According to the U.S. Department of Transportation’s National Highway Traffic Safety Administration (NHTSA), 37,461 people died in the United States in traffic crashes in 2016, a 5.6% increase in fatalities from 2015. It was the second consecutive year of increasing fatalities following an 8.4% increase from 2014 to 2015. This study applies random-effects generalized linear mixed modeling techniques to examine the association of changes in traffic fatality counts with changes in explanatory factors, by state, between 2005 and 2016. Three regressions modeled different outcomes:​ 1) passenger vehicle occupant fatalities, 2) pedestrian fatalities, and 3) motorcycle fatalities

Motor vehicle-related traffic fatalities were collected by year and by state using NHTSA’s Fatality Analysis Reporting System (FARS). A variety of sources provided measures on explanatory factors. The Fatality counts (outcome) and explanatory factors were then combined as panel data by year (2005-2016) and state (51 states including the District of Columbia). The models tested the association between fatalities and more than seventy explanatory factors including economic, exposure, behavioral and vehicle factors.

The study found that the increases in passenger vehicle fatality counts were associated with increases in vehicle miles traveled (exposure) and an improving economy. In addition, the increase in the population age 65 and older and an increase in the percent of this population in the workforce also was associated with increasing fatality counts. Several behavioral factors were associated with changes in fatality counts, including non-belt use and increased drunk driving. Conversely, improved vehicle safety design was associated with a decline in occupant fatalities. A rise in motorcycle fatalities was associated with increased exposure (motorcycle registrations and overall vehicle miles travelled) and an improving economy. Among pedestrian and motorcycle fatalities, there is some evidence that driver distraction plays a role.

While the quasi-experimental study design does not allow for inferences of causality, the models can be applied to forecast future fatality counts based on expected or observed environmental, behavioral and vehicle factors or to evaluate the potential impact of prospective interventions.

Increased exposure, the improving economy, and behavioral factors drove increases in fatality counts between 2005 and 2016. However, improved vehicle safety design substantially countered these effects, mitigating the increases.