Empirical Testing of Classification Relatives

There has been considerable discussion regarding the theoretical basis for insurance classifications and the calculation of classification rates. Rather than focusing on those issues, this paper presents some tests of the relative accuracy of competing rating methodologies. These tests are empirical in nature and involve comparing among classes the cost differences that actually have emerged with estimates of those differences using an alternative classification ratemaking methodology. In addition to tests of classification relativities, this paper also includes a test of the differences in excess loss experience among classes. These tests have been applied in practice and this paper includes examples of the corresponding calculations.