http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
정재욱 ( Jae Wook Chung ),여윤희 ( Yoon Hui Yeo ) 한국금융연구원 2011 금융연구 Vol.25 No.1
Using accident and illness insurance applicants` data of a domestic life insurance company for a certain time period (2009. 1~7), we empirically examine if there exists a certain relation between individual credit information and accident rate. In other words, we analyze, step by step, the marginal effects of using individual credit information as an underwriting factor, in addition to existing internal underwriting criteria of an life insurance company, on the accuracy of accident rate forecasting. As a first step, we statistically prove that, the higher the individual credit rating (in terms of high-level, medium-level, and low-level) is, the lower the accident rate is. Second, to measure and compare the effects of predicting accident rate, we primarily derive three different logistic stepwise regression models, based on the scope of data employed (that is, “model 1”-a life insurer`s internal data only; “model 2”-data compiled by the KFB, in addition to a life insurer`s internal data; “model 3”-data compiled by the KCB, in addition to a life insurer`s internal data as well as data compiled by the KFB). First, on “model 1,” all the five variables from internal data have been safely selected. Those are gender, age, risk class, current number of policies owned, number of previous claim payments. Second, on “model 2,” two variables from KFB data have been selected additionally, along with five variables from insurance company`s internal data. Those two from KFB data are loan amount in banks and time since oldest credit opened. Third, on “model 3,” four variables from KCB data have been selected additionally, along with five variables from insurance company`s internal data, while any variables from KFB data have not been selected. Those four from KCB data are secured loan amount, percent of lump sum payment used in last one year, total number of delinquency substitute payment, and income. In the mean time, the unique direction of each selected variables was maintained in three dif- ferent models, and all the selected variables turned out to be statistically significant. Finally, we compares the results of accident improvement in each model to the bottom 2%, 5%, and 10% of high-risk group respectively. Our result shows the ranking as an order of “model 3” > “model 2” > “model 1” in terms of improving the accuracy of accident rate forecasting. It implies that a life insurer may have positive effects of reducing the accident rate by using individual credit information as an underwriting factor, thus eventually improving management efficiency.