Religious Involvement, Moral Community and Social Ecology: New Considerations in the Study of Religion and Reentry

Abstract

Objectives

To examine the link between an individual’s religious involvement in prison and recidivism and assess how macro-level conditions in the counties to which individuals return shape this relationship.

Methods

Using data from 1362 previously incarcerated people, a series of hierarchical generalized linear models are used to examine the extent to which an individual’s religious involvement in prison relates to recidivism post-release. We also examine how county-level religious adherence, economic disadvantage, and potential social service assistance directly affect recidivism, and how each shape the relationship between religious involvement and recidivism.

Results

Findings show that county-level religious adherence was directly associated with lower recidivism, but individual-level religious involvement was not when assessing recidivism over longer periods of time post-release. Cross-level interactions revealed that county-level resource deprivation conditions the effect of individual religious involvement.

Conclusions

Findings have theoretical implications for the study of religion and reentry. Methodologically, failing to account for the religious context of counties, in addition to micro–macro linkages, harms individual level research on religion and reentry.

Introduction

Religion is traditionally thought of as a deeply personal experience, the salience of which may rise or decline throughout various moments and events in a person’s life. Several factors including adverse experiences and traumatic life events are associated with changes in religiosity across the life course (Ingersoll-Dayton et al. 2002; Uecker et al. 2007). Often this change can be positive, with studies finding that the experience of spiritual or religious growth after an adverse experience can provide a framework for people to help find meaning in their life and incorporate positive life changes (de Castella and Simmonds 2013). This individual narrative is often applied to persons entering prison as they look for guidance and support through issues of isolation, deprivation, and feelings of guilt and shame. Narratives of change derived from studies in criminology often include stories of religious conversion, identity change, and “quantum change” in spiritual virtues such as empathy, altruism and hope for the future (Hallett et al. 2017; Kerley et al. 2011; Rigsby 2018).

A rich sociological tradition reminds us, however, that religion also operates at a group-or community-level (Durkheim [1897] 1951; Stark and Bainbridge 1996), whereby the values and beliefs of a community can help to sustain and support the growth of individual values and beliefs. Reflective of a normative climate, this notion of “moral community” has been considered by some a greater source of control than individual beliefs and values (Stark and Bainbridge 1996; Stark et al. 1980), one that exerts influence through community interaction (Harris and Ulmer 2017). While many studies have articulated the individual experience of religion in prison, most have ignored the broader set of ecological realities to which persons return after prison, including the normative climate. As an example, a growing body of literature has sought to quantify the relationship between “religious involvement”, used here to capture individual attendance and participation in faith-based services (Aranda 2008; Giordano et al. 2008), and recidivism (Dominey and Lowson 2017; Duwe and King 2013; Giordano et al. 2008; Johnson 2004, 2011; Johnson et al. 1997; Mowen et al. 2017; O’ Connor and Bogue 2010; Stansfield et al. 2017; Willison et al. 2010; Young et al. 1995).

These studies posit a negative relationship between religious involvement and recidivism, guided by reentry scholarship that highlights the critical importance of social support for helping previously incarcerated individuals cope with the many challenges they face after release (Bales and Mears 2008; Berg and Huebner 2011; Boman and Mowen 2017; Cochran 2014; Duwe and Johnson 2016; Duwe and King 2013; Visher 2005). These expectations are not always confirmed in quantitative studies assessing the relationship between religious involvement and criminal reoffending over the long-term, however (Johnson 2004; Johnson et al. 1997; Giordano et al. 2008; Stansfield et al. 2017; Willison et al. 2010). The difficulty in drawing firm conclusions about the role of religion point to some additional challenges after release from prison that are not being captured; or, that there are other processes—dynamics occurring at the macro-level—that influence the link between individual religious involvement (a micro-level process) and recidivism that research has yet to unpack.

While there are a significant number of individual factors that likely link religious involvement to desistence (e.g., Giordano et al. 2008; Mowen et al. 2017 Stansfield et al. 2017), few—if any—studies have sought to consider the importance of community-level measures of religion and related social support (e.g., macro-level ecological conditions) on individual recidivism (e.g., a micro level outcome) following release from prison. And when macro-level religious contexts are considered, individual religious involvement is rarely considered simultaneously (Harris and Ulmer 2017; Longest and Uecker 2017; Ulmer et al. 2008). This failure to consider ecological characteristics of ones’ broader environment is a particularly important oversight given that the long-term processes of religious involvement and personal change occur within the context of faith communities and ecological conditions after a person is released from prison. Highlighting broader limitations to criminological research in the 2016 Sutherland Address, Matsueda (2017, p. 513) finds that, “ignoring social interaction effects and complex micro–macro linkages when they are in fact present will have negative consequences for individual-level research”. While Matsueda’s (2017) comments were also aimed at larger issues of causality within the micro–macro link in criminology, his timely critique offers an important framework for researchers examining the link between religious involvement/support and recidivism. Clearly, the structural context to which individuals return plays an important role in the reentry process. Moreover, given the iterative and participatory dynamics of long-term religious involvement occurring both within and outside of prison, more work is needed to examine how broader macro-level ecological conditions interact with religious involvement to shape longer-term behavior of the individual.

In this article, continuing the emphasis on the potential support offered by religion and spirituality, we consider whether the macro-level social context of moral community, economic resources, and potential social service support condition the effect of individual religious involvement on recidivism for adult males undergoing reentry from prison. We present results of multilevel generalized linear mixed-effects models that incorporate data from a variety of sources to examine recidivism among a sample of men admitted to the Oregon Department of Corrections (ODOC) in 2004. Results add to prior work on individual-level religion–recidivism linkages by incorporating data from the counties to which persons were released to examine the role of macro-level characteristics within this process. Findings have implications for the growing interest in humanist, spiritual and religious programs in prison, as well as faith-based reentry initiatives.

Background

In a recent review of the religion–crime relationship, Adamczyk et al. (2017) noted that social control theories are commonly invoked to posit a negative relationship between religious involvement and criminal behavior. Using this intuitive framework, scholars have posited that the development of bonds to a religious institution, attachment to a prosocial network of peers, as well as more structured conventional activities, will deter a person from entering criminal activity. Scholars have also suggested that religion provides a critical entrée to prosocial connections and social support networks, which renders individuals less likely to recidivate (Giordano et al. 2008; O’Connor et al. 2006; Schroeder and Frana 2009; Stansfield et al. 2017). Prosocial connections and support can of course be attained in the absence of religion, with family often identified as a critical support resource after prison (Taylor 2016). Furthermore, when participation in religious programs is combined with a high level of peer interaction, religious involvement can have negative behavioral consequences (Mowen et al. 2018). But for many people, “religion can form the basis for an institutional support network through which recently released prisoners can build and repair relationships, find jobs, and establish social support” (Stansfield et al. 2017, p. 525). And for the most isolated persons, on-going connections with faith-informed mentors and volunteers may provide a critical level of support and accountability otherwise missing during the transition back to the community (Dominey and Lowson 2017; Duwe and King 2013; Johnson 2011; Tomczak and Thompson 2017), in addition to the social learning of prosocial norms and values (Lee and Ousey 2005).

Although easily viewed from a control perspective, the pro-social effects of faith-based services have more recently been considered a catalyst for positive identity transformation that allows a person to develop hope for the future and provide resources and strategies for coping with stress (Agnew 2001). For example, through interviews of inmate graduates of Louisiana State Penitentiary’s seminary program, Hallett et al. (2017) uncovered how religious behavior and spiritual learning helped individuals reshape their sense of purpose and belonging to society (see also, Hallett and McCoy 2015; Maruna et al. 2006). Seminary participation ultimately helped graduates of the program, as well as members of inmate-led churches, avoid prison misconduct (Hallet et al. 2017). Using social control and identity theories as frameworks, research continues to demonstrate that spiritual and religious programming can empower people in prison towards positive identity change and a reduction in reoffending (Hallett et al. 2017; Lee et al. 2016; Pagano et al. 2015), and that a continuum of mentoring and social support may maximize recidivism benefits (Duwe and Johnson 2016; Duwe and King 2013). To date, however, each study that we are aware of has focused on individual program participation and religious behavior, without consideration of other resources and avenues of support available, and the moral community within which people reenter.

Moral Communities

In proposing a theory of moral communities, Stark et al. (1980) argued that religion facilitates conformist behavior through interaction with others in the larger context, whereby prevailing attitudes and beliefs of religious groups, regions, and even nations, sustain and support the effect of personal religious beliefs on behaviors. Several studies have posited a link between “moral communities” and adverse behavioral and health outcomes including delinquency and drug use (Adamczyk 2012; Adamczyk and Palmer 2008; Harris and Ulmer 2017; Regnerus 2003; Ulmer and Harris 2013) and homicide (Lee and Bartowski 2004). Fewer still have linked “moral communities” with criminal justice outcomes, including support for the death penalty (Unnever and Cullen 2006) and sentencing decisions (Ulmer et al. 2008). Nevertheless, most considerations of the contextual effects of moral communities are from the sociology of religion field, with very few macro-level criminology studies integrating the moral communities’ thesis (Admacyzk et al. 2017). Furthermore, prior criminal justice studies of moral community were unable to measure both the (macro-level) moral community and the (micro-level) individual person religiosity at the same time.

Criminological studies of the more specific relationship between religiosity and recidivism have also underappreciated the wider contextual influences of religiosity. This limitation is due, in part, to the difficulty of obtaining micro-level data on religious and spiritual involvement of individuals in prison combined with concurrent difficultly in collecting macro-level measures of religiosity and spirituality in the communities to which individuals return. Nevertheless, in line with Stark and Bainbridge’s (1996) emphasis on exploring the interaction between personal religiosity and religiosity in the larger context, it is possible that personal religiosity is less salient as a source of support than religiosity and support available in the social context of local counties. In a recent study, Mowen et al. (2017) utilized cross-lagged dynamic panel models to examine the joint influences of between-individual differences in “religious support” measured in prison, and within-person change in religious support after release on criminal recidivism. In finding a significant interaction of religious support in prison (a “between-person” effect) combined with post-release changes in support (a “within-person” effect), the authors revealed the importance of continued and strengthening support from religious services after release. While noting the importance of the personal change individuals can go through in prison through participation in spiritual or religious services, pro-social support, structural bonds and opportunities are also necessary to complete the integration of individuals back into society. Despite an increase in the number of studies assessing change in individual religious or spiritual involvement after incarceration, no study that we are aware of has considered how the social ecological conditions, including the availability of community social support, moderate the effect of religious/spiritual involvement and recidivism.

Social Ecology and Social Service Support

There has been a long tradition in criminology of examining the relationship among social ecology, individual well-being, and crime, with ample studies in health and criminal justice literatures also considering the mediating, or moderating, effects of social support resource availability (Hipp et al. 2010; Wenzel et al. 2002; Xie et al. 2012). There has also been a slow rise in the number of reentry and recidivism studies incorporating both individual characteristics and ecological conditions including studies in Oregon (Kubrin and Stewart 2006), Florida (Mears et al. 2008) and Iowa (Tillyer and Vose 2011). Studies in Oregon and Florida found that the risk of recidivism was higher for persons living in disadvantaged neighborhoods or counties after release. These studies emphasized that people are embedded within social structures that determine the availability of jobs, drug and alcohol treatment, and the provision of resources and larger support structures. Social service providers can form a network of institutions that can assist people with a past incarceration (Xie, Lauritsen, and Heimer 2012). Some studies have also sought to measure the availability or proximity to social and health services directly (Hipp et al. 2008, 2009, 2010). As an example, Hipp et al. (2010) found that although racial and ethnic minority parolees in California had more social services nearby, services in low income areas faced much greater demand. Higher potential demand for services was associated with an increased risk for recidivism (Hipp et al. 2010).

Although studies examining the importance of spirituality and religion in the lives of previously incarcerated individuals have rarely considered social ecology, some recent studies provide interesting guidance. Given that the black Protestant church has long been considered a mechanism through which disadvantaged communities collectively combat social problems, Harris and Ulmer (2017) examined whether black Protestant adherence was negatively associated with homicide, robbery, burglary and larceny. The authors posited and found a protective effect of adherence, particularly in more disadvantaged locales, believed to be due to the black church’s role in promoting social organization and informal social control in the absence of other alternative institutions (see Lee and Thomas 2010). Using data from the Pathways to Desistance dataset, Stansfield (2017) also discussed the historical value of the black church in examining recidivism among a sample of serious adolescent offenders. Although black participants were more likely to report that religion was important in their lives, religiosity was only significantly associated with criminal desistance among white participants. The author posited two explanations. First, and in line with the work of Hipp et al. (2010), the author suggested that the greater demand for services in low-income areas may dilute these resources more so than in suburban or rural areas. Greater demand may make religious volunteers and workers less effective due to dilution. Alternatively, Stansfield (2017) suggested that individual religiosity may become fraught with difficulty and struggle in the face of the wider social problems and economic strains found in some areas. Although rarely considered by criminologists, the notion of “spiritual struggle” or “negative religious coping”, defined as spiritual discontent and a reappraisal of God, has often been considered in the religion literature as a possible response in times of major life stress and change, such as reentry from prison (Exline and Rose 2005; Pargament et al. 1998; Wortmann et al. 2011).

With these findings and ideas in mind, we suggest at least two pathways through which social ecology may condition the effect of religion on recidivism. Firstly, strain theories (Agnew 2001) posit that living in an area with limited economic opportunities prevents a person from the institutional engagement necessary to achieve conventional goals. Furthermore, social disorganization theories note that access to other structural mechanisms of social support is not equally distributed geographically, with both socially-disorganized and rural areas facing difficulties. Resource deprivation and collective efficacy (Sampson et al. 1997) may also affect the extent to which religious and spiritual viewpoints are reinforced throughout an area. Given the importance of the economic context of reentry and access to social service providers for reentry success (Hipp et al. 2010; Kubrin and Stewart 2006), there is reason to believe that religion may take on a different role for some people post-release.

On the one-hand, ecological stressors may lead to a reduction in the reliance of religion and spirituality, as frustrated individuals move from problem-solving in cooperation with God, to a greater self-reliance. Alternatively, in the absence of limited opportunities, religion may take on an even greater salience (e.g. Harris and Ulmer 2017), given that connections via religion and spirituality can provide tangible aftercare and assistance to ex-offenders (O’Connor et al. 2006), connect individuals to employment opportunities (Stansfield et al. 2017), in addition to providing a framework for coping with strain (Schroeder and Frana 2009). Many faith-based community reentry initiatives provide secular programs consisting of evidence-based treatments and work cooperatively with many groups to connect individuals to employment, housing, and other important social resources in the community (Duwe and Clark 2013; O’Connor and Duncan 2011).

As discussed, participation in humanist spiritual, or religious programs in prison may instill a sense of future hope and self-worth in a person, increasing the likelihood that a person after release can cope with stresses and strains without the use of crime, drugs or alcohol, while also facilitating an identity transformation (Jang et al. 2017). Yet wide variation exists in the ecological context to which prisoners return home. Scholars of recidivism and reentry have increasingly called on researchers to examine interaction effects between individual characteristics and features of the ecological context (see also Matsueda 2017). As numerous ecological and larger support structures may interact with individuals coping through religion and spirituality, we examine independent and interdependent influences of social ecological conditions, including the availability of social service support and economic disadvantage, and religious involvement on recidivism among returning men and women.

Current Study

Collectively, our discussion highlights the importance of further investigating the effects of individual religious and spiritual involvement in prison on recidivism, with concurrent attention afforded to ecological conditions including the religious composition of the community, availability of social support, and presence of economic disadvantage. Towards this goal, we examine four related hypotheses which extend existing research into the role of religion and spirituality for reentry. Supporting recent work on the importance of humanist, spiritual, or religious (HSR) programming in prison (e.g., Hallett et al. 2017; Mowen et al. 2017; O’Connor and Duncan 2011), we expect that (Hypothesis 1) incarcerated individuals with higher levels of participation in humanist, spiritual, or religious programs will have a lower likelihood for recidivism, net of assessed individual risk for recidivism and demographics. Moving this research forward to consider the religious context in which individuals return and based within the moral community thesis, we expect (Hypothesis 2a) that the proportion of religious adherents in the county to which a person returns from prison will also have a direct protective effect on recidivism. Related, we expect that areas with more adherents will strengthen individual religious commitment and encourage pro-social values and norms; thus, we expect that (Hypothesis 2b) county-level religious adherence will condition the effect of religious involvement on recidivism.

In accord with recent studies of social service provision and economic disadvantage, our third hypothesis suggests that (Hypothesis 3a) individuals released into areas with greater social service assistance will report significantly lower recidivism and (Hypothesis 3b) individuals released into areas with greater levels of economic deprivation will report significantly more recidivism. Finally, we expect these same ecological conditions to moderate the influence of HSR participation on recidivism such that (Hypothesis 4a) the combination of individual religious involvement and available social service assistance will be associated with a lower probability of recidivism and (Hypothesis 4b) economic deprivation will weaken the relationship between religious involvement and recidivism.

Data and Methods

Data

Individual level data were supplied by Correctional Services of the Oregon Department of Corrections (ODOC). The unit is responsible for overseeing intake and assessment, institutional services such as health, drug and alcohol, education, volunteer and Humanist, Spiritual, and Religious (HSR) services, in addition to reentry services across all prisons in Oregon. The unit began recording the amount of time spent in humanist, spiritual and religious programs in 2004. The sample provided for this study included the 1362 males admitted to the Oregon state prison system in 2004 who remained incarcerated for at least 1 year and who were released back to the community within 5 years. The sample excluded all people admitted to the ODOC in 2004 who were not released, in addition to those transferred to another state prison after admission, persons released to an immigration detainer after identification by Immigration Customs and Enforcement (ICE), and deported, and persons who died. These criteria meant that when follow-up data were collected at the end of 2017, each person had a follow-up period of at least 8 years.

This dataset, which included information on the county to which a person was released, was combined with county-level data obtained from two sources. Social ecology data were obtained from the 2010 U.S. Census, and accessed via the National Historical and Geographic Information System (NHGIS). The 2010 U.S. Religion Census: Religious Congregations and Membership Study provided county-level data on moral communities and was accessed via the Association of Religion Data Archives (ARDA). Although a large ecological unit, social, religious, and economic conditions vary significantly between counties. And as noted by Mears et al. (2008), political and criminal justice practices are typically organized at the county-level, suggesting that individuals who return to the same county are more likely to experience similar reentry and criminal justice resources, than individuals who return to another county. In Oregon, each person released from prison is released under supervision by a county community corrections department. Additionally, the county represents a meaningful context for religious institutions, religious group-norms, and resources (Borgonovi 2008; Dyreng et al. 2012), and has been used in the few existing studies assessing the role of religious contexts in criminal justice (Lee and Batowksi 2004; Longest and Uecker 2017; Ulmer et al. 2008).

Dependent Variable

Recidivism

There are many ways to measure recidivism, each with its own assumptions and criteria for behavior, and each with its own limitations (Lyman and LoBuglio 2006). Given that rearrests may include incidents for which charges are dropped or a person is found not guilty, we follow prior studies (e.g. Hipp et al. 2010; Mears et al. 2008) by defining recidivism as an individual being reconvicted for any new offense (including statutory, property and violent offenses) after release from prison. This binary measure captures whether a person was reconvicted at all (1 = yes, 0 = no) within the first 3 years post-release, 5 years post-release, and 8 years post-release. As displayed in Table 1, 30% of individuals were reconvicted within 3 years post-release, 39% by the end of their 5th year, and 47% by the end of their 8th-year post-release.

Table 1.

Descriptive characteristics of all variables used in the study (n = 1362)

Variable	Mean	SD	Min	Max
Individual-level
3 year recidivism	0.300	0.458	0	1
5 year recidivism	0.391	0.488	0	1
8 year recidivism	0.471	0.499	0	1
HSR monthly attendance	1.486	2.539	0	22
Education	0.850	0.356	0	1
Job preparation	0.967	0.179	0	1
Cognitive	0.675	0.469	0	1
Substance abuse	0.322	0.467	0	1
Ethnicity (white = 1)	0.845	0.362	0	1
Age	33.264	10.393	16	76
Known gang member	0.080	0.272	0	1
Time served (years)	1.557	0.515	1.014	4.606
Sanctions	1.830	2.213	0	7
Current offense
Violent	0.211	0.408	0	1
Drug	0.270	0.444	0	1
Sex	0.133	0.339	0	1
Other	0.386	0.418	0	1
Assessed risk	0.245	0.186	0	0.851
County-level
Adherence rate (lg)	5.662	0.220	5.293	6.364
Economic disadvantage	0.000	0.950	− 1.875	1.932
Family poverty	0.105	0.023	0.061	0.152
Public assistance	0.030	0.006	0.016	0.069
Female headed households	0.155	0.026	0.108	0.204
Median income	44,726.61	6728.178	35,974.000	62,574.000
Social service	1.920	0.345	1.411	3.656
Law enforcement rate (lg)	2.700	0.525	1.946	3.829
Population (lg)	10.815	1.382	7.534	13.508
Gini coefficient	0.019	0.011	0.004	0.056
Metropolitan area	0.222	0.421	0	1

Individual Covariates

Religious Involvement

Our key individual-level explanatory variable captured the average monthly attendance at a humanist, spiritual, or religious (HSR) program. This included any program provided in prison by a chaplain or faith-motivated volunteer.¹ This was calculated by summing the total number of attendances during the entire prison stay and dividing by the total number of months served. This measure presents a picture of someone’s overall level of participation, rather than isolating their participation at one point in time, and has a mean of 1.486, standard deviation of 2.539, and ranges from 0 (no religious involvement) to 22 (very high proportion of religious involvement). While prior evaluation studies have captured program participation with a dichotomous measure, assessing the average monthly religious involvement more accurately captures the ongoing commitment and extent to which an individual is involved and exposed to program messages (Duwe and Clark 2017).

Needs Program Participation

While HSR services are open to all people in prison, due to a scarcity of programming resources, many needs-based programs are reserved for those with the highest risk to recidivate. There is inevitably overlap in program participation, however, especially among persons with the greatest motivation to engage in treatment. To ensure that participation in HSR services is not confounded with participation in other empirically-based rehabilitation programs, we also control for whether an individual participated in any of the following needs-based programs: alcohol and other drug treatment, cognitive-based therapy, education programs (including GED preparation), job skills development, and on the job training. Dichotomous measures of each of these variables were included as control variables. Most individuals had at least some hours spent in job preparation (97%) and education programs (85%), but there were less individuals who spent any amount of time in cognitive or substance use programs.

Recidivism Risk

The Oregon DOC uses an Automated Criminal Risk Score (ACRS), a validated risk measure (Henning et al. 2013), to assess the risk of recidivism for an Oregon inmate. The ACRS is based on scores from 7 items: age, earned time, sentence length, prior revocation, number of prior incarcerations, prior theft convictions, and the current offense severity. The ACRS is a continuous measure ranging from 0 to 1.

Prisoner Conduct

Considering research that has indicated a relationship between prison misconduct and recidivism among adults (Caudill and Trulson 2016; Cochran et al. 2014), we capture the average number of sanctions received per year of a person’s prison stay. The count of sanctions exhibited significant right skew (Skewness = 2.538, Kurtosis = 11.750), with some outliers identified at 25 sanctions or higher. To reduce the skew with this measure, we truncated the number of sanctions at 7 incidents. A little over 8% of the sample had more than 7 infractions per year of their stay. Additionally, we include a dichotomous measure indicating whether the prisoner was a known gang member. We also include the total time served in prison (measured in years) because longer time incarcerated is correlated with a lower risk of recidivism (Caudill and Trulson 2016). Although many reentry studies also control for whether a person is released on parole or other supervision conditions, all persons released from the Oregon DOC are on post-prison supervision.

Demographic Characteristics

Finally, we include other characteristics of the individual shown to have implications for recidivism risk including age at intake and ethnicity (ethnic minority = 0; non-Hispanic White = 1). The sample was largely homogenous with so few cases of non-Hispanic White that attempts to further define racial/ethnic categories were not possible due to extremely low cell frequencies. We also control for current offense type, as the difficulties of reentry process may be altered by the nature of a person’s offense and criminal record. To account for this possibility, we include a dummy variable indicating whether an individual was convicted of a violent offense (21.1%), or a drug offense (27.0%), or a sexual offense (13.3%) in contrast to some other offense (38.6%).

County‑Level Covariates

In line with the moral community hypothesis, our overall goal of this study is to test the direct effect of religious community on individual recidivism. Consistent with other recent studies (i.e. Longest and Uecker 2017), we obtain county-level estimates of moral community from the 2010 U.S. Religion Census. The U.S. Religion Census is unique in that it provides a portrait of county-level religious behavior and affiliation nationwide. Details of the full methodology can be obtained from the US Religion Census (Grammich 2012), but in short, data collection relied on contact persons for each religious body listed in the Yearbook of American and Canadian Churches to compile data by county for all their congregations. For our study, we created a measure of the number of religious adherents as a percentage of the county population. Adherents may include all persons with an affiliation to a congregation, including children, members, and attendees who are not members. The adherent figure is designed to be the most complete count of people affiliated with a congregation, and the most comparable across counties (Grammich 2012). Data from the US Religion Census reveal that Oregon has one of the lowest rates of adherence to any religion (31.2% of the population, second only to Maine) of all 50 states. Although significant variation still exists within Oregon, it suggests that our examination will be limited in generalizability, potentially underestimating the effect of a moral community compared to states that report adherence rates above 60% (e.g. Alabama, Louisiana, North Dakota, Utah).

Our second goal was to examine whether wider economic resources and social service support conditions the effect of individual-level religious involvement on recidivism. As with prior studies examining the criminogenic consequences of economic disadvantage (Phillips 2002; McCall et al. 2008), factor analysis was used to develop a composite measure reflective of disadvantage and limited economic resources. Our factor included four items measured at the county-level: median income, the percentage of families living in poverty, the percentage of family households headed by a female with no husband present, and the percentage of families receiving public assistance (Eigenvalue = 2.41; all factor loadings exceeded .75).

Informed by the work of Xie et al. (2012), we also created a county-level measure of the rate of social service assistance. Social service assistance was obtained from Table B24010 of the American Communities Survey 2008–2010 estimates and calculated via the percentage of employees who were in an occupation of “Community and social service occupations”, divided by the county-population over the age of 16 and in the labor force. Community service occupations include two categories defined in the American Communities Survey: “Counselors, social workers, and other community and social service specialists”, and “Religious workers”. These positions reflect the availability of social service assistance and support, which are a critical component of prisoner reentry success (Morenoff and Harding 2014; Hipp et al. 2009; Visher 2005). Indeed, recent surveys suggest that the fastest growing position in this occupation category involves rehabilitation and drug abuse counseling (Bureau of Labor Statistics 2017).

We also included several other county-level measures to help further refine the ecological context of each county. Informed by Mears et al. (2008), and consistent with prior studies capturing police force size (Marvell and Moody 1996; McCall et al. 2008), all models also included a measure of police force size, obtained from Police Employee (LEOKA) data to calculate the number of full-time sworn officers per capita. As is often the case, the measure was log transformed to reduce skew associated with the measure (McCall et al. 2013). We also include a measure of the total county population (log transformed), the level of income inequality as measured by the Gini coefficient, and a dichotomous measure indicating the presence of an Metropolitan Area, defined by the Office of Management and Budget.

Analysis

This study assesses recidivism as a function of religious involvement and other covariates, while considering that individual desistance processes are also dependent on the social ecology to which an ex-offender is released after prison. Significant variation exists in the follow-up periods used by studies of recidivism, including periods of 15-months (Boman and Mowen 2017), 3 years (Mears et al. 2008), or longer (Huebner et al. 2010; Kurlychek et al. 2012). The current analysis proceeded to examine the effect of religious involvement on rearrests within fixed-follow up periods of 3 years, 5 years and 8 years post-release. Using these three periods, logistic regression models are estimated to examine the relation between religious involvement and recidivism over short-term and long-term intervals, controlling for individual demographics and characteristics.

Next, to move beyond examining only the micro-level context in which religious involvement relates to recidivism, and to explore expected insignificance of religious involvement over the long-term (Giordano et al. 2008), we then examine the direct and moderating roles of religious context and social ecology on the link between religious involvement and recidivism during each follow up period. To accomplish this, we estimate a series of hierarchical models introducing the level-two covariates. As the dependent variable is dichotomous (1 = recidivism), we employ a series of mixed-effect hierarchical generalized linear models (HGLMs, see Rabe-Hesketh and Skrondal 2012) that relax several traditional assumptions about the distribution of regression residuals through the introduction of a random intercept, and incorporate our county-level predictors.² Specifically, the inclusion of a random intercept accounts for a lack of independence across observations as individuals (at level one) are nested within counties (at level two).

Prior to analysis, we examined bivariate correlations among our covariates to reduce concerns of multicollinearity. No correlation between individual-level variables exceeded .174. Higher correlations were observed at the county-level, between economic disadvantage and social service assistance (r = .367) and between economic disadvantage and religious adherence (r = .374). County-level variables were entered in our HGLM models one at a time to ensure that the magnitude and direction of estimates did not change significantly. Given consistency in the county-level coefficients, we present mixed-effects models with all county-level variables entered simultaneously.

Results

Logistic Regression

Prior to testing our main hypotheses, we used our level-1 variables to assess the relationship between religious involvement and recidivism using a logistic regression model, estimating the likelihood of reconviction within 3 years, 5 years, and 8 years post-release, as a function of individual level variables only. Results are displayed in Table 2. Consistent with prior literature on recidivism, higher assessed risk for recidivism and more sanctions in prison were positively associated with recidivism within the first 3 years after release. Older age was associated with a lower likelihood of recidivism. Higher levels of religious involvement, measured by average monthly participation in HSR services, was associated with a marginally significant reduction in the likelihood of recidivism within 3 years post-release. Given the higher likelihood of recidivism within the first 2–3 years after release, this protective benefit of HSR involvement supports numerous other studies that have found religious identification and participation in faith-based services can help promote desistance. As many studies of recidivism routinely use longer follow-up periods, we reestimated the model using 5- and 8-year follow-up periods. Sanctions and assessed risk remained significantly associated with recidivism within longer follow up periods, while a significant and negative relationship between non-Latino White ethnicty and recidivism also emerged. With a longer follow-up period post-release, HSR involvement was not significantly associated with recidivism. These findings reflect the conclusion of several recent studies on the relation between religious involvement and recidivism, conclusions are difficult to draw (Power et al. 2014) and the relationship is not always sustained over the long term (Giordano et al. 2008). We explore whether there are other dynamics occurring at the macro-level that are influencing the link between individual religious involvement (a micro-level process) and recidivism using hierarchical models below.

Table 2.

Logistic regression of recidivism controlling for all individual level covariates (n = 1362)

	3 year recidivism		5 year recidivism		year recidivism
	Coef	SE	Coef	SE	Coef	SE
HSR	− 0.067*	0.031	− 0.049+	0.026	− 0.037	0.024
Cognitive	0.064	0.143	0.129	0.133	0.099	0.128
Job skills	− 0.015	0.387	0.242	0.375	0.128	0.359
Education	− 0.022	0.196	0.063	0.182	0.161	0.177
AOD	0.061	0.145	0.193	0.135	0.135	0.132
Sanctions	0.364**	0.031	0.326**	0.028	0.306**	0.028
Assessed RISK	1.431**	0.367	1.626**	0.344	1.643**	0.337
Gang member	0.107	0.228	0.176	0.223	0.053	0.225
Curr off: violent	− 0.223	0.174	− 0.206	0.162	− 0.302	0.156
Curr off: drug	− 0.096	0.157	− 0.047	0.147	− 0.093	0.143
Curr off: sexual	− 0.133	0.203	− 0.026	0.189	− 0.065	0.184
Time served	0.111	0.152	− 0.041	0.121	− .011	0.118
Age	− 0.015*	0.007	− 0.010	0.006	− 0.016**	0.006
White	− 0.093	0.179	− 0.398*	0.168	− 0.473**	0.167
Intercept	− 1.840	0.510	− 1.145	0.451	− 0.373	0.432
Chi2	218.59		216.83		212.36
Log likelihood	− 718.36		− 798.92		− 831.87

SE Standard error

^*p < .05;

^**p < .01;

⁺p < .10

HGLM Analysis

Table 3 reports the results of our HGLM models estimating the likelihood of reconviction using a 5-year follow up period (the first point at which the association between individual HSR involvement and reconviction is no longer significant by conventional standards). This table builds upon Table 2, model 2, by introducing county-level covariates in model 1, in addition to the inclusion of a county-level random intercept. As before, higher assessed risk and sanctions in prison were consistently and positively associated with recidivism. By contrast, older age and non-Hispanic white ethnicity were both negatively associated with recidivism. Turning to our county-level estimates, the results in Table 3, model 1 reveal that two of our county-level measures had direct effects on recidivism. The normative moral climate was significantly associated with lower recidivism while economic disadvantage was significantly associated with higher recidivism, such that individuals released into areas with higher disadvantage had a significantly higher likelihood of reconviction.

Table 3.

Hierarchical generalized linear regression modeling examining reconviction within 5 years (n = 1362)

	Model 1		Model 2		Model 3		Model 4
	Coef	SE	Coef	SE	Coef	SE	Coef	SE
HSR attendance	− 0.048	0.027	0.041	0.763	− 0.049	0.032	− 0.047	0.027
Program: cognitive	0.074	0.137	0.075	0.137	0.070	0.137	0.077	0.137
Program: jobs	0.167	0.388	0.166	0.388	0.174	0.388	0.157	0.388
Program: education	0.121	0.191	0.121	0.191	0.119	0.192	0.124	0.191
Program: AOD	0.177	0.142	0.177	0.142	0.169	0.142	0.174	0.142
Sanctions	0.341**	0.031	0.341**	0.031	0.342**	0.031	0.342**	0.031
Assessed risk	1.591**	0.359	1.590**	0.359	1.596**	0.359	1.586**	0.359
Gang member	0.142	0.235	0.142	0.235	0.141	0.235	0.137	0.235
Current offense: violent	− 0.219	0.168	− 0.219	0.168	− 0.219	0.168	− 0.225	0.168
Current offense: drug	− 0.011	0.152	− 0.010	0.153	− 0.017	0.152	− 0.005	0.152
Current offense sexual	− 0.042	0.196	− 0.042	0.196	− 0.048	0.196	− 0.037	0.196
Time served	0.187	0.147	0.188	0.147	0.188	0.147	0.185	0.147
Age	− 0.011+	0.007	− 0.011+	0.007	− 0.012+	0.007	− 0.012+	0.007
White	− 0.439*	0.175	− 0.438*	0.175	− 0.439*	0.175	− 0.438*	0.175
County-level
Adherence	− 0.955**	0.356	− 0.934*	0.398	− 0.953**	0.355	− 0.931**	0.356
Disadvantage	0.177*	0.077	0.177*	0.077	0.139	0.087	0.168*	0.082
Social service	− 0.119	0.274	− 0.118	0.275	− 0.123	0.275	− 0.018	0.258
HSR*adherence			− 0.016	0.133
HSR* disadvantage					.151**	0.057
HSR*social service							− 0.091+	0.046
Random variation	0.08		0.09		0.08		0.09
chi2	181.20**		181.19**		181.61**		182.74**

All models include controls for county population, police force size, inequality, and the presence of a metropolitan area

SE Standard error

^*p < .05;

^**p < .01;

⁺p < .10

To test our remaining hypotheses, the next three models in Table 3 (models 2, 3 and 4) introduce interaction terms encompassing each of the three ecological conditions along with religious involvement. In model 2, the interaction encompassing county-level religious adherence and individual religious involvement failed to reach significance suggesting that community religious adherence is independent—and not interdependent—in its effect on recidivism. In model 3, the interaction encompassing disadvantage and HSR participation reached significance suggesting county-level disadvantage and HSR participation interact in their effects on recidivism. Finally, in model 4, the interaction encompassing social service and HSR participation failed to reach conventional significance but was marginally significant (p < .10).

The intuitive interpretation of interaction coefficients from linear regression models does not apply in nonlinear models. As discussed by Norton et al. (2004), statistical significance cannot be interpreted from z statistics in the regression output, and the direction of the interaction term coefficient may differ across observations. In order to more accurately understand the interactive nature of economic disadvantage and individual religious involvement on the probability of reconviction, we plot marginal effects (Williams 2018) shown in Fig. 1 below by fixing disadvantage at below and above median values, varying HSR involvement and holding all other covariates constant. As illustrated by Fig. 1, higher than median disadvantage reduces the prosocial influence of HSR on reconviction.

Fig. 1.

Predicted margins of disadvantage in 5 year follow-up period

Additionally, based on the finding of a significant and negative interaction found by Harris and Ulmer (2017), signifying that black Protestant adherence dampened the criminogenic effects of structural disadvantage on crime, we also tested an interaction (not disaplyed) between county-level adherence and county disadvantage. This interaction was not significant. This is likely a consequence of the research setting of this study, whereas Harris & Ulmer’s study focused on Black Protestant adherence in 733 predominantly black counties, where the church has been particularly influential throughout history.

Supplementary Analysis

A number of additional tests were conducted to increase confidence in our substantive findings. Reentry research incorporating survival analyses has become more popular as a way to increase the accuracy of conclusions (Spohn and Holleran 2002; Huebner et al. 2010) and, as a result, we also employed a mixed-effects survival model to examine reconviction. The models in “Appendix 1” combine the multilevel modeling of Table 3 with the parametric analysis of survival-time outcomes, where “failure” is reconviction within 5 years post-release, and the time to rearrest (that results in a reconviction) is measured in days after release from the ODOC. A two-level mixed-effects Weibull regression extends a typical survival model by incorporating a county-level random intercept. Focusing on our key explanatory variables, a higher level of participation in HSR programs was not associated with the rate of reconviction once we controlled for assessed risk, individual, and county-level characteristics. Results demonstrated that county adherence and county economic disadvantage were both directly associated with the rate of recidivism. Interaction terms encompassing the joint effects of religious adherence, social service availability and economic disadvantage with HSR were entered individually in Models 2 through 4. We again found that the effect of religious involvement on reconviction differs depending on the levels of county economic resource deprivation.

In using a 5 year follow up period for our multilevel analyses, we interpreted the significant interaction between HSR involvement and economic disadvantage as evidence that higher county-level disadvantage takes on greater significance over longer periods post-release, dampening the protective effects of HSR involvement. To confirm this assumption, HGLM models were replicated using 3- and 8-year follow-up periods to see whether this interaction is consistent across follow-up periods. As revealed in “Appendix 2”, the protective effects of HSR involvement are roughly equal across levels of county economic disadvantage using a shorter follow-up period. Using a longer follow-up period, however, again shows a significant interaction between HSR involvement and economic disadvantage.

Discussion

The present research addressed two primary objectives. First, it estimated the extent to which individual-level involvement in humanist, spiritual, or religious (HSR) services was associated with recidivism. Second, the analysis examined the extent to which the important macro-level community factors of religious adherence, economic disadvantage, and social service availability conditioned the estimated effect of religious involvement. Unlike prior studies which assessed the religiosity–recidivism relation using individual-level measures only, the present research estimated the effect of individual and county-level characteristics on the likelihood reconvictions to more clearly test micro–macro linkages (e.g., Matsueda 2017).

Our first hypothesis predicted a negative association between individual HSR participation in prison and reconviction. This hypothesis was supported over the critical and shorter 3 year follow up period. This association dissipated when examining recidivism over longer follow-up periods, however, after controlling for other individual-level covariates. As numerous scholars have suggested (Clear et al. 2000; Giordano et al. 2008; Stansfield et al. 2017), the failure to find a significant association between services in prison and longer-term recidivism benefits does not mean religious services do not serve a vital mission. Those who reject or ignore religion in conversations about penal reform neglect the role religion plays in making prison tolerable (Green 2013; Hallett et al. 2017). The pro-social “gospel model” as highlighted by Hallett et al. (2017) is less about what works, and more about what is morally right. Indeed, our data reaffirmed the high demand in prison for humanist, spiritual and religious services. As such, ensuring volunteers can get into prisons and offer services is important as incarcerated men and women look for spaces where they can make meaning in their lives during incarceration. However, such a finding does suggest more theoretical work is needed on the relation between religious involvement and recidivism. Although narratives of identity change often include stories of religion and spirituality, quantitative assessments often do not bare these findings out over the long-term (Giordano et al. 2008). This suggests that there are some religious and non-religious challenges that individuals face over the long-term desistance journey that are not accounted for in the research. Furthermore, the challenges and distress faced by reentering prisoners may in turn create spiritual struggles and anger towards God (Exline and Rose 2005; Wortmann et al. 2011), potentially undoing the pro-social effects of religion developed during prison.

Although we did not find an effect of HSR at the individual-level, results for the second hypothesis, that the proportion of religious adherents in the county would relate to decreased recidivism (hypothesis 2a) and that this macro-level factor would condition the effect of HSR on recidivism (hypothesis 2b) received mixed-support. In support of this hypothesis, we found that individuals who returned to counties with greater proportions of religious adherents reported significantly reduced odds of reconviction than individuals who returned to areas with lower proportions of religious adherents. Scholars have debated whether there is something specific about faith that provides recidivism benefits, or whether the negative associations between religion and recidivism are solely a function of the support and connections to services that religion can provide. Our finding that the moral community is associated with recidivism, and that this relationship is largely independent of the individual’s participation in HSR, signifies the salience of faith, whether individuals are aware of its presence or not. Findings from the United Kingdom suggest that not everyone with access to services from the voluntary sector will accept help (Dominey and Lowson 2017). Yet faith-derived values and norms can transcend individual program participation.

The significance of moral communities is consistent with prior studies linking religious contexts and negative behavioral outcomes (Harris and Ulmer 2017; Lee and Bartowski 2004; Regnerus 2003; Ulmer and Harris 2013), which suggest that moral communities may promote values and norms which permeate social boundaries in a community, and civically-engaged denominations will promote denser prosocial ties in a community (Lee and Bartowski 2004). This study was unable to assess potential negative effects of moral communities. As suggested by Ulmer et al. (2008), some moral communities (in their case, rural Pennsylvania) may demand a certain level of action and punishment by criminal justice officials which may make rehabilitation and desistance more difficult for individuals returning from prison. We were unable to capture data reflecting these political and judicial values which may be inherent in a moral community, but like prior scholars (Savelsberg 2004; Ulmer et al. 2008), we suggest more work is needed that captures religious contexts, in addition to further work exploring how individual religious characteristics may interact with religious contexts in both positive and negative ways.

Our next set of hypotheses, that the ecological conditions of social assistance (hypothesis 3a) and economic disadvantage (hypothesis 3b) would significantly relate to recidivism as main effects, also received mixed support. Economic disadvantage was consistently associated with higher recidivism, but wider social service availability (which is higher in communities of economic disadvantage) was not. Instead, we found greater support for hypothesis 4a and 4b as both ecological characteristics conditioned the relationship between religious involvement and recidivism. To this end, our findings of a non-association between participation in HSR services and recidivism appeared to be the result of variation in wider economic deprivation and, to a lesser extent, social assistance availability in the counties to which people returned after prison. Although we are unable to examine change in individual religious involvement or commitment post-release, it is possible that initial optimism and positive identity changes incited by faith in prison are undermined and reversed by difficulties in finding support and economic resources after release. To this end, criminologists must do a better a job of keeping pace with the religion literature. As noted by Jang and Franzen (2013), criminologists have largely considered religion through a one-time measure of participation. But a fruitful area of future research may be to consider research into spiritual struggles, recognizing that difficulties faced by previously incarcerated individuals in employment, social standing and emotional distress, may alter the way one interacts with religion.

This finding also has practical implications for members of the faith-community who conduct outreach and support. The faith community and religious workers have a well-documented history of aiding with housing, employment, and access to social and health services; however, as documented by Hipp et al. (2008, 2009), demand in poor areas may exceed supply. As an example, the size of the African American ex-offender population has placed enormous pressure on predominantly Black communities to absorb these individuals, with more limited donations and local funding available to support outreach (Brown 2010). Greater collaboration between volunteers and faith-providers across county-lines may help to equalize the supply and demand ratio.

In addition to the policy and theoretical implications, this study caries important methodological considerations. Although the field of criminology has long recognized the importance of macro-level factors on micro-level behavioral outcomes (and, of course, vice versa), specifying this link has been difficult to do in practice, particularly in studies on reentry and religion. Though we concentrate on only three macro-level conditions (religious adherence, social support, and economic deprivation) at the county-level, we believe this study builds upon existing literature by demonstrating that although individual-level religious support (measured as HSR in the current study) fails to significantly relate to recidivism, the importance of religion as a protective factor against recidivism materializes as an important broader macro-level characteristic of the community to which one returns. This finding becomes more salient given the well-established literature on the link between religion and offending that stresses how individual identity transformations occur within the broader religious and faith-based community. Moreover, negative structural conditions like economic deprivation can undermine the prosocial influence of religious support for returning individuals while the presence of social support can increase the influence of religious support during the reentry process. At best, failure to consider this important micro–macro link overlooks three seemingly important dimensions that affect recidivism. At worst, this oversight results in a complete misunderstanding of the meaning and role of religious support for men and women as they face the challenges of returning to their communities.

Apart from the findings, the present research has some limitations. As noted, we do not measure change in individual religion after release, rather we assume that the level of involvement in HSR events in prison is predictive of on-going participation after release. This limitation is important as prior work has demonstrated that changes in religious involvement influence offending during reentry (e.g., Mowen et al. 2017). Future research should seek to overcome this limitation because stressful life events can precipitate spiritual struggles, in which individuals come to see religion and meaning in a new light, leading to a reappraisal of God (Pargament et al. 1998). Relatedly, this study does not assess potential harms of either religious involvement or of moral communities. While the focus on the positive effects of religious involvement is not dissimilar from the body of work on religion in the lives of people incarcerated or recently released, this does limit our ability to advocate for greater faith-informed program opportunities for people in prison.

There are also limitations to the data. Given that the sample was from Oregon, we have a relatively small sample of counties, and the counties exhibited a lower rate of adherence to any religion than any other state in the US, except for Maine. Both prison and county populations also lacked diversity in terms of race and ethnicity. This is important as prior research has found racial/ethnic differences in the extent to which individuals come to rely on religion for support (e.g., Stansfield 2017), connected to historical differences in the social, political and economic benefits provided by black Protestant churches to community members (Harris and Ulmer 2017). As a result, findings from this study may not be generalizable to other demographics and requires replication across a larger number of aggregate units (neighborhoods or counties). Additionally, there are numerous macro-level characteristics that influence individual behavior and future research should examine how other measures of collective efficacy and social disorganization effect the link between individual-level religious support and recidivism.

Nevertheless, the theoretical and policy implications of these findings are important. The results emphasize the importance of considering the normative climate and moral community when studying the relationships between religious involvement and recidivism. And while the availability of humanist, spiritual and religious programs in prison is important, the recidivism benefits may be maximized when combined with a continuum of support and access to economic resources after release. Future data collection efforts should on centered on the demand and supply of faith-based services and social support in high disadvantage areas, in addition to a focus on those who turn to the voluntary sector for support. This may help us better understand how the faith community is serving ex-offenders after release, whether service providers are over-burdened in some communities, and how one’s faith responds in the face of these reentry challenges.

Appendix 1

See Table 4.

Table 4.

Mixed effects proportional hazard models of days to reconviction

	Model 1		Model 2		Model 3		Model 4
	HR	SE	HR	SE	HR	SE	HR	SE
HSR involvement	0.971	0.018	0.957	0.497	0.972	0.019	0.971	0.02
Adherence	0.557*	0.130	0.556*	0.131	.559*	0.131	0.557*	0.130
Disadvantage	1.107*	0.056	1.107*	0.056	1.083	0.061	1.108*	0.056
Social service	0.933	0.157	0.925	0.156	0.921	0.156	1.101	0.204
HSR*adherence			1.002	0.090
HSR* disadvantage					1.102**	0.040
HSR*social service							0.959	0.028
Chi2	245.38		245.38		246.11		246.73

All models include controls for all Level 1 covariates displayed in Tables 2 and 3, in addition to Level 2 controls for controls for county population, police force size, inequality, and the presence of a metropolitan area

HR Hazard ratio, SE standard error

^*p < .05;

^**p < .01**;

⁺p < .10

Appendix 2

See Fig. 2.

Fig. 2.

Predicted margins of disadvantage in 3 and 8 year follow-up periods