在本研究中，以跟政府政策與民眾生活息息相關的八大公股行庫為研究對象，採用資料包絡分析法探討其經營績效，樣本期間為2009至2011年，三個年度共計24個決策單位（Decision Making Unit；DMU）。選取股東權益、利息支出及薪資費用為投入變數，而以利息收入與非利息收入為產出變數。研究結果發現，在本研究中以臺銀98之表現最佳，總效率為100％且被參考次數達五次之最，可見各DMU應以臺銀98為標竿，做為效法學習的對象。屬於金控子公司的行庫，其總效率、純粹技術效率及規模效率皆優於非金控子公司之行庫。由差額變數分析可知，各個未達效率的DMU都有其應改善的項目。進一步以乘數分析與敏感度分析來探討，可知在產出項目中以利息收入對效率值的影響最大，在投入項目中則以利息支出影響最大。 The thesis uses the data envelopment analysis (DEA) to evaluate the operation efficiency for that close to the government policy and public life of the eight government banks. The period for the study is from 2009 to 2011, amount to three years and 24 decision-making units (DMU). Select the shareholders' equity, interest expense and salary costs for the input variables , and taking interest income and non-interest income for the output variables. The results show that the performance of the BOT (Bank Of Taiwan ) 98 is the best. The technical efficiency of the BOT98 is 100%, and there is reference to the number of five. Each DMU should be BOT98 as the benchmark, as a follow learning objects. For the banks of the financial holding group, the technical efficiency, pure technical efficiency and scale efficiency are better than the independent bank. The slack variables analysis shows that each inefficient DMU has projects that should be improved. Further to the multiplier analysis and sensitivity analysis to explore, we know that interest income in the output items is the most influential, and interest expense in the input items is the most influential.
