Current empirical estimates of racial and other unwarranted disparities in sentencing suffer from two pervasive flaws. The first is a focus on the sentencing stage in isolation. Studies control for the “presumptive sentence” or closely related measures that are themselves the product of discretionary charging, plea-­‐bargaining, and fact-­‐finding processes. Any disparities in these earlier processes are built into the control variable, which leads to misleading sentencing-­‐disparity estimates. The second problem is specific to studies of sentencing reforms: they use loose methods of causal inference that do not disentangle the effects of reform from surrounding events and trends.

This Article explains these problems and presents an analysis that corrects them and reaches very different results from the existing literature. We address the first problem by using a dataset that traces cases from arrest to sentencing and by examining disparities across all post-­‐arrest stages. We find that most of the otherwise-­‐ unexplained racial disparities in sentencing can be explained by prosecutors’ choices to bring mandatory minimum charges. We address the problem of disentangling trends using a rigorous method called regression discontinuity design. We apply it to assess the effects of the loosening of the U.S. Sentencing Guidelines in United States v. Booker. Contrary to prominent recent studies, we find that Booker did not increase disparity, and may have reduced it.


Criminal Law | Law | Law and Economics

Date of this Version