Data has been collected from a structured sample of just over 18500 drivers using a postal questionnaire, to determine the relationship between the accident liability of these drivers, and factors such as age, driving experience, sex, socio-economic group (SEG), and annual mileage by type of road. Accident liability is defined as the expected number of accident involvements per year. Generalised linear modelling techniques have been used to develop a statistical model which will predict the accident liability for an individual driver as a function of relevant explanatory variables. The model suggests that accident liability is dependent mainly on exposure (total annual mileage), the driver's age and his or her driving experience measured as the number of years since passing the test. Predicted accident frequencies are not directly proportional to annual mileage, and are dependent on the proportion of driving done in the dark and on different types of road (built-up, rural and motorway). Accident liability falls with increasing age and driving experience. The form of the age and experience relation means that the proportional change inliability with increasing age or experience is larger for younger drivers than for older drivers; this is particularly marked for experience - i.e., the learning curve is steep. Men have higher accident liabilities than women at all ages. The differences between socio-economic groups are relatively small.
Abstract