Efficiency difference of electric grid utilities considering heterogeneity factor: Empirical analysis between China and Japan

ABSTRACT

Different economic development models, resource endowment conditions, and reform and regulatory strategies in China and Japan affect the performance of electric grid utilities significantly. This study combined the meta-frontier model with the four-component SFA model to investigate the sources of efficiency bias for electric grid utilities in China and Japan from 2006 to 2020. The results indicated significant differences between the efficiencies of utilities in China and Japan. China’s average technical efficiency was 58%, while Japan’s was 75%. Japan’s efficiency is higher than that of China. Although largely caused by long-term and structural issues, the factors underlying the inefficiency of grid companies vary between China and Japan. After China implemented electricity reform in 2015, the efficiency gap between China and Japan has been gradually narrowing. By dividing the grid firms in China and Japan into eight groups, we found significant variations in efficiency between regions. These findings suggest that the reform of the power system should be carried out according to local conditions and there is a need for continued deepening of the electricity reform.

KEYWORDS:

• Efficiency difference
• meta-frontiers
• overall efficiency
• persistent efficiency
• transient efficiency

Acknowledgements

We appreciate the financial support from the National Natural Science Foundation of China under grants Nos. 72243009, 72174141 71874121 and 72173016, and the support from Major Projects of the National Social Science Fund of China under Grant No. 22&ZD104.

Disclosure statement

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

Notes

¹ This arises from decreasing cost in developmentalism instead of the claim of increasing cost in neoclassical economics.

² Data sourced from the World Bank: https://www.worldbank.org/en/home.

³ The translog function is more flexible than the Cobb-Douglas function. The translog function does not violate the curvature characteristics of the output space and is available for multiple outputs.

⁴ Due to missing data for some years, the panel dataset is unbalanced. The data for China related to 30 provincial-level companies and were sourced from the China Electric Power Yearbook, China Electric Power Industry Statistical Data Collection, China Meteorological Yearbook, China Statistical Yearbook and State Grid Limited Company Yearbook. Because data were not available, the Tibet Power Grid Corporation was not included. Hong Kong, Macao, and Taiwan were not included in this study because they are characterized by sizable differences in their corporate governance structure and operation mode. The data corresponding to the Inner Mongolia Electric Power Co., Ltd. and the State Grid Corporation in eastern Inner Mongolia were combined, and these data are hereinafter referred to as Inner Mongolia to keep all statistical data consistent during the period.

⁵ The data of 10 power T&D companies in Japan were obtained from the companies’ securities reports, Japan Statistical yearbook, Electricity Statistics Database and other publicly issued materials. The Japanese electric power companies investigated in this study include Hokkaido Electric Power Company (EPCo.), Tohoku EPCo., Tokyo EPCo., Chubu EPCo., Hokuriku EPCo., Kansai EPCo., Chugoku EPCo., Shikoku EPCo., Kyushu EPCo. and Okinawa EPCo., listed geographically from north to south.

⁶ The estimated result is achieved through Stata. The estimation steps were based on Kumbhakar, Lien, and Hardaker (2014).

⁷ We conducted LR tests on the Hicks neutral hypothesis for g1, g2, pool, and meta model. The chi-square statistic values were 101.66, 90.29, 109.40, 256.22, respectively, which was greater than the critical value of 18.47, even at the 1% significant level, so we rejected the assumption of technological neutrality.

⁸ The results of the Hausman test are detailed in Appendix Table A1.

⁹ Due to space limitations, the coefficient results are not reported in-text, see Appendix Table A2 for details.