Differences in the Trip Attributes of Drivers with High and Low Accident Rates

In a large survey project to create a microdata base on vehicle ownership and usage in Michigan, 7581 respondents were grouped by a multivariate procedure into meaningful population subgroups with differing accident rates. Age, sex and residence location were the most effective of several biographical factors in distinguishing groups with mean rates ranging from 3.8 to 15.4 accidents per million miles of driving. These groups were analyzed for differences in driving trip attributes: light condition, road type, vehicle class, passenger load, trip purpose and perceived public transportation substitutability. It was found that of the six, passenger load and trip purpose had the least relevance to accident rate. Vehicle size seemed a promising factor, but despite high small car exposure in (especially youthful) high accident rate groups, disaggregate data showed the subgroup using small cars to have lower rates. It was concluded that no simple combination of qualitative exposure measures, even when collected with the most detailed microdata techniques, promises to predict accident rate with high success, and that some conventional assumptions about the relative hazardousness of certain types of driving should be re-evaluated.