本研究採用Fried et al.(2002)所提出之三階段資料包絡分析法(data envelopment analysis, DEA)，衡量台灣地區19家產險公司之經營效率。三階段DEA法可改善傳統DEA法無法衡量隨機干擾因素之缺失，且可同時考慮環境變數與隨機干擾因素之影響。第一階段先以原始產出投入變數資料進行DEA分析，估算出各項效率值。第二階段以隨機邊界法(Stochastic Frontier Analysis, SFA)迴歸模型分析環境變數對投入差額變數之影響，調整投入變數資料。第三階段再以調整後之投入變數資料與原始產出變數資料，重新利用DEA分析產險公司之經營效率。 研究結果發現，調整環境變數與隨機干擾因素後，第三階段之各項效率值與第一階段之效率值有很大之差異，且第三階段結果發現造成總技術無效率之主要原因皆為規模無效率，而非第一階段來自純技術無效率或規模無效率。另外，調整後大部份產險公司皆為IRS，亦即此產險公司生產規模未處在最適規模下，應擴大生產規模，以達規模效率。在考慮環境變數與隨機干擾因素後，可真正反映各產險公司之經營效率，且可供給產險公司未來經營策略之參考。 This paper employs three-stage data envelopment analysis (DEA) of Fried et al. (2002) to measure the efficiency of 19 domestic non-life insurance companies in Taiwan during the period from 2003 to 2007. The three-stage DEA can be improved the defects of the traditional DEA that can’t evaluate the statistical noise, and it can decompose environmental effects and statistical noise effects from efficiency. In the first stage, a traditional DEA is applied to obtain initial efficiency value. In the second stage, a Stochastic Frontier Analysis (SFA) model is used to regress first stage’s slacks on a set of environment variables and adjust the input data in the second stage. In the third stage, DEA is used again to estimate efficiency value after employing the adjusted input variables. The empirical results show that the third stage efficiency is different from the first stage efficiency, after adjustment of environmental parameter and statistical noise. In the third stage, the reason of technical inefficiency comes from scale inefficiency, not pure technical inefficiency, And most of the non-life insurance companies are in the IRS condition. After considering environmental and statistical noise effects could estimate real efficiency of domestic non-life insurance companies. Besides, the conclusions of the empirical results may provide them with some information to make decision in the future.
