Identifying Hierarchical Interactions amongst Pretrial Risk Factors

• In Event: Pretrial Release Assessments & Bail

Abstract

Pretrial risk assessment tools are often used to inform pretrial decisions. A minority of validation studies have examined biases at the risk item-level. Terranova & Ward (2020) constructed a pretrial risk assessment tool conducting an in-processing calibration of the tool at the item-level. This study identified risk factors that may be contributing predictive biases and used these estimates for calibrating the tool. The current study explores biases in individual risk factors by fitting sparse hierarchical interaction models with demographic characteristics, race/ethnicity and sex (Bien, Taylor & Tibshirani, 2013; Bien & Tibshirani, 2022). A retroactive sample of defendants arrested, detained, and released pretrial that includes a variety of common pretrial risk assessment factors is used. Findings provide support for risk factor prediction and associations with a pretrial defendant’s protected sub-group membership. This will provide implications for furthering the on-going study of pretrial risk assessment, predictive biases, and pretrial release decision-making.

Author

• Victoria Terranova, University of Northern Colorado