![Sample our Law journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5942/TOC-Subject-Sample-2021-Law.jpg"})
Abstract
This study examined the effects of school‐based crime prevention strategies aimed at reducing criminal opportunity. Results are mixed as to the effectiveness of such efforts in reducing violent victimization among students. Further, few studies have examined the effects net of student‐level risk factors. Finally, it is unclear as to whether such measures agitate or placate students’ risk perception and fear. Guided by a multilevel opportunity perspective, this study used self‐report data from 2,644 seventh‐grade students nested within 58 schools to test whether such efforts reduce students’ victimization, risk perception, and fear of violence at school. Hierarchical logistic models were estimated to control for individual‐level opportunity for victimization. Net of compositional differences, the prevention practices did not significantly reduce the likelihood of experiencing violent victimization or perceptions of risk, and only one measure, metal detectors, significantly reduced fear. Implications for school crime prevention are discussed in light of the findings.
Acknowledgments
This research was sponsored in part by grant DA‐11317 (Richard R. Clayton, principal investigator) from the National Institute on Drug Abuse. We thank Richard R. Clayton, Graham C. Ousey, Scott A. Hunt, Michelle Campbell Augustine, Shayne Jones, Kimberly Reeder, Staci Roberts, and Jon Paul Bryan for their contributions to the Rural Substance Abuse and Violence Project, which provided the data analyzed here.
Notes
1. Standard practice in the “fear of crime” literature is to distinguish between “risk perceptions” and “fear” as related, yet distinct, concepts. Risk perception refers to the perceived likelihood of experiencing victimization, whereas “fear of crime” typically refers to worry or concern about experiencing victimization (e.g., Ferraro, Citation1995; Ferraro and LaGrange, Citation1987).
2. Beyond school‐level misconduct and disorder, studies have also operationalized exposure with measures of school size, school type (i.e., middle school, high school, etc.), public v. private designation, and rural/urban location under the assumption that larger, urban, public schools, for instance, would provide less student homogeneity and thus greater levels of aggregate exposure or proximity to motivated offenders. There is, generally speaking, much inconsistency in the effects of such proxy measures across studies (cf. Anderman & Kimweli, Citation1997; Benbenishty & Astor, Citation2005; Burrow & Apel, Citation2008; Campbell Augustine et al., Citation2002; George & Thomas, Citation2000; Payne et al., Citation2003; Schreck et al., Citation2003). Since all of our subjects were seventh graders from public schools, several of these proxy measures are not pertinent to our data. In preliminary models not reported here, however, we did control for school size. Its effect was non‐significant.
3. The authors were reluctant to suggest that such findings were an indication that communal schools did not matter for student victimization. One potential alternative explanation offered by the authors for the different effects on student versus teacher victimization is that the study relied on teacher indicators of communal school organization.
4. Once parental consent was obtained, sample attrition was largely due to either student absenteeism on the day of survey administration or student transfer to another school. Tracking procedures allowed investigators to obtain about half of the absentee and transfer students. Less than 1% of student attrition was due to outright refusal to participate. As reported elsewhere (e.g., Wilcox et al., Citation2009), comparison of sample characteristics with demographic data on the participating schools from Kentucky Department of Education showed that the sample is similar to the targeted population in terms of race, but it slightly under‐represents male students.
5. Though we do not know for certain whether sample bias resulted from non‐participation in either the student or principal survey among the 74 targeted schools, there was not a clear, systematic pattern behind such non‐response/attrition. Wilcox, May, and Roberts (Citation2006, p. 510) report that “refusals came from large, mid‐size, and small districts/schools, ranging in setting from rural to urban and from various regions of the state.”
6. Incomplete (i.e., missing) data from the student survey were largely confined to one variable—peer delinquency.
7. With the exception of the items used to construct self‐reported criminal behavior and delinquent peers, each set of items loads on one unique factor and these factor loadings range from moderate to high (see Appendix for specific values). Unlike some multi‐item measures that are created to indicate one latent construct (i.e., school efficacy), the self‐reported criminal behavior and delinquent peers measures were created to control for average levels of students or their peers engaging in different illegal behaviors, respectively. Whether they represent one latent factor is not relevant for these two measures because our theoretical interest is to capture the essence of a range of different behavior that students and their peers might engage in. Hence, if the respective items load on one or more factors is not relevant to the estimation of statistical models for hypothesis testing.
8. We initially controlled for socioeconomic status, which was measured as the highest educational attainment of the respondent’s mother or father, scored on a seven‐point scale. However, there was substantial missing data on this variable and it was not significant across any of the models. We then conducted the analyses without socioeconomic status using the same sample (i.e., only those cases with valid values on all study variables, including socioeconomic status) and compared the results. No coefficients changed in direction or significance when socioeconomic status was removed from the analyses. Further, equality of coefficients tests revealed no significant differences across the models. Therefore, socioeconomic status was ultimately not included in the estimation of the final models reported here to retain cases.
9. Initial models which included the individual‐ and school‐level variables described above also controlled for percent of school on free or reduced lunch, total enrollment, percent male, and percent non‐white at the school level (data were obtained from the Kentucky Department of Education). Because these variables were not significant across the victimization, risk perception, and fear models, they were removed for parsimony and to ensure model stability.
10. The independent variables were grand‐mean centered and robust standard errors were estimated for all models.
11. Due to the small sample size at Level 2 (N = 58 schools), we report significance at both p < 0.10 and p < 0.05 for Level 2 variables.
12. PATHS—Promoting Alternative Thinking Strategies—is just one example of a relatively successful social competency program reviewed by Gottfredson (Citation2001).
