Faere et al. ( 1994 ) used distances relative to DEA frontiers to calculate
Malmquist TFP change. Fuentes et al. ( 2001 ) and Orea ( 2002 ) used stochastic
frontier based on translog distance function to calculate Malmquist TFP index.
Compared with DEA, SFA has two obvious advantages:first, it’s able to explain
the white noise term; second, it can be used to conduct conventional test of
hypothesis. While the disadvantages are also very clear: on the one hand, the
necessity to specify a distributional form for the inefficiency term; on the other
hand, the necessity to specify a functional form for the production function (or cost
function). Moreover, in terms of consistency, the results of SFA are much more
consistent, not easily affected by outliers, thus more suitable for larger sample.
While DEA requires strong sample homogeneity, and its results are vulnerable to
outliers. Whereas an extraordinary advantage of DEA is that it’s rendered to deal
with multi-output situations, it is extremely complicated for SFA to deal with
multi-output, which needs to aggregate multiple output into one comprehensive
output or apply distance function.
and Productivity....................................... 2.4 Empirical Studies in Evaluating University Research Efficiency
Efficiency and Productivity
2.4.1 Empirical Studies Outside China.....................
In recent 30 years, more and more researchers are trying to apply SFA and DEA
methods to assess the research efficiency and productivity in higher education.
These international studies on measuring research efficiency and productivity in
universities are mainly focused on U.S., U.K., and Australia. Nevertheless, till
recently, there are not too many studies targeting research efficiency and produc-
tivity evaluations in universities, of which even fewer deal with the evaluations by
SFA. For instance, Izadi et al. ( 2002 ) used SFA to assess the technical efficiency
and cost efficiency in 99 British universities. Similarly, Horne and Hu ( 2008 ) used
the same method to estimate the technical efficiency and cost efficiency in 33
Australian universities. Stevens ( 2005 ) used the same method to assess the effi-
ciency of 80 England and Walsh universities from academic years of 1995/96 to
1998/99, and discussed the impact of characteristics in staff and students on the
efficiency. Some newly empirical studies came from Kempkes and Pohl ( 2010 ) and
Daghbashyan ( 2011 ), both studies applied SFA to examine the efficiency changes
in Germany and Sweden universities, and analyze the influencing factors.
Compared with empirical studies using SFA method, the non-parametric DEA
method is much more preferred by researchers due to its strength in handling
multi-product units. Table2.1summarizes some important literature. In the study of
Johnes and Johnes ( 1993 ), they used different input-output indicator systems to run
DEA models, aiming for assessing the research efficiency of economic departments
in British universities. Theirfindings show that the DEA results are not so sensitive
18 2 Evaluation on University Research Efficiency and Productivity...