The Impact of Industrial Subsidies and Enterprise Innovation on Enterprise Performance: Evidence from Listed Chinese Manufacturing Companies

Abstract

Governments worldwide have introduced various tax mechanisms to foster enterprise innovation, which in turn affect enterprise performance. To promote the innovation level of domestic enterprises, China has adopted an innovation-driven strategy policy. Based on China’s manufacturing company data from 2007 to 2017, this article constructs a mediating effect model to study the direct effect of tax incentives and government subsidies on enterprise performance and the mediating effect of innovation on enterprise performance. We use RIF regression to investigate the difference between the effect of industrial subsidies on promoting technological innovation and enterprise performance. The study finds that tax incentives and government subsidies encourage enterprise performance through innovation, with the mediating effect accounting for about 34.5% and 16.8%, respectively. Industrial subsidies play a more obvious role in improving the innovation performance of high-tech enterprises. There is no significant difference in tax incentives on the performance of large enterprises and small and medium-sized enterprises. Moreover, government subsidies play an essential role in promoting the performance of large enterprises. Furthermore, with the increase in quantile, the impact of tax incentives and government subsidies on innovation is getting more extensive and more significant. Innovation has an increasing effect on enterprise performance, the effects of tax incentives on enterprise performance are becoming less and smaller, and the nexus of government subsidies and enterprise performance is generally unchanged. Therefore, implementing appropriate tax incentives and government subsidies for enterprise innovation is essential for improving enterprise performance, especially for high-tech enterprises. Enterprise size should not be used as a criterion for the government to implement tax incentives, although government subsidies tend to support large enterprises.

1. Introduction

Most studies have long identified innovation as an essential factor driving economic growth [1,2,3,4]. In order to promote sustained economic growth, many countries have implemented preferential tax policies to encourage innovation enterprise, which is a market-oriented approach with high efficiency [5,6]. As evidence, as of 2017, 30 out of 35 OECD countries, 21 out of 28 EU countries, and a few non-OECD economies offer tax relief on innovation expenditure [7]. However, the high risk of corporate innovation activities and the positive spillover characteristics of innovation results lead to insufficient corporate innovation. This is actually why the government should intervene to provide guarantees for enterprises to carry out innovation activities. Among them, government subsidies and tax incentives are the two most commonly used measures [8].

The 19th National Congress of the Communist Party of China put forth that innovation is an important driving impact on development and strategic support for creating a new economic system and an important guarantee for the sustainable development of the country. To promote the innovation level of domestic firms and implement its innovation-driven strategy, China has implemented industrial subsidy policies for enterprise innovation. In 2008, China implemented the New Enterprise Income Tax Law, which decreased the introductory income tax rate from 33% to 25%. High-tech enterprises’ income tax rate was decreased to 15%, 40% lower than the benchmark income tax rate of 25%. In addition, tax preference is given to enterprises according to the additional deduction of R&D expenses. In order to promote enterprise innovation, the Chinese central government is encouraging many enterprises, especially listed companies, to apply for local government subsidies. The Fifth Plenary Session of the 19th CPC Central Committee denotes that achieving sustained high-quality development is innovation driven. However, China’s innovation capacity at present does not meet the requirements of high-quality development. We explore the effect of industrial subsidy policies on innovation and innovation on enterprise performance from the perspectives of tax incentives and government subsidies. It provides relevant policy enlightenment in terms of adopting effective industrial policies to promote independent innovation and high-quality economic development.

The increase in the credit rate increased the R&D expenditures of high-tech enterprises in response to taxes, but has little effect on the R&D expenditures of non-high-tech enterprises [9]. According to the review, other countries use various tax mechanisms and initiatives to facilitate small and medium-sized enterprises’ (SMEs’) innovation capabilities and how universities integrate the R&D efforts with industries through grants and incentives. SMEs could benefit from tax mechanisms’ spillover from financing their R&D collaboration [10]. In addition, existing research mainly focuses on domestic policies directly addressing appropriateness and financing issues and the second-order impacts on enterprises’ innovation output, and less attention has been paid to the difference between government subsidies and tax incentives [11].

This paper provides the following three contributions: Firstly, we use the mediating effect model to investigate the direct effect of industrial subsidy and the mediating effect of promoting enterprise performance by promoting innovation. Similar previous studies lack this perspective. For instance, Luo et al. [12] measured the potential relationship between governmental subsidies and financial performance at a micro level; Yu et al. [13] studied the relationship between government subsidies and new energy vehicle companies, and pointed out that the adjustment effect of smart transformation has a significant effect on government subsidies; and Liu et al. [14] used a panel threshold model to prove that energy consumption, technological innovation, and supply chain management show a non-linear relationship with enterprise performance and have a significant technological innovation threshold effect.

Our paper shows that innovation output plays a partial mediating effect on the nexus of industrial subsidy and enterprise performance. Following the literature, the government’s financial incentives have nothing to do with the patents of high-tech companies or ordinary companies [15]. We can calculate that the total marginal effect of tax incentives on enterprise performance is 0.252, where the mediating effect is 0.087. The total marginal effect of government subsidies on enterprise performance is 0.116, of which the mediating effect is 0.020. This shows that the promotional effect of government subsidies and tax incentives on enterprise performance through innovation is manifested as a mediating effect whose proportion of tax incentives is relatively higher, and the proportion of the mediating effect of government subsidies is relatively low.

We further use the mediating effect model to explore the difference between the promotional effect of industrial subsidies in high-tech enterprises and non-high-tech enterprises, large enterprises, and SMEs. Following the literature, as the tax incentives are found to benefit the research network mainly in large companies, they should become the base for future policies to emulate and apply to R&D in SMEs to enrich their innovation and product commercialization [10]. This paper explores the differential influence of government subsidies and tax incentives based on the above research methods. It provides relevant policy suggestions for rationally using the two fiscal and tax policy tools, improving the efficiency of industrial subsidy policies and promoting the innovation-driven strategy to promote high-quality economic development.

Finally, to understand the influence of industrial subsidies on the innovation of enterprises with various innovation levels, we empirically explore the heterogeneity of industrial subsidies on firm innovation level and firm performance through RIF regression. In line with previous results, R&D subsidies have an obvious effect on promoting enterprises’ exploratory innovation, and local R&D subsidies have a more significant role in increasing enterprises’ exploratory innovation [16]. Moreover, we find that innovation output plays a part in the mediating role between industrial subsidies and enterprise performance. Still, it cannot test the different impacts of industrial subsidies on the innovation level between enterprises. It also cannot check the difference between industrial subsidies and innovation output on the performance level between enterprises.

Below, we discuss the related theories and previous studies. In Section 2 and Section 3, we propose the theoretical framework of this paper. In Section 4, we show the empirical framework and data sources. In Section 5, we summarize the whole paper and put forward policy implications. Finally, we put forward limitations and future research.

2. Theoretical Background

As for research on the impact of taxes on innovation, Mukherjee et al. [17] studied the impacts of taxes on innovation based on US interstate panel data and found that higher tax rates reduce future innovation output, a conclusion that remains under a series of robustness tests. Cai et al. [18] empirically investigated the influence of tax revenue on enterprise innovation by using data on Chinese manufacturing enterprises. Research has indicated that tax incentives can simultaneously improve the quantity of innovation and innovation quality and play a more obvious role in promoting enterprises with higher financing constraints. Karmaker et al.’s [19] results point out that environmental taxes can improve technological innovation. Uyar et al. [20] found that innovation capacity and intellectual property rights may be two effective mechanisms in alleviating tax evasion. Akcigit et al. [21] produced empirical research based on relevant data concerning enterprises and patent inventors. Their results show that increasing personal income tax and corporate income tax is not conducive to enterprise innovation. Chen et al. [22] showed that China’s preferential tax policies for high-tech enterprises have significantly increased enterprises’ productivity and R&D investment. The results of Labeaga et al. [23] show that the possibility of stopping applications for tax credits has diminished over time. More domestic studies examined how preferential tax policies affect enterprise innovation [24,25,26]. Wan et al. [27] indicated that preferential tax policy has an apparent crowding-in impact on R&D technical efficiency and a pronounced crowding-out impact on business transformation efficiency. Most of these studies support the positive role of tax incentives on enterprise innovation. Scholars in China and foreign scholars have also investigated the positive relationship between enterprise innovation and government subsidies [28,29,30,31]. However, contrary to the results of most research that support tax incentives to promote enterprise innovation, some studies have found a negative effect of government subsidies on innovation. In addition, some scholars have found that a few social factors can also affect corporate innovation. For example, Ober and Kochmańska [32] explored the potential impact of communication within Polish IT firms on their innovation adaptation. Wisdom et al. [33] took a narrative synthesis approach, identifying the theoretical features used to increase innovation. Ramos-Hidalgo et al. [34] examined the impact of the technology model that foreign companies initially adopt when entering China and their subsequent decisions to improve the technology model; their results suggest that innovation as a stimulus for internal change is an effective way to achieve sustainable growth in international markets.

In related research that innovation affects enterprise performance, Qiao et al. [35] studied the influence of industry association network members on the innovation and corporate performance of Chinese SMEs, and their results show that innovation increases the corporate performance of SMEs. Using a sample of 204 small firms and a hierarchical regression model, Chege and Wang [36] pointed out that technological innovation has a positive effect on enterprise performance. Scholars such as Romer [4] and Aghion and Howitt [2] have introduced R&D into endogenous growth models, demonstrating that innovation can promote economic growth. We find that R&D subsidies will be conducive to enterprises’ exploratory innovation, and the effect of local R&D subsidies is more significant. In addition, in highly specialized industrial agglomerations, the positive impact of subsidies will benefit the beneficiaries. Empirical studies by Aghion et al. [37] and Aghion et al. [38] showed that competition ensures economic growth by promoting innovation. Pece et al. [39] explored empirical studies based on data from Central and Eastern European countries, and the research results show a positive effect of innovation on economic growth. Energy consumption has an inverse s-shaped relationship with corporate performance, and the threshold effect of technological innovation is significant.

Santos [40] found that enterprises with higher performance in the pre-intervention stage are more likely to receive the subsidies. Outward foreign direct investment (OFDI) has positive effects on the innovation performance of Chinese SMEs’ subsidiaries, and it is more substantial when the OFDI is directed towards developed rather than emerging countries [41]. The non-R&D subsidy, enterprise size, and age do not directly affect firm-level innovation [42]. Mulier and Samarin [43] evaluated the impact of a pan-European innovation funding program on corporate growth and innovative output and found that the impact of subsidies is highly heterogeneous in sectors with different R&D or knowledge intensity and competition levels.

In related research on domestic innovation affecting enterprise performance, Zhang et al. [44] explored the impact of patent output on economic growth based on provincial panel data from 1985 to 2012 and indicated that the nexus of invention patents and per capita GDP growth rate is U-shaped, whereas the utility model and appearance patents do not have a significant promotional effect. Jia et al. [45] investigated domestic innovation from the micro-enterprise level, based on the data on Chinese listed companies from 2006 to 2010, and showed that innovation significantly promotes economic growth. The authors further confirmed that this impact is non-linear. In general, Gao et al. [16] found that R&D subsidies have a promoting effect on firms’ exploratory innovation, and the effect of local R&D subsidies is more significant. In particular, in highly specialized industrial clusters, the positive impact of subsidies is more beneficial to the beneficiaries [42]. Shao et al. [46] confirmed that government subsidies in the home country negatively regulate the nexus of a firm’s facial recognition technology capability and domestic sales performance.

To sum up, the vast majority of related studies investigated the impact of industrial subsidy policies on firms’ innovation and innovation on enterprise performance, respectively or separately. This cannot effectively test the mechanism brought by industrial subsidy policies to promote enterprise performance. In addition, most previous studies have focused on the horizontal effect, and there is a lack of corresponding studies investigating the gap between enterprises. Research from the perspective of the overall impact effect and the gap between enterprises can identify the effect of industrial subsidies in promoting different enterprises and provide empirical evidence for improving the effectiveness of industrial subsidy policies with limited fiscal and tax resources.

The impact of industrial subsidies and corporate innovation on enterprise performance proves the necessity of this research. Based on the above, our premise is that innovation plays a mediating effect on the nexus between industrial subsidies and corporate performance. Therefore, three hypotheses are proposed:

Hypothesis 1.

Industrial subsidies are positively related to corporate innovation.

Hypothesis 2.

Corporate innovation has a positive effect on corporate performance.

Hypothesis 3.

Industrial subsidies are positively related to corporate performance; industrial subsidies can improve corporate performance by promoting innovation.

3. Research Methods

3.1. Sample and Data Collection

The data we used for estimation came from GuoTai, a CSMAR database. After screening, the final sample used in this paper were the non-balance panel data from listed companies from 2007 to 2017. Considering that innovation is more critical among the manufacturing enterprises, the data used in this article are from the listed manufacturing companies. In addition, the annual loss (negative net profit) samples of listed companies were eliminated.

3.2. Variable Selection and Description

The core explanatory variable Ind selected includes preferential tax and government subsidy ln(Gov). This article measures tax incentives by net profit/operating profit. A more considerable value indicates a higher degree of tax incentives (this article also measured tax incentives by net profit/total profit, but the difference between total profit and operating profit in the data used in this paper was not big, and the estimated results were basically the same). Gov represents the amount of government subsidies to listed companies. The mediating variable is innovation output inno. Referring to most documents, the number of patent applications is represented as the proxy of innovation. Enterprise performance laborp (total revenue of listed companies/the number of employees) is calculated by labor productivity.

The control variables selected include lnL (the L is calculated by the number of enterprise employees), capital input lnK (the K is measured by the net fixed asset value), Age of a listed company, and financial constraints Fin (operating cash flow/total assets)—the larger the number is, the lower the degree of financing constraints is. The asset–liability ratio Lev is measured by the proportion of total liabilities to total assets, the TobinQ of listed companies is calculated by the ratio of market value to total assets, and the dummy variable is ownership Own. When the listed company is recorded as 1, it is a state-owned enterprise; otherwise, it is 0.

3.3. Basic Model

The benchmark regression model used in this paper was the mediation effect model. In addition to the core control variables of industrial subsidies and innovation output level, the influencing factors of the mediation effect model also include enterprise input and other enterprise characteristic indicators. The mediating effect model is set as follows:

$$\begin{array}{l}\ln (labor{p}_{it+1})={\alpha }_{10}+{\alpha }_{11}In{d}_{it}+{\alpha }_{12}\ln L_{it}+{\alpha }_{13}\ln K_{it}\\ +{\alpha }_{14}Fi{n}_{it}+{\alpha }_{15}Le{v}_{it}+{\alpha }_{16}Tobin{Q}_{it}+{\alpha }_{17}Ow{n}_{it}+{\lambda }_{2s}+{\delta }_{2d}+{\eta }_{2t}+u_{2it+1};\end{array}$$

(1)

$$\begin{array}{l}\ln (inn{o}_{it+1})={\beta }_0+{\beta }_{1}In{d}_{it}+{\beta }_2\ln L_{it}+{\beta }_3\ln K_{it}\\ +{\beta }_{4}Fi{n}_{it}+{\beta }_{5}Le{v}_{it}+{\beta }_{6}Tobin{Q}_{it}+{\beta }_{7}Ow{n}_{it}+{\lambda }_{1s}+{\delta }_{1d}+{\eta }_{1t}+u_{1it+1}\end{array}$$

(2)

$$\begin{array}{l}\ln (labor{p}_{it+1})={\alpha }_{20}+{\alpha }_{21}In{d}_{it}+{\alpha }_{22}\ln (inn{o}_{it+1})+{\alpha }_{23}\ln L_{it}+{\alpha }_{24}\ln K_{it}\\ +{\alpha }_{25}Fi{n}_{it}+{\alpha }_{26}Le{v}_{it}+{\alpha }_{27}Tobin{Q}_{it}+{\alpha }_{28}Ow{n}_{it}+{\lambda }_{3s}+{\delta }_{3d}+{\eta }_{3t}+u_{3it+1};\end{array}$$

(3)

Among them, the industry subsidy variable Ind is the key explanatory variable, including preferential Tax and government subsidy Gov. This paper also estimated the two industry subsidy variables separately; inno is the output variable for innovation and laborp is the productivity of labor, and is used to measure corporate performance levels. λ, δ, and η represent industry fixed-effect, regional fixed-effect (provincial fixed-effect in this article), and annual fixed-effect, respectively. Considering the lagging impact of industrial subsidies on enterprise innovation and performance, this paper studied the policy effect of industrial subsidies. It treated this variable in the first phase, weakening the endogenous existence of possible two-way causal problems between industrial subsidies and the dependent variables. Each control variable took a one-period lag.

Model (1) is used to examine the effect of industrial subsidies on enterprise performance but cannot separate and test the mediating effect of industrial subsidies by promoting innovation. Therefore, this paper further constructed Models (2) and (3) to examine this mediating effect. Whereas Model (2) is used to examine the impact of industrial subsidies on enterprise innovation output, Model (3) is a model after the addition of mediating variables to Model (1).

Under Models (1)–(3), the mediating effect of industrial subsidies through affecting innovation output can be tested. The specific steps are as follows: The first step investigates whether industrial subsidies affect enterprise performance according to Model (1); if significant, go to the second step to test whether industrial subsidies affect innovation output according to Model (2). If substantial, enter the third step and add both the industrial subsidy variables and innovation output variables to the model according to Model (3). If the influence of innovation on enterprise performance is significant and if the effects of industrial subsidies are significant, it is a partial mediating effect, and if not, the full mediating impact will be significant.

In addition, under the setting of Model (2) and Model (3), the marginal impact of industrial subsidies on enterprise performance can be expressed as:

$$d\ln (laborp)/dInd={\alpha }_{21}+{\alpha }_{22}{\beta }_1;$$

(4)

Among them, α₂₁ is the direct effect of industrial subsidies affecting enterprise performance (in the direct and mediating effects decomposed here, the direct effect is not a “direct effect” in the absolute sense, but calls the other effects other than the channel of affecting enterprise performance through innovation); α₂₂β₁ is the mediating effect of industrial subsidies, which then affects enterprise performance through innovation output; β₁ is the marginal impact of industrial subsidies on enterprise innovation output; and α₂₂ is the marginal effect of innovation output on enterprise performance.

However, although the mediating effect model can test the effect of industrial subsidies on the enterprise innovation level, industrial subsidies that promote enterprise performance and promote enterprise performance through innovation cannot test the preferential tax difference between enterprise innovation levels. It also cannot test the effect of industrial subsidies and the innovation output difference between enterprise performances. This paper is based on RIF regression, and the RIF regression equation is as follows:

$$\begin{array}{l}v(F)=E_y(RIF(y;v))={\lambda }_0+{\lambda }_{1}Ind+{\lambda }_2\ln L+{\lambda }_3\ln K+\\ {\lambda }_{4}Fin+{\lambda }_{5}Lev+{\lambda }_{6}TobinQ+{\lambda }_{7}Own+{\lambda }_s+{\delta }_d+{\eta }_t+u_{it+1};\end{array}$$

(5)

The respective explanatory variables have the same meaning as those described above, and v can be a variety of statistics characterizing the distributed F(y), including the mean, quantile, Gini coefficient, variance, etc. This is also the advantage of RIF’s return relative to the traditional regression. That is, flexible statistics can be constructed to give a more thorough description of the problem. Considering the error ranges of the RIF, Firpo et al. [47] recommended a consistent estimation of White heterovariance. In this paper, this refers to the number of innovation output levels and labor productivity.

4. Empirical Analysis

4.1. Estimated Results of OLS

In this paper, Models (1)–(3) were first estimated by the OLS estimation method, which investigates the relationship between industrial subsidies, technological innovation, and enterprise performance. The results are represented in

Table 1. Estimated results of the basic model.

	(1)	(2)	(3)	(4)	(5)	(6)
	ln(laborp)	ln(inno)	ln(laborp)	ln(laborp)	ln(inno)	ln(laborp)
Tax	0.247 ***	0.888 ***	0.165 ***
	(3.85)	(7.43)	(2.59)
ln(Gov)				0.115 ***	0.222 ***	0.097 ***
				(14.58)	(16.07)	(12.09)
ln(inno)			0.098 ***			0.088 ***
			(13.88)			(12.38)
lnL	−0.364 ***	0.533 ***	−0.423 ***	−0.450 ***	0.444 ***	−0.496 ***
	(−22.39)	(21.63)	(−24.87)	(−27.22)	(17.56)	(−29.50)
lnK	0.279 ***	0.123 ***	0.271 ***	0.245 ***	0.023	0.247 ***
	(18.30)	(5.76)	(18.19)	(15.40)	(1.03)	(15.75)
Age	0.004 ***	−0.009 ***	0.005 ***	0.004 ***	−0.010 ***	0.005 ***
	(2.66)	(−3.22)	(3.16)	(2.69)	(−3.79)	(3.23)
Fin	0.653 ***	0.702 ***	0.589 ***	0.681 ***	0.777 ***	0.618 ***
	(5.22)	(3.32)	(4.79)	(5.58)	(3.70)	(5.13)
Lev	0.639 ***	−0.107	0.649 ***	0.643 ***	−0.181 **	0.660 ***
	(10.48)	(−1.20)	(10.66)	(10.66)	(−1.98)	(10.88)
TobinQ	−0.016 ***	0.017 *	−0.016 ***	−0.021 ***	0.013	−0.022 ***
	(−3.14)	(1.74)	(−3.45)	(−4.24)	(1.38)	(−4.47)
Own	0.082 ***	0.004	0.081 ***	0.110 ***	0.051	0.105 ***
	(4.18)	(0.12)	(4.22)	(5.64)	(1.39)	(5.53)
Industry/region/years	control	control	control	control	control	control
Sample size	7252	7252	7252	6942	6942	6942
R2	0.386	0.400	0.408	0.427	0.422	0.443

***, **, and * are significant at the 1%, 5%, and 10% significance levels, respectively, and the bracketed values are t-statistics.

. (On the basis of the Cobb–Douglas production function and the Napierian logarithm, these models can be derived in the following ways: ln(Y/L) = lnA + (a − 1)lnL + blnK. Therefore, when the dependent variable is the natural logarithm of labor productivity, the coefficient of the labor input variable lnL is a − 1 rather than the output elastic a of labor input, and using labor productivity as a dependent variable of enterprise performance is more meaningful than total output, which also facilitates the RIF regression analysis below.)

Table 1 (1) shows that the coefficient of the variable Tax was significantly positive, which indicates that moderate tax incentives help improve corporate performance. With the intensity of tax incentives increasing by 0.1, enterprise performance will increase by about 2.47%. Wu et al. [48] also drew the above conclusions using panel data from China. The possible reason is that the implementation of preferential tax policies has raised the disposable funds of enterprises, which may use this part of the funds for technology research and development, production equipment renewal, etc., to further improve their level of performance. However, whether tax incentives are adopted through innovation to promote enterprise performance cannot be tested with column (1).

From column (2) of Table 1, the estimated coefficient of the tax preference variable Tax was significantly positive, which shows that the higher the tax preference, the higher the innovation output level of the enterprise, and improving the degree of tax preference will increase the innovation output of the enterprises. This is because tax preference can transform the positive externalities of product innovation into internal enterprises, increase development benefits, and encourage investment in product innovation [49]. As column (2) of Table 1 shows, the estimation coefficient of the tax preference variable Tax was significantly positive, which indicates that the higher the tax preference, the greater the innovation output level of the enterprise, and improving the degree of tax preference will increase the innovation output of enterprises. Among them, for every 0.1 increase in the intensity of tax incentives, the innovation output of enterprises will increase by about 8.9%. The high estimation coefficient of the tax preference variable Tax in column (1) indicates that the marginal effect of tax preference on innovation output is less than that on enterprise labor productivity. This indirectly shows that the marginal effect of innovation output promoting enterprise labor productivity is negligible.

As shown in column (3) of Table 1, after adding the innovative output variable ln(inno), the estimated coefficient of the tax preferential variable Tax remained positive and significant, but less than for this variable in column (1). The coefficient of the innovation output variable ln(inno) was significantly positive, which indicates that the greater the enterprise innovation output, the higher the enterprise performance level, and appropriately improving enterprise innovation output helps to improve enterprise performance.

According to column (4) of Table 1, the estimated coefficient of the government subsidy variable ln(Gov) was positive and at the 1% significance level, which shows that government subsidies can also increase the performance level and tax incentive for enterprises. For each 5% increase in government subsidies, corporate performance levels will increase by about 0.5%. In Table 1 (6), after the addition of the innovative output variable ln(inno), the government subsidy variable ln(Gov) remained positive and significant, but less than for this variable in column (4). The estimated coefficients of the innovation output variable ln(inno) remained significantly positive, indicating that improving enterprise innovation output helps to improve enterprise performance. Enterprise innovation may change the organizational form of the enterprise, improve the competitiveness of the enterprise, and then increase the performance of the enterprise [50].

Based on the regression results of the above three models and the test steps of the mediating effects model, it can be seen that innovation output plays a partial mediating role between industrial subsidy and enterprise performance; that is, the greater the industrial subsidy, the higher the innovation output of the enterprise, and thus, the higher the level of enterprise performance. The “partial” mediating effect is because industrial subsidies will also affect enterprise performance through channels other than innovation. Based on Model (4), we can calculate that the total marginal effect of tax incentives on enterprise performance is 0.252, whereas the mediating effect is 0.087. The total marginal impact of government subsidies on corporate performance is 0.116, of which the mediating effect is 0.020, accounting for about 16.8%. This shows that the promotion effect of government subsidies and tax incentives on enterprise performance through innovation is manifested as a mediating effect, whose proportion of tax incentives is relatively higher, and the proportion of the mediating effect of government subsidies is relatively low.

4.2. Heterogeneity Test

China has formulated relatively more preferential treatment for high-tech enterprises and SMEs at this stage. The industrial subsidy policies may have significant differences between high-tech and non-high-tech enterprises, large enterprises, and SMEs. The high-tech enterprises in this paper include the following areas: pharmaceutical manufacturing, instrument manufacturing, computers, communication, other electronic equipment manufacturing, and aerospace, railway, ship, and other transportation equipment manufacturing, etc. Enterprises in other industries are classified as non-high-tech enterprises. Therefore, this article tested the heterogeneity. The samples of high-tech enterprises and non-tech enterprises were estimated. The impacts of tax incentives are represented in

Table 2. Differences between high-tech enterprises and non-high-tech enterprises with tax incentives.

	High-Tech Enterprises			Non-High-Tech Enterprises
	ln(laborp)	ln(inno)	ln(laborp)	ln(laborp)	ln(inno)	ln(laborp)
Tax	0.320 ***	1.122 ***	0.206 **	0.197 **	0.759 ***	0.132
	(3.10)	(5.90)	(2.06)	(2.41)	(4.96)	(1.62)
ln(inno)			0.104 ***			0.093 ***
			(10.20)			(9.59)
Controlled variable	Control	Control	Control	Control	Control	Control
Industry/region/years	Control	Control	Control	Control	Control	Control
Sample size	3322	3322	3322	3930	3930	3930
R2	0.362	0.409	0.389	0.429	0.415	0.447

*** and ** are significant at the 1% and 5%, significance levels, respectively, and the bracketed values are t-statistics.

, and the estimates of government subsidies are shown in

Table 3. Differences between high-tech enterprises and non-high-tech enterprises with government subsidies.

	High-Tech Enterprises			Non-High-Tech Enterprises
	ln(laborp)	ln(inno)	ln(laborp)	ln(laborp)	ln(inno)	ln(laborp)
ln(Gov)	0.130 ***	0.254 ***	0.108 ***	0.101 ***	0.195 ***	0.085 ***
	(11.18)	(12.23)	(9.26)	(9.43)	(10.47)	(7.86)
ln(inno)			0.090 ***			0.085 ***
			(8.89)			(8.88)
Controlled variable	Control	Control	Control	Control	Control	Control
Industry/region/years	Control	Control	Control	Control	Control	Control
Sample size	3322	3322	3322	3930	3930	3930
R2	0.362	0.409	0.389	0.429	0.415	0.447

*** is significant at the 1% significance level, and the bracketed values are t-statistics.

.

As shown in Table 2, the impact of tax incentives for high-tech enterprises on innovation output and enterprise performance was more significant than that of non-high-tech enterprises. The marginal effect of innovation output of high-tech enterprises’ performance was also more significant than that of non-high-tech enterprises, but the difference was relatively small. Table 3 shows that the influence of government subsidies on innovation output and enterprise performance of high-tech enterprises was more significant than that of non-high-tech enterprises, which also shows the more substantial role of the innovation output of high-tech enterprises in promoting enterprise performance. In general, both tax incentives and government subsidies suggest that industrial subsidies have a more considerable role in fostering high-tech enterprises.

Furthermore, samples of large enterprises and SMEs were estimated, and the results are shown in

Table 4. Differences between high-tech enterprises and non-high-tech enterprises in large enterprises.

	Large Enterprises			Small- and Medium-Sized Enterprises
	ln(laborp)	ln(inno)	ln(laborp)	ln(laborp)	ln(inno)	ln(laborp)
Tax	0.218 **	0.649 ***	0.152 *	0.209 **	0.949 ***	0.156 *
	(2.55)	(3.82)	(1.78)	(2.35)	(5.82)	(1.75)
ln(inno)			0.108 ***			0.059 ***
			(10.30)			(7.01)
Controlled variable	Control	Control	Control	Control	Control	Control
Industry/region/years	Control	Control	Control	Control	Control	Control
Sample size	3254	3254	3254	3998	3998	3998
R2	0.394	0.446	0.421	0.420	0.233	0.428

***, **, and * are significant at the 1%, 5%, and 10% significance levels, respectively, and the bracketed values are t-statistics.

and

Table 5. Differences between high-tech enterprises and non-high-tech enterprises in SMEs.

	Large Enterprises			Small- and Medium-Sized Enterprises
	ln(laborp)	ln(inno)	ln(laborp)	ln(laborp)	ln(inno)	ln(laborp)
ln(Gov)	0.107 ***	0.229 ***	0.086 ***	0.072 ***	0.166 ***	0.063 ***
	(8.53)	(10.30)	(6.81)	(7.62)	(9.38)	(6.50)
ln(inno)			0.095 ***			0.058 ***
			(9.25)			(6.93)
Controlled variable	Control	Control	Control	Control	Control	Control
Industry/region/years	Control	Control	Control	Control	Control	Control
Sample size	3125	3125	3125	3817	3817	3817
R2	0.436	0.465	0.457	0.462	0.244	0.470

*** is significant at the 1% significance level, and the bracketed values are t-statistics.

, respectively. Among them, the average fixed-asset pair value in the sample used was about 12.78. In this article, the fixed-asset pair value was more significant than, or equal to, the average (i.e., 12.78), which was defined as large enterprises, and the remaining enterprises as SMEs.

Table 4 shows that for both large businesses and SMEs, the mediating effect of tax incentives to promote enterprise performance through innovation was manifested as a partial mediating effect. Column (1) and column (2) show that tax incentives significantly promoted the output of SME innovation, but smaller incentives were given to SMEs than large enterprises. The results in column (3) indicate that the direct effect between tax incentives promoting the performance of SMEs and large enterprises was small. Although tax incentives boosted SMEs’ innovation output, SMEs’ ability to turn the innovative output into actual output was relatively weak. As a result, tax incentives to promote the performance of big enterprises and SMEs were not significantly different.

Table 5 shows that government subsidies that improve enterprise performance through innovation were partially mediated among large enterprises and SMEs. The influence of government subsidies on large enterprises in improving innovation output and enterprise performance as better than that of SMEs. This indicates that government subsidies play a more considerable role in large enterprises, unlike tax incentives.

4.3. Robustness Test

In this paper, we tested for robustness in two aspects: selecting innovative output variables and the robustness test based on the choice of lag period. First, the innovation output variable chosen in this research was the number of patent applications. Still, some researchers believe that the number of patent licenses is a more reasonable innovation output index than the number of patent applications. Second, the estimation of the model took a one-period lag. However, taking a different lag period may have had significant effects on the estimated results in this paper. Therefore, this article also took a two-period lag estimation. This could test whether the choice of lag period significantly affected the results and further weakened possible endogenous problems caused by bidirectional causality between tax incentives among innovation output and enterprise performance. The robustness test results show that neither the substitution innovation output index nor the substitution lag period substantially impacted the basic conclusions of this paper. The robustness results show that the conclusions of the benchmark regression in this paper are reliable.

4.4. Further Study: RIF Quantile Regression

4.4.1. The Impact of Industrial Subsidies on Innovation Output

According to Model (2), the RIF method was used for the estimation. The quantiles—including 0.1, 0.3, 0.5, 0.7, and 0.9—and the variance were selected as the statistics for analysis. The effects of tax incentives and government subsidies were estimated separately, with the estimates shown in

Table 6. Effects of tax incentives on innovation output: RIF quantile regression results.

Statistics	Quantile					Variance
	0.1	0.3	0.5	0.7	0.9
Tax	0.786 ***	0.837 ***	0.925 ***	0.838 ***	0.915 ***	0.115
	(3.00)	(4.65)	(6.26)	(5.72)	(3.79)	(0.38)
Control	Control	Control	Control	Control	Control	Control
Industry/region/years	Control	Control	Control	Control	Control	Control
Sample size	7252	7252	7252	7252	7252	7252
R2	0.078	0.181	0.248	0.276	0.250	0.115

*** is significant at the 1% significance level, and the bracketed values are t-statistics.

and

Table 7. Effect of government subsidies on innovation output: RIF quantile regression results.

Statistics	Quantile					Variance
	0.1	0.3	0.5	0.7	0.9
ln(Gov)	0.116 ***	0.167 ***	0.192 ***	0.233 ***	0.321 ***	0.241 ***
	(3.71)	(7.58)	(11.23)	(14.14)	(12.87)	(5.92)
Control	Control	Control	Control	Control	Control	Control
Industry/region/years	Control	Control	Control	Control	Control	Control
Sample size	7252	7252	7252	7252	7252	7252
R2	0.078	0.181	0.248	0.276	0.250	0.115

*** is significant at the 1% significance level, and the bracketed values are t-statistics.

, respectively.

According to Table 6, for the five sub-loci of 0.1, 0.3, 0.5, 0.7, and 0.9, the coefficients of Tax preference variable were 0.786, 0.837, 0.925, 0.838, and 0.915, respectively, and the estimated coefficients of each sub-locus were approximately between 0.8 and 0.9, showing an overall upward trend. It can be seen that there were specific differences in the effect of tax incentives on the patents of listed companies under the innovation output level. When the variance was used as a statistic for RIF estimation, the coefficient of the preferential variable Tax was positive. This shows that tax incentives can somewhat promote the difference in innovation output level among enterprises. However, the effect was not apparent and did not pass the significance test.

From Table 7, for the five quantiles of 0.1, 0.3, 0.5, 0.7, and 0.9, the coefficients of the government subsidy variable ln(Gov) were 0.116, 0.167, 0.192, 0.233, and 0.321, respectively, showing a faster rising trend in the estimated coefficients at each quantile. It can be seen that the greater the innovation output level, the higher the effect of government subsidies on their innovation output, which can widen the gap in the innovation output level among enterprises. Compared with tax incentives, the effect of government subsidies on different innovation capacity enterprises was more diverse. According to the last column in Table 6, the coefficient of the government subsidy variable ln(Gov) was positive and at the 1% significance level, further indicating the significant facilitation influence of government subsidies on the gap in innovation output levels among enterprises.

4.4.2. Impact of Industrial Subsidies on Enterprise Performance

We continued to study the effect of enterprise performance from tax incentives and government subsidies using RIF quantile regression based on Model (1), using the same quantile as described above. Estimates are represented in

Table 8. Effect of tax incentives on corporate performance: RIF quantile regression results.

Statistics	Quantile
	0.1	0.3	0.5	0.7	0.9
Tax	0.445 ***	0.225 ***	0.201 ***	0.142	0.133
	(4.16)	(2.97)	(2.84)	(1.62)	(0.89)
Control	Control	Control	Control	Control	Control
Industry/region/years	Control	Control	Control	Control	Control
Sample size	7252	7252	7252	7252	7252
R2	0.170	0.237	0.256	0.261	0.200

*** is significant at the 1% significance level, and the bracketed values are t-statistics.

and

Table 9. Effect of government subsidies on corporate performance: RIF quantile regression results.

Statistics	Quantile
	0.1	0.3	0.5	0.7	0.9
Ln(Gov)	0.105 ***	0.114 ***	0.098 ***	0.100 ***	0.109 ***
	(9.09)	(13.04)	(11.65)	(9.71)	(5.84)
Control	Control	Control	Control	Control	Control
Industry/region/years	Control	Control	Control	Control	Control
Sample size	6942	6942	6942	6942	6942
R2	0.180	0.264	0.278	0.274	0.214

*** is significant at the 1% significance level, and the bracketed values are t-statistics.

, respectively.

According to Table 8, when a quantile was located between 0.1 and 0.5, the estimation coefficient of the tax preference variable Tax was significantly positive, and tax preference substantially improved the enterprise performance level. Especially at the 0.1 quantiles, the estimated coefficient was significant, at 0.448. Especially at 0.1, there was a substantial gap in the 0.3–0.5 quantiles, whereas at the 0.7–0.9 quantiles, the estimated coefficient was relatively small. This shows that tax incentives positively impacted low-performance enterprises and had less impact on high-performance enterprises, which demonstrates that implementing preferential tax policies for enterprises with a low-performance level can improve the enterprise performance level more significantly. This may be because weak enterprises lack funds and face more significant financing constraints and capital demand, so tax incentives can alleviate these problems, significantly improving the enterprises’ performance. However, the degree of financing constraints among strong enterprises is low, and the capital needs are relatively low, so the promotion effect of tax incentives will be reduced.

According to Table 9, when the quantile was between 0.1 and 0.9, the estimated coefficient of the government subsidy variable ln(Gov) was positive and at the 1% significance level. Under each quantile, government subsidies significantly increased the enterprise’s performance level. The estimated coefficients of the government subsidy variable ln(Gov) varied little among the quantiles, which were all around 0.1, indicating that there is no obvious difference in the impact of government subsidies on the performance gap between enterprises as the enterprise performance level improves.

4.4.3. Impact of Industrial Subsidies on Enterprise Performance: Mediating Effect Mechanism

The above quantile estimation results analyzed the effects of industrial subsidies on innovation output and industrial subsidies on enterprise performance. Then the RIF estimation was represented with Model (3). RIF as used to test the direct effects of industrial subsidies and innovation affecting enterprise performance between quantiles, and the regression results are shown in

Table 10. Tax incentives, innovative output, and enterprise performance: RIF quantile regression results.

Statistics	Quantile
	0.1	0.3	0.5	0.7	0.9
Tax	0.380 ***	0.159 **	0.132 *	0.064	0.032
	(3.56)	(2.09)	(1.86)	(0.72)	(0.21)
ln(inno)	0.077 ***	0.078 ***	0.083 ***	0.094 ***	0.120 ***
	(7.51)	(9.56)	(10.66)	(10.13)	(7.55)
Control	Control	Control	Control	Control	Control
Industry/region/years	Control	Control	Control	Control	Control
Sample size	7252	7252	7252	7252	7252
R2	0.177	0.247	0.268	0.273	0.208

***, **, and * are significant at the 1%, 5%, and 10% significance levels, respectively, and the bracketed values are t-statistics.

and

Table 11. Government subsidies, innovative outputs, and enterprise performance: RIF quantile regression results.

Statistics	Quantile
	0.1	0.3	0.5	0.7	0.9
ln(Gov)	0.089 ***	0.100 ***	0.083 ***	0.083 ***	0.083 ***
	(7.55)	(11.23)	(9.62)	(7.93)	(4.44)
ln(inno)	0.074 ***	0.066 ***	0.073 ***	0.081 ***	0.123 ***
	(6.95)	(7.86)	(9.01)	(8.56)	(7.60)
Control	Control	Control	Control	Control	Control
Industry/region/years	Control	Control	Control	Control	Control
Sample size	6942	6942	6942	6942	6942
R2	0.187	0.270	0.287	0.283	0.222

*** is significant at the 1% significance level, and the bracketed values are t-statistics.

, respectively.

As shown in Table 10, when the quantile gradually changed from 0.1 to 0.9, the estimated coefficient of the tax preferential variable Tax was positive and showed a decreasing trend. Among them, the estimated coefficients of the 0.1, 0.3, and 0.5 quantiles passed the significance test, and the estimated coefficient at the 0.7 and 0.9 quantiles failed it. This indicates that the direct effect of tax incentives on corporate performance when a quantile is lower, and not on corporate performance at higher quantiles, is similar to the coefficients shown in Table 8. However, the coefficients in Table 10 are relatively small due to part of the effects reflected in terms of the mediation effects. This shows that the direct impact of tax incentives on enterprise performance can reduce the gap in the level of performance among enterprises. The innovative output variable ln(inno) significantly positively estimated the coefficients at different quantiles. Specifically, as the quantile improved, the estimated coefficients also increased. This shows that innovation output significantly impacts enterprises with a high-performance level, whereas the promotional effect on enterprises with a low performance level is relatively small; that is, innovation can widen the performance gap among enterprises.

As seen in Table 11, the estimated coefficient of the government subsidy variable ln(Gov) at each quantile was positive and at the 1% significance level. Under each quantile, government subsidies significantly promoted enterprises’ performance level. The government subsidy variable ln(Gov) estimates did not vary widely across quantiles, which were all around 0.09. With the improvement of the enterprise performance level, there were still no significant gaps in the direct effect of government subsidies on the performance gap between enterprises. The innovation output variable ln(inno) significantly positive estimated the coefficients at different quantiles, and innovation widened the performance gap between enterprises. We can see that innovation widened the performance gap among enterprises, and there were some differences in the estimation coefficients due to differences in the sample size.

To sum up, with the change in quantiles, there were obvious differences in the influence of preferential government tax subsidies on innovation output and enterprise performance. With the increase in quantiles, there were more and more tax incentives, government subsidies for innovation output, and tax incentives for enterprise performance. In contrast, the influence of government subsidies on enterprise performance was generally unchanged. After introducing the innovation output variable and increasing quantiles, the impact of tax incentives and government subsidies’ enterprise performance still showed similar characteristics.

5. Conclusions and Policy Implications

To increase the innovation level of domestic enterprises and the implementation of innovation-driven strategy, China has introduced many industrial subsidy policies for enterprise innovation. However, the Fifth Plenary Session of the 19th CPC Central Committee put forth that China’s current innovation capacity does not meet the requirements of high-quality development. Therefore, achieving sustained high-quality development is innovation driven. This research investigated the influence of industrial subsidy policies on innovation and innovation on enterprise performance from the perspectives of tax incentives and government subsidies. It provides relevant policy enlightenment on how to adopt effective industrial policies to promote better independent innovation and high-quality economic development. Based on the data from 2007 to 2017 on Chinese-listed manufacturing companies, this paper constructed a mediating effect model to investigate the two policies that offer tax incentives and government subsidies to promote enterprise performance and mediate the promotion of enterprise performance through innovation. This article used RIF regression to examine the different tax incentives and government subsidies allocated to encourage innovation and corporate performance across quantiles. The major results obtained in this article are as follows:

(1)

Tax incentives and government subsidies have promoted innovation that has improved corporate performance. Innovation plays a partial mediating effect in the nexus of tax incentives, government subsidies, and enterprise performance, and the partial mediating effect accounts for about 34.5% and 16.8%, respectively. This shows that the mediating impact is more significant when tax incentives promote corporate performance than government subsidies.

(2)

Both tax incentives and government subsidies played a considerable role in improving innovation and performance among high-tech enterprises, indicating that implementing industrial subsidy policies for high-tech enterprises can achieve beneficial results. There was little difference between the direct impacts of tax incentives on the performance of SMEs and large enterprises. Although tax incentives had a more significant effect on improving the innovation output of SMEs, the ability of these SMEs to transform innovation output into actual output was relatively weak, resulting in little difference between the performance of large enterprises and SMEs’ performance as a result of tax incentives. However, government subsidies showed better characteristics in promoting large enterprises.

(3)

With the increase in quantiles, innovation output increasingly promoted enterprise performance, which indicates that the positive effect of innovation output on high-quantile enterprise performance is higher than that of low-quantile enterprise performance. The higher the enterprise performance level, the greater the promotion effect of innovation on enterprise performance, which significantly widens the performance gap between enterprises. With the increase in quantiles, the impact of tax incentives and government subsidies became increasingly significant in terms of innovation output. The effect of tax incentives on enterprise performance became less and less. In contrast, the impact of government subsidies on enterprise performance was generally unchanged. After introducing the innovation output variable and increasing quantiles, tax incentives and government subsidies’ direct effect on enterprise performance was similar.

We also put forward some policy implications: (a) Since the intermediary effect of tax incentives to promote enterprise performance through innovation accounts for about 34.5%, the intermediary effect of government subsidies in promoting enterprise performance through innovation reaches 16.8%. This shows that the innovative implementation of appropriate tax incentives and government subsidy policies for enterprises and improving the market transformation capacity of scientific and technological achievements is an essential channel for enhancing the performance of scientific and technological initiatives. This shows that implementing preferential tax and government subsidy policies to promote enterprise innovation and thus improve market transformation through scientific and technological achievements is an essential channel for improving enterprise performance. (b) Tax incentives and government subsidies will play a more obvious role in promoting high-tech enterprises’ innovative output and performance. Therefore, more tax incentives and government subsidy policies for high-tech enterprises can be implemented, which will improve their promotional effect on innovative output and economic growth in the economy. (c) There is no significant difference in the role of tax incentives in promoting the performance of large enterprises and SMEs. However, government subsidies have a more substantial role in increasing the performance of large enterprises, so the scale of enterprises should not be used as a selection standard for the government to implement preferential tax policies. Therefore, government subsidies tend to support large-scale enterprises. (d) With the increase in quantiles, the impact of tax incentives on enterprise performance became less and less, whereas the influence of government subsidies on enterprise performance was generally unchanged. Therefore, more preferential tax policies can be implemented for low-performing enterprises (low per capita primary business income). Enterprise performance should not be used as a selection standard to enforce government subsidy policies.

6. Limitations and Future Research

Although we fully measured the impact of tax incentives and government subsidies on enterprise performance, there are still some research limitations that can be further explored in the future. First, the research object of this paper is Chinese manufacturing enterprises, which has limitations. In the future, this achievement can be extended to various industries in China and abroad. Second, this paper only selected the mediating variables of innovation, and we can explore multiple paths in the future to help enterprises improve their performance effectively. Third, the model settings in this paper are relatively simple, and the introduction of spatial models and dynamic models can be considered in the future.