PURE PREMIUM PRICING FOR AUTO INSURANCE USING GENERALIZED LINEAR MODELS (GLM)

Abstract

Pricing pure premium for auto insurance usually based on risk of the auto. There are a lot of methods used to pricing auto insurance pure premium. One of them is Generalized Linear Models that will be discussed here. The used of Generalized Linear Models for pricing auto insurance pure premium considers the auto risk given the observable characteristics of the auto. This model combines the conditional expectation of the claim frequency with the expected claim cost. The case study here uses auto insurance data is not real. The data also contain the characteristics of the auto. The parameter estimation method is used here is maximum likelihood estimation. The random variable distribution of variable Y is a particular family of distribution, namely the exponential family. From the result of checking distribution, the distribution of claim frequency is Poisson distribution and claim cost is Gamma distribution. Both claim frequency and claim cost modelling use natural logarithm link function. In this case, the characteristics of auto that influence the pure premium are type of vehicle, age of insured, and the profession of insured. Keywords: pure premium, claim frequency, claim cost, Poisson distribution, Gamma distribution, maximum likelihood estimation method, Generalized Linear Models.