Determinants of seat belt use: A regression analysis with FARS data corrected for self-selection

We develop a methodology to use FARS data as an alternative to NOPUS in estimating seat belt usage. The advantages of using FARS over NOPUS are that (i) FARS is broader because it contains more variables relevant for policy analysis, (ii) FARS allows for easy multivariate regression analysis, and finally, (iii) FARS data is more cost-effective.

We apply a binary logit model in our analysis to determine the likelihood of seat belt usage given various occupant, vehicle, and built environment characteristics. Using FARS data, we derive coefficient estimates for categories such as vehicle occupants' age and night time seat belt use that observational surveys like NOPUS cannot easily provide.

Our results indicate that policies should focus on passengers (as opposed to drivers), male and young vehicle occupants, and that law enforcement should focus on pick-up trucks, rural roads, and nights. We find evidence that primary seat belt laws are effective.

Although this is primarily a methodological paper, we present and discuss our results in the context of public policy so that our findings are relevant for road safety practitioners, researchers, and policymakers.