Regulatory Stringency and Industry Performance: Evidence from RegData U.S. 6.0

Abstract: This paper studies the economic incidence of broad federal regulation across U.S. industries using RegData U.S. 6.0. We measure regulatory stringency using binding language density--expected regulatory restrictions per 1,000 words--constructed from the text of the Code of Federal Regulations and mapped to NAICS industries using supervised relevance weights. Using the 1970-2025 regulatory panel, we document three stylized facts: (1) binding density evolves through many small year-to-year adjustments and a recent plateauing in 2025; (2) short-run changes are across industries rather than concentrated in a small set of sectors; and (3) annual changes are weakly related to baseline industry size and productivity. We also validate the normalized measure by showing that treated industries exhibit clear and persistent increases in binding density around major legislative events. Linking within-industry variation in binding density to Bureau of Economic Analysis measures of real output and real value added for three-digit NAICS industries over 1998-2024, we find that a one-unit increase in restrictions per 1,000 words is associated with approximately a 0.7 percent decline in real output and a 1.2 percent decline in real value added. The relationship persists in distributed-lag specifications and is robust across alternative functional forms. These findings suggest that normalized regulatory text captures economically meaningful shifts in the regulatory environment and provides a scalable framework for studying their economic effects.