تأثیر دوستان در خصوص سرقت، نوجوانان و بزهکاری جزئی: آزمون اثرات رشد به گزارش همسالان

Abstract Most theories about adolescent involvement in delinquent behaviors afford a prominent role to peers or friends and their behaviors. In this article I explore the age-graded role that actual friends’ behavior plays in explaining respondents’ theft and minor delinquency during middle and later adolescence, paying special attention to the potentially mitigating effects of social control measures. Results suggest that the influence of friends’ own reported theft and minor delinquency in explaining respondents’ behavior is relatively modest despite employing several different measures of friends’ behaviors. Family influences proved weaker than expected during middle adolescence. The influence of friends’ behavior was only modestly influential, and slightly more noticeable during later adolescence. The sheer amount of time spent with friends, on the other hand, was influential throughout adolescence. The results reinforce skepticism concerning the accuracy of self-reports about peer behavior, and suggest that perceptions of peers’ behavior may be more influential than the behaviors themselves.

. Introduction It is well documented that peers and friends are an important influence on adolescent behavior. Peer attitudes and behaviors are often considered the most important external predictors of adolescent delinquency (Akers et al., 1979; Aseltine, 1995; Elliott et al., 1985; Matsueda, 1982). However, most studies to date have involved surveys in which the respondent self-reports their own and their friends’ delinquent activity. Correlations between behaviors of one’s own and one’s peer group are often high (Kandel, 1996; Sampson and Laub, 1993; Warr, 1993a). Too high for several researchers (e.g., Gottfredson and Hirschi, 1990; Kandel, 1996), who suggest that method overlap in asking an adolescent about their own behavior and then about the number of their friends who act similarly invites a likely imputation of one’s own behavior to others. This problem would artificially inflate regression estimates, causing researchers to credit peers with too much influence on delinquency (Kandel, 1996; Zhang and Messner, 2000). Additionally, peers and friends are conceptually distinct: friends enjoy proximity and interaction with each other, whereas peers need not have a personal relationship at all (Moffitt, 1993). With a few exceptions, most research about peer influence appears to actually focus on friends, persons with whom an adolescent is thought to share an amicable, perhaps reciprocated, relationship. Survey questions typically ask a respondent about the behaviors in which they think their friends are involved. An adolescents’ peers, on the other hand, may not be considered their friends, yet may still influence their behavior (Moffitt, 1993). This study examines actual friends’ reports of participation in delinquency and evaluates this influence on respondents’ self-reports of two types of delinquency over the course of adolescence, using data from a nationally representative, longitudinal survey of adolescents. The research questions this paper addresses are these: First, Are actual friendship group traits and behaviors influential in predicting delinquency, as previous research on their perceived counterparts has suggested? Second, do institutional (e.g., family, school) and internalized (e.g., self-image, propensity to offend) social control mechanisms mitigate the relationship between friend-reported and self-reported delinquency? In answering these questions, this analysis also explores how different types of friendship group influences may shape adolescent delinquency. Since evidence from social and developmental psychologists strongly suggests that peer and family effects vary across the life stage of adolescence, I focus particularly on the possible age-graded nature of friends’ influence ( Brown et al., 1986; Kandel, 1996).

5. Results Table 1 displays descriptive statistics for all variables, split by age groups (14–16 and 17–19 at second wave). Several numbers stand out. First, middle adolescents self-report occurrences of theft slightly more than do older adolescents, perhaps indicating the age-gradedness of this form of delinquency. Nevertheless, the inter-group differences on the theft index are not statistically significant. There is little difference in the propensity to offend (for both dependent variables) between groups. The delinquent quality of their friendship networks (by all three measures) is not remarkably different, either. Older adolescents, however, appear to spend more time with their friends, likely a reflection of the greater level of autonomy they enjoy. Two psychological variables-positive self-image and family satisfaction—are modestly lower among older adolescents. Table 1. Descriptive statistics for respondent and peers’ traits and behaviors Variables Range Overall Means Middle adolescents Older adolescents Behaviors or characteristics of the respondent N=1648 N=567 N=1081 Theft index (number of events in past year) 0–12 .72 .85 .65 Low propensity to steal (no offenses at first interview) 0,1 .66 .69 .65 High propensity to steal (more than one offense at first interview) 0,1 .24 .23 .25 Minor delinquency offense index (number of events in past year) 0–12 1.12 1.22 1.06 Low propensity for minor offenses (no offenses at first interview) 0,1 .40 .42 .40 High propensity for minor offenses (>1 offense at first interview) 0,1 .35 .33 .36 Intact/broken biological family (0=broken, 1=intact) 0,1 .61 .60 .61 Male 0 .50 .45 .53 Family economic advantage 0–5 3.07 3.30* 2.96 African–American 0,1 .14 .16 .13 Age at wave 1 13–18 15.89 14.42* 16.65 Family satisfaction 3–15 10.99 11.36* 1.79 Self-image 10–30 24.24 24.71* 24.00 Level of autonomy 0–6 4.33 4.09* 4.46 School trouble 0–16 4.05 4.09 4.04 Behaviors or characteristics of respondent’s peers Number of friends who steal 0–7 1.34 1.41 1.30 Mean theft level among friendship network 1–4 1.27 1.24 1.29 Proportion of friends nominated that steal 0–1 .30 .26 .32 Number of friends who are minor delinquents 0–7 2.44 2.60 2.35 Mean minor delinquency level among friendship network 1–4 1.39 1.36 1.41 Proportion of friends nominated that are minor delinquents 0–1 .52 .50 .53 Average time spent with friends 1–6 3.30 2.96* 3.47 * Significant difference between groups at .05 level. Table options Table 2 displays unstandardized coefficients from the negative binomial regression of self-reported theft and minor delinquency on the friends’ delinquency measures only. The delinquency of one’s friends appears to be only slightly less influential during middle adolescence. The coefficient for the number of delinquent friends on theft displays a significant difference (at the .10 level) between age groups, influencing older adolescents to a greater degree than their younger counterparts. Nevertheless, the overall effect of friends’ theft on respondent theft is not remarkable, given the low proportion of variance explained. 2 Indeed, the independent influence of these measures of delinquency among one’s network of friends is surprisingly minimal. In correlation coefficients between the delinquency, friends’ delinquency, and offender class measures both for the total sample and the age-graded samples (not shown), the correlations between friends’ and respondents’ behavior appear slightly stronger among older adolescents. Yet among neither age class of adolescents does any correlation between friends’ theft/minor delinquency and that of the respondents exceed .17. This consistent finding reinforces Kandel’s (1996) suggestion that studies which rely on self-reported peer delinquency typically display correlations 2–3 times the size of those from studies employing peer reports. 3 Table 2. Negative binomial regression estimates of delinquency (theft index and index of minor delinquency) on different peer delinquency (theft or minor) measures—split by age categories (no controls, coefficients from six different models) Effect Theft Minor delinquency Middle Older Middle Older Peer delinquency measure Number of “delinquent” friends .092** .229** .048*** .042*** (Standard error) (.03) (.07) (.01) (.01) Pseudo R-square .001 .004 .001 .001 OLS R-square .004 .010 .004 .002 Proportion of “delinquent” friends .707** 1.274*** .479*** .474*** (.23) (.35) (.15) (.06) .003 .008 .003 .003 .009 .019 .009 .008 Mean “delinquency” of friends .592* 1.101*** .653*** .623*** (.27) (.16) (.13) (.04) .002 .008 .005 .009 .007 .026 .014 .029 N 567 1081 567 1081 Unstandardized coefficients appear above their standard errors (in parentheses). Below these appear the pseudo R-square for these models and an R-square obtained from parallel OLS estimates (model results not shown). Estimates appearing in boldface indicate a significant difference (at .10 level) between age groups. ** p<.01 (Two-tailed tests). *** p<.001 (Two-tailed tests). * p<.05 (Two-tailed tests). Table options The first two models of Table 3 display unstandardized coefficients from the negative binomial regression of self-reported theft on control variables only (the respondent’s own characteristics and behaviors). The pseudo R-square term employed here to augment the −2 log-likelihood statistic indicates a quite modest proportion of variance (about 2%) in delinquency explained by these several measures. Nevertheless, it is a substantial jump from those displayed in Table 2. Individuals in both age groups appear negatively influenced by the trouble they have with or in school. No gender influence appears during middle adolescence, while the anticipated effect of being male appears to contribute to theft among older adolescents. Indeed, the expected delinquency count for males is more than 4 times 4 that of females. One lingering theme among both age groups is the pronounced randomness of delinquency. Table 3. Negative binomial regression estimates of delinquency (theft index) on respondent and peer group characteristics and behaviors Effect Middle #1 Older #1 Middle #2 Older #2 Middle #3 Older #3 Peer delinquency measures a 1. Number of delinquent friends .132* .240*** .083+ .161** (.05) (.06) (.05) (.05) 2. Proportion of friends that are delinquent .685 ** 1.170 *** .192 1.226 *** (.25) (.24) (.35) (.21) 3. Mean delinquency of friends .713 * .927 *** .524 .859 *** (.31) (.11) (.42) (.07) Offender class Low propensity to offend −1.224* −1.156*** (.52) (.32) High propensity to offend .482 .605*** (.35) (.16) Respondent measures Average time spent with friends .355*** .088** .351** .099* .219* .091+ (.11) (.03) (.11) (.05) (.10) (.05) Male .257 .426*** .303 .427** .500 .299* (.38) (.13) (.39) (.13) (.44) (.14) Intact biological family .038 −.005 −.008 .006 .144 −.146 (.19) (.25) (.22) (.22) (.26) (.20) African–American .120 .117 .168 .204 .219 .148* (.18) (.12) (.18) (.14) (.16) (.06) Family economic advantage −.027 −.002 −.038 .014 −.088 .038 (.06) (.03) (.06) (.03) (.08) (.06) Family satisfaction −.050 −.084** −.038 −.094*** −.036 −.027 (.06) (.03) (.07) (.03) (.06) (.03) School trouble .123*** .123*** .129*** .116*** .092* .085*** (.05) (.01) (.03) (.01) (.04) (.01) Self-image .015 −.034** .012 −.040** .045* −.023 (.02) (.01) (.03) (.01) (.02) (.02) Level of autonomy −.010 .041 −.029 .031 −.025 .071 (.05) (.06) (.05) (.05) (.04) (.06) Model fit statistics −2 Log-likelihood 1241.36 2026.70 1238.66 2015.56 1182.04 1916.38 Pseudo R-square .018 .021 .020 .026 .065 .074 OLS R-square (model results not shown) .045 .053 .047 .061 .175 .165 N 567 1081 567 1081 567 1081 Standard errors appear below the coefficient, in parentheses. Coefficients appearing in boldface indicate a significant difference between age groups. a The italicized peer delinquency coefficients were taken from separate regression models, to avoid collinearity effects. All other coefficients appearing here come from the model including “Number of Delinquent Friends” as the peer delinquency measure. They are very similar to the regression coefficients from models that substitute the other two peer delinquency measures. * p<.05 (Two-tailed tests). *** p<.001 (Two-tailed tests). + p<.10 (Two-tailed tests). ** p<.01 (Two-tailed tests). Table options Among both groups, simply spending time with friends contributed to greater delinquency. However, it is significantly less influential among older adolescents, despite the fact that older respondents spend more time with friends. Likewise, the influence of a positive self-image appears to shield older adolescents from delinquency, but has no effect among younger adolescents. A minor family satisfaction influence also appears among older adolescents. For most of these coefficients, the small sample sizes or large standard errors prevented between-group significance differences in the coefficients. The second set of models in Table 3 shows the influence of adding the friends’ delinquency measure(s). As in Table 2, their contribution is modest. Their magnitude and significance are essentially comparable across groups, though slightly larger among older adolescents. Even when controlling for these friends’ delinquency measures, the time spent with friends remains a predictor of delinquency (and more so for middle adolescents). Basically, the addition of the friends’ delinquency measures changes very little in the models, suggesting minimal indirect influence of respondent traits through the delinquency of their friends. The third set of models adds the two propensity-to-offend measures, which clearly account for the largest share of variance explained. Their addition triples its magnitude. Yet propensity-to-offend is only mildly significant during mid-adolescence. Those mid-adolescent respondents who had not offended at wave one reported on average about 70% fewer delinquent events at wave two than those who had one offense. Older adolescents with a high propensity to offend report almost twice as many instances of theft as do those who were considered of average propensity to offend. Additionally, accounting for an adolescent’s propensity-to-offend reduces the influence of spending time with one’s friends, as well as the influence of friends’ delinquency (by all measures) among middle-adolescents. The friends’ delinquency measures remain significant among older adolescents. Table 4 displays coefficients from models identical to those in Table 3, but using a minor delinquency index as the dependent variable. I will focus on key differences between Table 3 and Table 4. Compared with theft, there are more between-group differences in effects on minor delinquency, and this form of delinquency shows a bit less randomness than theft. The effect of being male and having trouble in school is more aggravating among older adolescents, and the effect of being African-American more protective during mid-adolescence. Beyond these, many of the relationships are comparable across delinquency outcomes. Time spent with friends is positively related to minor delinquency among both age groups. The proportion of friends that are minor delinquents appears less influential than its counterpart on theft. The mean delinquency of one’s friends is strong in magnitude across models, though less so in real significance. As with theft, the propensity to offend (or be a minor delinquent at Wave I) remains a key predictor. When accounting for these, friends’ delinquency appears again to influence older adolescents slightly more than middle adolescents. Table 4. Negative binomial regression estimates of minor delinquency index on respondent and peer group characteristics and behaviors Effect Middle #1 Older #1 Middle #2 Older #2 Middle #3 Older #3 Peer delinquency measures a 1. Number of delinquent friends .039* .034*** .019 .009 (.02) (.01) (.05) (.01) 2. Proportion of friends that are delinquent .358 ** .189 + .250 + .204 ** (.14) (.11) (.15) (.07) 3. Mean delinquency of friends .631 *** .412 *** .442 * .344 *** (.18) (.10) (.18) (.08) Offender class Low propensity to offend −.822*** −1.147*** (.12) (.11) High propensity to offend .311* .429*** (.13) (.03) Respondent measures Average time spent with friends .177*** .075** .175** .075*** .098** .051+ (.06) (.03) (.06) (.02) (.04) (.03) Male .327** .590*** .335** .590*** .277* .448*** (.11) (.03) (.11) (.03) (.12) (.03) Intact biological family −.099 .074* −.108 .075* −.035 .006 (.10) (.03) (.10) (.04) (.09) (.04) African–American −.218** −.054* −.176+ −.013 −.069 −.150*** (.08) (.02) (.09) (.03) (.13) (.04) Family economic advantage .018 .002 .013 −.004 −.016 −.006 (.07) (.02) (.06) (.02) (.07) (.02) Family satisfaction −.042 −.019* −.040 −.020* −.003 .014 (.04) (.01) (.04) (.01) (.03) (.01) School trouble .089*** .137*** .089*** .136*** .071** .103*** (.02) (.01) (.02) (.01) (.02) (.01) Self-image .019 −.007 .018 −.009 .027* .006 (.01) (.02) (.01) (.02) (.01) (.01) Level of autonomy .022 −.005 .014 −.006 .015 −.018 (.04) (.01) (.04) (.01) (.03) (.01) Model fit statistics −2 Log-likelihood 1687.56 2915.72 1685.99 2914.10 163.03 277.43 Pseudo R-square .027 .040 .028 .041 .060 .088 OLS R-square (model results not shown) .078 .113 .079 .114 .148 .195 N 567 1081 567 1081 567 1081 Standard errors appear below the coefficient, in parentheses. Coefficients appearing in boldface indicate a significant difference between age groups. a The italicized peer delinquency coefficients were taken from separate regression models, to avoid collinearity effects. All other coefficients appearing here come from the model including “Number of Delinquent Friends” as the peer delinquency measure. They are very similar to the regression coefficients from models that substitute the other two peer delinquency measures. * p<.05 (Two-tailed tests). *** p<.001 (Two-tailed tests). ** p<.01 (Two-tailed tests). + p<.10 (Two-tailed tests).