Bayesian and credibility estimation for the chain ladder reserving method

Gisler and Wuthrich [8] describe how to calculate reserve estimates by means of Credibility and Bayesian estimators based on the development factors from different lines of business. This approach allows combining individual and collective claims information to get better estimations of the unknown reserves. In this paper we compare the reserves estimates and the mean square error of prediction from two different models: Credibility and Bayesian ones. The objective is to show how the reserve estimates of these models are similar to the classical chain ladder models under certain distributional assumptions. The work includes a way of implementing the Bayesian model using Markov Chain Monte Carlo methods with the programming tool WinBUGS [15].