This study examined the predictive efficacy of 10 rationally derived, archivally coded variables for assessing recidivism risk among extrafamilial child molesters.
Follow-up data on 111 child molesters who were discharged from the Massachusetts Treatment Center between 1960 and 1984 were used. Data were obtained from each offender's prison record and criminal record. Data for coding all of the risk-assessment variables were obtained from the offender prison record. The variables were amount of contact with children, degree of sexual preoccupation with children, impulsivity, juvenile antisocial behavior, adult antisocial behavior, frequency of prior sexual offenses, paraphilias, alcohol-use history, social competence, and victim's gender. All postrelease charges against subjects in the sample were encoded in chronological order for each individual, beginning in 1960 and ending on January 1, 1985. For each charge, researchers coded the date, the description of the charge, the State in which it occurred, and whether it was a sexual offense. The variables related to a prediction of recidivism for a sexual offense were sexual preoccupation with children, paraphilias, and number of prior sexual offenses. Juvenile and adult antisocial behavior, paraphilias, and low amount of contact with children predicted nonsexual victim- involved and violent recidivism. The area under the receiver- operating characteristics curves in all three analyses showed discrimination that was substantially better than chance, ranging from .73 to .79. 6 tables, 1 figure, and 39 references