More Government Subsidies, More Innovation of New Energy Firms? Evidence from China

Abstract

This paper evaluates the causal relationship between government subsidy and the innovation performance of new energy firms through count models using 2007–2021 data from China’s listed new energy companies. By looking at the subsidy for listed new energy firms and the number of granted patents, we find government subsidy policies significantly boost firms’ innovation performance. We estimate that a tenfold increase in government subsidy would lead to an increase of 7.11 in the total number of granted patents for new energy firms. Furthermore, a heterogeneity analysis shows such an effect varies depending on the nature of property rights, subsidy scale, and region for new energy firms. To be specific, state-owned firms are more dependent on government subsidy, the effect on innovation is generally higher in the high-subsidy group than in the low-subsidy group while being higher in the low-subsidy group when it comes to low-tech design patents, and firms in the eastern region are most sensitive to government subsidy. This paper also assesses the role of R&D investment in how government subsidy policies boost firms’ innovation performance; that is, by increasing their R&D funding investment rather than R&D manpower investment. These findings illustrate that in developing countries, government subsidy is effective in boosting new energy firms’ innovation performance.

1. Introduction

In recent years, new energy has gradually become a focal point of global industrial competition [1,2]. New energy can help mitigate the overexploitation of natural resources and environmental pollution problems caused by rapid economic development [3,4] and suits the needs of sustainable development and the world’s transition to low-carbon energy [5,6]. The Renewable Energy Law is a law enacted by the Chinese government to encourage economic entities of all forms of ownership to participate in the development and utilization of renewable energy and achieve sustainable economic and social development. The implementation of the law marks the development of renewable energy into a standardized stage. With the implementation and revision of the law, the Chinese government has paid more attention to the development of the new energy industry and has successively promulgated industrial subsidy policies and support policies, which has greatly improved the quantity and quality of China’s new energy enterprises. Since the Renewable Energy Law came into effect on the first of January 2006, China has seen continued advances in new energy technology and equipment, but the country still depends on imports when it comes to core technology as well as key parts and equipment. China is in urgent need of independent innovation to end its reliance on external technology. However, to innovate, new energy firms need to bear high costs and face huge risks over a long investment cycle [7], with their innovation outcomes producing significant positive externalities in technological advances [8,9], environmental protection, and energy security, among other aspects. This challenge, coupled with the fact that both technology and the market are immature in the new energy industry, will discourage firms from innovating. In this context, support from the national government becomes essential.

Government subsidy is a form of direct subsidy, using grants, soft loans, and other ways to directly support enterprises’ R&D and innovation. In China, the regulators hope to improve firms’ research and development (R&D) and innovation by directly providing them with financial resources. A total of CNY 2.796 trillion was invested in research and experimental development (R&D) in 2021. Energy conservation and environmental protection accounted for CNY 2.4 trillion in the 13th Five-Year Plan. Government subsidies became a crucial support for enterprises to grow innovation-driven R&D. In the future, high-level R&D and innovation will turn China from the world’s factory into a global innovation powerhouse. As one of the important policy tools designed to boost the independent innovation capability of firms, government subsidies have long been adopted in China, with some remarkable results. Government subsidies can help address market failures brought about by unavoidable knowledge spillovers and non-complete exclusivity in the process of R&D and innovation [10]. This is particularly true for today’s new energy industry requiring huge investments. While firms invest themselves, government subsidies provide an important additional source of financing which motivates them to continually work on R&D and innovation. However, it cannot be ignored that there is also much literature in existing studies that focuses on the negative effect of government intervention on enterprise innovation activities. For example, Blanes and Busom believe that government subsidies may distort the price of innovation inputs, and some companies may even conduct rent-seeking to obtain subsidy programs [11]. Additionally, government subsidies may also phase out R&D inputs from firms themselves and hinder firms’ innovation to some extent [12,13].

Weaknesses of developing countries in independent innovation, including in the massive new energy industry, make the targeted implementation of government subsidy policies a high priority. Different from the development of traditional industries, R&D and innovation is the key for the new energy industry to keep vital. However, focusing on new energy companies, there is still a lack of systematic evidence of how much government subsidy can effectively spur their innovation and development. In a major study, Li et al. found that there exist complex nonlinear relationships between government innovation and non-innovation subsidy and new energy firms’ innovation inputs and outputs, and only at certain intervals can those subsidies play a role in boosting firms’ innovation [14]. The study also argues that because of innovative inertia, firms can still maintain a certain innovation even if the role of government subsidy is not immediately visible. Wu et al. found an inverted U-shaped relationship exists between the scale of subsidy to new energy firms and their innovation investment, believing government intervention is the best choice [14].

However, some researchers believe that problems such as information mismatch [15] and R&D failures [16] create obstacles to new energy firms’ innovation and development [17,18]. To solve those problems, external financial and technological support is crucial for their innovation and development [7]. However, problems exist in implementing subsidy policies in reality, a topic that has drawn wide attention from scholars. Existing studies discuss the question from the perspective of different countries. Peters et al. discusses the impact of demand-pull and technology-push policies on innovation in the photovoltaic industry of Germany [8]. Sung uses data on South Korean new energy firms to investigate the impact of government policies on firm-level innovation [19]. In fact, whether government subsidies can effectively promote new energy firms’ innovation remains a controversial question.

In addition, enterprise R&D funding and manpower investment provide financial and intellectual support for the innovation activities of enterprises and are the decisive factors to improve the independent innovation ability of enterprises. When carrying out innovation activities, enterprises invest R&D funds to improve the supporting hardware and software facilities required for R&D activities and also invest in high-quality R&D personnel to realize enterprise knowledge accumulation and achievement creation. Earlier studies mostly paid attention to the role of R&D investment as part of government subsidies in boosting innovation [20,21,22] without distinguishing between funding investment and manpower investment. This highlights the need to investigate the effect of government subsidy in innovation and how it influences innovation in developing countries.

This paper will further build on existing research to fill that gap by gauging the effect of government subsidies on innovation with new evidence recorded from China, the largest developing country. Therefore, we take China’s new energy-listed companies as an example to explore the answers to the following questions: Taking direct financial subsidies as an example, do government subsidies have an impact on the innovation performance of new energy enterprises? Is the impact positive or negative? What paths do government subsidies affect enterprise innovation performance? What are the differences in the innovation effect of government subsidies between different corporate ownership structures? In addition, we also discuss the subsidy size differences and regional differences, which may be related to affecting the innovation performance of new energy enterprises.

Our study begins with an eye-catching event: the Chinese government’s direct financial subsidy policy for new energy firms. More precisely, this paper sheds light on the causal relationship between government subsidy policies and new energy firms’ innovation performance using empirical models. The empirical analysis starts with consolidating the data about government subsidy and listed new energy firms’ innovation performance between 2007 and 2021. The focus of this study is on estimating the influence of government subsidy on the number of granted patents for listed new energy firms during the 15-year period. Based on the characteristics of patent data, we use the fixed-effect Poisson regression model to analyze the effect of subsidy on innovation. The empirical model assesses the average impact of government subsidy on each firm’s patent grants in different years. In the meantime, we introduce R&D investment as a variable to investigate the mediating effect of R&D funding investment and R&D manpower investment on how government subsidy influences a firms’ innovation. The analysis also includes other control variables that manifest the level of a firm. Furthermore, this paper also takes into account the heterogeneous effect of subsidy scale and firm ownership on how government subsidy influences innovation. In the end, a series of robustness tests are conducted to validate the robustness of the baseline regression result.

The main contributions are as follows: First, this paper offers the latest findings about how government subsidies influence firms’ innovation. Previous studies have contradictory conclusions, so this paper mainly focuses on the impact of direct financial subsidy on the number of patents of enterprises, and further subdivides the innovative effect on the subsidy policy of the government. Second, we examine the role of R&D investment in government subsidies and enterprise innovation. Government subsidies promote most corporate innovation through R&D funding investment, rather than R&D manpower investment. Third, focusing on new energy enterprises, we explore the innovation benefits of government subsidies, which provides a reference for the sustainability of the new energy industry. The remaining sections of this paper are organized as follows: Section 2 presents the theoretical analysis and hypotheses. Section 3 describes the data, defines key variables, and introduces our empirical models. Section 4 discusses the results of our empirical analysis, including baseline regression, mechanism analysis, heterogeneity analysis, and robustness analyses. Section 5 discusses our findings. Section 6 presents the managerial applications of this paper. Section 7 presents the main conclusions of this paper.

2. Theoretical Analysis and Research Hypotheses

2.1. The Relationship between Government Subsidies and New Energy Enterprise Innovation

Innovation efficiency, which is essentially the process of producing knowledge, is a reflection of a company’s capacity for research and development. Innovation plays a crucial role in company growth. Government subsidies affect firms’ innovation efficiency in complex ways. According to previous literature, government subsidies influence new energy enterprises’ innovation efficiency primarily through two mechanisms.

The first is the facilitation mechanism caused by externality theory [10,21]. Externalities theory provides the economic theoretical basis for government intervention in enterprises’ R&D and innovation activities. Firms would face challenges such as capital shortages, financing constraints, R&D personnel, and technical knowledge to innovate [23]. In China, even large firms would face resource restrictions, especially in the new energy sector. There is also uncertainty and risk associated with R&D activities. Therefore, firms are less likely to engage in innovative activities. At this point, government subsidies are an effective way to compensate for market failures. Government subsidies can provide more debt financing and venture capital funding for firms through a demonstration effect [24,25] and increase the availability of overall resources to support firms’ R&D and innovation [26]. Government innovation subsidies offer capital flow, information flow and technology flow to firms’ innovation activities through the resource allocation effect [14] and signaling mechanism [27]. Moreover, government non-innovation subsidies can bring new concepts, knowledge, and technologies to a firm by attracting investors, including foreign investors [28,29,30].

According to the analysis above, we argue that government subsidies promote R&D activities and innovation performance through external interventions. We suggest the following hypothesis:

H1 (a).

Government subsidies have a positive impact on the innovation performance of new energy enterprises.

Second, government subsidies crowd out firms’ own R&D investments, resulting in a disincentive effect [11,12]. From the perspective of resource acquisition, Yu et al. found that excessive subsidies will squeeze out an enterprise’s own R&D investment due to government subsidies crowding out the effect of enterprise subsidies on innovation performance [31]. Catozzella and Vivarelli found that government innovation subsidies have a negative impact on firms’ innovation productivity after analyzing a sample of Italian firms through the bivariable endogenous switching model [32]. Keupp and Gassmann found that firms’ innovation activities often take place when resources are limited because firms with limited resources usually harness them for R&D and innovation to the greatest extent by putting them into a field they are familiar with [33]. However, some studies found that even with the help of adequate government subsidies and other external resources, firms would see their innovation performance drop if they did not invest internal resources in R&D [34,35]. It is more attractive for an enterprise to generate excess income by engaging in rent-seeking activities, increasing the difference between the factor market and the product market. In this case, many of the government subsidies acquired by the enterprise will be transferred to rent-seeking activities, which will squeeze out the process innovation and product innovation that are needed.

According to the analysis above, we argue that government subsidies inhibit corporate innovation by the crowding out effect. We suggest the following hypothesis:

H1 (b).

Government subsidies have a negative impact on the innovation performance of new energy enterprises.

2.2. Pathways through Which Government Subsidies Affect Innovation in New Energy Enterprises

The increase in innovation quantity and quality requires the support of R&D investment, which includes R&D funding investment and R&D manpower investment. For R&D funding investment, government subsidies will enable enterprises to have sufficient funds to purchase R&D equipment and materials and improve their innovation performance. Based on German manufacturing firms, Hussinger found that there exists a U-shaped relationship between government subsidy and firm R&D intensity [36]. Before reaching a certain threshold, government subsidy has a substitution effect, and beyond that threshold, government subsidy is helpful for greater firm R&D intensity. Guo et al. also found government subsidy can significantly encourage firms to enlarge their independent R&D spending and thus boost their innovation performance [37]. Therefore, we propose the following hypothesis:

H2. 

R&D funding investment is the path for government subsidies to affect the innovation of new energy enterprises.

On the one hand, for R&D manpower investment enterprises to obtain subsidy qualification is conducive to improving credit rating, so as to attract the inflow of scientific research talents and improve innovation performance. On the other hand, government subsidies increase the disposable income of enterprises and guide enterprises to attract more R&D talents by increasing salaries or developing education and training so as to improve their innovation performance. However, the existing research on government subsidies, R&D manpower, and innovation is scarce. We suggest the following hypothesis:

H3. 

R&D manpower investment is the path for government subsidies to affect the innovation of new energy enterprises.

2.3. Heterogeneity of the Innovation Effect of Government Subsidies

2.3.1. Differences in Property Rights

Most state-owned enterprises are in a monopoly position in the new energy industry, so they have less pressure for survival and development. In China’s economic transformation, a special organization has been formed with corporate ownership characteristics, which directly affects its R&D performance and subsequent innovation activities. Compared with other types of enterprises (such as private enterprises), state-owned enterprises emphasize how to innovate effectively and efficiently, because they have huge political resources and are quite conducive to obtaining government subsidies [38]. Hence, the R&D incentives of subsidies to state-owned enterprises were greater than those of non-state-owned enterprises. We suggest the following hypothesis:

H4. 

The positive impact of government subsidies on the innovation performance of state-owned enterprises is greater than that of non-state-owned enterprises.

2.3.2. Differences in Subsidy Scale

The subsidy scale plays an important role. If the subsidy intensity is too low and leads to insufficient innovation incentives, subsidized enterprises may not be able to obtain enough innovation resources, and innovation willingness cannot be effectively stimulated, resulting in low levels of innovation investment by enterprises [20]. As a result, substantive innovations have not significantly increased [39,40]. On the contrary, if the subsidy intensity is too high, it may induce the “rent-seeking” behavior of enterprises. Enterprises pursue the number of innovation results to seek government financial subsidies. In the long run, it will inevitably lead to low innovation efficiency [27,41]. Therefore, we propose the following hypothesis:

H5. 

Different subsidy scales have different effects on new energy enterprises’ innovation performance.

2.3.3. Differences in Regions

Regional factors also play a crucial role. There is a regional imbalance in government R&D support [42]. Public R&D investment is a supplement to the R&D expenditure of enterprises with low technological innovation capacity. Different regions differ in the prerequisites, activities, processes, and networks of innovation. Therefore, there are regional differences in the intensity and effect of government subsidies. The agglomeration of regional innovation resources will also have an important impact on innovation performance [43]. The eastern coastal area is a region with concentrated innovation resources in China, with convenient transportation and more talent. The eastern coastal areas of China have concentrated innovation resources, convenient transportation, and more talent. In contrast, enterprises in the central and western regions have large space to improve technological innovation efficiency. Therefore, we propose the following hypothesis:

H6. 

The positive impact of government subsidies on enterprises’ innovation performance in eastern regions is greater than that in central and western regions.

3. Data and Methodology

3.1. Database

To validate our hypotheses, we developed a unique panel dataset consisting of government subsidy and innovation performance of China’s listed new energy firms between 2007 and 2021. According to the industry classification of the China Securities Regulatory Commission, this paper defines the companies in the field of photovoltaic power generation or photovoltaic product manufacturing, wind power generation or wind power product manufacturing, biomass power generation, new energy product manufacturing and new energy vehicles as new energy enterprises. Our sample originally included all listed companies in the sample period. The sample data was taken from the China Stock Market and Accounting Research (CSMAR) Database, with patent data from estimates of business innovation in the domestic and foreign patent application form in the CSMAR Database. The patents included design, utility models, and invention patents. Next, we obtained the sample of all listed new energy firms by matching those data with the list of new energy firms published on the websites of East Money and Huaxi Securities.

Then, with a unique company code, we cross-linked and sorted information taken from different sources and screened the data according to the following standards: (1) removing firms the shares of which carry ST (special treatment, shares are placed under ‘risk alert’) and * ST tags; (2) removing firms for which important financial data, including operating income, gross assets, and gross liabilities, are missing; and (3) removing firms for which reporting periods are missing. In the end, we collected a study sample of 146 firms. We chose 2007 as the starting year because it was the year when new accounting standards were introduced for China’s listed companies, making some data before 2007 not comparable to data after 2007. Data about other company-level variables came from the CSMAR and Wind databases.

3.2. Measures

3.2.1. Dependent Variable

This study takes innovation performance as the dependent variable. Existing studies measure business innovation performance mainly through two approaches. One is patent numbers, including patent applications, patent grants and patent citations [44]. The other is the sales revenue of new products. Given the difficulty in accessing new product data, this paper only uses patent numbers to measure firms’ innovation performance. According to the Chinese patent law, patents include invention patents, utility model patents, and design patents. Of the three types, invention patents are the most difficult to develop and approve, and also patents with the greatest innovation value, so they are most representative of innovation outputs. Some cities do not support utility model and design patents [45], and patent subsidy across China is mainly targeted at invention patents. Meanwhile, the process from making an invention patent application to getting a grant takes three to four years in China [45], not to mention that not all applications will be granted. Therefore, this paper takes the number of granted patents for firms as the proxy variable for innovation performance, including the total number of annual patent grants, as well as the number of inventions, utility models, and design patents for firms [46,47].

3.2.2. Independent Variable

We use direct financial subsidies as a measure of government subsidies. Direct financial subsidy refers to the use of national and local governments to provide financial donations, income, prices, and other support to certain specific targets, groups, or organizations. This paper uses the actual subsidy amount disclosed in annual reports. To display the relationship between data more clearly, this paper will analyze data after logarithmic processing of the subsidy amount.

3.2.3. Control Variable

Control variables include size, age, CI, SR, and Tobin’s Q, as shown in

Table 1. Descriptive statistics.

Variables	Describtion	N	Mean	S.D.	Min	Max
Gain_Total	Total number of patents jointly obtained that year	1699	4.570	19.98	0	340
Gain_Invention	Number of inventions jointly obtained that year	1699	1.520	10.44	0	180
Gain_UM	Number of utility models jointly obtained that year	1699	2.760	13.21	0	268
Gain_Design	Number of designs jointly obtained that year	1699	0.290	2.690	0	72
Subsidy	Log (Government subsidy + 1)	1671	7	2.090	0	12.39
R&D Investment	Log (R&D investment + 1)	1306	8.530	1.960	0	13.66
Size	Log (Annual total assets)	1699	22.58	1.320	19.60	26.67
CI	Capital-Intensity (Fixed assets/Total assets)	1699	0.340	0.420	0	12.83
Age	Enterprise age	1699	23.03	5.340	10	37
SR	Redundant resources	1699	1.390	1.580	0.0300	18.22
Tobin’s Q	Tobin’s Q	1699	2.150	0.820	0	3.430

. These company-specific characteristics may influence a firm’s innovative activities. Firstly, smaller and younger companies are more innovative [48,49], and the government R&D subsidy program has a significant impact on the number of patent applications, more markedly in the case of smaller firms [50]; however, older companies have more time to accumulate knowledge and experience to support innovation, so we controlled the size and age of companies. We used the natural logarithm of companies’ gross assets to measure their size. Secondly, we define age as the number of years for which a company has been listed [32]; define CI as the ratio of a company’s fixed assets to gross assets [51]; measure SR, which will increase a firm’s explorative activities and therefore boost its innovation performance [52], by the liquidity ratio, i.e., (current assets-inventories)/current liabilities [51]; and define Tobin’s Q as the ratio of a company’s market capitalization to its gross assets. Lastly, we include time as a dummy variable to control characteristics associated with time-fixed effects.

3.2.4. Mediating Variable

Business R&D investment, which measures the ability of a firm in R&D investment, is regarded as the mediating variable. This paper uses R&D Investment (absolute amount of R&D investment + 1) to calculate the R&D funding investment of firms and the number of technical staff to calculate the R&D manpower investment of firms.

Table 1 presents an overview of measures about the listed new energy firms included in the sample, i.e., total number of granted patents (Gain_Total), number of granted invention patents (Gain_Invention), number of granted utility model patents (Gain_UM), number of granted design patents (Gain_Design), subsidy, R&D investment, size, CI, age, SR, and Tobin’s Q. In total, we include 1699 observation results in our analysis.

3.3. Estimation Methods

In this study, the dependent variable is measured by the number of patents granted to listed new energy firms each year. For this reason, the linear regression method (LRM) may result in biased estimates. As patent grants often vary greatly from company to company, a simple logarithmic transformation cannot solve this problem. First, since the variance (19.98) in our data is less different from the mean (4.570), we assume they are approximately equal. Therefore, we used a Poisson regression model for the estimation. Furthermore, the difference between variance and mean can be much larger in a Negative Binomial regression model. Hence, the Negative Binomial (NB) regression model is more appropriate. To provide a better fit to our data, we next used a Negative Binomial regression model to further test our hypothesis. Third, we tested the benchmark regression results with ordinary least square (OLS), generalized linear model (GLM), and zero-inflation Negative Binomial (ZINB) model to ensure its robustness.

We first use the Poisson regression model to determine the impact of government subsidy on innovation performance of new energy firms. Assuming innovation performance variable $Y_{it}$ (subscript i represents the particular firm under observation, t represents the year) also obeys the Poisson distribution with parameter ${\lambda }_{it}$, the probability that the number of patent applications is $Y_{it}$ is:

$$P(Y_{it}=y_{it})=\frac{e^{-{\lambda }_{it}}{\lambda }_{it}{}^{y_{it}}}{y_{it}!} y=0,1,2\dots$$

(1)

where, ${\lambda }_{it}$ > 0 represents the mean times the event happens, determined by a series of dependent variables $x_{it}$. To ensure that ${\lambda }_{it}$ is not negative, we can transform the conditional expectation function $E(Y_{it}|x_{it})$ into the exponential function $\exp ({x^{\prime }}_{it} {\beta }_i)$. With a given sample $\left\{\left(x_{it},y_{it}\right):i=1,2,\dots N\right\}$, we obtain the maximum likelihood estimation of β through the maximum log likelihood function. The log likelihood function is expressed as:

$$\ln L(\beta )={\displaystyle \sum _{i=1}^N{\displaystyle \sum _{i=1}^T\left[-\exp ({x^{\prime }}_{it}{\beta }_i)+y_{it}{x^{\prime }}_{it}{\beta }_i-\ln (y_{it}!)\right]}}$$

(2)

We think that as a firm’s size, age, CI, SR, and Tobin’s Q change, its innovation capability will change accordingly. Therefore, this paper controls these variables in the model, establishing the following baseline regression model:

$$Y_{it}=\alpha +\beta Subsid{y}_{it}+\delta X_{it}+{\lambda }_t+{\epsilon }_{it}$$

(3)

where, $Y_{it}$ represents innovation performance of firm i in year t, including Gain_Total, Gain_Invention, Gain_UM, and Gain_Design; $Subsid{y}_{it}$ represents the government subsidy firm i gets in year t; $X_{it}$ includes all control variables in the model; and ${\epsilon }_{it}$ represents the random disturbance term. The Poisson regression model uses the maximum likelihood estimation method to calculate the maximum likelihood estimate of $\beta$ and $\delta$. In this study, our focus is on parameter $\beta$. If $\beta$ is significantly larger than 0, it shows that government subsidy can significantly boost firms’ innovation.

4. Empirical Results

4.1. Baseline Results

Table 2. Correlations.

Variable	(1)	(2)	(3)	(4)	(5)	(6)	(7)
(1) Gain_Total	1.000
(2) Subsidy	0.123 ***	1.000
(3) Size	0.157 ***	0.466 ***	1.000
(4) CI	−0.003	0.150 ***	0.242 ***	1.000
(5) Age	−0.002	−0.013	0.133 ***	−0.023	1.000
(6) SR	−0.042 *	−0.161 ***	−0.333 ***	−0.160 ***	−0.191 ***	1.000
(7) Tobin’s Q	0.115 ***	0.114 ***	0.407 ***	0.004	0.528 ***	−0.313 ***	1.000

*** p < 0.01, * p < 0.1.

illustrates the correlation between each variable and the others. The variables have quite a low correlation, with the mean value of the variance inflation factors (VIFs) being 1.38, much lower than the acceptable 10 [53]. This indicates multicollinearity can be safely ignored.

Table 3. Basic results.

	(1)	(2)	(3)	(4)	(5)	(6)
	Logit	Probit	Poisson	Poisson	Poisson	Poisson
	Gain_Total	Gain_Total	Gain_Total	Gain_Invention	Gain_UM	Gain_Design
Subsidy	0.108 ***	0.0625 ***	0.189 ***	0.274 ***	0.129 ***	0.475 ***
	(3.40)	(3.38)	(27.11)	(20.08)	(15.77)	(12.68)
Size	0.140 ***	0.0847 ***	0.207 ***	0.367 ***	0.141 ***	−0.286 ***
	(2.59)	(2.61)	(19.32)	(19.15)	(10.14)	(−5.94)
CI	−0.302	−0.178	−0.519 ***	−0.571 ***	−0.357 ***	−0.860 ***
	(−1.54)	(−1.58)	(−10.03)	(−6.31)	(−5.68)	(−3.58)
Age	−0.0236 *	−0.0154 *	−0.0459 ***	−0.116 ***	−0.0188 ***	−0.0383 ***
	(−1.75)	(−1.91)	(−14.68)	(−18.43)	(−4.86)	(−3.36)
SR	−0.0160	−0.0122	−0.0172	0.0560 ***	−0.0593 ***	−0.198 ***
	(−0.38)	(−0.49)	(−1.49)	(3.54)	(−3.54)	(−3.28)
Tobin’s Q	0.0450	0.0331	0.747 ***	1.258 ***	0.675 ***	0.0917
	(0.45)	(0.57)	(25.42)	(20.77)	(18.21)	(1.08)
Year FE	√	√	√	√	√	√
_cons	−5.276 ***	−3.093 ***	−6.326 ***	−11.07 ***	−5.543 ***	1.468
	(−4.45)	(−4.43)	(−23.79)	(−23.52)	(−15.95)	(1.39)
N	1671	1671	1671	1671	1671	1671

*** p < 0.01, * p < 0.1.

shows the baseline regression results. In the Poisson regression model, coefficients of innovation performance are all positive and significant, meaning government subsidy has a significant positive influence on innovation performance of China’s new energy firms.

We conduct a Hausman test on the model and find that the fixed effect model is more suitable. Therefore, we choose a fixed-effect model for our estimation. Regression in the count model comes in two steps. The first step is to determine whether a new energy firm has gained patent grants. Column 1 in

Table 4. Results of the ordinary least square estimate.

	(1)	(2)	(3)	(4)
	Gain_Total	Gain_Invention	Gain_UM	Gain_Design
Subsidy	0.982 ***	0.452 ***	0.434 **	0.0963 ***
	(3.66)	(3.20)	(2.43)	(2.62)
Size	1.283 ***	0.821 ***	0.502	−0.0409
	(2.69)	(3.27)	(1.58)	(−0.63)
CI	−1.837	−0.950	−0.769	−0.119
	(−1.53)	(−1.51)	(−0.96)	(−0.72)
Age	−0.127	−0.114 *	−0.000489	−0.0119
	(−1.09)	(−1.86)	(−0.01)	(−0.75)
SR	0.194	0.173	0.0466	−0.0254
	(0.58)	(0.97)	(0.21)	(−0.55)
Tobin’s Q	1.726 **	0.731 *	0.959 *	0.0364
	(2.05)	(1.65)	(1.71)	(0.32)
Year FE	√	√	√	√
_cons	−32.82 ***	−18.74 ***	−14.75 **	0.664
	(−3.24)	(−3.52)	(−2.19)	(0.48)
N	1671	1671	1671	1671
R2	0.0453	0.0328	0.0325	0.0104

Robust clustered t statistics are shown in parentheses and clustered at the city level. *** p < 0.01, ** p < 0.05, * p < 0.1.

explains the variable Patent_gain = {0, 1}. If the number of patent grants is larger than 0, the variable equals 1; if not, the variable equals 0. At the same time, we use the logit model and probit model, with results shown in Columns 1 and 2 of Table 3, respectively. The two coefficients are both very significant at the 1% level, indicating that an increase in government subsidy significantly boosts innovation by new energy firms.

The second step is to determine the specific number of patent grants. To do so, we first use the Poisson for regression and then use the OLS and Negative Binomial regression method to test the differences between coefficients. Table 3 shows the regression results of the mixed Poisson model. Columns 3–6 show the coefficients of government subsidy are all positive at a significance level of 1%. We also calculate the marginal effect of government subsidy, as shown in

Table 5. Mean marginal effect results of the Poisson regression.

	(1)	(2)	(3)	(4)
	Gain_Total	Gain_Invention	Gain_UM	Gain_Design
	Poisson dy/dx	Poisson dy/dx	Poisson dy/dx	Poisson dy/dx
Subsidy	0.711 ***	0.389 ***	0.250 ***	0.131 ***
	(22.45)	(18.16)	(11.35)	(10.69)
Size	1.293 ***	0.626 ***	0.639 ***	−0.0654 ***
	(25.70)	(20.34)	(16.48)	(−4.70)
CI	−2.684 ***	−0.917 ***	−1.236 ***	−0.277 ***
	(−11.07)	(−6.54)	(−6.83)	(−3.89)
Age	−0.380 ***	−0.203 ***	−0.171 ***	−0.0201 ***
	(−26.68)	(−21.82)	(−16.08)	(−6.07)
SR	−0.0482	0.0962 ***	−0.128 ***	−0.0498 ***
	(−0.94)	(3.97)	(−2.94)	(−3.01)
Tobin’s Q	5.020 ***	2.162 ***	3.033 ***	0.0944 ***
	(35.52)	(23.47)	(27.40)	(3.91)
N	1671	1671	1671	1671

*** p < 0.01.

. The variable of the mean government subsidy is significantly positive (0.711) at a significance level of 1%, albeit slightly smaller than the OLS regression coefficient (0.982). That is, when mean government subsidy increases tenfold, the total number of granted patents to new energy firms would increase by 7.11. The results indicate that government subsidy policies have a positive impact on the number of Gain_Invention, Gain_UM, and Gain_Design of new energy firms. Table 4 and

Table 6. Results of the Negative Binomial regression estimate.

	(1)	(2)	(3)	(4)
	NB Model	NB Model	NB Model	NB Model
	Gain_Total	Gain_Invention	Gain_UM	Gain_Design
Subsidy	0.147 ***	0.154 ***	0.109 **	0.609 ***
	(3.60)	(3.54)	(2.40)	(5.48)
Size	0.276 ***	0.505 ***	0.191 **	−0.237
	(3.75)	(6.43)	(2.22)	(−1.56)
CI	−0.582 **	−0.399 *	−0.741 *	−2.829 ***
	(−2.06)	(−1.81)	(−1.92)	(−3.03)
Age	−0.0337	−0.0614 ***	−0.0139	−0.0630
	(−1.57)	(−2.78)	(−0.64)	(−1.10)
SR	−0.131 **	−0.0682	−0.119	−0.136
	(−2.32)	(−1.57)	(−1.59)	(−1.15)
Tobin’s Q	0.555 ***	0.671 ***	0.657 ***	0.0936
	(4.41)	(5.18)	(4.79)	(0.30)
Year FE	√	√	√	√
_cons	−7.071 ***	−13.13 ***	−6.132 ***	0.252
	(−4.10)	(−7.15)	(−3.15)	(0.07)
N	1671	1671	1671	1671

*** p < 0.01, ** p < 0.05, * p < 0.1.

show the OLS and Negative Binomial regression results, respectively. Although there are slight differences between the coefficients and those in the mixed Poisson model, the coefficients of key variables are positive at a significance level of 1%, validating the robustness of the benchmark regression.

4.2. Mechanism Analysis

Based on the MacKinnon et al. method to identify the mediating role [54], this paper takes R&D investment as the mediating variable, consisting of R&D funding investment and R&D manpower investment. Relevant data come from the CSMAR Database. We take the following steps to validate the mediating effect of R&D funding investment and R&D manpower investment: (1) Regress the independent variable government subsidy and the dependent variable innovation performance. If the regression coefficient is significant, it shows government subsidy influences innovation performance; (2) regress the independent variable government subsidy and the mediating variable. If the regression coefficient is significant, it shows government subsidy influences the mediating variable; and (3) if the two conditions are met, regress the independent variable government subsidy, the mediating variable, and the dependent variable innovation performance simultaneously. If the regression coefficient of the independent variable government subsidy falls or becomes insignificant, it shows the effect of government subsidy on innovation performance is transmitted in part or in whole through the mediating variable.

According to the aforesaid steps, we establish the following empirical model:

Step 1, examine the total effect of government subsidy on innovation performance.

$$Y_{it}=c_0+cSubsid{y}_{it}+c_1{X}_{it}+{\lambda }_t+{\epsilon }_{it}$$

(4)

Step 2, examine the relationship between government subsidy and the mediating variable.

$$R\&D\_investmen{t}_{it}=a_0+aSubsid{y}_{it}+a_1{X}_{it}+{\lambda }_t+e_{it}$$

(5)

Step 3, control the mediating variable and examine the direct effect of government subsidy on innovation performance.

$$Y_{it}=b_0+b_{1}Subsid{y}_{it}+bR\&D\_investmen{t}_{it}+b_2{X}_{it}+{\lambda }_t+{\epsilon }_{it}$$

(6)

The formula below is used to calculate the proportion of the mediating effect, where a is the coefficient of $Subsid{y}_{it}$ in Formula (5); b is the coefficient of $R\&D\_investmen{t}_{it}$ in Formula (6); and c is the coefficient of $Subsid{y}_{it}$ in Formula (4):

$${\mathrm{Mediating}\_\mathrm{effect}}_{\mathrm{R}\&\mathrm{D} \mathrm{Investment}}=\frac{\mathrm{a}\times \mathrm{b}}{c}$$

(7)

Table 7. Mediating effect model: R&D Funding Investment.

	(1)	(2)	(3)
	Poisson	Fixed Effects Model	Poisson
	Gain_Total	R&D Funding Investment	Gain_Total
Subsidy	0.189 ***	0.0671 ***	0.111 ***
	(27.11)	(3.22)	(16.90)
R&D Funding Investment			0.374 ***
			(34.08)
Size	0.207 ***	0.489 ***	0.0261 *
	(19.32)	(7.23)	(1.87)
CI	−0.519 ***	0.0264	−0.0113
	(−10.03)	(0.32)	(−0.36)
Age	−0.0459 ***	−0.0404 *	−0.0194 ***
	(−14.68)	(−1.89)	(−6.05)
SR	−0.0172	0.0789 ***	−0.0649 ***
	(−1.49)	(2.74)	(−4.63)
Tobin’s Q	0.747 ***	0.0115	0.862 ***
	(25.42)	(0.11)	(27.01)
Year FE	√	√	√
_cons	−6.326 ***	−4.866 ***	−7.949 ***
	(−23.79)	(−3.33)	(−7.70)
N	1671	1297	1297

*** p < 0.01, * p < 0.1.

illustrates the mediating effect of R&D funding investment. The regression result in Column 2 shows a positive and significant coefficient of government subsidy, indicating that government subsidy boosts firms’ R&D funding investment. After the mediating variable R&D funding investment is introduced into the model, as shown in Column 3, the coefficient of government subsidy is still positive and significant. This shows that R&D funding investment has a mediating role in the relationship between government subsidy and innovation performance. According to Formula (7), about 13.3% of the total effect is mediated by R&D funding investment.

Table 8. Mediating effect model: R&D Manpower Investment.

	(1)	(2)	(3)
	Poission	Poission	Poisson
	Gain_Total	R&D Manpower Investment	Gain_Total
Subsidy	0.189 ***	0.0201 ***	0.198 ***
	(27.11)	(46.23)	(28.10)
R&D Manpower Investment			0.0000566 ***
			(21.39)
Size	0.207 ***	0.737 ***	0.101 ***
	(19.32)	(1011.74)	(7.92)
CI	−0.519 ***	−0.178 ***	−0.282 ***
	(−10.03)	(−108.90)	(−5.56)
Age	−0.0459 ***	−0.0177 ***	−0.0411 ***
	(−14.68)	(−78.67)	(−12.58)
SR	−0.0172	0.0746 ***	−0.0271 **
	(−1.49)	(96.42)	(−2.18)
Tobin’s Q	0.747 ***	−0.189 ***	0.887 ***
	(25.42)	(−112.93)	(27.11)
Year FE	√	√	√
_cons	−6.326 ***	−9.621 ***	−4.178 ***
	(−23.79)	(−572.99)	(−15.30)
N	1671	1309	1309

*** p < 0.01, ** p < 0.05.

looks at the mediating effect of R&D manpower investment. The regression result in Column 2 shows a positive and significant coefficient of government subsidy, indicating that government subsidy boosts firms’ R&D manpower investment. After the mediating variable R&D manpower investment is introduced into the model, as shown in Column 3, the coefficient of government subsidy is still positive and significant but the number increases. This shows that R&D manpower investment does not have a significant mediating role in the relationship between government subsidy and innovation performance. According to Formula (7), only 0.0006% of the total effect is mediated by R&D manpower investment. The results confirm that the R&D manpower investment’s intermediary utility is not significant.

4.3. Heterogeneity Analysis

4.3.1. Differences in Property Rights

For new energy firms, the effect of government subsidy on their innovation performance may vary depending on the nature of their property rights [55,56,57]. To validate this assumption, it is necessary to analyze different natures of property rights, as presented in

Table 9. Analysis of the regulatory role of the nature of property rights.

	(1)	(2)	(3)	(4)
	Gain_Total	Gain_Invention	Gain_UM	Gain_Design
Subsidy	0.114 ***	0.0746 ***	0.110 ***	0.427 ***
	(15.18)	(4.96)	(12.41)	(10.33)
SOE	−2.567 ***	−3.214 ***	−1.744 ***	−2.418 ***
	(−22.12)	(−15.85)	(−12.16)	(−3.95)
Subsidy × SOE	0.259 ***	0.435 ***	0.105 ***	0.202 ***
	(19.47)	(18.64)	(6.21)	(2.98)
Size	0.214 ***	0.223 ***	0.225 ***	−0.229 ***
	(18.64)	(10.84)	(15.24)	(−4.57)
CI	−0.419 ***	−0.520 ***	−0.211 ***	−0.778 ***
	(−8.33)	(−5.92)	(−3.59)	(−3.24)
Age	−0.0460 ***	−0.0994 ***	−0.0237 ***	−0.0371 ***
	(−14.55)	(−15.94)	(−6.02)	(−3.24)
SR	−0.00789	0.0567 ***	−0.0545 ***	−0.202 ***
	(−0.64)	(3.48)	(−2.97)	(−3.23)
Tobin’s Q	0.931 ***	1.178 ***	0.946 ***	0.255 ***
	(29.59)	(19.17)	(23.33)	(2.77)
Year FE	√	√	√	√
_cons	−5.956 ***	−6.717 ***	−7.169 ***	0.659
	(−20.72)	(−12.89)	(−19.43)	(0.59)
N	1671	1671	1671	1671

*** p < 0.01.

. The full sample regression result in Columns 1–4 illustrates the regulatory role of the nature of property rights. The coefficient of the interaction term between government subsidy and the nature of property rights is significant at a 1% confidence level, indicating that the impact of government subsidy on firms’ innovation performance significantly differs depending on the nature of property rights, which has a positive regulatory role in the effect. In

Table 10. Heterogeneity analysis: Ownership.

	(1)		(2)		(3)		(4)
	Gain_Total		Gain_Invention		Gain_UM		Gain_Design
	SOE	NSOE	SOE	NSOE	SOE	NSOE	SOE	NSOE
Subsidy	0.371 ***	0.106 ***	0.523 ***	0.0734 ***	0.216 ***	0.102 ***	0.710 ***	0.382 ***
	(25.58)	(14.51)	(23.31)	(5.10)	(11.35)	(11.75)	(7.73)	(9.20)
Size	0.0669 ***	0.325 ***	0.0715 **	0.414 ***	0.0374	0.323 ***	−0.221 **	−0.0778
	(3.40)	(22.05)	(2.47)	(13.41)	(1.33)	(18.30)	(−2.09)	(−1.29)
CI	−0.0694 *	−0.386 ***	−0.203 **	−0.546 ***	0.0638	−0.311 ***	0.331	−1.063 ***
	(−1.67)	(−5.41)	(−2.44)	(−3.52)	(1.34)	(−3.77)	(1.29)	(−3.17)
Age	−0.202 ***	0.0134 ***	−0.224 ***	−0.00538	−0.171 ***	0.0173 ***	−0.570 ***	−0.0112
	(−26.12)	(3.59)	(−19.11)	(−0.65)	(−16.06)	(3.86)	(−7.57)	(−0.88)
SR	−0.0380 **	−0.0421 **	0.0129	0.107 ***	−0.0936 ***	−0.0781 ***	−0.0971	−0.300 ***
	(−2.49)	(−2.21)	(0.67)	(3.32)	(−3.45)	(−3.19)	(−1.25)	(−4.08)
Tobin’s Q	1.672 ***	0.565 ***	1.583 ***	0.726 ***	1.773 ***	0.650 ***	6.484 ***	−0.0939
	(21.98)	(15.52)	(15.04)	(8.71)	(15.42)	(14.48)	(7.37)	(−0.98)
Year FE	√	√	√	√	√	√	√	√
_cons	−3.230 ***	−8.995 ***	−4.389 ***	−13.71 ***	−3.257 ***	−9.458 ***	−19.79	−1.678
	(−6.94)	(−23.90)	(−6.57)	(−11.48)	(−4.75)	(−21.29)	(−0.01)	(−1.29)
N	649	1022	649	1022	649	1022	649	1022

*** p < 0.01, ** p < 0.05, * p < 0.1.

, we divide the full sample into state-owned firms and non-state-owned firms and find the coefficient for state-owned firms is larger than that for non-state-owned firms at a 1% confidence level. One possible reason is that state-owned firms are more dependent on government subsidies than non-state-owned firms.

4.3.2. Differences in Subsidy Scale

According to a study by Wu et al., when government subsidy is large in scale, new energy firms would make short-sighted investment decisions, i.e., getting subsidies only by expanding production instead of investing in innovative activities [58]. To further verify whether differences in subsidy size affect innovation performance, we divided the sample into high-subsidy and low-subsidy groups. This was based on whether government subsidies received by companies each year were higher or lower than the median in one year for all companies. Columns 1–8 in Table 3 show the role of subsidy scale on the number of Gain_Total, Gain_Invention, Gain_UM, and Gain_Design, respectively.

By looking at the regression result in

Table 11. Heterogeneity analysis: subsidy scale.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Gain_Total		Gain_Invention		Gain_UM		Gain_Design
	Low	High	Low	High	Low	High	Low	High
Subsidy	0.0721 ***	0.244 ***	0.0259	0.379 ***	0.0836 ***	0.140 ***	0.293 ***	0.193 ***
	(5.77)	(16.82)	(1.06)	(15.62)	(5.69)	(7.11)	(2.79)	(3.29)
Size	0.537 ***	0.0477 ***	0.380 ***	0.268 ***	0.597 ***	−0.0697 ***	0.384 ***	−0.323 ***
	(24.70)	(3.32)	(8.53)	(10.94)	(23.24)	(−3.59)	(3.08)	(−5.76)
CI	0.146 *	−0.783 ***	0.517 ***	−0.624 ***	0.0346	−0.635 ***	−0.448	−1.166 ***
	(1.70)	(−12.18)	(3.21)	(−6.07)	(0.34)	(−7.52)	(−0.72)	(−4.19)
Age	0.0405 ***	−0.0752 **	0.0151	−0.151 ***	0.0458 ***	−0.0452 ***	0.0907 **	−0.0673 **
	(7.25)	(−19.26)	(1.33)	(−19.58)	(6.88)	(−9.15)	(3.22)	(−5.24)
SR	0.0354 **	0.0214	−0.00152	0.200 ***	0.0284	−0.0821 ***	0.132 ***	−0.664 ***
	(2.05)	(1.08)	(−0.04)	(8.58)	(1.30)	(−2.73)	(2.96)	(−6.18)
Tobin’s Q	−0.0970 **	1.056 ***	0.132	1.635 ***	−0.134 **	1.050 ***	−0.686 ***	0.121
	(−2.18)	(27.58)	(1.44)	(21.31)	(−2.50)	(20.67)	(−3.83)	(1.26)
Year FE	√	√	√	√	√	√	√	√
_cons	−13.92 ***	−2.746 ***	−10.63 ***	−11.69 ***	−16.21 ***	−0.131	−29.32	6.645 ***
	(−25.62)	(−7.97)	(−10.20)	(−10.61)	(−23.88)	(−0.29)	(−0.02)	(5.64)
N	850	821	850	821	850	821	850	821

Robust clustered t statistics are shown in parentheses and clustered at the city level. *** p < 0.01, ** p < 0.05, * p < 0.1.

, it is not difficult to understand that the effect of the subsidy on innovation is generally more significant in the high-subsidy group than in the low-subsidy group, as shown by the number of patent grants in Columns 1–6. It shows that the larger the scale of government subsidies is, the more significant the effect on promoting enterprises’ innovation performance. There is only one exception, which is the effect of subsidy on innovation is more significant in the low-subsidy group than in the high-subsidy group when it comes to the number of Gain_Design. One reason may be that design patents are relatively low-tech. When the subsidy scale is high, new energy firms tend to use it for high-tech innovative areas rather than invest in lower innovation activity. This is in line with common sense. Therefore, the innovation benefits of government subsidies in the new energy industry are heterogeneous.

4.3.3. Differences in Regions

To investigate the regional heterogeneity regarding the effect of government subsidy on new energy firms’ innovation performance, we divide the sample data into different regions to conduct group regression. In China, there are three geographical regions: the eastern region, the central region, and the western region.

Table 12. Heterogeneity analysis: regions.

	(1)	(2)	(3)
	Gain_Total
	East	Middle	West
Subsidy	0.232 ***	−0.0337	0.00195
	(29.18)	(−0.86)	(0.10)
Size	0.0961 ***	−0.484 ***	1.120 ***
	(8.62)	(−3.74)	(15.06)
CI	−0.203 ***	0.120	−0.467 **
	(−3.93)	(0.52)	(−1.97)
Age	−0.0692 ***	0.0388	0.141 ***
	(−21.19)	(1.32)	(7.23)
SR	−0.0615 ***	−0.870 ***	−2.275 ***
	(−4.92)	(−4.20)	(−7.40)
Tobin’s Q	1.060 ***	0.984 ***	0.296 **
	(32.46)	(3.86)	(2.11)
Year FE	(8.57)	(0.02)	(0.02)
_cons	−3.882 ***	−7.749	−42.52
	(−14.37)	(−0.01)	(−0.06)
N	1227	253	182

Robust clustered t statistics are shown in parentheses and clustered at the city level. *** p < 0.01, ** p < 0.05.

presents the regression result of regional heterogeneity. Clearly, government subsidy only has a significant role in boosting the innovative performance of new energy firms in the eastern region but does not significantly spur innovation by new energy firms in the central and western regions. This phenomenon can be attributed to two reasons. Firstly, most new energy firms are concentrated in the relatively developed coastal areas in east China while the number is relatively small in the central and western regions. Secondly, in the central and western regions, scientific research talent and institutes are scattered, without much motivation to innovate, resulting in a low return rate for government subsidies.

4.4. Robustness Check

4.4.1. Model Transformation

Considering that the independent variable, patent count, is a discontinuous integer with a large number of zero values (accounting for 71.51% of all observational results), a mixed Poisson regression model and a Negative Binomial model may lead to biased results. We conduct Vuong testing (Z-score > 0) and find the zero-inflation Negative Binomial (ZINB) regression model is more suitable [59]. For this reason, we adopt the ZINB model and the generalized linear model to examine differences in coefficients.

Table 13. Results of the zero-inflation Negative Binomial regression estimate.

	(1)	(2)	(3)	(4)
	Gain_Total	Gain_Invention	Gain_UM	Gain_Design
Subsidy	0.275 ***	0.167 ***	0.452 ***	0.107 ***
	(6.77)	(3.40)	(3.41)	(2.96)
Size	−0.0402	−0.0391	−0.0787	0.0300
	(−1.48)	(−1.14)	(−1.05)	(1.24)
CI	−0.336	−0.298	−2.017	−0.236
	(−0.93)	(−0.71)	(−1.63)	(−0.69)
Age	−0.0532 **	−0.0113	−0.112	−0.0743 ***
	(−2.43)	(−0.43)	(−1.55)	(−3.84)
SR	−0.166 **	−0.209 **	−0.157	−0.180 ***
	(−2.31)	(−2.49)	(−0.70)	(−2.74)
Tobin’s Q	0.680 ***	0.621 ***	0.182	0.856 ***
	(4.78)	(3.79)	(0.52)	(7.06)
_cons	−13.52	−1.813	−11.61	−15.39
	(−0.03)	(−0.55)	(−0.01)	(−0.02)
N	1671	1671	1671	1671

*** p < 0.01, ** p < 0.05.

and

Table 14. Results of the generalized linear model estimate.

	(1)	(2)	(3)	(4)
	Gain_Total	Gain_Invention	Gain_UM	Gain_Design
Subsidy	0.189 ***	0.274 ***	0.129 ***	0.475 ***
	(27.11)	(20.08)	(15.77)	(12.68)
Size	0.207 ***	0.367 ***	0.141 ***	−0.286 ***
	(19.32)	(19.15)	(10.14)	(−5.94)
CI	−0.519 ***	−0.571 ***	−0.357 ***	−0.860 ***
	(−10.03)	(−6.31)	(−5.68)	(−3.58)
Age	−0.0459 ***	−0.116 ***	−0.0188 ***	−0.0383 ***
	(−14.68)	(−18.43)	(−4.86)	(−3.36)
SR	−0.0172	0.0560 ***	−0.0593 ***	−0.198 ***
	(−1.49)	(3.54)	(−3.54)	(−3.28)
Tobin’s Q	0.747 ***	1.258 ***	0.675 ***	0.0917
	(25.42)	(20.77)	(18.21)	(1.08)
Year FE	√	√	√	√
_cons	−6.326 ***	−11.07 ***	−5.543 ***	1.468
	(−23.79)	(−23.52)	(−15.95)	(1.39)
N	1671	1671	1671	1671

*** p < 0.01.

present regression results based on the ZINB model and the generalized linear model, respectively. In both models, the coefficient of government subsidy is significantly positive, another indication of the robustness of the baseline regression results.

4.4.2. Variable Substitution

We also use different methods to measure innovation performance. Innovation performance can be measured by patent numbers, including patent applications, patent grants, and patent citations [45]. As patent citations are not easily accessible, we use the number of patent applications as a substitution variable for the number of patent grants. Next, we conduct a mixed Poisson regression with the number of patent applications. The results from Columns 1–2 in

Table 15. Results after dependent variable replacement.

	(1)	(2)	(3)	(4)
	Apply_Total	Apply_Invention	Apply_UM	Apply_Design
Subsidy	0.309 ***	0.482 ***	0.120 ***	0.374 ***
	(43.43)	(44.71)	(13.01)	(9.37)
Size	0.140 ***	0.106 ***	0.135 ***	−0.219 ***
	(14.68)	(8.00)	(9.25)	(−4.24)
CI	−1.038 ***	−1.770 ***	−0.207 ***	−0.959 ***
	(−22.23)	(−25.94)	(−3.69)	(−3.77)
Age	−0.0361 ***	−0.0428 ***	−0.0224***	−0.0260 **
	(−13.83)	(−11.62)	(−5.74)	(−2.16)
SR	−0.00169	0.0564 ***	−0.120 ***	−0.203 ***
	(−0.20)	(6.02)	(−6.39)	(−3.40)
Tobin’s Q	0.648 ***	0.754 ***	0.582 ***	0.0649
	(28.48)	(22.85)	(17.08)	(0.75)
_cons	−4.975 ***	−6.324 ***	−4.462 ***	−0.874
	(−22.66)	(−20.22)	(−13.35)	(−0.62)
N	1671	1671	1671	1671

*** p < 0.01, ** p < 0.05.

show that regardless of how we measure firms’ innovation, the coefficient of government subsidy is always significantly positive at a 1% confidence level.

4.4.3. Different Control Variables

In Table 15, other control variables are added to ensure the robustness of the results. To rule out the possibility that the baseline regression results are caused by factors that influence new energy firms’ operation, a set of variables related to business operation are added to

Table 16. Results after adding control variables.

	(1)	(2)	(3)	(4)
	Gain_Total	Gain_Invention	Gain_UM	Gain_Design
Subsidy	0.181 ***	0.335 ***	0.0888 ***	0.472 ***
	(22.62)	(21.30)	(9.35)	(12.15)
Size	0.270 ***	0.404 ***	0.232 ***	−0.187 ***
	(21.48)	(17.39)	(14.39)	(−3.59)
CI	−0.654 ***	−0.621 ***	−0.493 ***	−1.217 ***
	(−11.21)	(−6.48)	(−6.62)	(−4.43)
Age	−0.0488 ***	−0.121 ***	−0.0214 ***	−0.0287 **
	(−14.49)	(−17.44)	(−5.18)	(−2.38)
SR	−0.198 ***	−0.0169	−0.301 ***	−0.599 ***
	(−9.89)	(−0.70)	(−10.63)	(−6.07)
Tobin’s Q	0.717 ***	1.310 ***	0.628 ***	−0.148
	(22.52)	(19.52)	(15.71)	(−1.59)
DAR	−1.976 ***	−1.633 ***	−2.176 ***	−2.993 ***
	(−19.17)	(−8.92)	(−16.56)	(−6.76)
RRTA	3.872 ***	6.354 ***	2.760 ***	2.472 ***
	(32.27)	(33.57)	(16.66)	(5.37)
TAT	0.301 ***	0.269 ***	0.368 ***	−0.180
	(9.08)	(4.13)	(9.20)	(−1.16)
ORGR	−0.0873 ***	0.0795 *	−0.189 ***	−0.270 *
	(−2.70)	(1.85)	(−4.00)	(−1.80)
Year FE	0	0	0	0
_cons	−6.895 ***	−12.40 ***	−6.211 ***	1.656
	(−23.84)	(−23.38)	(−16.58)	(1.51)
N	1560	1560	1560	1560

Robust clustered t statistics are shown in parentheses and clustered at the city level. *** p < 0.01, ** p < 0.05, * p < 0.1.

. They include the debt-to-asset ratio (DAR), rate of return on total assets (RRTA), total asset turnover (TAT), and operating revenue growth rate (ORGR). The regression results are similar to the baseline regression results, an indication that the newly added factors related to business operation will not change the earlier research findings and further eliminate their interference in the models in this paper because of their impact on innovation. As shown in Columns 1–4, the coefficients fall but are still statistically significant.

4.4.4. The Lagged Effect of Government Subsidy

The previous assumption was that the impact of government subsidies on the number of patents is instantaneous. It takes time to accumulate and translate new knowledge into innovative commercial applications, as well as to apply for and approve patents. Therefore, we lag both subsidies and all control variables to test the lagged effect of government subsidies. The regression results are shown in

Table 17. The lagged effect of government subsidy.

	(1)	(2)	(3)	(4)
	Gain_Total	Gain_Invention	Gain_UM	Gain_Design
Subsidyt−1	0.198 ***	0.317 ***	0.129 ***	0.438 ***
	(28.88)	(23.37)	(16.05)	(12.59)
Sizet−1	0.185 ***	0.296 ***	0.126 ***	−0.199 ***
	(17.50)	(16.09)	(9.12)	(−4.34)
CIt−1	−0.534 ***	−0.828 ***	−0.231 ***	−1.153 ***
	(−10.10)	(−8.74)	(−3.79)	(−4.65)
Aget−1	−0.0420 ***	−0.106 ***	−0.0177 ***	−0.0280 **
	(−13.90)	(−17.84)	(−4.70)	(−2.52)
SRt−1	−0.0544 ***	0.0390 **	−0.129 ***	−0.156 ***
	(−4.28)	(2.46)	(−6.55)	(−2.97)
Tobin’s Qt−1	0.651 ***	1.116 ***	0.600 ***	−0.0881
	(23.87)	(20.11)	(17.29)	(−1.16)
Year FE	√	√	√	√
_cons	−6.713 ***	−11.13 ***	−5.871 ***	0.303
	(−24.27)	(−23.18)	(−16.12)	(0.28)
N	1670	1670	1670	1670

*** p < 0.01, ** p < 0.05.

. It is not difficult to find that the regression coefficients are all significantly positive, indicating that government subsidies do have a time lag in promoting corporate patents. However, it does not affect the conclusion of benchmark regression, in essence, again verifying the robustness of benchmark regression.

5. Discussion

Firstly, according to the results in Table 3, government subsidy has a direct role in boosting new energy firms’ innovation performance, and this result is consistent with H1 (a). However, that role is mediated by R&D funding investment, not by R&D manpower investment. Today China has elevated the development of the new energy industry to the status of national strategy. Due to the high costs of development and use, as well as external impacts associated with R&D, the industry’s development relies on government policy support. Our study reveals the general positive influence of government subsidy; that is, government fiscal support can give a boost to the development of new energy firms. To that end, the government should set up funds dedicated to providing financial support for new energy firms that boast greater innovation potential and benefit from a more favorable environment. Furthermore, local governments should offer more assistance to local new energy industries in achieving technological innovation through local special funds, special bonds, and the like.

Another finding of this study is that the role of government subsidies in boosting innovation performance varies considerably among different regions and is especially prominent in the eastern region. To address the gap, China should take necessary measures designed to make government guidance fairer and make sure subsidies and support are offered to firms in different regions and on different scales. The government should provide more subsidies for firms in regions taking a lead in innovation and increase the rate of return on government funding support. In addition, for the central and western regions, the government should not only employ policies to facilitate cross-region exchanges and communication over innovation activities but also increase subsidies to local governments there, through science-based macroregulation, in a bid to solve the problem of funding shortages and promote innovation in the new energy industry across China.

6. Managerial Applications

Our findings have important implications for the governments of developing countries. First of all, the innovative development of Chinese new energy enterprises in the largest developing country cannot be separated from the government’s support and help. Second, government subsidies for enterprise capital investment are the key to sustainable enterprise innovation development. Third, the government subsidy policy should tilt more toward vulnerable groups and protect weak enterprises at low economic levels. Fourth, due to the time lag of subsidy policies, policymakers and government agencies can develop subsidy plans for many years to help new energy enterprises innovate and achieve sustainable development. In general, solving the problems brought about by capital can improve efficiency, reduce costs, and improve competitiveness, making government subsidies become a key factor in realizing the innovation and development of enterprises.

7. Conclusions

7.1. Findings

R&D and innovation are important drivers of economic transformation and upgrading. As an essential fiscal policy, government subsidies are meant to encourage firms to work on R&D and innovation amid efforts to transition from “Made in China” to “Created in China”. Based on both theoretical research and empirical analysis, this study offers a feasible economic explanation for the innovation of China’s new energy firms. This study investigates the causal relationship between government subsidies and innovation performance using a sample that includes data from listed Chinese new energy firms between 2007 and 2021. Furthermore, this study examines the mediating role of R&D funding investment in the relationship and analyzes the heterogeneity of innovation benefits from government subsidies. At the same time, a series of robustness tests are conducted to prove the robustness of the baseline regression results. The empirical results lead to the following conclusions:

First, the mixed Poisson regression model is established to make an in-depth analysis of the causal relationship between government subsidy and the innovation performance of new energy firms. Empirical testing points to the significant role of government subsidy increases in boosting innovation performance. The marginal effect of the Poisson regression is then compared with the OLS regression results to validate the model selection correctness. The benchmark regression is further tested using a Negative Binomial regression.

Secondly, this study analyzes how government subsidy increases new energy firms’ innovation performance from the perspective of R&D investment, or to be more specific, its mediating role in the relationship between government subsidy and innovation performance. It is found that R&D funding investment mediates the relationship between government subsidy and innovation performance, i.e., government subsidy spurs new energy firms’ innovation by increasing their R&D funding investment, while the mediating effect of R&D manpower investment is not significant.

Finally, a heterogeneity analysis of firms points to significant differences in the effect of government subsidies on innovation performance depending on the nature of property rights, subsidy scale, and region. State-owned firms are more reliant on government subsidies. The effect on innovation is generally higher in the high-subsidy group than in the low-subsidy group, although the effect on innovation is higher in the low-subsidy group when it comes to low-tech design patents. It is also found that only innovation in new energy firms in the eastern region is sensitive to government subsidies.

7.2. Sustainable Development

Based on the above findings, we draw some policy implications for sustainable development. First of all, from the perspective of sustainable new energy enterprises development, government subsidies should continue to be maintained or increased. From the research perspective, government subsidy is positively related to enterprise innovation performance. High government subsidies are conducive to the output of innovation performance of new energy enterprises. In the market economy, government subsidies are a direct incentive way to ensure the survival of new energy enterprises and promote their development, which is conducive to enterprises reducing resource pressure, better-optimizing resource allocation, and improving their innovation ability.

Furthermore, we standardize the innovation and promotion path of new energy enterprises after receiving government subsidies. Through R&D funding investment, government subsidies improve innovation performance. In order to relieve financial pressure, some new energy enterprises transfer them to non-R&D investment after receiving the government subsidy which leads to the inability to improve innovation performance, thus making the government subsidies fail. Therefore, after the introduction of national follow-up policies and regulations, it is necessary to refine the use path of new energy enterprises after receiving government subsidies.

Moreover, the allocation of resources to different new energy enterprises should be based on the characteristics of their property rights. The conclusion of the research shows that new energy enterprises with different proprietary rights have significant differences in the impact of government subsidies on innovation performance. Therefore, the government can focus on non-state-owned enterprises to promote their higher innovation performance when subsidizing incentives.

Finally, given that government subsidies vary by geographical location, governments should gradually provide targeted financial support and adjust subsidy measures.

7.3. Limitations and Further Research Suggestions

First, due to the availability of data, this paper only considers direct fiscal subsidies as government subsidies. Other types of government subsidies, such as tax incentives and fiscal interest discounts, are not discussed. Future research could more fully discuss government subsidies’ innovative effects. Secondly, when exploring influence paths, the intermediary roles of R&D capital investment and manpower investment are discussed. Future research may explore other impact pathways in more detailed ways, such as the study of mediating and regulatory effects, or within the framework of a more general structural equation model. Finally, the classification of subsidy size is not necessarily optimal. Future research may utilize a threshold effects model to more precisely determine the extent of government subsidies.