Ecologies of juvenile reoffending: A systematic review of risk factors

Highlights

• Studies of residential context (i.e., ecological factors) in relation to youth reoffending have dramatically expanded.

• Concentrated disadvantage, the most frequently tested ecological factor, predicts re-arrest (pooled OR = 1.09, p = 00.01).

• As for other ecological factors, existing research is limited to few contexts and samples, and has yielded mixed results.

• Future research that enhances measurement, diversifies samples, and tests theoretical relationships is warranted.

Abstract

Purpose

Research on juvenile reoffending has experienced an ecological turn, marked by an impressive expansion of studies that test the relation between elements of residential context and reoffending. Yet, to date, little consensus exists regarding what ecological factors matter, how they affect reoffending, and for whom they matter most. To address this gap, this study takes stock of research that tests the relationship between ecological factors and reoffending among youth.

Methods

A systematic review, in accordance with PRISMA-P guidelines, of quantitative studies (k = 27) was conducted. Evidence was synthesized quantitatively (i.e., meta-analytically, tabularly) and qualitatively (i.e., narratively).

Results

A variety of ecological factors have been tested, but results are inconsistent and reflect relatively few contexts and samples. The most frequently tested factor, concentrated disadvantage (k = 15), is a predictor of re-arrest (pooled OR = 1.09, p = .01). Inconsistent findings regarding other factors seem to reflect sample and study characteristics.

Conclusions

Research to date does not indicate summarily rejecting or accepting ecological factors as risk factors for reoffending. To further clarify the ecology-reoffending relationship and inform recidivism reduction interventions, future research should sample from unexamined contexts and test theoretically meaningful relationships via approaches that strengthen causal inference.

Introduction

Repeat offending among justice-involved youth is a sizable problem with public safety consequences. Nearly half of youth adjudicated for a delinquent or criminal act will reoffend, and those who continue involvement in crime as adults ultimately account for the majority of all crime (Bullis, Yovanoff, Mueller, & Havel, 2002; Moffitt, Caspi, Harrington, & Milne, 2002; Moffitt, Caspi, Rutter, & Silva, 2001). Yet, not all youth reoffend. Identifying differences between those who do and do not reoffend—and related risk and protective factors for reoffending—is one of criminology's central aims.

Researchers have identified risk and protective factors for reoffending to inform risk assessment instruments and, hypothetically, to guide recidivism reduction interventions (Kennealy, Skeem, & Hernandez, 2017; Singh et al., 2014). Much of this work has focused on individual (e.g., anti-social cognitions) and relational (e.g., peer associations) risk factors for reoffending. Despite progress in these areas, the ability to predict and prevent reoffending remains limited (Schwalbe, 2007). If the individual and relational factors examined to date only partially explain reoffending among youth, what other factors may have predictive utility? Social structural, economic, and systemic ecological factors, also referred to as “macro,” “environmental,” or “neighborhood” effects, represent other potential sources of risk for reoffending.

In the past decade, numerous studies of reoffending among youth have emerged as part of a broader “cottage industry” (Sampson, Morenoff, & Gannon-Rowley, 2002) in ecologically-oriented research. In the wake of this ecological turn, consensus has yet to be reached regarding what ecological factors matter for reoffending, how they unfold to affect reoffending, or for whom they matter most. This paper takes a step toward filling this gap by systematically reviewing research on the ecology-reoffending relationship among youth, and qualitatively and quantitatively assessing the effects of ecological factors on youth reoffending.

As we suggest above, research on risk factors for juvenile reoffending has been dominated by a focus on individual and relational characteristics. Several reviews provide useful summaries of this literature (see Cottle, Lee, & Heilbrun, 2001; Dowden & Brown, 2002; Loeber & Dishion, 1983; Piquero, Jennings, Diamond, & Reingle, 2015; Wibbelink, Hoeve, Stams, & Oort, 2017). Generally, age, gender, and race predict reoffending (Piquero et al., 2015), with younger, male, and Black youth being more likely to reoffend– though there is some evidence that the effect of race disappears when conditioned on socioeconomic status (Cottle et al., 2001). Reoffending is also often, though not always, predicted by delinquent history, negative use of leisure time, family and parenting factors, association with anti-social peers, substance use, neurodevelopmental factors, and mental health problems (c.f. internalizing and severe disorders; Basto-Pereira, Começanha, Ribeiro, & Maia, 2015; Cottle et al., 2001; Wibbelink et al., 2017). The relevance of some of these factors to reoffending also varies along demographic dimensions (Basto-Pereira et al., 2015).

Prior research on individual and social factors highlights some important indicators for risk prediction and prevention. Yet, a systematic review of standardized risk assessment tools (k = 28; Schwalbe, 2007) indicated many instruments informed by this literature are marginally more accurate in predicting reoffending than flipping a coin (mean AUC = 0.64, where AUC = 0.50 indicates the instrument is as likely to incorrectly predict recidivism as it is to correctly predict reoffending). In the absence of tools with greater predictive utility and interventions that target known risk factors, repeat offending among youth continues.

In 1942, Ernest Burgess wrote, “If we wish to reduce delinquency, we must… think of its causes more in terms of the community and less in terms of the individual” (in Shaw & McKay, 1942, pp. xiii). Nearly 80 years after Burgess wrote about the relationship between residential context and delinquency, scholars of juvenile reoffending are heeding Burgess' words. This ecological turn, or return, is evidenced by a significant increase in the number of publications focused on ecological factors for juvenile reoffending (see Fig. 1). Contemporary research on the ecology-reoffending relationship is primarily built upon prior research on the ecology-crime relationship and the ecology-delinquency relationship. We describe these two areas of research below, highlighting the ways in which their theoretical bases and empirical findings may and may not support the ecology-reoffending relationship. Where possible, we turn to prior systematic and narrative reviews, which offer important distillations of these foundational bodies of research but have yet to be completed for research focused on the ecology-reoffending relationship.

Research on the ecology-crime relationship has long established that crime and delinquency rates are elevated in communities that share certain social and structural features (Pratt & Cullen, 2005; Sampson et al., 2002). Theories of social disorganization and, relatedly social capital, cohesion, and control, are among those most popular for explaining the relationship between community features and crime and delinquency rates. In the most basic sense, theorists in this vain argue that disorganized communities reflect the intersection of poor economic conditions, residential turnover, and elements of demographic and family composition (e.g., ethnoracial diversity and concentration, concentration of youth, female-headed households; e.g., Sampson, 1986a; Shaw & McKay, 1942). When these factors combine, their “concentration effects” produce crime and other negative social outcomes (Wilson, 2012). Social capital, collective efficacy, and control theorists (Bursik Jr & Grasmick, 1993; Sampson, Raudenbush, & Earls, 1997), argue that the relationship between these social and structural features and crime rates are at least in part explained by related reductions in formal and informal social capital and control, which are necessary to collectively control behavior and guard against the occurrence of crime (see also Wilson, 2012). This latter group of theories, with their ability to connect macro- and micro-forces to crime and delinquency, have increasingly garnered attention (see, e.g., Groff, 2015; Sampson, 2012; Warner, 2014).

Empirical research provides support for the ability of social disorganization-related theories to explain variation in crime and delinquency rates between communities, while some research indicates alternative theories may also be relevant. In a meta-analysis of “macro” predictors of crime (k = 216), Pratt and Cullen (2005) found neighborhoods with higher crime rates tended to be distinct in dimensions relevant to social disorganization and social capital. Specifically, high crime neighborhoods were high in concentrated disadvantage (poverty, racial composition, unemployment, and family disruption) and residential mobility, and tended to be low in collective efficacy and social ties (measures of social interactions and strength of social ties; e.g., social capital; mean effect sizes = 0.17–0.41). Pratt and Cullen also found unsupervised local peer groups and inequality (e.g., the Gini Index or measures of group differences in socioeconomic status) were relatively strong predictors of crime rates, indicating theories other than social disorganization, such as theories of relative deprivation and/or contagion (see Dishion & Tipsord, 2011; Odgers, 2015), may also help explain variation in crime rates between communities.

Beyond research on macro-level causes of community crime and delinquency rates, research on the ecology-reoffending relationship is informed by a corpus of scholarship on the cross-level effects of ecological factors on individual-level outcomes, including delinquency. In a review of 40 observational studies, Sampson et al. (2002) found that, much like crime and delinquency rates, indicators of youth wellbeing were predicted by structural factors of concentrated disadvantage and residential mobility, as well as social ties and norms (including social cohesion, social control, and surveillance), institutional resources (e.g., presence of schools, social, or health services), and routine activities (i.e., everyday spaces that present opportunities for delinquency, like transportation nodes; see Felson, 2013). Three studies included in Sampson et al.'s review focused on delinquency. Two of these studies found neighborhood disadvantage was associated with delinquency among some youth, depending on age and individual risk factors (see Seidman et al., 1998; Wikström & Loeber, 2000). The third study found neighborhood conditions were not associated with delinquency, but that more “proximal” social factors (e.g., exposure to delinquent peers) were highly predictive (see Lanctot & Smith, 2001).

In addition to these observational studies, several papers from the Moving to Opportunities experiment (MtO; see Sanbonmatsu et al., 2011) have focused on delinquency. In MtO, low income families were randomized to one of three groups: 1) those who received a housing voucher for use in a low-poverty neighborhood, 2) those who received a housing voucher for use in any neighborhood, and 3) those who stayed in their current public housing (the control). For behavioral measures of delinquency (parent-reported or self-reported delinquent behavior), early comparisons between those who relocated to less disadvantaged communities and control group members were mixed. Some MtO analyses indicated moving to a less disadvantaged context reduced boys' behavior problems, and others found no evidence of effects on behavior problems (Katz, Kling, & Liebman, 2001; Leventhal & Brooks-Gunn, 2003). Later results indicated movement to less disadvantaged contexts increased delinquency among boys who moved as teens, and produced no statistically significant effects for girls or younger boys (Schmidt, Krohn, & Osypuk, 2018). For administrative measures of delinquency (i.e., arrests), girls in MtO who moved to less disadvantaged contexts had fewer arrests for property and violent crimes compared to control group members, while their male counterparts had reductions in arrests for violent crimes but increased arrests for property crimes (Sciandra et al., 2013; see also Kling, Ludwig, & Katz, 2005; Ludwig, Duncan, & Hirschfield, 2001a). These effects attenuated over time, along with changes in neighborhood conditions.

Together, macro-level research on crime and cross-level research on delinquency provides some support for the relationship between ecological factors and reoffending, and some insight as to what ecological factors may matter and how. Ecological factors, especially those related to disadvantage, residential instability, social capital and cohesion, the presence of unsupervised peer groups, and inequality, differentiate crime rates across communities. In some cases, these ecological factors also predict individual level delinquency. Studies on the ecology-delinquency relationship also suggest ecological effects unfold through more proximal forces, interact with age and gender (and related differences in social networks and individual experiences), and are contemporaneous (i.e., as residential contexts change, delinquency risk changes; effects are situational, and not the result of an indelible effect made during a specific developmental period; Sciandra et al., 2013). They also suggest that effects may vary depending on how delinquency is operationalized (i.e., as self-reported or observed behavior verses administratively recorded criminal justice outcome, like arrest).

Though macro-level research on crime rates and cross-level research on delinquency have offered these important insights, their application to research on reoffending also faces potential limitations. First, assuming theory and evidence from research on the ecology-crime relationship can be directly applied to the individual behavior of youth amounts to the ecological fallacy (i.e., inferring delinquent behavior among residents from their community's crime and delinquency rates). Second, research on the ecology-delinquency relationship may differentiate delinquent behavior across the general population but not within a sub-population of previously adjudicated youth. Ecological factors have heterogeneous effects. Thus, it is possible, for example, that ecological factors have stronger effects on offending or reoffending among those who possess fewer or weaker individual risk factors, while those who possess more or stronger individual risk factors may offend or reoffend regardless of residential context (see Wikström & Loeber, 2000). By focusing on youth from the general population or from low-income families, estimates of ecological effects on delinquency from the studies described above may be greater than the effects of ecological factors on reoffending for justice-involved youth, who likely possess more (or more significant) individual and relational risk factors by virtue of their prior system involvement. In the most basic sense, questions of whether or how ecological factors predict crime rates or delinquency in the general population, are distinct from questions of whether or how ecological factors predict reoffending among justice involved youth. In light of these distinctions, understanding and strengthening research on the ecology-reoffending relationship requires an independent synthesis of evidence.

Beyond a dedicated focus, distilling research on ecological risk factors for reoffending requires attention to the methodological specificities of both risk factor research and research that tests relationships across units of analysis (i.e., from macro factor to individual behavior). The traditional risk factor framework, proposed by Kraemer et al. (1997) and adapted to offending by Murray, Farrington, and Eisner (2009) (see also Monahan & Skeem, 2013), defines risk factors as correlates that temporally precede outcomes and defines causal risk factors as factors that vary and alter outcomes as a result of their variation. While experiments are well-suited to establishing causality, well-designed observational studies may closely approximate their findings. Observational studies that approximate experimental studies tend to (a) be well-controlled (i.e., measure potential confounders prior to risk factors, and use comparison groups or statistical procedures to account for confounders, e.g., propensity scoring, multivariate regression), and (b) measure within-individual change in risk factors. The ability of such observational studies to identify risk factors and likely causal relationships between risk factors and reoffending is important, given experiments are often difficult or impossible to conduct where risk factors for reoffending are concerned.

Identifying ecological risk factors for reoffending brings additional challenges. Scholars of ecology face an array of methodological complications, which have been discussed in depth elsewhere (see Diez Roux & Mair, 2010). These challenges include the modifiable areal unit problem (MAUP) and uncertain geographic context problems (UGCP), which speak to the introduction of error dependent on how geographic units are defined and differential exposure of individuals to those units (Fotheringham & Wong, 1991; Kwan, 2012). Scholars of ecology also face challenges in establishing causal risk factors (Diez Roux & Mair, 2010). By virtue of their very scale, exposure to macro-level factors is not easily manipulated and, as a result, experiments are seldom employed. In the absence of such techniques, ecological research is challenged by the potential for social selection, whereby the factors that lead people to live in certain contexts may confound the association between the qualities of those contexts and outcomes (Cheshire, 2012). As previously discussed, one experimental study (MtO) has examined the relationship between moving to less disadvantaged contexts and delinquency. We know of no experimental study that has examined residential movement's effect on reoffending among youth.

Still, there is good reason to take stock of what research does exist. First, as previously discussed, well-designed observational research can speak to causal risk factors for reoffending and, in turn, inform interventions to curb reoffending. Second, identification of risk factors for reoffending, even if not causal, has relevance to risk assessment and the identification of youth likely to reoffend. Studies that identify ecological risk factors can also provide insight into the complex relationships between contexts and outcomes; they can identify which elements of environments are associated with outcomes, assess specific patterns of interaction between individuals and their environments, and test group differences across these associations and patterns (Sharkey & Faber, 2014). As noted by Sharkey and Faber (2014), researchers should avoid reducing such complexity to questions of whether or not residential context “matters,” and instead pursue research that can answer “when, where, why, and for whom…residential contexts matter” (p. 562). The current study heeds this advice, while responding to calls for systematic reviews that speak to risk factors and possible causes of reoffending.

The aim of this paper is to synthesize and assess research on the ecology-reoffending relationship. Specifically, we systematically review studies to answer two research questions: (1) What, if any, ecological factors predict reoffending among youth? And, (2) where research supports the relationship between ecological factors and reoffending, what is the nature of that relationship (i.e., how and for whom do ecological factors matter for reoffending)? In the remainder of this paper, we synthesize studies to identify likely socio-structural, economic, and justice system-related ecological risk factors for reoffending. We then assess what these studies say about where, when, why and for whom contexts matter for reoffending, and whether current research provides correlational or causal support. We conclude by discussing lessons learned, emphasizing implications for future research that seeks to strengthen risk prediction and inform interventions to reduce reoffending and related collateral consequences.