A Re-assessment of Road Accident Data-Analysis Policy: Applying Theory from Involuntary, High-Consequence, Low-Probability Events like Nuclear Power Plant Meltdowns to Voluntary, Low-Consequence, High-Probability Events like Traffic Accidents

More people are injured and die annually from motor vehicle accidents than from less commonly occurring events like nuclear power plant meltdowns. Unlike motor vehicle accidents, however, incidents at nuclear power plants and in commercial aviation are thoroughly scrutinized and analyzed, and the information fed back to operators, to determine how such disasters can be prevented. Roughly parallel systems should be in place in the traffic safety system, where both the professional driver and the average driver need to be more aware of road hazards and the decisions they should make to avoid them. This report examines the literature on involuntary, high-consequence, low-probability (IHL) events like nuclear power plant meltdowns to determine what can be applied to the problem of voluntary, low-consequence, high-probability (VLH) events like motor vehicle accidents. It examines five closely related literatures on IHL events: ?normal? accident theory, system reliability theory, high reliable organizations theory, complexity and tight coupling theory, and a theory of feedback and learning (band-of-accident theory). Based on these theories, the researchers developed and tested a series of propositions to explain traffic injuries and fatalities. They carried out logistic regression analyses, examining driving conditions and decisions drivers make as factors that can lead to fatalities and injuries, then characterized and described the models, found in state crash data publications, that traffic safety officials use for understanding fatalities and injuries. These models were compared with the instructional material that is used in state driving educational manuals in order to investigate how to improve the collection and use of road traffic safety data based on analysis of the existing data and its use. Through the investigation, the researchers found that the most significant condition leading to a fatality or an injury was driving on a rural road, and the most significant decision was choosing not to use a seat belt. How factors combine to cause fatalities and injuries was also examined. For example, a combination of risky driver behavior at stop and yield signs was significantly related to both fatalities and injuries. Similarly, a combination of illegal speed and alcohol use was significantly related to both fatalities and injuries. Overall, the fatality model explained about 2 percent of the variance and the injury model explained about 12 percent of the variance. In the investigation of state driving instruction manuals, the researchers discovered that about one-third of the pages in a typical manual were devoted to factors that traffic safety officials consider to be the main reasons for fatalities and injuries. Although the current data collected in Minnesota, when analyzed, provided a number of powerful predictors of fatalities and injuries relating to the conditions a driver faces and the actions that drivers take, overall the data?s ability to explain crash severity could be better. Improved theory can inform data collection and result in more powerful predictive models that could be used in programs to educate drivers.