Predictors of Religiosity among US Prisoners

Abstract

Criminological research on religiosity among prisoners has focused on the effects of or outcomes associated with religiosity. Studies have discovered that faith-based programs can reduce recidivism and that religiosity facilitates adaptation to imprisonment and is associated with reductions in serious misconducts. Criminologists have yet to examine the predictors of religiosity among prisoners. In this study, I examine individual- and facility-level predictors of inmate religiosity to uncover the relationship between individual demographic and criminal justice characteristics and religiosity among prisoners. I use national data sets, the Survey of Inmates in State Correctional Facilities and the Census of State Correctional Facilities, and multilevel modeling techniques to examine these relationships. Findings at the individual level indicate that the same factors that are important influences on religiosity in the general population are also significant predictor of religiosity among prisoners, and that the criminal justice/criminal history characteristics of prisoners are also important influences on religiosity. At the facility level, prisons in the Southern region of the US had the highest rates of religiosity among prisoners.

1. Introduction

The historical role and influence of religion, particularly Christianity, in US prisons is widespread. The first penal institutions in the US were designed by the Quakers as alternatives to brutal and public corporal punishment for law violators (O’Connor 2002; Sumter et al. 2018). The goal of these first prisons, from the Quaker perspective, was to isolate criminals from harmful social influences and, through solitary prayer and meditation, to allow individuals to repent or become penitent for their sins, hence the term “penitentiary” (Clear and Sumter 2002; Sumter et al. 2018). More recently, faith-based programs in prison have garnered political support and popular attention as a way to rehabilitate prisoners, with some prisons even providing seminary schooling for offenders (Jang et al. 2020; Mears 2007; Mears et al. 2006).

Empirical research on the effectiveness of religiosity in prison indicates that religiosity can reduce serious forms of inmate misconduct, but the evidence that religiosity is associated with rehabilitation/reduced recidivism among inmates is mixed. Studies consistently find that religiosity is associated with a reduction in serious, violent misconduct (i.e., institutional rule violations). The relationship between religiosity and less serious forms of misconduct is much less clear (Kerley et al. 2006; Meade 2014; Steiner et al. 2014; Sturgis 2010). Research has also uncovered that the potential of religion and faith-based programs to reduce recidivism is limited, with few programs indicating promising results (Johnson 2004; Johnson and Larson 2003; Johnson et al. 1997; Stansfield 2017). Criminological research exclusively frames religiosity as an independent variable. That is, scholars hypothesize and test whether variation in religiosity of individuals or prisoners produces observed variation in other outcomes, such as crime, substance use, misconduct, and/or recidivism. An important extension of criminological research is to examine religiosity as a dependent variable and identify those factors that produce observed variation in levels of religiosity among an incarcerated sample.

In this study, I utilized national data sets of prisons and prisoners (2000 Census of State Correctional Facilities [CSCF] and the 2004 Survey of Inmates in State Correctional Facilities [SISCF]) to examine individual- and facility-level predictors of inmate religiosity. I also examined the impacts of demographic and criminal history measures on the religiosity of prisoners.

2. The Effect of Religiosity in Prison

Research on religion in prison has often focused on the impact of religion on behavioral outcomes of prisoners, specifically misconduct in prison and recidivism after release (Stansfield et al. 2017). Studies have found that religious prisoners are less likely to commit misconduct, particularly serious and violent misconduct, than nonreligious prisoners (Camp et al. 2008; Jang et al. 2018; Kerley et al. 2006; Meade 2014; Sturgis 2010). A systematic review of studies conducted between 2000 and 2013 utilizing multivariate analyses to estimate the impact of religiosity on prisoner misconduct found that 68 percent of studies found a significant, inverse relationship between religiosity and prisoner misconduct, while 32 percent of studies found no significant relationships (Meade 2014). More recently, Jang et al. (2018) found that a religious conversion in prison was directly associated with reductions in the probability of having disciplinary infractions, and religiosity was indirectly related to prison misconduct through cognitive/identity transformations. One consistent finding emerging from these studies is that religiosity in prison is consistently associated with lower prevalence and incidence of violence in prison, but religiosity is much less likely to be significantly associated with minor forms of misconduct (Camp et al. 2008; Kerley et al. 2006; Meade 2014; Meade and Bolin 2018; Steiner and Wooldredge 2008; Steiner et al. 2014; Sturgis 2010).

Camp et al. (2008) found that participation in a faith-based program called the Life Connections Program (LCP) was associated with lower odds of committing serious misconduct (homicide or escape), but not with lower odds of less serious misconduct. Kerley et al. (2006) also discovered that participation in a prison ministry program reduced the likelihood of arguing and fighting with other inmates compared to those inmates who were not involved. A number of studies have utilized the 2004 Survey of Inmates in Correctional Facilities to estimate the relationship between measures of religiosity and inmate misconduct (Meade 2014; Meade and Bolin 2018; Steiner and Wooldredge 2008; Sturgis 2010). Studies utilizing this data also consistently find that religiosity is associated with lower odds of violent assaults against other inmates and staff members, while less serious forms of misconduct (substance and nonviolent rule violations) are not influenced by measures of prisoner religiosity. This finding among these data was even replicated after utilizing propensity score matching to produce samples of religious and nonreligious inmates from the data that were statistically balanced on relevant covariates of religiosity and misconduct (Meade 2014).

Research evaluating the effectiveness of religion on recidivism (rearrests, reconviction, incarceration, etc.) are less clear compared to the literature on religiosity and misconduct. For instance, Stansfield (2017) discovered that religiosity was not consistently associated with self-reported serious offending in the Pathways to Desistance data but was associated with reduction in drug use. Research in Pakistan with Muslim individuals has discovered that religiosity is associated with lower odds of recidivism when measured as probation failure/revocation (Bhutta et al. 2019; Bhutta and Wormith 2016). Much of the literature on religiosity and recidivism takes the form of faith-based program evaluations. Many evaluations of faith-based programs have produced no statistically significant differences between program participants and samples of inmates who did not participate in faith-based programs (Haviv et al. 2020; Johnson 2004; Johnson et al. 1997; Johnson and Larson 2003; Stansfield et al. 2017). On the other hand, some evaluations have produced promising results for faith-based programs. Duwe and King (2013) found that participation in a faith-based program called the Inner Change Freedom Initiative (IFI) was associated with reductions in rearrests, reconvictions, and reincarcerations. One similar finding across studies is that program graduates or those who participate in faith-based programs to a high degree have lower recidivism rates than nonparticipants and/or noncompleters (Camp et al. 2006; Daggett et al. 2008; Mears 2007). The implication of these findings is that inmates who select into faith-based programs or are motivated to succeed may have lower odds of recidivism, regardless of religiosity. These findings underscore the importance of understanding the correlates of inmate religiosity, as well as the reasons why prisoners may become involved with religion in prison in the first place.

Qualitative research with prisoners has shed some light on this question. Framed within life-course theories, a religious conversion in prison may provide a spiritual transformation that can aid in rehabilitation processes and situate prisoners to change their thinking patterns and seek help (Jang et al. 2018; Johnson 2011; Maruna et al. 2006). Spiritual transformations may promote individual growth and moral development and may function as a cognitive shift (Giordano et al. 2008; Jang et al. 2018). Jang et al. (2020) found that exposure to prison peer ministers was associated with an increase in prosocial attitudes and virtue and a reduction in self-reported aggression. Other qualitative studies have revealed the value that faith can have for those incarcerated (Clear et al. 2000; Dammer 2002). Religion helps prisoners find a new identity of self-worth and value, cope with loss of freedom and autonomy, deal with their guilt, and find forgiveness for their crimes. Religious traditions emphasizing the love of God can be valuable for prisoners who feel rejected by their communities and isolated from society. Religious narratives emphasizing the will of God in prisoners’ lives help them find meaning in their incarceration and lack of freedom. Religious inmates also frequently accept responsibility for their crimes, and God’s forgiveness helps with the guilt associated with their prior wrongdoings (Clear et al. 2000; Dammer 2002; Gul and Asad 2018; Jang et al. 2018; Johnson and Larson 2003).

Prisoners also have disclosed that religion and religious activities can provide extrinsic benefits as well (Clear et al. 2000). For instance, Clear et al. (2000) uncovered that prisoners report being involved with religious communities and activities because they can provide protection and community for unpopular or maligned individuals in the prison culture (e.g., sex offenders), as well as opportunities for socialization. Religious services and programs give prisoners access to goods (snacks, books, etc.) and opportunities that nonreligious inmates do not have. Opportunities that religious activities afford inmates include access to outside volunteers, particularly women, and socialization opportunities (Clear et al. 2000). On the other hand, while incarceration offers unique motivations and opportunities for individual involvement in religion, the factors that predict religiosity among prisoners may not be dissimilar from predictors of religiosity among the general population.

3. Predictors of Religiosity

In terms of demographic variables, individuals who are older, who are women, and who are members of racial/ethnic minorities, particularly African Americans, may be more likely to be involved in religious activities than their counterparts (Desmond et al. 2010; Gunnoe and Moore 2002; Wallace et al. 2003). As individuals age, they are more likely to be concerned about ultimate questions, such as the meaning of life and death, and are more likely to turn to religion (Jong et al. 2018). Religiosity may also be a mechanism for individuals to cope with deteriorating health due to aging (Ewen et al. 2020). Older individuals may also turn to religion to fulfill desires for social support and integration, particularly in light of ageism and societal perceptions devaluing aging individuals (e.g., diminished productivity in a capitalist society, burden to society, etc.) (Ewen et al. 2020; Sheerkat and Ellison 1999). Studies frequently find that women are more religious than men (Beeghley et al. 1981; Zhai et al. 2007). Explanations for why women are more religious than men in the Western world emphasize gendered socialization. Women are socialized to exemplify values more congruent with religiosity, such as gentleness, submissiveness, and nurturing (Miller and Hoffmann 1995). Women’s historical positions in society are also factors that may lead them to be more involved in religiosity. For example, exclusion from the labor force provided more time and opportunity for women to be involved in religious activities, and religious involvement as a component of child-rearing indicates a concern for the well-being of children and family. Finally, scholars have argued that the emphasis on the importance of religiosity for women is another mechanism through which social control is exerted on women and girls (Miller and Hoffmann 1995). In terms of race/ethnicity, religion has been a means for African Americans to cope with societal inequality of slavery, segregation, and discrimination (Stansfield 2017). In addition, religious participation for African Americans historically provided one of the only means for social integration and community, and Black culture, particularly in the rural South, places great social pressure and expectations on individuals to attend church (Cavendish et al. 1998; Stansfield 2017).

Family structure is also an important predictor of religious involvement. Getting married and having children often leads to an increase in religiosity, again emphasizing religion as a mechanism to improve family well-being (Beeghley et al. 1981; Desmond et al. 2010; Sheerkat and Ellison 1999; Wallace et al. 2003). Measures of socioeconomic status have also been important correlates of religiosity. Individuals with greater social capital and connections to other social institutions, such as employment and education, are more likely to be involved in religion (Cavendish et al. 1998; Desmond et al. 2010; Wallace et al. 2003). In terms of contextual variables, some scholars posit that ecological hardship and environmental subsistence may promote reliance upon religion/religiosity for norm enforcement and coping mechanisms (see Botero et al. 2014). Research in the US consistently finds that levels of religiosity are consistently higher in rural areas and in the Southeastern region of the US (Gunnoe and Moore 2002; Sheerkat and Ellison 1999; Wallace et al. 2003; Zhai et al. 2007).

Consistent with predictors of religiosity in the general US population, then, factors that are associated with religiosity at the individual level among prisoners may also include age, gender, and race/ethnicity. Specifically, older inmates, female inmates, and inmates who are African American may be more likely to be involved in religious activities in prison than their counterparts. In addition, inmates who have children and are married may be more likely to be involved in religious activities. Education and employment prior to arrest and work assignments in prison may also be important predictors of religiosity among inmates. At the facility level, prisons that are located in the Southeastern region of the US and prisons that are smaller may have higher levels of religiosity among their populations.

While qualitative research has explored facets of incarceration that motivate or explain inmate involvement in religious activities (Clear et al. 2000; Dammer 2002; Johnson and Larson 2003), quantitative studies have yet to explore criminal justice measures or aspects of incarceration that may be associated with religious participation among inmates. Drawing upon research related to adjustment to prison, which includes outcomes such as misconduct and mental health among inmates, as well as findings pertaining to the data utilized in this study, a number of relevant factors may be hypothesized to influence the religious activities of prisoners (Adams 1992; Meade and Steiner 2013; Steiner and Wooldredge 2008; Steiner et al. 2014; Toch et al. 1989). Adjustment may be related to religiosity in prison, as studies have uncovered a link between religiosity and misconduct (Camp et al. 2008; Kerley et al. 2006; Meade 2014; Meade and Bolin 2018; Steiner and Wooldredge 2008; Sturgis 2010), and the research cited above also describes ways that religiosity facilitates coping with and adjustment to the experience and process of confinement (Clear et al. 2000; Dammer 2002; Johnson and Larson 2003). At the individual level, these studies have found that offense type, criminal history, a history of drug use, associating with antisocial peers, program participation, having a work assignment, and time served in prison are all important predictors of inmate adjustment/maladjustment (see Steiner et al. 2014). At the facility level, the security level of the prison, overcrowding, and the composition of the prison population (particularly the proportion of violent offenders) are important predictors of maladjustment that may have relevance for levels of inmate religiosity.

Therefore, this study sought to address the gap in the literature related to predictors of inmate religiosity. Specifically, the research questions for this study were: (1) What is the relationship between predictors of religiosity among the general population and the religiosity of prisoners? and (2) How do prisoners’ criminal justice characteristics and background exert an influence on religiosity? I examine these questions both at the individual (i.e., prisoner) level, and facility level, utilizing multilevel modeling techniques.

4. Materials and Methods

4.1. Data and Sample

The data used for this study come from the 2000 Census of State and Federal Correctional Facilities (CSCF) and the 2004 Survey of Inmates in State and Federal Correctional Facilities (SISCF). Both data sets were collected by the US Bureau of the Census on behalf of the US Bureau of Justice Statistics. The CSCF was the sixth enumeration of the census of correctional facilities and contains information on the size, composition, staff, characteristics, and operation of facilities. The data were collected by way of mailed surveys to all prisons in the US. The 2004 SISCF is a nationally representative sample of inmates in US prison facilities, and includes information on inmate demographics and characteristics, as well as criminal history, offense variables, and experiences during incarceration. The data for the SISCF was collected through a multistage cluster sample, with prisons in the US being selected at stage 1, and inmates from selected prisons being sampled at stage 2. Respondents selected at stage 2 were interviewed with computer-assisted personal interviewing (CAPI). Inmates were provided with informed consent, both in writing and verbally, assuring them that participation was voluntary, individual identification from the data would be impossible, and their identity would be kept confidential by the interviewers.

Although a newer wave of data was collected in 2016 (the 2016 Survey of Prison Inmates), the number of measures/variables collected was more limited than the 2004 SISCF, and there are no measures related to religion or religiosity in the 2016 survey. In addition, the waves of data are not technically longitudinal/panel data, as separate samples were collected in 2004 and 2016, and thus it is not possible to merge measures from the 2004 and 2016 data. Given the more limited items available in the 2016 data, scholars still utilize the 2004 SISCF due to the greater richness of measures (Daquin and Daigle 2021; Felson and Krajewski 2020; Grosholz and Semenza 2018). Since the 2000 CSCF was used as the sampling frame for the 2004 CISCF, the survey contained a measure of the population size of the facility from the 2000 data, as well as a unique facility identification number. This allowed me to match inmates to the correctional facility in which they were housed and create a multilevel data set of inmates housed in correctional facilities.

The original sample for the 2004 SISCF was comprised of 14,499 inmates housed in 283 correctional facilities. To create the sample for this study, I first excluded federal facilities and facilities that were designated as community correctional facilities, due to the potential for unmeasured differences between state and federal and community and secure confinement facilities. After deleting federal and community facilities, I deleted cases with missing data on the measures to be modeled. This left a final sample size of 12,040 inmates housed in 242 secure, confinement correctional facilities. Cases deleted due to missing data were distributed proportionally across facilities, and no single institution produced an unusual number of cases with missing data. Analysis between the final sample and full sample revealed no significant differences in means among the variables to be modeled. An analysis of tolerance and VIF values indicated that multicollinearity was not present among the individual- or facility-level predictors, and sample weights provided in the data were normalized to the reduced sample and applied to all analyses.

4.2. Dependent Variables

The dependent variables in the analyses include two measures of religiosity. The first dichotomous measure of religiosity represents prisoners’ responses to a single item asking if they had “participated in any religious activity in the past week such as religious services, private prayer or meditation, or Bible reading or study.” This measure includes only behavioral indicators of religiosity and not affective measures, such as self-rated importance of religion. On the other hand, this measure does include common dimensions of religiosity utilized in measures in previous studies, such as church attendance, prayer, and scripture study (Adamczyk et al. 2017; Bhutta et al. 2019; Evans et al. 1995; Johnson et al. 2000; Kerley et al. 2011; Stark 1996; Sumter et al. 2018). It is also important for readers to note that this measure of religiosity comes from a single-question item, so it is impossible to decompose these different dimensions of religiosity. Research indicates that multidimensional items measuring religiosity are preferred and tend to result in more robust findings pertaining to religiosity (Johnson and Larson 2003); thus, there are several limitations pertaining to this measure of religiosity. The second measure of religiosity comes from a follow-up question to the item previously described, asking respondents to report the number of hours they spent involved in those religious activities in the past week. Answers ranged from 0–99 h spent in religious activities in the past week. Due to the skewed nature of this measure, answers were top-coded at 40 h, with less than 1% of the sample reporting spending more than 40 h in the past week engaged in religious activities. Descriptive statistics for these measures, as well as all measures used in the analysis, are provided in

Table 1. Descriptive statistics.

Variable	Mean	SD
Inmate-Level Predictors
Religiosity	0.57	0.50
Religiosity (hours)	3.71	7.01
Age	35.60	10.44
Female	0.20	0.40
White	0.37	0.48
African American	0.39	0.49
Hispanic	0.17	0.38
Other race/ethnicity	0.06	0.24
Child(ren)	0.69	0.46
Married	0.17	0.37
High school diploma	0.28	0.45
Employed in month before arrest	0.69	0.46
Incarcerated for violent offense	0.51	0.50
Used drugs daily in month before arrest	0.44	0.50
Prior incarceration(s)	0.56	0.50
Delinquent peers growing up	0.57	0.50
Work assignment	0.66	0.47
Program participation	0.57	0.50
Time served (months)	53.98	65.58
N = 12,040
Facility-Level Predictors
Prop. inmates max. security	0.20	0.27
Average daily population	1817.12	1361.79
Southern region	0.46	0.50
Crowding (ADP/design capacity)	1.29	0.47
Prop. incarcerated for violent offense	0.51	0.19
N = 242

.

4.3. Individual-Level Predictors

Individual level measures in the analyses came from the SISCF. Age was measured in years, and gender was measured with a dichotomous indicator (female = 1, male = 0). Race/ethnicity was measured with a series of dichotomous variables (African American, Hispanic, other race/ethnicity), with white individuals serving as the reference category. Child(ren), marriage, education, and employment were also measured with dichotomous indicators. Respondents were asked whether they were married (1 = yes, 0 = no), had children (1 = yes, 0 = no), had a high school diploma (1 = yes, 0 = no), or were employed or owned a business in the month before their arrest (1 = yes, 0 = no). Another series of dichotomous indicators represented respondents’ criminal justice histories and prison experiences, including whether they were incarcerated for a violent offense (1 = yes, 0 = no), used drugs in the month before their arrest (1 = yes, 0 = no), had been previously incarcerated (1 = yes, 0 = no), associated with delinquent peers when growing up (1 = yes, 0 = no), had an institutional work assignment (1 = yes, 0 = no), or participated in prison programming since their incarceration (1 = yes, 0 = no). Finally, time served was measured in months. Due to the skewed nature of the distribution of time served, the natural logarithm of this measure was modeled in all analyses.

4.4. Facility-Level Predictors

At the facility level, the security level of facilities was measured with the proportion of inmates in each facility housed in maximum security confinement. The size of the facility was modeled as the average daily population (ADP). A dichotomous measure of whether the facility was located in the Southeastern region of the US was included. The measure of crowding represents the average daily population (ADP) divided by the design capacity of the facility. Measures of more than 1 indicate a facility was overcrowded. Finally, a measure of the proportion of the facility population incarcerated for a violent offense was included. All of the facility-level measures came from the 2000 CSCF, with the exception of the proportion of the population incarcerated for a violent offense, which was an aggregation to the facility level of the item from the 2004 SISCF of whether inmates were incarcerated for a violent offense.

4.5. Analytical Plan

Multilevel modeling was used to estimate the impact of individual- and facility-level covariates on religiosity. Using the software package HLM 7.03, I used hierarchical generalized linear modeling (HGLM) to estimate the results of the individual- and facility-level covariates in the same regression model (Raudenbush and Bryk 2002). HGLM is a modeling technique that addresses statistical problems of including covariates from different levels of analysis into a single-level regression model. Specifically, HGLM addresses problems of correlated error and heteroskedasticity that can occur when pooling different levels of covariates into a single-level regression model. In addition, HGLM bases null hypothesis testing at level 2 (prison level) on the appropriate number of level 2 observations (versus level 1 Ns in single-level regression models). Finally, HGLM can control for unmeasured variation across level 2 units by centering level 1 predictors around the group mean (versus grand mean). Due to potential influence of unmeasured variation across level 2 units, I centered all individual-level covariates (with the exception of female) around the group mean. I used Bernoulli regression for the dichotomous religiosity outcome and Poisson models for the hours of religious activity in the past week. Poisson regression is appropriate for count outcomes, and I corrected these models for overdispersion (see descriptive statistics). The process for HLGM involves first estimating unconditional models for each outcome. Following this, level 1 fixed effects models with level 1 covariates were estimated. Finally, intercepts as outcomes models were estimated for each outcome to examine the impact of facility-level predictors on the religiosity of prisoners.

5. Results

An analysis of descriptive statistics indicated that 57 percent of the sample reported engaging in religious activities in the past week. Of those who reported engaging in religious activities, individuals reported spending an average of 3.71 h engaged in religious activities. The results of the multilevel, multivariate models are depicted in

Table 2. Inmate-level predictors of religiosity.

Variable	Religious Activity Past Week		Hours of Religious Activity Past Week
	b(se)	Exp(b)	b(se)	Exp(b)
Intercept	0.29 **	1.34	1.25 **	3.51
	(0.03)		(0.03)
Age	0.03 **	1.03	0.02 **	1.02
	(0.01)		(0.01)
Female	0.65 **	1.91	0.29 **	1.33
	(0.08)		(0.06)
African American	0.56 **	1.75	0.30 **	1.35
	(0.05)		(0.05)
Hispanic	0.35 **	1.42	−0.06	0.95
	(0.07)		(0.06)
Other race/ethnicity	0.56 **	1.75	0.22 *	1.25
	(0.04)		(0.08)
Child(ren)	0.06	1.06	0.08	1.08
	(0.04)		(0.04)
Married	0.25 **	1.28	0.12 *	1.13
	(0.05)		(0.04)
High school diploma	0.28 **	1.33	0.19 **	1.21
	(0.05)		(0.04)
Employed before arrest	0.26 **	1.30	0.14 **	1.15
	(0.05)		(0.04)
Incarcerated for a violent offense	0.13 *	1.14	0.20 **	1.22
	(0.05)		(0.04)
Used drugs daily in month before arrest	−0.04	0.96	0.02	1.02
	(0.05)		(0.04)
Prior incarceration	−0.16 **	0.85	−0.09	0.92
	(0.04)		(0.04)
Delinquent peers (growing up)	0.04	1.04	0.07	1.07
	(0.05)		(0.04)
Work assignment	0.13 *	1.14	0.05	1.05
	(0.05)		(0.05)
Program participation	0.43 **	1.53	0.20 **	1.22
	(0.05)		(0.04)
Time served in months (ln)	−0.11 **	0.90	−0.05 **	0.95
	(0.02)		(0.02)
N = 12,040
χ2	659.57		570.30

Notes: Maximum likelihood coefficients reported (with standard errors in parentheses), ** p ≤ 0.001, * p ≤ 0.01.

(individual level findings) and

Table 3. Facility level predictors of religiosity.

Variable	Religious Activity Past Week		Hours of Religious Activity Past Week
	b(se)	Exp(b)	b(se)	Exp(b)
Intercept	0.06	1.07	1.00 ***	2.72
	(0.13)		(0.11)
Prop. max security	−0.37 **	0.69	−0.11	0.90
	(0.12)		(0.10)
Average daily population	0.00001	1.00	0.00002	1.00
	(0.0002)		(0.00002)
Southern region	0.36 ***	1.43	0.26 ***	1.30
	(0.06)		(0.05)
Crowding	−0.06	0.94	−0.06	0.94
	(0.07)		(0.06)
Prop. Incarcerated for violent offense	0.41 *	1.50	0.39 *	1.47
	(0.16)		(0.16)
N = 242
χ2	527.35		464.95

Notes: Maximum likelihood coefficients reported (with standard errors in parentheses). *** p ≤ 0.001, ** p ≤ 0.01, * p ≤ 0.05.

(facility-level findings), and findings are interpreted and presented based on odds ratios for the Bernoulli models and event rate ratios for Poisson models. Based on the results presented in Table 2, age and gender (female) were significantly and positively associated with religiosity. Specifically, older inmates were more likely than younger inmates to report engaging in religious activities [exp(b) = 1.03], and older inmates reported significantly more time spent in religious activities [exp(b) = 1.02]. Similarly, women were significantly more likely than men to engage in religious activities [exp(b) = 1.91], as well as spend more time in religious activities [exp(b) = 1.33].

Race/ethnicity was also significantly associated with religiosity, with white inmates reporting less religiosity compared to African Americans, Hispanics, and other races/ethnicities. Specifically, African Americans reported more religious activities [exp(b) = 1.75] and more hours engaged in religious activities than white inmates [exp(b) = 1.35]. Hispanic inmates were significantly more likely than white inmates to report engaging in religious activities [exp(b) = 1.42], but the findings indicated no statistically significant differences in hours spent in religious activities between Hispanic and white inmates. Finally, inmates of other races/ethnicities were more likely than white inmates to report engaging in religious activities [exp(b) = 1.75] and spending more hours in religious activities [exp(b) = 1.25].

Turning to the impact of family structure and SES on religiosity, the models indicate that having children was not significantly associated with either measure of religiosity among respondents, but inmates who were married were significantly more likely to engage in religious activities [exp(b) = 1.28] and spend more time in religious activities than inmates who were not married [exp(b) = 1.13]. Education and employment were significantly associated with greater levels of religiosity. Having a high school diploma was significantly associated with both religious activities [exp(b) = 1.33] and hours spent in religious activities [exp(b) = 1.21]. Being employed before arrest was also associated with greater odds of engaging in religious activities [exp(b) = 1.30] and spending more hours in religious activities [exp(b) = 1.15].

An examination of the results of the criminal history measures indicates that offense type was a significant predictor of religious activities. Individuals who were incarcerated for a violent offense were significantly more likely than their counterparts to engage in religious activities [exp(b) = 1.14], as well as spend more time in religious activities [exp(b) = 1.22]. The measures of drug use prior to arrest and association with antisocial peers growing up were not significantly associated with either religiosity measure. Those respondents who had been previously incarcerated were less likely to report engaging in religious activities than those inmates who were incarcerated for the first time [exp(b) = 0.85], but a prior incarceration was not associated with hours spent in religious activities.

Inmates with a work assignment were more likely to engage in religious activities than inmates without a work assignment [exp(b) = 1.14], but having a work assignment was not significantly associated with the number of hours spent in religious activities. Inmates who participated in programs were more likely to engage in religious activities [exp(b) = 1.53] and spend a greater number of hours in religious activities [exp(b) = 1.22] than inmates who were not involved in prison programs. Finally, at the individual level, time served was inversely associated with both measures of religiosity. Serving more time in prison was associated with a reduction in the odds of engaging in religious activities [exp(b) = 0.90], as well as a reduction in the number of hours spent engaging in religious activities [exp(b) = 0.95].

The facility-level results are depicted in Table 3. Facilities with a greater proportion of maximum security inmates were associated with lower average odds of inmates engaged in religious activities [exp(b) = 0.69], but the proportion of maximum security inmates was not associated with the average hours spent in religious activities within facilities. The size of the population of prisons was not associated with either measure of religiosity. Similarly, overcrowding was not a predictor of engagement or time spent in religious activities either. Facilities that were located in the Southeastern region of the US had higher average odds of engaging in religious activities [exp(b) = 1.43], as well as a greater number of average hours spent engaged in religious activities. Finally, the proportion of the population incarcerated for violent offenses was positively associated with both measures of religiosity [religious activities exp(b) = 1.50; hours of religious activity exp(b) = 1.47].

6. Discussion and Conclusions

In this study, I examined the impact of a number of covariates on the religiosity of prisoners. Utilizing multilevel analyses, I modeled variables related to demographics, family structure, socioeconomic status, criminal history, and incarceration experiences on the odds of engaging in religious activities in the past week, as well as the number of hours spent engaged in religious activities in the past week. For those measures that were statistically significant, nearly all were associated with both outcome measures of religiosity, with only a few exceptions. Most notably, no measures that were significantly, inversely related to the odds of engagement in religious activities were associated with the hours engaged in religious activity (prior incarceration at the individual and proportion of maximum security at the facility level). On the other hand, the measures of Hispanic and work assignment were positively associated with engagement in religious activities, but were not significantly related to the hours spent in religious activities. Further discussion of these measures is presented below.

Consistent with research on religiosity among the general population, age, gender, race/ethnicity, marriage, education, and employment were significantly related to greater levels of religiosity, both in terms of the odds of engaging in religious activities and the hours spent in religious activities. Similar to the general population, levels of religiosity were higher among older inmates and women. This might suggest that the same processes at work related to aging and gender and religiosity among the population take place in prisons as well (Jong et al. 2018; Miller and Hoffmann 1995). Recall that scholars have speculated that aging results in greater religiosity as individuals begin to ponder the end and significance of their lives and gendered socialization places a greater emphasis on religiosity for girls and women (Beeghley et al. 1981; Ewen et al. 2020; Sheerkat and Ellison 1999; Zhai et al. 2007). In addition to these perspectives, religiosity may be more salient for these populations as coping mechanisms upon incarceration (see Clear et al. 2000; Dammer 2002).

The results related to the race/ethnicity measures demonstrated that non-white individuals have higher levels of religiosity than white prisoners. This is also consistent with research among the general population. Much of the literature discusses religiosity among African Americans, but the results of this study found that levels of religiosity were equally as high compared to white inmates among inmates of other races/ethnicities. In the US, Christianity has been the focus of the study of black American’s religiosity and the ways the Christian church has been integral to Black culture (Cavendish et al. 1998). In contrast to much of Christianity, Judaism and Islam require daily adherence to rituals and routines (diets, prayers, etc.) that may make religiosity and engagement in religious activities in prison more salient for non-white non-Christians (Bhutta et al. 2019; Bhutta and Wormith 2016; Haviv et al. 2020). Non–Judeo-Christian religious practices, including Buddhism, Native American spiritual practices, or other forms of spiritual/mindfulness practices, may also be salient cultural and coping mechanisms for individuals of other races/ethnicities (Auty et al. 2017). Unfortunately, the SISCF did not collect information on religious affiliation from participants, so this is mere speculation. Finally, Hispanic inmates were more likely to engage in religious activities than white inmates, but there were no significant differences in the hours spent in religious activities between Hispanic and white inmates.

In terms of the family structure and SES measures, all were associated with religiosity, with the exception of children. Individuals with children in the general population may use participation in religious activities/services as a way to increase family bonds and enhance family/child well-being. Incarceration separates parents from children, so incarcerated parents may be less likely to rely upon engagement with religion to enhance family well-being when they cannot participate in the same religious activities with their children. On the other hand, being married, having a high school diploma, and being employed before arrest were all related to increased levels of religiosity. Just as in the general population, those individuals in prison with more social capital and connection to social institutions may be more likely to engage in religious activities and programs in prison (Cavendish et al. 1998; Desmond et al. 2010; Wallace et al. 2003).

In sum, the findings related to the demographic, family structure, and SES measures of prisoners are reflective of the processes and findings related to religiosity in the general population. Some correctional scholars who have studied adjustment to prison have argued that individuals import their antisocial values and behavior into prison, which in turn affects adjustment and violence in prison (Irwin and Cressey 1962; Steiner et al. 2014). Other scholars have extended importation theory to examine the impact of demographic and socioeconomic factors prior to prison on violence and misconduct (Harer and Steffensmeier 1996; Irwin and Cressey 1962; Jiang and Fisher-Giorlando 2002; Wooldredge 1991). In a similar vein, prisoners may import their prosocial characteristics into prison. Thus, religious engagement and cultural emphases on religiosity prior to prison would also impact religiosity while incarcerated and reflect the same processes that motivate religious engagement in community.

The results of the criminal history and incarceration experience variables were not as great in magnitude (see odds/event rate ratios in Table 2 and Table 3) as the demographic, family structure, and SES measures. Nonetheless, this study revealed important findings related to criminal history and imprisonment. One of the more surprising findings was that being incarcerated for a violent offense was associated with greater levels of religiosity, both in terms of the odds of engaging in religious activities and the hours spent in religious activities. One might expect that individuals with a history of violence may be less likely to be religious. On the other hand, the literature on religion in prison indicates that religiosity among inmates is consistently related to lower odds of serious and violent misconduct across studies (Camp et al. 2008; Meade and Bolin 2018; Sturgis 2010). From the perspective of spiritual transformations and cognitive shifts, it could be the case that individuals with violent pasts who experience a religious conversion find particular value in religion as a means to seek redemption for their past crimes (Jang et al. 2018). Jang et al. (2018) found, for instance, that inmates who experienced a religious conversion in prison were more likely to recognize the harm of their past conduct and express a desire to change. Religiosity provides a framework for individuals to reject their previous violent behavior and selves. In addition, it allows them to seek forgiveness and provides a script for a new way of life that rejects the violent values and behaviors that are emphasized in the prison subculture (Gul and Asad 2018; Jang et al. 2018, 2020; Johnson 2011; Maruna 2001).

Individuals who had been previously incarcerated were less likely to engage in religious activities than individuals who were incarcerated for the first time. This finding could represent the value of religion for coping with imprisonment for the first time, in contrast to inmates who had been to prison previously and were familiar with the incarceration experience and inmate subculture (Clear et al. 2000; Dammer 2002). In light of the findings pertaining to religiosity and recidivism, which suggest that religiosity and faith-based programming are not strongly associated with successful reentry (Johnson 2004; Johnson et al. 1997; Johnson and Larson 2003; Stansfield et al. 2017), individuals who are returning to prison may be less likely to turn to religiosity if it had not been successful in preventing their return to prison the first time. On a similar note, time served in prison was also significantly inversely related to religiosity. That is, individuals who had been in prison for longer periods of time were less likely to engage in religious activities and spend fewer hours in religious activities than individuals who had spent shorter amounts of time in prison. Again, this may reflect the value of religiosity for initial coping with imprisonment, similar to the explanations of the finding of prior incarcerations on religiosity.

Finally, at the individual level, having a work assignment and participating in prison programs were significantly and positively associated with religiosity. Inmates with a work assignment were more likely to engage in religious activities than inmates without a work assignment, but having a work assignment was not associated with hours spent in religious activities. From an opportunity perspective, it seems likely that religious inmates with a work assignment would have less time to engage in religious activities than religious inmates without a work assignment. Nonetheless, work assignments and program participation may function in prison similarly to work and education outside of prison. Namely, they provide social capital and connections to prosocial individuals and institutions, and thus are more predictive of greater levels of religiosity (Cavendish et al. 1998; Desmond et al. 2010; Jang et al. 2020; Wallace et al. 2003).

At the facility level, neither the population size of the prison nor the measure of overcrowding was associated with levels of religiosity across prisons. These measures were grounded in the literature indicating that levels of religiosity were higher in more rural and less densely populated areas (Gunnoe and Moore 2002; Sheerkat and Ellison 1999; Wallace et al. 2003; Zhai et al. 2007). Based on these findings, then, the size of prison facilities does not represent the same social processes at work when considering the association between rurality and religiosity (Botero et al. 2014). Religious participation in rural areas provides opportunities for social networking and integration when other opportunities may be sparse (Chalfant and Heller 1991; Lee 2006; Stark 1996; Stark et al. 1982). Prison facilities, regardless of absolute size, confine a relatively large number of people in constant close quarters, so the need for social connection to others that exists in rural areas would be lacking in prison. In fact, individuals in prison may be more likely to seek opportunities for privacy and separation from others (e.g., the association of crowding on facility-level religiosity was negative, although not statistically significant).

The proportion of individuals incarcerated for a violent offense was positively associated with levels of religiosity across facilities, including the odds of engaging in religious activities, as well as the hours spent in religious activities. This could be simply an aggregation of the individual-level processes related to the relationship between histories of violent offending and religiosity discussed previously. The measure of security level of the facility (proportion of inmates housed in maximum security housing) was negatively associated with the average odds of engaging in religious activities, but was not significantly related to the hours spent in religious activities. Given that part of the measure of religiosity included attendance at services and/or Bible study, inmates housed in maximum security have fewer opportunities to engage in religious activities. More restrictive housing assignments/rules would by definition limit the activities in which inmates can participate, as well as religious items that inmates may possess (books, clothing, and other ceremonial items/rituals). Although this is speculation, the influence on religiosity of the measure of security level most likely represents correctional constraint and more limited opportunities to engage in religious activities, rather than a higher-risk population held in higher-security facilities, particularly in light of the findings related to a history of violence and the proportion of a facility’s population comprised of violent offenders.

The most salient finding (from the magnitude of the odds and event rate ratios) at the facility level was the location of the prison in the Southeastern US. Research has extensively documented the emphasis in Southern culture on religiosity, so it not surprising that prisons in the South had higher levels of religiosity than other regions of the US (Chalfant and Heller 1991; Ellison et al. 2003; Lee 2006). Again, this most likely represents the importation of cultural values and practices into prisons (Harer and Steffensmeier 1996; Irwin and Cressey 1962).

This study contributes to the scholarship on religiosity in prison. Most research has treated religiosity as an independent variable and has examined the impact of religiosity particularly on behavioral outcomes. I add to the literature by examining those factors, both before and during incarceration that are statistically associated with religiosity. Understanding the forces that shape religiosity for those who are incarcerated may help shed light on findings pertaining to the impact of religiosity on such factors as misconduct, adjustment, reentry, and recidivism. This study is not, however, without limitations. Firstly, the data used for this study were collected in 2000 and 2004. Although the data may be dated and questions may arise about generalizability, the subsequent wave(s) of the data sets do not contain measures related to religious activities or practices of respondents. Future data collections on a national (or smaller scale) should include questions about religiosity and should continue to examine those factors that are associated with religiosity among prisoners. The other major limitation related to this study regards the measure of religiosity. While the measure included in this study does ask about dimensions of religiosity (service attendance, scripture reading/study, prayer) that have been found to be important aspects of the measurement of religiosity (see Adamczyk et al. 2017; Sumter et al. 2018), these elements of religiosity were asked in a single item and could not be decomposed. In addition, affective measures of religiosity, such as the importance of religiosity to one’s daily life or decision-making were not a component of the data or questions asked. It would also have been valuable to have a measure of the religious affiliations of individual respondents (Bhutta et al. 2019; Bhutta and Wormith 2016; Haviv et al. 2020), but these data were not collected either. Religion is an important aspect of the lives of individuals and prisoners, and researchers collecting data in prison should take care to craft valid and reliable items to measure the religiosity of prisoners. This study is a first step in examining an overlooked aspect of the religious activities and practices of prisoners in the US, and a better understanding of religiosity among prisoners is important for prisoner well-being, coping, and adjustment to prison, and has implications for the safety and administration of prisons.