'What Works?' Revisited Again: A Meta-analysis of Randomized Field Experiments in Rehabilitation, Deterrence and Prevention

• NCJ Number: 170149
• Author(s): A. J. Petrosino
• Date Published: February 1999
• Length: 308 pages
• Agencies: NIJ-Sponsored

Annotation

One hundred fifty studies of individual-level interventions with offenders were examined by means to meta-analysis to assess the overall impact of such interventions on subsequent offending.

Abstract

The studies were selected by means of the alternation technique from 300 studies that used a randomized experimental design, were written in English, were available during 1950-93, and included a quantifiable outcome measure of crime. Each report was coded using a 196-item instrument. Intercoder reliability was assessed on a random subset of studies; the overall rate of agreement was 80 percent. The intervention ranged from treatment programs to deterrence strategies to juvenile delinquency prevention approaches. One hundred fifteen of the studies were rehabilitative in focus, 23 were focused on deterrence, and 9 were delinquency prevention programs. The analysis indicated that the overall average effect size was highly unstable. However, a homogeneity test indicated considerable variability in effect size across rehabilitation experiments. Studies with 10-100 participants reported much larger effects than experiments with more than 300 participants. In addition, experiments with youths under age 18 were more effective than programs for adults. Recommendations for future research and the strengthening of meta-analysis, tables, figures, footnotes, appended methodological information, author biography, and approximately 500 references