本研究利用Banker, Charnes and Cooper(1984)提出資料包絡分析(DEA)，以及Aigner, Lovell and Schmidt(1977)提出隨機邊界模型(SFA)，運用2003年至2006年金融控股公司之銀行14家為研究對象，針對二種方法進行成本效率之比較分析。 實證結果發現：以SFA方法平均成本效率值高於以DEA方法，舊銀行效率值皆高於新銀行，再區分群組以銀行或非銀行為主體，其銀行為主體，效率值高於非銀行，有官股介入銀行其技術效率值，DEA與SFA方法效率值呈現一致，惟配置效率、成本效率值以官股介入銀行較高，結果DEA與SFA方法效率值呈現反方向。 DEA方法結果發現，無論技術效率、配置效率、成本效率，舊銀行效率值皆高於新銀行；金融控股公司以銀行為主體效率值皆高於非銀行為主體；技術效率值無官股介入優於官股介入，而配置效率、成本效率有官股介入優於無官股介入。 This research carries out a comparative analysis for the cost efficiency of 14 holding company in Taiwan from 2003 to 2006 using the Data envelope analysis (DEA) proposed by Banker, Charnes and Cooper (1984) as well as the stochastic boundary model (SFA) proposed by Aigner, Lovell and Schmidt's (1977). The empirical result showed that the average cost efficiency value from the SFA method is higher than that from the DEA method and the old bank efficiency value is higher than the new bank. Dividing banks into two groups (bank and the non-bank), we find that efficiency value of bank is higher than the non-bank. Our finding also support that banks with the official stock, gained from the DEA model, present a consistent technology efficiency value with the SFA method while the allocation efficiency, the cost efficiency value present a reversed direction. Results from DEA also show that old bank efficiency value is higher than the new bank when it comes to technical efficiency, allocation efficiency, and cost efficiency; The financial holding company's efficiency value is higher than the non-bank when for the bank as the main body; Banks without official stock involvement show a better technical efficiency value than those with official stock while banks with official stock represent a better allocation efficiency and cost efficiency compared to those without official stock involvement.
