Second, concerning the quantitative approaches used in related studies, two of
them are the most popular. One is multivariate regression analysis, and this kind of
approach usually uses indicators like per capita research output to measure research
efficiency. The other is two-stage approaches like DEA-Tobit or DEA with
bootstrap. The later kind of approach uses thefirst-stage efficiency scores from
DEA model as dependent variable, and then explore what impacts on the these
efficiency scores. For sure, the application of two-stage approaches should pay
attention to ensure the robustness of results by using different techniques.
Third, concerning the selection and measurement of indicators, one type of
variable sometimes can be measured by different indicators. For instance, the
measurement of research efforts in universities could be measured by the ratio of
postgraduates to total student, or the ratio of research expenditure to total expen-
diture. Thus, it’s necessary to make comparison and select more appropriate vari-
ables into the model. For some indicators hard to be measured objectively, like
autonomy, competition, and reputation, one way is tofind suitable proxy variables,
the other way is to use some questionnaires or scales.
Forth, concerning the collection and utilization of data, since research production
is a cumulative process, panel data would be more accurate and comprehensive to
reflect this process, and much more helpful in making causal inference.
Fifth, concerning the selection of sample, different disciplines tend to use their
own production mode to produce knowledge, thus the influencing factors differs
among disciplines. Therefore, it’s necessary to explore these factors according to
discipline feature. While this approach is much difficult due to data accessibility.
Alternatively, we can divide universities into different types by referring to certain
standards, and explore the influencing factors at the sub-sample level.
Research Efficiency..................................... 6.2 Econometric Analysis of Factors Impacting on 211Us’
Research Efficiency
Based on the literature review on the factors impacting on the university research
efficiency in the previous section, we willfirst construct the indicator system of
influencing factors for the second stage analysis of DEA-Tobit models. The indi-
cator system would depend on the basis of constructing a theoretical indicator
system and the accessibility of data. Then, we will run the DEA-Tobit models to do
the second stage analysis for NEAM and HSS disciplines separately. Lastly, we will
provide some policy implications for NEAM and HSS disciplines respectively,
based on their separate results from econometric models.
6.1 Analytic Framework of Factors Impacting... 271