Simple models of fatality trends revisited seven years later

The author's original paper presented simple linear models of fatalities based on data through 1982 and appeared in this journal in 1984. The earlier paper reported good fits through 22 years of fatality counts using population and employment data, after adjusting for the short-term effect of the oil shortage (in 1974) and the long-term effect of the 55 mph speed limit (beginning in 1974). Attempts (reported here) to refit the model through seven additional years were less successful. It may be that the relationships originally reported have changed, that important new factors have been introduced, or that the original fits were partly luck. The recent emphasis on behavioral solutions to safety problems (including programs to decrease alcohol use and increase safety belt use) is one important change, so the effect of the benefits of these programs on the model fits is explored. The original model began to overpredict fatalities just as safety benefits were beginning to accrue from increased safety belt use and decreased driver-alcohol involvement, and both the model overprediction and the benefits of these programs increased over the seven years of data added to the original model. Thus, previous government estimates of the number of lives saved by these behavioral programs are supported by the observed change in the statistical relationship between fatality counts and economic factors.