رسیدگی به رابطه تجربی بین زمان با همسالان، دوستی و بزهکاری

Abstract Purpose Much of the research on peer influence has examined the relationship between peer associations and delinquency. Relatively little empirical research has addressed the effects of delinquent behavior on peer intimacy and time spent with peers. Our research attempts to fill these gaps in the literature as we hypothesize that, net of peer delinquency, delinquents spend more time with their peers but are less closely attached to their peers. Methods Using data from two waves of the National Youth Survey (NYS), we present two sets of regression models to account for selection bias resulting from whether respondents reported having friends. To assess the stability of our findings, we supplement our presented findings with extensive use of alternate estimation strategies. Results Conclusions regarding our hypotheses do not vary by estimation strategy. Delinquents spend more time with their peers, but delinquents and non-delinquents do not report differences in closeness to their peers. Conclusions Given our control variables, our finding introduces complexity in the causal priority between time spent with peers and delinquency. Prior delinquency may be a predictor of more time with peers, but partly as an avenue for opportunities for crime, not for the sake of friendship.

Introduction Scholars have repeatedly demonstrated that delinquency is an inherently social activity involving associates (e.g., Breckenridge and Abbott, 1912 and Thrasher, 1927; see also Felson, 2003, Reiss, 1986 and Warr, 2002). In fact, early research showed that over 90% of juvenile court cases involves two or more youths (Shaw & Myers, 1929, p. 662). The social nature of delinquency and the well-established finding that peer delinquency affects delinquency has led researchers to study different dimensions of the peer bond. Two particularly consequential dimensions of the peer bond are time spent with peers and peer intimacy, or friendship quality. Research has found that more time spent with peers, regardless of the behavior of the peers, has a positive effect on delinquency (Haynie and Osgood, 2005, Meldrum et al., 2009 and Regnerus, 2002). This relationship is especially salient when the time spent with peers is in unstructured activities (e.g., Haynie and Osgood, 2005, Osgood et al., 1996 and Wallace and Bachman, 1991). Evidence has also shown that attachment to peers has a positive effect on delinquency, independent of time spent with peers and the level of peer delinquency (Agnew, 1991, p. 62). Moreover, research has demonstrated that the effect of peer delinquency on delinquency is moderated by the quality of the friendship bond and time spent with peers (Agnew, 1991, Elliott et al., 1989, Giordano et al., 1986 and Vásquez, 2010). However, relatively little research has addressed either the causes of time spent with peers and peer intimacy or the effect of delinquent behavior on these aspects of the peer bond. In other words, friendship and peer interaction measures have generally been modeled as causes rather than as effects of crime. We hypothesize that delinquency will have a positive effect on time spent with peers but a negative effect on closeness to peers. That is, since delinquency is an inherently social activity, we expect delinquents to spend more time with their peers. On the other hand, we do not expect delinquents to be more attached to their peers or to have more meaningful friendships simply because criminal behavior tends to involve associates. We discuss elements of the peer bond below and then discuss the rationale for our hypotheses. Finally, we test our hypotheses using longitudinal data from the National Youth Survey (NYS). We estimate two sets of regression models to account for selection bias originating from whether respondents reported having friends prior to answering questions about peer bonding.

Conclusion and discussion Our research contributes to the literature by investigating the possibility that two distinct dimensions of the peer bond are influenced by delinquency. Specifically, we examine whether delinquents spend more time with their peers but are less closely attached to their peers. The fact that not all respondents have friends, however, presents a methodological issue for testing our hypotheses. We consequently examine our hypotheses using two sets of regression models that differ in their ability to distinguish respondents with friends from those without friends. Our standard regression models address this selection bias by including dummy variables for friendlessness, while Heckman selection models explicitly account for the skip pattern. We give considerable attention to this issue because selection bias has meaningfully adverse effects on model estimates. In addition, using various estimation strategies is a key practice for assessing the sensitivity of results to model specification (Chay & Powell, 2001, p. 34). With regard to the control variables in our study, we see inconsistencies in the findings across measure and method. Inconsistencies generally weaken the evidence for conclusions because powerful explanatory variables tend to have statistically significant effects regardless of estimation strategy (Osgood et al., 2002, p. 320). With regard to our primary hypotheses, however, we see conclusion-level invariance across estimation strategy and measure. The evidence consistently supports our first hypothesis that delinquents spend more time with their peers. The evidence related to our second hypothesis is also consistent across measure and method, but the conclusion is that delinquency has no detectable effect on closeness to peers, contrary to our hypothesis. A limitation of this study (and related studies) concerns the measurement of our peer bond variables. Measuring time spent with peers may seem a simpler task than measuring closeness to peers, but underlying complexities exist for both. As opposed to a pure time-spent-with-peers measure, for example, there may be value in a relative measure of time spent with peers, such as the percent of free time adolescents spend with their peers. It is possible that youths with very little free time could spend virtually all of their free time with peers, but still spend fewer hours with their peers than youths who have more free time. The goal could then become measuring the quality of the time spent with peers, as opposed to simply the quantity. And measuring closeness to peers, or the intensity of the friendship bond, remains a problematic task, as discussed above. With these limitations in mind, our analyses have several implications for future research. Our finding that delinquency increases time spent with peers, controlling for peer delinquency and a lagged measure of time spent with peers, introduces complexity in the causal priority between time spent with peers and delinquency. The effect of time spent with peers on delinquency has been examined and replicated repeatedly, and time spent with peers has significant explanatory power for delinquency (e.g., Osgood and Anderson, 2004, Osgood et al., 1996 and Regnerus, 2002). Whether delinquents simply have different interactions with peers, which reverses the causal priority, represents a relatively unexplored direction in understanding adolescent interactions. Future research should continue to explore the relationship between time spent with peers and peer delinquency using longitudinal data and appropriate methods that can attempt to establish true temporal ordering. In terms of policy, the results suggest that research should consider where time with peers takes place. For example, in the presence of capable guardianship, time spent with peers, in itself, is harmless. It is the convergence of offenders in settings with a lack of capable guardianship where criminologists ought to focus attention ( Felson, 2003). This setting is where likely offenders find one another, and resources should be devoted to disrupting or neutralizing potential offender convergence settings. Moreover, policies should target co-offending, especially among juvenile delinquents. This view coincides with the longstanding observation in criminology that “In fact, there is scarcely a type of delinquent boy who is not associated with others in his wrongdoing” ( Breckenridge & Abbott, 1912, p. 35). Our analyses also indicate that delinquents and non-delinquents do not differ in how they report and perceive attachment to their peers. Although this finding is in contrast to the “cold and brittle” hypothesis offered by Hirschi (1969), literature on the role of friendship quality in peer influence research is unclear (Berndt, 2002). Some studies suggest that the process of learning delinquent behavior from peers requires close attachments to those peers, while other studies suggest that delinquents are loosely attached to delinquent peers whom they select. The stability in our findings, given our different estimation strategies and sensitivity analyses, are particularly noteworthy because there are contradictory findings both within and across studies (Berndt, 2002 and Giordano et al., 2010). We hope that our study contributes to the growing research base on the friendship quality of delinquents, and thus brings us one step closer to disentangling the complex ways in which friendship quality affects behavior. As noted in Boman et al. (2012), most research regarding friendship quality focuses on only segments of a friendship, when empirical investigations need to examine the totality of friendship quality. Our analyses also highlight the need to account for selection bias resulting from filter questions common to large-scale secondary data sources utilized by researchers across disciplines. Accounting for problems such as selection bias can represent a considerable layer of complexity for those estimating empirical models, and the process often involves learning conceptually and technically intricate statistical techniques. If, however, one begins with statistical complexity, as opposed to theory, the major challenge becomes justifying potentially contradictory findings based on misspecified models. Empirical complexity should be introduced grudgingly and only when resting solidly on theory. For example, we began with two relatively straightforward hypotheses (i.e., does delinquency influence time spent with peers and closeness to peers?), but answering these questions required dealing with selection bias resulting from friendlessness, which is of theoretical importance to our study. We used two sets of regression models to handle selection bias, and we encourage others to use sensitivity analyses and cross-model comparisons to confirm that findings are not due to model specification. Finally, our analyses are consistent with research that has begun to rely on the well-established relationship between peer delinquency and delinquency and move towards studying different dimensions of the peer bond. Our analyses indicate that the friendships of delinquents and non-delinquents vary in interesting and complex ways. That is, delinquents and non-delinquents do not differ in their feelings toward their friends, but they do differ in how much time they spend with their friends. Our analyses thus affirm Warr’s (2002) encouragement for a more thorough investigation of the dimensions of the peer bond, while underscoring the finding that time spent with peers, regardless of the behavior of the peers, has a positive effect on delinquency (Osgood et al., 1996, Regnerus, 2002 and Wallace and Bachman, 1991).