Hierarchical bayesian modeling of health insurance claims

The purpose of this project is to propose a statistical model for health insurance total claim amounts classified by age group, region of residence and time horizon of the insured population under Bayesian framework. This model can be used to predict future total claim amounts and thus to facilitate premium determination. The prediction is based on the past observed information and prior beliefs about the insured population, number of claims and amount of claims. The insured population growth is modelled by a generalized exponential growth model (GEGM), which takes into account the random effects in age, region and time classifications. The number of claims for each classified group is assumed Poisson distributed and independent of the size of the individual claims. A simulation study is conducted to test the effectiveness of modelling and estimation, and Markov chain Monte Carlo (MCMC) is used for parameter estimation. Based on the predicted values, the premiums are estimated using four premium principles and two risk measures.