A Practical Model for Bayesian Trend Credibility

This paper demonstrates a Bayesian approach to the estimation of trend. This model allows for credibility weighting of observed results against other industry data or economic indices. Major features of the approach include (1) explicit recognition of the impact of claim counts, policy limits, line of business, and layering on trend estimates; (2) the incorporation of relevant severity distributions to the line of business reviewed; (3) as an output, the Bayesian posterior weights applicable to the prior trends modeled. The paper also applies the model to realistic scenarios in which the assumptions required for standard regression-based techniques are violated. Address for Correspondence: jranney12@yahoo.com