent results. LCP participants were just as likely as the comparison subjects to be involved in all forms of misconduct considered simultaneously. When all misconduct was parsed into the categories of serious misconduct and less serious misconduct, then an effect for LCP participation on serious misconduct was demonstrated. The consistency in ndings across the three matching strategies lends credibility to the nding. The present study was the most comprehensive to date in terms of using a wide variety of observed variables to create matched comparisons that were then used to evaluate the impact of faith-based programs, at least with respect to prison adjustment as measured by prison misconduct. Especially notable in terms of the covariates used were covariates that captured ve of the six dimensions of religiosity described by Johnson, De Li, Larson, and McCullough (2000) and a scale measuring motivation to change. The ndings provide limited support for the effectiveness of the LCP, but there is the hanging question of why LCP participants did not differ from the comparison subjects on the less-serious forms of misconduct, what prison employees often call nuisance misconduct. First, it is important to understand that these types of misconduct include such offenses as failing to follow a work order, being in an unauthorized area, being unsanitary or untidy, and smoking in an unauthorized area. There are some more general offenses among the 300 and 400 levels in the BOP that are not specic to being in prison, such as gambling and indecent exposure, but these offenses account for a very small part of the categories and counts of less-serious offenses. It is possible that program participation has the perverse effect of raising the likelihood of receiving a ticket for these offenses because of increased surveillance from program staff and perhaps even custody staff while in the program. Future studies should try to tease out this possibility, although this is a difcult topic to address. Do these ndings indicate that faith-based programs work? This is a very good question, but one that is not yet answered satisfactorily. On a single dimension, albeit an important one, participating in a faith-based program reduced serious misconduct during the program, but less-serious misconduct was apparently left unchanged. Further, there is no way to disentangle whether the reduction in serious misconduct was a result of the content or practices of the LCP or some other factor, such as living among like-minded individuals in a separate housing unit. Will this effect upon committing serious infractions carry over after release when LCP participants are no longer part of a larger group? Only empirical analysis can answer this question. The conclusion of this study is modest: there is reason to believe that faith-based programs may cause positive changes in the lives of inmates, and it is important to keep examining this issue. Acknowledgements The opinions represented in this article are those of the authors and do not reect the ofcial positions or policies of the Federal Bureau of Prisons or the United States Department of Justice.

581 586 580 579 592

15.77% 17.65% 17.99% 16.47% 16.76%

17.14% 16.26% 15.77% 15.48% 15.08%

- 1.37% 1.39% 2.22% 0.99% 1.68%

0.660 0.660 0.483 0.750 0.581

1,590 1,590 1,590 1,590 1,590 1,590

NA NA NA NA NA 16.25%

NA NA NA NA NA 14.04%

5.00% 4.47% 4.72% 4.78% 4.72% 2.21%

0.044 0.067 0.055 0.049 0.051 0.263

Notes: NA = not available. The NNMATCH procedure does not provide base rates for the two groups.

the comparison groups. For these matches, slightly over 5 percent of the inmates in the LCP had an instance of serious misconduct, where for the comparison group, the number was closer to 11 percent. The NNMATCH procedure produced matched comparisons where the difference between the LCP participants and comparisons was less pronounced, although still signicant. For the comparison based upon all members of the respective groups, with no attempt to match, a statistically signicant difference was demonstrated. As with the comparisons based on matched groups, the LCP participants were less likely to have been found guilty of an infraction. It is important to note that the power of this test was less than for the matched groups, however, and the difference in the percentages of inmates involved in serious misconduct was attenuated by the heterogeneity of the groups when they were not matched. The results for less serious misconduct are presented in Table 4. The ndings here are more ambiguous. The comparisons based on data produced by the PSMATCH2 and Parsons' greedy algorithms suggest small differences in incidence for the LCP and comparison groups. None of the differences approaches statistical signicance. This result is also observed when there is no attempt to control for differences between the two groups. When the matches were based upon the NNMATCH algorithm, statistical signicance was achieved in two of the data sets created by multiple imputation, and the results approach signicance in the other three data sets. The safest conclusion, though, is that the groups did not differ. In part, this conclusion is based on lack of a difference for the other matching methods, but statistically the NNMATCH result is not signicant at p = .05 if the results for the ve imputations are averaged. Clearly, the NNMATCH procedure produces somewhat different results than the other two matching procedures, as was also noted for any misconduct and serious misconduct. Discussion Matching techniques were used in this study to produce comparable distributions on observed and theoretically important variables for the treatment and comparison groups. The three techniques were not entirely comparable, but it was not the purpose of this study to disentangle which technique was most appropriate for the given data situation. Instead, the purpose of the analysis at hand was to

S.D. Camp et al. / Journal of Criminal Justice 36 (2008) 389395

395

Note
1. We thank one of the anonymous reviewers for this insight.

References
Abadie, A., Drukker, D., Herr, J. L., & Imbens, G. W. (2004). Implementing matching estimators for average treatment effects in Stata. The Stata Journal, 4, 290311. Akers, R. (1977). Deviant behavior: A social learning approach. Belmont, CA: Wadsworth. Camp, S. D., Gaes, G. G., Langan, N. P., & Saylor, W. G. (2003). The inuence of prisons on inmate misconduct: A multilevel examination. Justice Quarterly, 20, 501533. Camp, S. D., Klein-Saffran, J., Kwon, O. K., Daggett, D., & Joseph, V. (2006). An exploration into participation in a faith-based prison program. Criminology and Public Policy, 5, 529550. Clear, T. R., Hardyman, P. L., Stout, B., Lucken, K., & Dammer, H. R. (2000). The value of religion in prison: An inmate perspective. Journal of Contemporary Criminal Justice, 16, 5374. Clear, T. R., & Myhre, M. (1995). A study of religion in prison. IARCA Journal on Community Corrections, 6, 2025. Clear, T. R., Stout, B., Dammer, H. R., Kelly, L., Hardyman, P. L., & Shapiro, C. (1992). Prisoners, prisons and religion: Final report. Newark, NJ: Rutgers University. Clear, T. R., & Sumter, M. T. (2002). Prisoners, prison, and religion: Religion and adjustment to prison. In T. B. O'Connor & N. J. Pallone (Eds.), Religion, the community, and the rehabilitation of criminal offenders (pp. 127159). New York: Haworth. Cullen, F. T., & Gendreau, P. (2000). Assessing correctional rehabilitation: Policy, practice, and prospects. In J. Horney, J. Martin, D. L. MacKenzie, R. D. Peterson, & D. Rosenbaum (Eds.), Policies, processes, and decisions of the criminal justice system (pp. 109175). Washington, DC: National Institute of Justice. Dix-Richardson, F., & Close, B. R. (2002). Intersections of race, religion, and inmate culture: The historical development of Islam in American corrections. In T. B. O'Connor & N. J. Pallone (Eds.), Religion, the community, and the rehabilitation of criminal offenders (pp. 87107). New York: Haworth Press. French, S. A., & Gendreau, P. (2006). Reducing prison misconducts: What works! Criminal Justice and Behavior, 33, 185218. Gendreau, P., Goggin, C., & Law, M. A. (1997). Predicting prison misconducts. Criminal Justice and Behavior, 24, 414431. Hagan, J., & McCarthy, B. (1998). Mean streets: Youth crime and homelessness. Cambridge, MA: Cambridge University Press. Hirschi, T. (1969). Causes of delinquency. New Brunswick, NJ: Transaction. Homant, R. J., & Witkowski, M. J. (2003). Prison deviance as a predictor of general deviance: Some correlational evidence from project gangmill. Journal of Gang Research, 10, 6575. Johnson, B. R. (1984). Hellre and corrections: A quantitative study of Florida prison inmates. Unpublished doctoral dissertation, Florida State University, Tallahassee. Johnson, B. R. (2002). Assessing the impact of religious programs and prison industry on recidivism: An exploratory study. Texas Journal of Corrections, 28, 711. Johnson, B. R. (2004). Religious programs and recidivism among former inmates in Prison Fellowship programs: A long-term follow-up study. Justice Quarterly, 21, 329354. Johnson, B. R., De Li, S., Larson, D. B., & McCullough, M. (2000). A systematic review of the religiosity and delinquency literature. Journal of Contemporary Criminal Justice, 16, 3252. Johnson, B. R., & Larson, D. B. (2003). The InnerChange Freedom Initiative: A preliminary evaluation of a faith-based prison program. Philadelphia: Center for Research on Religion and Urban Civil Society. Johnson, B. R., Larson, D. B., & Pitts, T. G. (1997). Religious programming, institutional adjustment and recidivism among former inmates in Prison Fellowship programs. Justice Quarterly, 14, 145166. Johnson, B. R., Tompkins, R. B., & Webb, D. (2002). Objective hope: Assessing the effectiveness of faith-based organizations: A review of the literature. Philadelphia: Center for Research on Religion and Urban Civil Society.

Kerley, K. R., Matthews, T. L., & Blanchard, T. C. (2005). Religiosity, religious participation, and negative prison behaviors. Journal for the Scientic Study of Religion, 44, 443457. Krohn, M. D., & Massey, J. L. (1980, Autumn). Social control and delinquent behavior: An examination of the elements of the social bond. The Sociological Quarterly, 21, 529543. Laub, J., & Sampson, R. (2003). Shared beginnings, divergent lives: Delinquent boys to age 70. Cambridge, MA: Harvard University Press. Leuven, E., & Sianesi, B. (2003). Psmatch2: Stata module to perform full Mahalanobis and propensity score matching, common support graphing, and covariate imbalance testing. Retrieved from Institute for Fiscal Studies Web site: http://ideas.repec.org/c/ boc/bocode/s432001.html Myers, M. (2000). Toward theoretical integration. Theoretical Criminology, 4, 222225. O'Connor, T. P., & Perreyclear, M. (2002). Prison religion in action and its inuence on offender rehabilitation. In T. P. O'Connor & N. J. Pallone (Eds.), Religion, the community, and the rehabilitation of criminal offenders (pp. 1133). New York: Haworth Press. Parsons, L.S. (2001, April). Reducing bias in a propensity score matched-pair sample using greedy matching techniques. Paper presented at the annual conference of SAS Users Group International, Long Beach, CA. Pisciotta, A. W. (1994). Benevolent repression: Social control and the American reformatory-prison movement. New York: New York University Press. Rosenbaum, P. R. (1995). Observational studies. New York: Springer-Verlag. Rosenbaum, P. R., & Rubin, D. B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70, 4155. Rosenbaum, P. R., & Rubin, D. B. (1985). Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. American Statistician, 39, 3338. Rothman, D. J. (1980). Conscience and convenience: The asylum and its alternatives in progressive America. Boston: Little Brown. Rothman, D. J. (1995). Perfecting the prison: United States, 1789-1865. In N. Morris & D. J. Rothman (Eds.), The Oxford history of the prison: The practice of punishment in Western society (pp. 11129). New York: Oxford University Press. Samenow, S. E. (1998). Straight talk about criminals: Understanding and treating antisocial individuals. Northvale, NJ: Jason Aronson. Samenow, S. E. (2004). Inside the criminal mind, (Rev. ed.). New York: Crown. Sampson, R., & Laub, J. (1993). Crime in the making. Cambridge, MA: Harvard University Press. Schafer, J. L. (1997). Analysis of incomplete multivariate data. New York: Chapman and Hall. Sumter, M.T. (1999). Religiousness and post-release community adjustment: Executive summary. Unpublished doctoral dissertation, Florida State University, Tallahassee. Sumter, M. T., & Clear, T. R. (2002). What works in religious programs (Monograph No. 2). Washington, DC: International Association of Community Corrections. Sundt, J. L., Dammer, H. R., & Cullen, F. T. (2002). The role of the prison chaplain in rehabilitation. In T. B. O'Connor & N. J. Pallone (Eds.), Religion, the community, and the rehabilitation of criminal offenders (pp. 5986). New York: Haworth Press. Sutherland, E. H., & Cressey, D. R. (1947). Principles of criminology (5th ed.). Chicago: J. B. Linnincott. Wainer, H. (Ed.). (1986). Drawing inferences from self-selected samples. New York: Springer-Verlag. Yochelson, S., & Samenow, S. E. (1976). The criminal personality. New York: Jason Aronson. Young, M. C., Gartner, J., O'Connor, T., Larson, D., & Wright, K. (1995). Long-term recidivism among federal inmates trained as volunteer prison ministers. Journal of Offender Rehabilitation, 22, 97118. Zamble, E., & Poporino, F. J. (1988). Coping, behavior, and adaptation in prison inmates. New York: Springer-Verlag.

Вам также может понравиться