Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Jan 1.
Published in final edited form as: J Fam Violence. 2019 Sep 12;36(1):51–62. doi: 10.1007/s10896-019-00098-4

Youth Reentry from Prison and Family Violence Perpetration: the Salience of Family Dynamics

Thomas J Mowen 1, Benjamin W Fisher 2
PMCID: PMC8277155  NIHMSID: NIHMS1706826  PMID: 34267417

Abstract

The central role of family within the process of juvenile reentry from a term of incarceration has been well documented by researchers and practitioners alike. However, family violence among previously incarcerated youth remains alarmingly high across the United States. Drawing from differential coercion and social support theory, we examine how family dynamics may simultaneously promote and/or inhibit family violence perpetration among youth undergoing the process of reentry. Four waves of panel data from the male-only youth subsample of the Serious and Violent Offender Reentry Initiative are analyzed using a series of dynamic panel data models. Findings demonstrate that both pre- and post-release levels of family conflict are significantly associated with increased family violence during reintegration. Mechanisms of family support, however, are not associated with post-release family violence. Results from this study highlight the salience of family conflict in understanding family violence perpetration among recently released juveniles and their families.

Keywords: Family violence, Youth reentry, Family conflict, Family support, Incarceration

In 2016, just over 45,000 juveniles were in residential placement facilities in the United States (Office of Juvenile Justice and Delinquency Prevention [OJJDP] 2018b). Trends in rates of juvenile incarceration show that the average age within this population is about 17 years, and racial/ethnic minorities are incarcerated at a rate that is about 2.7 times higher than White youth (OJJDP 2018a). Despite a decrease in the number of youth incarcerated over the past decade (OJJDP 2018a), youth reentry – the process of leaving a detention facility and returning to the community – has remained a pressing issue among researchers, practitioners, and policy makers alike. Juvenile incarceration is associated with a wide range of negative outcomes following release including worse physical and mental health (Barnert et al. 2017), diminished psychosocial maturity (Dmitrieva et al. 2012), decreased employment prospects (Apel and Sweeten 2010), and increased contact with the criminal justice system (see Lambie and Randell 2013 for an overview). Research has also shown that juvenile incarceration inhibits educational outcomes (Bullis et al. 2002), and can change important social relationships with families (Sullivan 2004). As Steinberg et al. (2004) note, one major issue faced by adjudicated youth is poor family relationships, and removing youth from their home via incarceration can exacerbate existing issues within the family such as family violence.

Family violence occurs with striking regularity among high-risk and adjudicated youth. Findings from the Survey of Youth in Residential Placement (SYRP) show that about 67% of juvenile males incarcerated for a violent offense know or are related to their victims (Sedlak and Bruce 2010). Although the SYRP do not include data on violence perpetration, survey data show that just under 40% of youth report some form of abuse prior to incarceration with the majority of these experiences occurring within the family unit. Understanding the correlates of family violence are important for researchers and policy makers alike as prior studies have shown that incarcerated youth who experience or engage in family violence are placed at a much greater risk of engaging in other violent behaviors relative to incarcerated youth who do not engage in family violence (Spaccarelli et al. 1995). Overall, these trends and findings mesh with prior research highlighting that family relationships within this population tend to be highly unstable and tenuous (Steinberg et al. 2004).

Perhaps somewhat paradoxically in light of the relatively high rates of family violence among incarcerated youth, the family is often cited as the most important resource for youth as they leave incarceration and undergo the process of reentry (Mowen and Boman 2018; Panuccio et al. 2012, see also OJJDP 2018c). Families are critical sources of social support for returning youth (Martinez and Abrams 2013) and provide tangible (e.g., housing) and intangible (e.g., love and emotional support) support. Yet, juvenile reentry is commonly accompanied with tumultuous family relationships and interactions (Moore et al. 2013; Martinez and Abrams 2013). Although family support has been tied to increased reentry success among juveniles (Martinez and Abrams 2013), conflictual family relationships have been linked to recidivism and reincarceration (Braman and Wood 2003; Cottle et al. 2001; Mowen and Boman 2018). Scholarship is far less clear, however, on the relationship between family dynamics and family violence perpetration among returning youth.

Clearly, understanding factors that relate to the perpetration of family violence among previously incarcerated youth is pressing both in terms of theory and policy. Consequently, the goal of this study is two-fold. Drawing on differential coercion and social support theory (Colvin et al. 2002), we explore the extent to which family relations during incarceration and after release relate to family violence perpetration among returning youth. Using four waves of data from the youth subsample of the Serious and Violent Offender Reentry Initiative (a sample comprising male youth only), we explore how family conflict and support relate to family violence perpetration post-release. Following, we provide an overview of trends in juvenile incarceration and reentry. Then, we discuss the role of family within this process placed within the context of differential coercion and social support theory.

Juvenile Reentry and Family

As the Office of Juvenile Justice and Delinquency Prevention (2017) notes, national recidivism rates for juveniles do not exist, but the OJJDP estimates that about 55% of all juveniles will be rearrested within the first year of release and about 24% will be reincarcerated within the first year (see also Snyder and Sickmund 2006). Other studies mirror this conclusion. For example, a study by Calley (2012) found recidivism rates of about 24% and Bullis et al. (2002) reported recidivism rates of about 40% within the first year of release among this population. At the same time, researchers have called into question the efficacy of youth incarceration; a study sponsored by the U.S. Department of Justice following approximately 1300 youth over five years found that incarceration did not significantly reduce offending (Mulvey 2011).

These trends in recidivism among adjudicated youth have led researchers to explore a wide range of underlying mechanisms for the relatively high rates of reentry failure. Perhaps more so than any other factor, prior research has highlighted the important role of the family within the reentry process. Research has found, for example, that increased family contact during incarceration relates to more prosocial outcomes (Monahan et al. 2011), and family members provide “ties that bind” (Martinez and Abrams 2013) that ultimately lead to desistance from offending. Yet, family relationships can become strained within this process. Family members may be a significant source of stress and conflict (Braman and Wood 2003), and family conflict has been shown to lead to delinquent peer associations and recidivism (Mowen and Boman 2018). Highlighting this, Cottle et al. (2001) found that problematic family relations were one of the strongest predictors of recidivism among previously incarcerated youth. This duality is perhaps best summarized by Martinez and Abrams (2013) who find that although family members are typically excited about reuniting with their relatives, they can also be conflicted and report negative interactions and emotions including “anger, a sense of betrayal, and disappointment” (p. 173).

The importance of understanding how family dynamics relate to family violence among returning youth is well grounded in theory. Differential coercion and social support theory (DCSS; see Cullen 1994; Colvin et al. 2002) offers the possibility that individuals’ social networks and relationships can function both as (a) a coercive force that encourages criminal behavior because of the anxiety or fear associated with these relationships; and, (b) a source of support that restrains criminal behavior because of increased access to resources. Coercion refers to aversive inter-exchanges between the youth and their family members, perhaps best typified by family conflict. Social support, on the other hand, refers to receiving assistance from family members including love and a sense of belonging. Thus, rather than assessing the mere presence or absence of the interpersonal relationships of a given person, DCSS suggests that the nature of these relationships is a critical point of inquiry (Colvin et al. 2002; Colvin 2007). Additionally, both coercion and support can exist within the same relationships, where family members provide some forms of social support while simultaneously being a source of conflict. Similarly, it is possible for neither social support nor coercion to exist in a given relationship; the absence of one does not indicate the presence of the other. However, to date, research has not examined how the presence of family conflict and support – individually or in tandem – are related to family violence among juveniles.

From the perspective of DCSS, family conflict – as a coercive force – is likely to contribute to family violence while family emotional or instrumental support – forms of social support – likely restrain or prevent family violence. For example, juveniles may gain instrumental benefits from their family members such as food, shelter, transportation, and help finding employment (Martinez and Abrams 2013; Western et al. 2015). They may also reap other benefits such as having people with whom they may comfortably discuss their problems, and family members may provide access to social capital both within and outside the family (Western et al. 2015). According to DCSS, these social supports are expected to reduce the likelihood of offending, including engaging in family violence. On the other hand, families may act as a coercive force in which family conflict generates stress for juveniles that may increase their likelihood of engaging in family violence. This may arise from strained interpersonal relationships, financial hardships, or even family members’ criminal involvement. Although DCSS has been applied to the examination of successful reentry among adults (Mowen and Boman 2018), to our knowledge, no existing research has examined how families may incite family violence through conflictual relations or restrain family violence through social support among returning youth.

Clearly, the family plays a multifaceted and complicated role in the process of reentry. On one hand, family members are likely the most significant source of social support and thus, desistance from family violence. On the other hand, returning youth may experience conflict with their family members which may push them into recidivism, including family violence. While existing studies have tied family relationships to some outcomes including reincarceration and reoffending (Cottle et al. 2001; Mowen and Boman 2018), prior research has failed to examine how these family dynamics may contribute to family violence perpetration among youth. Family dynamics may be particularly salient for juveniles who may not have had the time or life experiences that would allow them to build broader networks with social capital that they might access to support their successful reentry. Instead, juveniles are likely to rely on their families for both their instrumental (e.g., housing) and emotional (e.g., love and encouragement) support. As such, examining the extent to which family dynamics influence family violence during reentry among juveniles may be particularly useful for understanding, and ensuring, prosocial outcomes following incarceration.

Current Study

As highlighted above, the family plays a central role for youth as they undergo the process of reentry from incarceration. Although family violence appears to be an all too common experience among this population, prior research offers no insight into the relationship between family dynamics and family violence perpetration. To address this important gap in the literature, we use data from the male-only youth subsample of the Serious and Violent Offender Reentry Initiative (SVORI) to examine how family support and family conflict relate to family violence perpetration. Following the preceding discussion of DCSS, we offer two hypotheses:

Hypothesis 1: youth who report higher levels of family support will report lower levels of family violence perpetration post-release

Hypothesis 2: youth who report higher levels of family conflict will report higher levels of family violence perpetration post-release.

Methods

Data

Data for this project come from the youth subsample of the Serious and Violent Offender Reentry Initiative (SVORI, Lattimore and Visher 2009). A federally funded initiative, the goal of SVORI was to examine the effect of enhanced reentry programming on a variety of post-release outcomes in the areas of criminal justice, education, housing, health, and employment. Enhanced reentry programming referred to a wide-range of programming, classes, and treatment aimed at addressing the needs and risk of returning individuals such as substance abuse treatment, anger management prevention classes, reentry preparation, education, and skills training (see Visher et al. 2017 for a detailed overview). As a quasi-experimental design, the SVORI data contain individuals who were identified as SVORI participants as well as individuals who received “business as usual” (Lattimore and Steffey 2009). Generally, individuals were selected to participate in SVORI – both as participants and non-participants – based on criminal history as well as risk (for a detailed description on participant recruitment and selection, see Lattimore and Steffey 2009, p. 36–37). For the youth subsample, individuals were included in the SVORI sample across four different sites with data collection beginning in 2004/2005 and ending in 2007.

The total sample size at wave one includes 337 male youth (average age at wave one = 16.8 years, range = 14 to 22 years) with four distinct waves of panel data.1 Wave one data were collected from each respondent via interview prior to release, typically about 30 days before the scheduled date of release. Wave two data were collected approximately 3 months post-release; wave three data about 9 months post-release; and wave four data were collected about 15 months following release from prison/juvenile detention. About 72% of individuals participated in each wave of data collection (Lattimore and Steffey 2009, p. ES-8). At each wave, respondents were asked a series of questions about their personal experiences and beliefs, family relationships, criminal offending, as well as questions about programming and treatment (see Lattimore and Steffey 2009 for a broad overview of SVORI). In the current analysis, we draw data from all four waves.

Measures

Dependent Measure: Post-Release Family Violence

The dependent measure in the current study is family violence post-release captured at waves two, three, and four. The SVORI data contain five questions assessing family violence. Specifically, respondents were asked, since the previous interview, how often they had engaged in the following behaviors: 1) threatened to hit someone with a fist or anything else that could hurt them, 2) thrown anything at someone that could hurt them; 3) push, grab or shove someone; 4) slap, kick, bite, or hit someone with a closed first; or, 5) threatened to use a weapon on someone. Response categories followed along a six-point ordinal scale (0 = never, 1 = once, 2 = a few times, 3 = about once a month, 4 = a couple of times a month, 5 = once a week, 6 = several times a week). The average Cronbach’s alpha across each post-release wave is approximately .901 indicating high inter-item reliability (Cronbach 1951). To ensure these items captured a unidimensional construct, we also performed an exploratory factor analysis (Kim and Mueller 1978). All individual items loaded at .75 or higher, and the Eigenvalue (3.402) indicated the items loaded onto a single factor. Each item was summed to create a scale capturing post-release family violence. This measure has a mean of 4.042, standard deviation of 6.083 and ranges from 0 (no family violence) to 30 (very high levels of family violence). As this measure varies across time, the within-individual standard deviation is 4.071 highlighting a great deal of within-person change in family violence during the three post-release waves. Due to the significant positive skew in the data, we use the natural logarithm. Descriptive statistics for this measure, as well as all other measures used in the analyses, are shown in Table 1, below.

Table 1.

Descriptive statistics of the serious and violent offender youth sample

Variables Mean SD Range Between SD Within SD n
Family violence
 Post-release family violence 4.042 6.085 0–30 4.991 4.071 337
 Pre-incarceration family violence 6.767 7.396 0–30 – – 337
Incarcerated family measures
 Conflict 6.661 1.704 3–12 – – 330
 Emotional support 10.406 1.775 3–12 – – 332
 Companionship support 11.012 1.359 3–12 – – 334
Post-release family measures
 Conflict 6.348 1.693 3–12 1.475 0.938 289
 Emotional support 9.486 1.771 3–12 1.423 1.101 289
 Companionship support 10.022 1.497 3–12 1.272 0.859 289
 Instrumental support 6.526 1.131 3–8 0.939 0.692 289
Programming
 Personal relationships 0.365 0.481 0, 1 – – 337
 Criminal attitude 0.749 0.433 0, 1 – – 335
 Domestic violence support group 0.095 0.294 0, 1 – – 336
 Mentoring 0.395 0.489 0, 1 – – 337
 Anger prevention 0.570 0.495 0, 1 – – 337
 Education 0.482 0.499 0, 1 – – 332
Additional control measures
 Non-white 0.801 0.399 0, 1 – – 337
 Family visits 2.880 1.134 1–5 – – 334
 Gang member 0.134 0.341 0, 1 – – 336
 Parole violation 0.401 0.49 0, 1 – – 337
 Disciplinary infractions 0.575 0.495 0,1 – – 334
 Middle school 0.246 0.431 0, 1 – – 337
 Self-efficacy 14.421 2.478 4–20 – – 335
 SVORI participant 0.451 0.498 0, 1 – – 337
Open in a new tab

SVORI, Serious and Violent Offender Reentry Initiative; SD, Standard Deviation; n = Valid Cases

Pre-Release Family Violence

Though we discuss our analytic strategy in a subsequent section, to ensure our findings are robust, we include a measure capturing pre-incarceration levels of family violence. This measure was created using identical items as the post-release measure of family violence, but asked the respondent about family violence perpetrated in the six months leading up to incarceration (collected at wave one). Like the dependent measure, a Cronbach’s alpha (.916) as well as a confirmatory factor analysis (all factor loadings exceeding .75) indicated these measures represented a unidimensional and consistent construct. These items were summed to create a scale of pre-release family violence. This measure has a mean of 6.767, standard deviation of 7.396, and ranges along the same scale as the dependent measure (0 to 30).

Family Conflict and Support

To create a measure of family conflict, we draw data from three measures used in prior research (Mowen and Boman 2018). Specifically, respondents were asked to respond to the following three statements: 1) I fight a lot with my family members, 2) I feel like I disappoint my family; and, 3) I am criticized a lot by my family. Respondents could answer along a four-point Likert scale (1 = strongly disagree, 2 = disagree, 3 = agree, 4 = strongly agree). These items were then summed together to create a scale of family conflict, though we separate pre-release and post-release responses (alpha across all four waves = .802). The pre-release (wave one) measure of family conflict has a mean of 6.661, standard deviation of 1.704, and ranges from 3 (low family conflict) to 12 (high family conflict). The post-release measure of family conflict (waves two, three, and four) has a mean of 6.348, standard deviation of 1.693, and ranges from 3 (low family support) to 12 (high family support). The within-individual standard deviation in post-release family conflict is .938 suggesting that levels of family conflict change within-persons across time after release from prison. About 47.7% of the variation in family conflict is between-individuals while 52.3% is within-individuals across time.

In addition to family conflict, we also include measures that encompass family support. Specifically, we draw data from a series of questions assessed on a four-point Likert scale (1 = strongly disagree, 2 = disagree, 3 = agree, 4 = strongly agree) that asked respondents the extent to which they had someone in the family to provide: 1) advice on a place to live, 2) help finding a job, 3) financial support, 4) transportation; questions that asked if they had someone in their family 5) to talk to about their problems, 6) who understands their problems, 7) to turn to for suggestions, 8) who loves me; as well as questions that asked if they: 8) feel close to their family, 9) want family involved in their life, and 10) are a source of support for their family. To examine whether these measures represented unior multi-dimensional measures of family support (e.g., Martinez 2006), we performed an exploratory factor analysis with orthogonal rotation.

The rotated factor analysis, shown in Table 2, revealed that these ten items loaded distinctly onto three factors. The first factor captures family instrumental support and is comprised of measures indicating that the respondent had someone in their family to provide advice on a place to live and help finding a job (alpha = .886). The second factor captures emotional support and is comprised of the three items indicating that the respondent has a family member to talk to about problems, who understands their problems, and to turn to for suggestions (alpha = .867). The final measure captures companionship support and is comprised of three items indicating that the youth feels close to their family, wants family involved in their life, and is a source of support for their family (alpha = .750).

Table 2.

Rotated factor analysis for family support measures

Factor 1 Factor 2 Factor 3
Have someone in my family…
to provide advice on a place to live 0.776 0.314 0.196
to provide help finding a job 0.781 0.254 0.127
to provide financial support 0.327 0.248 0.232
to provide transportation 0.372 0.324 0.156
to talk to about problems 0.294 0.672 0.094
who understands my problems 0.303 0.722 0.219
to turn to for suggestions 0.297 0.744 0.215
who loves me 0.363 0.342 0.288
I feel close to my family 0.250 0.303 0.543
I want family involved in my life 0.295 0.263 0.570
I am a source of support for my family 0.246 0.275 0.461
Eigenvalue 2.250 2.209 1.150
Cronbach’s Alpha 0.886 0.865 0.750
Open in a new tab

Bold indicates the item was used in the corresponding scale

Like the measure of family conflict, we include pre- and post-release measures of family support as separate covariates. Specifically, the pre-release wave one measure of family emotional support has a mean of 10.406, standard deviation of 1.775 and ranges from 3 (low emotional support) to 12 (high emotional support). The mean for pre-release companionship support is 11.012, standard deviation of 1.359, and ranges along the same metric. Because the respondents were incarcerated at wave one, there is no measure of pre-release instrumental support.

The post release (waves two, three and four) measure of emotional support has a mean of 9.486, overall standard deviation of 1.423, and ranges from 3 (low support) to 12 (high support). The within-individual standard deviation is 1.101 and captures the extent to which this measure varies across time. The post-release measure of companionship support has a mean of 10.22, standard deviation of 1.771 and also ranges from 3 (low support) to 12 (high support). The within-individual standard deviation of companionship support is .859. The post-release measure of instrumental support has a mean of 6.526, standard deviation of 1.131, and ranges from 2 (low support) to 8 (high instrumental support). The within-individual standard deviation of instrumental support is .692.

Programming

Although prior research examining the programmatic elements of the SVORI sample has provided mixed-results on the efficacy of programming on reentry outcomes (see Visher et al. 2017), we control for a number of programs which were all assessed at wave one. We include a series of binary measures (1 = yes, 0 = no) capturing whether or not the respondent participated in personal relationship training (36.5% of the sample), programming to change criminal attitudes (74.9% of the sample), domestic violence support groups (9.5% of the sample), received mentoring (39.5% of the sample), participated in anger prevention classes (57.0% of the sample), and/or received educational programming (48.2% of the sample). In the analysis, we contrast those who received these programming elements (1 = yes) compared to those who did not receive these programs (0 = no).

Additional Control Measures

We also control for a variety of demographic characteristics, prior criminal justice history, and personal experiences that may affect family violence. Each were collected at wave one. Specifically, we include a binary measure for race capturing non-white respondents (80.1% of the sample) in contrast to white respondents (19.9% of the sample). We control for family contact during incarceration by including an ordinal measure indicating the extent to which the respondent was visited during incarceration (1 = never, 2 = a few times, 3 = monthly, 4 = weekly, 5 = daily). The mean of this measure is 2.880 indicating that, on average, youth were visited between a few times and monthly during incarceration. The standard deviation of this covariate is 1.134, with a range from 1 (no visits) to 5 (daily visits).

We also include a binary measure capturing whether or not the respondent was a gang member at wave one (1 = yes). Overall 13.4% of respondents indicated they were a member of a gang at wave one. We also include binary measures – also from wave one – that capture whether the respondent had a parole violation (1 = yes) with 40.1% of respondents indicating they had a prior parole violation, and whether the respondent had received a disciplinary infraction during incarceration (1 = yes) with 57.5% of respondents reporting having received a disciplinary infraction. We also include a binary measure indicating whether or not the respondent was enrolled in middle school in contrast to those who reported being in, or have completed, high school. Overall, 24.6% of the sample reported being enrolled in middle school compared to 76.4% who had completed or were enrolled in high school.

We also control for self-efficacy at wave one as some prior work has found that self-efficacy is particularly important for returning persons (Bahr et al. 2010). To create this measure, we identified a total of 14 measures in the data that asked youth a variety of questions concerning beliefs about their abilities, future, and attitudes using a four-point Likert scale (1 = strongly agree, 2 = agree, 3 = disagree, 4 = strongly disagree). We performed an exploratory factor analysis to determine the total number of factors. Results demonstrated the presence of only one factor with five items loading onto this measure. Specifically, these items asked youth how much they agreed/disagree with the following statements: 1) there is no way you can solve some of the problems you have, 2) you feel like you are being pushed around in life, 3) you often feel helpless dealing with the problems of life, 4) you have little control over the things that happen to you, and 5) your life has gone out of control. These items were summed to create a scale of self-efficacy. This measure has a mean of 14.421, standard deviation of 2.478, and ranges from 4 (very low efficacy) to 20 (very high efficacy). Finally, we control for whether or not the respondent was a SVORI participant. Overall, 45.1% of respondents were identified as SVORI program participant.

Analytic Strategy

The SVORI data are longitudinal panel data whereby respondents were repeatedly sampled across four waves of data. This nested design requires an approach that accounts for a lack of independence and correlated error term across time. As Allison (2015) notes, mixed- and fixed-effects models – likely the most popular forms of longitudinal data analysis – are unable to include a lagged measure of the dependent variable in the case of longitudinal panel data due to correlated error terms across waves. Yet, there is often a need to account for a lagged measure of the dependent variable as a control because “baseline” levels of a dependent variable may be a particularly robust predictor of future changes in that measure. A dynamic panel model uses structural equations to meet the assumption of independence thereby allowing for the inclusion of a lagged measure of the dependent variable as a control in the analysis. Specific to our analysis, we account for family violence prior to incarceration to ensure that findings on changes in post-release family violence are not simply an outcome of omitting this key covariate. The dynamic panel model approach uses a series of changed equations whereby the lagged dependent variable at wave one shares a pathway with each progressive outcome (t + 1). The equation for the dynamic panel data model is:

yit=λyit−1+xii˙β+wiδ+αi+εt+vit

Within this approach, yit–1 is the outcome for each individual i at time t, yi0 is the exogenous baseline observation of yit, xit is the vector of time variant measures, wi is the vector of the time invariant measures, δ is the unit of time, αi is the fixed-effect at each wave, εt represents unobserved similarity across waves, and vit is the variant error term (Williams et al. 2018).

Using the dynamic panel model, we proceed in multiple steps. First, we examine the relationships across the measures through correlations and variance inflation factors. Second, we then present the results of the dynamic panel models by examining the relationship between pre-incarceration family violence and family violence post-release. Next, we step in the time invariant measures including the pre-release measures of family relationships, programmatic variables, and all control measures. We then present the full model that includes both pre-release and post-release measures. Finally, we note that there is attrition in the SVORI youth sample. About 72% of the 337 youth at wave one participated in each of the three follow-up waves (Lattimore and Steffey 2009). To retain data, we use full-information maximum likelihood estimation (Allison et al. 2017).

Results

Prior to presenting the results, we first examined the correlations and variance inflation factors among these measures. Results demonstrate that variance inflation is relatively low across all measures with a mean variance inflation factor of 1.39, and no variance inflation actor exceeding 2.11 (O’Brien 2007). Likewise, a correlation matrix revealed that correlations – even among pre- and post-release measures of family conflict and support – were relatively low (approximately .3 across all pre- and post-release family measures).

Results of the dynamic panel data models are shown in Table 3, below. The insignificant chi-square value, as well as the comparative fit index (above .90) indicates that the model fits the data well (Williams et al. 2018). For this baseline model, we note that the root mean squared error of approximation (RMSEA) value of .059 is slightly above the preferred cut off of .05, but the RMSEA improves in subsequent models once covariates are added. Overall, results of this model demonstrate that pre-incarceration measures of family violence do not significantly relate to changes in family violence post-release.

Table 3.

Fixed-effect dynamic panel model examining post-release family violence

Model 1
 Model 2
 Model 3
Variables b SE b SE b SE
Pre-incarceration family violence 0.024 0.050 0.021 0.051 0.012 0.049
Incarcerated family measures
 Conflict – – 0.081 0.030* 0.067 0.032*
 Emotional support – – −0.044 0.031 −0.061 0.032
 Companionship support – – 0.034 0.041 0.018 0.044
Post-release family measures
 Conflict – – – – 0.086 0.043*
 Emotional support – – – – 0.055 0.037
 Companionship support – – – – 0.039 0.060
 Instrumental support – – – – 0.042 0.062
Programming
 Personal relationships – – 0.142 0.098 0.105 0.100
 Criminal attitude – – 0.052 0.117 0.039 0.120
 Domestic violence support group – – 0.220 0.160 0.206 0.163
 Mentoring – – 0.011 0.093 −0.007 0.095
 Anger prevention – – 0.181 0.098 0.194 0.100
 Education – – −0.086 0.090 −0.103 0.091
Additional control measures
 Non-white – – −0.352 0.114* −0.333 0.117*
 Family visits – – −0.039 0.039 −0.037 0.040
 Gang member – – 0.175 0.152 0.164 0.156
 Parole violation – – 0.209 0.095 0.184 0.096
 Disciplinary infractions – – 0.161 0.094 0.140 0.096
 Middle school – – 0.004 0.102 0.020 0.104
 Self-efficacy – – 0.007 0.021 0.013 0.021
 SVORI participant – – 0.108 0.091 0.116 0.093
 Chi-square 10.822 51.083 60.041
 RMSEA 0.059 0.030 0.029
 CFI 0.972 0.946 0.945
Open in a new tab
*

p ≤ .05

b, Unstandardized Coefficient; SE, Standard Error; RMSEA, Root Mean Squared Error of Approximation; CFI, Comparative Fit Index

In model two, we introduce the pre-release measures of family support, programming, and additional control measures. The measures of fit indicate the model fits the data well. The substantive results demonstrate that family conflict during incarceration is significantly related to within-person increases in family violence post-release (p< .05), while family supports are not associated with post-release family violence. Specifically, a one unit higher score on pre-release family conflict is associated with an 8.4% higher score in post-release family violence. Non-white respondents, relative to white respondents, report significantly lower levels of family violence (p < .05). To more intuitively illustrate the effect size of pre-release family conflict on post-release family violence, Fig. 1 presented below plots the predicted value on the non-logged post-release family violence index for individuals with mean, low (two standard deviations below the mean), and high levels of family conflict (two standard deviations above the mean).

Fig. 1.

Fig. 1

Open in a new tab

Bar graph plotting mean, low, and high levels of pre-release conflict and post-release family violence

Model three introduces the post-release family measures. Results of this model demonstrate that post-release family conflict is significantly associated with increased family violence (p < .05). Substantively, this finding shows that within-person increases in family conflict are associated with significant within-person increases in family violence. Specifically, a one unit increase in post-release conflict is associated with an 8.9% increase in post-release family violence. However, post-release measures of family supports are not significantly associated with post-release family violence. To examine the effect sizes of post-release family conflict on family violence, Fig. 2, shown below, shows the expected score on the family violence index for an individual with mean levels of family conflict at wave two who experiences a one standard deviation increase in post-release family violence across waves two, three, and four.

Fig. 2.

Fig. 2

Open in a new tab

Bar graph plotting one standard deviation increase in post-release family conflict and post-release family violence

Overall, these findings suggest that individuals with higher levels of family conflict prior to release experience significantly higher levels of family violence post-release, and increases in family conflict post-release are also associated with higher levels of family violence.

Discussion

Drawing from differential coercion and social support theory (Colvin et al. 2002), the goal of this research was to examine how family support and family conflict related to family violence perpetration among youth undergoing reentry from incarceration. Specifically, we used four waves of data from the youth subsample of the Serious and Violent Offender Reentry Initiative to examine how specific types of pre- and post-release family support and family conflict related to family violence. We now return to our hypothesis and extant literature to explore the findings of the analysis.

Our first hypothesis, that family support would relate to lower levels of post-release family violence perpetration, is not supported. Specifically, results demonstrated that pre-release levels of family emotional and companionship support were not significantly associated with post-release family violence after taking into account a variety of controls. Further, post-release family emotional support, family companionship support, and family instrumental support also failed to significantly relate to family violence after adjusting for a variety of additional factors. This finding stands in contrast to prior research that has highlighted that family support is predictive of more successful reentry experiences among formerly incarcerated juveniles, although prior research has not specifically examined outcomes related to family violence (Martinez and Abrams 2013).

Our second hypothesis premised that family conflict would relate to higher levels of post-release family violence. This hypothesis is supported as the results of a series of dynamic panel data models demonstrated that both pre- and post-release levels of family conflict related to significant increases in family violence during the process of reentry. Specifically, higher levels of family conflict immediately prior to release related to significant increases in post-release family violence. Similarly, youth who experienced increases in family conflict once released reported significant increases in family violence perpetration. This finding is consistent with extant research that has found that family conflict does not promote successful reentry, but instead predicts recidivism (Braman and Wood 2003; Cottle et al. 2001; Mowen and Boman 2018). Although prior studies have not examined family violence as an outcome, our results demonstrate the underlying process affecting recidivism is similar for family violence.

When placed within the context of DCSS (Colvin et al. 2002), we find support for the notion that family conflict – as a mechanism of coercion – is significantly associated with increased risk of family violence perpetration among returning youth. This finding meshes with Martinez and Abrams (2013) who noted that frustration and conflict can lead to youth withdrawing from their prosocial relationships with their family members. Our results suggest that part of this withdrawal may manifest in the form of threatened or actual physical violence against family members. On the other hand, we find little support for the notion that family support reduces the risk of family violence for returning youth. One potential explanation for this is that the outcome investigated in this study – family – may be affected differently by family support than other outcomes that have been studied in prior research such as other forms of reoffending, arrest, and incarceration. However, research on adults has shown that family conflict appears to undermine the protective effects of family support on recidivism (Mowen and Boman 2019b), and research suggests that family conflict may be particularly salient for returning youth (Mowen and Boman 2018). Finally, one area that was not specifically examined in this study was the possible interactive effect of family conflict and support. That is, the influence of family conflict on family violence may be contingent on the presence, or absence, of family support for youth undergoing the process of reintegration (e.g., Mowen and Boman 2019a). Modeling these more complex family dynamics could be a fruitful direction for future research.

Findings from this study carry important policy implications. Perhaps most broadly, findings from this study suggest that family conflict may be a particularly salient factor in promoting deleterious outcomes among returning youth. This finding is bolstered by the fact that our analytical models accounted for levels of family violence prior to incarceration. That is, the influence of pre-incarceration and within-person changes in post-release family conflict on family violence perpetration appears to be independent of pre-existing levels of family violence perpetration. This finding echoes prior work that has marked negative family relationships as particularly problematic for returning youth (e.g., Cottle et al. 2001). From a policy standpoint, research has shown that youth reentry is typified by strained family connections (Braman and Wood 2003; Cottle et al. 2001; Martinez and Abrams 2013) and incarcerated youth often have difficulty in maintaining quality ties to their family members (Monahan et al. 2011). Although we account for the number of times the youth was visited by family members, it is possible that simply being visited is not enough to counteract the formation of family conflict. Instead, it is possible that programs aimed at increasing the quality of contact between the juvenile and their family may lower levels of both pre- and post-release conflict and, therefore, reduce family violence during reentry. Additionally, some prior research has found that reentry programming that contains strong mentoring components has been successful in reducing recidivism among returning youth (Bouffard and Bergsetg 2008), which may include reducing family violence. Although our results demonstrated that participation in a mentoring program was not significantly associated with family violence, it is possible that mentoring programs specifically aimed at reducing family conflict may prove fruitful in reducing family violence.

Perhaps more poignantly, research on multisystemic therapy for adjudicated youth has shown promising results in reducing perpetration of violence (Henggeler et al. 1996). Using a randomized clinical trial design, Henggeler et al. (1993) found that youth who received multisystemic family preservation therapy were far less likely to be re-arrested that youth did not receive multisystemic family preservation therapy. While the study did not specifically analyze family violence as an outcome, other research projects have also found that multisystemic therapy that involves family members promotes prosocial outcomes among high-risk youth (Sawyer and Borduin 2011, see also Perkins-Dock 2001). In this vein, Letourneau et al. (2009, p. 91) note that effective interventions for antisocial behavior in adolescents “share a family-based focus.” To our knowledge, the youth in the SVORI sample did not receive this type of family-based therapy. Based on the wide range of research finding ample support in the effectiveness of family-based therapy in reducing offending among high-risk youth, an additional policy recommendation based on our study is to improve access to family-based interventions among incarcerated youth. Such interventions may effectively reduce family conflict because family functioning and relations are a core focus of this type of therapy. Future research should examine the efficacy of family-based interventions on reducing family conflict during reentry.

Despite the contributions of this project, there are a number of notable limitations. First, the sample used in this study is not representative of all juveniles. For instance, the sample encompassed only male juveniles and, as a result, findings may not be generalizable to female youth. Future research should examine the joint roles of family support and family conflict on family violence among adjudicated female youth. Second, this study’s sample included serious offenders only, and thus, the findings may not generalize to less serious juvenile offenders returning from a term of incarceration. Third, we are unable to account for neighborhood and community characteristics, which is also a notable limitation. For example, in an analysis of data encompassing 562 male and female juveniles undergoing reentry in Florida, Kennedy et al. (2018) noted that while family dynamics were significantly associated with post-release outcomes, neighborhood characteristics were the most robust predictors of reentry success and failure. Future research should explore how neighborhood characteristics might mitigate, or magnify, the relationship between family dynamics and family violence perpetration.

There are additional limitations concerning the amount of time the SVORI data encompass as well as limitations concerning the measurement of key variables in the study. Towards the former, the SVORI data are now over a decade old and thus may not represent current dynamics that juveniles face post-release. Additionally, our study was limited to the first 15 months-post release and, as a result, our findings do not speak to longer term outcomes for youth as they transition into adulthood. Towards the latter, the measure of family violence used in this study is somewhat limited in scope and does not encompass all possible forms of family violence (Gelles 1985). Likewise, some of the measures of family support – such as advice on finding a place to live, or help finding a job – may not be as relevant to youth as their adult counterparts. Future research should consider additional measures of family support that may pertain more explicitly to adolescents. Relatedly, wave one (pre-incarceration) levels of family violence were collected from respondents when they were incarcerated which means that this measure of pre-incarceration family violence may suffer from recall bias. Moreover, the use of a single reporter (i.e., the youth participants) may introduce bias into the variables (e.g., social desirability bias) that could have been improved by triangulating the youths’ responses with other data sources. Finally, while our findings demonstrate that pre-incarceration levels of family violence are not significantly associated with post-release changes in family violence, it is possible that the effect of pre-incarceration family violence may be more salient earlier in the reentry process or more influential among individuals who are incarcerated for shorter amount of time. Future research should explore this possibility.

Conclusion

Overall, findings from this study highlight the pressing need to reduce family conflict among incarcerated and recently released youth to curb family violence perpetration. Yet, the existing focus regarding the role of family within the juvenile justice system is almost entirely on the role of family support. For example, a recent literature review provided by the Office of Juvenile Justice and Delinquency Prevention (2018c) on family engagement in the juvenile justice system focused almost exclusively on establishing and maintaining supportive familial relationships. Yet, as highlighted by this study, the presence of family support is not the same as the absence of family conflict. Researchers, practitioners, and policy-makers alike must recognize that family conflict is a significant risk factor for family violence perpetration among high-risk and adjudicated youth, and that this risk factor exists independently of family support. As highlighted by this research, failure to take effective steps in reducing family conflict among previously incarcerated youth and their families may result in the continuation of high rates of family violence within this population.

Acknowledgements

This research was supported in part by the Center for Family and Demographic Research, Bowling Green State University, which has core funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (P2CHD050959).

Footnotes

1

A subsample that included female youth was initially collected by SVORI researchers, but was not released due to the small sample size and concerns over anonymity.

References

  1. Allison P. (2015). Don’t put lagged dependent variables in mixed models. Statistical horizons. Available at: https://statisticalhorizons.com/lagged-dependent-variables. Accessed 8 Jan 2019. [Google Scholar]
  2. Allison PD, Williams R, & Moral-Benito E. (2017). Maximum likelihood for cross-lagged panel models with fixed effects. Socius: Sociological Research for a Dynamic World, 3, 1–17. 10.1177/2378023117710578. [DOI] [Google Scholar]
  3. Apel R, & Sweeten G. (2010). The impact of incarceration on employment during the transition to adulthood. Social Problems, 57, 448–479. 10.1525/sp.2010.57.3.448. [DOI] [Google Scholar]
  4. Bahr ST, Harris L, Fisher JK, & Armstrong AH (2010). Successful reentry: What differentiates successful and unsuccessful parolees? International Journal of Offender Therapy and Comparative Criminology, 54, 667–692. 10.1177/0306624X09342435. [DOI] [PubMed] [Google Scholar]
  5. Barnert ES, Dubovitz R, Nelson BB, Coker TR, Biely C, Li N, & Chung PJ (2017). How does incarcerating young people affect their adult health outcomes? Pediatrics, 139, 1–9. 10.1542/peds.2016-2624. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bouffard JA, & Bergsetg KJ (2008). The impact of reentry services on juvenile offenders’ recidivism. Youth Violence and Juvenile Justice, 6, 295–318. 10.1177/1541204007313384. [DOI] [Google Scholar]
  7. Braman D, & Wood J. (2003). From one generation to the next: How criminal sanctions are reshaping family life in urban America. In Travis J. & Waul M. (Eds.), Prisoners once removed: The impact of incarceration and reentry on children, families, and communities (pp. 157–188). Washington, DC: Urban Institute. [Google Scholar]
  8. Bullis M, Yovanoff P, Mueller G, & Havel E. (2002). Life on the “Outs” – Examination of the facility-to-community transition of incarcerated youth. Exceptional Children, 69, 7–22. 10.1177/001440290206900101. [DOI] [Google Scholar]
  9. Calley NG (2012). Juvenile offender recidivism: An examination of risk factors. Journal of Child Sexual Abuse, 21, 257–272. 10.1080/10538712.2012.668266. [DOI] [PubMed] [Google Scholar]
  10. Colvin M. (2007). Applying differential coercion and social support theory to prison organizations: The case of the penitentiary of New Mexico. The Prison Journal, 87, 367–387. 10.1177/0032885507304774. [DOI] [Google Scholar]
  11. Colvin M, Cullen FT, & Vander Ven T. (2002). Coercion, social support, and crime: An emerging theoretical consensus. Criminology, 40, 19–42. 10.1111/j.17459125.2002.tb00948.x. [DOI] [Google Scholar]
  12. Cottle CC, Lee RJ, & Heilbrun K. (2001). The prediction of criminal recidivism in juveniles: A meta-analysis. Criminal Justice and Behavior, 28, 367–394. 10.1177/0093854801028003005. [DOI] [Google Scholar]
  13. Cronbach LJ (1951). Coefficient alpha and the internal structure of tests. Psychometrika, 16, 297–334. [Google Scholar]
  14. Cullen FT (1994). Social support as an organizing concept for criminology: Presidential address to the academy of criminal justice sciences. Justice Quarterly, 11, 527–559. 10.1080/07418829400092421. [DOI] [Google Scholar]
  15. Dmitrieva J, Monahan KC, Cauffman E, & Steinberg L. (2012). Arrested development: The effects of incarceration on the development of psychosocial maturity. Development and Psychopathology, 24, 1073–1090. 10.1017/S0954579412000545. [DOI] [PubMed] [Google Scholar]
  16. Gelles RJ (1985). Family violence. Annual Review of Sociology, 11, 347–367. 10.1146/annurev.so.11.080185.002023. [DOI] [Google Scholar]
  17. Henggeler SW, Melton GB, Smith LA, Choenwald SK, & Hanley JH (1993). Family preservation using multisystemic treatment: Long-term follow-up to a clinical trial with serious juvenile offenders. Journal of Child and Family Studies, 2, 283–293. 10.1007/BF01321226. [DOI] [Google Scholar]
  18. Henggeler SW, Cunningham PB, Pickrel SG, Schoenwald SK, & Brondino MJ (1996). Multisystemic therapy: An effective violence prevention approach for serious juvenile offenders. Journal of Adolescence, 19, 47–61. 10.1006/jado.1996.0005. [DOI] [PubMed] [Google Scholar]
  19. Kennedy TD, Edmonds WA, Millen DH, & Detullio D. (2018). Chronic juvenile offenders: Exploring risk factor models of recidivism. Youth Violence and Juvenile Justice (Online first) at: 10.1177/1541204018770517. [DOI] [Google Scholar]
  20. Kim J, & Mueller CW (1978). Factor analysis: Statistical methods and practical issues. New York: Sage. [Google Scholar]
  21. Lambie I, & Randell I. (2013). The impact of incarceration on juvenile offenders. Clinical Psychology Review, 33, 448–459. 10.1016/j.cpr.2013.01.007. [DOI] [PubMed] [Google Scholar]
  22. Lattimore PK, & Steffey DM (2009). The multi-site evaluation of SVORI: Methodology and analytic approach. U.S. Department of Justice. Document 230424. [Google Scholar]
  23. Lattimore PK, & Visher CA (2009). Multi-site evaluation of SVORI: Summary and synthesis. Raleigh: RTI International. Document 230421. [Google Scholar]
  24. Letourneau EJ, Henggeler SW, Borduin CM, Schewe PA, McCart MR, Chapman JE, & Saldana L. (2009). Multisystemic therapy for juvenile sexual offenders: 1-year results from a randomized effectiveness trial. Journal of Family Psychology, 23, 89–102. 10.1037/a0014352. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Martinez DJ (2006). Informal helping mechanisms: Conceptual issues in family support of reentry of former prisoners. Journal of Offender Rehabilitation, 44, 23–37. 10.1300/J076v44n01_02. [DOI] [Google Scholar]
  26. Martinez DJ, & Abrams LS (2013). Informal social support among returning youth offenders: A metasynthesis of the literature. International Journal of Offender Therapy and Comparative Criminology, 57, 169–190. 10.1177/0306624X11428203. [DOI] [PubMed] [Google Scholar]
  27. Monahan KC, Goldweber A, & Cauffman E. (2011). The effects of visitation on incarcerated juvenile offenders: How contact with the outside impacts adjustment on the inside. Law and Human Behavior, 35, 143–151. 10.1007/s10979-010-9220-x. [DOI] [PubMed] [Google Scholar]
  28. Moore T, McArthur M, & Saunders V. (2013). Young people talk about transitioning from youth detention to the community: Making good. Australian Social Work, 66, 328–343. 10.1080/0312407X.2012.752020. [DOI] [Google Scholar]
  29. Mowen TJ, & Boman JH IV. (2018). A developmental perspective on reentry: Understanding the causes and consequences of family conflict and peer delinquency during adolescence and emerging adulthood. Journal of Youth and Adolescence, 47, 275–289. 10.1007/s10964-017-0794-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Mowen TJ, & Boman JH IV. (2019a). Do we have it all wrong? The protective roles of peers and criminogenic risks from family during prison reentry. Crime & Delinquency, 65, 681–704. 10.1177/0011128718800286. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Mowen TJ, & Boman JH IV. (2019b). The criminogenic influence of family on substance use during reentry: A life-course perspective on between individual differences and within individual changes. Justice Quarterly, 36, 841–869. 10.1080/07418825.2018.1439518. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Mulvey EP (2011). Highlights from the pathways to desistance: A longitudinal study of serious adolescent offenders. Washington D.C: Office of Juvenile Justice and Delinquency Prevention. Available at: https://www.ncjrs.gov/pdffiles1/ojjdp/230971.pdf. Accessed 8 Jan 2019. [Google Scholar]
  33. O’Brien RM (2007). A caution regarding rules of thumb for variance inflation factors. Quality & Quantity, 41, 673–690. 10.1007/s11135-006-9018-6. [DOI] [Google Scholar]
  34. Office of Juvenile Justice and Delinquency Prevention. (2017). Literature review: A product of the model programs guide, Juvenile reentry. Washington D.C. Available at: https://www.ojjdp.gov/mpg/litreviews/Aftercare.pdf. Accessed 8 Jan 2019. [Google Scholar]
  35. Office of Juvenile Justice and Delinquency Prevention. (2018a). Juveniles in corrections: Demographics. Washington D.C. Available at: https://www.ojjdp.gov/ojstatbb/corrections/qa08203.asp?qaDate=2015&text=no&maplik=link3. Accessed 8 Jan 2019. [Google Scholar]
  36. Office of Juvenile Justice and Delinquency Prevention. (2018b). Statistical briefing book. Washington D.C. Available at: https://www.ojjdp.gov/ojstatbb/corrections/qa08201.asp?qaDate=2016. Accessed 8 Jan 2019. [Google Scholar]
  37. Office of Juvenile Justice and Delinquency Prevention. (2018c). Family engagement in juvenile justice. Washing D.C. Available at: https://www.ojjdp.gov/mpg/litreviews/Family-Engagement-in-Juvenile-Justice.pdf. Accessed 8 Jan 2019. [Google Scholar]
  38. Panuccio EA, Christian J, Martinez DJ, & Sulivan ML (2012). Social support, motivation, and the process of juvenile reentry: An exploratory analysis of desistance. Journal of Offender Rehabilitation, 51, 135–160. 10.1080/10509674.2011.618527. [DOI] [Google Scholar]
  39. Perkins-Dock RE (2001). Family interventions with incarcerated youth: A review of the literature. International Journal of Offender Therapy and Comparative Criminology, 45, 606–625. 10.1177/0306624X01455006. [DOI] [Google Scholar]
  40. Sawyer AM, & Borduin CM (2011). Effects of multisystemic therapy through midlife: A 21.9-year follow-up to a randomized clinical trial with serious and violent juvenile offenders. Journal of Consulting and Clinical Psychology, 79, 643–652. 10.1037/a0024862. [DOI] [PubMed] [Google Scholar]
  41. Sedlak AJ, & Bruce C. (2010). Youth’s characteristics and backgrounds: Findings from the survey of youth in residential placement. Office of Juvenile Justice and Delinquency Prevention. D.C.: Washington: Available at: https://www.ncjrs.gov/pdffiles1/ojjdp/227730.pdf. Accessed 8 Jan 2019. [Google Scholar]
  42. Snyder HN, & Sickmund M. (2006). Juvenile offenders and victims: 2006 National Report. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention. [Google Scholar]
  43. Spaccarelli S, Coatsworth JD, & Bowden BS (1995). Exposure to serious family violence among incarcerated boys: Its association with violent offending and potential mediating variables. Violence and Victims, 10, 163–181. 10.1891/0886-6708.10.3.163. [DOI] [PubMed] [Google Scholar]
  44. Steinberg L, Chung HL, & Little M. (2004). Reentry of young offenders from the justice system: A developmental perspective. Youth Violence and Juvenile Justice, 2, 21–38. 10.1177/1541204003260045. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Sullivan ML (2004). Youth perspectives on the experience of reentry. Youth Violence and Juvenile Justice, 2, 56–71. 10.1177/1541204003260047. [DOI] [Google Scholar]
  46. Visher CA, Lattimore PM, Barrick K, & Tuller S. (2017). Evaluating the long-term effects of prisoner reentry services on recidivism: What types of services matter? Justice Quarterly, 34, 136–165. 10.1080/07418825.2015.1115539. [DOI] [Google Scholar]
  47. Western B, Braga AA, Davis J, & Sirois C. (2015). Stress and hardship after prison. American Journal of Sociology, 120, 1512–1547. 10.1086/681301. [DOI] [PubMed] [Google Scholar]
  48. Williams R, Allison PD, & Moral-Benito E. (2018). Linear dynamic panel-data estimation using maximum likelihood and structural equation modeling. In Forthcoming in The Stata Journal Available at: https://www3.nd.edu/~rwilliam/dynamic/SJPaper.pdf. Accessed 8 Jan 2019. [Google Scholar]

RESOURCES