ABSTRACT Methods of standardization of ‘raw’ data from official accident -reports that were used in the study of car accidents in the West German state of Nordrhein- Westfalen in 1980 are described.

The larger study aimed to identify technical car parameters which contributed to accidents and required the identification of groups of drivers who have an average risk of causing accidents.

The analysis involved the collection of additional data, that would allow the standardization of ‘raw’ driver-age accident frequency and car-model accident frequency data to account for the relative ‘exposure’ or degree of participation of driver and car populations in normal traffic. In the case of the car models, this involved the collection of data on the abundance of car models, and a calculation of the average annual mileage for different models. For drivers, the age distribution in normal traffic was approximated by the age distribution of drivers who were involved in, but not the cause of accidents, according to the official accident reports. The exposurecorrected age distribution of drivers who caused accidents reveals a much higher involvement of older drivers (above 54 years) in accidents than would be expected from their contribution to the total driver population in normal traffic. Using this approach for different accident types clearly shows the extremly high involvement of 1) the 18-22 year age group in causing longitudinal and single-car type accidents and 2) drivers above 55 years of age in accidents occuring at intersections.

The distribution showing the accident involvement of specific car models when exposure to traffic is taken into account is significantly different from the raw accident frequency distributions and shows that many car tyPes with low accident frequencies have a high probability of accident involvement.

Speed data were also collected in a roadside study to examine the relationship between driver age, car type and choice of speed. Data for separate observation days showed the existence of significantly different speed distributions thereby precluding the analysis of the data for all days as a single group. However, by calculating a daily 86-percentile speed and the relative numbers of each age group exceeding it, the data could be combined. Drivers aged 25 - 46 years chose the highest speeds and speeds decreased for higher ages.