MODELLING CLAIMS SEVERITY USING PEAK OVER THRESHOLD METHOD (POT) IN PROPERTY INSURANCE

• Main Author: Marisa
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/39111
• Institution: Institut Teknologi Bandung

Insurance companies, as providers of insurance products, need to do a risk management control in order to avoid the error in measuring the risk. The substantial risk or loss faced by insurance policyholders is referred to as a claims severity or broad claims. One of the most commonly used risk gauges is the Value at Risk (VaR) by assuming the regular occurrence. The severity of claims can be extensive when the frequency is minimum or can be called as extreme events that give a huge impact. Therefore, in this thesis, the distribution model of extreme value theory is used to analyze extreme event data with a Peak Over Threshold (POT) approach which generates the distribution Generalized Pareto. In specifying an extreme value using the POT method, will assigned some candidates of a limit of threshold (u) from the result of the mean excess function plot. The threshold candidate will be tested using the Kolmogorov Smirnov test to get a threshold value that matches with the Generalized Pareto distribution. The matching value will be used in performing calculations of the value at risk that will be used later in calculating a reserved claim. In applying this method, will use broad data claims on property insurance that occurred in the year 2010-2016.