Differential Racial-Ethnic Predictive Validity

This study examines the predictive validity of the Los Angeles County Needs Assessment Instrument (LAC) on a sample of African American and Hispanic juvenile probationers. It focuses on analyzing predictive error, which is often overlooked in risk assessment research.

BACKGROUND

Recent research suggests including White offenders in risk assessment samples can bias the predictability for minority groups. This study addresses that limitation by examining the LAC’s predictive validity specifically for African American and Hispanic juveniles, without the influence of White offenders. It also examines classification errors, which are rarely studied in offender assessments.

KEY FINDINGS

  1. Bias in Risk Assessments: Including White offenders in sample groups biases the predictability of risk and needs assessment tools, leading to inaccuracies for minority groups like African American and Hispanic juvenile probationers.
  2. Instrument Performance: The Los Angeles County Needs Assessment Instrument (LAC) predicted better outcomes for Hispanic juveniles compared to African American juveniles, with a 16% better-than-chance classification rate for Hispanics.
  3. Predictive Errors: Predictive error analysis showed that the LAC’s accuracy varied by race, with Hispanic probationers having more accurate predictions compared to African Americans, where predictive validity was nonsignificant.
  4. Study Limitations: The study highlighted the need for further research into the factors causing differential predictive validity and called for periodic reassessment of tools to address these biases.
  5. Policy Implications: Recommendations include revalidating assessment instruments regularly and ensuring they are free of racial/ethnic biases to equitably predict probation outcomes for all groups.

RECOMMENDATIONS

– Periodically reassess youth to determine the impact of interventions on reoffending and adjust service intensity as needed.

– Revalidate the LAC to minimize racial/ethnic predictive disproportionalities.

– Consider using multilevel modeling techniques to account for community-level factors that may influence predictive validity across racial/ethnic groups.

– Examine alternative outcome measures beyond rearrest, such as reconviction or recommitment.

CONCLUSION

The LAC demonstrates better predictive validity for Hispanic juvenile probationers compared to African Americans. The study highlights the importance of examining classification errors and differential prediction across racial/ethnic groups. Further research is needed to understand the factors contributing to these differences and to improve risk assessment instruments’ equitable predictive validity across diverse populations.

Sample and Demographics

The study included 480 male juvenile probationers from Los Angeles County (139 African American, 341 Hispanic). The average age was 16 years old.

Predictive Validity

For Hispanic probationers, the LAC achieved an Area Under the Curve (AUC) of 0.66, indicating it predicted rearrest 16% better than chance. For African American probationers, the AUC was 0.58 and not statistically significant.

Significant Predictors

For Hispanic probationers, gang affiliation, age, and supervision level were significant predictors of rearrest. For African American probationers, only age was a significant predictor.

Classification Errors

There were no significant differences in classification errors between African American and Hispanic probationers across different cutoff scores.

Comparative Performance

The LAC demonstrated better predictive validity for Hispanic probationers compared to African American probationers. Including Hispanic probationers improved the overall predictive accuracy of the instrument.