Experience rating in the classic Markov chain life insurance setting

We consider experience rating within multi-state life insurance using an empirical Bayes approach for a multivariate frailty extension of the classic Markovian setting. In particular, a search for simple methods and benchmark models is initalized. Under quadratic loss and mutually independent conjugate Gamma priors, we relate parameter estimation to multivariate negative binomial regressions and obtain a well-known credibility formula. Throughout the talk, links to practice are highlighted.