*4.2. The Impact of Government Incentive on Tax Avoidance*

According to the theoretical model, when the government's GDP incentive is large, the government will increase preferential treatment for enterprises, which will lower the actual tax burden of enterprises, and the government's preferential policies favors more to automobile manufacturing enterprises, which contribute more to capital accumulation. Table 3 shows the regression results of the impact of government incentives on the actual tax burden and the differences of this impact between automobile manufacturing and maintenance enterprises. We did not list the coefficients of intersection terms between the annual dummy variables and the calculated profit *PRO* in the table, and the intersection between provincial dummy variables and the calculated profit *PRO* are also omitted for simplicity. All regressions used the robust standard error to avoid heteroscedasticity problems. No government incentive variable is added In Column (1) of Table 2. The coefficient of intersection term of *VEHM* and *PRO* represents the difference of actual tax burden between automobile manufacturing enterprises and maintenance enterprises, and the coefficient is not significant, which means the actual tax burden difference between the two enterprises is not significant if the influence of government incentives is not controlled. The regression results of Column (2–4) show the regression results controlling the government incentives. The coefficients of the *GOV* × *PRO* in the column (2–4) are insignificant or negative, which indicates that the greater the government's GDP incentive is, the larger preferential to the enterprises, and thus, the actual tax burden of enterprises will be reduced. However, we care more about the coefficients of the triple intersection terms of *VEHM* × *GOV* × *PRO*. In the Column (2–4), the coefficients of these terms of are all significantly negative, which shows that although the government's GDP incentive leads to the reduction of the actual tax burden of enterprises, however, what more important is that the government's intervention is biased, and the government gives more preferential subsidies to firms of automobile manufacturing industries. The government's GDP incentive plays a bigger role in reducing the actual tax burden of enterprises for automobile manufacturing enterprises than for automobile maintenance enterprises. For example, the coefficient of the triple interaction term in Column (4) is −0.4934, which indicates that the negative effect of government incentives on firm's tax burden is about 0.5 lower in the automobile manufacturing firms compared with the automobile maintenance firms. These results confirm the hypothesis and prove the conclusion of theoretical model empirically as well.


**Table 3.** The influence of government incentive on the sensitivity of reported profit and real profit.


**Table 3.** *Cont.*

Note: *t* statistics in parentheses, \* *p* < 0.10, \*\* *p* < 0.05, \*\*\* *p* < 0.01.

#### *4.3. The Influence of Government Incentive Changes on Tax Evasion of Di*ff*erent Types of Enterprises*

#### 4.3.1. Enterprises Ownership

The impact difference in government incentives varies in different types of enterprises. For example, it is generally believed that government interventions are different in enterprises with different ownership in the literature. On the one hand, state-owned enterprises will receive more preferential treatment and subsidies from the government. On the other hand, state-owned enterprises (SOEs) are mostly in the upstream industries, especially the manufacturing industry [59]. According to the theoretical model, enterprises in the automobile manufacturing industry, which is more likely to be the upstream industries, are more easily affected by the government incentives. So we conjecture that the difference of impact of government's GDP incentive tax avoidance between automobile manufacturing

enterprise and automobile maintenance enterprise will be more significant in state-owned enterprises. Therefore, we divided the samples into SOEs and non-SOEs to compare this impact difference. We define the SOEs as the firms whose registration types are state-owned, collective or wholly state-owned firms. Table 4 showed the regression results of Equation (28) in SOEs and non-SOEs. We can see that the coefficients of triple interaction term *VEHM* × *GOV* × *PRO* are all significantly negative in the group of SOEs, which is Column (1) (3) and (5), and the absolute value is larger than the coefficients in Table 3. However, the coefficients of triple interaction term *VEHM* × *GOVPRO* are not significant in the group of non-SOEs, i.e., Column (2) (4) and (6), which indicates that the bias of preferential government subsidies caused by GDP incentives is more obvious in SOEs. The government, which is inspired by GDP incentive, gives more subsidizes to upstream enterprises, especially state-owned enterprises. This behavior will induce a larger amount of investment in industries that are more helpful to capital accumulation, which will promote economic growth. These results further prove our conclusion in the theoretical model.

**Table 4.** Regression results of different ownership enterprises.



**Table 4.** *Cont.*

Note: *t* statistics in parentheses, \* *p* < 0.10, \*\* *p* < 0.05, \*\*\* *p* < 0.01.

### 4.3.2. Firm Size

Large enterprises playing a greater role in capital accumulation, and the government tend to "seize the large enterprises and release the small ones" [60], so large enterprises are more like to be the government's target and can promote GDP growth better. Therefore, if the government conducts a differential intervention on enterprises in different industries, due to the GDP incentive, we conjecture that the effect of a government tax incentive on reducing the actual tax burden of manufacturing enterprises will be more obvious in large enterprises. Therefore, we group the samples by firm size disclosed in the industrial enterprise database. The large firms include the enterprises with the label of "large scale", "super scale", "large type 1" and "large type 2", and the small firms include the enterprises with the label of "medium", "medium type 1", "medium type 2" and "small". Table 5 shows the regression results of the Equation (28) in firms with different size. We can see that all the coefficients of triple interaction terms are significantly negative, and the values are large in large firms' group. These results show that the phenomenon of the difference of intervention caused by government incentive is more apparent in large size enterprise, and the biased subsidies and preferential policies are given by the government to the automobile manufacturing industry are more likely to occur in large enterprises. These results prove our hypothesis, as well.

**Table 5.** Regression results of enterprises of different sizes.



**Table 5.** *Cont.*

Note: *t* statistics in parentheses, \* *p* < 0.10, \*\* *p* < 0.05, \*\*\* *p* < 0.01.

#### 4.3.3. Firm Belongs

The short-term GDP incentive of the government mainly refers to the incentive of local governments. The government tends to implement policies more effectively for enterprises that are close to it considering asymmetric information; thus, the enterprises far away are more likely to be decentralized to lower level of local governments [61]. Therefore, the local governments tend to give more intervention to enterprises that are in their jurisdiction for GDP growth. So we grouped the samples into local and non-local enterprises groups according to the level of enterprise registration authorities disclosed in the industrial enterprise database. Local enterprise refers to the firms whose enterprise registration authority is not a state. Table 6 shows the regression results in the different groups. We can see that the coefficient of triple interaction terms is all significantly negative in the local group, which are the Column (2) (4) and (6). However, they are insignificant in the group of central government authorized firms, which are the Column (1) (3) and (5). The results also suggest that the phenomenon of differential intervention caused by government incentives is more obvious in local firms. This may because the biased subsidies and preferential policies that the government gives to the automobile industry are more likely to occur in the local firm under jurisdiction.


**Table 6.** Regression results of central and local enterprises.


**Table 6.** *Cont.*

Note: *t* statistics in parentheses, \* *p* < 0.10, \*\* *p* < 0.05, \*\*\* *p* < 0.01.

#### **5. Robustness Check**

We do robustness check to make sure that our results are invalid. We used only the sample firms whose business status is normal operation. The business status of enterprises can be divided into business, closure, preparation, the closed, the bankruptcy, and other six states, according to the database of China's industrial enterprises. The enterprise in the state of abnormal operating may have biased tax avoidance, and the only enterprise in the state of normal operating can reflect the actual level of government intervention. Therefore, we only keep the firms who are in normal business status. There are 1492 companies are dropped, and Table 7 shows the regression results. We can see that the regression results are similar to the results in Table 3. For example, the first coefficient of triple interaction term *PRO* × *VEHM* × *VAT* in Column (2) is −0.8267, which is significant. These results show that the intervention on automobile manufacturing firms is biased as well, which is consistent with our previous results.



**Table 7.** *Cont.*

Note: *t* statistics in parentheses, \* *p* < 0.10, \*\* *p* < 0.05, \*\*\* *p* < 0.01.

#### **6. Conclusions**

The automobile manufacturing industry developed rapidly in China; however, the automobile maintenance industry less developed. In addition to the factors from the demand and supply side in the market, we tried to explain this fact from the perspective of government incentive and intervention in China. Firstly, we established a two-sector model with government incentives and government intervention, and then we analyzed the game between the government and market, and the optimal subsidy policy under the government's GDP incentive is obtained. The results of the theoretical model show that the government gives more preferential policies to automobile manufacturing firms compare to firms in automobile maintenance industries under short term GDP incentive. That's why the development of the automobile manufacturing and maintenance industry is unbalanced. Secondly, we use three indicators to represent the government GDP incentive under fiscal decentralization and test the differential impact of GDP incentive on tax avoidance of the two kinds of firms empirically. The empirical results show that the GDP incentive of the government caused by fiscal decentralization induced more preferential treatment to automobile manufacturing enterprises, and thus, increase their tax evasion degree, which proves the mechanism of government incentive in our theoretical model.

Understanding the incentive and implementation of industrial policy can help us understand the evolution mechanism of China's automobile industrial policy and automobile industrial structure better. Well-developed automobile maintenance industrial can improve customer loyalty that can help the automobile manufacture company survive in this highly competitive industry; with the increase of new kinds of vehicles and customer's requirements, it is necessary for the automotive service industry to constantly improve its development mode and introduce new technologies; and the balanced growth of automobile manufacturing and maintenance industry is one of the driving forces for market sustainability of automobile industry and sustainable regional growth. Based on these facts, we propose that, in an initial phase of the development of the automotive industry, the automobile manufacturing industry, which is upstream industry, should be encouraged to develop more than automobile maintenance industry, which is downstream industry, because the latter would have no reason to exist if the upstream industries did not exist. However, China may pay more attention to rebalancing the weight of the two industries after passing this first phase as economic develops. In this new phase, some measures" such as reducing the short term GDP incentive of local government and making the performance evaluation more diversified, will lead to better policy that promotes the transformation and upgrading of the whole automobile industrial structure, even the whole industrial structure optimization and economic growth.

**Author Contributions:** All authors were involved in preparing the manuscript. T.W. and L.M. contributed to the design of the research framework, L.M. contributed to the establishment of the theoretical model and the writing of the manuscript. Q.D. and T.W. conducted empirical analysis.

**Funding:** This research was funded by the National Natural Science Foundation of China, grant number [71703022 and 71804181], the Fundamental Research Funds for the Central Universities in UIBE, grant number [15QD21, CXTD7-04 and CXTD8-06], School of Banking and Finance, UIBE, and the National Center for Mathematics and Interdisciplinary Sciences, CAS.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A**

Proof of the result of Equation (18).

**Proof.** The consumer's budget constraint when *t* = 0 is:

$$w\_0 - w\_0 L + T\_0 = p\_0 \left| \frac{r\_0}{p\_0} K\_0 + (1 - \delta) K\_0 - K\_1 \right|. \tag{A1}$$

The consumer's budget constraint when *t* = 1 is:

$$w\_1 - w\_1 L + T\_1 = p\_1 \left[ \frac{r\_1}{p\_1} K\_1 + (1 - \delta) K\_1 - K\_2 \right]. \tag{A2}$$

The consumer's budget constraint when *t* = 2 is:

$$w\_2 - w\_2 L + T\_2 = p\_2 \left[ \frac{r\_2}{p\_2} K\_2 + (1 - \delta) K\_2 - K\_3 \right]. \tag{A3}$$

From the Equation (5), we can get

$$\frac{\beta u'(c\_{t+1})}{u'(c\_t)} = \frac{p\_t}{p\_{t+1}} \frac{1}{r\_{t+1}/p\_{t+1} + (1 - \delta)}.\tag{A4}$$

Therefore, we can know that <sup>β</sup>*u*(*c*1) *<sup>u</sup>*(*c*0) <sup>=</sup> *<sup>p</sup>*<sup>0</sup> *p*1 1 *<sup>r</sup>*1/*p*1+(1−δ). Multiply the left hand of Equation (A2) by β*u*(*c*1) *<sup>u</sup>*(*c*0) and multiply the right hand of Equation (A2) by *<sup>p</sup>*<sup>0</sup> *p*1 1 *<sup>r</sup>*1/*p*1+(1−δ), and then we can get:

$$\left| \frac{\beta u'(c\_1)}{u'(c\_0)} (c\_1 - w\_1 L + T\_1) = p\_0 \right| K\_1 - \frac{K\_2}{\frac{r\_1}{p\_1} + (1 - \delta)} \Big|. \tag{A5}$$

Multiply the left hand of Equation (A3) by <sup>β</sup>2*u*(*c*2) *<sup>u</sup>*(*c*0) and multiply the right hand of Equation (A3) by *<sup>p</sup>*<sup>1</sup> *p*2 1 *<sup>r</sup>*2/*p*2+(1−δ) <sup>×</sup> *<sup>p</sup>*<sup>0</sup> *p*1 1 *<sup>r</sup>*1/*p*1+(1−δ), and then we can get:

$$\frac{\beta^2 u'(c\_2)}{u'(c\_0)}(c\_2 - w\_2 L + T\_2) = p\_0 \left[ \frac{K\_2}{r\_1/p\_1 + (1 - \delta)} - \frac{K\_3}{[r\_2/p\_2 + (1 - \delta)][r\_1/p\_1 + (1 - \delta)]} \right],\tag{A6}$$

and so on. We add up all the results from Equation (A1), i.e., (A1) + (A5) + (A6) + ... , and we can get

$$\sum\_{t=0}^{\infty} \beta^t \frac{\mu'(c\_t)}{\mu'(c\_0)} (c\_t - w\_t L + T\_t) = [r\_0 K\_0 + p\_0(1 - \delta) K\_0]. \tag{A7}$$

This is the same with Equation (18) in Section 2. -

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
