抄録
This paper develops a multinomial logit (MNL) model to identify important factors affecting the selection of automobile insurance policies (AIPs), which are characterized with high similarity among the bundled alternatives. Various nested logit (NL) model specifications are attempted to elucidate the choice behaviors for these highly similar auto insurance products. Some 3,000 voluntary automobile insurance records, randomly drawn from a non-life insurance company in Taiwan, are used to conduct the empirical study. The estimation result of the preferred MNL model shows that vehicle usage, engine capacity, imported/domestic vehicle, and number of claims are significant factors influencing one's selection of AIPs. Owners of newer vehicles, of larger engine capacities, of imported vehicles, and of more insurance claims tend to purchase wider coverage of insurance packages. The NL models confirm that the bundled AIPs are similar and highly correlated products. The empirical results also provide evidence of adverse selection in automobile insurance market -- the wider the insurance coverage, the less motivation the insured would reduce the number of claims (or equivalently, prevent the accidents).
