【作者】 陈浩；

【导师】 王晓红；

【作者基本信息】 哈尔滨工业大学 ， 技术经济及管理， 2012， 博士

【摘要】 高等学校是我国科研活动的重要力量，尤其在基础研究活动中占有重要地位，我国政府对高校的科研创新一直都很重视。作为一个发展中国家，在国内纷纷提出建设世界一流大学的时代背景下，尤其在当前国家急需创建科技创新体系的形势下，如何充分利用我国高校有限的科研资源，提高我国高校在全球的学术竞争力，建设世界一流的大学，对高校的科研效率进行科学合理的评价显得尤为重要和迫切。科研评价方法有很多种，考虑到高校科研活动具有多投入多产出的特征，以及数据包络分析方法（Data Envelopment Analysis, DEA）在解决多投入多产出的综合评价问题所具有的优点，国内外许多学者都选择了DEA方法进行研究，取得了许多研究成果。但是，已有的研究成果都忽视了一个问题——大部分对高校的科研效率评价都是基于传统DEA模型，而传统DEA模型则隐含假设“投入产出发生在同一个时间段”，即高校当年投入当年得到产出，这显然与“高校科研投入产出分布在多个时间段”的现实情况不相符，即高校的科研经费是分多个年份拨入高校，并且高校的科研成果在连续多年里按照科研进度分阶段发表。此外，在现实的许多研究中，需要经常对决策单元（Decision MakingUnit, DMU）进行多个时期的考察，在此基础上进行面板数据分析来研究DMU的效率的波动和变化情况，一些学者选择应用DEA视窗分析模型（WindowAnalysis, WA）来进行动态分析。然而，在应用DEA视窗分析模型进行动态分析的研究过程中，仍然存在一些问题：学者基于主观选择的不同窗宽所得出的结果存在比较大的偏差；不同窗宽的结果存在偏差，如何确定理想的窗宽使得偏差最小；投入产出之间存在一定的滞后期，而滞后期对于视窗分析结果存在比较大的影响，这一点在已有的研究成果中都没有引起重视。针对上述这些问题，引出了本文所研究的问题。由于已有研究成果的科研评价指标比较零散，并且主观选择的随意性比较大，不能够较全面的涵盖高校科研活动，具有一定的片面性。所以，在第一章介绍国内外DEA模型研究现状的基础上，本文在第二章基于科研评价指标体系的建立标准和构建步骤，建立了我国高校的科研评价指标体系；其次，在第三章，基于我国高校的科研评价指标体系，应用DEA视窗分析模型，对我国高校在2003-2007年、2003-2008年和2003-2009年的科研效率进行动态评价，并对不同窗宽的分析结果进行了对比分析；在第四章，通过对比不同窗宽的分析结果，揭示了“窗宽的确定”对DEA视窗分析模型结果有重要影响，先从理论上进行分析，研究确定理想窗宽，建立理想窗宽的DEA视窗分析模型，对我国高校多年的科研效率进行动态分析；最后在第五章，增加考虑高校科研投入产出的滞后期，基于理想窗宽的DEA视窗分析模型，定量分析滞后期对我国高校科研效率的影响。研究结果表明：随着窗宽的增加，我国各省市高校的技术效率（TechnicalEfficiency, TE）和纯技术效率（Pure Technical Efficiency, PE）呈现递减的变化趋势，而规模效率（Scale Efficiency, SE）呈不规律的变化；科研效率（包括TE，PE和SE）在不同年份对窗宽变化的敏感程度不同，即在某些年份，科研效率值对窗宽的变化比较敏感，表现为不同窗宽下的效率值相差比较大，在某些年份，科研效率值对窗宽的变化不敏感，表现为不同窗宽下的效率值相差不大；选择不同的窗宽会使得研究者对科研效率的长期变化趋势得出大相径庭的结论。通过对比理想窗宽的DEA视窗分析模型与传统DEA模型的分析结果，发现两者之间存在很大的偏差，这进一步说明了建立新的DEA视窗分析模型，对我国高校科研效率进行动态分析是很有必要的，此DEA视窗分析模型的分析结果更符合科研工作的实际情况。考虑滞后期与不考虑滞后期的分析结果存在很大的偏差，说明滞后期对我国高校科研效率评价具有很大的影响，在高校科研评价中必须引起重视。本文基于理想窗宽的DEA视窗分析模型的分析结果，对我国高校的科研评价提出了一些对策建议。更多还原

【Abstract】 Colleges and universities are important forces in scientific research activitiesof China, and especially occupy an important position in basic research activities.The government of China always pays attention to the scientific research andinnovation activities of universities. As a developing country, in the background ofconstructing some top-level universities in the world at home, and especially in theurgent situation of establishing Chinese Science and Technology Innovative System,how to make the best use of our limited scientific research resources in order toenhance our universities’ academic competitiveness in the world, and to constructsome world class universities, it is particularly important and pressing to evaluateour universities’ scientific research efficiencies in a scientific and reasonable way.There are many methods in scientific research evaluation. Considering thecharacteristics of multi-input and multi-output (MIMO) in universities’ scientificresearch activities, and the obvious advantages that Data Envelopment Analysis(DEA) occupies to solve comprehensive evaluation problems with MIMO, somescholars used DEA to evaluate universities’ efficiencies. But one question is ignoredin existed studies such as most of these studies used conventional DEA models, thatsuppose inputs and outputs happen in the same period, for example someuniversities put inputs and gain the outputs in the same year, which does not inaccord with the reality. The inputs and outputs of scientific research in universitiesdistribute in many periods, for example scientific research funds are allocated touniversities in several years, and scientific research outputs are published in manyyears according to the scientific research phases. Moreover, some researches inreality need to study decision making units (DMU) in many periods, based on whichto have a panel data analys is to study the fluctuation and changes in DMUs’efficiencies. Some scholars used DEA window analysis (WA) model to have adynamic research. Many questions existed in the dynamic analysis process byapplying DEA WA model. There are bigger deviations in the results which obtainedbased on different window length that scholars choose subjectively. In considerationof these deviations, how to determine an optimal window length to make minimumthese deviations is a question. Last but not the least, time lags that existed inscientific research inputs and different outputs have an obvious effect on WA results, which are ignored in existing studies. These questions above-mentioned introducethe problems of this dissertation.Because the scientific research evaluation indexes in existing studies arescattered and choose subjective ly that cannot contain scientific research activities inuniversities completely, which have some one-sidedness to some extent. So, basedon introduce DEA research status at home and aboard in chapter1, this dissertationat first set up the scientific research effic iencies evaluation index system ofuniversities in China based on building standards and steps in chapter2, then useWindow Analysis to study the dynamic scientific research efficiencies ofuniversities of China during2003-2007,2003-2008and2003-2009in chapter3.The analys is results are compared between different window lengths. Based on thein-depth studies on the importance of the determination of window length inchapter4, this dissertation at first study the determining process in theory,determine an optima l window length, and then set up a WA model which have theoptimal window length to have a dynamic study on scientific research efficienciesof universities in China. Considering the time lag between scientific research inputsand outputs of universities in chapter5, the influence that time lag have onscientific research efficiencies are studied based on the WA model with an optimalwindow length quantitatively.The results show that technical efficiency (TE) and pure technical efficiency(PE) at universities in different provinces of China are decrease and scale efficiency(SE) is varied irregularly with the increase of window lengths. Scientific researchefficiencies are sensitive to the change of window length in different years. Whilewindow length changes, the changing trend of scientific research efficienciesstudied by scholars in a long time will change radically. There are great variancesbetween the results based on the Window Analysis model with a optimal windowlength and based on conventional DEA models, that further prove it is necessary toset up the Window Analysis model with a optima l window length to have a dynamicstudy and the results are more in accord with the reality of scientific researchactivities. There are great variances between the results considering time lag or notwhich proved that time lag have a great influence on scientific research efficienciesat universities of China. The scholars must pay attention to time lag in scientificresearch evaluation of universities. Based on the results above-mentioned, thisdissertation gives some suggestions on scientific research evaluation of universities in China.更多还原