Hypothesis

Facilities subject to periodic federal inspection (OSHA, CMS, EPA, FDA) exhibit compliance degradation that follows a power-law distribution rather than the linear decay assumed by current inspection scheduling models.

Observable Prediction

Using publicly available federal inspection databases (OSHA IMIS, CMS CASPER/OSCAR, EPA ECHO, FDA inspection databases), the time-to-first-violation after a clean inspection should cluster according to a Pareto distribution with shape parameter α ≈ 1.5–2.0, meaning:

1. ~20% of facilities account for ~80% of post-inspection violations within 12 months
2. The hazard rate is NOT constant — it spikes in months 2–4 post-inspection, plateaus months 5–12, then accelerates again after month 18
3. This pattern should be domain-invariant across regulatory agencies (OSHA workplace safety, CMS nursing homes, EPA emissions, FDA drug manufacturing)

Falsification Criteria

If time-to-first-violation data from any two of the four federal databases (OSHA IMIS, CMS CASPER, EPA ECHO, FDA) fits an exponential or normal distribution better than a power-law (using Kolmogorov-Smirnov goodness-of-fit, p < 0.05), the hypothesis is falsified.

Implications

If confirmed, current fixed-interval inspection schedules (annual, biennial) are systematically misallocating enforcement resources. A power-law-informed inspection model would front-load inspections in the 2–4 month post-inspection window and use Bayesian updating on facility-specific α parameters to dynamically adjust inspection frequency.

Data Sources

All data required to test this hypothesis is freely available:

• OSHA: Establishment Search via enforcement data (public API)
• CMS: Nursing Home Compare, Hospital Compare (data.cms.gov)
• EPA: ECHO Enforcement and Compliance History (echo.epa.gov)
• FDA: Inspection Classification Database (fda.gov/inspections-compliance)

No proprietary data needed. Fully reproducible with federal open data.