جنسیت عنوان یک تعدیل کننده و فشار همسالان ادراک شده به عنوان یک میانجی رابطه برون سازی-بزهکاری: آزمون تئوری مسیر های جنسیتی

Abstract Purpose The current study sought to determine whether sex moderated peer mediation of the externalizing-delinquency relationship as part of a larger test of the gendered pathways theory of crime. Methods Data gathered from 4,144 (2,079 males and 2,065 females) members of the National Longitudinal Survey of Youth-Child sample were subjected to simple correlational and moderated mediation analysis. Results Externalizing behavior and delinquency correlated equally in boys and girls but in testing a full moderated mediation model it was discovered that sex moderated the mediating effect of perceived peer pressure on the externalizing–delinquency relationship. Whereas externalizing behavior predicted delinquency in both boys and girls, perceived peer pressure only mediated the externalizing-delinquency relationship in boys. Conclusions These results support the gendered pathways to delinquency model to the extent that the relationship between childhood externalizing behavior and delinquency was mediated by perceived peer pressure in males but not females. The implications of these results for theoretical refinement of the gendered pathways approach and crime prevention and intervention are discussed.

Introduction Much of what we know about crime is based on male data. Although male offenders have traditionally outnumbered female offenders by a ratio of 9 or 10 to 1, the gap has narrowed in recent years (Lauritsen, Heimer, & Lynch, 2009). Unfortunately, our understanding of female offending has lagged behind their growing presence in the criminal justice system. As such, it is imperative that we evaluate the generalizability of research findings obtained on male juvenile and adult offenders to female juvenile and adult offenders. Efforts to clarify the well-documented difference in crime prevalence rates for men and women have contributed to the development of several different theoretical models, to include the gendered pathways theory of crime initiation, maintenance, and desistance. The pathways model attempts to explain the gender gap in offending by identifying sex-based similarities and differences in offending and then integrating these points of overlap and contrast into a coherent theoretical perspective (Belknap, 2007 and Chesney-Lind and Palko, 2004). Whereas more female than male offenders present with a history of physical and sexual abuse (Harlow, 1999) and more male than female offenders commit stranger violence (ABS, 2013), male and female offenders are similarly responsive to age of crime onset (Moffitt et al., 2008 and Urban and Walters, 2014) and developmental context (Odgers et al., 2008). One variable that has frequently been used to explain the gender gap in crime is association with deviant peers. Research has fairly convincingly demonstrated that delinquent peer associations mediate the relationship between sex and future offending in that males are more likely to be negatively affected by these associations than females (Augustyn and McGloin, 2013, Fagan et al., 2007, Jensen, 2003 and Piquero et al., 2005). According to the results of a study by Mears, Ploeger, and Warr (1998) this is probably the result of both differential vulnerability and differential exposure in the sense that boys tend to have more delinquent peer associations than girls (differential exposure) and are also more likely to be negatively affected by these associations (differential vulnerability). Whereas the majority of studies on gender and crime have employed sex as the independent variable in a simple mediation analysis, it may be more productive to conceptualize sex as a moderator variable. Negriff, Ji, and Trickett (2011) discovered that exposure to delinquent peers fully mediated the relationship between earlier pubertal maturation and later delinquent behavior and that sex had no moderating effect on this relationship. Weerman and Hoeve (2012) also found that sex had little bearing on the ability of peers to influence subsequent delinquency. Early externalizing behavior may form a direct link to subsequent delinquent and criminal behavior or it may follow an indirect path through deviant peer associations. Early adolescents in one study who displayed externalizing behavior were at increased risk for associating with deviant peers in mid-adolescence and engaging in crime and violence in late adolescence and early adulthood (Brook, Brook, Rubenstone, Zhang, & Saar, 2011). Research denotes that while boys typically report higher levels of externalizing behavior than girls, early externalizing behavior is an effective predictor of subsequent delinquency in both sexes (Green et al., 2008, Leadbeater et al., 1999 and Simonoff et al., 2004). As was noted earlier, the gendered pathways model of crime holds that there are both similarities and differences in the mechanisms leading to male and female criminality. It would appear that one of these points of similarity may be continuity in antisocial behavior over the life-course in the form of a strong connection between early externalizing behavior and subsequent delinquency and crime. Peers have been found to mediate the relationship between early risk factors and subsequent delinquency. However, this effect may be stronger in males than in females. In fact, the results of several studies indicate that while peers have a significant impact on delinquency in boys and men, they have no appreciable effect on delinquency in girls and women (Jensen, 2003 and Piquero et al., 2005). Moreover, parental monitoring may moderate or mediate the peer-delinquency relationship. Tilton-Weaver, Burk, Kerr, and Stattin (2013), for instance, determined that parental monitoring of peer relationships formed a complex relationship with future delinquency in offspring. Under some circumstances parental monitoring inhibited the delinquency-promoting effects of deviant peer associations but under other circumstances parental monitoring actually accelerated the delinquency-promoting effects of deviant peer associations. In an earlier study, Bowman, Prelow, and Weaver (2007) had found that maternal monitoring of peer relationships suppressed the mediating effect of deviant peers on delinquent behavior in girls but not in boys. These results suggest that parental monitoring may explain sex moderation of peer mediated relationships. Whether parental monitoring inhibits or promotes deviant peer associations, findings from both the Tilton-Weaver et al. (2013) and Bowman et al. (2007) studies indicate that parental monitoring of peer relationships impacts on peer mediation of the externalizing-delinquency relationship and should probably be controlled in any studies on this relationship. There were two distinct but interconnected hypotheses tested in this study: 1. Childhood externalizing behavior is just as effective in predicting delinquency in females as it is in predicting delinquency in males. 2. Sex moderates the externalizing → perceived peer pressure → delinquency relationship such that a significant mediation effect should occur in males but not in females, after controlling for basic demographic variables and parental monitoring of time spent with friends.

Results Descriptive statistics and simple correlations Descriptive statistics for the nine variables included in this study, along with their correlations, can be found in Table 1. Age, race (white), parental limits, BPI external, perceived peer pressure, and delinquency comparisons between male and female participants revealed that female participants were slightly older than male participants and that boys achieved significantly higher BPI external, perceived peer pressure, and delinquency scores than girls (see Table 2). With respect to the first hypothesis, externalizing behavior and delinquency correlated .12 (p < .001) in males and .14 (p < .001) in females. Collinearity between the eight predictor variables included in this study (age, race, parental limits, sex, BPI external, perceived peer pressure, and the two interaction terms) was tested and found to be within tolerable limits (tolerance = .884–.998; variance inflation factor [VIF] = 1.002–1.132). Table 1. Descriptive Statistics and Correlations for the Variables Included in this Study Variable N/n M SD Range Race Parent Sex External Peer SxE SxP Delinq Age 4144 9.50 0.50 9–10 .01 .03 .04⁎ -.00 -.01 .02 .00 .01 White 4144 0.49 0.50 0–1 .01 .01 -.07⁎⁎ -.05⁎ .01 .01 -.04⁎ Parental Limits 3569 2.05 1.02 0–3 .02 -.04⁎ .02 .02 -.01 .02 Sex 4144 1.50 0.50 1–2 -.15⁎⁎ -.09⁎⁎ -.02 -.01 -.10⁎⁎ BPI-External 4144 7.49 6.27 0–40 .11⁎⁎ -.06⁎⁎ -.19⁎⁎ .14⁎⁎ Peer Pressure 3399 0.07 0.26 0–1 -.29⁎⁎ -.06⁎⁎ .12⁎⁎ Sex x External 4144 − 0.15 0.98 − 5.12–4.72 .10⁎⁎ -.08⁎⁎ Sex x Peer 3399 − 0.09 0.97 − 3.49–3.55 -.02 Delinquency 3839 0.08 0.28 0–1 Note. Age = age at time the Behavior Problem Index was administered; White represents race and is coded as white (1) versus nonwhite (0); Parental Limits = how often parents limit time with friends on school nights; Sex is coded as male (1) versus female(2); BPI-External = Behavior Problems Index externalizing score; Peer Pressure = feels pressure from friends to engage in crime and is coded as yes (1) versus no (0); Sex x External = interaction between sex and the BPI externalizing score; Sex x Peer = interaction between sex and peer pressure; Delinquency = convicted, placed on probation, or incarcerated in jail or prison in the last 2 years and is coded as yes (1) versus no (0); N/n = number of participants with complete data on that particular variable; M = mean, SD = standard deviation, Range = range of scores in current sample. ⁎ p < .05. ⁎⁎ p < .001. Table options Table 2. Male–female Differences on the Variables Included in this Study Males Females Comparison Variable n M SD n M SD t p Age 2079 9.48 0.50 2065 9.51 0.50 − 2.36 .018 Parental Limits 1787 2.03 1.04 1782 2.08 1.00 − 1.50 .134 BPI External 2079 8.42 6.79 2065 6.55 5.54 9.71 .000 N Yes No n Yes No χ2 p White 2079 48.8% 51.2% 2065 49.6% 50.4% − 0.23 .622 Peer Pressure 1693 9.4% 90.6% 1706 4.8% 95.2% 27.12 .000 Delinquency 1919 11.0% 89.0% 1920 5.6% 94.4% 37.14 .000 Note. Age = age at time the Behavior Problem Index was administered; Parental Limits = how often parents limit time with friends on school nights; BPI-External = Behavior Problems Index externalizing score; White represents race and is coded as white (1) versus nonwhite (0); Peer Pressure = feels pressure from friends to engage in crime and is coded as yes (1) versus no (0); Delinquency = convicted, placed on probation, or incarcerated in jail or prison in the last 2 years and is coded as yes (1) versus no (0); n = number of participants with complete data for that particular variable; M = mean, SD = standard deviation, Yes = condition present; No = condition absent; t = t-test with 4142 degrees of freedom (3567 for parental limits); χ2 = chi square statistic with 1 degree of freedom; p = significance of the t-test or chi square results. Table options Moderated mediation analysis The second hypothesis was tested with moderated mediation by merging Preacher et al.’s (2007) Models 2 and 3 with Edwards and Lambert’s direct, first stage, and second stage models to create a single regression equation. As indicated by the results outlined in Table 3, sex moderated the direct effect of externalizing behavior on delinquency (delinquency on sex x external) and the indirect effect of perceived peer pressure on delinquency by means of the second stage or b pathway (delinquency on sex x peer). Sensitivity testing revealed that an unobserved confounding covariate would need to correlate .29 with both the mediator (peer) and outcome (delinquency) variables to reduce the mediating effect in the total sample to zero. Table 3. Moderated Mediation of the Externalizing-Delinquency Relationship by Sex and Peer Pressure Predictor b(95% CI) β t p Peer Pressure on Age − 0.018(− 0.155, 0.114) − 0.009 − 0.26 .798 White − 0.203(− 0.341, -0.067) − 0.100 − 2.92 .004 Parental Limits 0.060(− 0.005, 0.127) 0.059 1.75 .081 Sex − 0.105(− 0.169, -0.034) − 0.051 − 3.06 .002 BPI External 0.023(0.012, 0.034) 0.138 4.15 .000 Sex x External − 0.020(− 0.093, 0.047) − 0.019 − 0.57 .571 Delinquency on Peer Pressure 0.182(0.051, 0.309) 0.176 2.74 .006 Age 0.076(− 0.057, 0.211) 0.036 1.11 .265 White − 0.146(− 0.286, -0.010) − 0.069 − 2.05 .040 Parental Limits 0.037(− 0.028, 0.103) 0.035 1.10 .271 Sex − 0.312(− 0.455, -0.165) − 0.148 − 4.25 .000 BPI External 0.034(0.022, 0.046) 0.198 5.84 .000 Sex x External 0.097(0.027, 0.167) 0.088 2.69 .007 Sex x Peer − 0.096(− 0.167, -0.023) − 0.089 − 2.64 .008 Total Effect 0.039(0.027, 0.050) 0.222 6.70 .000 Direct Effect 0.034(0.022, 0.046) 0.198 5.84 .000 Indirect Effect 0.004(0.001, 0.009) 0.024 2.24 .025 Note. Peer Pressure on = regression equation with peer pressure as the outcome; Delinquency on = regression equation with delinquency as the outcome; Age = age at time the Behavior Problem Index was administered; White represents race and is coded as white (1) versus nonwhite (0); Parental Limits = how often parents limit time with friends on school nights; Sex is coded as male (1) versus female (2); BPI-External = Behavior Problems Index externalizing score; Peer Pressure = feels pressure from friends to engage in crime and is coded as yes (1) versus no (0); Sex x External = interaction between sex and the BPI externalizing score; Sex x Peer = interaction between sex and peer pressure; Total Effect = sum of the direct and indirect effects; Direct Effect = direct effect of the BPI externalizing score on delinquency; Indirect Effect = indirect effect of the BPI externalizing score on delinquency as mediated by peer pressure; b(95% CI) = unstandardized coefficient and the lower and upper limits of the 95% confidence interval for the unstandardized coefficient (in parentheses); β = standardized coefficient; t = asymptotic t-test; p = significance level of the asymptotic t-test; N = 3328. Table options The results of the moderated mediation analysis did not change when the parental limits control variable was treated as a mediator variable or was left out of the analysis entirely. When the parental limits variable was included in the regression equation as a mediator, a significant indirect effect (b = .004[.001–.009], β = .022, t = 2.06, p < .05) capable of explaining 10% of the total effect was uncovered. When the parental limits variable was left out of the equation, a significant indirect effect (b = .004[.001–.009], β = .024, t = 2.21, p < .05) surfaced that accounted for 11% of the total effect. Simple slope analysis of the conditional indirect effect of externalizing behavior on delinquency revealed that while perceived peer pressure successfully mediated the externalizing-delinquency relationship in males (see Table 4), it failed to mediate the externalizing-delinquency relationship in females (see Table 5). In the male analysis, perceived peer pressure was found to account for 31.5% of the variance in the total effect, but because the direct effect remained significant, the mediating effect is classified as partial. A sensitivity analysis of this mediating effect determined that a confounding covariate would need to correlate .36 with both perceived peer pressure and delinquency to completely eliminate the mediating effect observed in the male subsample. Table 4. Simple Slopes for the Peer-Mediated Externalizing-Delinquency Relationship in Males Predictor b(95% CI) β t p Peer Pressure on Age − 0.078(− 0.246, 0.102) − 0.038 − 0.88 .379 White − 0.174(− 0.354, -0.001) − 0.085 − 1.97 .049 Parental Limits 0.057(− 0.024, 0.141) 0.058 1.37 .171 BPI External 0.029(0.017, 0.041) 0.190 4.70 .000 Delinquency on Peer Pressure 0.282(0.149, 0.401) 0.282 4.39 .000 Age 0.086(− 0.078, 0.255) 0.042 1.00 .317 White − 0.225(− 0.395, -0.055) − 0.110 − 2.61 .009 Parental Limits 0.018(− 0.057, 0.096) 0.018 0.47 .640 BPI External 0.018(0.005, 0.030) 0.115 2.80 .005 Total Effect 0.026(0.014, 0.037) 0.168 4.26 .000 Direct Effect 0.018(0.005, 0.030) 0.115 2.80 .005 Indirect Effect 0.008(0.004, 0.014) 0.053 3.12 .002 Note. Peer Pressure on = regression equation with peer pressure as the outcome; Delinquency on = regression equation with delinquency as the outcome; Age = age at time the Behavior Problem Index was administered; White represents race and is coded as white (1) versus nonwhite (0); Parental Limits = how often parents limit time with friends on school nights; BPI-External = Behavior Problems Index externalizing score; Peer Pressure = feels pressure from friends to engage in crime and is coded as yes (1) versus no (0); b(95% CI) = unstandardized coefficient and the lower and upper limits of the 95% confidence interval for the unstandardized coefficient (in parentheses); β = standardized coefficient; t = asymptotic t-test; p = significance level of the asymptotic t-test; n = 1787. Table options Table 5. Simple Slopes for the Peer-Mediated Externalizing-Delinquency Relationship in Females Predictor b(95% CI) β t p Peer Pressure on Age 0.035(− 0.170, 0.254) 0.017 0.32 .747 White − 0.231(− 0.445, -0.027) − 0.114 − 2.15 .031 Parental Limits 0.062(− 0.048, 0.174) 0.061 1.09 .277 BPI External 0.018(0.001, 0.035) 0.100 2.08 .038 Delinquency on Peer Pressure 0.145(− 0.113, 0.331) 0.140 1.28 .200 Age 0.061(− 0.146, 0.285) 0.029 0.57 .571 White 0.071(− 0.148, 0.290) 0.034 0.64 .525 Parental Limits 0.051(− 0.052, 0.158) 0.049 0.95 .343 BPI External 0.052(0.034, 0.070) 0.274 5.76 .000 Total Effect 0.055(0.038, 0.072) 0.288 6.31 .000 Direct Effect 0.052(0.034, 0.070) 0.274 5.79 .000 Indirect Effect 0.003(− 0.001, 0.010) 0.014 1.01 .312 Note. Peer Pressure on = regression equation with peer pressure as the outcome; Delinquency on = regression equation with delinquency as the outcome; Age = age at time the Behavior Problem Index was administered; White represents race and is coded as white (1) versus nonwhite (0); Parental Limits = how often parents limit time with friends on school nights; BPI-External = Behavior Problems Index externalizing score; Peer Pressure = feels pressure from friends to engage in crime and is coded as yes (1) versus no (0); b(95% CI) = unstandardized coefficient and the lower and upper limits of the 95% confidence interval for the unstandardized coefficient (in parentheses); β = standardized coefficient; t = asymptotic t-test; p = significance level of the asymptotic t-test; n = 1782.