During the last years, the interest in rating the interior safety of different cars and constructions have increased. Several methods have been used, based on either laboratory crash tests or on real life accident data. Several limitations are related to both methods, and there is no universally accepted method to rate cars in relation to each other.

One technique that has been proved to be powerful in car safety rating is the paired comparison method proposed by Evans. In this method, the most important exposure parameter, accident severity, can be controlled. The method has, however, some fundamental drawback and that is the fact that only a part of the accident population can be used. Single accidents, and accidents with trucks, buses etc must be omitted.

In this presentation a technique to expand the accident data is showed. It is also shown how aggressiveness can affect the results in safety rating, together with some other results of using paired comparisons.