A Bayesian Interpretation of Data Trimming to Remove Excess Claims. Excess Claims and Data Trimming in the Context of Credibility Rating Procedures

摘要

The effect of excess or catastrophic claims is well recognized in insurance. For example, in experience rating it is customary to truncate the data to minimize the effect of such outliers; Gisler has recently proposed a credibility formula using such data trimming. This paper develops a model of the excess claims process and finds the exact Bayesian forecast. The resulting forecast form is approximately at data trim, thus justifying the simpler, heuristic approach. (Author)