“Spillover Effects of Regulatory Inspections: Evidence from OSHA” by Matthew S. Johnson, David I. Levine and Michael W. Toffel, NBER Working Paper 35646 and Harvard Business School Working Paper 27-014, released September 2026

The Occupational Safety and Health Administration long kept a yearly list of roughly 3,500 workplaces reporting the highest injury rates, and on the lists from 2001 to 2010, whether a workplace landed just above or just below the line was close to chance. A paper by Matthew Johnson, David Levine and Michael Toffel, released this month, uses that line to ask what an inspection did for the workplaces nobody walked into. The Labor Department, which houses OSHA, helped fund it.

Quite a lot. One inspection cut injuries serious enough to keep someone off work by about 11 percent at other establishments in the same ZIP code, averaged over the next four years. That comes to roughly three fewer of those injuries a year across the neighbors. Against their earlier estimate of about half an injury a year at the inspected site, the authors put the spillover at several times the direct effect.

The obvious comparison, workplaces with more inspected neighbors against those with fewer, finds essentially nothing. The authors trace that to inspections going where injury rates were unusually high, plausibly because of local conditions the neighbors shared.

The scribe study earlier this week had the same shape, since its first reading undercounted what the tool would give back. I think this is the costlier kind of mistake, because an audit judged only by what changed where it looked will read as overhead the first time somebody hunts for savings. The authors do not test why the neighbors improved; one candidate they list is managers expecting to be next. My bet is that a good share of what an audit function accomplishes works that way.

The inspections are at least fifteen years old, so I would not carry the exact ratio into a budget meeting. I would carry the direction.