Fiscal decentralization and interregional Capital misallocation: evidence from China

Abstract

Misallocation of factors of production has been recently viewed as a promising explanation accounting for the large difference in total factor productivity (TFP) across countries. This article differs from previous studies by concentrating on interregional capital misallocation and by focusing on the role of fiscal decentralization in shaping misallocation. Using a city level panel data set, we measure intra-provincial and inter-city capital misallocation in China over 2003–2018. The empirical results based on provincial level panel data suggest that fiscal decentralization can lower inter-city capital misallocation while revenue decentralization performs better than expenditure decentralization. We further find that this positive effect is more significant and much larger when it is the market rather than government intervention that is driving the flow of capital. The results are robust to subsample regressions, IV estimations, spatial autoregressions and alternative measurement of interregional misallocation.

Keywords:

• Fiscal decentralization
• interregional misallocation
• China

JEL Classification::

• H11
• H23
• H70

Acknowledgements

We thank three reviewers for excellent comments. We are very grateful to Yongzheng Liu and Baoyun Qiao for their constructive suggestions. All errors are ours.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 This mechanism may not appear to directly affect the dispersion of MPK across regions, but it does imply that capital may be less efficient or productive than it could be when allocated elsewhere. To see this, consider a centralized government planning to allocate a budget of 1 million dollars to provide public services in two regions, jurisdictions A and B, where citizens in A prefer better school quality while citizens in B are in favor of better natural environment. However, decision making by the centralized government may not be able to incorporate this information, so each of these two regions obtains 0.25 million for constructing schools and 0.25 million for building parks. Suppose that resources for building schools and parks cannot be converted to each other. It is evident that it is possible to attain a more efficient allocation by reallocating resources across regions A and B. That is, it would be better to shift budget funds for schools from region B to region A or/and reallocate investment in parks from region A to region B. The same logic applies when it comes to the producer efficiency discussed later, for which, however, we may expect a more direct association of centralized/decentralized decision making with the dispersion of MPK.

2 For more evidence on local protectionism in China, please refer to Eberhardt, Wang, and Yu (2016) and Barwick, Cao, and Li (2017).

3 We realize that most studies measure misallocation based on micro firm-level data. In particular, relevant studies on China use data from the Chinese Annual Survey of Industrial Firms, which is a census of all non-state firms with more than 5 million yuan in revenue plus all state-owned firms (for example, Hsieh and Klenow 2009). However, using this firm-level dataset would result in some other problems. First, the dataset is not available after the year of 2013, and actually most studies focus on the sample period before 2009 considering the quality of the data. Therefore, it cannot provide enough observations of inter-city misallocation for us to conduct panel regressions. Second, measuring interregional misallocation based on firm-level data can be quite complicated. In the context of this paper, misallocation comprises inter-city misallocation and misallocation within each city, but we focus only on the former. Furthermore, distinguishing between these two types of misallocation relies on the assumption that the factors of production are first allocated within the city to attain the optimal and then allocated across cities, which is not necessarily the case in reality. Third, this dataset covers only manufacturing firms, hence we cannot gain a full picture of misallocation with this dataset, especially when the share of the service sector in China has increased from 39% to 60% over the sample period. In fact, when it comes to measuring spatial misallocation, aggregate data is often employed by researchers (see Brandt, Tombe, and Zhu 2013; Hsieh and Moretti 2019).

4 Banerjee and Duflo (2005) present a comprehensive review of existing methods for measuring MPK.

5 In Caselli and Feyrer (2007), they separate natural resources such as land from reproducible capital to correct the capital share and find substantial change in results. However, we simplify the measurement due to the lack of data in our sample.

6 Earlier studies adopted lower rates, for example, 5% in Perkins (1988) and 6% in Young. The specific rate adopted in Shan (2008) is 10.96%.

7 In China, the government invests a large amount of money in infrastructure, but not all these are included in the government’s official budgets. Many of these investments are usually conducted by large state-owned enterprises. On the other hand, a good part of government expenditure is used to provide regular public goods and services, which should not be viewed as “interventions” in the local economy. Therefore, our measure excludes expenditure on the provision of health, education and social insurance services, the most common and important types of “justified” local public services.

8 In 2000, 21 regions were reformed to be prefecture-level cities and 2 county-level cities were upgraded to prefecture-level cities. In 2001, 6 regions were changed to prefecture-level cities and then in 2002, 10 more regions were authorized to be prefecture-level cities.

9 We also use GLS approach to estimate the model and the obtained results are reported in Table A5.

10 We thank the reviewer for pointing this out.

11 We realize that the accountability argument may have a reduced validity in China because of the absence of competed democratic elections of local officials.

12 To further confirm this, we collect data on provincial level fiscal transparency issued by Shanghai University of Finance and Economics to capture local governance and run a regression by replacing intervention with fiscal transparency in columns (1) and (2) Table 3. The corresponding results are presented in Table A6. We find that the interaction term between fdrev and fiscal transparency is negative and statistically significant at the 5% level and the coefficient on fdrev is insignificant.

13 The three measures in Bartelsman, Haltiwanger, and Scarpetta (2013) are the within-industry standard deviations of labor productivity and of TFP and the correlation between firm size and the average product of labor based on the intuition that productive firms should be big in size.

Additional information

Notes on contributors

Zheng Li

Zheng Li is a Ph.D. candidate in economics in Department of Economics, Georgia State University. His field of interest is public economics and applied microeconomics.

Jorge Martinez-Vazquez

Jorge Martinez-Vazquez is Director of International Center for Public Policy at Andrew Young School of Policy Studies, Georgia State University. His field of interest is public economics.