Replication Data for: One Threshold Doesn’t Fit All: Tailoring Machine Learning Predictions of Consumer Default for Lower-Income Areas

This replication package contains code and supplementary data for "One Threshold Doesn't Fit All: Tailoring Machine Learning Predictions of Consumer Default for Lower-Income Areas" (Meursault, Moulton, Santucci, & Schor, Journal of Policy Analysis and Management). The package includes scripts for data preparation for this research, machine learning model training, and threshold analysis, organized in three main directories. This package provides supplementary data including rolling window specifications, FIPS codes, and Community Reinvestment Act Low and Moderate Income area designations. The master script _00_run_all.R orchestrates the complete replication workflow. Note that certain variable names have been removed per contractual limitations on sharing data content or attributes. The data that support the findings of this study are from the Federal Reserve Bank of New York Consumer Credit Panel/Equifax data (CCP). Access to the CCP microdata is limited to Federal Reserve System researchers, and their coauthors, due to contractual limitations. Individual analyses also use the Federal Reserve Board’s Capital Assessments and Stress Testing (Y-14M) report and the HMDA-McDash-CRISM dataset: a combination of Home Mortgage Disclosure Act (HMDA) data, Black Knight McDash (McDash) data, and Equifax Credit Risk Insight Servicing data, that is linked to the McDash data (CRISM). Access to these datasets is also restricted.