Non-random assignment, vehicle safety inspection laws and highway fatalities

In this study, the distribution of inspection laws across states is endogenously determined by the relative strengths of lobbying groups within states. Previous studies that treat the laws as exogenous and find a 5 to 15 percent reduction in fatalities may have produced biased results. A selection bias model is developed in which non-random assignment is taken into account. Two equations are estimated: one explaining how many inspections are required, and the other explaining the effects of the inspections on fatalities per capita. Using single-equation techniques, results are obtained that are similar to prior studies. In the two-equation model that accounts for non-random assignment, inspection laws are not found to significantly reduce fatalities per capita.