PRICING OF VOLUNTARY MOTOR INSURANCE USING GENERALIZED LINEAR MODELS

Abstract

This research aimed to create a Generalized Linear Model (GLMs) for determining Voluntary Motor Insurance Premiums using Voluntary Motor Insurance data that began coverage in 2022, specifically Motor Insurance Class 1 that provide coverage for private passenger cars with no more than 7 people (Car Code 110) from the Office of Insurance Commission (OIC). The researcher considered 14 quantitative and qualitative independent variables, such as vehicle group, vehicle age, capacity engine, and vehicle shape variables, grouped by make and model according to ISO 3833:1977 standards. In preparing the data to build the model, the researcher eliminated some variables with high correlation to avoid the problem of multicollinearity. From the results of the Generalized Linear Model study under different distributions of dependent variables, in the case where the premiums are Gamma Distribution with a Logarithmic Link Function gave the best prediction results. Considering the lowest AIC, BIC and Deviance value and the highest Concordance Index value, which reflects the appropriateness of the model with the data with a right-skewed distribution. Although the analysis in this research found that all independent variables were statistically significant (p-value < 0.0001) due to the big data (n = 700,858), which is a common characteristic in using big data. However, the practical determination of premiums does not only consider the level of statistical significance, but also Effect Size of the variable's impact, such as the Regression Coefficient and the Relativity.