Studying Biases in New York's Bail Reform Laws using Bayesian Statistics
New York is the only state in America that requires judges to set bail exclusively based on a defendant’s income and risk of failing to appear for court with no consideration for their perceived “dangerousness” since it can induce racial biases. Following New York’s controversial bail reform laws in January 2020 that relaxed rules around charges where bail could be set, July 2020 amendments reinstated stricter bail requirements. This analysis employs hierarchical Bayesian models to evaluate how bail amounts are set across the population of defendants and by judge on standard criteria such as criminal history and estimated income, as well as the effect of the July 2020 rule changes and judges’ racial biases. In studying population-level effects to bail amounts, the conditional effects of bail on various key variables, and investigating judge-specific behaviors in setting bail, clear and concerning racial discrepancies exist in how bail is set across a variety of metrics in New York state, which the July 2020 reforms do not alter. This raises concerns around racial equity and greater standardization needed in how judges set bail, especially as the recent midterm elections promise sweeping changes to current bail rules.
The raw data for this project can be found here: https://ww2.nycourts.gov/pretrial-release-data-33136
The code for cleaning and analyzing the data can be found in the /src folder The visualization outputs can be found in the /outputs folder
Please message me on LinkedIn if you would like to read the full thesis with full methodology, assumptions, and results.
