Abstract: Conventional DEA models assume that the inputs of a specific period are consumed to produce the outputs of the same period. This underlying assumption may not be valid in some performance evaluation, that outputs of a specific period can be thought to be produced by consumption of not only inputs in the same period but also the inputs in multiple previous periods. In order to evaluate the efficiency of two-stage processes more objectively, this paper proposes a two-stage DEA model which considers the output time-lag effects of intermediate variable. And it applies the Malmquist productivity index to know the dynamic changes of time series finally, so as to find out the reason of efficiency change. The proposed method is applied to evaluate the innovation efficiency of 30 provinces of China. The results show that the rank of considering the output time-lag effects of intermediate variable model is more objective, the efficiency of innovation is not only closely related to the efficiency of R&D stage and economic transformation stage, but also closely related to the influence of external factors.

Key words: 时滞性, 两阶段数据包络分析（DEA）, Malmquist生产率指数

摘要：

传统DEA模型假设特定时期的产出是通过消耗同一时期的投入产生的，这个潜在的假设不适用于某些绩效评价，即一个特定时期的产出不仅需要消耗该时期的投入，而且需要消耗之前多个时期的投入。为了更加客观地评价两阶段过程的效率值，构建了考虑中间变量产出时滞性的两阶段DEA模型，同时将其与Malmquist生产率指数相结合，寻求时间序列上的动态变化，从而找到效率变化的原因。并将提出的方法应用到中国30个省份创新效率的评价中，结果表明，考虑中间变量产出时滞性的模型排名更客观；创新效率不仅与研发阶段和经济转化阶段的效率密切相关，而且还受外部因素的影响。

关键词: 时滞性, 两阶段数据包络分析（DEA）, Malmquist生产率指数