AN EMPIRICAL PORTRAIT OF COMMUNITY REENTRY AMONG SERIOUS JUVENILE OFFENDERS IN TWO METROPOLITAN CITIES

Abstract

This study examined the community reentry process among 413 serious adolescent offenders released from juvenile court commitments in two metropolitan areas. Data are provided about postrelease court supervision and community-based services (CBSs) during the first 6 months in the community as well as indicators of antisocial activity, formal system involvement, school attendance, and employment. Findings indicate that a far greater proportion of offenders reported receiving supervision than CBSs, but when utilized, the frequency of CBS use was high, and intensive services reduced the odds of formal system involvement. In addition, court supervision increased the likelihood of positive adjustment during community reentry. These results held after controlling for social context variables, including peer deviance, parental monitoring, and contact with caring adults.

According to the most recent figures available, 140,700 delinquency cases in 2004 resulted in out-of-home residential commitments, an increase of 34% since 1985 (U.S. Department of Justice, in press). This process of removing and eventually returning this many young people to the community creates formidable challenges for the juvenile justice system and its affiliated service providers. However, despite growing interest in the study of offender reentry and reintegration, the immense need for successful strategies continues to outpace research on the implementation and evaluation of effective reintegration programs (Altschuler & Brash, 2004).

There is no doubt that adolescents making the transition from residential placements back to the community need a variety of supportive services. These adolescents are usually high-risk offenders who struggle with multiple problems known to predict poor outcomes. Using data from the Census of Juveniles in Residential Placement (CJRP), Snyder (2004) described candidates for reentry programs as predominantly male (86%), minority (39% Black, non-Hispanic; 17% Hispanic), and older youth adjudicated for violent acts (38%). He depicted a high-risk population in need of rehabilitative programs including educational, mental health, and drug-treatment services. In previous studies, researchers have documented that nearly 58% of reentry offenders could be certified with a special education disability (30% for an emotional disturbance and more than 22% for a learning disability; Bullis, Yovanoff, & Havel, 2004). Others have reported that two thirds of committed males meet diagnostic criteria for at least one mental health disorder, and roughly half can be diagnosed with a substance use disorder (see Grisso, 2004).

THE PROBLEM OF REENTRY FOR JUVENILE OFFENDERS

Unfortunately, community reentry is often associated with failure (Krisberg, 1997). Formerly incarcerated youth show a number of poor short- and long-term outcomes after returning to the community: low rates of youth going back to school and getting a high school diploma (Gemignani, 1994); high rates of unstable employment and welfare dependence in adulthood (Grogger, 1995); and disturbing rates of rearrest within a few years of discharge from institutional settings. Although estimates suggest that about 50% of young offenders will return to the juvenile justice system (Lipsey, Wilson, & Cothern, 2000), two recent studies found that 91% of youth exiting Delaware’s Serious Juvenile Offender Program and the California Youth Authority were rearrested within 2 and 3 years, respectively (Byrnes, Macallair, & Shorter, 2002; Rodriguez-Labarca & O’Connell, 2004). In fact, 60% of the Delaware offenders were rearrested within 6 months of their discharge date.

The first few months following release appear to be critical for these young offenders. This is when these adolescents are first without the structure, supervision, and support of institutional settings. In addition, they are often returning to families and communities that may have trouble accommodating them even under the best of circumstances. It is during this period right after release when engaging young offenders and reintegrating them in law-abiding activities are particularly important for reducing the chances that they will reoffend (Altschuler & Armstrong, 1994).

THE ROLE OF SUPERVISION AND SERVICES IN COMMUNITY REINTEGRATION

The term aftercare typically describes a period of time following release from a residential facility when young offenders remain under juvenile justice authority. This usually represents a 6-month period during which the provision of postrelease aftercare supervision and community services are thought to be crucial for promoting successful community adjustment (Juvenile Court Judges’ Commission, 2001). Specifically, it has been argued that these aftercare services will be effective to the extent that they can balance mandates of the juvenile justice system with the developmental needs of young offenders (Altschuler & Brash, 2004; Chung, Little, & Steinberg, 2005). This type of developmental perspective considers both risk and protective factors that contribute to adjustment and highlights the need to examine outcomes other than rearrest or antisocial behavior, such as schooling and employment. It also emphasizes factors outside of the individual (e.g., family and living arrangements, peer affiliations) that may help to explain why the facility-to-community transition is associated with poor outcomes. Similar to Altschuler and colleagues (Altschuler & Armstrong, 2001; Altschuler & Brash, 2004), we use the term community reintegration to stress a holistic perspective of how youth function within their families, communities, the justice system, and the larger society.

Unfortunately, the available research on the provision of aftercare services and their impact on community reintegration outcomes is rather limited, especially in light of the significant challenges posed. This is because most studies of reentry have focused on documenting criminal outcomes and not on understanding young offenders’ achievement of positive adjustment (e.g., Byrnes et al., 2002). A notable exception is a study of youth exiting the Oregon Youth Authority conducted by Bullis and his colleagues (2004). Importantly, the authors found that youth who used mental health and other social services (e.g., vocational rehabilitation) within 6 months of release were about 2 times more likely to be engaged—defined as employed and/or enrolled in school and not arrested or placed back in the justice system—than participants who used no services at all. Moreover, the authors confirmed the importance of “getting started right,” as youth who were engaged at 6 months were almost 2.5 times more likely to be engaged at 1 year. However, Bullis et al. (2004) did not assess postrelease supervision, and the study focused predominantly on Caucasian males. Thus, it is unclear what role supervision played on engagement and whether findings generalize to minority youth who are more representative of the national reentry population.

To date, the most comprehensive data come from experimental evaluations of the intensive aftercare program (IAP) model (Altschuler & Armstrong, 1994). This model is based on findings that postrelease supervision alone does not reduce juvenile recidivism (see Altschuler, Armstrong, & MacKenzie, 1999) and recommends the provision of frequent and coordinated supervision and community-based services (CBSs) before and after juvenile offenders are released from residential settings. It also advocates for studying areas of youth functioning that are theoretically and empirically linked to recidivism (e.g., peer deviance). The IAP model has received a great deal of theoretical attention, but recent evaluations have reported inconclusive results because of study limitations (e.g., small sample sizes) and have left a number of unanswered questions about the impact of aftercare activities on community reintegration (Wiebush, Wagner, McNulty, Wang, & Le, 2005). It is unclear, for example, whether postrelease supervision and services are related in the same way to different outcomes, such as recidivism or getting engaged in school or work, or whether the influence of aftercare activities on individual adjustment can be explained by intermediate outcomes such as peer deviance or support from adults.

PRESENT STUDY

To date, studies of offender reentry have provided limited data on reintegration outcomes and the impact of aftercare supervision and services on individual adjustment. Naturalistic studies have generally focused on criminal outcomes (e.g., Byrnes et al., 2002; Rodriguez-Labarca & O’Connell, 2004), and more comprehensive experimental studies have had trouble drawing conclusions about the effects of aftercare activities (e.g., Wiebush et al., 2005). Furthermore, most experimental studies have focused on high-intensity programs and cannot speak to the importance of aftercare activities that provide other levels of duration and intensity.

The current study examines the facility-to-community transition among 413 serious adolescent offenders released from juvenile court commitments in Maricopa County, Arizona, and Philadelphia County, Pennsylvania. Looking at community reentry in these two locations may be particularly informative from a policy and planning perspective because the two counties manage the reentry process and the provision of aftercare services in appreciably different ways (Griffin, 2005). First, the two states differ regarding the age to which juveniles remain under juvenile court control (ends at age 18 in Arizona but can extend to age 21 in Pennsylvania). Second, different entities are responsible for supervising juveniles following their release from state commitments. Arizona, like most states, uses a corrections control model in which the agency that oversees the state’s juvenile corrections is entirely responsible for supervising reentry. Typically, a juvenile is subject to parole-style supervision but has no further contact with the local court that ordered the original commitment. In contrast, Pennsylvania uses a local court control model in which the supervision of reentry offenders is always a local matter, overseen by probation officers who work as agents of the court. Compared to corrections control, this model is thought to yield greater access to resources (e.g., service providers) and a more flexible response to the needs of individual youth (e.g., range of sanction options). Philadelphia County, in particular, is involved in a state-initiated multi-year effort to improve aftercare services (Griffin, 2004).

The present study tries to provide a broad view of functioning during young offenders’ first 6 months in the community on multiple indicators of adjustment—antisocial activity, formal system involvement, school attendance, and employment. We also assessed their participation in postrelease court supervision and CBSs and present data about the duration and intensity of these activities. Finally, analyses investigated which dimensions of supervision and service activities were related to adjustment after controlling for (a) demographic and risk characteristics and (b) levels of postrelease peer deviance, parental monitoring, and support from caring adults, three factors that might explain the effects of aftercare programs on outcomes. This is the first study to examine such comprehensive questions among reentry offenders soon after they return to the community, when “getting started right” might have a strong impact on promoting and maintaining successful long-term adjustment.

METHOD

PARTICIPANTS

Participants (N = 413) were drawn from the sample enrolled in the Pathways to Desistance study, an ongoing longitudinal investigation of adolescents who were recently adjudicated of a serious crime (almost entirely felony offenses) in the juvenile or adult court systems in Philadelphia (Philadelphia County) and in Phoenix (Maricopa County), Arizona. In total, the Pathways study enrolled 1,355 adolescents aged 14 to 18, representing approximately 1 in 3 youth adjudicated on the enumerated charges in the two locales during the recruitment period (December 2000 through February 2003). Information regarding the theoretical foundation for the Pathways study is presented in Mulvey et al. (2004), and details regarding recruitment and full sample characteristics are discussed in Schubert et al. (2004). Overall retention rates have been high during the course of the study, with youth completing approximately 92% of expected interviews through 24 months; 2% of participants dropped out of the study, and 2% died during this follow-up period.

Similar to Snyder (2004), the reentry population was identified as youth released from custody facilities following a juvenile court commitment regardless of the type of facility (youth processed in adult court were excluded from these analyses). During the enrollment phase of the Pathways study, 547 offenders were processed in the juvenile system and sent to institutional placements as a result of their index arrest (see Cauffman, Piquero, Kimonis, Steinberg, & Chassin, in press, for how these youth differ from those placed on probation). The current analyses were limited to 413 of these youth who had at least 6 months of reentry data available following release from their initial court placements; the release date could not be determined for 74 cases, and less than 6 months of postrelease data was available for 60 cases.

The final sample was predominantly male (92%, n = 378) and 16.50 (SD = 1.10) years of age at the baseline interview. Also, 58% (n = 240) were African American, 23% (n = 94) were Hispanic (mostly Mexican American), 15% (n = 63) were non-Hispanic Caucasian, and 4% (n = 15) were other ethnicities. Most were from Philadelphia County (77%, n = 316) and came from single-parent homes (52%, n = 215), where the mother was the primary parent figure. The highest level of parental education achieved was typically a high school diploma (82%, n = 338). The average number of total prior petitions to juvenile court was three (SD = 2.3), and the most serious adjudicated charge for which youth were enrolled were crimes against persons (37%, n = 153), property crimes (29%, n = 120), drug offenses (18%, n = 74), and weapons offenses (11%, n = 45). A comparison between the 413 youth included in these analyses and the 134 excluded cases revealed no significant differences with respect to age, gender, ethnicity, parental education, and adjudicated charge. The included group, however, averaged more total prior petitions (2.64 versus 2.07, t[544] = 2.36, p < .05) and had more youth from homes where the mother was the sole parental figure (37% versus 26%, χ²(1) = 5.13, p < .05).

IDENTIFYING THE INITIAL COURT PLACEMENT AND AFTERCARE PERIOD

The aim of the larger longitudinal study is to describe the adjustment of serious adolescent offenders across multiple domains of functioning. Accordingly, a baseline interview was completed within 75 days of a youth’s adjudication hearing in juvenile court, and follow-up interviews were conducted every 6 months thereafter (additional information regarding study procedures can be found in Schubert et al., 2004). For the current analyses, we examined only the period of time that reflected the 6-month period of reentry after an adolescent’s initial institutional placement—from this point forward, this is called the aftercare period.

Two preliminary data management tasks had to be addressed before undertaking analyses of the aftercare period. First, the length of the initial residential placement and the date of release had to be identified; this involved integrating information from court records (e.g., date of disposition hearing, location of the commitment facility) with data obtained from participant interviews. Second, because the Pathways study was designed to track youth every 6 months after a baseline interview (and not following release from a residential setting), the data had to be restructured so that the first month of each offender’s reentry period represented the first month that he or she returned to the community (and not necessarily the first month covered by a follow-up interview). The strategy used to complete each of these tasks is described in some detail below because these procedures affect the types of data available for later analyses.

During each follow-up interview, youth were asked certain questions about the entire recall period (e.g., “How would you describe the relationship with your parents during the past 6 months?”) and other questions about each separate month (e.g., “Where were you living in Month 1, Month 2, etc. of the recall period?”). The monthly data were obtained using a life-event calendar approach (Belli, 1998; Caspi, Moffitt, Thornton, & Freedman, 1996), whereby the research participant is provided with a visual calendar that contextualizes the recall of research data by anchoring information to salient events (e.g., a birthday). Previous work has found that this method produces reasonably accurate information about youths’ participation in antisocial behavior and criminal offending (Horney, Osgood, & Marshall, 1995). More important for the current analysis, an earlier examination of our own data revealed that the life-event calendar approach generated reliable information about the occurrence and timing of residential stays for Pathways participants from Philadelphia (see Mulvey, Schubert, & Chung, 2007).¹

To determine the length of the initial court placement and date of release, the living situation for each youth was first plotted across the first 2 years of the study (e.g., Facility X in Month 1, Facility Y in Month 2) and then matched to the calendar month and year that corresponded to each of the 24 months. Facility names from the monthly data were then compared with information in court records to identify the initial dispositional commitment for each youth and the specific months he or she entered and exited the placement.² The 413 participants stayed an average of 253 (SD = 125) days in their initial placements and were released from the following types of settings: contracted residential facility (66%, n = 274); state-operated training school (15%, n = 63); contracted residential mental health facility (11%, n = 46); jail (4%, n = 17); and drug and alcohol facility (3%, n = 13).

Because of varying lengths of institutional placement, youth returned to the community at different points across the data collection periods: 22% (n = 90) between the baseline and 6-month interview; 39% (n = 162) between the 6- and 12-month interview; 31% (n = 132) between the 12- and 18-month interview; and 7% (n = 29) between the 18- and 24-month interview. As such, it was possible for a youth’s aftercare period to span different portions of two follow-up recall periods. For example, if a participant was released at the start of the recall period covered by the 18-month follow-up interview, his or her entire aftercare period was likely captured by data collected at the 18-month interview. However, if the participant was released toward the end of this follow-up period, we restructured the data to use portions of both the 18-month interview (for the initial part of the aftercare period) and 24-month interview (for the remaining months). Importantly, the flexibility of monthly data allowed us to combine and examine data across aftercare periods that did not neatly overlap with the fixed interval of the follow-up interviews.

MEASURES

Four types of data that could be related to successful adjustment were examined: (a) demographic characteristics, (b) factors indicating future risk for offending or other negative outcomes, (c) types of postrelease court supervision and CBSs, and (d) youths’ relationships to peers, parents, and other adults during the aftercare period. Because different data were used to assess these final sociocontextual variables, these measures are described in a later section. We also constructed indicators of adjustment using multiple measures of behavior during the aftercare period.

In general, demographic and risk factors reflected offender characteristics at or during the year prior to the baseline interview before youth went to their placements. In contrast, measures of court supervision and CBSs, as well as indicators of community reintegration, reflected participants’ circumstances after release from institutional settings. These latter measures were constructed from monthly follow-up data representing the aftercare period.

Demographic Characteristics

Demographic factors included age, gender, ethnicity, and site. Age represented each offender’s age at the start (i.e., Month 1) of the aftercare period.

Risk Factors

These included five type of factors previously identified as salient risk characteristics that can impact adjustment outcomes during community reentry (Hawkins et al., 1998; Wiebush et al., 2005): neighborhood characteristics, parental education, history of antisocial behavior, academic or cognitive functioning, and history of mental health problems.

Neighborhood characteristics were based on the street address associated with a youth’s first month of reentry (i.e., address to which he or she returned in the community). To determine neighborhood residence, each participant’s address was matched to a 2000 census tract, and a composite score of neighborhood disadvantage was computed using four census variables: percentage of households living below the poverty line, percentage of households receiving public assistance, percentage of unemployed residents, and percentage of residents with less than a high school education. A principal component analysis confirmed that these variables represented one factor that accounted for about 78% of the variance, with higher scores reflecting greater levels of neighborhood disadvantage.

Parental education was assessed as the highest level of schooling completed by the mother and father at the baseline interview. The score was computed by averaging the highest level achieved by each parent (or the highest for one parent in the case of single-parent families). Scores ranged from 1 (grade school or less) to 6 (some graduate/professional school).

Youths’ history of antisocial behavior was characterized using four variables: age at first arrest, number of court petitions (past year), lifetime level of aggressive offending, and lifetime level of income offending. The first two were taken from court records and the others were derived from an adaptation of the Self-Report of Offending (SRO) inventory (Huizinga, Esbensen, & Weiher, 1991; Thornberry & Krohn, 2000). Each of the SRO scales was based on variety scores that represented the proportion of 10 (income offending) or 11 (aggressive offending) offenses that the youth reported committing at least one time.

Two variables were created to indicate academic or cognitive functioning: level of academic commitment (past year) and the presence or absence of a cognitive deficit. Academic commitment was the average value of standardized scores for grades, attendance, and two scales that assessed bonding to teachers and school orientation (Cernkovich & Giordano, 1992). A previously conducted CFA showed that this four-indicator model was a good fit to the Pathways data, with higher values reflecting greater academic commitment (Mulvey et al., 2007). To identify cognitive deficits, this study considered scores from two tasks administered during the baseline interview, the Wechsler Abbreviated Scale of Intelligence (WASI; Wechsler, 1999) and Stroop Color and Word Test (Golden, 1978). Although neither the WASI nor Stroop is used in isolation to diagnose learning disabilities or cognitive disorders, the measures are often used to assess educational needs or cognitive impairment that warrant rehabilitative attention. A cognitive deficit was indicated for a full scale IQ of 70 or less (at least two standard deviations below the mean of 100) or meeting the Stroop criterion for prefrontal disorder/pathology (suggesting cognitive and/or neurological deficits such as dyslexia or reading disabilities).

Mental health problems were characterized as the presence or absence of select mood/anxiety disorders (major depression, posttraumatic stress, dysthymia, manic episode), substance use disorders (alcohol abuse/dependence, drug abuse/dependence), or use of CBSs for drug, alcohol, or other kinds of emotional problems during the 12 months prior to baseline. Mental health disorders were evaluated with the Composite International Diagnostic Interview (CIDI; World Health Organization, 1990), which uses criteria outlined by the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV). Youth were asked about their involvement in the following CBSs using a modified version of the Child and Adolescent Services Assessment (CASA; Burns & Goldman, 1998, p. 82): individual sessions (psychologist, counselor, social worker, mentor, priest/religious figure); group sessions (community support group, self-help group, mental health treatment group); in-home services; partial hospitalization/day program; school-based services.

Postrelease Court Supervision and CBSs

In contrast to most of the demographic and risk factors, the supervision and service use variables reflect what happened to adolescents in the community after being released from their initial court commitments. During each month of the aftercare period, youth reported on the number of times that they participated in four types of court-supervision programs and six types of CBSs. Court-supervision activities included probation (face-to-face or telephone meetings with a probation officer), community-intensive supervision programs, court-ordered groups (e.g., victim awareness), and drug court. CBSs included the five services described earlier (individual, group, in-home, partial hospitalization/day program, school-based) and also job training and job placement.

These reports were used to create two dimensions of involvement, intensity and duration, to represent the level or dosage of postrelease supervision and services that participants received during the aftercare period. Similar to the strategy used in previous studies of the IAP model (e.g., Wiebush et al., 2005), intensity scores represented the mean number of supervision and/or service contacts received per month (sum of total contacts across the aftercare period divided by the number of months that youth participated in the activity). The second dimension, duration, represented the count of aftercare months that youth participated in a given activity (number of months that youth reported at least one supervision or service contact). Separate intensity and duration scores were computed for each of the four supervision and six CBS categories in addition to total scores that combined categories within these two groups.

Postrelease Adjustment Outcomes

Indicators of successful reintegration include staying out of legal trouble and reconnecting with traditional social and community institutions like school and employment. Based on previous research, five outcomes were selected to indicate positive and negative adjustment during each month of community reentry. Three indicators of positive adjustment included school attendance, employment, and engagement (a consolidated measure of school and work involvement). For each month of the aftercare period, participants were given credit for attending school if they reported being enrolled and not missing more than 5 days during the month. To code employment, youth were asked about legal jobs and the number of hours worked. They were considered employed in any given month if they worked at least 21 hours per week (i.e., more than part-time) for 1 week during the month.³ The two outcomes were also combined to create an overall indicator of engagement (whether youth attended school or were employed during each month of the aftercare period).

Two indicators of negative outcomes assessed self-reported antisocial activity and formal system involvement. Antisocial activity was indicated if youth reported engaging in two or more of the antisocial acts listed on the SRO inventory during any given aftercare month; the threshold was set at two acts to avoid detecting an isolated slip in behavior. Formal system involvement was indicated for youth who reported being arrested or having an overnight stay of at least 7 days in a residential or rehabilitative setting (e.g., detention center, drug and alcohol facility). Because this latter measure included admissions to both secure and rehabilitative settings, it assessed general adjustment difficulties that resulted in youth being removed from his or her community. Examples of precipitating events included a new arrest, probation violation, and referrals for mental health problems that warrant residential treatment.

RESULTS

The results are presented in four parts. The first two sections present empirical portraits of the reintegration outcomes and participants’ receipt of supervision and CBSs after they returned to the community. These data were then combined to ask whether the duration and intensity of aftercare services are related to outcomes after controlling for relevant demographic characteristics and risk variables. In the fourth and final section, we used a subset of offenders to validate study findings while considering postrelease levels of peer deviance, parental monitoring, and adult support; the final section further describes these postrelease factors and the strategy used to identify the reduced sample. Because only 23% (n = 96) of offenders were from Maricopa County, analyses lacked adequate power to examine research questions separately in the two counties. Site, however, was included as a dichotomous covariate in each of the analyses to help account for potential differences between the two research locations.

EMPIRICAL PORTRAIT OF REINTEGRATION OUTCOMES

As further described below, slightly different samples were used to examine reintegration outcomes. The first outcome, formal system involvement, considered all 413 offenders, but a reduced sample of 355 was used for all other outcomes. This is because 58 of the youth who became involved with the system were readmitted to a residential facility within the first 3 months of the aftercare period and removed from the community. These youth were excluded from subsequent analyses because they did not have equal opportunity to achieve reintegration outcomes nor to experience community aftercare supervision and services. Compared to the 58 “early readmission” cases, the 355 offenders were generally older during the aftercare period (17.00 versus 16.55, t[411] = −2.68, p < .01) and also at the time of their first court petition (14.17 versus 13.53, t[411] = −2.85, p < .01), were more likely to be from Philadelphia County (80% versus 20%, χ²[1] = 12.44, p < .001), and had fewer petitions in the year before study participation (0.78 versus 1.25, t[410] = 2.82, p < .01). There were no significant differences in regard to gender, ethnicity, and the most serious enrollment charge.

Formal System Involvement

During the aftercare period 35% of 413 offenders (n = 143) became involved with the system. On average, this system involvement occurred 2.6 months (SD = 1.62) following release back into the community. For descriptive purposes, Table 1 presents bivariate comparisons (t tests for continuous measures, chi-square tests for categorical variables) of demographic and risk factors among those who did and did not achieve this outcome.⁴ As described in the table, both site and ethnicity were significant. Proportionally more offenders from Maricopa County (46% versus 31%) met the outcome, and with respect to ethnicity, 41% of Caucasian, 29% of African American, 46% of Hispanic, and 33% of other ethnicity offenders became involved with the system. This outcome was also more likely for offenders who were younger, had more prior petitions to court, had their first petition to court at an earlier age, and participated in a larger variety of both aggressive and income offenses. These youth also demonstrated less academic commitment, were more likely to have met criteria for a substance use disorder, and had been involved in CBSs during the year prior to entering the study.

TABLE 1. Comparison of Demographic and Risk Factors Across Community Reintegration Outcomes.

	System Involvement (N = 413)		Antisocial Activity (N = 355)		School Attendance (N = 281)		Engagement (N = 355)
	Yes (n = 143)	No (n = 270)	Yes (n = 96)	No (n = 259)	Yes (n = 198)	No (n = 83)	Yes (n = 285)	No (n = 70)
Variable	%	%	%	%	%	%	%	%
Demographic factors
Site	p < .05		p < .01		ns		ns
PA	69.2	80.7	69.8	83.4	81.3	73.5	80.0	78.6
AZ	30.8	19.3	30.2	16.6	18.7	26.5	20.0	21.4
Gender	ns		ns		ns		ns
Male	95.1	90.0	94.8	90.3	90.4	91.6	92.3	88.6
Female	4.9	10.0	5.2	9.7	9.6	8.4	7.7	11.4
Ethnicity	p < .05		ns		ns		ns
Caucasian	18.2	13.7	16.7	13.1	14.6	8.4	15.4	8.6
Hispanic	30.1	18.9	27.1	19.3	19.7	28.9	21.4	21.4
African American	48.3	63.7	51.0	64.4	62.1	59.0	59.3	67.1
Other	3.5	3.7	5.2	3.1	3.5	3.6	3.9	2.9
	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)
Mean age	16.74 (1.08)	17.03 (1.22)**	17.04 (1.23)	16.98 (1.17)	16.57 (1.16)	17.32 (1.08)***	16.89 (1.21)	17.41 (1.04)**
Risk factors
Mean neighborhood disadvantage	0.76 (0.92)	0.70 (0.86)	0.72 (0.94)	0.75 (0.85)	0.73 (0.87)	0.84 (0.93)	0.71 (0.86)	0.87 (0.89)
Mean parent education	2.71 (0.81)	2.66 (0.78)	2.77 (0.80)	2.65 (0.78)	2.68 (0.76)	2.61 (0.81)	2.72 (0.79)	2.50 (0.78)*
Mean age at first prior	13.83 (1.58)	14.21 (1.57)**	13.97 (1.60)	14.24 (1.50)	13.98 (1.50)	14.08 (1.58)	14.20 (1.53)	14.06 (1.54)
Mean no.of priors (past year)	1.09 (1.46)	0.71 (0.97)**	0.68 (0.96)	0.81 (1.10)	0.72 (0.98)	0.70 (0.89)	0.78 (1.09)	0.77 (0.97)
Mean aggressive offending	0.36 (0.20)	0.30 (0.19)**	0.31 (0.20)	0.31 (0.20)	0.29 (0.19)	0.32 (0.20)	0.31 (0.19)	0.32 (0.21)
Mean income offending	0.44 (0.24)	0.37 (0.24)**	0.37 (0.25)	0.37 (0.24)	0.34 (0.24)	0.37 (0.24)	0.38 (0.24)	0.36 (0.24)
Mean academic commitment	−0.09 (0.49)	0.02 (0.52)*	−0.01 (0.52)	−0.00 (0.51)	0.01 (0.51)	−0.05 (0.54)	0.01 (0.51)	−0.06 (0.52)
	%	%	%	%	%	%	%	%
Cognitive deficits	22.4	24.4	25.0	22.4	24.2	27.7	22.5	25.7
Mood/affective Dx	13.6	10.0	8.3	11.7	10.2	10.8	10.2	12.9
Substance use Dx	50.0	37.9*	41.7	39.3	34.5	48.2*	39.6	41.4
Prior CBSs	54.5	40.7**	45.8	43.2	45.5	43.4	43.5	45.7

Note. Dx = diagnosis; CBSs = community-based services.

^*p < .05.

^**p < .01.

^***p < .001.

Antisocial Activity

Among the 355 youth who were in the community at least 4 months, 27% (n = 96) reported being involved in antisocial activity during the aftercare period. On average, these youth met the criterion for only one sixth of the months (M proportion of months with antisocial activity = 0.12, SD = 0.23) and did so for the first time about 1 month after returning to the community (M months to antisocial activity = 1.40, SD = 1.61). Again, proportionally more youth from Maricopa (versus Philadelphia) County met this outcome (40% versus 24%). No other demographic or risk factors distinguished the two groups.

School Attendance

About 21% (n = 73) of 355 offenders reported that they already graduated by the start of the aftercare period. Of the 281 youth eligible to attend school, 71% (n = 198) reported attending for at least 1 month without missing 5 or more days. Most reported attending the same month they were released (M months to school attendance = 0.43, SD = 0.99), and 47% (n = 131) maintained this outcome for at least half of their aftercare period. As shown in Table 1, participants with regular attendance were generally younger and less likely to have had a substance use disorder during the year before their court commitment.

Employment

Some 42% (n = 151) of 355 offenders reported having a job at least 21 hours per week. In general, youth began working the same month as their release (M months to employment = 0.61, SD = 1.21) and maintained some type of employment at this level for over half of their aftercare months (M proportion of months employed = 0.52, SD = 0.29). Given uncertainties about the effects of employment on adolescent adjustment, we did not compare employed and nonemployed offenders on demographic and risk factors. Instead, such comparisons were reserved for the engagement outcome that is more likely to tap youths’ overall involvement in productive community activities.

Engagement

During the aftercare period, 80% (n = 285) of participants were either in school or working. They typically became engaged the same month they were released (M months to engagement = 0.72, SD = 1.21) and met the criterion during more than half of the followup months (M proportion of months engaged = 0.50, SD = 0.36). As described in Table 1, engaged youth were generally older and came from households with higher levels of parent education as compared to their nonengaged counterparts.

EMPIRICAL PORTRAIT OF POSTRELEASE COURT SUPERVISION AND CBSs

This section turns to a description of the supervision and CBS activities reported by participants who remained in the community for at least 4 months (N = 355). Table 2 describes the proportion and number of youth who participated in each type of activity, the number of months it took to first report it, and the intensity and duration of each. These data are presented separately for the four supervision activities and six CBSs, and also for the total measures that combine information across all types of supervision and CBSs, respectively.

TABLE 2. Use of Postrelease Court Supervision and Community-Based Services During the Aftercare Period (N = 355).

	Frequency	Months Until First Contact		Intensity		Duration
	%	M	SD	M	SD	M	SD
Court supervision
Probation (n = 302)	85.1	1.23	0.74	3.35	3.50	4.18	1.76
Intensive supervision (n = 37)	10.4	1.32	0.94	13.00	16.75	2.86	1.65
Court-ordered groups (n = 7)	2.0	1.74	1.50	6.64	8.30	1.86	1.86
Drug court (n = 2)	< 1.0	1.00	1.00	8.00	9.89	2.50	0.71
Total supervision (n = 303)	85.4	1.16	1.61	4.40	6.60	4.29	1.73
Community-based services (CBSs)
Individual (n = 61)	17.2	1.74	1.30	5.87	5.63	3.10	1.69
Group (n = 27)	7.6	1.84	1.49	5.30	4.39	3.00	1.73
In-home (n = 18)	5.1	2.17	1.62	3.68	1.92	3.00	1.71
Partial/day program (n = 10)	2.8	1.30	0.48	11.72	9.02	1.50	0.85
School-based (n = 29)	8.2	1.49	1.02	9.17	7.54	2.90	1.75
Vocational (n = 24)	6.8	2.63	1.70	8.67	9.96	1.71	1.00
Total CBSs (n = 124)	34.9	1.60	1.12	7.51	7.38	3.15	1.78

As indicated in the table, 85% (n = 302) of youth were under some type of court supervision, but only 35% (n = 124) reported participating in CBSs. The initiation of both types of services, however, was indicated within 2 months of release. Among those involved with supervision programs, the lowest rates were observed for specialized judicial services, such as court-ordered groups and drug court, and the highest for probation. Youth averaged 4.40 contacts per month across all programs and were usually supervised for the full duration of their aftercare period. Among the CBSs, individual sessions were used most often, followed by group services and school-based programs. Youth who used CBSs averaged 7.51 contacts per month across all types of services and reported using them for varying proportions of their aftercare period.

It was surprising to find that about 15% and 65% of reentry offenders reported no participation in postrelease supervision and CBSs, respectively. Table 3 describes demographic and risk factors that distinguished among those who did and did not participate in the services. Results of these bivariate comparisons (t tests for continuous variables, chi-square tests for categorical variables) revealed that youth were more likely to be supervised in Philadelphia versus Maricopa County (89% versus 72%). Youth receiving supervision also tended to be younger and report a lower variety of both income and aggressive offenses.

TABLE 3. Comparison of Demographic and Risk Factors among Offenders Who Did and Did Not Use Postrelease Supervision and Community-Based Services (N = 355).

	Supervision		Community-Based Services
	Yes (n = 303)	No (n = 52)	Yes (n = 124)	No (n = 231)
Variable	%	%	%	%
Demographic factors
Site	p < .001***		p < .01**
PA	82.8	61.5	71.0	84.4
AZ	17.2	38.5	29.0	15.6
Gender	ns		ns
Male	91.4	92.3	90.3	92.2
Female	8.6	7.7	9.7	7.8
Ethnicity	p < .01**		p < .01**
Caucasian	13.2	19.2	21.8	10.0
Hispanic	21.5	21.2	23.4	66.7
African American	62.0	53.8	50.0	20.3
Other	3.3	5.8	4.8	3.0
	M (SD)	M (SD)	M (SD)	M (SD)
Mean age	16.91 (1.21)	17.55 (0.89)*	16.56 (1.16)	17.23 (1.14)***
Risk factors
Mean neighborhood disadvantage	0.76 (0.89)	0.62 (0.76)	0.59 (0.84)	0.82 (0.88)*
Mean parent education	2.67 (0.77)	2.74 (0.92)	2.75 (0.82)	2.64 (0.77)
Mean age at first prior	14.15 (1.55)	14.31 (1.42)	14.15 (1.31)	14.18 (1.64)
Mean no.of priors (past year)	0.76 (0.99)	0.88 (1.41)	0.88 (1.24)	0.72 (0.95)
Mean aggressive offending	0.30 (0.19)	0.37 (0.23)*	0.32 (0.20)	0.31 (0.20)
Mean income offending	0.36 (0.24)	0.45 (0.24)**	0.38 (0.25)	0.37 (0.24)
Mean academic commitment	−0.01 (0.51)	0.01 (0.52)	−0.01 (0.51)	−0.00 (0.52)
	%	%	%	%
Cognitive deficits	24.1	17.3	13.7	28.1**
Mood/affective Dx	10.3	13.5	9.8	11.3
Substance use Dx	38.2	50.0	45.5	37.0
Prior CBSs	43.2	48.1	53.2	39.0*

Note. Dx = diagnosis; CBSs = community-based services.

^*p < .05.

^**p < .01.

^***p < .001.

Several features also distinguished those who did and did not use CBSs. On average, youth participating in CBSs were younger, from neighborhoods with less disadvantage, less likely to have a cognitive deficit, and more likely to have received CBSs in the year prior to study enrollment. Site and ethnicity were also associated with CBS use; proportionally more youth from Maricopa County (51% versus 31%) and more Caucasian youth (54%) reported using CBSs (compared to 29% of African American, 38% of Hispanic, and 46% of other ethnicity youth).

PREDICTING REINTEGRATION OUTCOMES

This section reports the results of logistic regression analyses that examined multivariate relations among demographic and risk factors, participation in postrelease supervision and services, and three separate reintegration outcomes: readmission, recidivism, and engagement. To address this issue in as parsimonious fashion as possible, analyses only included risk variables that were associated with a given outcome at p < .15 in previous bivariate analyses (see Table 1). In each regression analysis, all demographic variables were entered first, followed by select risk factors, and then the intensity and duration measures for supervision and CBSs (a score of 0 was used if youth did not report participating in a given postrelease activity).

For these analyses, the definition for duration and intensity (of supervision and services) was adjusted to reflect levels for only the period prior to the month in which a participant first reached a particular outcome. For example, if a youth reported antisocial activity for the first time in Month 4, all duration and intensity measures were computed to reflect levels of supervision and CBSs during only the first 3 months. This approach was adopted because the use of duration or intensity measures that reflected the whole aftercare period would be confusing for these purposes. If an adolescent, for instance, engaged in a high level of antisocial activity in Month 3 of the aftercare period, this behavior might precipitate more service involvement in the subsequent months. In this case, looking at the values for service involvement during the entire aftercare period might indicate a positive relation between antisocial behavior and service use. But it would be erroneous to conclude that more intense services increased the likelihood of antisocial activities, as the effect would be going in exactly the opposite direction.

Table 4 presents results of the logistic regression analysis for formal system involvement. Regarding demographic and risk factors, youth were less likely to be involved if they were older (odds decreased 23% with each additional year, p < .01) and used CBSs the year prior to study enrollment (odds decreased 46%, p < .05). Other factors that were significant at the bivariate level (site, ethnicity, substance use diagnosis) were no longer important for explaining system involvement. However, two dimensions of supervision and services—duration of supervision and intensity of services—significantly reduced the odds of formal system involvement: 44% reduction with each additional month of supervision (p < .001) and 12% reduction for each additional service contact per month (p < .05).

TABLE 4. Logistic Regression Models of System Involvement, Antisocial Activity, and Engagement.

	System Involvementa (N = 395)		Antisocial Activityb (N = 340)		Engagementc (N = 335)
Predictor Variable	B	OR	B	OR	B	OR
Demographic characteristics
Philadelphia County	−0.49	0.61	0.14	1.15	−1.00	0.37
Male	0.97	2.64)	1.21	3.36	0.87	2.39
Age	−0.35	0.71*	−0.09	0.91	−0.55	0.58**
Ethnicity	ns	ns	ns	ns	ns	ns
African Americand	−0.15	0.86	−0.25	0.78	−0.73	0.48
Hispanicd	0.52	1.68	0.48	1.62	−0.30	1.35
Select risk factors
Parental education	—	—	0.25	1.29	0.66	1.94**
Antisocial behaviore	0.20	1.22	—	—	—	—
Substance use disorder	−0.10	0.90	—	—	—	—
Academic commitment	−0.48	0.62	—	—	—	—
Previous community services	−0.62	0.54*	—	—	—	—
Supervision and services
Supervision—intensity	0.03	1.03	0.03	1.03	0.62	1.85***
Supervision—duration	−0.58	0.56***	−0.76	0.47***	−0.83	0.44***
Services—intensity	−0.13	0.88*	−0.02	0.98	0.26	1.30
Services—duration	0.18	1.20	−0.08	0.92	−0.29	0.75
Final model statistics
Chi-square (df)	98.62/13***		91.67/10***		83.56/10***
NR2 (% classified correctly)	.31/73.4		.34/77.1		.36/4.5

Note. Final analyses were limited to Caucasian, African American, and Hispanic participants because small numbers of youth in the “other” ethnic group caused problems with model convergence. B = logistic regression coefficient; OR = odds ratio; NR²= Nagelkerke R².

^aSample based on N = 413 and excludes 18 other ethnicity youth.

^bSample based on N = 355 and excludes 15 other ethnicity youth.

^cSample based on N = 355 and excludes 15 other ethnicity youth and 5 outliers with a Cook’s distance greater than 1.0.

^dReference group is Caucasian.

^eAntisocial behavior was a composite variable that averaged standardized scores for four indicators (previously conducted confirmatory factor analysis showed that this four-indicator model of antisocial behavior was a good fit to the data (data, Mulvey et al., 2007).

^*p < .05.

^**p < .01.

^***p < .001.

Similar to the analysis for system involvement, increased duration of court supervision reduced the odds of engaging in antisocial activity (reduced 53% with each additional month, p < .001). The intensity of services, however, was not important for explaining antisocial behavior. After controlling for all other factors in the model, no other demographic or risk characteristics emerged as significant predictors.

In contrast to the two previous analyses, the intensity of supervision emerged as an important predictor of engagement. Each additional unit of supervision intensity (i.e., one additional contact per month) increased the odds of engagement by 85% (p < .001). The duration of supervision was also important but in the unexpected direction (each additional month of supervision decreased the odds of engagement by 56%, p < .001). Finally, analyses revealed that engagement was less likely among older adolescents (odds decreased 42% with each additional year, p < .01) and more likely among youth whose parents had higher levels of education (odds increased 94% with each additional unit, p < .01).

Interestingly, a closer look at the negative link between supervision duration and engagement revealed that a prevalence of “early engagers” in the sample may help to explain the unexpected result. As noted earlier, youth typically got engaged the same month that they were released (184 offenders, or 52%, were attending school or working within the same month that they returned to the community). The duration score for these youth would be constrained to a maximum value of 1, and in essence, these cases would not provide a valid test of duration because they did not have sufficient time before the outcome to experience any meaningful amount of supervision. To test this hypothesis, “early engagers” were excluded from the sample and the data were reanalyzed (N = 161). As suspected, the link between the duration of supervision and engagement became nonsignificant, although the intensity of supervision continued to exert a positive effect (B = −.30, SE = .16, p < .01).

CONSIDERING CONTEXTUAL INFLUENCES IN THE PREDICTION OF REINTEGRATION OUTCOMES

The primary goal of this set of analyses was to validate our findings for antisocial activity and engagement after controlling for levels of parental monitoring, peer deviance, and support from caring adults. In particular, we wanted to confirm that supervision and service effects were not simply reflections of these sociocontextual factors producing more involvement with service systems and promoting positive outcomes. The system involvement outcome was not considered because 18% of the full sample (n = 74) was readmitted to a residential facility within the first 3 months of release and spent less than half of their aftercare period in the community.

Participants and Procedures

The current analyses were limited to a subset of 166 youth who provided valid data for parental monitoring, peer deviance, and caring adults. These variables were not valid for all participants because, in contrast to the assessment of aftercare activities, data on the sociocontextual factors were collected only at the recall-period level (i.e., youth provided average ratings that summarized experiences during the entire recall period). And because youth were released at different points between follow-up interviews, variables assessed at the recall level sometimes described only a small portion of the youth’s aftercare experience. To be considered for the present analysis, we required that at least 4 months of a participant’s aftercare period fall within the recall period of a single follow-up interview. This way, recall data would describe at least two thirds of the aftercare period, a proportion that seemed reasonable to represent youths’ postrelease experiences. A total of 166 participants met the 4-month criterion.

This subsample was predominantly male (93%), African American (63%; Caucasian = 11%, Hispanic = 25%), from Philadelphia County (77%), and was about 17.17 (SD = 1.12) years of age at the time of release. Compared to the 174 participants who were excluded from the analysis, the present sample was generally older (t[338] = −2.48, p < .05) but was not different in terms of site, parental education, gender, or ethnicity.

Measures

Three sociocontextual variables were considered for these analyses: parental monitoring, peer deviance, and diversity of caring adults. The 4-item parental monitoring scale described the extent to which parents supervise and have rules for their adolescent (Steinberg, Lamborn, Dornbusch, & Darling, 1992). Because preliminary analyses showed that more than 40% of participants reported being responsible for themselves during the aftercare period, this variable was recoded into three categories prior to analysis: 0 (no caregiver), 1 (low monitoring), or 2 (high monitoring). The median split score was used to distinguish low and high groups. Peer deviance was assessed with an adaptation of a widely used scale developed for the Rochester Youth Study (Thornberry, Lizotte, Krohn, Farnworth, & Jang, 1994); scores were averaged across the 10-item scale. The diversity of caring adults was assessed with questions from two sources (Institute of Behavioral Science, 1990; Nakkula, Way, Stauber, & London, 1990). The measure indicated the count of unique adults whom youth could access for support across eight domains (e.g., adults you admire and want to be like, adults you can talk to about trouble at home).

Results

Separate logistic regressions were conducted to predict antisocial activity and engagement. Given the reduced sample size, analyses only included demographic, risk, and contextual factors that showed significant bivariate relations with the outcomes at p < .15. This modified strategy was adopted because there was not enough power to include all demographic characteristics in the analyses.

The final model for antisocial activity included gender and the four supervision or services variables. No other demographic, risk, or contextual factors showed important bivariate relations with the outcome (the results of bivariate analyses are not presented here but are available from the authors on request). Consistent with our prior results, the duration of court supervision emerged as the only significant predictor (p < .05) of self-reported antisocial activity. The odds of this outcome were reduced by 55% with each additional month of court supervision.

The final model for engagement included age, parental education, and all three sociocontextual factors. Again, consistent with our prior results, parental education, age, and the intensity and duration of court supervision were all associated with engagement. Each unit increase in parental education increased the odds of engagement by 91% (p < .05), and each unit increase in age reduced the odds by 42% (p < .05). In addition, after controlling for sociocontextual variables, supervision intensity and duration still emerged as important predictors (odds of engagement increased 40%, p < .05, with each additional supervision contact per month and decreased 45%, p < .01, with each additional month of supervision). Importantly, contextual variables also seemed to make a difference. Each unit increase in peer deviance reduced the odds of engagement by 56%, and each additional adult mentioned increased the odds by 60%.

DISCUSSION

This study provides a comprehensive description and analysis of the community reentry process for serious juvenile offenders. Several aspects of the study make it particularly informative for efforts to improve practice and policy in this area. For one thing, the fine-grained nature of the data and the selectivity and size of the sample produces an unusually detailed picture of the reentry process in a group of offenders with considerable clinical and policy salience, much like those described by Snyder (2004) as the high-risk reentry population. In addition, the study focuses on the first 6 months after release, the period during which intensive aftercare programs try to have an impact. Importantly, this study is able to provide a picture of the relative influence of a broad range of factors—individual risk characteristics, supervision and service involvement, contextual influences—on both positive and negative adjustment during this period. Finally, the study considers the influences of peer deviance, parental monitoring, and caring adults during the aftercare period, three potentially powerful factors that have been the focus of considerable theoretical, but little empirical, attention.

A few limitations are also important to note. First, the study is based on self-reports, and there could be some shared method variance in the reports of activities like offending and involvement with services. The use of life calendars, structured to obtain monthly data during a relatively short recall period, is an effective way to obtain data of this type (Belli, 1998), but it could certainly be subject to systematic bias. Our analyses of the concordance of calendar data with official records for periods of institutional placements are encouraging, but self-reports are by nature subject to under- or overreporting in certain situations (Roberts & Mulvey, in press). Second, the outcomes of system involvement, antisocial activity, school and work involvement, and engagement are represented rather simplistically here (i.e., as binary outcomes). Although reasonable for summarizing and making comparable the detailed data underlying these behaviors, this general representation may be masking some important processes. Having a job at particular follow-up points, for instance, does not tell us whether an adolescent kept a particular job for any extended time, and the effects of the factors tested may be different on more differentiated outcomes. Finally, the depiction of supervision and service provision is rather global. The variables measured here indicate how much supervision or service involvement the adolescent reported receiving but not the quality of or circumstances associated with these interactions. Thus, although this study can comment on the potential effect of having these experiences, it cannot address why a sizable proportion of these adolescents did or did not receive these interventions (e.g., the services were never recommended, the adolescent did not show up for a recommended service) or the most effective ways to provide supervision or CBSs. These limitations notwithstanding, several notable findings across the various analyses have clear implications for structuring these types of services and understanding their impact on reintegration during the facility-to-community transition.

LEVELS OF SUPERVISION AND SERVICE INVOLVEMENT

The first notable finding is that, when supervision is provided, there is a rather high level of involvement reported in the first 6 months. Supervision lasted for at least three quarters of the aftercare period and the mean level of contacts was between four and five per month, a frequency consistent with the recommended standard of weekly contact (Griffin, 2004). Looking just at the adolescents who reported receiving intensive supervision, youth reported approximately 13 contacts per month for almost 3 months, again generally consistent with recommendations for this more intense model of supervision (Wiebush et al., 2005).

On the other hand, a sizeable proportion of young offenders (15% of the sample) reported no supervision during this period. Although the reasons for this cannot be determined from the available data, one explanation—aging out of the system—seems highly plausible. These offenders, compared to their counterparts who received supervision services, were generally older in both sites and disproportionately more were from Maricopa County. As noted earlier, juvenile court supervision can extend beyond age 18 under Pennsylvania, but not Arizona, statute. Thus, state-specific regulations regarding juvenile justice authority may help to explain why youth in Maricopa County were less likely to receive court supervision.

Consistent with previous reports, involvement with CBSs was considerably lower than involvement with supervision. Only about 35% of offenders reported using some type of CBS (primarily individual services) during the aftercare period. At the same time, when they were involved with a service, the adolescents used it rather intensely, averaging about eight contacts per month with a service provider (almost double the rate of supervision contacts). The lower rate of involvement with CBSs is understandable given that these services are recommended mainly when a particular problem has been identified and attendance is usually voluntary. However, a number of researchers have already argued for increased service provision for high-risk offenders as they return to the community (Altschuler & Armstrong, 2001). Similar to our findings for supervision involvement, older youth were less likely to have contact with a service provider at any point during the aftercare period. CBS involvement was also less likely among minority offenders and those returning to neighborhoods with higher levels of community disadvantage. However, because these comparisons were conducted for descriptive purposes and did not control for other variables, additional research is needed before drawing any firm conclusions about the significance of these findings.

LEVELS OF NEGATIVE AND POSITIVE COMMUNITY REINTEGRATION OUTCOMES

The proportions of adolescents with negative outcomes were about what might be expected from a sample of this sort. Thirty-five percent (n = 143) reported having formal system involvement—rearrest or readmission to a residential facility—within the first 6 months. In addition, among those who lived in the community for at least 4 months, more than 27% reported significant antisocial behavior. Clearly, a sizable proportion of adolescents were involved in antisocial activities without ending up back in institutional care. The rates of system involvement observed here appear to fall within the range of recidivism rates reported in previous studies with longer follow-up periods (20% to 85%; Greenwood, Deschenes, & Adams, 1993; Rodriguez-Labarca & O’Connell, 2004; Wiebush et al., 2005).

A particularly striking point about the reintegration outcomes is the high level of engagement with school and/or work that was achieved within a short time after offenders returned to the community. In general, 80% of the adolescents in the sample were either attending school or employed for more than 20 hours per week within the same month of their release, figures that are in line with 12-month follow-up rates reported in other studies of community reentry (Deschenes & Greenwood, 1998; Greenwood et al., 1993). Interestingly, a sizable proportion of adolescents who reported an arrest or involvement in antisocial activities also had these events occur within the first month of the aftercare period (38% and 65%, respectively). Given the potential value of “getting started right,” it would be important for more targeted investigations to further examine instances of early failure during community reentry.

PREDICTING COMMUNITY REINTEGRATION

Perhaps the most obvious initial finding here is that individual demographic and risk characteristics appear to have little power for explaining the observed outcomes. Consistent with the findings of previous studies, relatively few individual factors showed significant bivariate relations with either negative or positive outcomes. The ones that did typically became nonsignificant when other variables were taken into consideration. The main exception to this were the findings for age, which was negatively associated with both system involvement and engagement after considering other individual factors and involvement with aftercare activities; older youth were at reduced risk of formal system involvement but at increased risk of disconnecting from community-based institutions during the facility-to-community transition.

Analyses also revealed an interesting difference in the contribution of postrelease contextual factors to the achievement of positive and negative outcomes. Specifically, peer deviance and the support of caring adults were important for predicting engagement but not self-reported antisocial activity. These findings confirm that the qualities of the social contexts to which these young offenders return can significantly influence youths’ overall adjustment in the community. Although previous research has indicated the protective role of low peer deviance and high adult support among delinquent youth, no studies have linked them to community reintegration outcomes. Our pattern of findings seem to indicate that these contextual factors are more influential in promoting positive adjustment, whereas the avoidance of negative outcomes may be more dependent on the provision of specific supervision and CBSs.

Arguably the most consistent conclusion across the analyses is that postrelease supervision matters. Interestingly, different dimensions of these activities were important for positive versus negative outcomes—duration reduced the odds of system involvement and self-reported delinquent behavior, and intensity increased the odds of engagement. In each of these circumstances, supervision helped to promote successful community reintegration, and these findings held after considering postrelease sociocontextual factors. This result may seem contradictory to the findings of previous studies reporting that supervision was ineffective (see Altschuler et al., 1999), but the present study differed from these earlier reports in a number of important ways. First, we assessed outcomes during the first 6 months—much earlier than previous reports—and may have captured the impact of postrelease supervision at a particularly sensitive time during the reintegration process. Second, supervision activities in this study evidenced more variability than in previous reports, which allowed there to be adequate power to observe meaningful relations between the length of supervision and study outcomes.

Regarding CBSs, CBS involvement—specifically, the intensity of CBSs—was important for only one outcome: More frequent contact with CBS providers reduced the odds of formal system involvement. In interpreting the null findings for the other outcomes, it is important to note that the analyses for antisocial activity and engagement were conducted with different samples. As described earlier, they excluded a group of “early failure” youth who returned to a residential facility within 3 months, and it is among these relatively low-risk offenders that CBS involvement had no appreciable impact beyond the effect of supervision. With respect to engagement, it is not too surprising that court supervision, but not CBSs, emerged as a significant predictor. It seems reasonable that court-supervised activities—often designed to connect youth with important community institutions and resources—would promote school and work involvement more effectively than CBSs that likely focus on more general issues related to adjustment. Although this result seems inconsistent with the findings of Bullis et al. (2004), who documented a link between service use and engagement, the two studies used different definitions for the outcome (the present study focused solely on school and work participation and Bullis and colleagues also considered antisocial and criminal behavior).

FUTURE DIRECTIONS

This study highlights at least three issues that warrant continued refinement and research. First, the findings represent overall effects of supervision, services, and contextual factors on outcomes in the community. As such, it would be important for futureanalyses to examine the differential impact of these experiences for subgroups of offenders (e.g., who vary according to risk for reoffending). Such analyses should consider multiple reintegration outcomes, as our results suggest that the processes behind refraining from antisocial activities may be distinct from those behind engaging in positive community activities.

Second, this study did not address the larger issue of long-term adjustment after institutional care. Several studies (Greenwood et al., 1993; Wiebush et al., 2005) have found that engagement does not always last for long periods after release (i.e., an adolescent might enroll in school or start a job but then not stick with it), and the effects of supervision or service involvement on this longer term adjustment process may be different than what was observed for the short period examined here. A closer examination of longitudinal reciprocal effects among aftercare activities and reintegration outcomes could inform efforts to structure resources in ways that help young offenders achieve and maintain positive outcomes in the community.

Finally, given the significance of service intensity to reducing risk for system involvement, it is important to understand why more youth are not using services in the aftercare period. To do this, careful work needs to be done to monitor the quality of services provided, the barriers to involvement with services, and the appropriateness of matching adolescents to service activities. In addressing these issues, it would be important for studies to consider the philosophy of the system that governs the reentry process. As described earlier, different entities are responsible for supervising juveniles following their release from state commitments, with most states using a corrections control model in which the agency that oversees the state’s juvenile corrections is entirely responsible for overseeing reentry. There are good reasons to think, however, that reentry programs governed by court control (supervision of reentry offenders is a local matter) or shared control (i.e., shared by the local court and the agency that oversees the state’s commitment institutions) would facilitate increased service involvement and more positive reintegration outcomes (Griffin, 2005).

CONCLUSION

Overall, our findings indicate that reentry offenders are sensitive to close and extended supervision and service contacts during the aftercare period and that different components of these processes may be important for motivating restraint from negative behaviors and promoting engagement in school and/or work. The general findings about the importance of aftercare services in reducing the likelihood of recidivism mirror results from other investigations of long-term criminal outcomes with adult offenders on parole (Burdon, Messina, & Pendergrast, 2004; Martin, Butzin, Saum, & Inciardi, 1999; Travis & Visher, 2005). Before this study, however, such findings were inconclusive among adolescent offenders, had not been investigated during the first 6 months after institutional placement, and had not been tested when controlling for sociocontextual factors. Although the current study provides several clues about factors that promote short-term adjustment, much still needs to be done to understand the dynamic processes that affect the community reentry process among serious juvenile offenders.

TABLE 5. Follow-Up Logistic Regression Models of Offending and Engagement.

	Antisocial Activity (N = 166)		Engagementa (N = 163)
Predictor Variable	B	OR	B	OR
Select demographic and risk characteristics
Age	—	—	−0.55	0.58*
Gender	−1.58	0.21	—	—
Parental education	—	—	0.65	1.91*
Select contextual factors
Peer deviance	−0.15	1.45	−0.82	0.44**
Diversity of caring adults	−0.09	1.03	0.47	1.60*
Parental monitoring	−0.37	1.03	0.09	1.09
Supervision and services
Supervision—intensity	0.03	1.03	0.34	1.40*
Supervision—duration	−0.80	0.45***	−0.59	0.55**
Services—intensity	0.03	1.03	0.53	1.70
Services—duration	−0.06	0.95	−0.29	0.75
Final model statistics
Chi-square (df)	43./70 8***		48.55/9***
NR2 (% classified correctly)	.34/75.9		.38/79.8

Note. B = logistic regression coefficient; OR = odds ratio; NR² = Nagelkerke R².

^aSample based on N = 166, which excludes 3 outliers with a Cook’s distance greater than 1.0.

^*p < .05.

^**p < .01.

^***p < .001.