Incorporating expert opinion into a stochastic model for the chain-ladder technique

To date, much effort has been directed towards the development of stochastic models that are analogous to traditional deterministic methods. In practice, however, the traditional models are often altered to incorporate expert opinion. This paper considers the use of Bayesian models to allow practitioners to apply their judgement to the development factors in the chain-ladder technique. The implementation uses MCMC methods within winBUGS. In this way, it is possible to use stochastic models to obtain predictive distributions of reserves in a much wider range of situations.