In studies of both juvenile and adult recidivism, the institutional variables that have been investigated most often are parole prognosis (staff predictions of reoffending), institutional misconduct, personality measures, participation in work or education programs, and frequency of family contact. Overall, these studies rarely produced instruments that predict more than 15 percent of the variation in recidivism, whether they included institutional variables or not. Institutional misconduct appears to be the most reliable institutional predictor variable. Parole prognosis was often predictive, but this is a subjective judgment and it is difficult to know on what basis it was made. Personality measures can add to base expectancy scores for juvenile offenders, but some scales may be reflecting previous misconduct more than a personality dimension. The utility of program participation and family contact is less clear. Because most studies did not directly consider the extent to which institutional predictors added to the efficiency of predictions, and because of difficulties of comparability of definitions and predictor or criterion variables, the extent to which institutional variables add to known predictors of recidivism, such as previous criminal records, remains unclear. Tabular data on the studies and 78 references are included.