Methods of Analysis and Findings

Profile of Studies in the Database

The pool of studies selected for the meta-analytic database shared the following features:

• They were conducted in the United States by psychologists, criminologists, or sociologists and were published after 1970.
  
  
• The sample populations were largely male, mostly white or of mixed ethnicity, with an average age of 14 to 17 years. Most of the juveniles had prior offenses. In two-thirds of the samples, some or all of the juveniles had a history of aggressive behavior.
  
  
• In most of the samples, juveniles were under the supervision of the juvenile justice system and were receiving court-ordered intervention. In one-third of the sample groups, juvenile justice personnel administered treatment. In one-fifth of the groups, treatment was administered by mental health personnel in public or private agencies. In the remainder of the groups, it was administered by other counselors, laypersons, or researchers.
  
  
• For noninstitutionalized juveniles, the interventions studied included counseling, skill-oriented programs, and multiple services (combinations of services or treatments that involved several different approaches). For institutionalized juveniles, they included counseling, skill-oriented programs, and community residential programs. Treatments usually lasted 1 to 30 weeks and involved continuous contact or sessions that ranged from once or twice per week to daily, for 1/2 hour to 10 hours per week.
  
  
• Almost half of the studies used random assignment to experimental conditions; many of the others used some form of matching. Control groups typically received the usual treatment (e.g., regular probation or institutional programs). The recidivism outcome variables that were measured most frequently were police contact or arrest, court contact, or parole violations.

Recidivism Effect Size

Only one recidivism outcome measure was selected from each study. Police contact or arrest was selected if it was available; otherwise, officially recorded contact with juvenile court or offense-based probation violations were used because they are the most comparable to police arrest. The difference between the treatment and control groups on the selected recidivism measure was calculated for each study and standardized so that different measures could be compared.

Overall, juveniles who received treatment showed an average 12-percent decrease in recidivism. This result, while not enormous, was positive, statistically significant, and large enough to be meaningful. More important, however, was the large variability in effects across studies. The remainder of this Bulletin explores the characteristics of the interventions that produced the largest effects on recidivism.

Variation in Study Methods and Procedures

The differences in methods and procedures used in the studies are the first source of variability in effect size. The use of a multiple regression equation made it possible to estimate what the mean effect size over the 200 studies would be if all the studies were uniform in method and procedure. The method-adjusted effect sizes were then analyzed in terms of various treatment variables to identify those producing larger effects.

Effective Intervention for Serious Juvenile Offenders	Juvenile Justice Bulletin • April 2000