目前國內公共工程無論現行考核方式或既有之績效評估文獻,皆無法有效顧及工程執行機關之內、外在執行條件,且難以落實至一般基層單位;又由於工程執行環境不同於一般行政部門,故亟需建立有效可行之績效評估模式,以提高公共建設執行績效。本研究以「行政院原住民族委員會」九十三年度被查核過的36個公共工程執行機關為實證分析對象,為兼顧研究成果之客觀性與準確性,本研究將先以群集分析篩選工程執行環境具同質性之工程機關,之後再使用資料包絡分析法衡量各工程單位之績效。 依據群集分析之結果顯示,36個受評機關在預算、承辦人學經歷、主辦課課長學經歷、人口數、營造廠商家數、年平均降雨量,六個不同的執行條件上應分為「高預算支出類型」及「普通預算支出類型」兩類。而資料包絡分析之衡量則以預算及承辦人學經歷、主辦課課長學經歷為投入項,並以查核之施工品質及品管制度分數為產出項,經由分群與不分群兩種方式進行資料包絡分析後顯示,「普通預算支出類型」之機關個數,佔全體不分群前50%之比例高達72.22%,此乃由於「高預算支出類型」之機關多位處偏遠,當所有機關一起進行分析時,「普通預算支出類型」機關則較佔優勢,可見分群與否確實影響了各機關之績效表現;此外本研究亦利用資料包絡分析中之「差額變數」,計算各無效率單位長短期應改進之方向與幅度,研究成果期可提供相關工程主管機關改進現行評估模式之參考。
Considering the present performance evaluate method couldn’t attend to entire condition of the public work execute agency which execute in different environment from ordinary administrative departments in Taiwan area. Therefore it’s urgently to build up an efficacious performance evaluate method to improve the achievements of the public works. The study empirical analysis with 36 agencies that have been construction inspection by “ Council of Indigenous Peoples, Executive Yuan” in 1994. In order to complete the study achievements, it would sift the agencies with the same execute construction environment by “Cluster Analysis” and then measure the performance of each agency by “Data Envelopment Analysis(DEA)”. According to the result of cluster analysis, 36 agencies in total budget, undertaker academic and experience background, undertaker chief academic and experience background, population, contractor, annual average rainfall these conditions should be separated for two groups “Highly Expenditure Agencies” and “General Expenditure Agencies”. Afterward utilizes the total amount as input, and the amount of cases, the score of construction quality and quality control system in construction inspection as output. After compare the analysis of two groups and whole agencies, “General Expenditure Agencies” has get ahead from the other one account for 72.22% in the whole agencies analysis by first 50%. This is owing to the remote of “Highly Expenditure Agencies” which is a disadvantage in whole agencies analysis. It prove swarm off or not influence the performance measurement of agencies. Besides this study utilizes “Slack” to calculate the short-term and long-term improvement strategies for those sections with less efficient. The results provide managerial suggestions for improving the current evaluation procedures.