Estimating vehicle types effects on auto insurance premiums in São Paulo City: a GAMLSS Approach

Abstract

This paper aims to identify how the different types of vehicle models impact insurance premiums for robbery and theft coverage in the city of São Paulo and considers as covariates the years of vehicle use, the risk classification, gender, and the driver’s age. To achieve this aim, we classify car models by using the database provided by the São Paulo Public Security Department; this database includes all the police reports registered for robbery and theft of vehicles in the city. We obtain the data for estimating the variables by using AUTOSEG, made available by SUSEP on its website, and use GAMLSS to estimate the frequency and severity parameters for the negative binomial and gamma distributions, respectively. As a result, unlike what was expected, the risk classifications were not significant for the average frequency of vehicle robbery and thefts, and it was possible to infer that vehicles classified as being at greater risk have a lower severity; that is, these vehicles are classified as popular.