Financial Structure and Income Inequality: Evidence from China

ABSTRACT

This study investigates the relationship between financial structure and income inequality in China and explores a channel for changes of financial structure to influence income inequality. Our results suggest that, relative to total bank credit, an increase in the raised capital from the stock market reduces income inequality, whereas a rise of turnover in the stock market augments income inequality. Financial structure affects income inequality by influencing the development of medium-sized enterprises. Our evidence supports the financial structure relevancy view. To reduce income inequality, the Chinese government should help to promote equity financing and decrease excessive speculation on the stock market.

KEY WORDS:

• China
• financial structure
• income inequality
• medium-sized enterprises

JEL CLASSIFICATION:

• D3
• P3
• O1

Acknowledgments

We greatly appreciate the comments and suggestions given by the anonymous referees, Wentao Ma, Cheng Li, Honggang Xue, Shun Zhang, and Caoyuan Ma. The opinions expressed in this article are those of the authors and do not necessarily represent those of the PBC. All errors remain our responsibility.

Notes

^1. Stock exchanges were established in Shanghai and Shenzhen in 1990 and 1991, respectively, which are considered an important benchmark in the financial market development in China. Before this event, banking was almost the only financial sector responsible for financial resource allocation. For more details regarding the development and transition of China’s financial system, please refer to Allen et al. (2012).

^2. According to ‘China Capital Markets Development Report’ over years provided by China Securities Regulatory Commission, many non-state-owned enterprises, such as Suning Appliance, have developed rapidly after going public in China’s stock market.

^3. SMEs, both in urban and rural areas, contribute more than 50% of the tax revenue and 65% of patents and inventions. Nevertheless, many studies have found that SMEs in China are significantly more financially constrained than their larger counterparts (Zheng & Chen 2007).

^4. The fact that large state-owned enterprises usually obtain lower cost credits from banks is one of the characteristics of dual structure of China’s finanical system (Allen et al. 2012).

^5. In 2012, “Several opinions on deepening the reform of income distribution system” was made and implemented by National Development and Reform Commission who had authorization from State Council of the People’s Republic of China, which gave rise to a considerable change in Chinese income distribution structure and the whole economic system. Therefore, the data after 2011 are excluded from our sample in order to capture a more precise relationship between financial structure and income inequality in China.

^6. Market value traded is a better measure of the actual size of China’s stock market compared to ‘market capitalization’ because the latter includes non-tradable shares. Market fund raised measures the activity of the stock market in social capital allocation, given that we cannot obtain the data of ‘float supply of China’s market capital’. The theoretical mechanism in this study is really associated with the financial structure that is a ratio rather than the absolute value of traded value and raised fund. Following Levine (2002), Allen et al. (2012), and Lin and Xu (2011), total bank credit is taken as the denominator when we define the financial structure indicators in this study.

^7. We follow the income hierarchy defined in the Statistical Yearbook in various regions of China.

^8. The estimated Gini coefficients, especially the evolution pattern of these coefficients, are mainly consistent with Naughton (2007) and Benjamin et al. (2008). Actually, because of the various detailed calculation approaches and data sources, estimated Gini coefficients are not entirely coincident in the existing literature. But most of all, the evolution patterns of Gini coefficient estimates among these studies, including Xie and Zhou (2014), are all coincident.

^9. For the purpose of saving space, we do not provide the concrete form of a treatment effect model in this article. Please refer to Guo and Fraser (2010) for more modeling details about Heckit treatment effect models.

^10. The mean of the Gini coefficients is 0.377 in our sample period. We can calculate the amount of change in the mean of Gini coefficients on the basis of the coefficient estimate on StructureSize_it.

^11. As to the latter augmentation effect on income inequality, Chi (2012) reveals an illuminating channel that capital income mainly from stock trades is distributed the most unequally, and its contribution to total income inequality has been growing. This study aims to exploit another channel associated with the development of SMEs.

^12. Chinese SMEs usually are labor-intensive enterprises. To obtain additional outputs, they need to employ numerous low-skilled workers who belong to the low-income group.

^13. To eliminate the influence of the possible endogenous variables, we further run the regression excluding Privatization_i,t-1, Gintervention_i,t-1, and Laeduyear_i,t-1 respectively or simultaneously. The regression results are also reported in Table 2.

^14. It is worth mentioning that although the coefficient of Structure Size is insignificant in Column 4 of Table 2, the result in Column 1 is more credible because more control variables are included in the empirical model. Additionally, this coefficient is also significant in specifications (2), (3), and (5), respectively.

^15. These regressors are set according to the identified determinants of a nation’s financial structure by Levine (2002) and Allen et al. (2012). The exclusion variables are year dummy variables in Tables 4 and 5 since in the outcome equations, the cross-multiplication items between a row vector standing for a set of province dummies Regiondum and a variable of time trend are employed to control for provincial time trends, which is a more accurate approach to control for time effects for different provinces in China.

^16. That is, StructureIndex_it and Lpgdp_it in Table 4 and Gintervention_it and Privatization_it in Table 5.

^17. For the parsimony and identifiability of our structural form model, the cross-multiplication items between a row vector standing for a set of province dummies and a variable of time trend are not included in model (5), considering the limit of our sample size. We can obtain the same empirical conclusions if the dynamic panel models are employed as a robustness test, which include the cross-multiplication items, Regiondum × Trend_t.

Additional information

Funding

Financial support from National Natural Science Foundation of China (Grant No. 71473189) is gratefully acknowledged.