台灣汽車保險風險分類之實證分析

本文主要檢驗汽車保險現行採用及額外之風險因子是否為有效的風險分類？以台灣全體產險公司自用小客車三種車體損失險保單為研究對象，建立不同風險因素組合的預測模型，估計其預測變數係數，再透過估計之預測模型進行樣本內與樣本外的預測，以確認預測模型的準確性。本文的實證結果顯示，在現行採用的訂價風險因素，甲式保單顯示理賠係數是相對重要的風險分類因素；乙式保單則是理賠係數與從車因素；丙式保單則是被保險人的年齡。其次，在額外的風險分類分析，本文發現甲式及乙式保單均顯示前一保單年度的理賠經驗是重要的風險分類資訊，此意謂前三年的理賠係數應給予前一年之理賠經驗給予相對較高的權重，如此保費的訂定或許可以達到更公平。由於目前在實務上，三種保單的定價因素都相同，尤其在費率自由化之後，本文的實證結果，可建議保險公司進行實際的風險分類規劃時，針對不同保單可以考量不同的核保或訂價因素之參考依據，進行有效風險分類差異化的費率創新。 This study examines whether the characteristics of risk classification are effective by developing different models to predict claim occurrences. We use three unique types (Form A, Form B, and Form C) of coverage for vehicle damage insurance data in Taiwan. We first identify the basic characteristics of policyholders and vehicles which are currently applied by insurers for property damage to vehicles. We further investigate some supplementary information, which requires small information cost for the insurers to obtain from the insureds. Our evidence shows that, for the crucial risk classification based on the current official rating formula, claim coefficient is a relatively critical factor for Form A, while claim coefficient and car age are important for Form B, and policyholder age is important for Form C. We also find that some additional information such as the claim record in the previous policy year provides useful information for risk classification. Our findings imply that the official rating formula may consider various risk factors among different types of coverage for damage to vehicles. In fact, insurers could improve claim prediction by adding unused information such as the claim record in the previous policy year or other factors. Thus, the deregulation of automobile insurance rating can benefit from an efficient risk classification.