Understanding High Skill Worker Productivity Using Random Case Assignment in a Public Defender's Office
University of Pennsylvania Law School
University of Toronto - Faculty of Law
3rd Annual Conference on Empirical Legal Studies Papers
Measuring high skill worker productivity presents several challenges. High skill workers almost always select their tasks, and unobserved variation across tasks introduces estimation bias. In addition, output for high skill workers is difficult to measure. We exploit a natural experiment where cases are randomly assigned to attorneys within a public defender office to address the selection problem; the random assignment ensures that unobservables have the same distribution across attorneys. Using this data, we are able to investigate the efficiency of the labor market for attorneys, theories of human capital accumulation, and labor market discrimination. Despite a relatively flat wage distribution within cohorts, we find substantial heterogeneity in attorney productivity, as measured by case outcomes. A defendant assigned to a public defender at the 10th percentile of the productivity distribution has an expected sentence length 5.3 months (74% of the mean) longer than the defendant assigned to the public defender at the 90th percentile. Adding attorney fixed effects to a regression of sentence length on case and defendant characteristics doubles the explanatory power. While we do not find any impact of gender or law school quality, Hispanic attorneys in our data set have significantly higher productivity than non-Hispanic attorneys, as do attorneys with greater tenure. These findings suggest that there may be some discrimination in the labor market, and that the positive correlation of wages and tenure is due at least in part to human capital accumulation.
Number of Pages in PDF File: 43
JEL Classification: J15, J24, J44, J70, K40
Date posted: April 16, 2008
© 2016 Social Science Electronic Publishing, Inc. All Rights Reserved.
This page was processed by apollobot1 in 0.218 seconds