This report studies the ways in which national accident data can be analysed to compare the secondary safety of different car models. It considers the approaches of the Swedish Folksam insurance company and the British Department of Transport, the two principal approaches currently in use. The basic formulations are shown to be directly related, and to give rise to very similar model rankings. The report provides a theoretical and empirical justification for the British index, and shows that its ranking is preferable. The report also shows how to develop this index to compare directly the safety of cars of differing size. The risk of the driver being injured once a car is involved in an accident tends to an accident tends to reduce with the car's size, and it is seen that, on average, for every 100 kg added to the unladen mass of a car, the probability of its driver being injured falls by 0.045. Detailed analysis of accident data for several popular models tends to confirm that the reduction of the index is entirely the effect of mass, rather than of design. This relationship leads to an index which will compare the safety of car models independent of the effect of mass.
Abstract