Results from a Multi-Site Evaluation of the G.R.E.A.T. Program

Abstract Despite a long history of youth gang problems in the United States, there remains a paucity of evaluations identifying promising or effective gang prevention and intervention programs. One primary prevention program that has received limited support is Gang Resistance Education and Training (G.R.E.A.T.). An earlier national evaluation of the G.R.E.A.T. core middle school curriculum reported modest program effects but, importantly, found no programmatic effect on gang membership or delinquency. This manuscript presents results from a second national evaluation of the revised G.R.E.A.T. core curriculum that utilizes a randomized field trial in which classrooms were randomly assigned to treatment and control conditions. Approximately 4,000 students attending 31 schools in seven cities comprise the initial sample. Analyses of one-year post-treatment data indicate that students receiving the program had lower odds of gang membership compared to the control group. Additionally, the treatment group also reported more pro-social attitudes on a number of program-specific outcomes. Keywords: youth gangsgang preventionprogram evaluationrisk factors Acknowledgments This research was made possible, in part, by the support and participation of seven school districts, including the School District of Philadelphia. This project was supported by Award No. 2006-JV-FX-0011 awarded by the National Institute of Justice, Office of Justice Programs, US Department of Justice. The opinions, findings, and conclusions or recommendations expressed in this manuscript are those of the authors and do not necessarily reflect the views of the Department of Justice or of the seven participating school districts. We would like to express our appreciation to the students who made this project possible by completing the student questionnaires. And, this project would have been impossible without our team of colleagues and research assistants; special thanks to Adrienne Freng, Brad Brick, and Dena Carson. Notes 1. In another article, we have reported on the high level of program fidelity associated with delivery of the G.R.E.A.T. program in classrooms participating in the current evaluation, allowing outcome evaluation results to be attributed with confidence to the program (Esbensen, Matsuda, Taylor, the second principal would not agree to random assignment and withholding some students from the program. 8. One of the five originally-selected schools in Chicago (comprised of nearly 100% African-American students) agreed to participate in the evaluation but was unable to meet the requirements of the study and was dropped from the sample. Given time constraints (i.e. too late in the school year to select a comparable school and implement the program with fidelity), we were unable to replace the excluded school during 2006–2007. Thus, the resulting sample was disproportionately Hispanic and not representative of the district. To increase sample representativeness, we added two primarily African-American schools to the evaluation in the 2007–2008 school year, even though this meant that these schools would be one year behind other schools in the evaluation. 9. The analysis file includes data for 3,246 students with data for both Waves 2 and 3, another 368 for Wave 2 but not Wave 3, and 88 for Wave 3 but not Wave 2, for a total of 3,702 students with either or both. The 3,702 students represent an upper bound for the analyses because it counts youth with any data and does not take into account variable-specific missing data on any given outcome or cases lost when we control for Wave 1 (from being missing on the same variable). The analysis-specific counts of cases are for person/waves rather than people (as specified in MLwiN). In the basic model (without Wave 1 control) we lose cases only due to being missing on the outcome because the only other variables involved we have for everybody (wave, site, and treatment/control). With respect to missing data, the total dataset has 6,948 person/wave cases; the number included in the analyses with & without Wave 1 control varies from 6,611 and 6,180 (attitudes toward gangs) to 6,905 and 6,751 (school disorganization). 10. Analyses were also conducted separately by wave, to assess treatment effects at post-test and treatment effects at the one-year follow-up. For all but 5 measures, there was a significant treatment effect at both time points. For the five that differed, the difference in effect between Wave 2 and Wave 3 was not statistically significant, and there was a statistically-significant average treatment effect across the time periods. 11. In an unpublished report submitted to NIJ and in Esbensen et al. (Citation2011), we reported a 54% reduction in the odds of gang joining. The difference reported here is due to a change in the MLwiN program that now allowed the model to run with all variance terms included in the analysis. 12. In response to one reviewer’s concerns, the Wave 2 specific program effect was a 38.7% reduction in the odds of gang membership and 40.6% for Wave 3. aProgram effect as percent reduction. bNegative binomial model. cLogistic regression model. 13. We acknowledge that given the findings reported for the other two program goals and proximate program goals, it is surprising that there was no reduction in offending associated with the program. This is especially so, given the overlap in risk factors associated with gang membership and offending.