The Impact of R&D and Non-R&D Subsidies on Technological Innovation in Chinese Electric Vehicle Enterprises

Abstract

The effectiveness of government subsidies for electric vehicle (EV) enterprises and future improvements to subsidy policies to promote industry development have garnered widespread attention. Distinct mechanisms exist through which R&D and non-R&D subsidies impact enterprise innovation. This paper differentiates between R&D and non-R&D subsidies and uses data from listed companies and New Third Board companies in China from 2013 to 2022 to empirically analyze the effects of these two types of subsidies on the innovation of EV enterprises from the perspectives of innovation strategy and the industrial chain. The results show that both R&D and non-R&D subsidies effectively alleviate the inhibiting effects of financing constraints. R&D subsidies significantly incentivize innovation in EV enterprises, whereas the effect of non-R&D subsidies is not as pronounced. The incentivizing effect of R&D subsidies exhibits two distinct characteristics: first, R&D subsidies compel enterprises to choose an innovation strategy that prioritizes “quantity over quality”; second, R&D subsidies exert a more pronounced influence on enterprises in the upper and middle sectors of the EV industrial chain compared to downstream enterprises, which tend to engage in more strategic innovation behaviors.

1. Introduction

Governments worldwide have introduced various incentive programs to promote the adoption of electric vehicles (EVs) [1]. These initiatives aim to advance national technological progress [2] and address environmental issues associated with internal combustion engine vehicles (ICEVs) [3]. As the largest passenger car market globally, China’s prioritization of EV manufacturing is essential for transitioning from a major automotive nation to an automotive powerhouse. This strategy is also critical for addressing climate change, promoting green development, reducing dependence on oil imports, and enhancing the global competitiveness of China’s automotive industry [4]. Since being designated as a strategic emerging industry and a leading pillar industry in 2010, Chinese EV enterprises have received comprehensive government policy support at both central and local levels. These supports include financial subsidies, value-added tax (VAT) reductions, government procurement incentives, purchase tax exemptions, exemptions from traffic restrictions, no license plate lottery requirements, restrictions on traditional fuel vehicle emissions, green license plates, charging infrastructure subsidies, a “whitelist” for power batteries, and the “dual credit” policy [5,6]. Notably, the “dual credit” policy, inspired by California’s Zero Emission Vehicle (ZEV) program introduced in 1990, aims to promote EV development and reduce emissions from traditional fuel vehicles. Driven by these subsidy policies, China’s EV industry has rapidly developed, forming a complete industrial chain from upstream raw material supply, to key component R&D and production (such as power batteries, motors, and electronic control units), to vehicle design and production, supporting infrastructure construction and achieving initial industrialization.

According to the China Association of Automobile Manufacturers, China’s EV production reached 9.587 million units in 2023, ranking first globally [7]. The market share of EVs in China has also steadily increased, reaching 31.6% in 2023 [7], with the proportion of EV ownership rising to 6.07% [8]. Despite the rapid growth of China’s EV industry, more than 90% of automotive chips rely on imports [9], and advanced sensors and other key components are monopolized by developed countries. The industry still faces challenges such as lagging core technology levels [10] and insufficient, unbalanced development of the industrial chain [4].

The R&D activities of EV enterprises are characterized by long cycles, substantial capital investment, and high risk levels. Additionally, R&D outcomes have strong externalities [11], making enterprises reluctant to invest heavily in the fundamental research behind product innovation. Therefore, government intervention through industrial policy coordination is necessary [12]. Among various policies, government financial subsidies, which are most closely related to enterprise innovation activities, have garnered significant attention [13].

Scholars have conducted extensive research on government subsidies and technological innovation in EV enterprises, but consensus has not been reached. Proponents argue that government subsidies can compensate for the loss of technological spillovers during the innovation process and enhance managerial risk-taking capabilities, thereby incentivizing innovation [14,15]. Additionally, government subsidies serve as a signal, helping enterprises attract external investment and reduce financing costs [16,17,18].

Conversely, some scholars contend that government subsidy policies exhibit a strong tendency to replace market mechanisms with government choices, potentially leading to inefficient “rent-seeking” and “subsidy fraud” behaviors [19,20]. For instance, the large-scale “subsidy fraud” by new energy vehicle (NEV) companies in 2016 was a result of unscientific policy guidance, with “the number of vehicles involved in subsidy fraud reaching 76,374, accounting for 15.4% of total production, and the amount involved reaching 9.27 billion CNY, accounting for 27.7% of total central government subsidies” [21]. Moreover, government subsidies can crowd out enterprises’ own R&D investments, thereby reducing overall innovation efficiency [22].

Chinese EV enterprises have received substantial government subsidies from the early stages of their development, viewed as a crucial tool for supporting enterprise growth through industrial policy. Various types of subsidies have been provided to EV enterprises, from central to local levels, including both R&D subsidies and non-R&D subsidies such as local government investment attraction incentives (offering free or low-cost land), fiscal contribution rewards, VAT rebates, social security subsidies, export incentives, pollution control subsidies, relocation compensation, and financial distress assistance [23]. Non-R&D subsidies may not directly impact the technological innovation activities of EV enterprises [16,24,25]. Nevertheless, previous research has often analyzed the impact of government subsidies on EV enterprise innovation without distinguishing between R&D and non-R&D subsidies, potentially biasing regression results. Therefore, it is necessary to differentiate government subsidies and clarify the relationship between R&D and non-R&D subsidies and technological innovation in EV enterprises.

Under the influence of government subsidies, enterprises may either strive to improve the quality of innovation while pursuing innovation quantity, or they may prioritize innovation quantity at the expense of innovation quality [26]. Additionally, enterprises positioned differently within the industry value chain exhibit distinct technological characteristics and varying demands for R&D and non-R&D subsidies, potentially leading to differential effects of these subsidies [27]. This study meticulously collects and screens data on R&D and non-R&D subsidies provided to EV enterprises by the government, including both battery electric vehicles (BEVs) and plug-in hybrid electric vehicles (PHEVs), examining their impact on the innovation output and R&D strategies of these enterprises at a micro level. Furthermore, it investigates the heterogeneous effects of these subsidies on enterprises at different stages of the industry chain.

This paper contributes to the literature in several ways: First, it distinguishes between R&D and non-R&D subsidies, providing a more targeted analysis of their effects on technological innovation in EV enterprises. Second, it examines the impact of R&D and non-R&D subsidies on both the quantity and quality of innovation, addressing the strategic choices and trade-offs enterprises face between these two aspects, which is often overlooked in previous research. Third, from the perspective of the industry value chain, this study analyzes the heterogeneous effects of R&D and non-R&D subsidies on technological innovation in EV enterprises at different stages, offering insights for the precise adjustment of subsidy policies to promote technological innovation in the EV industry.

2. Theoretical Analysis and Research Hypotheses

2.1. R&D Subsidies, Non-R&D Subsidies, and Innovation Output in EV Enterprises

R&D-based technological innovation is crucial for the development of EV enterprises, but financing constraints may pose significant barriers to innovation. This is especially true for highly innovative, smaller EV enterprises, which require substantial funding to conduct socially beneficial and profitable R&D projects. However, internal funds are often insufficient, making external financing the primary source of R&D funding [28]. Due to issues like information asymmetry and insufficient collateral, these enterprises frequently face difficulties in securing necessary innovation funding. Firstly, because R&D activities can be discontinued at any stage from initiation to outcome conversion, external investors often find it challenging to fully understand these activities and make informed investment decisions [29]. Secondly, the high costs associated with evaluating investment projects by institutional experts lead investors to adopt a conservative stance towards innovative projects. Thirdly, most R&D expenditures are allocated to internal expenses such as salaries for R&D personnel, which cannot be easily used as collateral for bank loans, making it difficult for R&D-intensive enterprises to obtain such loans [30]. Lastly, although information disclosure is a method to reduce information asymmetry, enterprises are typically reluctant to disclose innovation-related information early on to prevent competitors from stealing trade secrets, further exacerbating financing constraints [29].

R&D subsidies can directly or indirectly alleviate the financial constraints of EV enterprises, positively impacting those reliant on external funding for innovation activities and thereby increasing R&D investment [31]. Firstly, R&D subsidies partially mitigate the limitation of insufficient internal funds, although they do not fundamentally solve the issue of sustainable innovation investment. Secondly, R&D subsidies carry a “signaling effect” [17,18,32], indicating to the market that an enterprise’s R&D activities have been professionally evaluated and endorsed by government departments. This reduces the evaluation costs for investors and addresses the issue of information asymmetry in external financing for EV enterprises, providing them with sufficient funds to engage in innovation activities [33,34]. Moreover, R&D subsidies may enhance the risk-taking capacity of management [35], encouraging managers to devote more effort and resources to R&D activities.

In addition to financing constraints, EV enterprises face the issue of R&D externalities [23]. Their innovation activities exhibit significant environmental and technological externalities. While the environmental externalities are challenging to quantify directly in economic terms, their long-term social and ecological value is immense, which is a crucial reason for government support in this sector. The R&D activities of EV enterprises typically drive technological advancements across the entire automotive industry, and these technological externalities can lead to market failures. R&D subsidies can effectively address market failures caused by technological spillovers, enhancing the motivation for R&D and promoting high-quality industry development. Based on the above analysis, R&D subsidies can both directly and indirectly relieve financial pressures on innovation activities and address market failures due to technological externalities. Therefore, we propose the following hypothesis:

H1a: 

R&D subsidies significantly incentivize the innovation output of EV enterprises.

Relative to R&D subsidies, non-R&D subsidies are primarily used for the daily operations and non-technical expenditures of enterprises. Although these subsidies can improve the overall financial status of a company and indirectly support R&D activities to some extent—such as by easing operational pressures and allowing more internal funds to be allocated to R&D—they are not explicitly aimed at technological innovation. As a result, their actual impact on innovation may be significantly less than that of R&D subsidies. For enterprises facing substantial operational pressures, non-R&D subsidies might help improve cash flow and reduce operational stress in the short term. However, since these funds are not specifically allocated for R&D activities, they do not have a substantial incentivizing effect on innovation output. Based on the above discussion, we propose the following hypothesis:

H1b: 

Non-R&D subsidies do not significantly incentivize the innovation output of EV enterprises.

2.2. R&D Subsidies, Non-R&D Subsidies, and Innovation Strategies in EV Enterprises

R&D subsidies might have a dual impact on the innovation strategies of EV enterprises. First, there is the rent-seeking distortion mechanism. Essentially, R&D subsidies represent a redistribution of government resources [19]. Local government officials, who control the allocation of these resources, may engage in rent-seeking behaviors, including fraudulent subsidy claims by EV enterprises [36] and corruption [20]. The costs associated with rent-seeking can deplete the funds available for genuine R&D, forcing enterprises to forego high-cost breakthrough innovations, thereby affecting the quality of innovation. Furthermore, once enterprises acquire substantial R&D subsidies through rent-seeking, they may become complacent and less willing to invest time and effort into pursuing high-quality innovation.

Second, from the perspective of the acceptance mechanism, enterprises receiving R&D subsidies are required to meet government-set innovation quantity targets. Failure to do so may result in the suspension or retraction of disbursed funds. Given the complexity and diversity of technological fields involved in the EV industry, governments may find it challenging to accurately assess innovation quality and thus may focus more on the quantity rather than the quality of innovations [37]. Under such circumstances, enterprises might adopt an innovation strategy that prioritizes quantity over quality to ensure compliance and secure continued funding.

It is worth noting that the application process for R&D subsidies in China is often complex and lengthy. Government officials may lack specialized industry knowledge, leading to subsidies being granted post-achievement rather than at the most critical stages of R&D [38]. Consequently, by the time subsidy funds are disbursed, enterprises might have already completed their R&D projects, diminishing the potential impact of these funds on innovation quality.

In summary, while the government’s intention in providing R&D subsidies is to spur enterprise innovation and accelerate industry development, market dynamics may lead enterprises to prioritize innovation quantity over quality under the influence of these subsidies. Therefore, we propose the following hypothesis:

H2a: 

Under the influence of R&D subsidies, enterprises may prioritize increasing the quantity of innovation over improving its quality.

Non-R&D subsidies are primarily allocated for daily operations and non-technical expenditures, such as market promotion and employee training. Although these subsidies cannot be directly used for R&D activities, they can enhance the overall operational environment of enterprises, alleviating financial pressures in other areas and enabling more internal funds and resources to be directed towards R&D. However, non-R&D subsidies also have potential drawbacks.

Firstly, non-R&D subsidies might lead to inappropriate resource allocation behaviors within enterprises. If enterprises become overly reliant on non-R&D subsidies, their motivation and drive for self-initiated innovation may diminish, increasing their dependency on subsidies. Additionally, given the broader usage scope of non-R&D subsidies, enterprises might allocate these funds to projects with quick returns rather than long-term R&D investments, potentially affecting their long-term innovation strategies.

Secondly, the application and review process for non-R&D subsidies is relatively simpler, which might encourage rent-seeking behaviors. Enterprises may find it easier to obtain these subsidies through improper means, leading to a focus on securing subsidies rather than pursuing effective innovation.

Based on the above discussion, we propose the following hypothesis:

H2b: 

Under the influence of non-R&D subsidies, enterprises may prioritize improving operational conditions rather than directly enhancing innovation quality.

2.3. The Impact of R&D and Non-R&D Subsidies on Innovation in EV Enterprises from the Perspective of the Industrial Chain

Although the EV industry chain forms a cohesive whole, enterprises located at different positions within the chain possess unique technological characteristics. Upstream enterprises supply raw materials for the production of core components of EVs, with primary products including lithium, electrolytes, cathode and anode materials, and separators. However, critical resources needed for power batteries, such as lithium and nickel, are heavily dependent on imports, and key technological breakthroughs in cathode materials and battery separators have yet to be achieved [10]. Midstream enterprises primarily produce components such as batteries, motors, and electronic controls, which account for over 60% of the total vehicle cost [27]. These enterprises are pivotal to technological R&D and innovation, yet such innovation is relatively challenging. Downstream enterprises are involved in complete vehicle integration, body technology, and chassis technology. Many of China’s downstream enterprises have transitioned from traditional vehicle manufacturers and possess extensive experience in automobile manufacturing, such as BYD and Great Wall, which have a foundational basis in vehicle integration technology and face relatively lower innovation difficulties.

Downstream enterprises are generally more mature compared to upstream and midstream enterprises; thus, the impact of R&D subsidies might be more significant for upstream and midstream enterprises. Firstly, apart from some mature nickel mining and processing companies, most upstream and midstream enterprises are emerging firms with shorter establishment periods, weaker financial strength, and lower risk tolerance, thereby necessitating greater support from R&D subsidies. Secondly, from the perspective of innovation characteristics, upstream and midstream enterprises face higher innovation difficulties, and the positive externalities of their technological innovations are stronger, necessitating government-provided R&D subsidies to address these externalities. Lastly, considering the financing environment, downstream enterprises are mostly spin-offs from traditional automobile manufacturers or are established by financially robust companies, making them relatively less reliant on R&D subsidies. It is noteworthy that downstream enterprises also include startups such as NIO, Xpeng, and Li Auto, which, being new, have a more urgent need for funding support. In contrast, upstream and midstream enterprises face greater challenges in acquiring external funding support.

In terms of non-R&D subsidies, upstream and midstream enterprises similarly require substantial support. Non-R&D subsidies can aid these enterprises in improving production equipment, enhancing management levels, expanding markets, and conducting employee training. While these measures are not directly aimed at R&D, they can significantly boost overall operational efficiency and competitiveness, thereby indirectly promoting innovation activities. For example, non-R&D subsidies can help upstream enterprises better cope with fluctuations in raw material prices, ensuring production stability and continuity. Midstream enterprises can utilize non-R&D subsidies to improve production processes and equipment performance, reduce production costs, and enhance product quality, all of which are crucial elements in the innovation process.

The demand for non-R&D subsidies by upstream and midstream enterprises also reflects their disadvantaged market positions. Given the high technical difficulty and weaker financial strength of these enterprises in the industry chain, support from non-R&D subsidies can alleviate their financial pressures in daily operations, allowing them to focus more on technological innovation. Additionally, non-R&D subsidies can help these enterprises gain a more favorable position in market competition by improving product quality and reducing costs, thereby enhancing their market competitiveness and further promoting technological innovation.

Based on the above analysis, we propose the following hypotheses:

H3a: 

R&D subsidies have a stronger incentivizing effect on the innovation of upstream and midstream enterprises in the EV industry chain compared to downstream enterprises.

H3b: 

Non-R&D subsidies have a stronger incentivizing effect on the innovation of upstream and midstream enterprises in the EV industry chain compared to downstream enterprises.

3. Materials and Methods

3.1. Data

We selected data from listed and New Third Board (also known as the National Equities Exchange and Quotations, NEEQ) companies in the EV industry from 2013 to 2022 as the sample. The selection is based on the following considerations:

• The implementation of the “Energy-Saving and New Energy Vehicle Industry Development Plan (2012–2020)” in 2012 emphasized a pure electric drive strategy, gradually increasing the mileage standards for subsidized vehicles [39], which led to rapid development in the EV industry. The New Third Board formally began operations in January 2013, so 2013 is chosen as the starting year for the sample. The central government subsidy policy for EV purchases, which lasted for 13 years, nominally ended on 31 December 2022, making the sample’s ending year.
• Given that subsidy standards are set by governments at various levels, from central to local, collecting comprehensive data directly from the subsidy issuance end is challenging. Therefore, data are collected from the subsidy receiving end to ensure comprehensiveness, effectiveness, and reliability.
• The organizational and management systems of existing listed companies and those on the New Third Board are well established and can represent the development of the industry. Moreover, companies listed on the New Third Board are often high-quality enterprises in emerging industries, which aligns with the focus of this study on the EV industry.
• Although the EV industry is an emerging industry without a clear definition, it can be analyzed using some existing capital market indices, such as the CSI NEVs Index, Wind NEVs Index, and Tonghuashun NEV Concept Stocks. These indices offer a basic way to judge which enterprises belong to the NEV industry and thus which enterprises may be suitably selected for this study’s sample.
• Considering the research needs, data on government subsidies, R&D subsidies, corporate R&D expenditures, and financial information are necessary. Listed and New Third Board companies have good disclosure mechanisms, allowing for the collection of the data needed for this study from their annual disclosure reports, Wind, and CSMAR databases.

To ensure the scientific validity of the empirical results, the study excluded companies registered in traditional tax havens such as the British Virgin Islands, Cayman Islands, Luxembourg, and Bermuda. After combining information on companies’ main business activities and revenue structure, invalid information and missing samples were excluded. Ultimately, a sample of 74 companies was selected for the years 2013–2022, including 18 upstream suppliers, 32 midstream component manufacturers, and 24 downstream vehicle manufacturers and charging pile enterprises.

3.2. Variables

• Dependent Variable: Innovation output (Patent). Ying and He [26] define innovation output using patent applications. Given the time lag between patent application and approval, the number of patent applications effectively reflects the innovative achievements of enterprises. To mitigate endogeneity issues, this study examines the impact of current subsidies on the innovation of enterprises in the subsequent period. To eliminate the influence of company operation scale, this indicator is set as the ratio of the company’s patent applications to every CNY 10 million in sales revenue. The indicator is further divided into invention patents (Patent_Inv) and non-invention patents (Patent_other).
• Independent Variables: R&D subsidy intensity (RDsub) and non-R&D subsidy intensity (NRDsub). RDsub and NRDsub are measured by the ratio of the amount of R&D subsidies and non-R&D subsidies to the current period’s sales revenue per CNY 100, respectively. Previous research typically considers government subsidies as a whole to analyze their impact on enterprise innovation. However, government subsidies include non-R&D subsidies that are not directly related to technological innovation activities [16,24,25]. To distinguish the effects of R&D subsidies from non-R&D subsidies on EV enterprise innovation, we reference the works of Chen et al. [40] and Ying and He [26], manually collecting data related to fiscal subsidies for research, development, science, technology, innovation, patents, projects, and high-tech from the financial statement notes in the annual reports of listed companies. After careful reading, we selected every R&D subsidy, summed all types of R&D subsidies for each company in the current year, and calculated the R&D subsidy intensity. The non-R&D subsidy amount was calculated by subtracting the R&D subsidy from the total government subsidy.
• Control Variables: Referring to the research by Ying and He [26] and Zhang et al. [41], factors that may influence the effectiveness of government subsidies were selected as control variables. These include return on assets (ROA), total assets (Size), debt-to-asset ratio (Lev), net cash flow from operations (CFO), proportion of R&D staff (Rate_Staff), shareholding ratio of the largest shareholder (Rate_Top1), shareholding ratio of institutional investors (Rate_Inst), proportion of independent directors (Rate_Board), dual role of CEO and chairman (Dual), and the nature of enterprise ownership (Sta).

The definitions and calculation methods of all variables involved in this study are presented in

Table 1. Definitions of variables.

Variables	Meaning	Calculation Method	Source
Patent	Patent Application Quantity	The number of Patent Applications/Sales Revenue (per 10 million)	CNRDS and Wind database
Patent_Inv	Invention Patent Application Quantity	The number of Invention Patent Applications/Sales Revenue (per 10 million)	CNRDS and Wind database
Patent_other	Other Patent Application Quantity	The number of Other Patent Applications/Sales Revenue (per 10 million)	CNRDS and Wind database
RDsub	R&D Subsidy Intensity	100×R&D Subsidy/Sales Revenue	Firm annual reports
NRDsub	Non-R&D Subsidy Intensity	100×Non-R&D Subsidy/Sales Revenue	Firm annual reports
ROA	Return on Assets	(Total Profit + Financial Expenses)/Average Total Assets	Wind database
Size	Total Assets	ln (Total Assets)	Wind database
Lev	Asset-Liability Ratio	Total Debt/Total Assets	Wind database
CFO	Net Cash Flow Cash	Net Flow from Operating Activities/Total Assets	Wind database
Rate_Staff	R&D Staff Ratio	The number of R&D Staff/The number of Employees	Wind database
Rate_Top1	Shareholding Ratio of the Largest Shareholder	Number of shares held by the largest shareholder/Total shares of listed companies	Firm annual reports and Wind database
Rate_Inst	Shareholding Ratio of Institutional Investors	Number of shares held by institutional investors/Total shares of listed companies	Firm annual reports and Wind database
Rate_Board	Independent Director Ratio	Number of independent directors/Number of board members	Firm annual reports and Wind database
Dual	Dual Role	If the CEO also serves as the chairman of the company, equals to 1; otherwise 0	Firm annual reports and Wind database
Sta	Ownership	If the controller is a state-owned enterprise, equals to 1; otherwise 0	Wind database

.

3.3. Econometric Models

Given that government subsidies are not uniformly distributed across the entire industry but are targeted towards specific enterprises with developmental advantages, there is a tendency towards “picking winners” [19]. This selective behavior by the government can lead to endogeneity issues. To effectively address the endogeneity problem, we construct a structural equation model. This model comprises three equations: two models analyzing the allocation process of government R&D subsidies and non-R&D subsidies (Model 1 and Model 2) and a model examining the impact of R&D and non-R&D subsidies on the R&D output of EV enterprises (Model 3). The simultaneous equations model is estimated using the three-stage least squares (3SLS) method [42,43], providing consistent and reliable parameter estimates, thereby allowing for a more accurate evaluation of the impact of government subsidies on enterprise innovation output. To further investigate the mediating role of financial constraints in the effect of these two types of subsidies on enterprise innovation, we reference the study by [44] to construct Models (4) and (5), and we combine it with Model (3) for the mediation effect test.

$$RDsu{b}_{it}={\alpha }_1+{\displaystyle \sum Controls}+{\mu }_i+{\lambda }_t+{\epsilon }_{it},$$

(1)

$$NRDsu{b}_{it}={\alpha }_2+{\displaystyle \sum Controls}+{\mu }_i+{\lambda }_t+{\epsilon }_{it},$$

(2)

$$Paten{t}_{it+1}={\alpha }_3+{\beta }_{1}RDsu{b}_{it}+{\beta }_{2}NRDsu{b}_{it}+{\displaystyle \sum Controls}+{\mu }_i+{\lambda }_t+{\epsilon }_{it},$$

(3)

$$S{A}_{it}={\alpha }_4+{\beta }_{3}RDsu{b}_{it}+{\beta }_{4}NRDsu{b}_{it}+{\displaystyle \sum Controls}+{\mu }_i+{\lambda }_t+{\epsilon }_{it},$$

(4)

$$Paten{t}_{it+1}={\alpha }_5+{\beta }_{5}RDsu{b}_{it}+{\beta }_{6}NRDsu{b}_{it}+{\beta }_{7}S{A}_{it}+{\displaystyle \sum Controls}+{\mu }_i+{\lambda }_t+{\epsilon }_{it},$$

(5)

where i represents the enterprise, t represents time, α and β are the coefficients to be estimated, μ and λ represent individual and time fixed effects, respectively, and ε is the random error term. The SA index, proposed and constructed by Hadlock and Pierce [45], is calculated as $SA=-0.737\times Size+0.043\times {Size}^2-0.04\times Age$, where Size represents company size and Age represents company age. This index is robust as it is not influenced by endogenous financing variables and is widely used to measure the financing constraints faced by listed companies [33,46]. The SA value is negative, and the larger the absolute value, the more severe the financing constraints, implying greater difficulty in obtaining external financing.

In the benchmark regression, we use the panel OLS regression method, incorporating both individual and time fixed effects, with clustered robust standard errors. This approach allows us to control for individual and time characteristics, thereby reducing omitted variable bias and improving the reliability of the estimation results. In the robustness check, we replace the dependent variable in Model (3) with the number of patents. Given the count nature of patent data, we use a negative binomial regression (NBR). The regression analysis for this study was conducted using Stata 17.

4. Results

4.1. Descriptive Statistics

Table 2. Descriptive statistics.

Variables	Obs	Mean	Std.Dev	Min	Max
Patent	740	0.465	0.767	0	1.327
Patent_Inv	740	0.191	0.253	0	0.818
Patent_other	740	0.274	0.359	0	1.032
RDsub	740	0.758	0.962	0	7.362
NRDsub	740	1.039	1.932	0	21.072
ROA	740	0.063	0.171	−0.439	0.441
Size	740	23.649	2.964	18.852	29.324
Lev	740	0.466	0.335	0.032	2.719
CFO	740	0.045	0.081	−0.191	0.284
Rate_Staff	740	0.107	0.073	0.007	0.384
Rate_Top1	740	0.347	0.147	0.091	0.693
Rate_Inst	740	0.348	0.269	0	0.823
Rate_Board	740	0.373	0.054	0.333	0.667
Dual	740	0.284	0.427	0	1
Sta	740	0.245	0.478	0	1

presents the descriptive statistics of the main variables. The mean value of innovation output intensity (Patent) is 0.465, with a standard deviation of 0.767, indicating significant variation in R&D output among the sample enterprises, suggesting room for improvement in overall innovation levels. The mean of invention patents (Patent_Inv) is 0.191, while the mean of other patents (Patent_other) is 0.274, indicating that the proportion of invention patents is relatively low. The mean value of R&D subsidy intensity (RDsub) is 0.758, with a standard deviation of 0.962, while the mean value of non-R&D subsidy intensity (NRDsub) is 1.039, with a standard deviation of 1.932. This suggests that enterprises generally receive a certain level of subsidies, with non-R&D subsidies being more prevalent than R&D subsidies, but there is significant variability in subsidy levels across enterprises.

4.2. Cross-Sectional Dependence (CD) Test

Before conducting panel data analysis, it is crucial to test for cross-sectional dependence. We used two methods for this purpose: the Baltagi–Pinnoi LM test and the Pesaran CD test [47,48]. As shown in

Table 3. Results of cross-sectional dependence tests.

Test	Statistics	Prob.
Baltagi–Pinnoi LM test	227.819 ***	0.000
Pesaran CD test	8.631 ***	0.000

Note: Null hypothesis is that no cross-sectional dependence exists within the panel data. *** p < 0.01.

, the results indicate that both tests reject the null hypothesis of “no cross-sectional dependence” at the 1% significance level. Therefore, in the subsequent empirical analysis, we use a fixed effects model with clustered robust standard errors to account for the effects of cross-sectional dependence.

4.3. Benchmark Regression Results

This study is based on data from 2013 to 2022, using the 3SLS regression method. The results are presented in

Table 4. Benchmark regression results.

	(1)	(2)	(3)	(4)	(5)
	RDsub	NRDsub	Patent	Patent_Inv	Patent_Other
RDsub			0.063 *** (3.638)	0.021 *** (3.709)	0.042 *** (3.798)
NRDsub			0.073 (1.441)	0.037 (1.359)	0.036 *** (3.127)
ROA	0.151 (0.422)	0.444 (0.125)	0.029 *** (7.719)	0.025 *** (11.783)	0.042 *** (7.263)
Size	−0.173 *** (−3.808)	−0.175 *** (−4.018)	−0.024 * (−1.775)	−0.075 ** (−2.271)	−0.023 * (−1.892)
Lev	−0.031 (−0.845)	−0.057 (−0.920)	−0.126 *** (−5.491)	−0.137 *** (−4.211)	−0.094 *** (−3.223)
CFO	0.291 *** (6.282)	0.395 *** (5.943)	0.473 *** (7.438)	0.302 *** (6.623)	0.501 *** (7.340)
Rate_Staff	0.046 (1.058)	0.027 (1.338)	0.011 *** (17.514)	0.016 *** (12.731)	0.014 *** (10.105)
Rate_Top1	0.236 (1.174)	0.154 (0.813)	0.023 * (1.971)	0.036 ** (2.223)	0.015 * (1.765)
Rate_Inst	0.227 (1.042)	0.155 (1.287)	0.622 (1.445)	0.451 (1.477)	0.039 ** (2.283)
Rate_Board	0.095 (0.793)	0.164 (0.431)	0.131 ** (2.382)	0.097 *** (3.679)	0.176 (0.980)
Dual	0.093 (1.134)	0.081 (1.117)	0.379 *** (5.286)	0.304 *** (8.463)	0.279 * (1.693)
Sta	−0.022 ** (−2.116)	−0.048 *** (−7.626)	−0.027 *** (−6.105)	−0.019 *** (−4.868)	−0.034 ** (−2.181)
FE	Y	Y	Y	Y	Y
_cons	0.904 *** (5.190)	0.956 *** (7.938)	3.626 *** (4.547)	1.596 *** (5.965)	3.789 *** (6.531)
N	740	740	740	740	740
Adj.R2	0.472	0.397	0.282	0.173	0.190

Note: t statistics are in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

. Columns 1 and 2 examine the factors influencing the allocation of R&D and non-R&D subsidies to EV enterprises, indicating the criteria the government consider when distributing subsidies. The regression results show that the government prioritizes well-funded, smaller non-state-owned enterprises. The preference for firms with ample liquid assets suggests that these companies are in a growth phase with significant development potential. Compared to larger firms, smaller companies are more sensitive to policy changes, aligning with the findings of Wang et al. [34]. Cao and Wang [49] argue that state-owned enterprises are more likely to receive policy support due to their close ties with the government. However, our results contradict this, showing a government preference for private enterprises.

Column 3 assesses the impact of subsidies on overall corporate innovation. The regression coefficient for R&D subsidies is 0.063, that is, every additional CNY 1 million in R&D subsidies motivates enterprises to increase their number of patent applications by 0.63 in the next year. However, the impact of non-R&D subsidies becomes insignificant, suggesting that non-R&D subsidies may not significantly promote technological innovation in enterprises. Previous studies that examined government subsidies as a whole to assess their impact on EV enterprise innovation might have some bias. The results demonstrate that R&D subsidies have a significant incentivizing effect on the innovation output of EV enterprises, whereas non-R&D subsidies do not. This supports hypotheses H1a and H1b. This finding aligns with real-world expectations that government R&D subsidies reduce the innovation costs for enterprises, encouraging them to increase innovation activities and promote industrial innovation and upgrading.

Columns 4 and 5 investigate the impact of R&D and non-R&D subsidies on enterprise innovation strategies. The coefficients for R&D subsidies are 0.021 and 0.042, respectively, indicating that for additional patent application by enterprises, 0.21 are invention patents and the remaining 0.42 are other types of patents. Invention patents represent substantive innovation by enterprises, while utility model and design patents represent strategic innovation. The regression results show that as the number of invention patents increases, the number of applications for other types of patents increases more, with non-invention patents being twice the number of invention patents. This suggests that some EV enterprises might be more inclined to innovate with the primary goal of obtaining subsidies rather than genuine R&D-driven innovation. This result supports hypothesis H2, indicating that while R&D subsidies significantly enhance the quantity of innovation, enterprises exhibit strategic innovation behavior. Moreover, non-R&D subsidies significantly promote the application for other types of patents but have an insignificant effect on invention patents, further supporting the notion of strategic innovation behavior among EV enterprises.

Additionally, the regression results reveal several key factors affecting innovation output. Larger enterprises (Size), those with higher financial leverage (Lev), and state-owned enterprises (Sta) are more averse to high-risk R&D activities, leading to correspondingly lower R&D output. This could be because they focus more on stable business strategies, avoiding the risks and uncertainties associated with heavy R&D investment. In contrast, enterprises where the CEO also serves as the board chairman (Dual), those with higher shareholding ratio of the largest shareholder (Rate_Top1), higher independent director ratios (Rate_Board), higher proportion of R&D staff (Rate_Staff), higher returns on assets (ROA), and higher cash flows (CFO) tend to have higher R&D output. This indicates that enterprises with more powerful general managers, a greater number of independent directors, sound financial conditions—especially sufficient cash flow—and abundant technical resources are more likely to undertake high-risk R&D investments to pursue higher future returns. The adequacy of these resources enables enterprises to invest in projects that may bring high returns in the future, promoting continuous transformation and upgrading, and investing more in potential innovation fields.

4.4. Heterogeneity Analysis

Referring to [27], we categorize EV enterprises based on their position in the industrial chain into upstream, midstream, and downstream sectors to further analyze the differential impacts of R&D and non-R&D subsidies on these enterprises. This analysis reveals the differences in innovation dynamics and strategies within the industrial chain. The results are presented in

Table 5. The impact of R&D and non-R&D subsidies on EV enterprises from the perspective of the industry chain.

		(1)	(2)	(3)
		Patent	Patent_Inv	Patent_Other
Panel A: Upstream enterprises	RDsub	0.085 *** (3.683)	0.057 *** (3.367)	0.028 *** (4.042)
	NRDsub	0.103 ** (2.315)	0.076 *** (4.003)	0.027 (0.993)
	Controls	Y	Y	Y
	FE	Y	Y	Y
	_cons	2.976 *** (5.916)	1.398 *** (3.986)	2.913 *** (4.577)
	N	180	180	180
	Adj.R2	0.332	0.356	0.278
Panel B: Midstream enterprises	RDsub	0.064 *** (4.088)	0.031 *** (3.162)	0.033 *** (3.285)
	NRDsub	0.079 (1.285)	0.042 ** (2.307)	0.037 (1.259)
	Controls	Y	Y	Y
	FE	Y	Y	Y
	_cons	0.897 *** (3.295)	1.504 *** (4.226)	0.807 *** (4.338)
	N	320	320	320
	Adj.R2	0.329	0.214	0.347
Panel C: Downstream enterprises	RDsub	0.049 *** (3.907)	0.008 *** (3.496)	0.041 *** (4.126)
	NRDsub	0.052 (0.865)	0.011 (1.427)	0.041 *** (3.226)
	Controls	Y	Y	Y
	FE	Y	Y	Y
	_cons	2.063 *** (5.427)	1.719 *** (3.865)	1.704 *** (4.982)
	N	240	240	240
	Adj.R2	0.275	0.253	0.377

Note: t statistics are in parentheses; ** p < 0.05, *** p < 0.01. To keep the table concise, the regression results for control variables have been omitted, displaying only the regression results for the main independent variables. Additionally, the regression results for the effect of government subsidies on different stages of the industrial chain are not shown.

.

The incentive effect of R&D subsidies is significantly higher for upstream and midstream enterprises compared to downstream enterprises. Specifically, for every additional CNY 1 million in R&D subsidies, upstream enterprises increase their patent applications by 0.85, midstream enterprises by 0.64, while downstream enterprises only increase by 0.49. Upstream enterprises typically face pressure related to technological transformation and the supply of intermediate products, playing a crucial role in technological innovation and core R&D within the industrial chain. Consequently, they are more capable of effectively utilizing R&D subsidies for substantive innovation activities. In contrast, downstream enterprises are more constrained by market demand and consumer preferences, with their innovation incentives driven more by market competition than government subsidies, resulting in a weaker response to R&D subsidies. These findings validate hypothesis H3a.

Statistically, non-R&D subsidies do not have a significant impact on the overall innovation of midstream and downstream enterprises but do significantly promote innovation in upstream enterprises. Economically, the incentive effect of non-R&D subsidies on upstream and midstream enterprises is higher than on downstream enterprises, and this effect is greater than that of R&D subsidies. Hypothesis H3b is not fully supported.

Further analysis reveals that within the technological innovations promoted by R&D subsidies, upstream enterprises primarily focus on invention patents, accounting for approximately 67%, while downstream enterprises tend to apply for non-invention patents, such as design patents and utility model patents, making up about 84%. This reflects differences in innovation strategies among enterprises positioned differently within the industrial chain. Compared to upstream enterprises, downstream enterprises exhibit more strategic innovation behavior. Upstream enterprises emphasize technological innovation, whether from R&D or non-R&D subsidies, utilizing all available resources for substantive innovation activities to establish their technological leadership in the industrial chain through invention patents. In contrast, downstream enterprises focus more on product design, usage methods, and other non-invention patents to enhance product recognition and competitiveness.

From a theoretical perspective, the response differences of enterprises in different positions within the industrial chain to R&D and non-R&D subsidies reflect the internal technology transfer and innovation dynamics within the chain. Government subsidies, as external stimuli, can promote innovation activities to a certain extent, but their effectiveness is constrained by factors such as technological linkages within the industrial chain, market demand, and competitive landscape. Therefore, in formulating subsidy policies, it is essential to consider the characteristics and innovation needs of enterprises at different positions within the industrial chain and adopt differentiated policy measures. This approach maximizes the innovation-stimulating effect of government subsidies and promotes continuous innovation and upgrading across the entire industrial chain.

4.5. Mediation Effect Analysis

In the previous sections, we explored the theoretical mechanisms through which R&D and non-R&D subsidies impact enterprise innovation, particularly within the EV industry. Financing constraints are a significant real-world barrier to innovation for enterprises, and both types of subsidies play a crucial role in alleviating this barrier. Their impact can be divided into direct effects and indirect effects [33,34]. The direct effect refers to subsidies providing enterprises with innovation capital directly, while the indirect effect refers to subsidies helping enterprises overcome information asymmetry in external financing, thus enabling them to secure sufficient funds for innovation activities.

The results in

Table 6. The mediation effect of R&D and non-R&D subsidies on EV enterprises.

	Full Sample		Upstream		Midstream		Downstream
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	SA	Patent	SA	Patent	SA	Patent	SA	Patent
RDsub	−0.213 *** (−4.058)	0.050 *** (3.612)	−0.370 *** (−3.330)	0.059 *** (4.165)	−0.187 ** (−2.181)	0.053 *** (3.976)	−0.105 (−0.985)	0.045 *** (4.677)
NRDsub	−0.362 *** (−3.161)	0.051 (1.024)	−0.271 *** (−3.039)	0.078 *** (3.182)	−0.379 *** (−3.604)	0.057 (0.876)	−0.517 *** (−3.475)	0.035 (1.280)
SA		−0.061 *** (−3.439)		−0.092 *** (−3.394)		−0.059 *** (−3.996)		−0.032 *** (−2.937)
Controls	Y	Y	Y	Y	Y	Y	Y	Y
FE	Y	Y	Y	Y	Y	Y	Y	Y
_cons	1.048 *** (3.803)	3.458 *** (3.265)	2.227 *** (4.563)	1.241 *** (4.372)	2.621 *** (3.692)	1.574 *** (3.148)	1.069 *** (4.691)	2.020 *** (3.327)
N	740	740	180	180	320	320	240	240
Adj.R2	0.149	0.256	0.137	0.282	0.129	0.258	0.148	0.267

Note: t statistics are in parentheses; ** p < 0.05, *** p < 0.01. To keep the table concise, the regression results for control variables have been omitted, displaying only the regression results for the main independent variables.

confirm the role of R&D and non-R&D subsidies in alleviating financing constraints, especially for upstream and midstream enterprises. Overall, both types of subsidies can incentivize enterprise innovation by mitigating financing constraints. The mediating effect of R&D subsidies accounts for 20.62% of the total effect (calculated as [(−0.213) × (−0.061)]/0.063 = 20.62%), while the proportion for non-R&D subsidies is 30.25%. Specifically, upstream enterprises are the most significantly affected by financing constraints, with the proportions of the innovation incentive effect due to alleviation of these constraints being as high as 40.05% and 24.21% for R&D and non-R&D subsidies, respectively. Midstream enterprises are relatively less affected, but there is still a noticeable mediating effect, accounting for 17.24% and 28.31% for R&D and non-R&D subsidies, respectively. Downstream enterprises, due to relatively sufficient funding, show no significant mediating effect from R&D subsidies; however, non-R&D subsidies exhibit a clear mediating effect on innovation, accounting for 31.82% of the total effect.

In summary, both R&D and non-R&D subsidies play crucial roles in alleviating financing constraints and promoting enterprise innovation but with different focal points. R&D subsidies effectively alleviate financing constraints for upstream and midstream enterprises, while non-R&D subsidies mainly improve the overall financial condition of enterprises to mitigate financing constraints. This type of subsidy has a similar effect on enterprises located at different positions within the industrial chain, as all enterprises can use these subsidies to increase cash flow, reduce financial risk, and enhance operational stability. This finding not only aids in understanding the mechanisms behind innovation activities in EV enterprises but also provides important references for the government in formulating more precise and effective policies to promote the sustainable development of the EV industry.

4.6. Robustness Check

In the baseline regression, we used the number of patent applications per CNY 10 million of revenue as the measure of technological innovation output. Now, we use the number of patent applications as the dependent variable and perform a robustness check using a count model. Given that the dependent variable (number of patents) is a non-negative integer with a mean of 23.35 and a standard deviation of 42.63, the variance is approximately 80 times the mean, indicating potential overdispersion. Although the panel Poisson regression remains consistent even with overdispersion, an NBR model may be more efficient. Therefore, we used a panel NBR model for the robustness check. The NBR model results are presented in

Table 7. Negative binomial regression results.

	(1)	(2)	(3)	(4)	(5)
	Patent	Patent	Patent	Patent_Inv	Patent_Other
RDsub	0.215 *** (3.566)		0.157 *** (3.318)	0.179 *** (3.924)	0.153 *** (3.415)
NRDsub		0.248 *** (2.981)	0.176 (1.192)	0.132 (1.388)	0.138 *** (3.992)
Controls	Y	Y	Y	Y	Y
FE	Y	Y	Y	Y	Y
_cons	5.713 *** (3.805)	4.163 *** (4.121)	4.906 *** (3.674)	5.066 *** (4.828)	4.559 *** (3.398)
N	740	740	740	740	740
pseudo.R2	0.192	0.181	0.137	0.085	0.093

Note: t-values in parentheses. *** p < 0.01. To keep the table concise, the regression results for control variables have been omitted, displaying only the regression results for the main independent variables. Additionally, the regression results for the effect of government subsidies are not shown.

and are largely consistent with the regression results in Table 4, thereby confirming the robustness of our empirical findings.

5. Discussion

This study delves into the multifaceted impact of government subsidies on the innovation dynamics within China’s EV industry. Our findings underscore the critical role that both R&D and non-R&D subsidies play in fostering technological advancements and enhancing innovation strategies among Chinese EV enterprises. These results not only corroborate the existing literature but also provide deeper insights into the mechanisms through which government interventions can stimulate sectoral innovation.

The Chinese government’s proactive stance on supporting the EV industry through a variety of incentive programs has been pivotal in driving the sector’s rapid development. As highlighted in [4], these initiatives are designed to transition China from a major automotive nation to an automotive powerhouse, thereby enhancing its global competitiveness. Our research aligns with the evaluation by [50] which emphasizes the comprehensive policy support provided to EV enterprises, including R&D subsidies that address market failures and non-R&D subsidies that alleviate financing constraints [51].

From a broader perspective, the development of China’s EV industry must be contextualized within the larger narrative of global automotive industry evolution, the ongoing energy crisis, and environmental protection imperatives. The transition towards electric mobility is a critical component of addressing climate change and promoting sustainable development [11]. As noted by Wu et al. [6], evolutionary policy incentives have significant strategic implications for the sustainable development of EVs in China. The shift from ICEVs to EVs not only reduces oil dependency but also mitigates the environmental impact of the automotive sector [11,14,15].

The global energy crisis underscores the urgency of accelerating the adoption of clean energy technologies. China’s commitment to green development through the promotion of EVs serves as a model for other nations grappling with similar challenges. By fostering a robust EV industry, China is not only contributing to global efforts to combat climate change but also positioning itself as a leader in the next wave of automotive innovation [15,50].

Future research should explore the long-term effects of these subsidies on the overall competitiveness of Chinese EV enterprises in the global market. Additionally, examining the interplay between different types of subsidies and their cumulative impact on innovation could provide more nuanced insights. As our study indicates, a holistic approach that integrates policy support with industry-specific needs is essential for sustaining the momentum of technological innovation in the EV sector.

6. Conclusions

Using a sample of listed and New Third Board EV companies in China from 2013 to 2022, this study empirically analyzes the impact of R&D and non-R&D subsidies on the innovation output of EV enterprises. The main conclusions are as follows: (1) R&D subsidies have a significant positive impact on the R&D output of EV enterprises. While non-R&D subsidies can improve the overall financial condition of enterprises, they do not have a significant direct impact on innovation output. In particular, R&D subsidies can effectively alleviate the constraints on enterprise innovation activities posed by financing constraints, providing more innovation capital. (2) R&D subsidies prompt enterprises to focus more on innovation quality, leading to a tendency to apply for more invention patents, whereas non-R&D subsidies do not significantly influence the choice of innovation strategies by enterprises. (3) R&D subsidies have varying impacts on enterprises at different positions in the industry chain. Upstream and midstream enterprises are more likely to be incentivized by R&D subsidies to pursue technological innovation, while downstream enterprises rely more on non-R&D subsidies to improve their operational status and are more inclined towards strategic innovation behaviors. (4) Both R&D and non-R&D subsidies indirectly promote technological innovation by alleviating financing constraints on enterprises. This effect is particularly significant for upstream enterprises, where the alleviation of financing constraints accounts for a substantial proportion of the mediating effect of R&D subsidies.

The findings of this study may have the following implications: (1) Establishing and improving policy supervision and evaluation mechanisms. As the government increases the subsidies each year, it should shift the acceptance criteria from quantity and form to quality and content, encouraging enterprises to pursue a balanced development of innovation quantity and quality. The government should uphold principles of fairness, justice, and transparency in the allocation and supervision of subsidies, limit intervention in economic activities, improve government services, and reduce rent-seeking opportunities. (2) Optimizing subsidy policies and focusing on the comprehensive development of the industry chain. When formulating industrial policies, the government should consider the development of the entire EV industry chain. Differentiated R&D subsidy policies should be developed for enterprises at different positions in the chain to promote overall industry development and enhance overall innovation capability.

While this study provides valuable insights, it also has some limitations. The analysis is primarily based on available data from 2013 to 2022, which may not fully capture long-term trends and effects. Additionally, the study focuses on Chinese EV enterprises, and the findings may not be directly applicable to other countries with different industrial and policy contexts. Future research should continue to explore the long-term effects of subsidies, optimize subsidy mechanisms to prevent rent-seeking behaviors, and compare subsidy policies across different countries to draw lessons for policy improvement in China.