Research On Measurement Models And Methods Of Non-life Reserve Risk

On January 1,2016,after 5 years,the Solvency II taking charge of insurance industry throught Europe took effect,at the same time China's insurance industry has been into the transitional period of "second generation of solvency" for a full year.Risk oriented insurance regulatory system requires a comprehensive and detailed measurement of insurance risk-Non-life reserve risk,as an important part of insurance risk,its accurate measure is of great significance for insurance company to strengthen the solvency and improve the level of risk management.Insurance regulators and insurance companies pay more and more attention to reserve risk measurement.However,existing research still forcus on reserve estimates and pay less attention on reserve risk measurement issues.This paper follow hot issues of the insurance regulatory system reform and do some research on measurement models and methods of non-life reserve risk.Specifically,this paper solve four problems:(1)the measurement of non-life one year reserve risk;(2)the measurement of reserve risk with residuals dependable structure;(3)the measurements of reserve risk in case of correlated claims data;(4)the influence of model uncertainty on reserve risk.Firstly,we clarify the meaning of non-life one year reserve risk through an intuitive example,introduce the concept of claims development result,and discuss simulation approach to obtain the predictive distribution of CDR based on the bayesian lognormal model.Secondly,in order to avoid reserve risk measurement errors caused by residual correlation,we use multiple hypothesis testing and the false discovery rate control process to solve this problem.Residual correlation is caused by data correlation,so on the basis of independent poisson model,we add the relevant random effects and develop the conditional autogressive poisson model which can describe the correlation in the claims data,and then estimate the model by bayesian method.Finally,we studies the influence of model uncertainty on the non-life reserve risk.Loglogistic growth curve model and Weibull growth curve model are common in claims reserving.We use bayesian model average to average the results of two models by their posterior model probabilities.In the end,we not only combine results of two models,but also obtain the reserve risk measurement under the perspective of model uncertainty.These work is expected to enrich the research of non-life reserve risk,further expand the field of claims reserving,and provide theoretical support and practical reference for the regulators and the insurance company.