Modelling Different Types of Bundled Automobile Insurance Choice Behaviour: The Case of Taiwan*

Automobile insurance policies (AIPs) are typically offered by the insurers in different bundled formats and some of which may be highly similar. This study proposes a two-component modelling system, which consists of choice of physical damage coverage and choice of non-physical damage coverage. Both multinomial logit (MNL) and paired combinatorial logit (PCL) models are attempted to explain the choice behaviours of AIP alternatives. The proposed models are tested with a large data set drawn from a Taiwanese non-life insurance company. It is found that the PCL model is structurally superior to the MNL model in analysing the choice of physical damage bundled AIP alternatives. In the context of non-physical damage coverage choice, however, the assumptions from the MNL model hold, suggesting that the use of PCL model is not required. Based on our estimation results, the insurance providers can develop marketing strategies to refine their existing AIPs or to develop new AIPs to better serve their customers.