The search for high risk groups requires the computation of accident data and risk exposure data, often expressed as kilometres travelled. The aim of this paper is to analyze, from a car diary survey, drivers' behaviour with regard to distance covered and driving speed, in order to get basic data for accident risk assessment. The survey was carried out by INSEE (Institut National de la Statistique et des Etudes Economiques) in 1981-1982. A car diary was provided for 2,615 cars, and the driver(s) had to fill it in during a week. He (she) noted the characteristics of each trip (reason, hour, length, features of the driver and the passengers, etc). This article first presents the survey methodology, and then describes the data analysis method used in order to establish a typology of drivers' exposure to risk. Homogeneous classes of risk exposure, that is kilometres travelled, have been identified from the links between different parameters in a multicriteria analysis. The process in its initial stage consisted in a correspondence analysis. Then, a hierarchical classification method is applied directly to the factors from the correspondence analysis. Categories are obtained which are markedly differentiated by the most prominent contributory factors which are essentially the trip purpose (home-to-work trip, professional trip, shopping, leisure, etc) and the social characteristics of drivers (profession, age, sex, type of the car used). In each category, car use is very different, and may be a basis for accident analysis.
Abstract