Predicting Sexual Recidivism

Area of Science:

• Criminology
• Forensic Psychology
• Juvenile Justice Research

Background:

• Adolescent sex offending presents significant public safety concerns.
• Understanding recidivism patterns is crucial for effective intervention and risk management.
• Existing research highlights the difficulty in accurately predicting sexual reoffending.

Purpose of the Study:

• To examine sexual reoffending patterns in adolescents with prior sex offense histories.
• To identify predictors of sexual recidivism within two years of a previous offense.
• To evaluate the predictive accuracy of statistical models using official juvenile justice data.

Main Methods:

• Utilized inductive statistical models (random forest) on archival data from the Florida Department of Juvenile Justice.
• Included social, offense, placement, and risk assessment history data for referred youth.
• Assessed predictive accuracy using receiver operator characteristic (ROC) curves and area under the curve (AUC) metrics.

Main Results:

• The number of prior felony and misdemeanor sex offenses emerged as the strongest predictor of sexual recidivism.
• Area under the curve (AUC) values ranged from 0.65 to 0.71, indicating modest predictive accuracy.
• The findings align with existing literature on the complexities of predicting sex offense recidivism.

Conclusions:

• Accurately predicting adolescent sexual recidivism remains a significant challenge.
• The study underscores the importance of prior offense history in risk assessment.
• Results contribute to the literature on risk assessment factors utilized by official institutions.