PREDICTING THE CLAIM AMOUNT OF A MOTOR VEHICLE INSURANCE USING GENERALIZED LINEAR MODEL (GLM)

• Main Author: JUSMAN , ABDILLAH
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/14588
• Institution: Institut Teknologi Bandung

In this nal project report(laporan tugas akhir), the risk of estimating the outstanding claim liability using Chain Ladder method is analyzed. The data used as a case study are settled and paid claims data from an reinsurance company XYZ, for accident year 2004 to accident year 2009. The Monte Carlo simulation is used to generate the random variable of aggregate claim amounts (severity) in each cell of the incremental run-o triangle. The probability model for the aggregate claim amounts in each cell of the incremental run-o triangle is obtained from the probability model of the individual claim amount and the assumed probability model of ultimate aggregate claim numbers. From the Anderson-Darling and Cramer-von Mises tests, with = 5% signicance level, it is found that the individual claim amount follows a lognormal distribution with parameters = 15; 637 and = 2291. The ultimate aggregate claim numbers is assumed to follow zero-truncated negative binomial distribution with parameters r = 0:272961 and = 3690:54569. Using the probability models of the individual claim amount and the aggregate claim numbers, the deviance of the Chain Ladder estimate of the outstanding claims liability and the estimate of the outstanding claim liability using a probabilistic model is analyzed. For the data used as a case study in this finnal project report, it is found that the Chain Ladder method has the risk of estimating the outstanding claim liability too low(under-reserved) compare to the estimate obtained using a probabilistic model.