آیا سیاست های انضباطی مدرسه، اثرات اجتماعی مثبت دارد؟ بررسی اثرات شدت از سیاست های مدرسه بر رابطه بین بزهکاری شخصی و همسالان

Abstract Purpose Empirical research has yet to demonstrate that strict school disciplinary policies deter student misconduct. However, underlying the null and negative effects observed in prior research may be competing social impacts. What is missing from prior research is an acknowledgement that the deviance amplification effects of criminogenic risk factors may be partially offset by the general deterrence effects of strict school sanctions. Methods Using data from the school administrator questionnaire, the in-school interview, and the in-home interview from the National Longitudinal Study of Adolescent Health, this study employs logistic hierarchical models to investigate whether strict school sanctions condition the relationship between personal and peer smoking, drinking, and fighting. Results Results indicate that the effects of peer smoking, drinking, and fighting on corresponding respondent delinquency are attenuated in schools with strict sanction policies for these behaviors. Conclusions Results suggest that school policies can aid in preventing crime in unanticipated ways, for example, by reducing the crime-inducing effects of having delinquent peers. Prior research may therefore be unintentionally discounting the general deterrence effects of school disciplinary policies by neglecting the moderating mechanisms through which these policies operate.

Introduction In 1994, Congress adopted the Gun-Free Schools Act, mandating a one year out-of-school suspension for gun possession and the referral of students carrying weapons to the juvenile justice system. This policy, largely a response to the epidemic of youth violence in the late 1980s and the highly publicized school shootings in the 1990s (Cook et al., 2010, Dodge et al., 2006 and Hirschfield, 2008), was aimed at deterring gun possession and preventing gun-related deaths. After the passage of this federal mandate, schools across the country swiftly extended harsh disciplinary sanctions to less serious rule violations such as cigarette smoking, alcohol and drug use, and fighting (see Gorman and Pauken, 2003, Hirschfield, 2008, Maimon et al., 2012, Payne and Welch, 2010 and Simon, 2006). Behaviors considered relatively normative during adolescence (Arnett, 1992, Baumrind, 1987, Dryfoos, 1991, Elliott et al., 1989, Johnston et al., 1991 and Moffitt, 1993) are now met with punitive sanctions such as home suspension and expulsion (APA, 2008, p. 852). Unfortunately, research has not demonstrated that these punitive sanctions deter future misconduct, and there is evidence that harsh punitive policies have unanticipated negative consequences (for an overview, see Cook et al., 2010). However, prior research may obscure the particular contexts and individuals for whom school disciplinary policies are relatively effective. What is missing from the research base on harsh school sanctions is an acknowledgement that harsh school sanction policies may deter student deviant behavior indirectly by moderating the effects of salient individual risk factors on delinquency. That is, the deviance amplification effects of criminogenic risk factors may be partially offset by the general deterrence effects of strict school sanctions. For example, emerging evidence (e.g., Maimon et al., 2012, Matjasko, 2011 and Novak and Clayton, 2001) suggests that, rather than restraining individuals directly through the threat of punishment, strong deterrent cues such as harsh sanction policies may work by overriding the motivational forces that encourage individuals to engage in deviance (Wikström, 2006 and Wikström, 2010). It is such moderating mechanisms through which sanctions operate in which we are interested. Specifically, this is the first study to investigate whether strict school sanctions deter delinquent behavior by attenuating the relationship between personal and peer smoking, drinking, and fighting. We focus on peer influence because the delinquent behavior of one’s friends is one of the most consistently recognized motivational forces in the etiology of delinquency during adolescence (e.g., Akers, 1998, Elliott et al., 1985, Jaccard et al., 2005, Matsueda and Anderson, 1998, Thornberry and Krohn, 1997, Warr, 1996, Warr, 2002 and Warr and Stafford, 1991). In addition, the inter-relationships among sanctions, peers, and delinquency is “one of the most important (yet) understudied” areas of deterrence research (see Nagin & Pogarsky, 2003, p. 185). We hypothesize that harsh disciplinary policies will reduce the social influence of delinquent peers. That is, harsh school disciplinary policies (e.g., expulsion and home suspension) may have “moral or educative effects” which are critical for the development of prosocial beliefs that counteract the influence of delinquent peers (Andenaes, 1971, p. 17). This reasoning suggests that positive social processes may underlie the null and negative effects observed in prior research. Ultimately, we argue that to fully understand the impact of strict sanction policies, research must take into account the direct and moderating processes through which sanctions operate. This research is critical to the future of deterrence theory, which depends on an examination of the ways in which formal sanctions are likely to influence individual behavior (see Pratt, Cullen, Blevins, Daigle, & Madsen, 2006, p. 386). We examine the moderating mechanisms through which formal school sanctions operate using data from the National Longitudinal Study of Adolescent Health (Add Health), a dataset particularly well suited for studying adolescents and their peers in the school context.

Results Baseline hierarchical models To examine if there was school-level variation in the delinquent outcomes, we first estimated three unconditional models without level-one or level-two covariates. The analyses (not shown here) indicated that there was significant school-level variability in youth self-reported smoking (τ00 = .25, p < .001), drinking (τ00 = .39, p < .001), and fighting (τ00 = .08, p < .001), justifying our multilevel modeling strategy. In Table 1, we add the key independent variables (peer delinquency, school sanction policy) to the unconditional models. Consistent with prior research, the results indicate that peer delinquency had a positive and significant effect on all types of measured delinquency (.17, p < .001 for peer smoking, .53, p < .001 for peer drinking, .42, p < .001 for peer fighting). The results also indicate that a harsh school sanction policy had a negative and significant association with respondent smoking (− .08, p < .05), but not with drinking (− .04, p > .05) or fighting (− .05, p > .05). This finding is consistent with prior research failing to find consistent negative effects of school sanction policies on antisocial behavior (see Maimon et al., 2012). Table 1. Hierarchical logistic regression models of Time 2 outcomes regressed on peer delinquency, lagged delinquency, and sanction policy Smoking a Drinking a Fighting a Variable β (SE) OR β (SE) OR β (SE) OR Peer delinquency b .17 (.05)*** 1.19 .53 (.04)*** 1.70 .42 (.05)*** 1.52 Sanction policy c –.08 (.04)* .92 –.04 (.03) .96 –.05 (.04) .96 Intercept − 1.56 (.03)*** .21 –.15 (.03)*** .86 − 1.05 (.03) .35 β Log-Odds Coefficient, SE Standard Error, OR Odds Ratio. a The models also control for lagged delinquency at Time 1, which is positively and significantly associated with the outcomes, b Peer delinquency represents peer smoking, peer drinking, and peer fighting, respectively, in the smoking, drinking, and fighting models, c Smoking, fighting policy: 1 = expulsion or out of school suspension, 0 = verbal warning, minor sanction, or detention, Drinking policy: 1 = expulsion, 0 = verbal warning, minor sanction, detention, or out of school suspension. *p < .05; **p < .01; ***p < .001 (two–tailed tests). Table options Do harsh school sanction policies moderate the association between peer and respondent delinquency? Next, we present a preliminary test of our motivating hypothesis by adding cross level interaction terms between peer delinquency and the corresponding school sanction policy to the models presented in Table 1. The simple slopes and interaction effects are presented in Table 2. The negative and significant cross-level peer delinquency × sanction policy interaction terms across all three models (− .32, p < .01 for smoking; –.16, p < .05 for drinking; –.38, p < .01 for fighting) provide strong evidence that harsh school sanction policies weaken the positive association between peer delinquency and respondent delinquency. Table 2. Simple slope and interaction analysis for association between Time 1 peer delinquency on Time 2 respondent delinquency, by school type Peer delinquency simple slope a Peer delinquency × sanction policy interaction effect b School type β (SE) OR β (SE) OR Smoking model Without policy .38 (.09)*** 1.46 –.32 (.10)** .73 With policy .06 (.06) 1.06 Drinking model Without policy .62 (.06)*** 1.86 –.16 (.07)* .85 With policy .46 (.06)*** 1.58 Fighting model Without policy .77 (.13)*** 2.16 –.38 (.14)** .68 With policy .39 (.05)*** 1.48 β Log-Odds Coefficient, SE Standard Error, OR Odds Ratio. a Peer delinquency represents peer smoking, peer drinking, and peer fighting, respectively, in the smoking, drinking, and fighting models, b Smoking, fighting policy: 1 = expulsion or out of school suspension, 0 = verbal warning, minor sanction, or detention, Drinking policy: 1 = expulsion, 0 = verbal warning, minor sanction, detention, or out of school suspension. *p < .05; **p < .01; ***p < .001 (two–tailed tests). Table options The results of a simple slope analysis in Table 2 demonstrate the peer delinquency effects in schools with and without strict sanction policies. For example, the association between peer and respondent smoking was highly significant in schools without a strict sanction policy for smoking (.38, p < .001), but did not reach significance in schools with a strict sanction policy for smoking (.06, p > .05). The peer drinking/respondent drinking relationship was highly significant in both types of schools, that is, in schools without (.62, p < .001) and with (.46, p < .001) harsh sanctions for drinking. However, the relationship between peer and respondent drinking was 26% [(.62 – .46)/.62] weaker in schools with a harsh sanction policy for drinking than in schools without a harsh sanction policy. The peer fighting/respondent fighting relationship was highly significant in both types of schools (.77, p < .001 in schools without a harsh sanction policy, .39, p < .001 in schools with a harsh sanction policy), but this relationship was 49% [(.77 – .39)/.77] weaker in schools with a harsh sanction policy for fighting than in schools without a harsh sanction policy. The interactions are presented graphically in Fig. 1, Fig. 2 and Fig. 3. Fig. 1 presents the predicted probability of respondent self-reported smoking as a function of peer smoking in schools with and without a strict sanction policy for smoking. The figure ranges from having no friends who smoke more than the respondent (low peer smoking) to having an entire network of friends who smoke more than the respondent (high peer smoking). The attenuating effect of attending a school with a strict smoking sanction policy is demonstrated by the similarity in the predicted probability of respondent smoking across levels of peer smoking in schools with a severe smoking policy: the predicted probability of smoking was nearly equivalent for youths with low (.16), middle (.16), and high (.17) levels of peer smoking. Conversely, in schools without a strict sanction policy for smoking, the predicted probability of smoking increased from .16 for youths with low levels of peer smoking to .22 for youths with high levels of peer smoking. Relationship between peer smoking and predicted probability of respondent ... Fig. 1. Relationship between peer smoking and predicted probability of respondent smoking in schools with and without strict smoking sanction policies. Figure options Relationship between peer drinking and predicted probability of respondent ... Fig. 2. Relationship between peer drinking and predicted probability of respondent drinking in schools with and without strict drinking sanction policies. Figure options Relationship between peer fighting and predicted probability of respondent ... Fig. 3. Relationship between peer fighting and predicted probability of respondent fighting in schools with and without strict fighting sanction policies. Figure options In Fig. 2, the attenuating effect of attending a school with a strict drinking sanction policy is demonstrated by a steeper rise in respondent drinking associated with peer drinking in schools without a strict drinking sanction policy, compared to schools with a strict drinking policy. The average student in a school with a strict sanction policy for drinking had a predicted drinking probability of .41 when peer drinking was low and a probability of .53 when peer drinking was high. The average student in a school without a strict drinking sanction policy had a predicted probability of .41 at a low level of peer drinking and a probability of .57 at a high level of peer drinking. In Fig. 3, the attenuating effect of attending a school with a severe fighting policy is illustrated by the differences in the slopes of the lines in schools with and without a strict sanction policy for fighting. The difference in the predicted probability of respondent fighting for youths with low and high levels of peer fighting was greater in schools without a strict fighting sanction policy (.15) than in schools with a strict fighting sanction policy (.08). In Table 3, we provide a more conservative test of our key hypothesis by controlling for an array of variables that previous research has indicated may be related to adolescent delinquency. As expected, the cross-level peer delinquency/school sanction interactions remain significant for all three outcomes (− .30, p < .01 for smoking, –.20, p < .01 for drinking, –.31, p < .05 for fighting), net of the control variables. Table 3. Hierarchical logistic regression models of Time 2 outcomes regressed on peer delinquency × sanction policy interaction and controls a Smoking Drinking Fighting Variable β (SE) OR β (SE) OR β (SE) OR Individual-level variables Peer delinquency b .17 (.08)* 1.19 .36 (.06)*** 1.43 .48 (.13)*** 1.62 Lagged delinquency c 2.70 (.04)*** 14.88 2.32 (.03)*** 10.18 1.67 (.04)*** 5.31 Male –.19 (.03)*** .83 –.23 (.03)*** .79 .76 (.02)*** 2.14 Black d − 1.10 (.07)*** .33 –.94 (.04)*** .39 .35 (.03)*** 1.42 Age .07 (.01)*** 1.07 .16 (.01)*** 1.17 –.05 (.01)*** .95 School attachment –.07 (.02)** .93 –.11 (.02)*** .90 –.03 (.02) .97 Respondent GPA –.30 (.03)*** .74 –.07 (.03)** .93 –.28 (.03)*** .76 Peer GPA –.25 (.06)*** .78 .01 (.05) 1.01 –.39 (.05)*** .68 Two parents –.04 (.04) .96 –.15 (.03)*** .86 –.02 (.05) .98 Parental attachment .04 (.02) 1.04 .02 (.02) 1.02 –.17 (.02)*** .84 Public assistance .01 (.15) 1.01 –.37 (.05)*** .69 .10 (.06) 1.11 Prior expulsion/suspension .79 (.04)*** 2.20 .44 (.03)*** 1.55 .71 (.03)*** .49 Missed school Time 1 to 2 .63 (.13)*** 1.88 .06 (.13) 1.06 –.24 (.13) .79 School-level variables Sanction policye –.05 (.03) .95 –.07 (.02)** .93 –.12 (.04)** .89 Public school –.52 (.06)*** .59 –.42 (.05)*** .66 .09 (.05)* 1.09 Parent teacher association .07 (.04) 1.07 .14 (.04)** 1.15 .14 (.04)*** 1.15 School delinquency f .36 (.04)*** 1.43 .31 (.04)*** 1.36 .30 (.07)*** 1.3 Cross-level interaction Peer delinquency × policy –.30 (.10)** .74 –.20 (.07)** .82 –.31 (.14)* .73 Intercept − 1.36 (.13)*** .26 .09 (.11) 1.09 − 2.25 (.11)*** .11 β Log-Odds Coefficient, SE Standard Error, OR Odds Ratio. Reference group: d White. a All models also control for other race, age2, and days between survey interviews, which were generally non-significant and did not exhibit patterns across models, b Peer delinquency represents peer smoking, peer drinking, and peer fighting, respectively, in the smoking, drinking, and fighting models, c Lagged delinquency represents Time 1 respondent smoking, drinking, and fighting, respectively, in the smoking, drinking, and fighting models, e Smoking, fighting policy: 1 = expulsion or out of school suspension, 0 = verbal warning, minor sanction, or detention, Drinking policy: 1 = expulsion, 0 = verbal warning, minor sanction, detention, or out of school suspension, f School delinquency represents average school-level smoking, drinking, and fighting, respectively, in the smoking, drinking, and fighting models. *p < .05; **p < .01; ***p < .001 (two–tailed tests). Table options Other important patterns also emerge. Pertaining to the individual difference covariates, males in our sample were less likely than females to smoke (− .19, p < .001) and to drink (− .23, p < .001), but were more likely than females to fight (.76, p < .001). In addition, Black youths were less likely than White youths to smoke (− 1.10, p < .001) and to drink (− .94, p < .001), but were more likely than White youths to fight (.35, p < .001); and age was positively associated with smoking (.07, p < .001) and drinking (.16, p < .001), but negatively associated with fighting (− .05, p < .001). Furthermore, respondents were less likely to smoke (− .07, p < .01) and to drink (− .11 p < .001) when they had higher levels of school attachment, and delinquency was inversely associated with personal academic achievement (− .30, p < .001 for the smoking model, –.07, p < .01 for drinking, –.28, p < .001 for fighting) and peer academic achievement (− .25, p < .001 for the smoking model, –.39, p < .001 for fighting). The results also show that respondents were more likely to engage in delinquency when they were previously expelled or suspended from school (.79, p < .001 for the smoking model, .44, p < .001 for the drinking model, .71, p < .001 for the fighting model). Conversely, the effects of age, parents’ marital status, parental attachment, public assistance, and dropping out or moving in between survey interviews were inconsistent across the models. Related to the school-level covariates, the results indicate that respondents attending public schools were less likely to smoke (− .52, p < .001) and to drink (− .42, p < .001), but were more likely to fight (.09, p < .05), than respondents attending private schools. In addition, the presence of a school parent teacher association tended to increase levels of respondent drinking (.14, p < .01) and fighting (.14, p < .001), and youths were more likely to engage in delinquency when they attended schools with higher average levels of student smoking (.36, p < .001), drinking (.31, p < .001), and fighting (.30, p < .001).