Explaining state-to-state differences in seat belt use: an analysis of socio-demographic variables

Abstract

Despite the extensive evidence about the benefits of seat belt use, there is a great deal of variation in use within the US. For example, the national average for seat belt use in 2009 was 84 percent while the state-level averages ranged from 68 percent in Wyoming to 98 percent in Michigan. The overarching goal of this project was to gain a better understanding of the socio-demographic variables (or factors) that influence statewide belt use rates. To the extent that these vary among states, they can partially account for differences in statewide belt uses. Previous studies have already identified some important factors that affect belt use rates: gender, age, race, vehicle type, seat-belt enforcement laws, and amount of fine for belt-use law violation. This project studied the influence of additional socio-demographic factors on state-level use rates. These factors were: education (percentage of high school educated population), racial composition (percentage of White), median household income, political leaning (percentage of Democrats), and a measure of religiosity. The analysis was based on data from the 2008 Fatality Analysis Reporting System (FARS) which has information on seat belt use on all vehicle crashes that resulted in at least one fatality. The use rates in FARS data were lower than that for the general population. However, our interest is on differences in rates between the states. To the extent that the state-to-state differences in FARS data are consistent with those in the general population, the findings are likely to hold for the population-at-large. Exploratory analysis showed that many of the use rate patterns in FARS data were in fact consistent with those found in other data sets. Of the five socio-demographic factors that were considered, three were identified as important: religiosity, race (percentage White), and political leaning (percentage Democrat). The other two – income and education – were not significant. Hold-out analyses confirmed that this conclusion was consistent across different subsets of data. A regression model that included these new factors accounted for a substantial amount of the state-to-state variation in seat belt use rates, providing further evidence of the usefulness of the results. The findings from this study are preliminary and have to be confirmed on other data sets. Nevertheless, they demonstrate that socio-demographic factors can be used to effectively explain state-to-state variation in seat belt use rates. If factors such as religiosity are indeed important, they can be used to develop appropriate programs for increasing belt use.