A Bayesian Monte Carlo Markov Chain Method for Loss Models and Risk Measure Assessments

Natural disasters are also known as catastrophes with low frequency but high damages. Typhoons and floods are the major catastrophes which lead to gargantuan losses in Asia. Once a disaster occurs, a broad region will be affected and this will result in huge social loss. If issuers or governments use the wrong loss models or risk measure indexes to price the related insurance products, they will get an inaccurate price and thus be insolvent to the claims. Previous researches often use a Log-Normal distribution to model a catastrophic loss. This is not appropriate since the characteristics of a loss distribution have some empirical facts, including the positive skewness and the heavy-tailed properties. Recently, some studies (McNeil and Frey, 2000; Rootzen and Tajvidi, 2000; Thuring et al., 2008) also point out that using Log-Normal distribution to model a characteristic loss is not suitable. Therefore, we build a typhoon and flood loss model with higher order moments and estimate the parameters through a Bayesian Monte Carlo Markov Chain method. According to the Kolmogorov-Smirnov test, we find that the Pareto distribution is more adaptive for modeling the loss of typhoon and flood. Further, we evaluate different kinds of risk measure indexes through simulating and numerical analysis. It gives the beacon to issuers or governments when they want to issue the insurance products about typhoon and flood loss.