Predictive Fairness of California-Youth Assessment and Screening Instrument on Infractions By Socioeconomic Status

Background and Purpose: Risk assessment tools are data-driven evaluations that estimate the likelihood that a person will commit a future offense. Proponents for the use of risk assessment tools in the juvenile justice system argue that they are more accurate, consistent, and objective measures for evaluating and sentencing offenders compared to individual decision-making unaided by a structured tool. Those who oppose the use of risk assessment tools argue that some risk factors used in these tools, such as criminal history, can serve as proxies for race and poverty. This research study evaluated the ability of the California-Youth Assessment and Screening Instrument (CA-YASI) to fairly predict serious institutional infractions among lower vs. higher socioeconomic status juveniles in the California juvenile justice system.

Methods: This study utilized an archival dataset that was originally created in a study assessing the predictive validity of the CA-YASI. The sample included 846 male youths in juvenile detention assessed using the CA-YASI. We conducted a series of logistic regressions to test for the potentially moderating effect of socioeconomic status (SES) on the relationship between total YASI score of an individual and whether or not they committed a serious infraction while incarcerated.

Results: Higher CA-YASI scores were associated with a greater likelihood that a juvenile would commit a serious infraction (OR=1.03, p<0.001), but the interaction between the moderating variable of SES and total CA-YASI score was not significant significant (OR=1.00, p=0.41). A likelihood ratio test to examine intercept bias found a significant difference between nested models 1) predicting institutional infractions from total YASI score and 2) predicting institutional infractions from total YASI score and SES score (X² (1) = 15.09, p<0.001). A likelihood ratio test to examine slope bias did not find a significant difference between nested models 1) predicting institutional infractions from total YASI score and SES score and 2) predicting institutional infractions from total YASI score, SES score, and the interaction between total YASI score and SES score (X² (1) = 0.69, p=0.40).

Conclusions and Implications: This study examined the primary risk assessment tool used in the juvenile justice system in California, and results can help guide policymakers and practitioners in their use of risk assessment tools. Results indicated that the CA-YASI predicts reoffending strongly and in similar form, regardless of a youth’s SES. Thus the results from the present study found that the CA-YASI is not predictively biased by socioeconomic status, suggesting that this risk assessment tool may serve its purpose of accurately evaluating risk levels regardless of a youth’s SES. If anything, the intercept bias results suggest that the CA-YASI underestimates the chance that lower SES youth would commit a serious infraction. Further research is needed to determine if using the CA-YASI has a disparate impact on those of a lower SES.