Further comments on the paper ``Setting a bonus-malus scale in the presence of other rating factors'' by Taylor (Q2356643)

This paper extends a previous analysis of taiwanese auto insurance data. The precedent study demonstrated that, due to the low variability of claim frequencies across mileage classes and the large concentration of policyholders in the lowest class in the bonus-malus scale independently of mileage, taking into account mileage differentials would hardly modify suggested bonus-malus levels, so that mileage information need not be used to correct discounts and penalties in the country. In the present study, an attempt is made to create more heterogeneous risk classes by adding to mileage, as a priori classification variables, driver age, car age, engine capacity, and territory. The spread of average claim frequencies increased. However, the variability of claim frequencies within each class did not decrease as expected. Consequently, the high overlap between risk classes is still present. Thus, the previous conclusion about bonus-malus levels would not be affected by taking other variables into account. Comment to [\textit{G. Taylor}, Astin Bull. 27, 319--327 (1997)]