Sturdy Inference: A Bayesian Analysis of U.S. Motorcycle Helmet Laws

Abstract

Motorcycle related fatalities continue to be a major concern for public health officials, economists, and policy makers interested in such matters. In 2006, 3% of all motor vehicles registered in the United States were 2-3 wheelers (motorcycle type vehicles), while riders of these vehicles accounted for 11% of vehicle related deaths. Such a disproportionate number of fatalities associated with motorcycles is certainly grounds for concern. Most studies of motorcycle fatalities attribute deaths to the avoidance of wearing helmets and the lack of helmet laws, speed, and alcohol usage. This study makes use of a rich panel data set for the period 1980 to 2010 by state and the District of Columbia to examine these factors and others. It is the first study to differentiate between the effects of universal and partial helmet laws on motorcycle fatalities. It also accounts for the effects of cell phone use, alcohol consumption, and suicidal propensities on these crashes after adjusting for a whole host of socioeconomic and driving related factors. The analysis is conducted using a new Bayesian technique, which examines the sturdiness of regression coefficients. This new method uses statistics referred to as S-values that addresses both estimation and model ambiguity. Results indicate that the variables we focus on, i.e., cell phones, alcohol consumption, and helmet laws affect motorcycle fatalities. Further, universal helmet laws appear to have a larger effect on such fatalities than partial helmet laws.