資料包絡分析法 (DEA)的有用性完全視其計算決策單位相對效率的能力而定，然而，當變數數目愈多時，該分析的敏銳度愈低。因此，使用 DEA模型昀重要的步驟之一就是篩選投入與產出變數。本文以 2005~2009年國內 27家銀行作為研究樣本，使用 Wagner and Shimshak(2007)所提出逐步選擇變數的方法，以篩選確認 DEA模型應該使用的變數，並探討逐步過程所得到的管理意涵；再者，本文檢視以逐步選取變數方法所得到的效率值與銀行獲利能力是否存在相關性，以及其是否能解釋銀行帳面價值與權益市場價值差異之原因，進而驗證逐步選取變數方法的有效性。實證結果發現，經過逐步過程篩選變數後的核心模型，純技術效率與獲利能力存在高度相關；再者，核心模型之純技術效率亦能解釋銀行權益市場價值與帳面價值差異之原因。

The Effectiveness of Data Envelopment Analysis (DEA) depends on the ability of measure decision making units’ efficiency. However, the greater the numbers of input and output variables in a DEA are, the less discernable is the analysis. Therefore, one of the most important steps in the application of modeling using DEA is the selection of input and output variables. Adopting the sample data of 27 banks from 2005-2009, we use the stepwise method of variables selection proposed by Wagner and Shimshak(2007) to decide the variables of DEA and discuss the new managerial insights gained from the stepwise procedure. In addition, we examine the relation between efficiency gained from the stepwise procedure and earning ability, and account for whether the efficiency is sufficient to bridge the gap between bank’s market and book values. Experimental results indicate that it is highly relative between the technical efficiency of core model gained from the stepwise procedure and earning ability. Furthermore, it can also explain the gap between market and book values.