questions where substantial empirical background information exists, conventional frequentist policy
analysis is hard to justify. Bayesian analysis
quantitatively incorporates prior knowledge, but is not often used in applied policy
We combined 2000–2016 data from the Fatal Analysis Reporting System with priors based on past empirical studies and policy
documents to study the impact of mandatory seat belt
laws on traffic fatalities. We used a Bayesian data augmentation approach to combine information from prior studies with difference-in-differences analyses of recent law changes to provide updated evidence on the impact that upgrading to primary enforcement of seat belt
laws has on fatalities.
After incorporating the evidence from past studies, we find limited evidence to support the hypothesis that recent policy
upgrades affect fatality rates. We estimate that upgrading to primary enforcement reduced fatality rates by 0.37 deaths per billion vehicle miles traveled (95% posterior interval -0.90, 0.16), or a rate ratio of 0.96 (95% posterior interval 0.91, 1.02), and increased the proportion of decedents reported as wearing seat belts by 7 percentage points (95% posterior interval 5, 8), or a risk ratio of 1.18 (95% posterior interval 1.13, 1.24).
Bayesian methods can provide credible estimates of future policy
impacts, especially for policy
questions that occur in dynamic environments, such as traffic safety.