French insurance companies usually classify their agents according to their results by branch and to this effect they calculate their respective claim ratios.

As regards motor insurance many agents have argued against this practice as they believed it was unfair to be adversely classified when they had the misfortune to incur a large claim.

In this article, the merits of various statistical systems for attenuating the effects of large claims on the portfolio are considered in order to come to the most equitable solution possible, i.e., to ascertain the best method and system for attenuating the effects of large claims. The conclusion is that the excess of loss technique is best suited to this effect.

To arrive at this conclusion the author of this article relates in detail the various stages of the study, the various systems envisaged and the tests which have allowed to select the most appropriate system. The main results obtained are also given together with their respective drawbacks.

At the end of the article the author mentions the still fairly limited bibliography which deals with this question.

Dans le but d'améliorer les résultats, les compagnies d'assurance ont l'habitude en France de publier un classement de leurs Agents Généraux basé sur le ratio S/P (sinistres/primes) et de tenir compte de ce classement pour l'attribution de certains avantages commerciaux.

Or, en assurance automobile, un malaise profond s'était emparé d'un certain nombre d'entre eux, et ils avaient mis l'accent sur un point qu'ils estimaient injuste:

“Nous ne pouvons pas être pénalisés” disaient-ils, “si un très gros sinistre, atteignant 500 ou 1000 fois la prime annuelle, frappe un de nos clients, car il s'agit alors d'un phénomène aléatoire dont nous ne sommes pas responsables”.

Telle est l'origine de la politique de l'écrêtement.