本文运用数据包络分析（DEA）的CCR和BCC模型，考虑到科技成果指标变化，对我国30个省市高校在1999-2002，2003-2006年两个时期的科研效率，catch-up效应进行比较分析；然后运用DEA的视窗分析（Window Analysis）研究科研效率的趋势和稳定性；最后在视窗分析结果的基础上，进一步进行相关分析。结果表明，科研相对产出与科研效率呈较大的正相关，尤其是人均科研相对产出。通过确定科学合理的科研产出指标，从重视绝对值的科研数量产出转向重视相对值的科研质量产出，完善科研激励制度，激发科研人员的科研热情，再加上合理的配置科研经费项目资源，有助于提高我国各省市高校的科研效率。

Based on Data Envelopment Analysis (DEA) models, such as CCR and BCC, considering the change of evaluation indicators for scientific and technology achievements, a comparative analysis on Universities’ R&D efficiencies and catch-up effect in 30 provinces of China during the period of 1999-2002 and 2003-2006 is conducted, and then the trend and stability of R&D efficiencies are studied by using DEA window analysis. At last, on the basis of window analysis results, an in-depth correlation analysis is discussed. The results show that R&D relative outputs, especially R&D relative outputs per capita, have an positive correlation relation with R&D efficiencies to a certain degree. By determining scientific and reasonable R&D output indicators, the focus is shifted from absolute R&D quantity outputs to relative R&D quality outputs, the R&D incentive system is improved and research staffs’ enthusiasm is boosted, and R&D funds and projects are rationally allocated. All of that can enhance the Universities’ R&D efficiencies in different provinces of China.