Stay or Go? Intra-government Tax Competition and Firms’ Location Decisions in China

ABSTRACT

The impact of the intra-government (vertical) tax competition on firms’ location choices is seldom studied in the literature. In this paper, we use the local corporate income tax revenue retention rate to represent the outcome of the vertical tax strategic interactions between the central and local governments in China. Theoretically and empirically, we find that a higher local corporate income tax revenue retention rate attracts newly-established firms, even when we simultaneously take into account the effect of the horizontal tax competition, i.e., the local effective corporate income tax rates. Such positive effect of the local corporate income tax revenue retention rate on firms’ location decisions is robust to a battery of sub-sample analyses and alternative model specifications, and is heterogeneous across firms with different ownership structures. This effect is more prominent in provinces with weaker agglomeration forces as well.

KEYWORDS:

• Corporate income tax
• vertical tax competition
• firm location

JEL CLASSIFICATION:

• H25
• H71
• R30

Acknowlegments

We thank Li Gan, Huanxiu Guo and Xiandeng Jiang for their constructive suggestions. We are grateful to Editor Ali Kutan and three anonymous reviewers for their helpful comments and advice. We also appreciate the feedback from the participants in 2018 IEF Conference on Financial and Economic Development in China in Nanjing Audit University. Chengrui Xiao acknowledges the financial support from [National Social Science Fund] under Grant [No. 16CRK002]; [Fundamental Research Funds for the Central Universities] under Grant [No. JBK171103]. All remaining errors are our own

Notes

1. In this paper, the terms intra-government tax competition and vertical tax competition are used interchangeably. The terms inter-government tax competition and horizontal tax competition are also used interchangeably.

2. In this paper, unless otherwise stated, the term “local governments” refers to the provincial governments in China.

3. Section 2.2 provides detailed discussions on the CIT revenue sharing system in China.

4. The FIEs are defined as firms receiving investment from foreign countries and from Hong Kong, Macau and Taiwan as well.

5. According to the 2007 standard issued by the National Bureau of Statistics of China, the above designated size manufacturing firms referred to those with annual sales above 5 million RMB (around US$750,000).

6. The documentations issued by the Qianjiang district is available on line: http://blog.sina.com.cn/s/blog_1835553e60102y3pz.html.

7. Theoretically, upper layers of governments have the priority regarding the retention of tax revenue. Each tax payment should be first transferred to the central government’s budget account and the central government can retain its proportion of tax revenue in the first place. Then, the remaining tax revenue would be transferred to and retained by the accounts of the provincial governments, and then, to the accounts of lower layers of governments. However, in order to obtain more tax revenue, the local governments and local tax administrators sometimes violate the retaining process and directly transfer the tax payments to the local governments’ budget account. Also, the local tax administrators would deliberately change the type of the tax payment from a shared/central tax to a local tax. To do so, the revenue does not have to be shared with the central government and can be exclusively obtained by the local governments.

8. To lower firms’ tax burden, the local governments could use various approaches. For example, the local governments would illegally extend tax exemption period and tax payment. They also can grant tax preferences, which are specifically designed for high-tech and eco-friendly enterprises, to some unqualified firms. Firm can negotiate the so-called “revenue loss” contracts with local governments to hide the true profits from the center. Moreover, local governments can deliberately reduce tax enforcement level and turn a blind eye to some tax evasion activities.

9. See Guriev (2004).

10. It can also be interpreted as the firm’s probability of being located in each province.

11. The ASIF is cleaned as follows. (a) The firms with invalid or duplicated identification code are excluded. (b) The firms with annual sales less than 5 million RMB and employees less than 10 are excluded. (c) The observations with missing values regarding the key financial information, such as total assets, net value of fixed assets, the number of employees, annual sales, are dropped out of the sample. (d) The observations violating the following General Accepted Accounting Principles (GAAPs) are excluded: (i) the net value of fixed assets must be smaller than the total assets; (ii) the liquid assets must be lower than the total assets; (iii) a firm’s paid-in capital, interest payment and export value must be non-negative; (iv) a firm’s profit ratio must lie between 0.1% and 99%; (v) a firm’s establishment time must be valid, i.e., the opening year cannot exceed the year when it was surveyed, and the opening month cannot greater than 12.

12. We first calculate a firm’s effective CIT rate at year t by using CIT payable/profit. To alleviate the influence of the outliers, this continuous variable is winsorized at the 1% and 99% levels. The effective CIT rate of province k at year t is obtained by averaging the effective CIT rates of all firms located within province k at year t.

13. The marginal effect of the control variable x₁ (with no interaction terms) is$\hat{p}\times (1-\hat{p})\times {\beta }_1$, where β₁ is the estimated coefficient, and $\hat{p}$ is the unconditional probability of location. In this paper, since firms choose to locate in one of the thirty provinces (other than Tibet) in China, the unconditional probability of location is around 0.0333 (= 1/30).

14. Including non-linear terms in the conditional logit model would complex the interpretation of the marginal effects of non-linear variable. Following Davies, Greenwood, and Li (2001), the marginal effects of ln(Infra) should be calculated by the formula $\hat{p}\times (1-\hat{p})\times ({\beta }_{infra}+2\times {\beta }_{infr{a}^2}\times \overline{\mathrm{l}\mathrm{n}(\mathrm{I}\mathrm{n}\mathrm{f}\mathrm{r}\mathrm{a})})$, where $\hat{p}$ is the unconditional probability of location, β_infra is the coefficient of ln(Infra), ${\beta }_{infr{a}^2}$ is the coefficient of ln(Infra)² and $\overline{\mathrm{l}\mathrm{n}(\mathrm{I}\mathrm{n}\mathrm{f}\mathrm{r}\mathrm{a})}$is the mean of ln(Infra) from the sample data.

15. It would be ideal if we could use the local CIT revenue retention rate changes induced by the exogenous shock, i.e., the 2002 CIT revenue sharing reform, to investigate the effect of intra-government tax competition on firms’ location decisions. In the previous literature, Wu et al. (2017) exploit the variations of the 11th Five-Year Plan water pollution reduction mandates, which were assigned as fixed targets to each province, as a quasi-natural experiment. They employ the difference-in-differences (DD) approach to investigate how environmental regulations affect firms’ location decisions. Unfortunately, even after 2002, the local CIT revenue retention rates tremendously varied cross provinces and time. We are unable to isolate the treatment and control groups for the 2002 CIT revenue sharing reform, so the DD approach is infeasible. However, we still use the CIT revenue attribution system before 2002 as a placebo test.

16. Follow Ai and Norton (2003), the marginal effect of the interaction term is calculated by $\hat{p}\times \left(1-\hat{p}\right)\times {\beta }_{12}+\hat{p}\times \left(1-\hat{p}\right)\times \left(1-2\hat{p}\right)\times ({\gamma }_1+{\beta }_{12}\times \overline{ln(\mathrm{N}\mathrm{\_}\mathrm{f}\mathrm{i}\mathrm{r}\mathrm{m}\mathrm{s})})\times ({\beta }_2+{\beta }_{12}\times \overline{\mathrm{l}\mathrm{n}(\mathrm{R}\mathrm{e}\mathrm{t}\mathrm{a}\mathrm{i}\mathrm{n})}).$ The marginal effect of ln(Retain) is calculated by $\hat{p}\times \left(1-\hat{p}\right)\times ({\gamma }_1+{\beta }_{12}\times \overline{ln(\mathrm{N}\mathrm{\_}\mathrm{f}\mathrm{i}\mathrm{r}\mathrm{m}\mathrm{s})})$. The marginal effect of ln(N_firms) is calculated by $\hat{p}\times \left(1-\hat{p}\right)\times ({\beta }_2+{\beta }_{12}\times \overline{\mathrm{l}\mathrm{n}(\mathrm{R}\mathrm{e}\mathrm{t}\mathrm{a}\mathrm{i}\mathrm{n})})$, where $\hat{p}$ is the unconditional probability of location, β₁₂ is the coefficient of the interaction term, γ₁ and β₂ are the coefficients of ln(Retain) and ln(N_firms), respectively. Since the marginal effect of the interaction term depends not only on the coefficients but also on the values of interacted variables, for simplicity, we use the mean of the interacted variables, $\overline{ln(\mathrm{R}\mathrm{e}\mathrm{t}\mathrm{a}\mathrm{i}\mathrm{n})}$ and $\overline{ln(\mathrm{N}\mathrm{\_}\mathrm{f}\mathrm{i}\mathrm{r}\mathrm{m}\mathrm{s})}$ to calculate the marginal effects of the interaction terms.

Additional information

Funding

This work was supported by the The National Social Science Fund [No. 16CRK002]; Fundamental Research Funds for the Central Universities [No. JBK171103].