Assessing the Relative Influence of Individual Attitudes,Social Supports,and Neighborhood Context on Reentry Outcomes: What Changes Matter Most?

Abstract

Using the Serious and Violent Offender Reentry Initiative (SVORI) data set, the current study examined the relationship between internal change factors, including agency, readiness for change, and legal cynicism, and four reentry outcomes (recidivism, reincarceration, drug use, and treatment participation). The study also assessed the impact of external change factors, such as family support and perceived neighborhood quality on reentry outcomes. Using a hybrid model approach, we found that within-individual changes in agency and family emotional support were significantly related to decreased self-reported recidivism and reincarceration over time. Within-individual changes in family emotional support were also significantly related to reductions in self-reported drug use and within-individual changes in readiness for change were positively associated with treatment participation across the postrelease waves of the SVORI data. Between-individual results further demonstrate the salience of change factors on reentry outcomes. Study findings inform desistance models and identify some promising targets for reentry programming.

Keywords

reentry social support recidivism substance use treatment agency

Introduction

Beginning in the 1970s, the United States embarked upon a grand social experiment by replacing decades-long indeterminate sentencing practices with “tough on crime” laws, resulting in the use of imprisonment as the primary response to crime control (Clear & Frost, 2014; Garland, 2001). A consequence of using incarceration as the main crime control strategy was an exponential increase in the U.S. prison population (Travis et al., 2014). With mass incarceration dominating both academic and political discourse over the last several decades (Clear, 2009; Travis et al., 2014; Wakefield & Wildeman, 2013), this overreliance on incarceration created a crisis of mass reentry. Since 2002, more than 600,000 individuals have been released to the community from state and federal incarceration annually (Carson & Golinelli, 2013).

Despite federal, state, and local efforts to improve reentry processes and outcomes, contemporary research continues to document exceedingly high recidivism rates. In a 9-year follow-up of state justice-involved individuals released in 2005, five in six (83%) formerly incarcerated individuals were rearrested at least once (Alper et al., 2018). As a result, the successful reintegration of formerly incarcerated individuals remains a profound policy challenge for America’s criminal justice system (Petersilia, 2003; Visher & Travis, 2003). Given high recidivism rates and increased spending on reentry programming, a need for continued research examining factors associated with successful reentry outcomes remains pressing.

A growing body of empirical research has begun to identify predictors of reentry outcomes and several theoretical models have been developed to aid in understanding individual change and complex reentry processes (Lloyd & Serin, 2012; Serin et al., 2010; Serin & Lloyd, 2009; Ward et al., 2004). Most extant models of change during reentry (e.g., the Multifactor Offender Readiness Model [MORM], and Integrated Model of Transition to Crime Desistance [IMTCD]) note the importance of both internal and external change factors; however, research has yet to fully establish the empirical validity of these models or integrate them within the broader literature on desistance. The current study explores the influence of internal change factors (agency, readiness for change, and legal cynicism) and external change factors (neighborhood quality and family support) on four reentry outcomes (recidivism, reincarceration, drug use, and treatment participation). This study examines whether within-individual changes in these variables are related to changes in reentry outcomes over time. The goal is to provide a partial test of untested reentry change models to inform programming and policy decisions. Identifying salient predictors of change during reentry and integrating these findings into theoretical frameworks of change and/or desistance is an important step toward building an empirical foundation that can adequately inform reentry policy and practice.

Literature Review

Models of Change

Existing conceptual frameworks grounded within the desistance and treatment literatures focus on how and why people change. Two models applicable to the current study, the MORM (Ward et al., 2004) and the IMTCD (Serin et al., 2010; Serin & Lloyd, 2009), supplement prior conceptualizations of the intricate and complex nature of desistance and reentry processes. Ward and colleagues’ (2004) MORM model views readiness to change as a function of both internal and external factors that interact. According to Ward and colleagues (2004), behavior change occurs when individuals possess certain cognitive/affective behavior properties in combination with external environmental support. Specific internal readiness conditions identified include cognitive factors (e.g., change beliefs, attitudes about treatment, self-efficacy), affective factors (e.g., emotional regulation, guilt, shame), volitional factors (e.g., goals, agency), behavioral factors (e.g., problem recognition, help-seeking, competencies), and identity factors (e.g., ability to recognize a prosocial identity). These internal readiness factors are predicted to interact with external readiness factors that include circumstances, location, opportunities, resources, interpersonal supports, and program characteristics (Ward et al., 2004). To date, this model has not been empirically validated (Mossière & Serin, 2014).

Serin and colleagues’ (2010; Serin & Lloyd, 2009) IMTCD depicts individual change as an interaction between internal and external factors known to influence desistance and reentry outcomes. This model presents a life-course perspective of desistance that recognizes the importance of commitment to change occurring at the start of the desistance process. Whether or not this commitment to change results in sustainable change depends on internal (i.e., agency/self-efficacy, attributions, outcome expectancies, identity, and change beliefs) and external (i.e., correctional interventions, proactive supervision, aftercare, positive relationships, and supportive community) change factors that must be present to sustain the intention to change (Serin et al., 2010). This model has not been subjected to rigorous empirical testing; although, an empirical base for the model is presented by Serin and colleagues (2010; Serin & Lloyd, 2009).

A lack of conceptual clarity and standardized measurement has limited the applicability of change models (Mossière & Serin, 2014). Consequently, empirical research testing the validity of individual change models remains limited (Casey et al., 2005; Day et al., 2010; Lloyd & Serin, 2012). Only recently have scholars attempted to empirically validate these models of change in justice-involved samples (Casey et al., 2005; Day et al., 2010). Despite the emergence of this research, many questions remain unanswered regarding the applicability and empirical validity of change models within the reentry context.

Internal Change Factors

Existing theoretical frameworks highlight the importance of internal and external change factors, which can be applied to understanding individual change during the reentry process. Although not explicitly defined, internal change factors have been referred to as “psychological and behavioral dispositions and states” (Ward et al., 2004, p. 650) or internal experiences, including “attitudes, beliefs, personality, and history” (Serin et al., 2010, p. 64), associated with changes in offending behaviors. The current study highlights three internal change factors empirically linked to desistance—human agency, legal cynicism, and readiness for change.

Human Agency

A considerable body of research implicates human agency in the desistance process (Bottoms et al., 2004; Laub & Sampson, 2001; Paternoster & Bushway, 2009; Paternoster et al., 2015). Human agency refers to the belief that an individual is capable of exerting influence upon themselves and their environment (France & Homel, 2006). Self-efficacy, defined as the belief in one’s self to perform well or produce certain effects (Bandura, 1997), represents a subcomponent of human agency also posited as a correlate of desistance. However, the nature of the relationship between agency and desistance outcomes is complex as agency is likely to interact with environmental factors to either maintain or change offending behaviors (Lloyd & Serin, 2012; Paternoster et al., 2015). For instance, prior research reports a strong association between justice-involved individuals’ self-efficacy and antisocial behaviors (Bandura et al., 1977; Hagger et al., 2002), and several studies find self-efficacy beliefs are positively related to the frequency of alcohol consumption (Maisto et al., 2000). In contrast, other research identifies a positive relationship between self-efficacy beliefs and motivation to change (McMurran et al., 1998) and a negative association between self-efficacy and recidivism (Benda, 2005; Maruna, 2001). These mixed findings indicate the need for further testing and theoretical refinement.

Legal Cynicism

A positive association between antisocial attitudes and recidivism has been well documented in prior research (Andrews et al., 1990; Bonta & Andrews, 2016; Walters, 2012) and desistance models identify antisocial attitudes as targets for change (Serin et al., 2010). One antisocial attitude that may fit under the umbrella of internal change factors is legal cynicism. Legal cynicism has been defined as skepticism toward the law and legal actors who are tasked with enforcing it (Sampson & Bartusch, 1998). For example, individuals socialized to hold negative views toward people in authoritative roles often view the rule of law as illegitimate (Kirk & Papachristos, 2011), and the justice system as reflecting the interests of those in power (Tyler & Huo, 2002). Addressing skepticism toward the law or perceptions of legal cynicism represents a critical component involved in the desistance and behavioral change process (Visher et al., 2004).

Readiness for Change

Readiness for change is conceptualized as a broad construct encompassing several specific factors (e.g., motivation, treatment readiness, treatment responsivity). Although readiness for change is often discussed in the desistance literature, a clear conceptual definition of this construct has yet to be agreed upon (Day et al., 2010; Mossière & Serin, 2014). Treatment readiness, a subcomponent of readiness for change, has been more thoroughly defined and studied. Serin and Kennedy (1997) conceptualized treatment readiness as a justice-involved persons’ level of motivation and their personal investment to engage or buy-in to therapeutic programming. Treatment readiness has been conceptually defined as “the presence of characteristics (states or dispositions) within either the client or the therapeutic situation, which are likely to promote engagement in therapy and that, thereby, are likely to enhance therapeutic change” (Ward et al., 2004, p. 647).

Previous research emphasizes individuals’ motivation for change in predicting substance use and other treatment engagement outcomes (Brocato & Wagner, 2008; Longshore & Teruya, 2006; Sloas et al., 2018). Treatment readiness has been positively associated with treatment engagement (Hiller et al., 2002), treatment retention (Longshore & Teruya, 2006), and reductions in drug use and driving under the influence of alcohol or drugs (DUI; Duvall et al., 2008). However, studies examining the effects of readiness for change on recidivism outcomes remain limited with most of the literature focused on substance abuse treatment engagement (Sloas et al., 2018).

External Change Factors

In addition to internal change factors, conceptual models of change emphasize the importance of external change factors on successful reentry (Serin et al., 2010) and treatment outcomes (Ward et al., 2004). External change factors refer to contextual (e.g., social supports, prosocial relationships) or therapeutic (e.g., program quality, program responsivity) factors that facilitate individual change (Day et al., 2010; Ward et al., 2004). Ward and colleagues (2004) specifically identified circumstances, location, opportunities, resources, support, and program/timing factors as important external change factors. Similarly, Serin and colleagues (2010) highlighted correctional interventions, proactive supervision, aftercare, positive relationships, and a supportive community as important external factors related to behavior change. Although a growing literature supports the need to consider these external change factors, the empirical status of these factors remains somewhat ambiguous (Serin et al., 2010).

Family Support

Several studies have documented a positive relationship between family emotional and instrumental support and successful reintegration (Barrick et al., 2014; Mowen et al., 2019; Taylor, 2016; Wallace et al., 2016). For example, Barrick and colleagues’ (2014) study of the effect of social ties on recently released females found family emotional support served as a protective factor decreasing the likelihood of recidivism. Taylor (2016) examined the direct effects of family emotional and instrumental support on formerly incarcerated individuals’ rearrest and reincarceration across multiple waves of the Serious and Violent Offender Reentry Initiative (SVORI) data. Results indicated that greater levels of emotional support were significantly associated with reductions in reoffending. Recently, Mowen and colleagues’ (2019) disaggregated different types of familial support and found only within-individual changes in instrumental family support (e.g., family assistance with housing, employment, transportation), not interactional or emotional support (e.g., feel close to family), were protective against recidivism and drug use across the postrelease SVORI waves. Taken together, existing research establishes the relevance of family support as a correlate of reentry outcomes; however, more research is needed to better understand the nuances of this relationship and specific mechanisms through which family support affects reentry outcomes (Mowen et al., 2019).

Neighborhood Quality

Securing affordable housing, employment opportunities, access to treatment services, and transportation are some of many obstacles returning individuals face upon release from incarceration (Rose & Clear, 2004; Travis, 2005; Visher et al., 2017). Such barriers can be compounded when returning to already unstable and socially disadvantaged neighborhoods, further increasing the likelihood of reoffending (Chamberlain & Wallace, 2016; Clear et al., 2003; Wallace & Papachristos, 2014). Research acknowledges that returning to socially disorganized communities with already limited access to opportunities and resources exacerbates reentry barriers and further reduces access to human and social capital (Clear et al., 2001; Rose & Clear, 2004). Consequently, formerly incarcerated individuals released to disorganized neighborhoods are potentially at a greater risk of engaging in antisocial behaviors and experiencing further criminal justice system involvement. Chamberlain and Wallace’s (2016) examination of mass reentry and neighborhood context revealed that concentration or clustering of parolees in neighborhoods increased individual-level recidivism. Results also indicated that parolees released to stable and less socially disadvantaged neighborhoods were less likely to reoffend (Chamberlain & Wallace, 2016). Although somewhat limited in scope, this theoretical and empirical evidence suggests the important role neighborhood context plays in facilitating or inhibiting successful reentry outcomes.

Current Study

The current study examined the relationship between internal change factors (human agency, legal cynicism, and readiness for change), external change factors (family emotional support, family instrumental support, and neighborhood quality), and key reentry outcomes, including recidivism, reincarceration, drug use, and treatment participation. The goal of this study was to examine within-individual changes in internal and external change factors and assess their relative salience for predicting reentry outcomes. Consistent with the MORM and IMTCD models, we hypothesized that within-individual changes in agency, legal cynicism, and readiness to change would be significantly related to changes in recidivism, reincarceration, drug use, and treatment participation. We predicted that changes in agency and readiness for change would be negatively related to recidivism, reincarceration, and drug use while changes in legal cynicism would be positively associated with these outcomes. The opposite effects were expected for each variable when predicting treatment participation. Regarding external change factors, we predicted that increases in family emotional support, family instrumental support, and neighborhood quality would be protective against recidivism, reincarceration, and drug use and positively related to treatment participation. These hypotheses are consistent with behavior change models that stress the importance of both internal and external correlates of change.

Method

Data

Study data were drawn from the SVORI data set (Lattimore & Visher, 2009). Data for this project were collected across 14 states between 2005 and 2007 to assess the impact of reentry programming on reentry outcomes, including employment, housing, recidivism, and desistance. A total of four waves of data were collected. Wave 1 data were collected approximately 30 days prior to respondents’ scheduled release from incarceration. Wave 2 data were collected approximately 3 months postrelease; Wave 3 approximately 9 months postrelease; and Wave 4 data were collected roughly 15 months postincarceration. Although juveniles (n = 337) were included in the original SVORI data, the current study only examines adult male and female incarcerated individuals (N = 2,054).

Given the study focus on the time-varying effects of internal and external change factors on postrelease outcomes, our analyses rely primarily on variables measured across Waves 2, 3, and 4 of the SVORI study. Our final analytic sample includes 1,148 individuals (55.9% of SVORI adults) who had at least two waves of complete data on our independent and dependent variables. At least two waves of data were needed to assess the effects of within-individual change on postrelease outcomes. The analysis sample is predominantly male (79%) with a mean age of 30 years (SD = 7.23). In terms of race and ethnicity, 36.1% were White, 51.0% were Black, 5.3% were Hispanic, and 7.6% self-identified as other non-Hispanic. The modal instant offense category was a violent offense (42.1%), followed by drug offense (25.0%), property offense (24.0%), and other conviction offense (8.9%). The mean number of total prior convictions was 4.97 (SD = 4.84). Finally, just over half of respondents were SVORI program participants (52.7%). Descriptive statistics for all time-varying and time-stable covariates are presented in Table 1.

Table 1:

Univariate Descriptive Statistics of Time-Variant and Time-Invariant Covariates Across Waves 2, 3, and 4

Variables	Wave 2			Wave 3			Wave 4
Variables	M	SD	Range	M	SD	Range	M	SD	Range
Time-variant measures
Recidivism	0.20	0.40	0−1	0.35	0.47	0−1	0.36	0.48	0−1
Reincarceration	0.05	0.21	0−1	0.17	0.37	0−1	0.16	0.37	0−1
Drug use	0.24	0.42	0−1	0.42	0.49	0−1	0.42	0.49	0−1
Treatment participation	0.48	0.49	0−1	0.46	0.49	0−1	0.39	0.48	0−1
Agency	2.91	0.50	1.3−4	2.87	0.54	1−4	2.89	0.52	1−4
Legal cynicism	2.03	0.54	1−4	2.06	0.54	1−4	2.04	0.55	1−3.8
Readiness for change	3.50	0.46	1−4	3.44	0.45	2−4	3.43	0.47	1.5−4
Family instrumental support	3.29	0.61	1−4	3.21	0.63	1−4	3.21	0.60	1−4
Family emotional support	3.25	0.47	1−4	3.18	0.50	1.2−4	3.19	0.48	1.5−4
Neighborhood quality	2.45	0.86	1−4	2.05	0.66	1−4	2.06	0.66	1−4
Age	30.14	7.23	18−69	30.67	7.20	19−70	31.37	7.27	19−70
Employment	0.74	0.43	0−1	0.81	0.38	0−1	0.78	0.40	0−1
High school/GED	0.65	0.47	0−1	0.68	0.46	0−1	0.70	0.45	0−1
Time-invariant measures
White	0.36	0.48	0−1	—	—	—	—	—	—
Black	0.51	0.50	0−1	—	—	—	—	—	—
Hispanic	0.05	0.22	0−1	—	—	—	—	—	—
Other/multiracial	0.07	0.26	0−1	—	—	—	—	—	—
Violent	0.42	0.49	0−1	—	—	—	—	—	—
Property	0.24	0.42	0−1	—	—	—	—	—	—
Drug	0.25	0.43	0−1	—	—	—	—	—	—
Other offense	0.08	0.28	0−1	—	—	—	—	—	—
Days incarcerated	6.47	0.89	2.1−9.15	—	—	—	—	—	—
Total convictions	4.97	4.84	1−20	—	—	—	—	—	—
SVORI participant	0.52	0.49	0−1	—	—	—	—	—	—
Male	0.79	0.40	0−1	—	—	—	—	—	—

Note. N = 1,148 across three time points (3,010 total observations). GED = general educational development; SVORI = Serious and Violent Offender Reentry Initiative.

Dependent Variables—Reentry Outcomes

Self-reported recidivism, reincarceration, self-reported drug use, and self-reported treatment participation are the dependent variables in the current study. Time-varying measures of self-reported recidivism were created from existing SVORI data items. SVORI respondents were asked to self-report if they committed nine types of crime (e.g., violent, drug, property, crimes against persons) since they were released from incarceration (at Wave 2) or since their last interview (at Waves 3 and 4). Individuals who reported involvement in any of the listed crimes at a given wave were coded as a recidivist at that wave. Separate dichotomous indicators of self-reported offending were created at each postrelease wave.

As an alternative indicator of recidivism, we also included a time-varying measure of respondents’ reincarceration status at the time of their SVORI interview. Reincarceration was assessed at each postrelease wave. This measure reflected whether the SVORI respondent was incarcerated in prison or jail at the time they were interviewed. This variable was coded dichotomously (1 = incarcerated; 0 = not incarcerated) with separate indicators created at Waves 2, 3, and 4. This measure was treated as time-varying because respondents who were incarcerated at the time of one interview may have been released prior to subsequent interviews or may have become incarcerated after not being incarcerated at a prior postrelease wave.

To create the self-reported drug use measure, data were combined from multiple items included in the original SVORI data set. At each postrelease interview, SVORI respondents self-reported their use of each of the 13 substances since their previous interview. These items were combined into dichotomous indicators of self-reported drug use at each of the three postrelease waves of the study. Separate indicators were created at Waves 2, 3, and 4.

Self-reported treatment participation was coded from six SVORI items repeated across the three postrelease waves. Respondents were asked whether they had received (a) “assistance with working on personal relationships”; (b) “training on how to change your attitudes related to criminal behavior”; (c) “anger management programs”; (d) “any educational services, such as GED or adult basic education classes”; (e) “mental health treatment or health care for emotional problems”; or (f) “any drug or alcohol treatment” since the time of their last interview. These six treatment services were selected because they focused on changing criminogenic needs rather than providing social support or other services (see Visher et al., 2017). Respondents who reported any treatment at a given wave were coded as treatment participants for that wave. Separate dichotomous indicators of treatment participation were created for each postrelease wave.

Independent Variables

The primary independent variables in this study were respondent beliefs about their ability and willingness to change. We identified three attitudes that reflected internal change beliefs in the SVORI data: human agency, legal cynicism, and readiness for change. The measure of agency utilized in the current study was derived from two attitudinal scales included in the SVORI data: self-efficacy and locus of control. Given the amount of conceptual overlap present in the two existing measures, we conducted a principal components analysis (PCA) to determine whether these two measures potentially reflected a broader theoretical construct of human agency. Initially, eight items were included in the PCA, but results demonstrated that two items did not load onto a single factor and were removed. The final PCA results supported a single-factor solution with factor loadings exceeding 0.55, an eigenvalue of 2.64, accounting for 44% of the variance. The human agency measure was created by summing and averaging the six scale items to create a total score with values ranging from 1 to 4. This scoring reflects the original Likert-type coding of the scale items (1 = strongly disagree to 4 = strongly agree). Higher scores indicate more reported agency. The alphas ranged from .74 to .77 from Wave 2 to Wave 4.

Consistent with prior research (Gau, 2015; Sampson & Bartusch, 1998), legal cynicism was captured by asking respondents questions about their attitudes toward the law and illegal behaviors. Legal cynicism was measured using a five-item scale with responses provided on a four-point Likert-type scale. Example items include “Laws are made to be broken,” and “It’s okay to do anything you want as long as you don’t hurt anyone.” The measure was created by summing and averaging the five scale items to create a total score with values ranging from 1 to 4. This scoring reflects the original Likert-type coding of the scale items and higher scores indicate more legal cynicism. The alphas ranged from .70 to .72 from Wave 2 to Wave 4.

The final independent variable was self-reported readiness for change. SVORI respondents were asked to report their level of agreement with questions pertaining to their motivation to change. Examples of these questions include “You are tired of the problems caused by the crimes you committed” and “You want to get your life straightened out.” The number of readiness for change items asked varied depending on whether the respondent was incarcerated or in the community at the time of the interview. To account for the differing number of items, the item responses were summed and then averaged to retain the original 1 to 4 scoring. The readiness for change measure was coded so that higher scores reflect more motivation to change. The alphas ranged from .74 to .78 from Wave 2 to Wave 4.

In addition to assessing internal change factors, this study assessed the relative salience of external change factors in predicting reentry outcomes. Given prior research demonstrating the importance of family support in the reentry process (Barrick et al., 2014; Mowen et al., 2019; Taylor, 2016; Visher et al., 2004), we incorporated measures of family instrumental and family emotional support. Family instrumental support was measured with five items, for example, “I have someone in my family who would provide help or advice on finding a place to live,” and “I have someone in my family who would provide help or advice on finding a job.” Response options ranged from strongly disagree (coded “1”) to strongly agree (coded “4”) and were coded so that higher scores indicate greater perceptions of instrumental support. Item responses were summed and then averaged to retain the original 1 to 4 scoring. The alphas were .88 across all postrelease waves.

Family emotional support was assessed using a 10-item, Likert-type scale with higher scores indicating greater perceived family emotional support. Example items include “I feel close to my family,” “I want my family to be involved in my life,” and “I consider myself a source of support for my family.” Item responses were summed and then averaged to retain the original 1 to 4 scoring. The alphas were .87 across all postrelease waves.

The final external change factor included was perceived neighborhood quality. To measure perceived neighborhood quality, data were drawn from a five-item, Likert-type scale. Example items include “It is hard to stay out of trouble in your neighborhood,” and “Drug selling is a major problem in your neighborhood.” Items were reverse coded so higher scores reflect greater perceived neighborhood quality. Item responses were summed and then averaged to retain the original 1 to 4 scoring. Cronbach’s alpha ranged from .76 to .82.

Control Variables

Prior research demonstrates that age, race/ethnicity, and employment are associated with desistance and reentry outcomes (Severson et al., 2012; Travis, 2005; Uggen, 2000). For this study, respondent age was a continuous variable measured in years. Race/ethnicity was self-identified by respondents. Dummy variables were constructed for White, Black, Hispanic, and other non-Hispanic. Education was assessed using a single item in which respondents self-reported completion of the 12th grade or general educational development (GED) equivalent, measured as a dichotomous variable (1 = high school diploma/GED, 0 = none). Sex was measured as a dichotomous variable (1 = male, 0 = female). Employment was a binary measure assessed using a single item in which respondents reported whether they had a job at any time since their previous interview (1 = employed, 0 = not employed). Finally, we included a measure of whether individuals were SVORI program participants measured as a dichotomous indicator (1 = SVORI program participant, 0 = non-SVORI program participant). In addition to controlling for demographic and study group variables, we also controlled for the total number of prior convictions, the instant offense type, and length of incarceration. For instant offense, we created a series of dummy variables with the categories of violent, property, drug, and other conviction. The total number of self-reported convictions was truncated at 20 to account for extreme outliers (fewer than 4% of values were recoded). Length of incarceration was measured as the total number of days the individual was incarcerated for the offense that immediately preceded their SVORI participation. We log transformed total number of days incarcerated to account for skewness.

Missing Data

Like most longitudinal and panel data sets, the SVORI data suffer from considerable attrition and missing data (Lattimore & Visher, 2009). In our analysis, we draw on a sample of 1,148 respondents or approximately 56% of the original adult sample. However, several studies relying on SVORI demonstrate that case attrition are largely missing at random with no significant differences found between Wave 1 and Wave 4 respondents (see Lattimore & Visher, 2009; Stansfield et al., 2017; Wallace et al., 2016). To ensure our estimates are not biased by case attrition, we performed multiple imputation with chained equations (MICE) and replicated our analyses in both the observed and imputed data sets (Azur et al., 2011). The number of imputations was set at 10 and data were imputed on all independent and dependent variables used in the current study. We chose to set the number of imputations at 10 as prior research indicates that larger thresholds may increase the likelihood of multicollinearity (Mowen & Culhane, 2017). We were able to fully impute data (N = 2,054) on all time-varying and time-stable covariates included in the study analyses.

Analytic Strategy

To test our hypotheses, we employed a hybrid model (or mixed-effects) approach (Allison, 2005). Although several methods exist for analyzing longitudinal panel data like SVORI, the hybrid model is appropriate for the current study because it allows for the estimation of the influence of both time-variant and time-invariant covariates on our dependent variables. This approach uses a multilevel model of repeated measures of time-variant and time-stable covariates nested within individuals. The hybrid model approach combines the advantages of fixed and random-effect models because it allows for the estimation of the influence of Level 2 data while still providing estimates of Level 1 variables (Schunck, 2013). Another key advantage is that the hybrid model permits the estimation of random and fixed effects simultaneously (Allison, 2009). This approach accounts for the nesting of observations within individuals and allows for the inclusion of both time-varying and time-stable independent variables. This approach allows the current study to extend prior research that has examined between-person differences and within-person changes on a variety of outcomes in the SVORI data (Boman & Mowen, 2018; Mowen et al., 2019).

Using Stata 15 (StataCorp, 2017), hybrid models were estimated in a series of steps. First, we created measures of between-individual and within-individual variation in our time-varying covariates. To do this, we first calculated the between-subject variation by averaging the scores of each individual across the waves of the repeated measure. The within-individual time-varying measures were then computed by subtracting the mean from the between-group values (Schunck, 2013). Finally, to test our hypotheses, we regressed each of the dependent variables onto our time-varying and time-stable independent variables using the xtlogit procedure in Stata. In our models predicting self-reported recidivism, self-reported drug use, and self-reported treatment participation, we included a control for whether the respondent was incarcerated at the time of their interview. This control was included in these models because respondents who were incarcerated at the time of their follow-up interview may have had fewer opportunities to commit new offenses, use drugs, and/or participate in treatment. Also, incarcerated respondents may have responded differently to interview questions given their incarceration status. Although these concerns created a need to control for incarceration status, respondents were not excluded from study analyses based on their incarceration status at any postrelease wave. This decision was made because the current study assessed the time-varying impact of theoretically relevant correlates of reentry outcomes and many respondents who were incarcerated at one postrelease wave were not incarcerated at a prior or subsequent wave. In fact, only 4.4% (n = 51) of our final analytic sample was incarcerated at multiple postrelease interviews. The low prevalence of multiwave incarceration limits concerns about respondents’ criminal opportunities.

Results

Hybrid Regression Findings

Self-Reported Recidivism

Table 2 presents the results of our hybrid models estimating the influence of internal and external change factors on four reentry outcomes. Results from the model predicting self-reported recidivism demonstrate that within-individual changes in agency (odds ratio [OR] = 0.59, p = .015) and family emotional support (OR = 0.49, p = .009) are significantly related to changes in recidivism across the postrelease waves. These findings indicate that the odds of self-reported recidivism are reduced for individuals who experience increases in agency and family emotional support over time. Age also demonstrated a significant relationship with reoffending (OR = 1.87, p < .001) although there was little time for within-individual changes in age during the study period. Within-individual changes in legal cynicism (OR = 1.63, p = .023), and reincarceration status (OR = 9.80, p < .001) were significantly associated with increased odds of self-reported reoffending. Among the time-stable covariates race, sex, offense type, and number of prior convictions all significantly predicted self-reported recidivism. Being Black, relative to White, was associated with a 57% reduction in the odds of recidivism (OR = 0.43, p < .001). Males (OR = 1.66, p = .017) had a greater likelihood of reoffending, relative to females, and individuals convicted of property crimes (OR = 1.65, p = .015) were more likely to self-report recidivism compared with individuals convicted of violent offenses.

Table 2:

Hybrid Model Estimating the Impact of Internal and External Change Factors on Recidivism, Reincarceration, Drug Use, and Treatment Participation

Variables	Recidivism		Reincarceration		Drug use		Treatment participation
Variables	Exp (B)	95% CI	Exp (B)	95% CI	Exp (B)	95% CI	Exp (B)	95% CI
Within-individual
Agency	0.59*	[0.386, 0.903]	0.51*	[0.302, 0.868]	0.70	[0.471,1.06]	0.94	[0.660, 1.35]
Legal cynicism	1.63*	[1.06, 2.51]	0.40**	[0.231, 0.702]	1.38	[0.915, 2.10]	0.78	[0.547, 1.12]
Readiness for change	1.16	[0.785, 1.72]	0.63	[0.389, 1.03]	1.03	[0.708, 1.52]	1.65**	[1.18, 2.31]
Family instrumental support	1.11	[0.738, 1.69]	0.57*	[0.332, 0.960]	0.93	[0.620, 1.39]	0.85	[0.600, 1.22]
Family emotional support	0.49**	[0.288, 0.835]	0.40**	[0.200, 0.782]	0.59*	[0.351, .992]	1.25	[0.793, 2.00]
Neighborhood quality	1.22	[0.961, 1.56]	0.93	[0.682, 1.27]	1.04	[0.824, 1.31]	0.90	[0.732, 1.11]
Self-reported recidivism	—	—	12.65***	[7.58, 21.1]	—	—	—	—
Incarceration	9.80***	[6.08, 15.8]	—	—	9.17***	[5.68, 14]	1.14	[0.775, 1.67]
Age	1.87***	[1.47, 2.38]	2.04	[1.49, 2.78]	2.08***	[1.64, 2.66]	0.56***	[0.450, 0.686]
Employment	1.11	[0.724, 1.71]	1.09	[0.651, 1.85]	0.94	[0.628, 1.43]	1.56*	[1.07, 2.25]
High school/GED	1.55	[0.773, 3.12]	1.12	[0.472, 2.66]	0.54	[0.280, 1.06]	1.13	[0.632, 2.05]
Between-individual
Agency	0.71	[0.457, 1.10]	0.71	[0.443, 1.15]	0.73	[0.467, 1.16]	0.52**	[0.351, 0.782]
Legal cynicism	1.87**	[1.26, 2.79]	0.79	[0.518, 1.22]	1.77**	[1.17, 2.67]	0.88	[0.624, 1.26]
Readiness for change	1.52	[0.946, 2.45]	0.61	[0.374, 1.02]	1.48	[0.905, 2.43]	2.46***	[1.60, 3.80]
Family instrumental support	2.09**	[1.28, 3.42]	1.12	[0.720, 2.05]	1.85*	[1.11, 3.06]	1.20	[0.775, 1.87]
Family emotional support	0.21***	[0.110, 0.400]	0.55	[0.281, 1.09]	0.21***	[0.108, 0.407]	0.71	[0.403, 1.26]
Neighborhood quality	1.97***	[1.45, 2.68]	1.52*	[1.09, 2.10]	1.89***	[1.37, 2.59]	0.69**	[0.522, 0.902]
Self-reported recidivism	—	—	3.58***	[2.14, 5.94]	—	—	—	—
Incarceration	7.08***	[3.80, 13.06]			6.98***	[3.65, 13.4]	1.50	[0.859, 2.62]
Age	0.95	[0.928, 0.973]	0.97*	[0.945, 0.996]	0.98	[0.962, 1.00]	1.01	[0.989, 1.02]
Employment	0.68	[0.423, 1.12]	0.52*	[0.314, 0.870]	0.63	[0.386, 1.06]	0.83	[0.534, 1.30]
High school/GED	1.57	[1.07, 2.31]	1.02	[0.679, 1.54]	0.95	[0.644, 1.41]	1.09	[0.779, 1.54]
Time-invariant measures
Black	0.43***	[0.300, 0.625]	0.96	[0.646, 1.43]	0.58**	[0.399, 0.848]	0.50***	[0.363, 0.697]
Hispanic	0.71	[0.344, 1.49]	0.89	[0.397, 2.00]	0.78	[0.361, 1.71]	0.61	[0.318, 1.17]
Other/multiracial	0.57	[0.321, 1.04]	0.99	[0.534, 1.86]	0.60	[0.323, 1.12]	0.80	[0.474, 1.36]
Property	1.65*	[1.10, 2.48]	1.24	[0.805, 1.92]	2.29***	[1.50, 3.52]	0.56**	[0.387, 0.815]
Drug	1.36	[0.915, 2.02]	0.86	[0.558, 1.33]	1.56*	[1.03, 2.35]	0.77	[0.545, 1.11]
Other offense	1.14	[0.642, 2.03]	1.23	[0.672, 2.27]	0.97	[.533, 1.12]	0.74	[0.443, 1.23]
Total convictions	1.09***	[1.06, 1.13]	1.01	[0.979, 1.04]	1.08***	[1.04, 1.11]	1.01	[0.981, 1.04]
Days incarcerated	0.85	[0.704, 1.03]	1.13	[0.910, 1.40]	0.83	[0.682, 1.01]	1.17	[0.992, 1.40]
SVORI participant	0.78	[0.575, 1.07]	1.31	[0.931, 1.84]	0.69*	[0.500, 0.960]	1.92***	[1.44, 2.56]
Male	1.66*	[1.09, 2.53]	1.80*	[1.11, 2.92]	1.52	[0.992, 2.34]	0.48***	[0.335, 0.701]
ρ (ICC)	.472		.391		.527		.460

Note. N = 1,148 with 3,010 total observations; Exp(B) = odds ratio; GED = general educational development; CI = confidence interval; SVORI = Serious and Violent Offender Reentry Initiative; ICC = intraclass correlation coefficient.

p < .05. **p < .01. ***p < .001 (two-tailed test).

When examining changes between individuals, legal cynicism (OR = 1.87, p = .002) was the only internal change factor that was significantly related to changes in the odds of recidivism. Individuals who scored higher on legal cynicism were at an increased risk of recidivism across the three postrelease waves. All three of the external change factors were significantly associated with changes in reoffending. Unexpectedly, neighborhood quality (OR = 1.97, p < .001) and family instrumental support (OR = 2.09, p = .003) were both positively related to changes in the odds of recidivism. However, family emotional support (OR = 0.21, p < .001) was negatively related to changes in the odds of recidivism over time.

Reincarceration

Results from the hybrid model predicting reincarceration (Table 2) reveal that within-individual changes in agency (OR = 0.51, p = .013), legal cynicism (OR = 0.40, p < .001), family instrumental support (OR = 0.57, p = .035), family emotional support (OR = 0.40, p = .008), and self-reported offending (OR = 12.66, p < .001) are significantly related to changes in reincarceration status across the postrelease waves. Consistent with the models predicting self-reported recidivism, results indicate that increases in agency and family emotional support reduce the odds of reincarceration over time. Although not significant in the model predicting self-reported recidivism, changes in family instrumental support are significantly and negatively related to reincarceration across the postrelease study waves. Not surprisingly, changes in self-reported offending were significantly and positively related to changes in reincarceration status over the postrelease waves. Inconsistent with our hypotheses and the self-reported offending models, increases in legal cynicism were related to reduced odds of reincarceration. Outside of the internal and external change factors, changes in age (OR = 2.04, p < .001) also significantly increased the odds of being reincarcerated. Within-individual changes in employment and education demonstrated no statistically significant association with reincarceration over time. Among the time-stable predictors, only sex was significantly related to reincarceration over time. Men (OR = 1.80, p = .016), relative to women, had a greater likelihood of reincarnation across the three postrelease study waves.

Between-individual changes in reincarceration were statistically related to self-reported offending (OR = 3.58, p < .001), individual perceptions of neighborhood quality (OR = 1.52, p = .012), age (OR = 0.97, p = .027), and employment (OR = 0.52, p = .012). Individuals who scored higher on family emotional support had reduced odds of being reincarcerated relative to those who reported less family emotional support. As seen with the self-reported recidivism model, individual perceptions of neighborhood quality were significantly and positively related to reincarceration.

Self-Reported Drug Use

Consistent with the results from the self-reported recidivism models, within-individual changes in family emotional support (OR = 0.59, p = .049) were significantly, negatively associated with self-reported drug use over time. This indicates that individuals who experience increases in family emotional support over time were at reduced odds of self-reporting drug use. Within-individual changes in reincarceration status were positively associated with increased odds of self-reported drug use (OR = 9.18, p < .001). Changes in age were significantly and positively related to changes in self-reported drug use (OR = 2.08, p < .001) across the postrelease waves. The time-stable effect of SVORI participation was negative and significant (OR = 0.69, p = .028), indicating that SVORI treatment participants were less likely to self-report drug use over the three postrelease waves. Conviction offense was also significantly associated with drug use. Individuals convicted of property and drug crimes were more likely to report drug use compared with persons convicted of violent offenses. Black respondents were significantly less likely to report drug use compared with White respondents (OR = 0.58, p = .005) and prior convictions (OR = 1.08, p < .001) were positively associated with drug use over time.

The time-varying effects of between-individual differences on self-reported drug use indicated that legal cynicism (OR = 1.77, p = .006), neighborhood quality (OR = 1.88, p < .001), and family instrumental support (OR = 1.85, p < .001) were all significant and positively associated with drug use over time. Between-individual differences in family emotional support (OR = 0.21, p < .001) were negatively associated with drug use over time. Compared with individuals who were not reincarcerated, individuals who were reincarcerated were significantly more likely to report using drugs (OR = 6.98, p < .001) over time.

Treatment Participation

The final model presents the hybrid regression results predicting self-reported treatment participation. Although not significantly related to recidivism or drug use, within-individual changes in readiness for change (OR = 1.65, p = .003) were significantly and positively related to treatment participation over time. Also unique to the treatment participation model, within-individual changes in employment (OR = 1.56, p = .018) were associated with increased treatment participation over time. Changes in respondents’ age demonstrated a significant, negative association with self-reported treatment participation (OR = 0.56, p < .001) over time. Within-individual changes in agency, legal cynicism, family instrumental support, family emotional support, and neighborhood quality were not significantly related to treatment participation.

In the between-individual models, readiness for change (OR = 2.46, p < .001) was a significant, positive predictor of treatment participation over time while agency (OR = 0.52, p = .002) and neighborhood quality (OR = 0.69, p = .007) were negatively related to treatment participation over time. Among the time-stable predictors, Black race (relative to White), male sex (relative to female), and an instant property offense (relative to violent) were negatively related to treatment participation while SVORI treatment group status was positively related to treatment participation over time.

Analysis of Imputed Data

Although prior research has largely demonstrated the missing data in the SVORI sample are largely missing at random, we addressed concerns about missing data using MICE. The use of MICE to impute data on all study independent and dependent variables allowed us to retain all 2,054 adult respondents in the SVORI study. Results of our hybrid models estimated on the imputed data are presented in Table 3. The general pattern of results from the observed data was partially replicated in the imputed models; however, some notable differences emerged. First, in the self-reported recidivism model, the positive, time-varying effect of legal cynicism was significant in the observed data, but not in the imputed data. Second, the negative, time-varying effect of family emotional support on self-reported recidivism and self-reported drug use was also significant in the observed data, but not the imputed data. Finally, within-individual changes in family instrumental support were significant in the observed data predicting reincarceration, but not significant in the imputed data. Within-individual changes in readiness for change emerged as a significant, negative correlate of reincarceration in the imputed model after failing to reach statistical significance in the observed model.

Table 3:

Multiple Imputed Hybrid Model Estimating the Impact of Internal and External Change Factors on Recidivism, Reincarceration, Drug Use, and Treatment Participation

Variables	Recidivism		Reincarceration		Drug use		Treatment participation
Variables	Exp (B)	95% CI	Exp (B)	95% CI	Exp (B)	95% CI	Exp (B)	95% CI
Within-individual
Agency	0.66*	[0.443, 1.00]	0.54*	[0.317, 0.932]	0.76	[0.484, 1.19]	1.05	[0.758, 1.46]
Legal cynicism	1.40	[0.908, 2.17]	0.46**	[0.289, 0.762]	1.15	[0.717, 1.85]	0.70	[0.467, 1.05]
Readiness for change	1.04	[0.757, 1.44]	0.59*	[0.388, 0.927]	0.87	[0.632, 1.21]	1.53**	[1.11, 2.10]
Family instrumental support	0.94	[0.549, 1.62]	0.90	[0.498, 1.65]	0.89	[0.544, 1.47]	0.79	[0.549, 1.16]
Family emotional support	0.67	[0.335, 1.34]	0.22***	[0.104, 0.495]	0.64	[0.355, 1.16]	1.38	[0.848, 2.26]
Neighborhood quality	1.17	[0.936, 1.47]	0.98	[0.669, 1.45]	1.01	[0.816, 1.25]	0.90	[0.767. 1.06]
Self-reported recidivism	—	—	6.33***	[4.13, 9.72]	—	—	—	—
Incarceration	4.82***	[3.30, 7.01]	—	—	5.97***	[4.01, 8.89]	1.02	[0.686, 1.52]
Age	1.20	[0.979, 1.47]	4.64***	[3.22, 6.71]	1.62***	[1.29, 2.02]	0.63***	[0.521, .776]
Employment	1.11	[0.837, 1.76]	0.77	[0.488, 1.21]	0.94	[0.616, 1.46]	1.51*	[1.01, 2.25]
High school/GED	1.55	[0.880, 2.42]	1.13	[0.447, 2.89]	0.49*	[0.274, 0.908]	1.12	[0.565, 2.23]
Between-individual
Agency	0.81	[0.593, 1.11]	0.77	[0.476, 1.26]	0.75	[0.508, 1.10]	0.53***	[0.373, 0.778]
Legal cynicism	1.75***	[1.34, 2.29]	0.63	[0.398, 1.00]	1.63**	[1.16, 2.30]	0.86	[0.585, 1.26]
Readiness for change	1.22	[0.885, 1.69]	0.60	[0.338, 1.09]	1.29	[0.893, 1.86]	2.13***	[1.54, 2.94]
Family instrumental support	1.58	[0.979, 2.56]	1.52	[0.748, 3.09]	1.48	[0.884, 2.49]	0.91	[0.650, 1.27]
Family emotional support	0.31***	[0.183, .528]	0.22***	[0.098, 0.517]	0.27***	[0.135, 0.568]	1.26	[0.825, 1.95]
Neighborhood quality	1.65***	[1.33, 2.05]	1.45*	[1.04, 2.03]	1.60**	[1.18, 2.15]	0.73**	[0.590, 0.920]
Self-reported recidivism	—	—	4.58***	[2.55, 8.24]	—	—	—	—
Incarceration	2.63***	[1.87, 3.74]	—	—	3.44***	[1.86, 6.35]	1.10	[0.713, 1.72]
Age	0.95***	[0.939, 0.973]	0.94***	[0.912, 0.968]	0.98	[0.962, 1.00]	1.00	[0.989, 1.02]
Employment	0.73	[0.515, 1.05]	0.27***	[0.163, 0.478]	0.79	[0.542, 1.15]	1.13	[0.809, 1.58]
High school/GED	1.43*	[1.07, 1.90]	0.96	[0.619, 1.50]	0.99	[0.700, 1.41]	1.03	[0.753, 1.42]
Time-invariant measures
Black	0.60***	[0.461, 0.792]	1.09	[0.681, 1.75]	0.76	[0.578, 1.02]	0.56***	[0.429, 0.731]
Hispanic	0.87	[0.506, 1.50]	0.66	[0.265, 1.67]	0.77	[0.379, 1.57]	0.72	[0.399, 1.31]
Other/multiracial	0.77	[0.528, 1.15]	1.57	[0.812, 3.06]	0.78	[0.481, 1.26]	0.91	[0.608, 1.38]
Property	1.57**	[1.17, 2.10]	1.43	[0.926, 2.21]	1.83***	[1.27, 2.63]	0.53***	[0.394, 0.711]
Drug	1.19	[0.902, 1.58]	0.98	[0.650, 1.48]	1.31	[0.928, 1.84]	0.79	[0.598, 1.06]
Other offense	1.09	[0.588, 2.05]	1.61	[0.934, 2.79]	1.03	[0.625, 1.71]	0.65	[0.410, 1.04]
Total convictions	1.07***	[1.04, 1.10]	1.01	[0.982, 1.04]	1.07***	[1.03, 1.11]	1.00	[0.977, 1.03]
Days incarcerated	0.90	[0.791, 1.03]	1.19	[0.948, 1.50]	0.84*	[0.715, 0.995]	1.04	[0.900, 1.21]
SVORI participant	0.87	[0.714, 1.07]	1.00	[0.717, 1.40]	0.83	[0.646, 1.07]	1.44**	[1.12, 1.86]
Male	1.22	[0.926, 1.62]	1.60*	[1.00, 2.57]	1.22	[0.885, 1.68]	0.62**	[0.462, 0.850]
ρ (ICC)	.350		.643		.492		.451

Note. N = 2,054 with 6,216 total observations; Exp(B) = odds ratio; CI = confidence interval; GED = general educational development; SVORI = Serious and Violent Offender Reentry Initiative; ICC = intraclass correlation coefficient.

p < .05. **p < .01. ***p < .001 (two-tailed test).

Discussion

The purpose of the current study was to explore the relative salience of multiple internal and external change factors as predictors of reentry outcomes to inform theory, policy, and practice. Consistent with the MORM and IMTCD models of individual change, we found support for the dynamic predictive validity of both individual attitudes (i.e., internal change factors) and contextual influences (i.e., external change factors). Given concerns about the appropriateness of interpreting between-individual changes in hybrid models (Allison, 2005), we focus primarily on the findings of our within-individual change models and the effects of time-stable covariates in this discussion.

For self-reported recidivism, reincarceration, and drug use outcomes, within-individual changes in family emotional support were identified as significant protective factors. Changes in family emotional support were significantly related to changes in the odds of self-reported offending, reincarceration, and self-reported drug use over the three postrelease waves of the SVORI study (approximately 15 months). Changes in family emotional support were not significantly related to self-reported treatment participation. While changes in family emotional support were significantly related to three of the four reentry outcomes, changes in family instrumental support were only significantly related to changes in incarceration status across the postrelease waves. Among the internal change factors included in our analyses, agency was identified as a significant protective factor against both self-reported recidivism and reincarceration. Readiness for change was the only internal or external change factor that was significantly related to changes in self-reported treatment participation across the three postrelease waves of the SVORI study. Interestingly, changes in legal cynicism were positively related to changes in self-reported recidivism but negatively related to changes in reincarceration status across study waves.

Finding family emotional support as a predictor of successful reentry outcomes is consistent with prior literature (Bales & Mears, 2008; Barrick et al., 2014; Mears et al., 2012; Travis, 2005), including previous studies conducted with the SVORI data (Barrick et al., 2014; Boman & Mowen, 2018; Taylor, 2016; Wallace et al., 2016). This finding adds to the growing body of literature indicating the importance of maintaining and/or strengthening family relationships during and after a period of incarceration (Barrick et al., 2014; Boman & Mowen, 2018; Taylor, 2016; Wallace et al., 2016). While within-individual changes in family emotional support were significantly related to both recidivism outcomes and self-reported drug use, within-individual changes in family instrumental support were only significantly related to reincarceration status. These findings diverge somewhat from prior research (Mowen et al., 2019; Taylor, 2016) and warrant further exploration. One difference between the current study and the work of Mowen and colleagues (2019) concerns the measurement of family instrumental and family emotional support. While our study relied on the original scales included in the SVORI data, Mowen and colleagues (2019) created measures of three different types of familial support (i.e., interactional, instrumental, and emotional) using factor analysis. Another potentially relevant difference between our study and the work of Mowen and colleagues (2019) concerns the inclusion of women in our models. The finding of a significant protective effect of family emotional support in our study is consistent with Taylor (2016) who also included both men and women in her study. Taken together, these findings suggest that different types of family support may be differentially salient across sex. This finding requires further empirical testing and theoretical exploration. In line with Mowen and colleagues (2019), we believe that there is a pressing need for further refinement of family support measurement in the study of reentry. Specifically, more work is needed to develop distinct conceptual and operational definitions of various types of social support and theoretical work is needed to more clearly articulate the relationship between these constructs and various reentry outcomes. Future research should be dedicated to further elucidating the mechanisms through which different types of social support affect specific reentry outcomes and exploring potential interactions with demographic or contextual variables. For instance, based on the differential impact of family emotional and family instrumental support observed in this study, research could explore why emotional support is generally more protective while instrumental support may be particularly salient for avoiding return to incarceration.

The relationship between within-individual changes in agency and improved reentry outcomes (i.e., reduced self-reported recidivism and reincarceration) observed in the current study is consistent with Paternoster and Bushway’s (2009) identity theory of desistance (ITD) that predicts that human agency is an important component of the desistance process. Although our findings appear consistent with the ITD, further empirical testing is needed given the potentially complex relationship between agency and desistance outcomes (Paternoster et al., 2015). This theoretical perspective suggests that agency interacts with identity shifts and may be either criminogenic or protective depending on other individual and contextual factors. Consistent with this interaction hypothesis, Lloyd and Serin (2012) also suggest the intention to change will interact with the perceived ability to change (i.e., agency) before meaningful behavior change occurs. Future research is needed to further probe the protective effect of agency observed in the current study. A particularly salient line of inquiry would assess whether agency is conditioned by identity as predicted by the ITD.

The positive relationship observed between within-individual changes in readiness for change and treatment participation over time during reentry indicates the salience of this construct as a predictor of treatment outcomes. However, the lack of significant relationships observed between readiness for change and either measure of recidivism or drug use suggests that although this construct is conceptualized as an important correlate of desistance (Serin et al., 2010; Serin & Lloyd, 2009; Ward et al., 2004), it may have limited utility for predicting individual change, at least when measured based on self-reported attitudes alone. More theoretical development and empirical testing is needed to elucidate the role that readiness attitudes play in shaping reentry outcomes. A promising line of future inquiry involves more complete tests of the conceptual models proposed by Serin and colleagues (2010; Serin & Lloyd, 2009) and Ward and colleagues (2004). This caveat aside, study findings do support readiness for change as a significant predictor of treatment participation that can be targeted in prerelease programming to improve postrelease treatment outcomes.

Study findings should be considered in the context of several relevant strengths and limitations. The SVORI data provided a relatively rich source of information on individual attitudes, social supports, and perceptions of reentry contexts (Visher et al., 2017). The data also included self-report information regarding offending, reincarceration, drug use, and treatment utilization. The longitudinal nature of the data allowed for the assessment of time-varying effects of behavior change correlates during reentry. And, despite sample attrition across waves, the SVORI sample was adequately large to answer study research questions.

Despite these strengths, some study limitations warrant consideration. Although hybrid models offer the advantage of simultaneously estimating time-variant and time-invariant effects, this technique also caries limitations worthy of discussion (Allison, 2014; Boman & Mowen, 2018). First, interpretation of between-individual effects should be made with caution given that the between-individual coefficients are captured as an average measure of the within-individual effects across multiple time points (Allison, 2005, 2014); This results in correlated error terms across the within- and between-person estimates and can hinder interpretation of temporal order. These methodological limitations might also partly explain discrepancies in our findings compared with other studies using the SVORI data set (e.g., Mowen et al., 2019). However, despite these limitations, other scholars utilizing the SVORI data set have relied on similar methods to assess change over time (Boman & Mowen, 2018; Mowen et al., 2019) and the hybrid approach remains a viable analytic approach that overcomes limitations of alternative modeling strategies (Allison, 2014).

As this study involved a secondary analysis of existing data, we were not able to locate measures of all relevant variables implicated in existing desistance frameworks; accordingly, the study provides only a partial test of these models. In addition, the SVORI measures of individual attitudes, our primary independent variables, have not been previously validated. The lack of external validation creates concerns about construct and other validity threats, although the included items do have face validity. Furthermore, the study measure of neighborhood quality reflects respondent perceptions rather than an objective indicator of neighborhood context. As such, respondents may perceive more or less disorder than is objectively present (Sampson & Raudenbush, 2004), which may pose a threat to the validity of this measure. However, extant research has established the reliability of self-reported measures of neighborhood characteristics (Echeverria et al., 2004). Given the possible discrepancy between individual perceptions and structural indicators of neighborhood context, future reentry research should leverage both perceptual and objective measures of neighborhood contexts. Another potential critique of this study concerns the reliance on self-reported measures of recidivism and drug use. While this critique is certainly relevant, prior research indicates that relying on self-reported offending data is an acceptable practice (Maxfield et al., 2000; Weis, 1986) and self-reported data offer the advantage of identifying offenses that did not come to the attention of criminal justice officials. Given this strength of self-reported offending data and the fact that official arrest data were not available for the SVORI study at the time we accessed them, we argue that the use of self-reported recidivism data was appropriate for testing our hypotheses. Sample attrition is another limitation of the current study; however, this limitation is addressed by the replication of our analyses after accounting for missing data through multiple imputation. Finally, the study sample, while appropriately large for the study analyses, is not representative of reentry populations in general. Given the SVORI emphasis on “serious” and “violent” formerly incarcerated individuals, study findings are not generalizable to other reentry populations; however, these findings generate relevant research questions and the study methodology can be replicated in other, more generalizable samples.

Despite these limitations, the current study findings have implications for practice and research. The finding that family emotional support was a significant protective factor against recidivism and drug use suggests that family support should be targeted in reentry programming efforts. As noted in prior research (Clear, 2009; Wakefield & Uggen, 2010), maintaining familial support during incarceration is challenging and must be addressed throughout the reentry process. The current study findings also identify agency as a potential target for reentry programming aimed at preventing recidivism and readiness for change as a target for programming aimed at improving postrelease treatment utilization. The findings related to agency are particularly intriguing from a programming perspective. Existing correctional treatment frameworks (e.g., the Risk, Needs, Responsivity model) do not consider agency as a criminogenic need to be targeted in rehabilitative programming. In contrast to this perspective, our findings suggest that interventions aimed at empowering returning individuals to become more agentic may reduce recidivism. This should be tested in future research, especially given the potential interaction between agency and identity implicated in Paternoster and Bushway’s (2009) ITD.

The partial test of existing desistance models (MORM and IMTCD) included in the current study highlights some directions for future research. Our findings indicate that continuing to refine and empirically assess conceptual models of change during reentry is essential. Several internal and external change factors implicated in existing frameworks were found to be related to reentry outcomes; however, support for the predictive validity of these constructs was not universal. Further research is needed to improve the measurement of theoretically relevant desistance and reentry correlates and more completely test the conceptual models identified in the current study. Without a better understanding of the complexity of the desistance process during reentry, treatment and policy initiatives are unlikely to achieve broad, cost-efficient success.

Footnotes

Authors’ Note:

The authors would like to thank the three anonymous reviewers and the Criminal Justice and Behavior editorial staff for their thoughtful insight and recommendations for preparation of this article. They would also like to acknowledge and thank Dr. Christy Visher for her willingness to grant them access to the Serious and Violent Offender Reentry Initiative (SVORI) data and Dr. James Ray and Dr. Thomas Baker for their feedback and recommendations regarding study methodology and analyses.

ORCID iD

Lucas M. Alward

Lucas M. Alward is a doctoral candidate in the Department of Criminal Justice at the University of Central Florida. His primary research interests include reentry and desistance, procedural justice and legitimacy, and the implementation of effective correctional interventions. His recent research has appeared in Criminal Justice Studies: A Critical Journal of Crime, Law and Society and Corrections: Policy, Practice and Research.

Michael S. Caudy, PhD, is an assistant professor of criminal justice at the University of Central Florida. His primary research interests include the design, implementation, and effectiveness of correctional interventions; the interaction between individual characteristics and correctional interventions in producing outcomes; risk and needs assessment practices; and readiness for change. His recent research has appeared in Criminal Justice and Behavior, Criminology & Public Policy, Justice Quarterly, and Journal of Criminal Justice.

Jill Viglione, PhD, is an assistant professor in the Department of Criminal Justice at the University of Central Florida. Her research focuses on examining the processes associated with the implementation of evidence-based practices in criminal justice settings. Her recent research has appeared in Criminal Justice and Behavior, Psychology, Crime & Law, and International Journal of Offender Therapy and Comparative Criminology.

References

Allison

P. D.

(2005). Fixed effects regression methods for longitudinal data using SAS. SAS Institute.

Allison

P. D.

(2009). Fixed effects regression models. SAGE.

Allison

P. D.

(2014, September 2). Problems with the hybrid method. Statistical Horizons. https://statisticalhorizons.com/problems-with-the-hybrid-method

Alper

Durose

M. R.

Markman

(2018). Update on prisoner recidivism: A 9-year follow-up period (2005-2014). Bureau of Justice Statistics.

Andrews

D. A.

Bonta

Hoge

R. D.

(1990). Classification for effective rehabilitation: Rediscovering psychology. Criminal Justice and Behavior, 17(1), 19–52.

Azur

M. J.

Stuart

E. A.

Frankgakis

Leaf

P. J.

(2011). Multiple imputation by chained equations: What is it and how does it work? International Journal of Methods in Psychiatric Research, 20(1), 40–49.

Bales

W. D.

Mears

D. P.

(2008). Inmate social ties and the transition to society: Does visitation reduce recidivism? Journal of Research in Crime & Delinquency, 45(3), 287–321.

Bandura

(1997). Self-efficacy: The exercise of control. Freeman.

Bandura

Adams

N. E.

Beyer

(1977). Cognitive processes mediating behavioral change. Journal of Personality and Social Psychology, 35(3), 125–139.

10.

Barrick

Lattimore

P. K.

Visher

C. A.

(2014). Reentering women: The impact of social ties on long-term recidivism. The Prison Journal, 94(3), 279–304.

11.

Benda

B. B.

(2005). Gender differences in life-course theory of recidivism: A survival analysis. International Journal of Offender Therapy and Comparative Criminology, 49(3), 325–342.

12.

Boman

J. H.

IV Mowen

T. J.

(2018). The role of turning points in establishing baseline differences between people in developmental and life-course criminology. Criminology, 56(1), 191–224.

13.

Bonta

Andrews

D. A.

(2016). The psychology of criminal conduct (6th ed.). Routledge.

14.

Bottoms

Shapland

Costello

Holmes

Muir

(2004). Towards desistance: Theoretical underpinnings for an empirical study. The Howard Journal of Criminal Justice, 43(4), 368–389.

15.

Brocato

Wagner

E. F.

(2008). Predictors of retention in an alternative-to-prison substance abuse treatment program. Criminal Justice and Behavior, 35(1), 99–119.

16.

Carson

E. A.

Golineeli

(2013). Prisoners in 2012: Trends in administration and releases, 1991-2012. Washington, DC: U.S. Department of Justice, Bureau of Justice Statistics.

17.

Casey

Day

Howells

(2005). The application of the transtheoretical model to offender populations: Some critical issues. Legal and Criminological Psychology, 10(2), 157–171.

18.

Chamberlain

A. W.

Wallace

(2016). Mass reentry, neighborhood context and recidivism: Examining how the distribution of parolees within and across neighborhoods impacts recidivism. Justice Quarterly, 33(5), 912–941.

19.

Clear

T. R.

(2009). Imprisoning communities: How mass incarceration makes disadvantaged neighborhoods worse. Oxford University Press.

20.

Clear

T. R.

Frost

N. A.

(2014). The punishment imperative: The rise and failure of mass incarceration in American. New York University Press.

21.

Clear

T. R.

Rose

D. R.

Ryder

J. A.

(2001). Incarceration and the community: The problem of removing and returning offenders. Crime & Delinquency, 47(3), 335–351.

22.

Clear

T. R.

Rose

D. R.

Waring

Scully

(2003). Coercive mobility and crime: A preliminary examination of concentrated incarceration and social disorganization. Justice Quarterly, 20(1), 33–64.

23.

Day

Casey

Ward

Howells

Vess

(2010). Transitions to better lives: Offender readiness and rehabilitation. Willan Publishing.

24.

Duvall

J. L.

Oser

C. B.

Leukefeld

C. G.

(2008). Readiness to change as a predictor of drug-related behaviors in a sample of rural felony probationers. The American Journal of Drug and Alcohol Abuse, 34(6), 741–748.

25.

Echeverria

S. E.

Diez-Roux

A. V.

Link

B. G.

(2004). Reliability of self-reported neighborhood characteristics. Journal of Urban Health, 81(4), 682–701.

26.

France

Homel

(2006). Societal access routes and developmental pathways: Putting social structure and young people’s voice into the analysis of pathways into and out of crime. Australian & New Zealand Journal of Criminology, 39(3), 295–309.

27.

Garland

(2001). The culture of control. Oxford University Press.

28.

Gau

J. M.

(2015). Procedural justice, police legitimacy, and legal cynicism: A test for mediation effects. Police Practice and Research, 16(5), 402–415.

29.

Hagger

M. S.

Chatzisarantis

N. L.

Biddle

S. J.

(2002). A meta-analytic review of the theories of reasoned action and planned behavior in physical activity: Predictive validity and the contribution of additional variables. Journal of Sport and Exercise Psychology, 24(1), 3–32.

30.

Hiller

M. L.

Knight

Leukefeld

Simpson

D. D.

(2002). Motivation as a predictor of therapeutic engagement in mandated residential substance abuse treatment. Criminal Justice and Behavior, 29, 56–75.

31.

Kirk

D. S.

Papachristos

A. V.

(2011). Cultural mechanisms and the persistence of neighborhood violence. American Journal of Sociology, 116(4), 1190–1233.

32.

Lattimore

P. K.

Visher

C. A.

(2009). Multi-site evaluation of SVORI: Summary and synthesis. RTI International.

33.

Laub

J. H.

Sampson

R. J.

(2001). Understanding desistance from crime. Crime and Justice, 28, 1–69.

34.

Lloyd

C. D.

Serin

R. C.

(2012). Agency and outcome expectancies for crime desistance: Measuring offenders’ personal beliefs about change. Psychology, Crime & Law, 18(6), 543–565.

35.

Longshore

Teruya

(2006). Treatment motivation in drug users: A theory-based analysis. Drug and Alcohol Dependence, 81, 179–188.

36.

Maisto

S. A.

Connors

G. J.

Zywiak

W. H.

(2000). Alcohol treatment changes in coping skills, self-efficacy, and levels of alcohol use and related problems 1 year following treatment initiation. Psychology of Addictive Behaviors, 14(3), 257–266.

37.

Maruna

(2001). Making good: How ex-convicts reform and rebuild their lives. American Psychological Association.

38.

Maxfield

M. G.

Weiler

B. L.

Widom

C. S.

(2000). Comparing self-reports and official records of arrests. Journal of Quantitative Criminology, 16, 87–110.

39.

McMurran

Tyler

Hogue

Cooper

Dunseath

McDaid

(1998). Measuring motivation to change in offenders. Psychology, Crime and Law, 4(1), 43–50.

40.

Mears

D. P.

Cochran

J. C.

Siennick

S. E.

Bales

W. D.

(2012). Prison visitation and recidivism. Justice Quarterly, 29(6), 888–918.

41.

Mossière

Serin

(2014). A critique of models and measures of treatment readiness in offenders. Aggression and Violent Behavior, 19(4), 383–389.

42.

Mowen

T. J.

Culhane

S. E.

(2017). Modeling recidivism within the study of offender reentry: Hierarchical generalized linear models and lagged dependent variable models. Criminal Justice and Behavior, 44(1), 85–102.

43.

Mowen

T. J.

Stansfield

Boman

J. H.

IV . (2019). Family matters: Moving beyond “if” family support matters to “why” family support matters during reentry from prison. Journal of Research in Crime & Delinquency, 56(4), 483–523.

44.

Paternoster

Bachman

Bushway

Kerrison

E. O.

Connell

. (2015). Human agency and explanations of criminal desistance: Arguments for a rational choice theory. Journal of Developmental and Life-Course Criminology, 1(3), 209–235.

45.

Paternoster

Bushway

(2009). Desistance and the “feared self”: Toward an identity theory of criminal desistance. The Journal of Criminal Law and Criminology, 99(4), 1103–1156.

46.

Petersilia

(2003). When prisoners come home: Parole and prisoner reentry. Oxford University Press.

47.

Rose

D. R.

Clear

T. R.

(2004). Who doesn’t know someone in jail? The impact of exposure to prison on attitudes toward formal and informal controls. The Prison Journal, 84(2), 228–247.

48.

Sampson

R. J.

Bartusch

D. J.

(1998). Legal cynicism and (subcultural?) tolerance of deviance: The neighborhood context of racial differences. Law & Society Review, 32, 777–804.

49.

Sampson

R. J.

Raudenbush

S. W.

(2004). Seeing disorder: Neighborhood stigma and the social construction of “broken windows.” Social Psychology Quarterly, 67(4), 319–342.

50.

Schunck

(2013). Within and between estimates in random-effects models: Advantages and drawbacks of correlated random effects and hybrid models. The Stata Journal, 13(1), 65–76.

51.

Serin

R. C.

Kennedy

(1997). Treatment readiness and responsivity: Contributing to effective correctional programming [Research report]. Correctional Service of Canada.

52.

Serin

R. C.

Lloyd

C. D.

(2009). Examining the process of offender change: The transition to crime desistance. Psychology, Crime & Law, 15(4), 347–364.

53.

Serin

R. C.

Lloyd

C. D.

Hanby

L. J.

(2010). Enhancing offender re-entry: An integrated model for enhancing offender re-entry. European Journal of Probation, 2(2), 53–75.

54.

Severson

M. E.

Veeh

Bruns

Lee

(2012). Who goes back to prison; Who does not: A multiyear view of reentry program participants. Journal of Offender Rehabilitation, 51(5), 295–315.

55.

Sloas

L. B.

Caudy

M. S.

Taxman

F. S.

(2018). Is treatment readiness associated with substance use treatment engagement? An Exploratory Study. Journal of Drug Education, 47(1–2), 51–67.

56.

Stansfield

’Mowen

T. J. O.

Connor

Boman

J. H.

(2017). The role of religious support in reentry: Evidence from the SVORI data. Journal of Research in Crime & Delinquency, 54(1), 111–145.

57.

StataCorp. (2017). Stata statistical software: Release 15.

58.

Taylor

C. J.

(2016). The family’s role in the reintegration of formerly incarcerated individuals: The direct effects of emotional support. The Prison Journal, 96(3), 331–354.

59.

Travis

(2005). But they all come back: Facing the challenges of prisoner reentry. Urban Institute Press.

60.

Travis

Western

Redburn

F. S.

(2014). The growth of incarceration in the United States: Exploring causes and consequences. National Academies Press.

61.

Tyler

T. R.

Huo

(2002). Trust in the law: Encouraging public cooperation with the police and courts through. Russell Sage Foundation.

62.

Uggen

(2000). Work as a turning point in the life course of criminals: A duration model of age, employment, and recidivism. American Sociological Review, 65(4), 529–546.

63.

Visher

C. A.

Lattimore

P. K.

Barrick

Tueller

(2017). Evaluating the long-term effects of prisoner reentry services on recidivism: What types of services matter? Justice Quarterly, 34(1), 136–165.

64.

Visher

C. A.

LaVigne

Travis

(2004). Returning home: Understanding the challenges of prisoner reentry. The Urban Institute.

65.

Visher

C. A.

Travis

. (2003). Transitions from prison to community: Understanding individual pathways. Annual Review of Sociology, 29(1), 89–113.

66.

Wakefield

Uggen

(2010). Incarceration and stratification. Annual Review of Sociology, 36, 387–406.

67.

Wakefield

Wildeman

(2013). Children of the prison boom: Mass incarceration and the future of American inequality. Oxford University Press.

68.

Wallace

Fahmy

Cotton

Jimmons

McKay

Stoffer

Syed

(2016). Examining the role of familial support during prison and after release on post-incarceration mental health. International Journal of Offender Therapy and Comparative Criminology, 60(1), 3–20.

69.

Wallace

Papachristos

A. V.

(2014). Recidivism and the availability of health care organizations. Justice Quarterly, 31(3), 588–608.

70.

Walters

G. D.

(2012). Criminal thinking and recidivism: Meta-analytic evidence on the predictive and incremental validity of the Psychological Inventory of Criminal Thinking Styles (PICTS). Aggression and Violent Behavior, 17(3), 272–278.

71.

Ward

Day

Howells

Birgden

(2004). The multifactor offender readiness model. Aggression & Violent Behavior, 9(6), 645–673.

72.

Weis

J. G.

(1986). Issues in the measurement of criminal careers. In Blumstein

Cohen

Roth

J. A.

Visher

C. A.

(Eds.), Criminal careers and “career criminals” (pp. 2–51). National Academy Press.