A Flexible Framework for Stochastic Claims Reserving

In this paper, a flexible framework for stochastic claims reserving is considered which includes several models proposed to date as special cases. The methodology is embedded within the generalized additive class of models (Hastie and Tibshirani [7]). The methodology is particularly useful since it allows smoothing of chain ladder development factors and estimation of tail factors automatically and easily as part of the model-fitting process, traditionally performed as an additional stage in the claims reserving process. The framework also provides estimates of reserve variability, which could prove useful in formulating and calibrating dynamic financial analysis (DFA) models