Abstract

In this paper we apply actuarial models to detailed, micro-level automobile insurance records. As we know, third party insurance is an important major for both policyholders and insurance companies. We modeling claim frequency, type and severity of third party insurance claims with incorporate different individual and vehicle risk factors such as vehicle age, vehicle usage, vehicle capacity and no of claim discount. This allows the actuary to differentiate prices based on policyholder characteristics. In addition, by using of various risk measures, including value at risk and tail value at risk to predict the insurance company capital requirement. Finally, we assessed the effects of dependence structure on these measures by using copula models. The result shows that the copula effect is increases with the percentile.   

 

Keywords: Third party liability insurance, Risk factors, Copula, Risk measures, Capital requirement,

Mathematics Subject Classification [2020]:  91G70, 62H05 

  1. Introduction

Insurance company that works with health and automobile insurance as a short-term policy, typically have massive amounts of in-company data. So, in this paper, we modeling types of losses in third-party liability insurance data. In fact, we use several characteristics to help explain and predict automobile accident frequency, type and severity. For this purpose, we consider 2014-2020 data consisting of policy and claims in three parts of ‘third party injury’, ‘third party property’ and ’personal injury protection’ payment files from a major Iran insurance company that are consist of 374,184 records. Our database consists of some risk factors such as vehicle age, vehicle capacity, vehicle usage, driver gender and No Claim Discount (NCD).

We consider: (1) losses for injury to a third party, (2) losses for property damage to a party other than the insured and (3) loss for injury to driver. It is not uncommon to have more than one type of loss incurred in each accident. So, occasionally perhaps, we have two or three kinds of loss for one event. We use statistical models to summarize micro-level data that subsequently need to be interpreted properly for financial decision-making.