Scientific and technological innovation efficiency in Chinese provincial higher education institutions: a three-division network DEA approach

Abstract

Scientific and technological (S&T) innovation in higher education institutions (HEIs) has become a major driving force behind China’s development. The Chinese government has attached great importance to S&T innovation in HEIs, creating the need for efficiency evaluation that can test the effectiveness of S&T policies and provide decision support for S&T innovation. The operational process of S&T innovation in HEIs can be decomposed into three divisions: research and development (R&D), technology transfer (TT), and technical service (TS). Furthermore, the intermediate products produced by the R&D division may be wasted in the TT division. Based on the internal structure and the utilization of intermediate products, this paper proposes a three-division network DEA to explore the efficiencies of S&T innovation in Chinese provincial HEIs. Based on empirical data from 2011 to 2020, the main findings are as follows. (1) The efficiencies of the S&T innovation in HEIs can be improved when the R&D and TT divisions cooperate to avoid wasting intermediate products. (2) The TT division is not only the main source of the overall inefficiency but also the main cause of the efficiency differences among areas. (3) The TT efficiencies of four areas tend to converge to the same steady state.

Keywords:

• Data envelopment analysis
• Higher education institutions
• Scientific and technological innovation efficiency
• three-division network structure
• Linkage efficiency

Notes

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 In the literature, the range of ${\alpha }_j$ is usually arranged by decision makers based on the assignment of the two divisions (Zhao et al., 2021). In this study, following Zhao et al. (2021), the range of ${\alpha }_j$ is given exogenously.

2 More ranges are provided to investigate the efficiency changes in Appendix C. The results verify that the changes in the assignment constraints do not have significant effects on the overall efficiency rankings. Thus, the effects of assignment constraints for the shared input on the efficiency results can be neglected in this study.

3 The efficiency results in the two cases are two paired-samples because they are obtained from the same set of provincial regions. In one situation, the efficiency results in the two cases do not satisfy the normal distribution (the normal distribution can be easily verified and thus omitted here), and the Wilcoxon test is appropriate for comparing such paired samples. In the other situation, the efficiency rankings in the two cases do satisfy the normal distribution, and the paired-samples t-test is appropriate for comparing such samples. Therefore, the Wilcoxon test and paired-samples t-test are used respectively to examine the efficiency differences and ranking differences between the two cases.

4 The linkage efficiencies in the cooperative case are all equal to one, and thus we omit these results in Table 3.

5 The Friedman test is appropriate for comparing several dependent samples. The efficiency results of the four areas are dependent samples, thus we use the Friedman test to examine the efficiency differences among the four areas. In addition, the Wilcoxon test is appropriate for comparing two dependent samples, so we use Wilcoxon test to investigate the differences between two groups in analyzing the four areas. For more details of the Friedman test see Hollander et al. (2013).

6 Due to space limitation, this study does not show the specific calculation process of the Dagum Gini coefficient. Interested readers may refer to the method of Dagum (1980, 1997).

7 Due to space limitation, this study does not show the specific calculation process of the $\sigma$ convergence model. Interested readers may refer to the method of Barro & Sala-I-Martin (1992) and Sala-I-Martin (1996).

8 Due to space limitation, this study does not show the specific calculation process of the $\beta$ convergence model. Interested readers may refer to the method of Barro & Sala-I-Martin (1992) and Sala-I-Martin (1996).

9 The New Economic Growth Theory believes that the convergence characteristics of economic variables differ due to different external conditions (Lucas, 1988; Romer, 1986). According to whether the convergence conditions are considered, $\beta$ convergence can be divided into absolute $\beta$ convergence and conditional $\beta$ convergence. Due to the availability of data, we only adopt the absolute $\beta$ convergence method to test TT efficiency. Alternatively, we could also use the conditional $\beta$ convergence method to analyze TT efficiency if the external conditions (e.g., degree of regional openness, intensity of financial support, and level of economic development) are considered. This can be an extension for future work.

10 The characteristics of division efficiencies are similar to those in Fig. C1, thus the analysis of division efficiencies is omitted.

Additional information

Funding

This work was supported by the National Natural Science Foundation of China (Grants 71701102, 71801132, and 71971205), 2022 Qinglan Project of Jiangsu Province in China, Key Program of the National Natural Science Foundation of China (71631006), and Major Program of the National Natural Science Foundation of China (72091210 and 72091215).