中国商业财产保险欺诈损失度量实证研究

Empirical study of measuring property insurance fraud loss in China

应用贝叶斯马尔科夫链蒙特卡洛(MCMC)方法估计参数,采用分段定义损失强度的损失分布法(PSD-LDA),测算了财产保险欺诈风险潜在损失TailVaR、经济资本和纯保费,并同超阈值(POT)方法、单一损失分布法等度量结果进行了比较.研究发现,财产保险欺诈损失尾部风险很大,PSD-LDA方法度量财产保险欺诈损失较为合理,为我国保险产品定价和保险欺诈风险管理决策提供理论依据.

Based on Bayesian Markov chain Monte Carlo （MCMC） method to estimate parameters, this paper proposes a loss distribution approach based on piecewise-defined severity distribution （PSD-LDA） to calculate the potential losses Tail VaR of property insurance fraud risk, its economic capital, and net premiums. We compare the results derived with other methods such as peaks over threshold （POT） method and single loss distribution approach. Empirical results show that property insurance fraud losses have fat tail risks, and the PSD-LDA model is more rational in measuring its fraud risk, which can provide a theoretical basis for the pricing of insurance products and the decisions of insurance fraud risk management.