經濟部科技專案自民國68年推動至今，已成為我國產業科技政策最重要的環節之一。自經濟部技術處於83年公佈科技專案之投入項目與成果指標資料庫後，科專執行成效之量化評估即有更具體依據。本研究以經濟部技術處資料庫中民國83年至91年之科技專案之主要投入項目與成果指標，同時考量縱斷面（跨年度）與橫斷面（技術領域別與執行機構別）兩個面向，以資料包絡分析法(Data Envelopment Analysis)進行科技專案執行效率之評估與比較。實證結果發現材料化工與通訊光電領域的技術效率、規模效率、配置效率皆相對較佳，而生技製藥與共通性領域應擴大投入規模，業界科專的執行效率則進步相當快速。在時間序列分析中，樣本期間的科專在整體技術效率上呈現微幅的衰退，但在技術變動上卻有相當大幅的進步，促成總要素生產力變動指數呈現出每年7.2%的成長。另在分別將「學習效果」與「經濟景氣」兩項外生變數抽離後，對於早期年份之技術效率均有負面影響。

The Science & Technology Development Program (STDP) initiated by the Ministry of Economic Affairs (MOEA) in 1979 has now become the core of Taiwan's policy measures in promoting industrial technological innovation. The prior studies based on the STDP input/output database of the Department of Industrial Technology have fried to measure the technical efficiency of the STDP either by project or by institution. This paper is aimed to re-explore the efficiency issues in a more extensive way by the Data Envelopment Analysis. The six technological domains of the STDP are chosen as the analytical units and the time-series input/output data of the STDP across 1994 through 2002 is employed to examine the technical efficiency, allocative efficiency, Malmquist total factor productivity index, etc. The results show some performance disparities among the six technological domains: the domains of Materials-Chemicals Engineering and of Communication-Electronics are in a relatively efficient operation; Biotechnology-Pharmaceutical domain and general domain are in the status of the increasing returns to scale; Industrial STDP domain reaches the efficiency frontier in a short period of time, revealing the R&D projects conducted by the industrial sector are more efficient than those by the research institutions. In the time-series analysis, the slight decline of the aggregate technical efficiency in the studied period is compensated by the gain of technological progress, both jointly making the annual growth of 7.2% of TFP changes of the STDP implementation during 1994～2002. Learning effect and business cycle are also added as exogenous variables respectively into the model to alleviate some negative effects inherently existing in the earlier years.