A Classification Tree Approach to Analyzing the Interactive Effects of Race on Juvenile Court Decisions

• In Event: Poster Session II

Abstract

Recent research on Disproportionate Minority Contact (DMC) in juvenile justice highlights the limitations of an exclusive focus on main effects in studying the relationship between race and court outcomes. By extension, race might be expected to have moderating effects on multiple legally-relevant factors. In this study, classification and regression trees—with youths’ race as a focal variable—are used to detect interactions with legally-relevant variables and to create prediction models for juvenile court decisions about detention and secure confinement. Basic regression fits a single formula to all cases included in the data; however, this can be problematic when there are many interactions and/or nonlinear relationships among the covariates. Classification models address these problems by partitioning the data multiple times based on the values of included covariates and then fitting prediction models within each partition. Data for this study come from official juvenile court records from multiple agencies in a large state and include measures of youths’ race, extralegal factors (e.g., sex), legally-relevant factors (e.g., prior record), and dispositions (e.g., youth placed in a secure confinement facility). The discussion focuses on the study’s findings in terms of identifying and understanding DMC and the method’s potential for research in this area.

Authors

• Shaun Gann, Boise State University

• Christopher Sullivan, University of Cincinnati