Improvement in Emotion Regulation While Detained Predicts Lower Juvenile Recidivism

Abstract

The goal of the current study was to investigate the relationships between observer-rated skills related to emotional and cognitive regulation post-admission and pre-release in a secure facility and official records of juvenile felony recidivism up to 1 year after release. Data came from a sample of 599 youth in a residential facility in Washington state (84% male; 38% White). Latent change score models indicated that both initial level of emotional regulation skills and improvement in emotion regulation skills while incarcerated were significantly related to lower recidivism. This pattern of findings remained when controlling for length of stay, among other covariates. Follow-up analyses indicated that the results for emotion regulation skills might be driven primarily by monitoring internal and external triggers. Additional research should investigate the connection between emotion regulation skills and juvenile recidivism, with a special focus on trigger monitoring and how to improve those skills.

Keywords

juvenile offending detention recidivism self-regulation

Introduction

Although incarceration of youth involved in the juvenile justice system has been on the decline in the U.S. in recent decades, it remains a persistent public health and safety issue. The one-day count of youth in residential placement facilities was nearly 44,000 in 2017, and this number has decreased from over 108,000 in 2000, a decline of roughly 60% (Sawyer, 2019). This drop in incarceration rate is due in part to efforts taken to reduce incarceration, including increased efforts to find alternatives to incarceration, changes in state policy, and shutdown of youth confinement facilities (NJJN, 2014). Despite these improvements, incarceration rates in the U.S. are much higher than rates in other countries (Laird, 2021), suggesting an over-reliance on incarceration as a means of addressing crime. Additionally, although incarceration rates have declined as a whole, there are considerable racial disparities in juvenile incarceration rates (Sawyer, 2019).

While incarceration is detrimental for everyone, juvenile incarceration is particularly harmful for youth due to its impact on their development. A prison environment fails to address vital psychological, educational, and health needs during adolescence (Barnert et al., 2016; Dmitrieva et al., 2012; Lambie & Randell, 2013). If the purpose of juvenile incarceration is to rehabilitate offenders, it often fails to live up to this promise in practice. There are few positive effects on reducing crime but significant negative effects on mental health and future behavior, including increased risk of reoffending (Aizer & Doyle, 2013; Lambie & Randell, 2013). Despite the recent reduction in juvenile incarceration rates, juvenile recidivism rates remain high (Seigle et al., 2014), motivating researchers and policy experts to better understand and reduce juvenile recidivism.

Understanding the risk factors that increase the risk of juvenile recidivism can inform prevention and intervention efforts aimed at reducing youths’ levels of these risk factors. A multitude of risk factors have been identified in past research, including past criminal behavior (e.g., number of offenses and age of first offense; Cottle et al., 2001), individual differences (e.g., personality characteristics: Cuevas et al., 2019; van Dam et al., 2005), and environmental influences (e.g., antisocial behavior of peers and family members: Cottle et al., 2001; Mulder et al., 2011). However, targeting environmental or individual risk factors within a secure facility can be difficult, as many are challenging to address within the context of a residential program (e.g., exposure to delinquent peers), and others may be more resistant to change (e.g., antisocial personality).

Comparatively, less research has examined skills that may serve as protective factors in reducing juvenile recidivism, despite the relative ease with which skills training can be incorporated into rehabilitation programs. For example, previous research has indicated that skills related to self-control and self-regulation (e.g., impulse control, temperance, mindfulness, identifying, and appropriately expressing emotions) are associated with decreased problem behaviors, including delinquency and recidivism among youth (Baglivio et al., 2016; Cauffman et al., 2005; Steinberg & Cauffman, 1996). Thus, these skills have been identified as important targets for interventions with offenders, particularly for trauma-informed care (Brazão et al., 2018; Ford & Blaustein, 2013). Previous research has indicated that self-regulation encapsulates many domains of functioning, especially cognitive (e.g., effortful control) and emotional (e.g., emotional lability or reactivity) regulation skills. If skills that are related to cognitive and emotional regulation can serve as protective factors against recidivism, improving these skills could potentially lead to decreased juvenile recidivism. The present study will examine how initial levels of and changes in cognitive and emotional regulation skills while incarcerated are associated with juvenile recidivism.

Self-Regulation and Crime

Self-regulation is defined as the ability to monitor, inhibit, and change behavior, attention, emotions, and cognition accordingly to achieve personal goals (Moilanen & Raffaelli, 2005). This is a broad definition that encompasses many characteristics and skills, including self-control (Mamayek et al., 2017). Self-control has a more behavioral focus and is defined as the ability to replace an action with one that helps achieve another goal (Carver & Scheier, 2010) or the tendency to avoid actions that have long-term negative consequences but positive short-term consequences (Hirschi & Gottfredson, 1994). Related constructs that overlap with self-control and may be encompassed by definitions of self-regulation include effortful control (Eisenberg et al., 2004), impulse control, and temperance, which are all concerned with the ability to regulate attention, behaviors, emotions, and thoughts.

Gottfredson and Hirschi (1990)’s general theory of crime is one of the prevailing theories in the criminology literature for explaining antisocial behavior. According to this theory, crime is caused by low self-control, which is characterized by impulsivity, risk-seeking, low tolerance for frustration, short-sightedness, and a tendency to be self-centered. This view proposes that criminal behavior can be appealing in providing immediate gratification and that only those individuals with sufficient self-control will be able to refrain from capitalizing on opportunities to commit crimes. This theory has been supported in research, with low self-control consistently emerging as a strong predictor of crime (Pratt & Cullen, 2000; Vazsonyi et al., 2017).

Although the general theory of crime originally proposed that self-control was established early in life and that levels of self-control remained relatively stable throughout one’s life (Gottfredson & Hirschi, 1990), recent research has provided evidence for within-individual variability in self-control over time (Burt et al., 2006; Hay & Forrest, 2006; Higgins et al., 2009; Ray et al., 2013). In addition, changing circumstances (e.g., parental socialization, exposure to prosocial, or deviant peers) may explain changes in self-control in adolescence (Burt et al., 2006; Hay & Forrest, 2006; Ray et al., 2013). Thus, if self-control is related to offending, and individuals’ self-control can improve, then it remains possible that improvements in self-control or related characteristics could be related to decreased risk of offending. Indeed, one study with youth in residential placements in Florida found that for female adolescents, improvement in self-control while detained was associated with reduced recidivism within 1 year of release, although no such effect was found for male offenders (Hay et al., 2018). Additionally, a meta-analysis of 34 studies with children ages 3–10 found a positive effect of self-control interventions on both self-control and delinquency (Piquero et al., 2010).

A related theory linking self-regulation and crime is the temperament-based theory of effortful control and negative emotionality advanced by DeLisi and Vaughn (Baglivio et al., 2016; DeLisi & Vaughn, 2014). This theory posits that effortful control is another term for self-control and that individuals with a temperament characterized by low effortful control and high negative emotionality tend to be more difficult and reactive and are subsequently more likely to engage in aggressive and delinquent behaviors. Another theory that has been more prominent in the developmental psychology literature is the dual systems model of adolescent brain development, which proposes two neurobiological systems: an early-emerging socioemotional system that is sensitive to rewards, and a later-emerging cognitive control system that is able to control impulses and regulate emotions (Steinberg, 2010). The gap between the development of these two systems, according to this framework, explains increased sensation seeking and risk-taking in adolescence. Thus, theories in both psychology and criminology agree that self-regulation can be a key predictor of juvenile recidivism and a potential target for intervention.

Cognitive and Emotion Regulation

While previous research has indicated that constructs related to self-regulation can be important predictors of juvenile recidivism (Pihet et al., 2011; Ray et al., 2016), few studies have attempted to differentiate aspects of self-regulation in order to understand whether some components are more strongly related to delinquency than others. Some aspects of self-regulation may be more related to cognitive regulation, such as future orientation and attentional control, while others may be more indicative of emotion regulation, such as coping with negative emotions. For example, in a study examining relationships between aspects of psychosocial maturity—a concept related to regulation that includes responsibility, perspective, and temperance—and risk-taking in emerging adults, impulse control and suppression of aggression were related to lower levels of self-reported risky behaviors such as criminal behavior or texting and driving, whereas responsibility and consideration of others were generally unrelated to risk-taking (Romaine, 2019). Another study examined developmental trajectories of antisocial behavior and aspects of psychosocial maturity from adolescence to young adulthood in a sample of 1170 male youth involved in the juvenile justice system and found that improvements in impulse control and suppression of aggression were associated with decreased antisocial behavior, whereas there were no clear associations for other aspects of psychosocial maturity, such as personal responsibility and resistance to peer influence (Monahan et al., 2009). Previous research has identified the unique contributions of alexithymia, impulsivity, and emotion dysregulation in predicting aggression (Garofalo et al., 2018), and implicated emotion (dys)regulation as a mediating process between other cognitive/attentional constructs (e.g., impulsivity, mindfulness, and moral disengagement) and aggression (Garofalo et al., 2016, 2018, 2019; Long et al., 2014).

Although many aspects of self-regulation will have both cognitive and emotional components, identifying the particular skills that might be associated with more cognitive- or emotional-focused regulation, and their associations with recidivism, can help to inform interventions. For example, cognitive regulation might result in improved ability to regulate attention and reign in impulsive thoughts, while emotional regulation would allow for greater control over negative emotions resulting from frustration. Therefore, the current study will differentiate between cognitive and emotional regulation in predicting recidivism.

The Residential Positive Achievement Change Tool (R-PACT)

The Residential Positive Achievement Change Tool (R-PACT) is a risk assessment designed to measure several static and dynamic risks and needs among juveniles in residential facilities. The R-PACT was developed in 2007 to modify an existing risk assessment (PACT) to obtain relevant information from youth who were living in facilities. The R-PACT has been used in juvenile facilities in Florida and Washington, and previous studies have generally examined its predictive validity among juveniles in Florida (Baglivio & Wolff, 2019; Hay et al., 2016). In addition, one study conducted in Washington examined the association between skills acquisition according to the R-PACT and recidivism after accounting for length of stay (Walker & Bishop, 2016). These studies have generally supported the validity of the R-PACT and found that skills assessed by the R-PACT predict lower risk of recidivism.

However, previous studies using the R-PACT have generally not separated skills that might be more closely related to emotional regulation from those that might be more closely related to cognitive regulation. One study that did separate some of the skills measured by a related assessment (the C-PACT) distinguished between skills related to effortful control and skills for dealing with social situations; however, this study did not use a factor analysis to determine how to separate items and did not compare associations of the two different types of skills with recidivism (Baglivio et al., 2016). Therefore, the current study will first conduct exploratory and confirmatory factor analyses to determine whether a set of skills measured by the R-PACT have a two-factor structure—with the factors corresponding to skills related to emotional and cognitive regulation—and compare the strength of the associations between each of these scales and recidivism, while controlling for other relevant factors.

The Current Study

This study examines how changes in skills related to cognitive and emotional self-regulation while incarcerated predict felony recidivism within 1 year of release among a sample of 599 youth at a juvenile detention facility. Skills related to self-regulation are measured using a modified version (the Integrated Treatment Assessment, or ITA) of a relatively widely used interviewer-rated risk measure in juvenile agencies, the R-PACT, both after admission to a secure facility and just prior to release from the facility. For the purposes of this study, and to maintain consistency with a previous study using these data (Walker & Bishop, 2016), we will refer to this measure as the R-PACT. Our goals are to (1) determine whether skills assessed by the R-PACT can be differentiated by whether they more closely relate to cognitive or emotional regulation and (2) to examine the associations of initial levels of and changes in skills related to cognitive and emotional regulation while incarcerated with felony recidivism within 1 year of release. Although there may be better potential measures of emotional and cognitive regulation than the R-PACT, the ability to use a widely administered assessment to examine emotional and cognitive regulation may prove useful for researchers and practitioners working with youth in residential facilities. We hypothesize that a two-factor solution will provide the best fit for skill items on the R-PACT, such that one factor will contain items more strongly related to cognitive regulation and the other factor will consist of items more strongly related to emotional regulation. We also hypothesize that greater initial levels of skills related to both cognitive and emotional regulation will predict lower risk of felony recidivism and that improvements in both sets of skills while incarcerated will also predict lower risk of recidivism, while accounting for potential confounds such as offense history and length of stay in the facility. While we acknowledge that there are other ways to measure recidivism, we chose felony recidivism within 12 months because this was readily available in the data and is consistent with similar studies of juvenile recidivism using the R-PACT (e.g., Baglivio & Wolff, 2019; Baglivio et al., 2016; Hay, 2013; Hay et al., 2016; Hay et al., 2018; Walker & Bishop, 2016).

Method

Participants and Procedures

The data for this study came from a sample of youth in a detention facility in Washington state and were collected for a previous study (Walker, 2016; Walker & Bishop, 2016). Institutional Review Board (IRB) approval was obtained for secondary analysis of these data from the authors’ institution, Bowling Green State University. This dataset combined data from Juvenile Justice Rehabilitation Administration (JJRA) records, the Administrative Office of the Courts (AOC) court contact database, and R-PACT assessments completed by 10 different raters, both within 30–45 days of intake and 30 days before release. JJRA records were used to obtain admission and release dates from secure facilities, AOC data were used to measure felony recidivism within 12 months post-release, and the R-PACT provided all other data used in this study.

Of the original 637 youth in the main study sample, all youth had R-PACT data, but only 599 had recidivism data for the full 12 months post-release. Therefore, data from all 637 youth were used in factor analyses for regulation skills, but data from 599 youth were used in models predicting recidivism. Admission dates ranged from December 5, 2008 to May 29, 2013, and release dates ranged from February 12, 2009 to August 1, 2013. Youth ranged from 11 to 19 years of age (M = 15.87, SD = 1.36). Most of the participants (84%) identified as male. Self-reported race/ethnicity information revealed the sample to be 38% White, 27% African American, 16% Hispanic, 3% Asian, 3% Native American, 2% other, and 11% reporting multiple racial/ethnic identities. 98% of youth reported English as their primary language. Almost all of the youth (n = 606) came in with a felony referral at admission, nearly all (n = 618) had a previous detention disposition, and about two-thirds (n = 419) had a felony referral for a crime against a person.

R-PACT data were obtained from a second cohort of youth who were admitted to facilities between April 2013 and February 2015 (N = 397). These youth did not have data on post-release recidivism, and therefore were not included in the main study sample or main analyses. However, they did have R-PACT data at intake and release. Therefore, we used data from this sample to conduct confirmatory factor analyses, after conducting exploratory factor analyses with data from the main sample. These youth appeared to be overall similar to youth in the main sample, as they were 15.59 years old on average (SD = 1.56; range = 10–19), most (84%) were male, and many identified as White (51%), African American (14%), or Hispanic (18%), with smaller numbers reporting multiple racial/ethnic identities (10%) or a different racial/ethnic identity (other; 7%). We subsequently refer to the 637 youth from the main analyses as cohort 1 and the additional 397 youth as cohort 2.

Measures

Recidivism

Recidivism was determined by whether the participants received a felony charge within 12 months of being released. This was coded as a binary variable, with 0 indicating no felony charge, and 1 indicating at least one felony charge within the 12-month period (although see sensitivity analyses for results with a count outcome).

Skills Related to Regulation

To assess skills related to regulation, we conducted exploratory and confirmatory factor analyses on a pool of 14 R-PACT items from the Skills domain that were related to cognitive and emotional self-regulation. These items included consequential thinking, goal setting, solving problems, situational perception, dealing with others, dealing with difficult situations, dealing with feelings or emotions, monitoring of internal triggers, monitoring of external triggers, control of impulse behaviors, control of aggression, dealing with distress, and staying in the moment. Each item was rated on its own 4-point scale, except for control of aggression and dealing with distress, which were each rated on their own 5-point scale. For example, the rater options for control of aggression were as follows: 5, no problem with aggression; 4, often uses alternatives to aggression; 3, sometimes uses alternatives to aggression; 2, rarely uses alternatives to aggression; and 1, lacks alternatives to aggression. Items were coded such that higher scores reflected higher levels of that particular skill. The factor analyses and the resulting scales that were used in analysis are described under Results.

Though interrater reliability on the R-PACT was not available for the current study (youth were assessed by only one rater at each time point), a previous study found the PACT’s interrater reliability on social history items (e.g., history of mental health problems, history of neglect) to be relatively high (Winokur-Early et al., 2012). Staff raters receive 1 day of training on how to administer and rate the R-PACT and have to demonstrate competence with a supervisor before administering the R-PACT.

Covariates

In models predicting recidivism, we included control variables that may be related to both regulation skills and recidivism, to test the unique predictive association between regulation skills and recidivism. These controls included demographic characteristics (e.g., age, gender (male or female), race/ethnicity (White, Black, Hispanic, multiracial, or other), offense history (e.g., age at first offense; under 13, 13–14, 15, or 16–17), felony history (0, 1, 2, or 3+ previous felony referrals), gang involvement (0 = no gang involvement, 1 = gang involvement), length of stay in months, and family and socioeconomic status (e.g., lack of positive adult relationships, whether a youth’s mother and father had been incarcerated, whether youth live with their mother and father, and annual household income [<$15k, $15–$35k, or >$35k]) as covariates in the present study. We also included other potential risk factors for recidivism, including history of alcohol use, history of drug use, victimization, witnessing violence, ADHD history, and history of mental illness. All covariates came from the R-PACT.

Analysis Plan

First, to create measures of skills related to cognitive and emotional regulation, we conducted exploratory factor analyses with the R-PACT skills data from cohort 1 (N = 637) at time 1 (admission) and time 2 (prior to release). Based on results of these analyses, we then conducted confirmatory factor analyses with both cohort 1 and cohort 2 (N = 397) at each time point, to establish a model that adequately fits the data. We expected that two factors—one factor consisting of skills related to cognitive regulation, and another factor consisting of skills related to emotional regulation—would emerge from the exploratory factor analyses and would fit the data well in confirmatory factor analyses, but we did not know a priori which of the 14 initial items would be associated with each factor. We then examined descriptive statistics and correlations for individual items, including mean levels at each time point and bivariate correlations with 12-month felony recidivism.

Next, to examine how initial levels of skills related to regulation at admission and change in these skills from admission to release predict recidivism, we conducted a latent change score model. Latent change score models (also known as latent difference score models) are conducted within a structural equation modeling framework and offer the ability to model both initial level and change over time, as well as the correlation among level and change, for multiple variables simultaneously (Kievit et al., 2018; McArdle & Grimm, 2010). In this study, we conducted latent variable or multiple indicator latent change score models, such that skills related to cognitive and emotional regulation at both time points were themselves latent variables measured by their respective manifest indicators, thereby reducing measurement error for these constructs. Data management and descriptive statistics were conducted in Stata 16 (StataCorp, 2019), and latent change score models were conducted in Mplus 8.0 (Muthén & Muthén, 1998-2017).

Results

Measurement Model

First, we used a combination of exploratory factor analysis and confirmatory factor analysis to develop a measurement model for our latent cognitive regulation and emotional regulation variables based on the observed items. The first step was running separate exploratory factor analyses based on polychoric correlation matrices for ordinal data at each time point in cohort 1, extracting two factors, and using varimax rotation to obtain rotated factor loadings. The factor loadings from these analyses are reported in Table 1. Because the indicator for solving problems had a stronger loading on the emotional factor at time 1 and a stronger loading on the cognitive factor at time 2, we decided to drop this item from further analysis. We also dropped the indicator for dealing with difficult situations, as it had relatively high (.45–.60) loadings on both factors at both time points.

Table 1.

Factor Loadings from Exploratory Factor Analysis at Both Time Points.

Indicator	Time 1		Time 2
Indicator	Emotional	Cognitive	Emotional	Cognitive
Consequential thinking	.225	. 5 16	.277	.676
Goal setting	.190	.701	.316	.618
Solving problems	.541	.409	.343	.733
Situational perception	.293	.677	.327	.723
Dealing with others	.257	.558	.475	.510
Dealing with difficult situations	.593	.515	.604	.448
Dealing with emotions	.657	.419	.617	.446
Monitoring of internal triggers	.801	.199	.739	.340
Monitoring of external triggers	.790	.156	.747	.236
Control of impulses	.499	.330	.589	.375
Control of aggression	.527	.340	.666	.293
Dealing with distress	.632	.445	.688	.385
Staying in the moment	.288	.687	.525	.560
Sum of squared loadings	3.591	3.095	4.016	3.417
Proportion of variance	.276	.238	.309	.263

Factor loadings greater than 0.40 are bolded.

Next, we conducted a confirmatory factor analysis separately for each cohort and each time point. Because the observed indicators were categorical, all models were estimated using a weighted least square mean and variance adjusted estimator (WLSMV) with delta parameterization and probit link. First, for each cohort at each time point we estimated a one-factor model for comparison. Then we estimated a two-factor model for each cohort at each time point, with the five-item latent cognitive factor (consequential thinking, goal setting, situational perception, dealing with others, and staying in the moment) and six-item latent emotional factor (dealing with emotions, control of impulses, control of aggression, dealing with distress, monitoring internal triggers, and monitoring external triggers) that we found from the exploratory factor analysis, and a covariance estimated between the two latent factors. Finally, we examined modification indices for each of the two-factor models and found large (>200) modification indices for including a residual covariance between the two monitoring trigger indicators, so we included this residual covariance in subsequent models. The fit indices from these confirmatory factor analyses are presented in Table 2 and indicate that the two-factor model with the residual covariance included provided the best fit to the data, although it had relatively large RMSEA values for cohort 1 at time 2 and for cohort 2 at both time points, and a significant chi-square test at both time points for both cohorts. Therefore, we chose to include the two-factor model with the residual covariance as our measurement model for the latent variables in subsequent analyses.

Table 2.

Confirmatory Factor Analysis Model Fit Statistics.

Model	Model fit indices
Model	χ² (df)	CFI	TLI	RMSEA	SRMR
One-factor model
Cohort 1 time 1	974.16 (44)***	.92	.90	.18	.09
Cohort 1 time 2	953.28 (44)***	.94	.93	.18	.06
Cohort 2 time 1	728.73 (44)***	.91	.89	.20	.09
Cohort 2 time 2	918.31 (44)***	.93	.92	.22	.11
Two-factor model
Cohort 1 time 1	560.27 (43)***	.95	.94	.14	.06
Cohort 1 time 2	693.30 (43)***	.96	.95	.15	.05
Cohort 2 time 1	590.27 (43)***	.93	.91	.18	.08
Cohort 2 time 2	832.27 (43)***	.94	.92	.22	.11
Two-factor model with residual covariance
Cohort 1 time 1	149.09 (42)***	.99	.99	.06	.03
Cohort 1 time 2	328.26 (42)***	.98	.98	.10	.03
Cohort 2 time 1	400.78 (42)***	.95	.94	.15	.07
Cohort 2 time 2	641.96 (42)***	.95	.94	.19	.10

Note. N = 637 for cohort 1 and 397 for cohort 2. Models were estimated using a weighted least squares mean variance adjusted (WLSMV) estimator, with delta parameterization and a probit link.

Descriptive Statistics and Correlations

Table 3 displays means and SDs for each of the 11 indicators for cognitive and emotional regulation and length of stay in months, as well as bivariate correlations with felony recidivism within 12 months of release, for cohort 1. None of the regulation skill indicators at time 1 was significantly correlated with recidivism, and neither was length of stay (all ps > .092). However, seven of the regulation skill items at time 2 were significantly negatively associated with felony recidivism: consequential thinking (r = −.09, p = .028), dealing with emotions (r = −.11, p = .009), dealing with distress (r = −.13, p = .002), control of impulses (r = −.12, p = .004), control of aggression (r = −.13, p = .002), monitoring internal triggers (r = −.19, p < .001), and monitoring external triggers (r = −.18, p < .001). Interestingly, only one of the items assessing a skill related to cognitive regulation at time 2 was significantly associated with felony recidivism, compared to six items assessing skills related to emotional regulation. Mean improvements in each of the 11 indicators appeared to be significant, as indicated by paired sample t-tests comparing the means (all t-values > 15, all ps < .001) as well as chi-square tests using the categorical responses (all chi-squares > 45, all ps < .001).

Table 3.

Descriptive Statistics and Correlations with Recidivism for Regulation Indicators.

	Time 1			Time 2
	Mean	SD	Correlation with recidivism	Mean	SD	Correlation with recidivism
Consequential thinking	2.23	.62	−.02	2.71	.72	−.09*
Goal setting	2.40	1.16	−.01	3.13	.85	−.06
Situational perception	2.15	.85	−.06	2.76	.76	−.05
Dealing with others	2.15	.71	.01	2.64	.73	−.06
Staying in the moment	1.99	1.01	−.00	2.80	.81	−.07
Dealing with emotions	1.69	.84	.02	2.65	.73	−.11**
Dealing with distress	1.58	.88	−.01	2.65	.79	−.16***
Control of impulses	1.28	.62	.01	2.08	.61	−.08
Control of aggression	1.84	1.11	.01	2.77	.94	−.13**
Monitoring of internal triggers	1.61	.76	−.01	2.61	.80	−.17***
Monitoring of external triggers	1.73	.74	.01	2.77	.73	−.17***
Length of stay in months	6.09	4.88	−.06	N/A	N/A	N/A
Recidivism (any conviction)	.29	.45	N/A	N/A	N/A	N/A

*p < .05, ** p < .01, ***p < .001. N = 637, except for recidivism descriptives and correlations (N = 599). All regulation variables ranged from 1 to 4, except for control of aggression and dealing with distress, which ranged from 1 to 5. Length of stay in month ranged from 0 to 39. Recidivism was a binary variable (0 = no felony conviction, 1 = any felony conviction). All differences in regulation variables from time 1 to time 2 were significant (p < .001), using either paired t-tests or chi-squared tests.

We also examined descriptive statistics for the regulation indicators in cohort 2, but could not examine correlations with recidivism because it was not measured in that sample. In general, means for each of the 11 indicators were higher in cohort 2 than in cohort 1, both at time 1 (p values for MANOVA and all t-tests comparing indicators across cohorts <.001; Cohen’s d ranged from .39 for dealing with emotions to 1.86 for dealing with others) and at time 2 (p values for MANOVA and all t-tests comparing indicators across cohorts <.001; Cohen’s d ranged from .25 for dealing with emotions to 1.49 for staying in the moment). Just as in cohort 1, there was significant mean improvement across all indicators from time 1 to time 2, as indicated by paired sample t-tests (all t-values > 5, all p-values < .001) as well as chi-square tests (all chi-square values >38, all p-values < .001). Cohort 2 did not contribute to further analyses because there were no recidivism data for this sample.

Multiple Indicator Latent Change Score Model

To estimate a multiple indicator latent change score model of cognitive and emotional regulation skills, we estimated two latent factors at each time point for cohort 1, consistent with the two-factor model described previously: a five-item cognitive factor and a six-item emotional factor. Residual covariances were estimated between the monitoring internal triggers and monitoring external triggers items at each time point, and residual covariances were also estimated for each item with itself at the other time point. Item thresholds and factor loadings were constrained to be equivalent across time points. We estimated a latent change score for each factor by: defining a latent change score by estimating a factor loading of 1 on the factor at time two; constraining the regression of the time two factor on its time one counterpart to 1; and constraining the time two factor variances to 0. We also estimated covariances between time one factor scores and between latent change scores, and regressed the change scores on both of the time one factors. Similar to confirmatory factor analyses, this model used the WLSMV estimator with delta parameterization and probit link. This model, without any controls or outcomes, appeared to fit the data reasonably well, except for a significant model chi-square test: χ² (231) = 1298.71, RMEA = .09, CFI = .96, TLI = .96, SRMR = .06. Significant results from this model indicated that cognitive and emotional factor scores were positively correlated at time one, and cognitive and emotional latent change scores were also positively correlated. Skills in cognitive regulation at time one positively predicted improvement in emotion regulation skills, and negatively predicted improvement in cognitive regulation skills (i.e., lower cognitive regulation skills at time one was associated with greater improvement from time one to time two). Emotion regulation skills at time one negatively predicted improvement in emotion regulation skills and did not significantly predict change in cognitive regulation skills.

Next, we included the felony charge indicator as an outcome in the model. We encountered convergence issues when regressing felony recidivism on cognitive and emotional regulation skills at time one and cognitive and emotional latent change scores; thus, we instead estimated a model that included correlations among felony recidivism, each regulation factor at time one, and each of the latent change scores. This model also seemed to fit the data reasonably well, except for a significant model chi-square test: χ² (249) = 1333.13, p < .001, RMSEA = .09, CFI = .96, TLI = .95, SRMR = .06. Results indicated that having a felony charge in the 12 months post-release was negatively correlated with improvement in emotion regulation skills (b = −.20, p < .001), but not significantly correlated with cognitive regulation skills at time 1 (b = −.03, p = .474), emotional regulation skills at time 1 (b = −.01, p = .840), or with improvement in cognitive regulation skills (b = −.06, p = .101).

Next, we included covariates in the model and regressed the felony charge outcome, as well as the regulation factor scores and change scores, on all of the covariates. The felony charge outcome was also regressed on initial regulation factor scores and change scores. This model continued to fit the data relatively well, except for a significant model chi-square test: χ² (765) = 1289.71, p < .001, RMSEA = .03, CFI = .97, TLI = .97, SRMR = .07. Table 4 displays the results from this model. Age was positively associated with initial cognitive and emotion regulation, but not with change scores. Males had lower initial cognitive regulation and less improvement in emotion regulation compared to females on average. Felony history was associated with lower initial emotion regulation. A lack of positive adult relationships was negatively associated with both cognitive and emotion regulation at time 1, as well as change in both cognitive and emotion regulation. Youth with a history of gang involvement were more likely to decrease in their cognitive and emotion regulation, while youth who were living with their biological father were more likely to increase in their cognitive and emotion regulation. Youth who reported a history of drug use had higher cognitive and emotion regulation at time 1 and greater improvement in cognitive regulation. Youth who did not report a history of victimization, witnessing violence, or mental health issues had higher initial levels of cognitive and emotion regulation. Cognitive regulation skills at time 1 were negatively related to change in cognitive regulation and positively related to change in emotion regulation skills. Emotion regulation skills at time 1 were negatively related to change in emotion regulation skills.

Table 4.

Initial Regulation and Change Scores Predicting 12-month Recidivism.

Predictor	Outcome
Predictor	Cognitive regulation at time 1	Emotion regulation at time 1	Change in cognitive regulation	Change in emotion regulation	Recidivism
Age	.07*	.18***	.04	.04	−.11
Gender (1 = male)	−.42***	−.17	−.12	−.29*	.13
Race/ethnicity (ref = White)
Black/African American	−.08	−.17	−.05	−.09	−.05
Hispanic	.11	−.07	−.02	.04	−.20
Multiracial	−.02	.11	−.02	−.14	−.25
Other	.29	.04	.01	−.03	−.17
Age at 1st offense (ref = <13)
13–14	−.07	−.28**	−.00	.04	.20
15	−.06	−.24	.02	.25	.54
16–17	.03	−.17	.06	−.00	−.02
Felony history (ref = 0)
1	.01	−.40*	−.32*	−.12	.47
2	−.03	−.50*	−.27	−.18	.21
3 or more	.01	−.42*	−.16	−.05	.67
Lack of positive adult	−.20**	−.25**	−.15*	−.16*	.25
History of gang involvement	−.13	−.18	−.16*	−.22*	.38*
Mother’s incarceration history	.02	−.02	.09	.06	−.12
Father’s incarceration history	−.00	−.00	.07	.07	.06
Living with biological mother	−.08	−.17	−.01	.02	−.14
Living with biological father	.04	.01	.17*	.30**	.03
Household income (ref = >$35k)
$15,000–$34,999	−.10	−.33*	−.05	−.03	.09
$0–$14,999	−.22*	−.48**	−.13	−.07	.37
Alcohol use	.12	.18	−.08	−.04	.17
Drug use	.22**	.28*	.25**	.11	−.55*
No victimization	.16*	.31**	.00	−.02	−.13
No witnessing violence	.30**	.34**	−.07	−.10	−.41
Any ADHD history	−.04	−.08	.01	.02	−.15
No history of mental illness	.20**	.23*	.11	.17	−.19
Cognitive regulation at time 1			−.31***	.44***	1.77**
Emotion regulation at time 1			−.09	−.96***	−1.60**
Cognitive regulation change					2.53**
Emotion regulation change					−2.03**

*p < .05, **p < .01, ***p < .001. N = 599. Coefficients are unstandardized. Model fit indices: χ² (765) = 1,289.71, p < .001, RMSEA = .03, CFI = .97, TLI = .97, SRMR = .07.

Recidivism, as measured by a felony charge in the 12 months post-release, was significantly predicted by initial regulation skill scores as well as latent change scores. Emotion regulation skills at time 1 (b = −1.60, p = .04) and improvement in emotion regulation skills from time 1 to time 2 (b = −2.03, p = .003) were both negatively and significantly associated with receiving a felony charge, such that youth with greater skills related to emotion regulation at the first assessment or with greater improvement from the first to the second assessment were less likely to receive a felony charge. However, cognitive regulation skills at time 1 (b = 1.77, p = .003) and improvement in cognitive regulation skills (b = 2.53, p = .009) were both positively associated with recidivism, such that youth with higher skills related to cognitive regulation at the first assessment or greater improvement from the first to the second assessment were more likely to receive a felony charge. In addition, youth with a history of gang involvement (b = .38, p = .047) and without a history of drug use (b = −.55, p = .042) were more likely to receive a felony charge.

Finally, to account for length of stay, we estimated the same model as above but regressed felony recidivism, change in cognitive regulation, and change in emotion regulation on length of stay in detention in months, and regressed length of stay on initial levels of cognitive and emotion regulation skills, as well as all covariates in the model. Figure 1 displays the results among regulation, length of stay, and recidivism from this model. This model continued to fit the data relatively well, except for a significant model chi-square test: χ² (783) = 1323.36, p < .001, RMSEA = .03, CFI = .97, TLI = .97, SRMR = .07. Results indicated that youth with higher levels of cognitive regulation skills at time 1 (b = .97, p = .042) and lower levels of emotion regulation skills at time 1 (b = −.83, p = .022) had a longer length of stay on average. However, length of stay did not significantly predict change in cognitive regulation skills (b = −.01, p = .084), change in emotion regulation skills (b = −.00, p = .811), or felony recidivism (b = .00, p = .982). The regression coefficients from initial levels of cognitive and emotional regulation skills, as well as change in cognitive and emotional regulation skills, to felony recidivism remained relatively unchanged from the previous model.

Figure 1.

Associations among cognitive and emotional regulation, length of stay, and felony recidivism. Note. *p < .05, **p < .01, ***p < .001. Unstandardized coefficients are displayed.

Sensitivity Analyses

We wanted to explore whether results would differ if we used a count of felony charges as an outcome rather than a binary indicator for any felony charge. However, due to the complexity of the multiple indicator latent change score model, we were unable to use a count outcome within this model. Instead, we conducted two different regression models in Stata using listwise deletion: a negative binomial model predicting the count of felony charges and a logit model predicting the binary indicator of any felony charge. We selected a negative binomial model rather than Poisson because the count of felony charges appeared overdispersed (M = .64, SD = 1.33). We included averaged scores for cognitive and emotional regulation at time 1, as well as differences between the average time 1 scores and average time 2 scores for the change in cognitive and emotion regulation. We retained all of the covariates from the original model in Table 4. The results from these models are displayed in Table 5. Compared to the model in Table 4, in the logit regression emotional regulation at time 1 was no longer significant. In the negative binomial model, only changes in cognitive and emotional regulation significantly predicted count of post-release felony charges, such that improvements in cognitive regulation and decreases in emotional regulation were associated with greater counts of felony charges.

Table 5.

Comparing Results from Binary and Count Models of Recidivism.

Predictor	Negative binomial regression (count)			Logit regression (binary)
Predictor	B	S.E.	Exp (b)	b	S.E.	Exp (b)
Cognitive regulation at time 1	.47	.26	1.60	.89**	.31	2.43
Emotion regulation at time 1	−.23	.28	.79	−.57	.33	.57
Cognitive regulation change	.74*	.29	2.10	.88*	.34	2.41
Emotion regulation change	−.94***	.26	.39	−1.15***	.28	.32

Note. *p < .05, **p < .01, *** p < .001. b = raw coefficient; S.E. = standard error; exp(b) = exponentiated coefficient (incidence rate ratio for negative binomial, odds ratio for logit). N for both models is 555. All of the control variables from the model in Table 4 were included in both models.

To further explore the different direction of associations of cognitive (positive) and emotional (negative) regulation with recidivism, we conducted some follow-up analyses. First, we estimated the final model from above in Figure 1 (recidivism regressed on initial level and change in regulation, as well as length of stay and all other covariates, etc.), but separately for cognitive and emotional regulation. In one model, only the cognitive regulation items and latent variables were included, and in the other, only the emotional regulation items and latent variables were included. Both models appeared to fit the data relatively well, with CFIs and TLIs above .97, RMSEAs below 0.03, and SRMRs below .07, although both had significant model chi-square tests (ps < .001). In the cognitive regulation skills model, neither initial level of cognitive regulation skills (b = .13, p = .183) nor change in cognitive regulation skills (b = −.16, p = .132) significantly predicted felony recidivism. In the emotion regulation skills model, initial level of emotion regulation skills did not significantly predict recidivism (b = −.03, p = .718), but increased emotion regulation skills were associated with decreased odds of felony recidivism (b = −0.29, p < .001).

As a final exploration of which indicators of regulation might be most strongly associated with felony recidivism, we estimated models in which the latent variables and latent change scores were replaced with observed initial values of each of the 11 skill indicators, as well as the observed change in each indicator (i.e., variable at time 2 - variable at time 1). Again, each of these models appeared to fit the data well, with CFIs above .95, TLIs above .84, an RMSEA of .013, and an SRMR of 0.060, as well as a non-significant model chi-square test (p =.284). In these models, none of the initial indicators was significant (all ps > .058). The only indicators in which change from time 1 to time 2 was a significant predictor of felony recidivism were impulse control (b = −0.20, p = .040), aggression control (b = −016, p = .022), dealing with distress (b = −.20, p = .008), monitoring internal triggers (b = −0.28, p < .001), and monitoring external triggers (b = −.30, p < .001). All relationships were in the negative direction, such that improvement on an indicator was associated with lower odds of felony recidivism. Note that if we were to adopt a significance level of .002 to account for the familywise error rate across 22 significance tests (.05/22 = .002) to mitigate against inflated Type I error, only the coefficients for monitoring internal triggers and monitoring external triggers would meet criteria for significance.

Discussion

This study examined data from juveniles in Washington state detention facilities to create and test a two-factor model of skills related to self-regulation and their association with later recidivism. The cognitive regulation factor consisted of consequential thinking, goal setting, situational perception, dealing with others, and staying in the moment, while the emotional factor consisted of dealing with emotions, control of impulses, control of aggression, dealing with distress, monitoring internal triggers, and monitoring external triggers. Higher levels of cognitive regulation skills at intake were associated with higher levels of emotion regulation skills. Changes in cognitive and emotional regulation skills from intake to pre-release were also positively associated with each other. Emotional regulation skills at intake and improvement in emotional regulation skills during incarceration were associated with lower risk for felony recidivism in the year following a juvenile’s release from a secure facility. Cognitive regulation skills at intake and improvement in cognitive regulation during incarceration were associated with higher risk for recidivism when emotional regulation skills were also included in the model, but when models were conducted separately, the only significant predictor of recidivism was improvement in emotional regulation skills. This change in results may be due to the direction and strength of relationships among the three variables, potentially resulting in a reversal paradox (Tu et al., 2008). Additionally, consistent with prior research predicting criminogenic cognitions (e.g., Tangney et al., 2017), emotion regulation might mediate the association between cognitive regulation skills and crime-related cognitions and behavior, such that the direct effect becomes negative after including the mediating pathway. Prior research has suggested that this may be due to a “dark side” to mindfulness (Tangney et al., 2017), although future research is needed to better explain these associations.

Our finding that likelihood of recidivism is lower after improvement in emotion regulation skills is consistent with literature showing that items from our emotional factor, like impulse control and aggression control, are associated with crime and recidivism (Gottfredson & Hirschi, 1990; Pratt & Cullen, 2000). This is further supported by impulse control and aggression control both individually having significant negative relationships with recidivism (although they did not remain significant predictors when adopting a stricter significance threshold to account for familywise error rate). In addition, our findings are in line with research supporting a link between emotion regulation and aggression (Garofalo, et al., 2018; Garofalo et al., 2019), as well as the effectiveness of interventions aimed at promoting emotional and cognitive regulation for reducing adult recidivism (Brazão et al., 2018).

Strengths and Limitations

One strength of the current study is the relatively large and racially diverse sample. An additional strength is the usage of 10 different raters for the R-PACT data, reducing the chance that ratings could be heavily skewed by the biases of a single rater. We also examined a wide variety of skills, decreasing the likelihood that any one skill was only related to recidivism due to its correlation with another skill. Finally, the use of latent variables within a structural equation modeling framework allowed us to reduce measurement error, and the use of multiple sources of data (R-PACT, official offense records) minimized shared method variance.

One limitation of this study is that the R-PACT—the source of data for all measures that make up the emotional and cognitive factors—has not yet been tested against other emotional regulation skills measures with well-supported construct validity (Hay, 2013). This could be a possible reason why when correcting for familywise error, only monitoring internal triggers and monitoring external triggers were significant negative predictors of recidivism. The current study only examines felony recidivism within 1 year of release. Thus, this study does not suggest that improvement in emotional regulation permanently decreases recidivism, as it is possible that improved emotional regulation skills only delay recidivism beyond the data collection window. It is not without precedent for something to delay but not reduce recidivism, as there is research supporting the idea that employment does just that (Tripodi et al., 2010). In addition, because we focused on felony recidivism, it is unclear whether cognitive and emotional regulation skills would have the same associations with lower-level (e.g., misdemeanor) recidivism. Finally, this was a sample of youth in detention in Washington state, and results might have limited generalizability to other samples (e.g., youth under community supervision in Washington and youth in detention facilities in other states).

Future Research Directions

Since this study has provided support that improved emotional regulation skills reduce short-term recidivism, one clear direction for future research would be to examine the effectiveness of possible emotion regulation training programs. Such research is still in its relatively early stages; for example, a recent paper provided a taxonomy of emotion regulation training programs and recommendations for how to study their effectiveness (Denny, 2020). While research has indicated that there are several possible options for emotion regulation training, it is not yet clear which are most likely to be effective. Such research is not just needed in general, but also in the specific setting of juvenile detention centers, because research with adults may not always be generalizable to youth in detention facilities. More research also may be needed to examine individual differences that could result in differing effectiveness of treatments for reducing recidivism. In addition, future studies could similarly examine subscales of cognitive and emotion regulation using the R-PACT, as we have done in the current study, to explore whether the factor structure and associations with recidivism in this study generalize to other samples.

Implications for Practice

The results of this study suggest that interventions aimed at improving youths’ emotional regulation skills may in turn reduce recidivism. Previous research has found that therapies, such as dialectical behavior therapy (DBT), that improve emotion regulation also improve outcomes for individuals in residential treatment (McCann et al., 2007; Tomlinson, 2018). Indeed, studies have indicated that DBT may be a promising treatment modality to reduce recidivism in juveniles in the justice system, although sample sizes have generally been small (N < 50), and researchers have not examined emotion regulation as a mediating factor (Quinn & Shera, 2009; Shelton et al., 2011). One study did find increases in use of skills related to mindfulness, distress tolerance, and emotion regulation among a sample of youth in detention after exposure to a DBT-based skills training group, but did not assess recidivism (Walden et al., 2019). Therefore, DBT and similar treatment programs could be an avenue for improving emotion regulation and reducing recidivism among youth in residential facilities, although more research is needed.

Conclusion

In this study, we used exploratory and confirmatory factor analyses to create measures of skills related to cognitive and emotional regulation and examined their association with juvenile recidivism. Improved emotional regulation skills from intake at a secure facility to release was associated with decreased recidivism, while improved cognitive regulation was either associated with increased recidivism (when both regulation measures were included in the model) or had no relationship with recidivism (when regulation measures were modeled separately). The only specific skills that significantly predicted recidivism when correcting for familywise error were monitoring internal triggers and monitoring external triggers. These findings suggest that additional research and resources aimed at improving emotion regulation skills—with a specific focus on trigger monitoring—may subsequently decrease juvenile recidivism.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Meagan Docherty

Author biographies

Meagan Docherty is an Assistant Professor at Bowling Green State University and an Adjunct Faculty Associate at University of Michigan. Her research has examined the development of externalizing behaviors from childhood to young adulthood, and has focused specifically on the development of callousness, gun carrying and use among adolescents and young adults, and risk and protective factors such as parenting quality and exposure to violence.

Andrew Lieman is a doctoral student at Bowling Green State University. His research interests include Industrial Organizational psychology with a focus on managerial resistance to empirical findings.

Brandon Lee Gordon, MA is a doctorate student of developmental psychology at Bowling Green State University. His research interests are varied including impulse control disorders in adolescence, human sexuality, and psychology of religion and spirituality. Currently, his research focuses on compulsive sexual behavior disorder in adolescence.

References

Aizer

Doyle

J. J.

Jr. (2013). Juvenile incarceration, human capital and future crime: Evidence from randomly-assigned judges. National Bureau of Economic Research, working paper 19102. http://nber.org/papers/w19102

Baglivio

M. T.

Wolff

K. T.

(2019). Predicting juvenile reentry success: Developing a global risk score and risk classification levels using the residential positive achievement change Tool. Youth Violence and Juvenile Justice, 17(3), 241–268. https://doi.org/10.1177/1541204018804870

Baglivio

M. T.

Wolff

K. T.

DeLisi

Vaughn

M. G.

Piquero

A. R.

(2016). Effortful control, negative emotionality, and juvenile recidivism: An empirical test of DeLisi and Vaughn’s temperament-based theory of antisocial behavior. The Journal of Forensic Psychiatry & Psychology, 27(3), 376–403. https://doi.org/10.1080/14789949.2016.1145720

Barnert

E. S.

Perry

Morris

R. E.

(2016). Juvenile incarceration and health. Academic Pediatrics, 16(2), 99–109. https://doi.org/10.1016/j.acap.2015.09.004

Brazão

Rijo

Salvador

M. D. C.

Pinto-Gouveia

(2018). Promoting emotion and behavior regulation in male prison inmates: A secondary data analysis from a randomized controlled trial testing the efficacy of the Growing Pro-Social Program. Law and Human Behavior, 42(1), 57–70. https://doi.org/10.1037/lhb0000267

Burt

C. H.

Simons

R. L.

Simons

L. G.

(2006). A longitudinal test of the effects of parenting and the stability of self-control: Negative evidence for the general theory of crime. Criminology, 44(2), 353–396. https://doi.org/10.1111/j.1745-9125.2006.00052.x

Carver

C. S.

Scheier

M. F.

, (2010). Self-regulation of affect and action. In Vohs

K. D.

Baumeister

R. F.

(Eds.), Handbook of self-regulation: Research, theory, and applications (pp. 3–23). New York: The Gilford Press.

Cauffman

Steinberg

Piquero

A. R.

(2005). Psychological, neuropsychological and physiological correlates of serious antisocial behavior in adolescence: The role of self-control. Criminology, 43(1), 133–176. https://doi.org/10.1111/j.0011-1348.2005.00005.x

Cottle

C. C.

Lee

R. J.

Heilbrun

(2001). The prediction of criminal recidivism in juveniles. Criminal Justice and Behavior, 28(3), 367–394. https://doi.org/10.1177/0093854801028003005

10.

Cuevas

Wolff

K. T.

Baglivio

M. T.

(2019). Dynamic risk factors and timing of recidivism for youth in residential placement. Journal of Criminal Justice, 60, 154–166. https://doi.org/10.1016/j.crimjus.2018.10.003

11.

DeLisi

Vaughn

M. G.

(2014). Foundation for a temperament-based theory of antisocial behavior and criminal justice system involvement. Journal of Criminal Justice, 42(1), 10–25. https://doi.org/10.1016/j.jcrimjus.2013.11.001

12.

Denny

B. T.

(2020). Getting better over time: A framework for examining the impact of emotion regulation training. Emotion, 20(1), 110–114. https://doi.org/10.1037/emo0000641

13.

Dmitrieva

Monahan

K. C.

Cauffman

Steinberg

(2012). Arrested development: The effects of incarceration on the development of psychosocial maturity. Development and Pyschopathology, 24, 1073–1090. https://doi.org/10.1017/S0954579412000545

14.

Eisenberg

Smith

C. L.

Sadovsky

Spinrad

T. L.

(2004). Effortful control: Relations with emotion regulation, adjustment, and socialization in childhood. In Baumeister

R. F.

Vohs

K. D.

(Eds.), Handbook of self-regulation: Research, theory, and applications (pp. 259–282). New York: Guilford Press.

15.

Ford

J. D.

Blaustein

M. E.

(2013). Systemic self-regulation: A framework for trauma-informed services in residential juvenile justice programs. Journal of Family Violence, 28, 665–677. https://doi.org/10.1007/s10896-013-9538-5

16.

Garofalo

Gillespie

S. M.

Velotti

(2019). Emotion regulation mediates relationships between mindfulness facets and aggression dimensions. Aggressive Behavior, 46, 60–71. https://doi.org/10.1002/ab.21868

17.

Garofalo

Holden

C. J.

Zeigler-Hill

, & Velotti

(2016). Understanding the connection between self-esteem and aggression: The mediating role of emotion dysregulation. Aggressive Behavior, 42(1), 3–15. https://doi.org/10.1002/ab.21601

18.

Garofalo

Velotti

Zavattini

G. C.

(2018). Emotion regulation and aggression: The incremental contribution of alexithymia, impulsivity, and emotion dysregulation facets. Psychology of Violence, 8(4), 470–483. https://doi.org/10.1037/vio0000141

19.

Gottfredson

M. R.

Hirschi

(1990). A general theory of crime. Stanford, CA: Stanford University Press.

20.

Hay

(2013). Residential positive achievement change Tool (R-PACT) validation study. http://www.djj.state.fl.us/docs/partners-providers-staff/r-pact_validation_study_executivesummary_june2013.pdf

21.

Hay

Forrest

(2006). The development of self-control: Examining self-control theory’s stability thesis. Criminology, 44(4), 739–774. https://doi.org/10.1111/j.1745-9125.2006.00062.x

22.

Hay

Widdowson

A. O.

Bates

Baglivio

M. T.

Jackowski

Greenwald

M. A.

(2016). Predicting recidivism among released juvenile offenders in Florida: An evaluation of the Residential Positive Achievement Change Tool. Youth Violence and Juvenile Justice, 16(1), 97–116. https://doi.org/10.1177/1541204016660161

23.

Hay

Widdowson

Young

B. C.

(2018). Self-control stability and change for incarcerated juvenile offenders. Journal of Criminal Justice, 56, 50–59. https://doi.org/10.1016/j.jcrimjus.2017.08.008

24.

Higgins

G. E.

Jennings

W. G.

Tewksbury

Gibson

C. L.

(2009). Exploring the link between low self-control and violent victimization trajectories in adolescents. Criminal Justice and Behavior, 36, 1070–1084. https://doi.org/10.1177/0093854809344046

25.

Hirschi

Gottfredson

M. R.

(1994). The generality of deviance. Routledge.

26.

Kievit

R. A.

Brandmaier

A. M.

Ziegler

Van Harmelen

A. L.

de Mooij

S. M. M.

Moutoussis

Goodeyer

I. M.

Bullmore

Jones

P. B.

Fonagy

Lundenberger

Dolan

R. J.

(2018). Developmental cognitive neuroscience using latent change score models: A tutorial and applications. Developmental Cognitive Neuroscience, 33, 99–117. https://doi.org/10.1016/j.dcn.2017.11.007

27.

Laird

(2021). Regional international juvenile incarceration models as a blueprint for rehabilitative reform of juvenile criminal justice systems in the United States. Journal of Criminal Law and Criminology, 111(2), 571–603. https://www.jstor.org/stable-48614947

28.

Lambie

, & Randell

(2013). The impact of incarceration on juvenile offenders. Clinical Psychology Review, 33(3), 448–459. https://doi.org/10.1016/j.cpr.2013.01.007

29.

Long

Felton

J. W.

Lilienfeld

S. O.

, & Lejuez

C. W.

(2014). The role of emotion regulation in the relations between psychopathy factors and impulsive and premeditated aggression. Personality Disorders: Theory, Research, and Treatment, 5(4), 390–396. https://content.apa.org/doi/10.1037/per0000085

30.

Mamayek

Paternoster

Loughran

T. A.

(2017). Self-control as self-regulation: A return to control theory. Deviant Behavior, 38(8), 895–916. https://doi.org/10.1080/01639625.2016.1206730

31.

McArdle

J. J.

, & Grimm

K. J.

(2010). Five steps in latent curve and latent change score modeling with longitudinal data. In van Montfort

Oud

, & Satorra

(Eds.), Longitudinal research with latent variables (pp. 245–273). Berlin, Heidelberg: Springer. https://doi.org/10.1007/978-3-642-11760-2_8

32.

McCann

R. A.

Ivanoff

Schmidt

Beach

(2007). Implementing dialectical behavior therapy in residential forensic settings with adults and juveniles. In Dimeff

L. A.

Koerner

(Eds.), Dialectical Behavior Therapy in clinical practice: Applications across disorders and settings. New York: Guilford Press.

33.

Moilanen

K. L.

Raffaelli

(2005). Parenting and self-regulation in adolescence: Associations with adolescent behavior. ProQuest Dissertations Publishing.

34.

Monahan

K. C.

Steinberg

Cauffman

Mulvey

E. P.

(2009). Trajectories of antisocial behavior and psychosocial maturity from adolescence to young adulthood. Developmental Psychology, 45(6), 1654–1668. https://doi.org/10.1037/a0015862

35.

Mulder

Brand

Bullens

van Marle

(2011). Risk factors for overall recidivism and severity of recidivism in serious juvenile offenders. International Journal of Offender Therapy and Comparative Criminology, 55(1), 118–135. https://doi.org/10.1177/0306624X09356683

36.

Muthén

L. K.

Muthén

B. O.

(1998-2017). Mplus user’s guide (8th ed.). Los Angeles, CA: Muthén & Muthén.

37.

National Juvenile Justice Network (NJJN) (2014). Policy platform: Reducing youth confinement. https://www.njjn.org/uploads/digital-library/2018%20NJJN%20Member%20Youth%20Justice%20Advances.pdf

38.

Pihet

Combremont

Suter

Stephan

(2011). Cognitive and emotional deficits associated with minor and serious delinquency in high-risk adolescents. Psychiatry, Psychology and Law, 19(3), 427–438. https://doi.org/10.1080/13218719.2011.598634

39.

Piquero

A. R.

Jennings

W. G.

Farrington

D. P.

(2010). On the malleability of self-control: Theoretical and policy implications regarding a general theory of crime. Justice Quarterly, 27(6), 803–834. https://doi.org/10.1080/07418820903379628

40.

Pratt

T. C.

Cullen

F. T.

(2000). The empirical status of Gottfredson and Hirschi's general theory of crime: A meta-analysis. Criminology, 38(3), 931–964. https://doi.org/10.1111/j.1745-9125.2000.tb00911.x

41.

Quinn

Shera

(2009). Evidence-based practice in group work with incarcerated youth. International Journal of Law and Psychiatry, 32, 288–293. https://doi.org/10.1016/j.jilp.2009.06.002

42.

Ray

J. V.

Jones

Loughran

T. A.

Jennings

W. G.

(2013). Testing the stability of self-control: Identifying unique developmental patterns and associated risk factors. Criminal Justice and Behavior, 40(6), 588–607. https://doi.org/10.1177/0093854812464222

43.

Ray

J. V.

Thornton

L. C.

Frick

P. J.

Steinberg

Cauffman

(2016). Impulse control and callous-unemotional traits distinguish patterns of delinquency and substance use in justice involved adolescents: Examining the moderating role of neighborhood context. Journal of Abnormal Child Psychology, 44(3), 599–611. https://doi.org/10.1007/s10802-015-0057-0

44.

Romaine

C. L. R.

(2019). Psychosocial maturity and risk-taking in emerging adults: Extending our understanding beyond delinquency. Emerging Adulthood, 7(4), 243–257. https://doi.org/10.1177/2167696818768013

45.

Sawyer

(2019). Youth confinement: The whole pie 2019. Prison Policy Initiative. https://www.prisonpolicy.org/reports/youth2019.html

46.

Seigle

Walsh

Weber

(2014). Core principles for reducing recidivism and improving other outcomes for youth in the juvenile justice system. New York: Council of State Governments Justice Center.

47.

Shelton

Kesten

Zhang

Trestman

(2011). Impact of a Dialectic Behavior Therapy-Corrections Modified (DBT-CM) upon behaviorally challenged incarcerated male Adolescents. Journal of Child and Adolescent Psychiatric Nursing, 24, 105–113. https://doi.org/10.1111/j.1744-6171.2011-00275.x

48.

StataCorp (2019). Stata statistical software: Release 16. College Station, TX: StataCorp LLC.

49.

Steinberg

(2010). A dual systems model of adolescent risk-taking. Developmental Psychobiology, 52, 216–224. https://doi.org/10.1002/dev.20445

50.

Steinberg

Cauffman

(1996). Maturity of judgment in adolescence: Psychosocial factors in adolescent decision making. Law and Human Behavior, 20(3), 249–272. https://doi.org/10.1007/BF01499023

51.

Tangney

J. P.

Dobbins

A. E.

Stuewig

J. B.

Schrader

S. W.

(2017). Is there a dark side to mindfulness? Relation of mindfulness to criminogenic cognitions. Personality and Social Psychology Bulletin, 43(1), 1415–1426. https://doi.org/10.1177/0146167217717243

52.

Tomlinson

M. F.

(2018). A theoretical and empirical review of dialectical behavior therapy within forensic psychiatric and correctional settings worldwide. International Journal of Forensic Mental Health, 17(1), 7295. https://doi.org/10.1080/14999013.2017.1416003

53.

Tripodi

S. J.

Kim

J. S.

Bender

(2010). Is employment associated with reduced recidivism? International Journal of Offender Therapy and Comparative Criminology, 54, 706–720. https://doi.org/10.1177/0306624X09342980

54.

Y.-K.

Gunnell

Gilthorpe

M. S.

(2008). Simpson’s paradox, Lord’s paradox, and suppression effects are the same phenomenon – the reversal paradox. Emerging Themes in Epidemiology, 5(2), 2. https://doi.org/10.1186/1742-7622-5-2

55.

van Dam

Janssens

M. A. M.

De Bruyn

E. E. J.

(2005). PEN, Big Five, juvenile delinquency and criminal recidivism. Personality and Individual Differences, 39(1), 7–19. https://doi.org/10.1016/j.paid.2004.06.016

56.

Vazsonyi

A. T.

Mikuška

Kelley

E. L.

(2017). It’s time: A meta-analysis on the self-control-deviance link. Journal of Criminal Justice, 48, 48–63. https://doi.org/10.1016/j.jcrimjus.2016.10.001

57.

Walden

A. L.

Stancil

Verona

(2019). Reaching underserved youth: A pilot implementation of a skills-based intervention in short-term juvenile detention. Journal of Prevention & Intervention in the Community, 47, 90–103. https://doi.org/10.1080/10852352.2019.1582147

58.

Walker

(2016). (2008-2015) Therapeutic change, length of stay, and recidivism in incarcerated juvenile offenders in Washington State, 2008-2015. (ICPSR 36226, Version V1) [Data set]. ICPSR. https://doi.org/10.3886/ICPSR36226.v1

59.

Walker

S. C.

Bishop

A. S.

(2016). Length of stay, therapeutic change, and recidivism for incarcerated juvenile offenders. Journal of Offender Rehabilitation, 55(6), 355–376. https://doi.org/10.1080/10509674.2016.1194946

60.

Winokur-Early

Hand

G. A.

Blankenship

J. L.

(2012). Validity and reliability of the Florida Positive Achievement Change Tool (PACT) risk and needs assessment instrument: A three-phase evaluation (validation study, factor analysis, inter-rater reliability). Tallahassee, FL: Justice Research Center.