Institute of Transportation, MOTC

Ming-Jyh Wang, 2008.01
Institute of Traffic and Transportation National Chiao Tung University

　　Car owners purchase appropriate automobile insurance policies (AIP) to provide coverage for property damages and personal injuries incurred by traffic accidents. Physical damage coverage, the most expensive policy, is the major source of revenues for non-life insurance companies in most countries today. However, new car owners are likely to purchase physical damage coverage in the first few years and then downgrade their insurance by either purchasing reduced physical damage coverage or not even buying any physical damage coverage in the subsequent years. As such, premium revenues for non-life insurance companies will be substantially reduced. The study is motivated by the importance of developing a modeling framework to gain insights into the insured’s choice for AIP.

 　　The research develops a model system that consists of two components: the first component is the decision to select different types of physical damage coverage; the second component is the choice of non-physical damage coverage involving third party liability as a basic protection with additional coverage. This study focuses on the first component and, particularly, explores repeated choices of different types of physical damage coverage. A discrete choice modeling framework including the choice of physical damage coverage type and the number of consecutive years that the insured has purchased the same type of coverage is further developed. Various discrete choice models including multinomial logit, nested logit, and paired combinatorial logit are attempted.

　　The proposed modeling framework is empirically tested using a panel data provided by a non-life insurance company in Taiwan. The results indicate that the repeated choices of physical damage coverage AIP are influenced by age of the driver, vehicle make, and engine capacity. The nested logit model statistically rejected the multinomial logit model, which demonstrates the statistical and structural superiority of the nested logit model in analyzing the insured’s repeated choices. Although the paired combinatorial logit model is more flexible than the multinomial logit or nested logit model, it is found that estimation of such model becomes very difficult when the number of alternatives gets large. The model framework developed in the study has improved our understanding of the repeated choices of AIP, and the estimation results have provided valuable implications for the insurer to modify existing automobile insurance policies or to develop marketing strategies so as to enhance the insured intension to repurchase the physical damage coverage AIP.