SE are around 0.19, 0.17 and 0.11 respectively, and those of TE or SE have minor
fluctuations, and that of PTE is basically stable. This reflects that the TE, PTE and
SE of NEAM disciplines in 211Us share the same pattern.
4.2.1.2 2007–2009 Results with MA Data
In order to avoid the estimation bias caused by irregular changes of time-series data,
and try to solve the problems of input effect and lag of output, we apply Moving
Average method combined with one-period lag to re-estimate the research pro-
duction efficiency in each year. Similarly, we can use CRS and VRS DEA models
to calculate the research production efficiency and its decomposition of the full
NEAM sample for 2007–2009. The detailed calculation results for every university
are presented in Tables4.46,4.47and4.48in Appendix. Here we summarize the
results into Table4.6.
Table4.6presents the numbers of DEA-efficient DMU and their operation
stages of full NEAM sample estimated in 2007–2009 by the method of MA with
one-period lag. Over the period, the number of technically efficient universities is
around 30 (or 30% of all 211Us), about 5% lower than the results using original
data in previous subsections. From the results of SE of university research pro-
duction, there are 32% of universities achieving optimum scale, and 10–15% of
Fig. 4.2 Changes of TE,
PTE and SE for NEAM
disciplines (2006–2010)
Table 4.6 Distribution of TE
and SE for NEAM disciplines
(2007–2009, MA)
2007 2008 2009
Overall technical efficiency
TE = 1 33 (33%) 32 (32%) 30 (30%)
TE < 1 67 (67%) 68 (68%) 70 (70%)
Scale efficiency
IRS 9 (9%) 15 (15%) 18 (18%)
DRS 57 (57%) 53 (53%) 49 (49%)
CRS 34 (34%) 32 (32%) 33 (33%)124 4 Evaluation on Research Efficiency of 211Us: The DEA Approach