Leveraging NLP and Large Language Models to Quantify Bureaucratic Discretion in U.S. Federal Agencies

• In Event: Bureaucratic Discretion: Measurement, Decisions, and Outcomes

Brief Overview

We use NLP and fine-tuned LLMs to analyze U.S. federal job specifications, quantifying bureaucratic discretion by scoring occupations. Aggregated bureau-level scores reveal how discretion influences policymaking at the administrative level.

Author

• William G. Resh, University of Southern California