MODELING OF PANEL COUNT DATA IN CALCULATING PURE PREMIUM FOR VEHICLE INSURANCE USING ZERO-INFLATED POISSON (ZIP) DISTRIBUTION

• Main Author: FEBRIYANA (NIM: 20814032), MELISA
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/28855
• Institution: Institut Teknologi Bandung

Data panel is an observational data of a few individuals who each observed in the time series. This research uses panel data in the form of zero-inflated Poisson (ZIP) distributed claims history data. ZIP distribution is a combination of indicators and Poisson distribution which divide the observations into two models, the first model is used to determine the probability of the observation’s value is equal to zero, while the second model is used to determine the probability of nonzero value . In this research, the pure premiums prediction for next period by modeling the claims frequency history that affected by random effect. The data used in this research is vehicle insurance from January to December 2013. In this research, the model was built to predict the pure premium for next period not only had considering the claim history, but also the period with no-claim filed from each policyholders that purpose to appreciate the policyholders who have no-claim by decreasing the premium they should be paid. The pure premium prediction was obtained by choosing the interest rate at 7.5%. That premium has relative increasing towards to the base premium, ie pure premiums charged to policyholders depends on the total number of claims filed.