On the Actuarial Simulation of the General Pareto Distribution of Catastrophe Loss

摘要

Precisely estimation of Cat-loss is not only the foundation of risk analysis, but also the premise of product design and the practice of insurance compensation. The law of large numbers generally assumes risks have normal distributions, which reduces the accuracy of damage assessment and influences the pricing of cat-insurance due to negligence of the extreme value at both sides of the distribution. Data of more than 100 million Yuan of earthquake disaster loss from 1969-2014 presents the characteristics of right skewed peak, excess kurtosis and heavy tail. Furthermore, the comparison of QQ plot, parameter estimation as well as test of model parameters between N-distribution, E-distribution, W-distribution and P-distribution shows that the generalized Pareto distribution fits the earthquake loss perfectly, and significantly improves the estimate precision.