Can Government Subsidies Promote the TFP of Enterprises? The Mediating Effect of R&D Decisions

Abstract

This article empirically analyses the effect of government subsidies on total factor productivity (TFP) based on the data of listed manufacturing companies in China. The results indicate that government subsidies increase total productivity directly as well as indirectly by increasing R&D investment. The positive effect of government subsidies on TFP is higher in non-state-owned enterprises (non-SOEs) than in state-owned enterprises (SOEs), higher in central SOEs than local SOEs and higher in enterprises with lower rather than higher TFP. Furthermore, the mediating effects of R&D decisions also differ among different enterprises. Therefore, the government should implement differentiated subsidy policies to promote enterprises’ TFP.

Keywords

Government subsidies enterprise R&D decision total factor productivity mediating effect

Introduction

For a long time, the extensive development mode of the manufacturing industry has not only promoted the rapid growth of China’s economy but also made macroeconomic development face increasingly severe resource and environmental constraints. The manufacturing industry in China is still in the middle and low end of the global industrial chain (Young, 2003). Moreover, China’s manufacturing industry not only has an advantage in terms of scale and labour costs but also a disadvantage regarding quality, science and technology, and efficiency. Although China is a large manufacturing centre, it is not a powerful one. Enterprises are the micro-foundation for industrial development. To promote China’s manufacturing industry to the high end of the global value chain, we must cultivate more world-class enterprises.

The key to cultivate world-class enterprises is to take innovation and technological progress as the core driving force to improve the quality and efficiency of enterprise development. In recent years, the deepening of China’s economic reforms has released the development power of enterprises. As a result, the total factor productivity (TFP) of enterprises has increased significantly (Brandt et al., 2017), which is inseparable from enterprises’ efforts to change the management mode and the development and application of new technologies, which in turn, require the support of the government and other external forces. Government support can promote the TFP of enterprises. On one hand, the promotion effect of government subsidies on TFP mostly takes effect through technological innovation (Wieser, 2005). The externality characteristics of R&D activities and efficiency improvement, as well as the inherent weakness of market mechanisms in solving external problems, lead to internal dependence on government support in the process of transformation. Government subsidies can make up for the weak part of market mechanisms and solve externality problems that enterprises face by constructing a dual-track resource allocation mode of ‘market mechanism + government’. Through a series of institutional arrangements and financial subsidies, the externalities of enterprise innovation and other production and operation activities can be internalised, effectively making up for the shortcomings of market mechanisms and improving the R&D incentives of enterprises (Bernini & Pellegrini, 2011). For example, Ren and Lv (2014) asserted that government subsidies committed to promoting innovation and development of enterprises and improving business performance are conducive to easing financing constraints, increasing scale efficiency and promoting TFP. Using the panel data of Belgian firms, Sergant and Van Cayseele (2019) also found that government subsidies would ease the financing constraints of enterprises and thus improve TFP.

However, some scholars have proposed that government subsidies negatively affect TFP (Yan & Yu, 2017). Howell (2017) pointed out that government subsidies have a negative impact on the production efficiency of the high- and low-tech industry sectors. Guan and Lansink (2016) found that the TFP of subsidised enterprises is generally lower than enterprises whose subsidies have been cancelled. Mary (2013) and Baráth et al. (2020) examined the impact of agricultural subsidies on farm TFP in France and Slovenia respectively, and found that agricultural subsidies are difficult to promote or even inhibit the improvement of farm TFP. Dvouletý and Blažková (2019) also found that government subsidies inhibit the improvement of TFP of Czech food enterprises. There are many reasons why government subsidies fail to improve the TFP of enterprises: the inconsistent R&D preferences between the government and enterprises, information asymmetry factors (e.g., effective information hiding and false signal transmission when enterprises apply for funding), the public property right attribute of government subsidies, the inefficiency of government decision-making and the imperfect characteristics of subsidy policies, the lack of tracking evaluation and the supervision mechanism of subsidy policies. Government subsidies cannot promote enterprises’ R&D and they produce a crowding out effect, making enterprises dependent on the policy and inhibiting the improvement of TFP (Yu, 2011; Xiao & Lin, 2014). The excessive attachment of government to the ‘sunk cost’ of zombie enterprises and the subsidy policies implemented to ‘salvage sunk costs’ also causes enterprises to suffer from subsidy dependence, resulting in a vicious circle in which, the more the subsidies are provided, the weaker the enterprises become. This phenomenon reduces the effect of subsidy policies and hinders the improvement of TFP. In addition, the dual attributes of Homo politicus and Homo economicus of government officials also make local officials pursue political and economic interests at the same time. In the process of subsidising enterprises, local officials have self-interested investment preferences, emphasising production and ignoring innovation, as well as rent-seeking and corrupt behaviours, which inhibits the effect of government subsidies, distorts government subsidy implementation and even reduces business efficiency and the TFP of enterprises (Weingast, 2009). In addition, based on the data of Italian enterprises, Colombo et al. (2011) found that in the absence of a good market competition environment, government subsidies can hardly promote the improvement of enterprises’ TFP effectively.

From the discussion above, we can conclude that controversy still exists about how government subsidies affect enterprises’ TFP. The reasons for the divergence of views are various, such as different measurement methods and indicators of the TFP of enterprises, sample selection differences, differences in empirical analysis methods and the screening methods of many confusing factors affecting TFP. In contrast to previous studies, this study takes manufacturing listed companies from 2008 to 2017 as research samples and empirically tests the impact of government subsidies on the TFP of enterprises and its mechanism based on a theoretical analysis. The conclusions have important implications for the formulation of subsidy policies regarding the manufacturing industry.

The marginal contribution of this study mainly concerns the following aspects. First, it supplements the relevant research on the effect of government subsidies on enterprise TFP. On one hand, previously, scholars used the Industrial Enterprise Database for their research samples. Although the database has many advantages, such as a large sample size, there are also some problems, such as the confusion of sample matching, abnormal index size and obvious measurement errors. On the other hand, the database contains many enterprises that differ greatly in terms of scale, financing capacity and R&D levels. When sufficient and detailed investigations on enterprise heterogeneity are lacking, the conclusions do not have a strong reference value for the formulation of subsidy policies for some specific types of enterprises. Therefore, this study features a detailed analysis of the research in this field based on the data of listed manufacturing companies, and the conclusion has more guiding significance for the formulation of subsidy policies for such companies. Second, this article discusses the mechanism of government subsidies affecting enterprises’ TFP, which provides theoretical guidance for the government to formulate reasonable subsidy measures. It is important to study the effect of government subsidies, explore the mechanism of government subsidies, and investigate why and how they can promote or inhibit the increase of the TFP of enterprises to provide a theoretical basis for the government to improve policy measures and subsidy programs. Therefore, for this study, we constructed a mediating effect model to analyse the mechanism that determines how government subsidies affect TFP. Third, this article focuses on the influence of government subsidies on the TFP of enterprises with different types of ownership and the differentiation characteristics of its mechanism, with an analysis of the reasons for the differentiation mechanism that enriches the research in this field.

Theoretical Analysis and Hypotheses

The Impact of Government Subsidies on the TFP of Enterprises

The development of enterprises requires the support of the government, China emphasises the important role of government power in the cultivation of world-class enterprises. As the most direct and universal form of government support for enterprises, subsidies have a significant impact on the operation and development of enterprises. First of all, government subsidies are a direct driving force for enterprises to form economies of scale. Enterprises inevitably encounter financing constraints in the production process (Benito and Hernando, 2007). The increasing difficulty of securing external financing and ever-higher financing costs make it difficult for enterprises to make optimal investment decisions, which limits the improvement of TFP (Ren & Lv, 2014). Government subsidies can ease the financing constraints of enterprises, fill gaps in funding, optimise investment decisions, meet the capital needs of scale expansion and promote the formation of scale economies and the improvement of TFP (Chen et al., 2014). Second, government subsidies can guide the direction of enterprises’ investments and optimise investment structure. The government has the total information advantage and co-ordination function in regional economic development and about potential market demand, and can judge the enterprise’s potential to become a world-class enterprise from a long-term perspective. It can formulate scientific subsidy policies based on the efficient processing of various pieces of economic information, considering the industrial development foundation, potentiality and the opportunities of enterprises and industries (Li et al., 2018). Government subsidies have a certain guiding role for enterprise investment, that is, they can lead the direction of enterprise investment to a certain extent, play a screening role, reduce the opportunity costs of enterprise market information collection, processing and the risk of decision-making errors, guide enterprises to optimise the allocation of internal resources and improve the timeliness and accuracy of enterprise investments. Finally, government subsidies guide the rational allocation of social resources among different enterprises. Subsidised enterprises can receive more financial support in a short time and improve their investment efficiency and TFP.

Due to the existence of information asymmetry and other factors, financial institutions often need to collect a great deal of information to lend to enterprises, which have relatively high information collection costs, including time and economic costs. To reduce information collection costs, on one hand, financial institutions tend to cater to the government’s investment preference. Government subsidies provide a kind of implicit guarantee for enterprise financing, which makes financial institutions more inclined to reduce the credit threshold for government-supported projects. On the other hand, the government often extends subsidies to enterprises with market prospects and development potential, sending out an economic guiding signal to financial institutions, such as banks, and helping financial institutions quickly identify enterprises with strong abilities and strong market prospects. This process improves the allocation efficiency of credit resources, shortens the approval time of enterprise credit financing, and increases investment efficiency (Wu, 2017). Based on the analysis above, we formulated the following hypothesis.

H1: Government subsidies can promote the total factor productivity (TFP) of enterprises.

The Mediating Effect of R&D Decisions

Technological progress is the most important driving force to improve the TFP of enterprises, and R&D and technological innovation are the core factors to promote technological progress. The impact of enterprise R&D investment on TFP has been examined in depth in previous studies, which widely agreed with the conclusion that R&D promotes the improvement of TFP (Cheng, 2017). Government subsidies may also indirectly promote the TFP of enterprises by influencing the R&D decisions of enterprises (Dong & Seo, 2013), that is, the decision to invest in R&D activities related to technology, craft and standards in the process of enterprise operation, which is mainly reflected in changes in R&D investment. The mediating effect of R&D decisions is that government subsidies have a significant impact on R&D investment and then indirectly affect the TFP of enterprises. Theoretically, the mediating effects of R&D decisions are mainly reflected in the following aspects. First, R&D subsidies constitute an important part of government subsidies, which can alleviate the financial constraints that enterprises face and improve their R&D investment capacity. Government R&D subsidies are also direct components of enterprise R&D expenditures and, to a certain extent, enrich the R&D funds that enterprises require. As a kind of implicit guarantee, government subsidies can transmit a positive signal to the market, which can guide external capital to concentrate on subsidised enterprises, improve the ability and level of R&D and then promote the productivity of enterprises (Ren & Lv, 2014). Second, government subsidies not only directly improve the scale of enterprise R&D but can also effectively encourage enterprises to engage in innovative activities, which have a crowding effect on R&D investment. Due to the externality effect of R&D activities and the natural weakness of market mechanisms in dealing with externalities, it is difficult for enterprises to achieve an optimal level of R&D investment. Government subsidies can solve the problems of enterprise innovation inertia that externalities cause, effectively reduce the cost of R&D activities, encourage enterprises to increase R&D investment (Lin, 2017), enhance companies’ innovation abilities and increase the TFP of enterprises (Ren & Lv, 2014). Third, government subsidies can guide the direction of enterprises’ R&D investments, reduce the cost and risk of market information collection, shorten decision-making times, improve the efficiency of R&D investments and promote TFP (Radas et al., 2015). As mentioned above, the government has the advantage of total information, allowing it to efficiently identify R&D fields with development potential and guide enterprises to gather R&D resources in those fields in the form of subsidies. In this way, the government can improve the pertinence and foresight of enterprise R&D activities, promote the quality of enterprises’ R&D, and positively impact TFP (Li et al., 2018). Based on the analysis above, we formulated the following hypothesis.

H2: Government subsidies can promote the expansion of the R&D scale and the improvement of R&D efficiency, as well as indirectly promote total factor productivity (TFP).

Research Design

Model Construction

This study attempts to explore the influence of government subsidies on TFP and the mediating effect of R&D decisions. First, to test H1, that is, to test the comprehensive impact of government subsidies on the TFP of enterprises, we constructed the basic model shown in Equation (1).

\begin{matrix} \ln T F P_{i t} = α_{0} + α_{1} \ln ​ S u b_{i t} + α_{2} A g e_{i t} + α_{3} S o e_{i t} + α_{4} M t b_{i t} + α_{5} L e v_{i t} + \\ α_{6} M a n_{i t} + α_{7} S c a_{i t} + α_{8} C f_{i t} + α_{9} O w n_{i t} + μ_{y e a r} + μ_{i n d u s t r y} + ε_{i t} \end{matrix}

(1)

In Formula (1), lnTFP represents the log value of the TFP of an enterprise and is the explained variable. lnSub is the logarithm of the proportion of government subsidies in the total assets of enterprises and is the core explanatory variable. If the regression coefficient of government subsidy is significant, it indicates that the government subsidy has a significant influence on the TFP of enterprises. For reference (Cheng, 2017), we selected enterprise age (Age), enterprise nature (Soe), growth (Mtb), financial leverage (Lev), management cost (Man), enterprise size (Sca), cash flow (Cf), ownership concentration (Own) and other variables as control variables of the multiple regression model. In addition, time dummy variables (μ_year) and industry dummy variables (μ_industry) were added to control the time fixed effect and industry fixed effect of the regression model.

Under the premise a₁ of significance, to test H2, that is, to test whether there is a mediating effect of enterprise R&D decisions, based on the method of Baron and Kenny (1986), we further constructed the multiple regression model displayed in Equations (2) and (3).

\begin{matrix} \ln R & D_{i t} = β_{0} + β_{1} \ln S u b_{i t} + β_{2} A g e_{i t} + β_{3} S o e_{i t} + β_{4} M t b_{i t} + β_{5} L e v_{i t} + \\ β_{6} M a n_{i t} + β_{7} S c a_{i t} + β_{8} C f_{i t} + β_{9} O w n_{i t} + μ_{y e a r} + μ_{i n d u s t r y} + ε_{i t} \end{matrix}

(2)

\begin{matrix} \ln T F P_{i t} = γ_{0} + γ_{1} \ln S u b_{i t} + γ_{2} \ln R & D_{i t} + γ_{3} A g e_{i t} + γ_{4} S o e_{i t} + \\ γ_{5} M t b_{i t} + γ_{6} L e v_{i t} + γ_{7} M a n_{i t} + γ_{8} S c a_{i t} + γ_{9} C f_{i t} + \\ γ_{10} O w n_{i t} + μ_{y e a r} + μ_{i n d u s t r y} + ε_{i t} \end{matrix}

(3)

Equation (2) examines the impact of government subsidies on enterprises’ R&D decisions, where lnR&D represents enterprises’ R&D decisions and is the explained variable. Compared with Equation (1), Equation (3) includes the intermediary variable of enterprise R&D decisions. From the coefficient of lnSub in Formula (2) and the coefficient of lnR&D in Formula (3), we can judge whether the intermediary mechanism of enterprise R&D decision exists or not. If both β₁ and γ₂ pass the significance test, there is a mediating effect of government subsidies on TFP by influencing enterprises’ R&D decisions, and its effect is β₁ × γ₂. At this point, if the regression coefficient γ₁ in Equation (3) remains significant, then the R&D decision of the enterprise is part of the intermediary variable. In other words, while the government subsidy directly affects the TFP of the enterprise, it also indirectly affects TFP by influencing the government’s R&D decision. If γ₁ in Equation (3) is not significant, then the R&D decision of the enterprise is a complete intermediary variable, indicating that the government subsidy has no direct impact on the TFP of the enterprise but only indirectly affects the TFP by influencing the R&D decision of the enterprise. If β₁ or γ₂ are not significant, the mediating effect of enterprise R&D decision does not exist.

Variable Setting and Data Characteristics

The TFP of enterprises is the explained variable in this article. Its main measurement methods are frontier analysis and non-frontier analysis. The measurement methods in the estimation process are mainly parametric method and semi-parametric method. The former is mostly used at the macro level, while the semi-parametric method is widely used in the micro field. We set the C–D production function as shown in Equation (4):

Y_{i t} = A_{i t} L_{i t}^{α} K_{i t}^{β}

(4)

where, Y represents the output, L represents the labour input, K represents the capital input, and A represents the TFP. Taking logarithm on both sides of Equation (4), we can obtain the bilateral logarithmic model:

\ln Y_{i t} = \ln A_{i t} + α \ln L_{i t} + β \ln ​ K_{i t} + μ_{i t}

(5)

By the estimation of Equation (5), the TFP of enterprises can be obtained. For the selection of estimation methods, among the semi-parameter methods, compared with the traditional ordinary least squares (OLS) regression method, the observed versus predicted (OP) regression method can better solve the simultaneity bias, selectivity bias (Olley & Pakes, 1996) and endogeneity problems in the model, and it is better than the LP method (Levinsohn & Petrin, 2003), which selects intermediate variables as instrumental variables for estimation. For this reason, we chose the OP method to calculate the TFP of enterprises. In terms of the selection of input–output indicators, we drew on previous studies (Giannetti et al., 2015) and selected the number of employees and the scale of capital expenditure as the labour input and capital input, selected the main business income as the enterprise output.

Government subsidy is the core explanatory variable of this article and the government subsidy level is represented by the logarithm of the proportion of the government subsidy amount in the total assets of an enterprise in its non-recurring profit and loss. The enterprise R&D decision is the intermediary variable in this article and the R&D investment is represented by the logarithm of R&D expense in the proportion of the total assets of the enterprise.

In the setting of control variables, the enterprise age is expressed by the difference between the different years and the year of establishment. The nature of enterprises was controlled using a dummy variable, that is, state-owned enterprises (SOEs) are 1, and non-state-owned enterprises (non-SOEs) are 0. The growth of enterprises was measured by the year-on-year growth rate of net assets in that year. The financial leverage is expressed by the asset–liability ratio and was measured by the ratio of the total liabilities and total assets of the enterprise. Management cost was assessed by the proportion of business management expenses in relation to total revenue. The enterprise size was determined by the logarithm of the total assets of the enterprise. Moreover, cash flow was measured by the ratio of the net cash flow generated from an enterprise’s operating activities to its operating income. Equity concentration is expressed by the total shareholding ratio of the top ten shareholders.

We employed listed manufacturing companies in the decade from 2008 to 2017 as samples and collected all the data from the Wind Economic Database. To eliminate the influence of extreme values, we winsorised the data to reduce the deviation caused by 1% extreme values on the estimation results. The statistical characteristics of each indicator are displayed in Table 1.

Table 1

Statistical Characteristics of Variables

Variables	Sample Size	Mean	SD	Minimum Value	Maximum Value
lnTFP	18,039	7.0826	1.1968	4.2898	10.3318
lnSub	16,982	−3.3836	1.2161	−12.6507	2.4584
lnR&D	16,694	−1.8019	1.1872	−12.8934	1.2959
Age	22,638	14.0167	5.8952	0.0000	58.0000
Soe	22,720	0.2267	0.4187	0.0000	1.0000
Mtb	22,720	0.2591	2.0099	−188.2250	84.4445
Lev	22,720	0.3778	1.1338	0.0000	96.9593
Man	22,720	0.0987	0.7203	0.0000	65.6063
Sca	18,788	2.8995	1.3583	−4.0570	8.8867
Cf	22,720	0.0652	0.4171	−24.8835	10.4891
Own	22,720	0.3588	0.3113	0.0000	1.0116

Source: Authors’ calculations.

Empirical Analysis

The Comprehensive Impact of Government Subsidies on the TFP of Enterprises

First, this article presents Formula (1) to test the comprehensive effect of government subsidies on the TFP of enterprises. In terms of method selection, we apply the OLS regression method and quantile regression method to estimate Formula (1), and the results are shown in Table 2. Among them, Model 1 only adds explanatory variables and control variables, while Models 2 and 3 successively control industry fixed effects and time fixed effects for the regression. As seen from the regression results, the regression coefficient of government subsidies is always positive at 1% significance level, indicating that government subsidies significantly improve the TFP of enterprises. Thus, H1 is confirmed. In addition, we also conducted robustness tests in the form of dummy variables: the subsidy item of enterprises receiving subsidies is 1, and the subsidy item of enterprises not receiving subsidies is 0. In this way, the regression suggests that the influence of subsidies on enterprises’ TFP is still positive. This effect may result from two mechanisms, namely, direct and indirect mechanisms of action. As for the direct action mechanism, on the one hand, government subsidies promote the capital deepening and scale expansion of enterprises, reduce enterprises’ production and operation costs, increase the efficiency of the enterprise scale and directly promote TFP. On the other hand, government subsidies can help guide enterprises to invest in more reasonable areas and improve their investment efficiency. In terms of indirect action mechanisms, government subsidies stimulate enterprises’ R&D, improve their innovation ability and also indirectly promote TFP. This article will identify these two mechanisms of action in the follow-up research.

Table 2

Test of the Comprehensive Impact of Government Subsidies on the TFP of Enterprises

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8	Model 9
	OLS	OLS	OLS	OLS	10%	25%	50%	75%	90%
lnSub	0.0664***	0.0633***	0.0721***	0.1450***	0.0983***	0.0829***	0.0692***	0.0559***	0.0480***
	(0.0038)	(0.0037)	(0.0036)	–0.0101	(0.0086)	(0.0041)	(0.0042)	(0.0037)	(0.0052)
Age	–0.0112***	–0.0135***	–0.00562***	–0.0063***	–0.0057***	–0.0041***	–0.0041***	–0.0027***	–0.0043***
	(0.0009)	(0.0009)	(0.0009)	(0.0009)	(0.0020)	(0.0010)	(0.0010)	(0.0008)	(0.0014)
Soe	0.0430***	0.0623***	0.0142	0.0155	–0.0105	0.0037	0.0304**	0.0077	–0.0132
	(0.0112)	(0.0108)	(0.0109)	(0.0109)	(0.0213)	(0.0157)	(0.0137)	(0.0102)	(0.0132)
Mtb	–0.0061***	–0.0068***	–0.0081***	–0.0074***	–0.0085***	–0.0166	–0.0174	–0.0175**	–0.0156***
	(0.0021)	(0.0020)	(0.0020)	(0.0020)	(0.0028)	(0.0113)	(0.0122)	(0.0068)	(0.0023)
Lev	0.328***	0.366***	0.274***	0.272***	0.2878***	0.2799***	0.2695***	0.2683***	0.2856***
	(0.0242)	(0.0234)	(0.0234)	(0.0236)	(0.0461)	(0.0332)	(0.0399)	(0.0282)	(0.0478)
Man	–0.138***	–0.149***	–0.108***	–0.103***	–0.7692***	–0.5649***	–0.5421***	–0.1908	–0.0911
	(0.0250)	(0.0238)	(0.0235)	(0.0236)	(0.1545)	(0.1332)	(0.1592)	(0.1502)	(0.0813)
Sca	0.790***	0.780***	0.802***	0.761***	0.7873***	0.7971***	0.7989***	0.8043***	0.8100***
	(0.0046)	(0.0044)	(0.0044)	(0.0049)	(0.0082)	(0.0052)	(0.0050)	(0.0046)	(0.0064)
Cf	0.182***	0.118***	0.146***	0.158***	0.1805***	0.1406***	0.1388***	0.0426	0.0322
	(0.0252)	(0.0240)	(0.0238)	(0.0239)	(0.0620)	(0.0437)	(0.0546)	(0.0579)	(0.0529)
Own	–0.176***	–0.159***	–0.170***	–0.185***	–0.1584***	–0.1738***	–0.1593***	–0.1525***	–0.1606***
	(0.0192)	(0.0182)	(0.0180)	(0.0181)	(0.0418)	(0.0209)	(0.0235)	(0.0199)	(0.0253)
Individual-fixed effect	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time-fixed effect	No	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant	5.077***	5.431***	5.695***	5.485***	5.5767***	5.724***	5.7321***	5.7563***	5.7600***
Constant	(0.0218)	(0.2250)	(0.2220)	(0.2220)	(0.1013)	(0.0849)	(0.0854)	(0.0872)	(0.0665)
R ²	0.751	0.778	0.785	0.783	0.4578	0.5074	0.5513	0.5904	0.6181
Sample size	16,537

Source: Authors’ calculations.

Notes: **p < .05; ***p < .01. OLS, ordinary least squares.

An OLS estimation describes the process in which the explanatory variable affects the conditional expectation value. However, in most conditions, the conditional mean model has relatively strict premise assumptions for sample distribution and model setting, and when these premise assumptions are not satisfied, the results obtained by traditional linear regression methods, such as OLS, will no longer be valid. Therefore, this article further introduces the quantile regression method to estimate Equation (1). Based on the conditional distribution of the dependent variable, this method fits the independent variable linear function and estimates the data of a specific distribution by estimating the values of the dependent variable at different points. Compared with the average marginal effect of the independent variable on the dependent variable examined by OLS estimation, the quantile regression estimation provides the marginal effect of the independent variable on the dependent variable at a certain point. In the analysis of this article, the advantage of quantile regression is that we can distinguish the influence of government subsidies on the TFP of enterprises under different TFP conditions. In the quantile regression analysis, five sub-loci of 10%, 25%, 50%, 75% and 90% were selected for this article and the results are shown in Models 5–9. We can see that, at different sub-points, the regression coefficient of government subsidies on enterprises’ TFP is always positive at 1% significance level, which once again demonstrates the promotion effect of government subsidies on enterprises’ TFP. In addition, the regression coefficient value of government subsidies in Models 5–9 gradually decreases, indicating that the higher the TFP of enterprises is, the smaller the marginal effect of government subsidies on TFP becomes. It is not difficult to understand this conclusion. When the TFP of an enterprise is at a high level, the scale economy of the enterprise has often been formed and the government subsidy has a very limited effect on the improvement of TFP. However, when the TFP of enterprises is relatively low, it may be because the scale economy has not yet been formed and there is a large funding gap for scale expansion, capital deepening and R&D and innovation capability improvement. In this case, the marginal effect of government subsidies on the TFP of enterprises is relatively high. From this point of view, it is better to give a helping hand than an icing on the cake.

From the regression results of the control variables, the influence of the enterprise’s operating years on its TFP is negative, that is, the longer the operating years, the lower the TFP of the enterprise becomes. This result may seem strange, but it is not difficult to understand. At present, many enterprises still use the management mode and production equipment when they were established. The longer the enterprise operates, the older the management mode and equipment will become, and the stronger the restrictions on TFP will be. After controlling the time and industry fixed effects of the model, the regression coefficient of the ownership nature of enterprises on TFP is not significant, indicating that there is no significant difference in TFP between listed SOEs and non-SOEs. Enterprise growth regression coefficients are negative and the most significant in the regression model, explaining that the promotion of an enterprise’s development is not accompanied by TFP. This phenomenon may be due to the fact that some enterprises blindly pursue quick development in the form of investment expansion, ignoring the intensive level of enterprise development, suppressing TFP improvements, and developing an inefficient growth phenomenon. The regression coefficient of financial leverage on TFP of enterprises is positive at 1% significance level in all models, indicating that enterprises with high debt tend to have high TFP, mainly because the higher the corporate debt, the stronger the corporate financing capacity becomes, which lays a capital foundation for the improvement of TFP. The management cost is negative to the regression coefficient of enterprise TFP at 1% significance level in Models 1–7, that is, the management cost limits the improvement of TFP. This phenomenon occurs because the excessive management cost squeezes out enterprises’ investment directly related to the improvement of TFP, such as scale economy mining and R&D investments, inhibiting the improvement of TFP. The variable of enterprise size is positive at 1% significance level in all models, indicating that scale expansion is conducive to the formation of scale economies and the improvement of enterprise TFP. The regression coefficient of enterprise cash flow in Models 1–7 is positive at 1% significance level, indicating that enterprise cash holdings can effectively alleviate capital shortages caused by other external factors in the enterprise operation process, thus promoting the improvement of enterprise TFP. The regression coefficient of ownership concentration on the TFP of enterprises is significantly negative, which may be because overly high ownership concentration tends to lead to decision-making errors while weakening enterprises’ R&D incentives, which is not conducive to the improvement of TFP.

In addition, there are differences in the ability and amount of government subsidies for enterprises of different ownerships. Differences also exist in the organisational structure, management efficiency, and enterprise scale of the enterprises themselves. The above-mentioned differences may cause government subsidies to have a differential impact on the TFP of enterprises. For this reason, this article further subdivides the total sample into samples of SOEs and non-SOEs, among which the SOEs sample is further subdivided into central and local SOEs. The four groups of samples were regressed and the results are displayed in Table 3. The regression coefficients of government subsidies to the TFP of enterprises with different types of ownership are all positive at 1% significance level. This finding indicates that government subsidies can significantly improve TFP in both SOEs and non-SOEs, again confirming H1.

Table 3

Sample Regression

	Model 1	Model 2	Model 3	Model 4
	Non-SOEs	SOEs	Local SOEs	Central SOEs
lnSub	0.0931***	0.0331***	0.0220***	0.0561***
	(0.0044)	(0.0064)	(0.0081)	(0.0104)
Control variables	Yes	Yes	Yes	Yes
Individual-fixed effect	Yes	Yes	Yes	Yes
Time-fixed effect	Yes	Yes	Yes	Yes
Constant	5.682***	5.322***	5.396***	5.347***
	(0.223)	(0.1750)	(0.108)	(0.199)
Sample size	11983	4554	2759	1795
R ²	0.740	0.798	0.799	0.811

Source: Authors’ calculations.

Notes: ***p < .01. SOEs, state-owned enterprises.

The results above indicate that there is no significant difference in the direction of government subsidies between the TFP of SOEs and non-SOEs. Furthermore, we tested the difference in the effect of government subsidies on the TFP of two types of enterprises. Therefore, based on Equation (1), we added the cross-multiplication of government subsidies and enterprise nature dummy variables for regression and the results are reported in Model 1 in Table 4. The regression coefficient of the cross-multiplication is negative at 1% significance level, indicating that the government’s subsidy effect on non-SOEs is better than that of SOEs. The reasons may be as follows: on one hand, non-SOEs make more intensive and efficient use of government subsidies; on the other hand, relative to the non-SOEs, SOEs have closer political associations, out of spoiling SOEs, government departments are more inclined to increase subsidies to SOEs, and the same as the main body, government officials and SOE executives are more prone to collusion and rent-seeking behaviours, affecting the effect of government subsidies.

In addition, we also examined the subsidies for central SOEs and local SOEs in terms of TFP effect size differences, namely in the further samples of the SOEs regarding the enterprise nature of virtual variables, including the local SOE assignment (1) and central SOEs (0), and government subsidies and the enterprise nature of virtual variables by a regression in type (1); the results are shown in Model 2 in Table 4. The regression coefficient of government subsidies is positive at 1% significance level, indicating that subsidies are conducive to improving the TFP of SOEs. The regression coefficient of the dummy variable of the enterprise nature is significantly negative, indicating that, compared with local SOEs, central SOEs have higher TFP. The regression coefficient of the cross-multiplier between the subsidy and the dummy variable of the enterprise nature is negative at 1% significance level, indicating that the government’s subsidy effect on the central SOEs is better than that of SOEs. The reasons may be as follows: On the one hand, due to local protectionism and other motives, local governments are more inclined to extend subsidies to local SOEs, but lack of follow-up evaluations and supervision mechanisms of subsidy funds leads to inefficient use of government subsidies by local SOEs; on the other hand, as mentioned above, compared with central SOEs, executives of local SOEs are more likely to collude with local government officials and seek rent from each other to inhibit the effect of subsidies, which affects the impact of subsidies on the improvement of TFP.

Table 4

Further Analysis of the Heterogeneity of Government Subsidy Effects

	Model 1		Model 2
lnSub	0.0906***	lnSub	0.0527***
	(0.0043)		(0.0100)
Soe	−0.198***	Soe_df	−0.0913*
	(0.0290)		(0.0484)
lnSub×Soe	−0.0607***	lnSub×Soe_df	−0.0324***
	(0.0077)		(0.0125)
Control variables	Yes	Control variables	Yes
Individual-fixed effect	Yes	Individual-fixed effect	Yes
Time-fixed effect	Yes	Time-fixed effect	Yes
Constant term	5.747***	Constant term	5.395***
	(0.221)		(0.177)
Sample size	16,537	Sample size	4554
R ²	0.786	R ²	0.799

Source: Authors’ calculations.

Notes: *p < .1; ***p < .01.

The Mediating Effect Test of Enterprise R&D Decisions

To test whether government subsidies affect TFP by influencing enterprises’ R&D decisions, we further regressed Equations (2) and (3) based on the regression of Equation (1). The results are shown in Table 5, where Models 1–3 correspond to Equations (1)–(3), respectively. As seen in Model 2, the regression coefficient of government subsidies on enterprises’ R&D investment is positive at 1% significance level, indicating that government subsidies significantly increase enterprises’ R&D investments, and subsidies have an incentive effect on enterprises’ R&D. Compared with Model 1, a variable of enterprise R&D investment was added to Model 3, and its regression coefficient is positive at 1% significance level, indicating that enterprise R&D investment has a relatively significant effect on TFP. However, the regression coefficient of government subsidies to the TFP of enterprises is still positive at 1% significance level. In addition, the Sobel test results also support the existence of a mediating effect. The results of a Bootstrap test indicate that the direct and indirect effects of subsidies on TFP are significant at 1% level. Therefore, while government subsidies directly promote enterprises’ TFP, they also encourage enterprises to increase R&D investments, improve enterprises’ innovation levels and indirectly increase TFP. Moreover, the mediating effect is 0.0148, accounting for 21.43% of the total effect (0.0689). Thus, H2 has been confirmed.

Table 5

Mediating Effect Test of Enterprise R&D Decisions

Dependent Variable	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Dependent Variable	lnTFP	lnR&D	lnTFP	lnTFP	lnR&D	lnTFP
lnSub	0.0689***	0.1523***	0.0541***	0.1013***	0.2387***	0.0706***
	(0.0039)	(0.0073)	(0.0038)	(0.0073)	(0.0145)	(0.0078)
lnR&D			0.0970***			0.0987***
			(0.0042)			(0.0046)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
Individual-fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Time-fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Constant term	5.7383***	−0.3164	5.7690***	5.4014***	0.5652	5.3649***
	(0.2145)	(0.4079)	(0.2185)	(0.2679)	(0.4983)	(0.2566)
Sobel inspection	0.0148 (z = 15.36, p = .000)
Bootstrap inspection (Direct effect)	0.0541 (z = 13.44, p = .000)
Bootstrap inspection (Indirect effect)	0.0148 (z = 13.25, p = .000)
Sample size	14,848	14,848	14,848	14,438	13,471	13,312
R ²	0.7950	0.5456	0.8019	0.783	0.289	0.797

Source: Authors’ calculations.

Note: ***p < .01.

The endogenous dependence between variables may influence the estimation results in this article. On the one hand, an important basis for the government choosing enterprises for subsidies is the operational efficiency of enterprises, that is, the government often purposefully chooses enterprises for subsidies according to their TFP. However, government subsidies have a significant impact on the TFP of enterprises, and a two-way influence relationship exists between government subsidies and the TFP of enterprises. To control the endogeneity problem caused by this two-way influence, for reference to relevant research, we selected the lag phase of government subsidies as the instrumental variable and the two-stage least squares estimation method. The reason why the lagging phase of government subsidies was chosen as the instrumental variable is mainly because the government subsidy preference has certain inertia characteristics. The government subsidy preference in a given phase is affected by the preference in the previous phase, further affecting enterprises’ R&D decisions and TFP. Moreover, the number of government subsidies in the subsequent period will not be affected by the TFP and R&D decisions of enterprises in the current period. Therefore, the application of government subsidies lagging behind the first period as an instrumental variable can better overcome the reverse causality problems in the model. The two-stage least squares estimation method was used to estimate Equations (1)–(3), and the regression results are displayed in Models 4–6 in Table 5. After controlling for the endogeneity problem, the government subsidy is still positive in Models 4 and 6 at 1% significance level; that is, the government subsidy can promote the TFP improvement of enterprises. At the same time, Model 5 shows that subsidies also encourage enterprises to increase R&D spending. In Model 6, enterprise R&D expenditure is also positive at 1% significance level, meaning that R&D promotes enterprises’ TFP. Therefore, on the whole, government subsidies not only directly promote enterprises’ TFP but also promote enterprises’ R&D while indirectly improving TFP. The estimation results above have a certain robustness.

In addition, to test the effect of government subsidies on enterprise TFP and the mediation effect of enterprise development decisions about whether differences exist between enterprises of different types of ownership, we further divided the samples into local SOEs, central SOEs, and non-SOEs; the results as displayed in Table 6. The Sobel test and bootstrap test results suggest that there is a mediating effect of enterprise R&D decisions in all three sample groups. The difference is that the direct effect of government subsidies on local SOEs is not significant. Thus, government subsidies not only directly improve the TFP of SOEs and non-SOEs but also encourage enterprises to increase R&D expenditure, thus indirectly improving the TFP of enterprises. However, in local SOEs, government subsidies can only indirectly promote TFP by encouraging enterprises to increase R&D expenditure, but the direct effect is not significant.

Table 6

Sample Test of the Mediating Effect of Enterprise R&D Decisions

	Local SOEs	Central SOEs	Non-SOEs
Sobel inspection	0.0030 (z = 2.45, p = .014)	0.0096 (z = 4.06, p = .000)	0.0225 (z = 16.08, p = .000)
Bootstrap inspection (direct effect)	0.0107 (z = 1.19, p = .234)	0. 0492 (z = 4.24, p = .000)	0.0625 (z = 13.10, p = .000)
Bootstrap inspection (indirect effect)	0.0030 (z = 2.27, p = .023)	0.0096 (z = 3.98, p = .000)	0.0225 (z = 12.13, p = .000)

Source: Authors’ calculations.

Note: SOEs, state-owned enterprises.

Similarly, this article also tests whether the heterogeneity of the size of the intermediation effect of R&D decisions are significant. To be specific, first of all, in Equation (2), the cross-multiplication term of enterprise nature (SOEs are 1, and non-SOEs are 0) and government subsidies are included for the regression. The results appear in Model 2 in Table 7. The regression coefficient of the cross-term between government subsidies and the enterprise nature is negative at 10% significance level, that is, the incentive effect of government subsidies on the R&D of non-SOEs is stronger, but the incentive effect on the R&D of SOEs is relatively small. This phenomenon may be due to the fact that, compared with SOEs, non-SOEs have a higher innovation preference (as seen from the regression results of the dummy variable of the enterprise nature) and are more likely to be influenced by the role of government subsidies and incentives (Luo et al., 2011).

Second, the cross-product term of the enterprise nature and government subsidies is further introduced in Equation (3) for regression to test the heterogeneous characteristics of the effect of enterprise R&D on TFP. The results are shown in Model 3 in Table 7. The regression coefficient of the enterprise nature and government subsidy multiplier is negative at 1% significance level, indicating that the promotion effect of the R&D in SOEs on enterprise TFP is lower than that of non-SOEs. The reasons may include three aspects: first, the R&D and innovation activities of non-SOEs are more intensive and efficient, which can better promote the improvement of the TFP of enterprises; second, SOEs bear more social responsibilities and their R&D and innovation activities are closer to the government’s innovation preference, meaning that they pursue the overall social value and long-term strategic significance of innovation activities rather than short-term economic interests. This idea is reflected in the short-term, low efficiency of SOEs’ R&D activities and their weak role in promoting TFP. Third, compared with non-SOEs, due to the multiple principal–agent problem, as well as the ‘institutional characteristics’ and ‘political personality’ of SOE executives, SOEs naturally lack the spirit of risk taking and incentives for R&D and innovation, making the effect of R&D activities on enterprise TFP smaller than that of non-SOEs (Yang et al., 2015). The incentive effect of subsidies on enterprise R&D and the promotion effect of enterprise R&D on TFP of SOEs are lower than non-SOEs, so the intermediary effect of the enterprise R&D of SOEs is lower than that of non-SOEs.

Third, to test government subsidies for central SOEs and local SOEs R&D incentive size for a significant difference, we joined the virtual variables of local SOEs and the cross-product term of local SOEs and government subsidies in Equation (3). The estimation is based on a sample of SOEs and the results of the model are shown in Table 7. The regression coefficient between government subsidies and the nature of enterprises is not significant, indicating that there is no significant difference between central and local SOEs in the incentive effect of government subsidies on enterprises’ R&D decision-making processes. In addition, the regression coefficient of the enterprise nature variable also failed the significance test, indicating that there was no significant difference in R&D preference between central SOEs and local SOEs.

Table 7

Further Analysis of the Heterogeneity of the Mediating Effects of R&D Decisions

Dependent Variable	Model 1	Model 2	Model 3	Dependent Variable	Model 4	Model 5	Model 6
Dependent Variable	lnTFP	lnR&D	lnTFP	Dependent Variable	lnTFP	lnR&D	lnTFP
lnSub	0.0906***	0.153***	0.0615***	lnSub	0.0527***	0.108***	0.0495***
	(0.0043)	(0.0084)	(0.0046)		(0.0100)	(0.0244)	(0.0105)
Soe	−0.198***	−0.230***	−0.257***	Soe_df	−0.0324***	0.0258	−0.0405***
	(0.0290)	(0.0579)	(0.0323)		(0.0125)	(0.0315)	(0.0137)
lnSub×Soe	−0.0607***	−0.0270*	−0.0308***	lnSub×Soe_df	−0.0913*	−0.0472	−0.199***
lnSub×Soe	(0.0077)	(0.0157)	(0.0083)	lnSub×Soe_df	(0.0484)	(0.1190)	(0.0545)
lnR&D			0.127***	lnR&D			0.0696***
			(0.0053)				(0.0111)
lnR&D×Soe			−0.0768***	lnR&D×Soe_df			−0.0392***
lnR&D×Soe			(0.0078)	lnR&D×Soe_df			(0.0127)
Control variables	Yes	Yes	Yes	Control variables	Yes	Yes	Yes
Individual-fixed effect	Yes	Yes	Yes	Individual-fixed effect	Yes	Yes	Yes
Time-fixed effect	Yes	Yes	Yes	Time-fixed effect	Yes	Yes	Yes
Constant term	5.747***	−0.236***	5.819***	Constant term	5.395***	−1.054***	5.705***
	(0.221)	(0.407)	(0.210)		(0.177)	(0.414)	(0.177)
N	16,537	15,259	14,848	N	4,554	3,820	3,773
R ²	0.786	0.310	0.804	R ²	0.799	0.316	0.811

Source: Authors’ calculations.

Notes: *p < .1; ***p < .01.

Finally, based on the sample of SOEs, we added the dummy variable of local SOEs and the cross-multiplication term of the R&D scale of local SOEs and enterprises to carry out a regression based on Equation (3) to test the difference of the impact effect of enterprise R&D on local SOEs and central SOEs. The results are displayed in Model 6 in Table 7. The regression coefficient of cross-multiplication between local SOEs and their R&D investment is negative at 1% significance level, indicating that the promotion effect of enterprise R&D on the TFP of central SOEs is significantly higher than that of local SOEs. This phenomenon may be due to the fact that, compared with local SOEs, the R&D activities of central SOEs face more extensive and effective supervision, possibly improving the intensification level of their R&D activities and innovation efficiency, as well as promoting the TFP of enterprises. There is no significant difference between central SOEs and local SOEs in the impact of subsidies on enterprise R&D decisions, and the effect of enterprise R&D on the improvement of the TFP of central SOEs is higher than that of local SOEs. Therefore, we can infer that the intermediary effect of the enterprise R&D decisions of central SOEs is higher than that of local SOEs.

Main Conclusions and Practical Implications

In recent years, China has put forward the goal of building a world-class enterprise, and through subsidies and other means to support the innovation and development of enterprises. Based on the data of listed companies in the manufacturing industry from 2008 to 2017, we analysed the impact of government subsidies on the TFP and the mediating effect of enterprise R&D decisions by using a multiple regression model, a quantile regression model, and an intermediary effect model. The main conclusions are as follows. First, government subsidies promote the improvement of TFP. However, the marginal effect of government subsidies on enterprises with low TFP is higher than those with higher TFP due to the differences in enterprises’ development foundation, production and operational efficiency, management ability and so on. Second, government subsidies not only directly promote the improvement of TFP but also encourage enterprises to increase R&D expenditure, enhance the innovation ability of enterprises and then indirectly improve TFP. Third, the effect of government subsidies on TFP varies among different ownership enterprises. Government subsidies play a stronger role in improving the TFP of non-SOEs. In terms of the mediating effect, government subsidies can promote the TFP of local SOEs by encouraging enterprises to increase R&D expenditure, while subsidies have no direct impact on TFP. In central SOEs and non-SOEs, government subsidies not only directly improve the TFP of enterprises but also indirectly promote TFP by encouraging enterprises to increase R&D expenditure. In addition, the mediating effect of R&D decisions in SOEs is lower than that in non-SOEs, and the mediating effect of R&D decisions in local SOEs is lower than that in central SOEs.

Based on the conclusions, this study suggests that the government should play a positive role in improving the TFP of enterprises. First, the government should optimise the allocation of its financial resources and allocate government subsidies reasonably. The results demonstrate that the marginal effect of government subsidies on enterprises with low TFP is higher than that of enterprises with high TFP, which indicates that the result of government subsidies to low-efficiency enterprises is better than that of high-efficiency enterprises. Therefore, in the process of implementing a subsidy policy, the government should, to a certain extent, put particular emphasis on low-efficiency enterprises with development potential, guide them to plan investments reasonably, promote the reasonable expansion of the enterprise scale and promote the improvement of TFP. However, in the process of subsidising enterprises with low TFP, ‘subsidy dependence’ and the formation of zombie enterprises should be avoided. At the same time, for high-efficiency enterprises, the tracking evaluation and supervision of government subsidies should be strengthened to improve the marginal effect of government subsidies.

Second, the incentive effect of government subsidies on enterprises’ R&D investment should be improved and enterprises should be encouraged to carry out R&D activities. R&D is the direct driving force for the improvement of TFP. Subsidies should be used to ease the financial constraints that enterprises face during their R&D activities, and higher R&D subsidies should be given to companies operating in R&D fields with market prospects and development potential. It is necessary to improve the transparency of government subsidy policies, reduce the information asymmetry in the subsidy process, guard against the false behaviour and speculation of enterprises in the process of applying for subsidies, effectively restrain and prevent rent-seeking and corruption and effectively improve the incentive and guiding role of government subsidies for enterprises’ R&D activities.

Third, policymakers should fully consider the differences in the effect of government subsidies on TFP in enterprises with different types of ownership. Government subsidies should be reasonably and optimally allocated among different ownership types. There are three special aspects to be considered as well. At first, the ‘paternal love’ between the government and SOEs should be prevented from interfering with the subsidy policy and the ‘subsidy dependence’ of SOEs should be avoided. Then, the subsidy policies for SOEs should not be excessively crafted according to the economic indicators alone but based on a reasonable evaluation of their strategic position and economic value in terms of economic and social development, social responsibility and the external role in the development of the national economy. Finally, reasonable subsidies should be given to non-SOEs, based on their practical needs to ensure that government subsidies lead to better economic and social benefits.

Footnotes

Acknowledgements

We are grateful to Professor Du Li from the Jilin University as a co-operative tutor for giving us guidance.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship and/or publication of this article: This study is supported by the Ministry of Education of China (Grant no. 20YJC790165), China Postdoctoral Science Foundation (Grant no. 2020T130245) and National Social Science Fund of China (Grant no. 18ZDA107).

References

Baráth

, Fertő

, & Bojnec

Š.

(2020). The effect of investment, LFA and agri-environmental subsidies on the components of total factor productivity: The case of Slovenian farms. Journal of Agricultural Economics, 71(3), 853–876.

Baron

, & Kenny

(1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51(6), 1173–1182.

Benito

, & Hernando

(2007). Firm behaviour and financial pressure: Evidence from Spanish panel data. Bulletin of Economic Research, 59(4), 283–311.

Bernini

, & Pellegrini

(2011). How are growth and productivity in private firms affected by public subsidy? Evidence from a regional policy. Regional Science and Urban Economics, 41(3), 253–265.

Brandt

, Biesebroeck

J. V.

, Wang

, & Zhang

(2017). WTO accession and performance of Chinese manufacturing firms. American Economic Review, 107(9), 2784–2820.

Chen

, Li

, Shapiro

, & Zhang

(2014). Ownership structure and innovation: An emerging market perspective. Asia Pacific Journal of Management, 31(1), 1–24.

Cheng

(2017). Technological innovation spillover and total factor productivity of enterprises: An empirical study based on listed companies. Economic Science, 222(6), 72–86.

Colombo

M. G.

, Grilli

, & Murtinu

(2011). R&D subsidies and the performance of high-tech start-ups. Economics Letters, 112(1), 97–99.

Dong

, & Seo

(2013). Government policy and technological innovation: A suggested typology. Technovation, 33(6–7), 173–179.

10.

Dvouletý

, & Blažková

(2019). The impact of public grants on firm-level productivity: Findings from the Czech food industry. Sustainability, 11(2), 1–24.

11.

Giannetti

, Liao

, & Yu

(2015). The brain gain of corporate boards: Evidence from China. The Journal of Finance, 70(4), 1629–1682.

12.

Guan

, & Lansink

A. O.

(2016). The source of productivity growth in Dutch agriculture: A perspective from finance. American Journal of Agricultural Economics, 88(3), 644–656.

13.

Howell

(2017). Picking ‘winners’ in China: Do subsidies matter for indigenous innovation and firm productivity? China Economic Review, 44, 154–165.

14.

Levinsohn

, & Petrin

(2003). Estimating production functions using inputs to control for unobservables. The Review of Economic Studies, 70(2), 317–341.

15.

, Yang

, & Lu

(2018). Can government participation improve regional innovation efficiency? Economic Review, 214(6), 3–14.

16.

Lin

(2017). Theoretical basis and development direction of the new structure of economics. Economic Review, 205(3), 4–16.

17.

Luo

, Zhao

, & Wang

(2011). Venturing abroad by emerging market enterprises. Management International Review, 51, 433–459.

18.

Mary

(2013). Assessing the impacts of pillar 1 and 2 subsidies on TFP in French crop farms. Journal of Agricultural Economics, 64(1), 133–144.

19.

Olley

, & Pakes

(1996). The dynamics of productivity in the telecommunications equipment industry. Econometrica, 64(6), 1263–1297.

20.

Radas

, Anić

, Tafro

, & Wagner

(2015). The effects of public support schemes on small and medium enterprises. Technovation, 38, 15–30.

21.

Ren

, & Lv

(2014). The financial constraints, the government subsidies and the factor productivity: A case study on the equipment-manufacturing enterprises in China. Management World, (11), 10–23.

22.

Sergant

, & Van Cayseele

(2019). Financial constraints: State aid to the rescue? Empirical evidence from Belgian firm-level data. Journal of Industry, Competition and Trade, 19, 33–67.

23.

Weingast

B. R.

(2009). Second generation fiscal federalism: The implications of fiscal incentives. Journal of Urban Economics, 65(3), 279–293.

24.

Wieser

(2005). Research and development productivity and spillovers: Empirical evidence at the firm level. Journal of Economic Survey, 19(4), 587–621.

25.

(2017). The signal effect of government R&D subsidies in China: Does ownership matter. Technological Forecasting and Social Change, 117, 339–345.

26.

Xiao

, & Lin

(2014). Government support, R&D management and technological innovation efficiency: An empirical analysis based on China’s industrial sector. Management World, (4), 71–80.

27.

Yan

, & Yu

(2017). Government subsidy and firms’ total factor productivity: A comparative analysis of emerging industry and traditional manufacturing industry. Industrial Economics Research, 86(1), 1–13.

28.

Yang

, Wei

, & Luo

(2015). Who innovate from the government subsidies? The joint moderate effects of the ownership and the factor market distortions. Management World, (1), 75–86.

29.

Young

(2003). Gold into base metals: Productivity growth in the People’s Republic of China during the reform period. Journal of Political Economy, 111(6), 1220–1261.

30.

(2011). Innovation cluster, government support and the technological innovation efficiency: Based on spatial econometrics of panel data with provincial data. Economic Review, 168(2), 93–101.