The Urban–Rural Gap of Chinese Household Finance

ABSTRACT

Using data from a large household survey, we investigate the size of China’s urban–rural gap in ownership of bank deposits, risky financial assets, and credit cards. We further examine the factors underlying the gap using decomposition analysis. Compared to their urban counterparts, rural Chinese are much less likely to own a variety of financial products. Both demand-side barriers and supply-side barriers to financial inclusion exist in China. More, we use instrumental variable analysis to address the endogeneity of the local supply of financial service. Above all, our study indicates that a large financial services vacuum in rural areas needs to be filled.

KEYWORDS:

• Chinese economy
• decomposition analysis
• financial inclusion
• household finance
• urban–rural divide

JEL CLASSIFICATION:

• G11
• G21
• D14

Supplementary Material

We present regression and non-linear decomposition results based on Probit model in this supplementary document.

Notes

1. For the English version of the CHFS questionnaires, see http://www.chfsdata.org/intro-14.html.

2. Wording of financial literacy questions is provided in the appendix.

3. Item non-response on economic variables such as income and is a typical problem in household survey data. The 2013 CHFS uses follow-up unfolding bracket questions to collect partial information when respondents failed to provide answers to open-ended questions on income and wealth. The survey team imputed exact amounts based on bracket information. We use the imputed values for income and wealth in our analysis.

4. Regression and decomposition results based on Probit models are provided in the online supplementary material.

5. The IHS transformation can be expressed as:$\mathrm{i}\mathrm{h}\mathrm{s}\left(\mathrm{x}\right)=log\sqrt{\left({\mathrm{x}}^2+1\right)+\mathrm{x}}$.

6. We thank one referee for pointing out this issue.

7. The CBRC provides in its website basic information (name, address, founding time, etc.) for all bank branches. Relevant information is collected by using a web crawler and addresses are matched to counties on Baidu Maps.

8. Ideally we would like to use distance instead of transit time. However, such information is not available in CHFS 2013.

9. The sample size is slightly reduced as for some households the information on county of residence is missing.

Additional information

Funding

This work was supported by the National Natural Science Foundation of China (No. 71471119), the China Postdoctoral Science Foundation (No. 2017M620382), the 111 project of China (No. B16040), and the Fundamental Research Funds for the Central Universities in China (No. JBK170148).