The always interesting and prolific Michael Sivak has a new paper, “Homicide Rate as a Predictor of Traffic Fatality Rate” — a pretty self-explanatory title — just out in the journal Traffic Injury Prevention.
Here’s the details:
Background: In the United States, traffic fatality rates per distance driven vary greatly from state to state, with the maximum rate being 2.9 times the minimum rate. This study was designed to examine factors associated with this variability.
Method: A multiple regression was performed on the 2006 state data. The dependent variable was the fatality rate per distance driven. There were 10 independent variables.
Results: The analysis identified seven statistically significant factors: homicide rate per capita (used in the analysis as a proxy for aggression), physicians per capita, safety-belt usage rate, proportion of male drivers, proportion of drivers over 64 years of age, income per capita, and deaths caused by alcohol related liver failures per capita (a proxy for the extent of intoxicated driving). These seven factors accounted for 71 percent of the variance in the traffic fatality rates. The strongest predictor of the traffic fatality rate was the homicide rate.
Conclusion: This finding suggests that social aspects of human interaction may play an important role in traffic safety.
In other words, states with a high murder rate are filled with aggressive people, many of whom take it out on the roads.
One issue that pops up in my head is the validity of statewide comparisons for both driving and murder rates, for the primary reason that the places where traffic fatalities occur are not necessarily the same places where murders occur (I’m imagining that the places with the higher per capita murder rates are urban counties, where driving exposure would be much less, but I may be very wrong on this). As William Lucy has argued, for example, in the state of Virginia there was a much higher risk of leaving home in sparsely populated suburbs — not because of would-be murderers but because of driving risk. Conversely, “cities that often are considered dangerous, like Richmond and Norfolk, ranked 19th and 30th in the number of traffic and homicide-by-stranger deaths among the 49 metropolitan-area jurisdictions included in the ranking.”
Or maybe some states just have a lot of murders and a lot of traffic fatalities, with no causal link between the two. Or perhaps all the police are all working on murder investigations, and too busy for traffic enforcement (though I have suggested the two should not be viewed is isolation).
I’m also not sure how deaths by liver disease as a proxy for alcohol-related driving deaths — there would seem to be any number of other things to consider (e.g., health-care quality in a given place, or density — Manhattan is filled with heavy drinkers but certainly must have fewer drunk drivers per capita).