论文部分内容阅读
Loss reserving has always been an important facet of insurance company risk management especially as of late where solvency concern is receiving increasing attention.For many types of general insurance products,the insurer has to set aside sufficient funds to cover losses for accidents that have already occurred,but either have not been reported or settled.The claims often observed are represented in the form of a loss triangle and the prediction process involves estimating the losses for development years that have yet to occur.In this paper,we focus on the innovative use of multivariate longitudinal data analysis in loss reserving for a general insurer with multiple lines of business.We find that if we structure the correlated loss triangles for several lines of business as a longitudinal framework,we are better able to understand the underlying dependencies among the lines and at the same capture the dynamic emergence of the losses.We employ copula techniques to model the multivariate nature of the losses for several lines of business and use random effects model to capture the time dependence that evolve over several accident years.To assess the performance of such multivariate longitudinal structure,we apply predictive approach to estimate the loss reserves and compare them with existing methods where the lines of business are usually independently estimated.This is a joint work with Priyantha H.Katuwandeniyage from University of Connecticut.