| 摘要: | 摘 要 近年來隨著資訊科技的進步與統計方法的改良，預測技術已不斷創新與改進，對預測準確度的要求也愈來愈高，準確的預測將可提高管理者決策與因應之效率。 本研究利用Box-Jenkins的ARIMA時間數列分析模型，加入影響保險費收繳率的各種相關因素與變項，分別建立單變量模式、介入分析模式與二元及多元轉換函數模式等三種模型，透過實証分析檢定來探討各種模型之預測能力，期能建立最適之全民健保保險費收繳率預測模型。本研究資料採用全民健康保險八十四年三月至九十年六月每月保險費之收繳率以SCA統計套裝軟體進行分析，實証結果在單變量ARIMA模型部分，依分析結果全民健保保費收繳率資料符合ARIMA(0,1,1)的模型，即全民健保保費收繳率受前一期殘差項之影響，因此在預估保費收繳率過程中，應考量前一期殘差項，以提高保費收繳率的預測效能。另外利用政策面各種因素之變項代入介入模型，依結果顯示全民健保保費收繳率除受前一期殘差項影響外，亦受第一次修訂（調降）平均眷口數的影響，故應考量平均眷口數調降之因素。再將經濟面、社會面變數代入轉換函數模型來探討，結果顯示保費收繳率除受前一期殘差項影響外，亦受前五期失業率的正向影響。 依本研究所建立之三種預測模型，經以RMSE及MAPE來作比較，其MAPE均小於10﹪，顯示各模型均具有高精確預測能力，其中以轉換函數模型較佳。 依本研究實証分析，對全民健保之政策建議如下： 一、為提高保險費收繳率，對欠繳保費者暫時拒絕給付是有效的措施，未來健保IC卡在設計上宜保留此機制。 二、為達到全民健保危險分擔與量能付費的精神，對弱勢民眾之保費負擔應加以重視與考量。 三、政府施政時，政策的公平性與採行數據之正確性應加以注意（如健保平均眷口數），以獲得民眾之認同，始能提高保費收繳率。 關鍵字：全民健康保險 National Health Insurance (NHI) 保費收繳率 Payment-Rate 平均眷口數 Average Number of Dependent 眷口數上限 Ceiling Number 失業率 Rate of Unemployment 經濟成長率 Rate of Economic Improvement 健保滿意度 Satisfied Rate of NHI 時間數列 Time-Series ARIMA模型 Model of ARIMA; Abstract As following the progress of computer technology and the improvement of statistic methods, the forecasting techniques have been innovated and improved，in order to help manager making decisions effectively, the demand of higher degree of accuracy of prediction is more and more serious. This research is set up the models to forecast the payment-rate of premium for the National health Insurance ( NHI) system, in addition to applying Box-Jenkins’ ARIMA Model of Time-Series, by considering those related factors and variates of the payment-rate of premium of NHI, Uni-variate model、Intervention Model、Two-variate Model and Mutiple-variate Transfer Function Model are also formulated. Through practical check and numerical analysis, we can identify the accuracy of prediction of each model. Our data are collected from each month’s payment-rate of premium of NHI from March 1995 to June 2001, and SCA software is used for analysis. All the predicted outputs of each model are as follows: １、According to the uni-variate ARIMA model, the payment-rate of NHI premium fit into ARIMA(0,1,1)，showing it is affected by the previous period of residual. To get more accuracy of forecasting the payment-rate of NHI premium, we should better think about the previous residual. ２、We should consider the influence of previous residual and the first amendment of the ACT in case of using the |Intervention Model to sep up the estimating model for payment-rate of NHI premium . ３、When the estimating mode is established by bivariate Transfer Function Model with economic and social factors, the result implies that payment-rate of NHI premium is affected by the previous residual and five periods’ rate of unemployment. The values of MAPE of these three models are all less than 10%, and this proves that the three models are good enough to fit the data. However, by comparing the RMSE and MAPE in each model did we find those values in Transfer Function Model proves to be the optimum one to generate more accurate estimating. Based on our research, the followings are recommended about the policy of the payment-rate of premium of NHI： １、For the purpose of raising the payment-rate of NHI premium, it is an effective way to reject the benefit before the premium is paid out. Therefore, it is better to keep this mechanism for the IC card system in the future. ２、 To reach the spirits of risk-sharing and the compatibility of which the premium- burden is based on one’s financial ability, the NHI system should have the responsibility to take care of those poor insured. ３、 The government should pay more attention on the fairness of the policies and the accuracy of the data such as the average number |
