Introduction

About two-thirds of state prisoners released are rearrested for a new crime within three years (Alper et al., 2018; Durose et al., 2014), and the 3-year rearrest rate has not changed much for several decades (Beck & Shipley, 1989; Langan & Levin, 2002). High recidivism among ex-prisoners seem to confirm the limited rehabilitative impact of incarceration. This is consistent with an assessment of prison chaplains in a national survey, where less than half (45.0%) of them reported a positive view on how their state correctional system was doing to prepare inmates for reintegration into the community (Boddie & Funk, 2012). Almost three-quarters (73.0%) said that access to religious programs in prison was “absolutely critical” to successful rehabilitation of inmates (see also Sundt & Cullen, 2002). This opinion is worthy of attention in a time of ever-tightening budgets, especially since volunteer-led and externally funded faith-based programs tend to be one of the few remaining resources available for promoting rehabilitation in prison.

Prior research shows that inmate involvement in religion or religiosity and participation in faith-based programs tend to be positively related to subjective well-being and inversely to prison misconduct (Clear et al., 2000; Dammer, 2002; Johnson, 2011; Kerley, Matthews, & Schulz, 2005; Kerley et al., 2011a; O’Connor & Perreyclear, 2002). These relationships—based mostly on cross-sectional data—imply the salutary effects of religion: cognitive, affective, and behavioral consequences of faith-based rehabilitation or reform that seek to change character traits, motivations, or disposition. These effects, however, are not the indicators or measures of rehabilitation itself, so the concept needs to be explicitly defined and observed separately from its outcomes. In this paper, we conceptualize rehabilitation as moral reform or “moral improvement” (Forsberg & Douglas, 2020) and operationalize the concept in terms of self-identity, existential belief, and character.

To empirically examine how religion contributes to reforming offenders, we conducted a quasi-experiment on a faith-based program, using a sample of male inmates housed at two correctional facilities in Texas: a state jail and a maximum-security prison. We hypothesized that program-increased religiosity contributes to inmate rehabilitation: that is, identity transformation, a new sense of meaning and purpose in life, and virtue development. We also hypothesized about key affective and behavioral outcomes of rehabilitation: religion-based reform reduces negative emotional states and the risk (i.e., probability) of interpersonal aggression among program graduates. To test these hypotheses, we applied manifest-variable structural equation modeling to analyze data from pretest and posttest surveys. Before describing our methodology and presenting findings, we begin with a conceptual discussion of the key concept, rehabilitation.

The Concept of Rehabilitation

Noting that offender rehabilitation (henceforth, rehabilitation) has not been adequately defined in the criminological as well as philosophical literature, McNeill (2012; 2014) offered a typology that consists of four forms of rehabilitation. Psychological or personal rehabilitation seeks to promote positive individual-level change in an offender by developing new skills or abilities and addressing deficits or problems, whereas legal or judicial rehabilitation concerns addressing the collateral consequences of the offender’s conviction by setting aside a criminal record and removing the stigma so they can requalify as a citizen (Maruna, 2011). Barriers to restoring social position as a citizen are moral as well as legal in that crime is not simply a legal but moral offense. Thus, the offender has to both improve moral capacities and seek redress for a wrongdoing as redemption is to be earned, which is the main concern of moral rehabilitation. The last form, social rehabilitation, is perhaps the most challenging to achieve because it involves not only “the restoration of the citizen’s formal social status and the availability of the personal and social means to do so” but also “the informal social recognition and acceptance of the reformed ex-offender” (McNeill, 2012:15; emphasis added).

More recently, Forsberg and Douglas (2020) developed an alternative taxonomy that distinguished five conceptions of rehabilitation based on the aims of rehabilitative measure and the means to achieve the intended end: rehabilitation as (1) anti-recidivism, (2) harm-reduction, (3) therapy, (4) moral improvement, and (5) restoration. The first two conceptions aim to reduce the likelihood of reoffending or engaging in conduct harmful to the well-being of others and an offender, using other means than reducing the offender’s capacity to reoffend or engage in such conduct (e.g., incapacitation), disincentivizing the offender’s reoffending or harmful conduct (e.g., deterrence), or incentivizing non-offending or less harmful conduct by the offender. The next conception intends “to cure or ameliorate a mental deficit … [whether] a mental illness or disorder, or … some defect in the capacities relevant for criminal responsibility” that caused an offender’s past offense and predisposes the offender to further offending.

The fourth and fifth conceptions correspond to McNeill’s last two forms of rehabilitation, though not exactly the same. McNeill’s “moral rehabilitation” includes an offender offering moral redress to the victim or the community, but, for Forsberg and Douglas, reparation is a part of rehabilitation as restoration that overlaps with McNeill’s “social rehabilitation.” On the other hand, their conception of rehabilitation as moral improvement focuses on making an offender morally better, while the nature and scope of moral improvement vary among the proponents of rehabilitation intended to have offenders become morally better. For example, Morris (1981:265) favors measures that help an offender develop an “identity as a morally autonomous person attached to … a moral good … that one feel contrite, that one feel the guilt that is appropriate to one’s wrongdoing, that one be repentant, that one be self-forgiving and that one have reinforced one’s conception of oneself as a responsible being.” Other scholars suggest that the scope should be narrower, like fortifying the moral capacities of offenders to reduce the likelihood of reoffending or targeting moral improvements relevant to crime that has been committed (Duff, 2001; Hampton, 1984; Howard, 2017).

In this paper, we conceptualize rehabilitation as moral reform—consistent with Forsberg and Douglas’ (2020) conception of rehabilitation as moral improvement—and suggest religion as a source of moral reform is well-positioned to have a wide-ranging rehabilitative impact on various domains of an offender’s life, including physical health and social relationships. We thus focus on three life domains: self-identity, existential belief, and character.

Religion and Rehabilitation

In our conceptualization of rehabilitation as moral reform, the term moral refers to “an orientation toward understandings about what is right and wrong, good and bad, worthy and unworthy, just and unjust, that are not established by our own actual desires, decisions, or preferences but instead believed to exist apart from them” (Smith, 2003:8). Since the matter of rightness, goodness, worthiness, and justice is determined by something outside of self, moral reform needs to be based on a system of self-transcendence rather than expediency or self-interest. Religion provides one such system and thus becomes a potential source of moral reform.

Self-Identity: Identity Transformation

Our conception of rehabilitation as moral reform assumes that offenders as humans are moral beings in that they are moral agents, one of whose “central and fundamental motivations for human action is to act out and sustain moral order” (Smith, 2003:8).Footnote 1 Of course, people do not always act morally or consistently live up to their own or others’ moral standards. Offenders are those who have failed to demonstrate reasonable firmness in response to criminogenic pressures they faced (Howard, 2017). This failure, especially when repeated, is likely to distort their understanding of who they are (i.e., moral beings) and lead them to adopt a criminal identity, while struggling to rationalize or make some sense of their own action by blaming others and society instead of owning moral responsibility. As a result, offenders may end up accepting that they are an automaton at the mercy of external forces rather than a morally autonomous person. Thus, rehabilitation as moral reform should aim at helping offenders discover their “true self” or “real me” (Maruna, 2001:88) in place of a criminal identity.

Religion offers an opportunity to replace an “old self” with a “new self” (James, 2007), helping offenders write a “redemption script,” a narrative that allows a new start built on the new self (Hallett & McCoy, 2015). Identity transformation via religion is a cognitive process that involves self-reflection and a change in self-concept, based on a new “living narrative” religion provides (Smith, 2003). It is also an affective process, which includes introspection and dealing with feelings of guilt over their wrongdoing and negative emotions (e.g., depression and anxiety) associated with criminal punishment (e.g., imprisonment) and the losses it caused (Clear et al., 2000). Identity transformation is the focus of identity theories of desistance from crime.

Giordano et al.’s (2002) symbolic interactionist theory posits that four types of “cognitive transformations” are necessary for desistance: (1) one’s openness to change (a general cognitive readiness for change), (2) one’s exposure to a particular hook (or set of hooks) for change, (3) one’s construction of a conventional “replacement self” or new identity, and (4) one’s perception of crime to be negative, unviable, or personally irrelevant. Identity transformation also involves “emotional transformations” that lead to “an increased ability to regulate their emotions in socially acceptable ways,” thereby reducing the likelihood to identify oneself with negative emotions (Giordano et al., 2007:1610). For Giordano et al. (2002), religion is a major hook for change among offenders, as it functions as a catalyst that provides a conventional replacement self and positive emotions (Giordano et al., 2008).

Paternoster and Bushway’s (2009) rational choice theory of desistance posits that offenders are committed to a criminal identity until they determine the cost of this commitment is greater than the benefit and perceive what they fear may become (“the feared self”) to be more likely than what they hope to become (“the positive possible self”). This perception is assisted by the “crystallization of discontent” (Baumeister, 1994), in which offenders see “failures or dissatisfactions across many aspects of [their] life [being] linked together and attributed to the criminal identity itself” (Paternoster & Bushway, 2009:1123). This cognitive process provides the initial motivation to change the self, and religion contributes to the process by helping offenders attribute their failures to their old self and offering a new self for a new start.

In a rare quantitative test of identity theories of desistance, using survey data from 2,249 inmates at America’s largest maximum-security prison, the Louisiana State Penitentiary (a.k.a., “Angola”), Jang et al. (2018b) found that religion played a role in contributing to identity transformation (see also Hallett et al., 2017). Specifically, they found religious conversion was positively related to cognitive transformation and crystallization of discontent, whereas inmate involvement in religion was positively related to emotional transformation. More recently, inmate participation in a faith-based program was found to increase religiosity, which in turn contributed to crystallization of discontent among prisoners in Colombia and South Africa (Jang et al., 2022a; Jang et al., 2022b; Johnson et al., 2021). In addition, in their qualitative study of 63 male inmates who had undergone a religious conversion, Kerley and Copes (2009) found that religion helped those inmates maintain their new identity through support networks (i.e., friendships with other religious individuals, whether inmates or local volunteers), formal and informal group activities (e.g., chapel services and Bible study or prayer meetings), “sharing” (whether evangelistic or altruistic), and personal reflection (e.g., “quiet time”).

Existential Belief: A Sense of Meaning and Purpose in Life

Offenders as moral beings have an orientation toward understanding what is worthy (Smith, 2003) or significant because humans are existential beings that have an innate need for a meaningful life, which largely derives from having purpose (a goal or goals) in life. A life of crime, particularly, common-law crime is hard to justify as meaningful given its destructiveness to others and the self, no matter how it is neutralized or rationalized (Sykes & Matza, 1957). As a result, offenders tend to lack a sense of meaning and purpose in life, which contributes to their criminal continuity. Thus, rehabilitation as moral reform should aim to help offenders how they can find a meaning and purpose in life for a change. According to Frankl (1984), the “true meaning of life” should be self-transcendent. While religion is a source of such meaning, self-transcendent meaning can also come from outside of religion, like close relationships with others (Costin & Vignoles, 2020) or a commitment to a cause, like environmental care or patriotism. In correctional institutions, however, religion is readily available to offer a time-honored system of meaning to offenders, helping them develop a new sense of meaning and purpose in life.

Prior research shows a positive association between religiosity and a sense of meaning and purpose in life among offenders. For example, in a study of male inmates at three maximum-security prisons in Texas, Jang et al. (2018a) found that inmate religiosity was positively related to perceived meaning in life (see also Jang et al., 2018b). Using data collected in a non-Western country, Jang et al. (2021) replicated the positive relationship (see also Jang et al., 2022a; Jang et al., 2022b). Specifically, analyzing data from a survey with male and female inmates housed in four South African prisons, they found that more religious inmates were more likely to report a sense of meaning and purpose in life than their less or non-religious peers. This positive relationship was found among both male and female inmates, showing that the relationship was gender neutral as well as cross-cultural.

Virtue Development

To the extent that crime is a result of limited moral capacities (Howard, 2017), rehabilitation as moral reform needs to aim at developing virtues among offenders. Since most religious traditions promote virtues like forgiveness, gratitude, accountability, and self-control (Emmons & McCullough, 2004; Evans, 2019; Rye et al., 2000), religious involvement is expected to increase virtues. First, religion not only emphasizes but also reveres virtues, teaching adherents to adopt and practice divine-like qualities (Rye et al., 2000). In Judaism, Christianity, and Islam, for example, forgiveness is a way to imitate God who forgives, carry out God’s plan beyond self-pity and resentment, and enhance one’s relationship with God. In Hinduism and Buddhism, forgiveness is a way to attain divinity or reach nirvana. Second, religion provides adherents with a spiritual or self-transcendent narrative, whereby virtue (e.g., self-sacrifice or forgiveness) has meaning even when it goes against human instincts (e.g., self-preservation) or counteracts a natural tendency (e.g., vengefulness). Finally, religious communities strive to stimulate virtue development as they collectively engage in practices (e.g., worship) that promote the connection between a transcendental narrative and virtuous behavior (Schnitker et al., 2019).

Prior research provides evidence that religion fosters virtues among individuals in the general population (Batson et al., 1999; Emmons & Paloutzian, 2003; Krause, 2018; McCullough et al., 2000; Rye et al., 2000). While research on religion and virtues among offenders is scant, Jang et al. (2018a) found that more religious inmates reported higher levels of forgiveness, compassion, and gratitude than their less or non-religious counterparts. Similarly, religiosity was found to be positively related to forgiveness, accountability, gratitude, and self-control among prisoners in Colombia and South Africa, both males and females (Jang et al., 2021; Jang et al., 2022a; Jang et al., 2022b).

Consequences of Rehabilitation

Faith-based rehabilitation as moral reform is likely to have affective and behavioral outcomes. First, identity transformation is expected to reduce negative emotions and deviant act among offenders, as it enables offenders to disassociate themselves from negative emotions that they used to identify with and to behave, consistent with the new self (Giordano et al., 2002; Giordano et al., 2007; Paternoster & Bushway, 2009). Second, a new sense of meaning and purpose in life is likely to decrease an offender’s negative emotions and misconduct as the new existential belief leads them to strive for conventional life goals and to manage their behaviors accordingly (Jang, 2016; McKnight & Kashdan, 2009; Steger & Frazier, 2005; Vanhooren et al., 2017). Finally, fostering virtues among offenders is expected to not only decrease deviance but also enhance emotional well-being (Emmons & McCullough, 2003; McCullough, 2000), since moral character is a central component of “eudaemonic” happiness (Ward & Maruna, 2007).

Research on rehabilitation as moral reform and its affective and behavioral consequences is limited, but three recent studies provide supportive evidence. First, Jang et al. (2018a) found crystallization of discontent and emotional transformation were inversely related to disciplinary convictions among prison inmates. They also found that inmates’ perceived presence of meaning in life and virtues (forgiveness, compassion, and gratitude) were inversely related to negative emotional states (depression and anxiety) and the likelihood of aggression toward another inmate. Second, the virtue of self-control was also inversely related to negative emotional states and the risk of aggressive misconduct (Jang et al., 2021; Jang et al., 2022b).

The Present Study: Hypotheses

To examine whether religion contributes to rehabilitation as moral reform, we conducted a quasi-experiment on a part of faith-based program, which is operating in seven units of Texas Department of Criminal Justice. The program, sponsored by a non-profit organization, called “Restoration Outreach of Dallas (ROD) Ministries,” consists of four in-prison Bible study classes (ROD I to IV) and aftercare following release from prison (ROD Ministries, 2015). The classes—each of which meets weekly for three months and cover 12 to 13 sessions—are facilitated not only by volunteers from local churches but also inmates who have both completed the classes and been trained to lead it. To complete each ROD class, inmates must attend at least nine sessions and are required to do homework. Following successful completion of a class and requirements, inmates participate in gradation and receive a certificate, becoming eligible for enrolling in the next class (e.g., ROD II after ROD I). The present study focuses on the first class (ROD I, henceforth, ROD program or, in short, ROD), comparing between inmates who completed the class (graduates) and those who did not complete (incompletes).Footnote 2

First, the ROD program was a Bible study and thus expected to increase inmate involvement in religion or religiosity, so we hypothesize the following.

• Hypothesis 1

• ROD graduates are more likely to report an increase in religiosity than the incompletes.

 Next, based on the literature reviewed above, we expect ROD-increased religiosity to contribute to rehabilitation. Since rehabilitation is a process of moral reform in terms of self-identity, existential belief, and character, it can be observed in terms of degree. Thus, inmates ahead of others in their progress in moral reform are likely to show signs of positive change compared to those who are making less progress. Thus, we hypothesize as follows:

• Hypothesis 2

• A change in religiosity is positively related to a change in (a) identity transformation, (b) a sense of meaning and purpose in life, and (c) virtues.

 Finally, we hypothesize about affective and behavioral consequences of religion-based rehabilitation as moral reform as follows:

• Hypothesis 3

• A change in identity transformation, a sense of meaning and purpose in life, and virtues are inversely related to a change in negative emotional states and the risk (i.e., probability) of interpersonal aggression.

Methods

Sample

We conducted a quasi-experimental study based on one-group pretest-posttest design to assess the effectiveness of ROD between July 2019 and March 2020. Two male-only units of the Texas Department of Criminal Justice (TDCJ) were selected for the study because the program had been operating there longer than other units. One was a state jail near the city of Dallas, and the other was a maximum-security prison, southeast of Dallas about 100 miles from the city.

Recruitment for the first ROD class was based on the facility-wide distribution of promotional flyers, which introduced inmates to the faith-based program and explained prerequisites for the class, including having at least six months left on their sentence and planning to reside in Dallas area upon release given that ROD aftercare is available only in that area. We visited the research sites to invite inmates screened and enrolled in the class to participate in our study. If they agreed, they signed a consent form and then completed a pretest survey.Footnote 3 After completing the class, graduates were asked to participate in a posttest survey, while those who failed to complete ROD were also asked to do the second survey. We obtained information about inmate’s sociodemographic and justice-related backgrounds from TDCJ.

A convenience sample of 231 inmates participated in the pretest survey, and 132 (57.1%) of them graduated with 99 not graduating.Footnote 4 Nearly half of them (109, 47.2%, 81 graduates and 28 incompletes) did the posttest.Footnote 5 Thus, the sample size for hypothesis testing was 109.

Measurement

The key exogenous variable, the ROD program completion, is dichotomous (0 = incomplete, 1 = graduate). Other exogenous variables were inmate’s backgrounds, including sociodemographic variables: age, race (dummy variables of Black, Hispanic, and Asian with the reference category of White), education (1 = 8th grade or less, 2 = 9th grade, 3 = 10th grade, 4 = 11th grade, 5 = 12th grade or GED, 6 = some college, 7 = college degree), intelligence (IQ score), marital status (dummy variables of being married including common-law marriage and divorced/separated/widowed [D/S/W] with the omitted category of being single), and religion (dummy variables of being Protestant, Catholic, Muslim, Jewish, an adherent of Eastern religion, and a follower of other religion with the omitted category of having no religion). Also included were justice-related backgrounds: a total number of prior incarcerations in prison (prior prison) and jail (prior jail) and current sentence length in year. In addition, a dummy variable, state jail, was created to control for any differences between the two research sites not only because one is a state jail and the other is a prison but also because ROD classes in the state jail were facilitated by local volunteers, whereas the program in the prison was led by inmates.

A first endogenous variable is inmate’s religious involvement or religiosity, which was measured by creating a scale summing standardized scores of five items (see Appendix A): two items of religious beliefs (perceived closeness to God and importance of religion) and three items of religious behaviors (frequency of religious service attendance, praying outside of religious services, and reading the Bible or other sacred text in private). Exploratory factor analysis generated a single-factor solution with moderate-to-high loadings, ranging from 0.490 to 0.779 at the pretest and from 0.545 to 0.759 at the posttest, and good inter-item reliability with Cronbach’s α being 0.827 and 0.780 at the pretest and posttest, respectively.

The next endogenous variables involve three domains of rehabilitation as moral reform: identity transformation, a sense of meaning and purpose in life, and virtues. First, identity transformation was operationalized by cognitive transformation, emotional transformation, and crystallization of discontent. The first and the last were measured by three items, whereas four items were used to measure the second (see Appendix A). Items of cognitive and emotional transformations loaded on a single factor with moderate-to-high factor loadings and acceptable-to-high internal reliability at both tests. On the other hand, the crystallization of discontent items had a poor internal reliability at both tests perhaps due to one item that had a low factor loading.

Second, to measure an inmate’s sense of meaning and purpose in life, we used Steger et al.’s (2006) four items of presence of meaning, which had a single-factor solution with high loadings and high-to-excellent internal reliability (see Appendix A).

Third, we created measures of seven virtues. To measure self-control, we used reverse-coded four items of Grasmick et al.’s (1993) Low Self-Control Scale, which had moderate-to-high loadings on a single factor and acceptable internal reliability at both pretest (from 0.472 to 0.709, α = 0.695) and posttest (from 0.498 to 0.651, α = 0.621). Compassion was measured by five items (Krause et al., 2016), which had a single-factor solution with moderate-to-high loadings and good inter-item reliability at both pretest (from 0.550 to 0.780, α = 0.795) and posttest (from 0.528 to 0.794, α = 0.777). Emmons et al.’s (2003) three items were used to measure gratitude based on the items that loaded on a single factor with moderate-to-high loadings and had acceptable-to-high internal reliability at both pretest (from 0.681 to 0.883, α = 0.835) and posttest (from 0.543 to 0.966, α = 0.692), whereas gratitude to God was measured by Krause’s (2006) two items that had an excellent inter-item reliability at both tests (α = 0.904 and 0.933).

The virtue of accountability (Evans, 2019) was measured separately for other people (human accountability) and God or a higher power (transcendent accountability), using 11- and 10-item scales, respectively (Witvlietet al., 2022a; Witvliet et al., 2022b). Items of human accountability loaded on a single factor with loadings higher than 0.600 with one exception (see Appendix A) and an excellent internal reliability at both tests (α = 0.890 and 0.923), whereas those of transcendent accountability had a single-factor solution with loadings higher than 0.800 with inter-item reliability being 0.964 and 0.961. To measure forgiveness of others, we used a single item asking inmates whether they had forgiven a person who recently hurt them. Regarding the person, we also measured their vengefulness, using five items of McCullough et al.’s (1998) Transgression-Related Interpersonal Motivations (TRIM) Inventory (see Appendix A), which loaded on a single factor with moderate-to-high loadings and high-to-excellent internal reliability at both pretest (from 0.594 to 0.842, α = 0.844) and posttest (from 0.775 to 0.921, α = 0.925).

Lastly, the ultimate endogenous variables, affective and behavioral outcomes of religion-based rehabilitation were measured in terms of two negative emotional states and behavioral intention. State depression was the average of six items from the CES-Depression Scale (Radloff, 1977), which loaded on a single factor with moderate-to-high loadings and had high internal reliability at both pretest (from 0.531 to 0.827, α = 0.849) and posttest (from 0.548 to 0.856, α = 0.862). Spitzer et al.’s (2006) 7-item generalized anxiety disorder scale (GAD-7) was used to measure state anxiety, and the items had a single-factor solution with moderate-to-high loadings and excellent inter-item reliability at both pretest (from 0.731 to 0.900, α = 0.929) and posttest (from 0.536 to 0.901, α = 0.909).

Behavioral intention was measured by an inmate’s self-reported probability of engaging in interpersonal aggression or, in short, intended aggression. To measure this construct, we used the vignette method, in which inmates were first asked to read the following scenario.

It’s Sunday afternoon. Mike is watching an NFL football game in the prison dayroom with other inmates. During a halftime break, Mike goes to the restroom. To reserve his seat, he asks a friend to “hold it down” for him. When Mike comes back, Joe is in his seat. Mike asks Joe to leave because it is his seat. Joe says he can sit anywhere he wants. Mike asks Joe to leave one more time. This time Joe ignores Mike. Meanwhile, everyone is watching what’s going on. Feeling not only dissed but also that he is right, Mike gets into an argument with Joe.

Then inmates were asked to indicate how likely it was that they would do the same as Mike (1 = not likely at all [0%], 2 = very unlikely, 3 = unlikely, 4 = likely, 5 = very likely, 6 = certainly [100%]).Footnote 6

Analytic Strategy

To test our hypotheses, we applied a manifest-variable structural equation modeling (SEM) approach to analyze data from the pretest and posttest. The modeling approach enabled us to not only simultaneously estimate for 16 endogenous variables (i.e., 12 mediating and three ultimate endogenous variables as well as religiosity), but also directly test the statistical significance of mediation, which path analysis would have not allowed us to. For model estimation, we employed Mplus 8 that incorporates Muthén’s (1983) “general structural equation model” and full information maximum likelihood (FIML) estimation. As concepts were measured by ordered categorical and continuous variables, we used the estimation option of MLR, which generates maximum likelihood estimates with standard errors that are robust to non-normality and non-independence of observations.

Next, to treat missing data, we used FIML, which tends to produce unbiased estimates similar to multiple imputation (Baraldi & Enders, 2010; Graham, 2009). Because of this missing data treatment method, the total number of observations Mplus used for model estimation was 321, who participated in the pretest survey, although 109 was the number of inmates who also participated in posttest survey. While SEM is a “large sample” method, either number indicated that our sample size was appropriate given that 100 to 150 is usually considered a minimum sample size for conducting SEM (Anderson & Gerbing, 1988; Ding et al., 1995; Tinsley & Tinsley, 1987). Finally, statistical significance (α = 0.05) was generally assessed using two-tailed tests, but we also applied one-tailed tests for the hypothesized relationships since their directions were a priori predicted.

Results

Table 1 presents descriptive statistics of variables measured at the pretest. The total sample (n = 109) included more ROD graduates (81, 74.3%) than incompletes (28, 25.7%). They were, on average, about 44 (44.06) years of age, with the youngest and oldest being 21 and 65, respectively (not shown in the table), and their racial backgrounds were White (37.6%), Black (40.4%), Hispanic (21.1%), and Asian (0.9%). The average education (5.69) fell between “12th grade or GED” and “some college,” and the inmates had, on average, a score (92.60) close to the lower end of “average intelligence” range (90–109) according to Wechsler Adult Intelligence Scale (Wechsler, 2008). While most (45.4%) of them were single, about a quarter (26.0%) were married or in common law marriage, whereas the remainder (28.6%) had post-marital status. Nine out of ten (89.6%) had Christian (77.3%) or other religion (3.8% Islam, 2.8% Judaism, 0.9% Eastern religion, and 4.7% “other religion”) with 10.4% reporting no religion.

Table 1 Descriptive Statistics of Variables Measured at the Pretest
Full size table

In addition, 55% of the sample were housed at the state jail at the time of pretest, and the study participants had been in prison, on average, about twice (1.93) prior to the current incarceration. Their average length of sentence was 24 (24.09) years. Results from t-test showed that graduates were serving shorter sentence than incompletes (19.98 vs. 36.00) because they were more likely to be inmates housed at the state jail (61.7%) than maximum-security prison (38.3%, not shown in the table) as we found earlier in the sample of pretest participants (n = 321; see footnote 4). However, these differences were found to be not significant, using the Bonferroni correction (α = 0.00185185…). Thus, the two groups were statistically equivalent at the pretest, that is, before they participated in the program.

Table 2 shows our model estimated for hypothesis testing (standardized coefficients are presented).Footnote 7 We found completion of ROD was positively related to religiosity at the posttest or Time 2 (0.206). Since religiosity’s Time 1 or previous (pretest) measure (religiosity T1) was controlled for, the positive relationship can be interpreted in terms of change: that is, the program completion increased inmate involvement in religion between the pretest and posttest. In other words, graduates were more likely to report an increase in religiosity than incompletes. Thus, Hypothesis 1 received empirical support. Next, the increased religiosity (religiosity T2) was positively related to a change in all three indicators of identity transformation—cognitive (0.275) and emotional transformations (0.479) and crystallization of discontent (0.349), perceived presence of meaning (0.554), and all seven virtues: self-control (0.502), compassion (0.335), gratitude (0.497), gratitude to God (0.481), human accountability (0.295), transcendent accountability (0.368), and forgiveness (0.382). In addition, the ROD-associated increase in religiosity was inversely related to vengefulness (‒0.495). That is, as hypothesized (Hypothesis 2), we found that the program had rehabilitative effects on inmates by increasing their religiosity, which in turn contributed to identity transformation, a sense of meaning and purpose in life, and virtue development.

Table 2 Estimated Structural Equation Model of Program Completion, Indicators of Rehabilitation, Negative Emotional States, and the Risk of Interpersonal Aggression (n = 109)
Full size table

The last three columns show that the religion-based rehabilitation significantly reduced negative emotional states and the probability of interpersonal aggression. Specifically, state depression was decreased by emotional transformation (‒0.233), presence of meaning (‒0.207), and self-control (‒0.183), and state anxiety was reduced by emotional transformation (‒0.369) and self-control (‒0.177). Next, the likelihood of engaging in aggression toward another inmate was lowered by an increase in emotional transformation (‒0.339) and the virtues of compassion (‒0.136), gratitude (‒0.110), gratitude to God (‒0.304), and human accountability (‒0.368).Footnote 8 In sum, Hypothesis 3 received partial support. It is worth noting that the program completion was inversely related to the probability of aggression (‒0.246), which indicates that ROD had rehabilitative effects that remained to be explained by other than what we included in the model.

A supplemental analysis was conducted to test the significance of indirect effects of the program completion and the program-increased religiosity on the secondary mediating and/or ultimate endogenous variables. We found that ROD significantly contributed to identity transformation, a sense of meaning and purpose in life, and the development of all seven virtues, while reducing vengefulness toward a person who caused pain in the past, by increasing inmate involvement in religion (see the first panel of Supplemental Table 4). Next, the increased religiosity significantly decreased negative emotional states and the risk of interpersonal aggression: specifically, state depression via religiosity-increased emotional transformation and perceived presence of meaning, state anxiety via emotional transformation and self-control, and intended aggression via emotional transformation, gratitude, gratitude to God, and human accountability (see the second panel). Taken together, the program was found to significantly decrease the negative emotional states and the risk of aggression by increasing religiosity, which in turn contributed to rehabilitation as moral reform: emotional transformation, presence of meaning, gratitude to God, and human accountability (see the bottom panel).

Another supplemental analysis was conducted to examine potential selection bias. Since ROD is a faith-based program, more religious inmates might have been drawn to and benefitted by the program than those less or not religious. To explore this issue, we used the medium of religiosity T1 (0.335) to divide the pretest sample (n = 229; two missing cases on religiosity T1) into low (n = 114) and high religiosity groups (n = 115), which included 53 and 55 inmates of the total sample (n = 108; one missing case), respectively. First, we found that the two groups had the same number of graduates (40 each) and did not significantly differ in graduation rate, while the low religiosity group’s rate (75.5%) was slightly higher than the high religiosity group’s (72.7%). Next, results from multi-group analysis revealed that the program completion increased religiosity among inmates who were not very religious at the pretest but had no significant effect among relatively religious, and religiosity was more likely to have rehabilitative effects on self-identity and character in the low than high religiosity group (see Supplemental Table 5). In sum, the faith-based program tended to contribute to rehabilitation by increasing religiosity among inmates who were not religious before the program compared to those who were already religious.

Discussion

Both rehabilitation and religion have long been linked to the original purpose of American penal system (Cullen et al., 2014). In 1790, Quakers pressured the Pennsylvania legislature to call for a renovation of local county jails, which eventually resulted in the creation of a separate wing of Philadelphia’s Walnut Street Jail to house felons in solitary cells, called “the penitentiary house.” This was a forerunner of the Pennsylvania state prisons—the Western and Eastern Penitentiaries, built in the early 19th century, and the penitentiary was a place for penance as inmates were meant to reflect on their wrongdoings and seek reform. Since the Pennsylvania system ended by the 1870s, however, prison reform efforts have not included religion because of the secularization of American society and the development of scientific disciplines concerned with human behavior, such as psychiatry, psychology, and sociology.

Nevertheless, religion remains an invaluable resource for American corrections, as religiously motivated volunteers continue to provide prisoners with non-religious (e.g., adult basic education, anger management, and entrepreneurship) as well as religious programs when prison administrators find it increasingly difficult to fund educational, vocational, and rehabilitative programs due to constricting budgets. Furthermore, correctional research empirically demonstrates the benefits of faith-based programs (Johnson, 2011), and an emerging body of evidence shows that inmate involvement in religion is related positively with emotional well-being and inversely with prison misconduct (Clear & Sumter, 2002; Jang et al., 2021; Kerley et al., 2011b; Kerley, Matthews, & Blanchard, 2005).

Despite the increasing evidence of rehabilitative effects of religion, prior research has been limited to examining the consequences of religion-based rehabilitation (e.g., a reduction in negative emotions and misconduct) rather than the rehabilitation per se, that is, how religion rehabilitates prison inmates and what changes happen to them. To address this gap in research, we conceptualized rehabilitation as moral reform and operationalized it in terms of positive changes in self-identity, existential belief, and character. To test hypotheses about religion-based rehabilitation and its consequences, we analyzed data from pretest and posttest surveys with male inmates who participated in a faith-based program operating in two correctional facilities in Texas. Results generally supported these hypotheses.

First, as expected, program graduates tended to report an increased involvement in religion between the two tests compared to inmates who participated but did not complete the program. Next, consistent with a second hypothesis, the program-increased religiosity was found to contribute to rehabilitation, enhancing identity transformation (cognitive and emotional transformations and a motivation for self-change à la crystallization of discontent), a sense of meaning and purpose in life, and virtues (self-control, compassion, gratitude, gratitude to God, human accountability, transcendent accountability, and forgiveness), while decreasing the vice of vengefulness. Finally, some indicators of faith-based rehabilitation were found to reduce state depression and anxiety and the probability of aggression toward another inmate, which provided partial support for the last hypothesis. A supplemental analysis revealed that the program’s indirect effects on rehabilitation via religiosity and on the emotional states and the risk of aggression via religiosity and its associated rehabilitation were statistically significant.

The notion that offenders need rehabilitation presumes that there is something wrong with them or they lack something, which led them to commit an offense. Consequently, rehabilitative measures aim to address the issue one way or the other. For example, McNeill’s (2012; 2014) “psychological or personal rehabilitation” and Forsberg and Douglas’ (2020) “rehabilitation as therapy” are concerned with fixing the problem (e.g., a mental illness or deficit) and having offenders develop new skills or abilities through job training or education. Our conception of rehabilitation as moral reform overlaps with their “moral rehabilitation” or “rehabilitation as moral improvement,” which is intended to morally improve offender who did not have “moral power” (Rawls, 2005). Specifically, it intends to address the two components of moral failure, epistemic (mistaken conclusions about whether certain conduct is permissible or wrongful) and motivational (non-compliance with moral duties).

Religion is a viable option to morally reform offenders because it provides them with philosophical reasons for not reoffending as well as teaching justice-related moral duties. For example, Christianity— from which the ROD program is based—teaches inmates that they are redeemable by the grace of God, and the redemption comes with a new identity (e.g., a child of God) that enables them to start a life anew. Inmates are also told that God has a specific plan for their lives, and that God’s purpose for their life will provide meaning and replace the desire for reoffending. In addition, the religion contributes to developing virtues among inmates because the religion teaches inmates to imitate God’s character, being compassionate and forgiving toward those who hurt them rather than taking revenge upon them, but the purpose-driven life is also likely to motivate inmates to practice self-control and be grateful for the gift of second chance and willing to be held accountable for their life by others as well as God. The present findings provide preliminary evidence for the rehabilitative effects of religion on prison inmates.

Our conception of rehabilitation as moral reform is based on two key premises: one concerns crime, and the other human nature. First, crime is not simply a legal but moral offense in that it is a violation of collective morality as well as a criminal law. However, the relationship between morality and criminal law is anything but straightforward because, while no behavior can become criminal without any moral basis or justification, behaviors deemed immoral cannot be criminalized without a certain level of moral consensus (Meier et al., 2006). Depending on the degree of agreement about immorality or seriousness of an act, crimes are distinguished between mala in se and mala prohibita or “consensus crimes” (e.g., murder or burglary) and “conflict crimes” (e.g., drug use or prostitution) (Hagan, 1985). Since the immorality of conflict crimes is contested (e.g., a nationwide debate over the legalization of marijuana), we acknowledge that our premise about crime being a “moral” offense is not politically neutral when it comes to conflict crimes. A relevant question here is whether rehabilitation as moral reform is more applicable to offenders incarcerated for consensus than conflict crimes given that the former offenders violated moral codes largely accepted in society, whereas the latter committed an act whose immorality is questioned by many (including the offenders themselves), thereby seeing no need to morally improve themselves. Although this is an empirical question for future research, we expect our concept of rehabilitation as moral reform to benefit both types of offenders because it focuses on positive changes in self-identity, existential belief, and character rather than targeting moral improvements relevant to reducing the likelihood of repeating crime that has been committed (Duff, 2001; Hampton, 1984; Howard, 2017), unlike “rehabilitation as anti-recidivism” (Forsberg & Douglas, 2020).Footnote 9

Next, we assumed that offenders as humans are morally autonomous beings, although they made a morally wrong choice by yielding to criminogenic pressures: thus, “contemptuous” punishment that fails to respect offenders as “moral persons,” who are capable of self-reform, undermines the prospect of their rehabilitation (Hoskins, 2013). Based on the same assumption, decrying the contemporary amoral penology, Cullen et al., (2014:74) proposed “the virtuous prison” to restore the moral purpose of American corrections—restorative rehabilitation—by using “offenders’ time of incarceration to cultivate moral awareness and the capacity to act virtuously.” For example, productive activities with a moral purpose that provide opportunities to be virtuous (e.g., using inmate wages to compensate victims and making toys for poor children) would help inmates redefine who they are, believe in meaning and purpose in life, and build character. Although a virtuous prison does not require religion, Cullen et al. illustrated the prospect of creating one with a faith-based prison (Johnson, 2014; see also Johnson et al., 2021). We agree that a virtuous prison can be based on a secular entity but cannot think of any better system ready to fill that space than religion.

While this study provides empirical evidence of how religion is likely to contribute to rehabilitation, it is necessary to acknowledge key limitations. First, we had no control group as our research was prematurely ended by the COVID-19 lockdown. Thus, we could not examine the rehabilitative effect of the ROD class based on observed differences between inmates who participated in the program and those who did not. Instead, we compared inmates who graduated the class with those who did not complete it. However, given that the graduates and incompletes were likely to have been similar in their motivation to participate in the program since they all voluntarily applied and that they were statistically equivalent at the pretest, the present findings provide at least preliminary evidence of the rehabilitative effect of religion.

A second limitation is nontrivial attrition: that is, about a half of pretest participants were not available for the posttest. While it was not surprising that posttest participants tended to be the program graduates and higher on self-control at the pretest compared to the non-participants, ROD’s impact on rehabilitation might have been overestimated to the extent that the former were more motivated to change themselves than the latter. Third, while we explored the possibility of rehabilitation as moral reform being more applicable to offenders who committed consensus than conflict crimes, we could not formally test whether the rehabilitative effect varies among offenders who committed different types of crimes because of our small sample size, which is a worthy topic for future research. Finally, while studying gender differences in the rehabilitative effect of religion is an important topic given that women tend to be more religious than men (Sherkat & Ellison, 1999), we could study only male inmates because the ROD program has not been extended yet to female facilities.

Despite these limitations, we believe our study contributes to the literature on offender rehabilitation by providing empirical evidence, though preliminary, of the rehabilitative effects of religion on inmate’s self-identity, existential belief, and character. The present study suggests that it would be prudent for correctional policy makers and prison administrators to be open to religious programs like the ROD class to not only protect an inmate’s constitutional right to practice religion but also help them achieve reform before returning to society.