Using Theory-Driven And Machine Learning Approaches To Examine The Ability To Pay Court-Imposed Fines

• In Event: Disparities in Punishment

Abstract

With the increasing use of fines in the U.S. and their negative impacts, including increased homelessness, decreased employability, and lengthier supervision, questions of equality and equity arise regarding their discretionary and inconsistent imposition. This study utilizes data from the Pennsylvania Commission on Sentencing from 2001 through 2012 to examine the implementation of fines related to individuals’ ability to pay, focusing on individuals with fines imposed as their only sentence (n=303,662). Based on a review of existing research, our theory-driven approach used fixed effects linear regression and logistic regression models including variables for gender, race, offense type, employment status, and median income by zip code. Preliminary analysis indicates that, while imposition of fines appear racially-neutral, there are racially-disparate impacts for individuals’ ability to pay. Additionally, fines imposed did not uniformly follow recommendations for lower and upper limits. Additionally, we used LASSO and Elastic Net methods to examine whether variables outside those established in previous studies influence the imposition of fines. We anticipate that the use of such machine learning techniques may uncover new variables previously not considered that help explain how and why courts impose fines and how these variables relate to defendants’ ability to pay their fines.

Authors

• Divya Ramjee, American University

• Jazmine Talley, American University

• Catherine Lee, George Mason University