Research on the Impact of Data Elements on the Innovation Capability of New Energy Vehicle Enterprises—Evidence from Chinese Listed Companies

Abstract

Based on panel data from 173 Chinese listed companies in the new energy vehicle industry from 2016 to 2023, this study constructs a two-way fixed effects model to examine the impact of data elements on corporate innovation capability. Systematically address issues and validate results through multiple measurement approaches, including robustness checks, instrumental variables methods, moderation effect analysis, and heterogeneity tests. The results indicate: (1) Data elements significantly enhance the innovation capability of new energy vehicle enterprises, and a conclusion that remains robust after a series of endogeneity and robustness checks. (2) Moderating effect tests reveal that human resources strengthen the relationship between data elements and corporate innovation capability. (3) Heterogeneity analysis shows that, in terms of capital sources, data elements have a more substantial promoting effect on the innovation capability of domestic enterprises compared to foreign-funded ones; regionally, the innovation-driven effect of data elements is more pronounced in eastern and central China than in the western region. This study not only offers practical guidance for new energy vehicle enterprises to allocate data elements and human resources effectively, but also provides an empirical basis for policymakers to formulate market-oriented data policies, thereby offering a new perspective for enhancing the innovation capabilities of new energy vehicle enterprises.

1. Introduction

1.1. Background and Motivation

As the world’s largest automotive market, China began deploying its new energy vehicle (NEV) industry in the early 21st century and has continually promoted its development through a comprehensive policy system. The New Energy Vehicle Industry Development Plan (2021–2035), issued in 2020, outlined key objectives such as strengthening the guiding role of leading enterprises and building world-class industrial clusters, thereby charting a clear course for industrial upgrading [1]. Driven by policy guidance and market forces, China’s NEV industry has achieved leapfrog development. By the end of 2024, the production and sales of NEVs in China reached 12.88 million and 12.86 million units, respectively, representing year-on-year growth of 34% and 35%. For ten consecutive years, China has led globally in production volume, sales volume, brand recognition, and competitive strength [2].

However, behind the rapid growth, the new energy vehicle industry is facing a series of severe challenges both domestically and internationally. Domestically, the market is experiencing a slowdown in growth, and the industry is falling into an internal competition phase, experiencing the dilemma of “revenue growth without profit growth” [3]. Internationally, the sector faces mounting obstacles in its global expansion, including elevated tariff barriers and stringent technical standards in key markets such as Europe and the United States, which collectively challenge the industry’s worldwide strategic layout [4]. At a deeper level, the industry’s inherent “big but not strong” issue persists, manifested in: persistent “chokepoint” risks in core foundational software and automotive-grade chips [5]; inefficient innovation coordination between upstream and downstream supply chain segments, which gives rise to isolated innovation silos [6]. Breaking through these bottlenecks requires the industry to shift from scale expansion to innovation-driven growth.

Against this backdrop, China is vigorously advancing its “Digital China” strategy, positioning data elements as a new engine driving economic and social development [7]. In this process, the data element, as a new key production element, is no longer just an auxiliary resource, but has use value through data collection, processing, integration and collaboration, and integrates with different production scenarios. Through resource integration, accelerating knowledge transfer, updating production technology and production concept, and is deeply embedded in technology research and development, production and manufacturing, industrial chain collaboration and enterprise decision-making, it significantly changes the form of production function, realizing industrial integration and value chain reconstruction, and thus becomes the core driving force to improve the innovation ability of new energy vehicle enterprises [8]. For the new energy vehicle industry, which is currently undergoing a critical phase of transformation, the value of data elements is particularly prominent. New energy vehicles themselves serve as massive carriers for data generation and application, with every stage of the R&D process relying on the driving force of massive data.

In summary, systematically analyzing how data elements influence the innovation capabilities of new energy vehicle enterprises, and thoroughly revealing their intrinsic mechanisms and pathways for enhancing corporate innovation capacity not only deepens and expands the theoretical understanding of data element value, but also the key to coping with the current industrial challenges, overcoming development bottlenecks, and empowering China’s new energy vehicle industry to build sustainable new competitive advantages on the global stage.

1.2. Literature Review

Existing research generally holds that data is not only a new type of production factor but also fundamentally alters the way of value creation through its unique attributes—notably non-exclusivity, economies of scale, renewability, and strong pervasiveness [9]. Chadeiaux [10] was the first to place data alongside traditional factors such as labor, capital, and land, emphasizing its pivotal role as a fundamental means of production. Yu and Wang [11] further demonstrated through their study of production factors and changes in production functions that data elements are as critical as land, labor, and capital. Lin and Meng [12] point out that data elements can help enterprises more efficiently acquire or utilize complementary assets, thereby enhancing dynamic innovation capabilities. Data production factors have penetrated all aspects of enterprise production and operation [13]. Enterprises can reduce the uncertainty they face and improve their productivity through data accumulation [14]. At the same time, the marketization of data elements is conducive to driving the digital transformation of enterprises [15], promoting integration between the digital and real economies [16], and enhancing economic resilience [17].

Empirical studies in recent years have provided strong evidence for the economic value of data elements. Müller [18] found that big data can increase average corporate productivity by 3% to 7%, with the speed of data circulation having a particularly significant effect on promoting innovation. Xu and Zhao [19] introduced endogenous growth models to analyze the direct impact and spillover effects of data elements on economic growth, confirming their potential to drive macroeconomic expansion. Farboodi and Veldkamp [14] incorporated data elements into growth models, revealing that data elements input enhances productivity by reducing uncertainty. Jones and Tonetti [20] explored how different data ownership models influence economic growth rates by integrating data elements into growth frameworks. Wei et al. [21] empirically demonstrated that improved data value capabilities significantly enhance the resilience of global value chains.

In the field of new energy vehicles, research on factors influencing corporate innovation primarily explores two dimensions: external environment and internal structure. Regarding external factors, scholars focus on the impact of policy environments, market demand, and competitive landscapes on corporate innovation. He and Cao [22] found an inverted U-shaped relationship between policy incentives and technological innovation in new energy vehicles, suggesting that moderate policy guidance facilitates corporate technological innovation advancement. Regarding internal factors, scholars have primarily examined the roles of corporate governance structures, resource allocation, and talent reserves. For instance, Gao [23] constructed a DEA model and found that six internal factors, including the shareholding ratio of the largest shareholder, significantly influence a company’s technological innovation efficiency.

Although the aforementioned studies have laid the groundwork for understanding the innovation mechanisms of new energy vehicle enterprises, most remain confined to the realm of traditional factors, with the critical role played by data elements yet to be systematically analyzed. Against the backdrop of deep integration between intelligent connectivity and electrification, data elements are profoundly reshaping the innovation pathways and competitive landscape of new energy vehicle enterprises. This transformation occurs through driving R&D iteration, empowering production optimization, and reconstructing business models. Therefore, incorporating data elements into the theoretical framework of innovation in new energy vehicle enterprises is not only theoretically necessary but also aligns more closely with the practical demands of industrial development.

1.3. Contributions

The contributions of this paper are primarily reflected in the following three aspects:

(1)

From a research perspective, this study integrates data element theory with the innovation practices of China’s new energy vehicle industry. While existing research has extensively explored data elements and the innovative development of new energy vehicle enterprises separately, most studies have primarily focused on traditional factors influencing innovation in these enterprises, with few combining data elements and innovation development in new energy vehicle enterprises. Based on large-scale automotive industry data, this study empirically examines the direct impact of data elements on the innovation capabilities of new energy vehicle enterprises for the first time, offering new insights into the development of innovation capabilities within the new energy vehicle sector.

(2)

From a theoretical perspective, this study identifies and validates the synergistic mechanism linking “data elements—human resources—innovation capabilities.” It not only confirms the innovative value of data elements for new energy vehicle enterprises but also reveals, through the introduction of human resources’ moderating effect, the pivotal bridging role of high-quality human capital in the processes of data interpretation, transformation, and application.

(3)

Methodologically and practically, this study adopts a variety of measurement methods, such as the instrumental variable method and grouping regression, to systematically deal with the endogenous problem. It further conducts heterogeneity analysis based on regional differences across eastern, central, and western China and firm ownership structures, thereby enhancing the rigor and applicability of its findings. The findings not only provide practical guidance for new energy vehicle enterprises on effectively allocating data elements and human resources but also offer empirical evidence for government departments to formulate differentiated policies for the marketization of data elements. This research holds significant reference value for promoting the high-quality development of new energy vehicle enterprises.

2. Theoretical Analysis and Research Hypotheses

As a core production factor in the digital economy era, the utilization of data elements within enterprises facilitates optimized production decisions [24], reduces societal innovation costs and trial-and-error risks [25], and drives economic growth [20]. Data can manifest as informational content or as new insights derived through technological processing tailored to specific contexts [26]. Only when data flows can it give full play to its capacity, this capacity stemming from the collection, synthesis, and application of these circulating entities [27]. The ability of real-time data collection and dissemination can narrow the information gap in the market, enhance resource allocation efficiency, and reduce operational uncertainties for enterprises, thereby stimulating increased innovation output [28]. Through in-depth analysis and mining of data, enterprises can identify emerging technological trends and market demands, thereby driving product innovation and industrial upgrading.

Currently, academic research on the impact of data elements on corporate innovation activities has not reached a consensus. Liu and Wang [29] propose that data elements enhance the enhancement of corporate innovation capabilities through four dimensions: the formation of digital innovation strategies, the evolution of innovation ecosystems, the restructuring of innovation organizational systems, and the advancement of innovation paradigms. Ma et al. [30] found that integrating data elements with traditional production factors can reduce innovation uncertainty, thereby boosting corporate innovation levels. Li and Zhang [31], adopting a resource orchestration theory perspective, propose that data elements significantly enhance the innovation capabilities of specialized, refined, distinctive, and innovative enterprises. Hu and Xia [32] contend that the utilization of data as a production factor enables numerous startups to leverage data for innovation, thereby reducing innovation costs and indirectly promoting increased innovation outputs. Therefore, the study proposes the following hypothesis:

Hypothesis 1.

Data elements exert a significant positive influence on the innovation capabilities of new energy vehicle enterprises.

Innovative talent emerges from the combined effects of educational cultivation and labor market utilization, while bursts of innovation result from the synergistic interaction of human capital accumulation and allocation [33]. As a vital repository of knowledge, inspiration, and skills, human capital not only serves as a key determinant of corporate innovation income but also constitutes the core component of innovation costs [34]. Human capital plays a central role in the R&D process, serving as a valuable resource for technological innovation. Strengths in human capital enhance regional technological innovation levels and play the driving role of human capital accumulation in the process of scientific and technological innovation [35].

According to the resource-based theory, a firm’s competitive advantage stems from the integration of heterogeneous resources. As a new strategic resource, the value release of data elements is highly dependent on the activation and transformation of human resources—particularly those with data analysis and technological R&D capabilities [36]. The aggregation of data elements facilitates talent mobility and allocation, thereby enhancing human capital development [37]. Specifically, the data elements themselves are only a collection of original information. Only through human agency—by integrating consumer demand data to strategize future product enhancements and cultivating big data insights—can enterprises target market opportunities and elevate manufacturing innovation capabilities [38].

In the technology-intensive and knowledge-intensive new energy vehicle industry, the impact of data elements on innovation capabilities particularly highlights dependence on talent. Highly skilled professionals possess strong creative thinking and extensive cutting-edge knowledge reserves [39], can more effectively identify valuable external knowledge, and increase the possibility of innovation through rapid absorption, utilization, and integration of knowledge [40]. Simultaneously, highly skilled professionals and R&D personnel possess robust technical expertise, providing strong technological support for sustained corporate innovation and more effectively converting resource investments into innovative outputs [41]. Innovation in new energy vehicles requires R&D teams to integrate massive vehicle condition data with user feedback. Rapid shifts in market demand compel enterprises to leverage human resources to transform user data into product innovation decisions, preventing the idling or misallocation of data resources. Therefore, Hypothesis 2 is proposed:

Hypothesis 2.

Human resources play a positive moderating role in the relationship between data elements and the innovation capabilities of NEV enterprises.

3. Research Design and Variable Description

3.1. Sample Selection and Data Sources

This study utilizes annual data from publicly listed new energy vehicle (NEV) enterprises between 2016 and 2023 as the research sample. The term “NEV enterprises” as defined herein primarily refers to listed companies whose core business encompasses the manufacturing of new energy vehicles, the R&D and production of core components (such as batteries, motors, and electronic control systems), and the supply of key materials. To ensure the validity of the sample data and the reliability of the research conclusions, the initial sample was screened based on the following criteria:

• Exclusion of financial and insurance companies;
• Removal of ST (Special Treatment) firms, *ST firms, and enterprises delisted during the observation period;
• Elimination of enterprises with substantial missing data;
• Winsorization of key continuous variables to mitigate the influence of outliers on regression results.

After processing, a total of 173 companies met the requirements.

For the two variables of corporate innovation capability (INN) and data elements (Data), since their raw data are count variables exhibiting numerous zeros and right-skewed distributions, we adopt the natural logarithmic transformation based on existing research [42,43]. This transformation addresses the mathematical definition issue of taking logarithms of zeros while effectively mitigating the impact of extreme values. It brings the variable distribution closer to normality and enhances the economic interpretability of subsequent regression coefficients. The NEV enterprise data were primarily sourced from: THS Concept Stocks, Chinese Research Data Service Platform (CNRDS), CSMAR Database, Annual report data were obtained from the CNINFO platform.

3.2. Model Specification and Variable Description

3.2.1. Model Construction

To verify the impact of data elements on the innovation capabilities of new energy vehicle (NEV) enterprises, based on existing research [44,45], the following dual fixed-effects model is constructed:

$${\mathrm{I}\mathrm{N}\mathrm{N}}_{\mathrm{i}\mathrm{t}}={\mathsf{\chi }}_0+{\mathsf{\chi }}_1{\mathrm{D}\mathrm{a}\mathrm{t}\mathrm{a}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\chi }}_2\sum {\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+\sum \mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}+\sum \mathrm{F}\mathrm{i}\mathrm{r}\mathrm{m}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(1)

In Equation (1):

The dependent variable ${\mathrm{I}\mathrm{N}\mathrm{N}}_{\mathrm{i}\mathrm{t}}$ represents the innovation capability of NEV enterprise i in year t; the core explanatory variable ${\mathrm{D}\mathrm{a}\mathrm{t}\mathrm{a}}_{\mathrm{i}\mathrm{t}}$ denotes the application level of data elements by NEV enterprise i in year, with its coefficient denoted by ${\mathsf{\chi }}_1$; $\sum {\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}$ refers to the control variables affecting the innovation capability of NEV enterprises, with its coefficient denoted by ${\mathsf{\chi }}_2$; $\sum \mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}$ and $\sum \mathrm{F}\mathrm{i}\mathrm{r}\mathrm{m}$ represent the year fixed effect and firm fixed effect, respectively; ${\mathsf{\chi }}_0$ denotes the conventional intercept; ${\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$ is the error term.

3.2.2. Dependent Variable

This study selected innovation capability as the dependent variable, denoted by INN. It is difficult to directly measure the economic or technological value of patents. A review of existing literature reveals that most scholars [43,46,47] employ patent volume as an indicator for evaluating innovation levels. Acs et al. (1992) [48] also demonstrated that patent metrics reliably reflect changes in innovation output. Compared to granted patents, patent applications provide a more timely and comprehensive reflection of a firm’s current innovation activities. The patent grant process is typically lengthy and uncertain, involving lags of three years or longer [49,50]. Using granted data introduces severe observational delays, failing to capture firms’ innovation decisions and resource allocation efficiency promptly. Furthermore, whether a patent is ultimately granted depends not only on innovation quality but also on non-capability factors such as corporate filing strategies (e.g., payment of fees, voluntary withdrawal) and external review factors (e.g., examiner strictness), introducing substantial noise into grant data. In contrast, patent application dates are closer to R&D completion times, which can more sensitively measure the innovation vitality, and the application behavior can also more directly reflect the initial attempt and output of enterprises to turn innovation into achievements. Therefore, this paper selects $\log (\mathrm{Number} \mathrm{of} \mathrm{patent} \mathrm{applications}+1)$ [51,52] as the dependent variable.

3.2.3. Explanatory Variable

This study selected data elements as the explanatory variable, denoted by Data. Given the difficulty in quantifying data elements within NEV enterprises and the lack of suitable measurement methods, it is challenging to characterize them through specific indicators. To reasonably depict the application level of data elements in enterprises, this study adopts the approach of Saunders and Tambe [53], using Python 3.12 programs to extract keywords reflecting the application of data elements from listed companies’ annual reports, calculate the total word frequency, and take the $\log (\mathrm{total} \mathrm{word} \mathrm{frequency}+1)$ as the explanatory variable.

To highlight the dynamic process of data transforming from a resource into a production factor and the multiple heterogeneous attributes of data elements, this study categorizes data element keywords into five dimensions based on the definition of data elements by the China Academy of Information and Communications Technology (CAICT) and the practices of Ma [54] and Chao et al. [55]: “Data Processing Dimension,” “Data Infrastructure Dimension,” “Data Technology Support Dimension,” “Data Application Scenario Dimension,” and “Data Security and Governance Dimension.” The details are presented in

Table 1. Composition of Data Element Feature Words for New Energy Vehicle Enterprises.

Dimension	Feature Words
Data Processing Dimension	Big data, Data collection, Driver behavior data, Customer data, Data preprocessing, Data mining, Data analysis, Intelligent data analysis, Simulation data analysis, Data visualization
Data Infrastructure Dimension	Data center, Data platform, Data analysis platform, Data transmission, Internet of Vehicles, Intelligent connected
Data Technology Support Dimension	Cloud computing, Distributed computing, Cognitive computing, Internet of Things, Cyber-physical system, Intelligent manufacturing technology, Autonomous driving technology
Data Application Scenario Dimension	Design optimization, Collaborative manufacturing, Production capacity analysis, Production scheduling optimization, Inventory cost control, Quality control, Autonomous driving, Remote monitoring and diagnosis, Energy management optimization, Driving habit evaluation, Personalized service, Precision marketing, Intelligent customer service, Customer satisfaction survey, Supply chain data analysis
Data Security and Governance Dimension	Data security and privacy protection, Network security, Encryption equipment, Digital currency, Data governance mechanism

.

3.2.4. Moderating Variable

This study selected human resources as the moderator variable, denoted by Rga. Following the approach of Li et al. [32], human capital is measured by the ratio of R&D personnel to total employees, with higher values indicating greater human capital endowment.

3.2.5. Control Variables

To enhance research precision, a series of control variables were incorporated, including: Firm size (Size, logarithm of total assets); Ownership concentration (Top1, shareholding proportion of the largest shareholder); Government subsidies (Gov); Fixed asset ratio (Fixed, net fixed assets to total assets ratio); Asset-liability ratio (Lev, total liabilities to total assets ratio).

3.3. Data Analysis Method

To systematically examine the impact of data elements on the innovation capabilities of new energy vehicle enterprises and ensure robust and in-depth research conclusions, this paper’s empirical analysis will proceed step-by-step according to the following strategy:

• Descriptive statistics will be conducted to preliminarily outline the distribution characteristics and fluctuation ranges of core variables, forming an initial understanding and assessment of the sample data to provide essential prerequisites for subsequent econometric analysis.
• Benchmark regression analysis will serve as the critical step for testing the core research hypothesis H1, focusing on identifying whether data elements exert a significant positive impact on the innovation capabilities of new energy vehicle enterprises, thereby providing direct evidence for Hypothesis 1.
• To enhance the reliability of the benchmark regression conclusions, this study conducts robustness tests and endogeneity treatment. By replacing variable measurement methods, adjusting the sample scope, and employing instrumental variable methods, we eliminate interference from model specification errors, measurement errors, and potential endogeneity issues, thereby demonstrating the reliability of the benchmark regression conclusions.
• After confirming the robustness of the main effects, this study further introduces a moderation analysis of human resources to examine whether human resources exert a positive moderating effect on the influence of data elements on the innovation capabilities of new energy vehicle enterprises, thereby testing Research Hypothesis (H2) and revealing the synergistic mechanism among “data elements-talent-innovation”.
• To uncover the differential manifestations of data elements’ impact on innovation capabilities across diverse groups and provide targeted, differentiated policy and management insights for governments and enterprises, we conclude with a heterogeneity analysis. This approach enhances the theoretical depth and practical value of the research.

4. Empirical Tests and Results Analysis

4.1. Descriptive Statistics

The study comprises 1384 valid statistical observations. The results of the descriptive analysis are shown in

Table 2. Descriptive Statistics Results.

Variable	Observations	Mean	Std. Dev.	Min	Max
INN	1384	3.124	1.431	0	6.436
Data	1384	3.099	0.908	1.099	5.684
Size	1384	22.57	1.030	20.50	25.16
Gov	1384	17.17	1.372	13.59	20.65
Fixed	1384	0.202	0.108	0.0125	0.519
Lev	1384	0.460	0.168	0.0988	0.884
Top1	1384	0.292	0.140	0.0479	0.685

Note: “INN” refers to Innovation Capability; “Data” refers to Data Elements; “Size” refers to Firm size; “Gov” refers to Government subsidies; “Fixed” refers to Fixed asset ratio; “Lev” refers to Asset-liability ratio; “Top1” refers to Ownership concentration.

:

For the overall sample, the average value of corporate innovation capability (INN) is 3.124 with a standard deviation of 1.431. The significant disparity between the maximum and minimum values indicates substantial variation in innovation capabilities across enterprises.

The average data element utilization level (Data) is 3.099 with a standard deviation of 0.908, ranging from 1.099 to 5.684. This suggests that while some enterprises have recognized the importance of data elements and begun applying them to business development and technological innovation, there remains considerable heterogeneity in data adoption levels across firms.

All other variables fall within reasonable statistical ranges, demonstrating the robustness of the dataset.

4.2. Baseline Regression Tests

The multicollinearity test results show that all VIF values are below 10, indicating the suitability for regression analysis. Therefore, we empirically test Hypothesis 1 (H1) using the OLS method to analyze the impact of data elements on the innovation capability of new energy vehicle (NEV) enterprises.

The baseline regression results presented in

Table 3. Baseline Regression Results.

	(1)	(2)	(3)
	INN	INN	INN
Data	0.498 ***	0.280 ***	0.131 ***
	(12.38)	(6.14)	(2.89)
_cons	1.580 ***	1.914 ***	−9.881 ***
	(12.16)	(13.76)	(−6.93)
Control Variables	NO	NO	YES
Year/Individual FE	NO	YES	YES
N	1384	1384	1384
R2	0.0998	0.140	0.226

Notes: *** indicate significance levels of 1%; “INN” refers to Innovation Capability; “Data” refers to Data Elements.

demonstrate that data elements consistently exert a significantly positive influence on corporate innovation capability after incorporating control variables. Specifically, when controlling for time and individual fixed effects, the coefficient for data elements is 0.280 (significant at the 1% level) without control variables, and becomes 0.131 (still significant at the 1% level) after including control variables. The coefficients for data elements remain significantly positive throughout these specifications. These results confirm H1, indicating that data elements indeed have a significantly positive impact on the innovation capability of NEV enterprises.

4.3. Robustness Tests and Endogeneity Treatment

4.3.1. Robustness Tests

To ensure the robustness of the research conclusions, the following methods were employed to conduct robustness tests on the results:

1.

Replace the dependent variable. Use $\log (\mathrm{Number} \mathrm{of} \mathrm{patents} \mathrm{obtained}+1)$ as the dependent variable. The primary reason lies in the fact that while patent application volume can promptly reflect innovation activities, it may include low-quality or strategic applications that fail to pass examination, introducing measurement errors. Patent grants, having undergone official substantive examination, better represent the quality and ultimate value of innovation outcomes. If the core findings persist after variable substitution, it indicates that the research conclusions simultaneously capture both the quantity and quality of innovation, further enhancing the credibility and explanatory power of the empirical results. The results are shown in Column (1) of

Table 4. Robustness Test Results.

	Alternative Dependent Variable	Excluding Municipalities
	(1)	(2)
Data	0.097 **	0.116 **
	(2.10)	(2.48)
_cons	−8.521 ***	−10.137 ***
	(−5.89)	(−6.86)
Control Variables	YES	YES
Year/Individual FE	YES	YES
N	1384	1241
R2	0.173	0.224

Notes: ** and *** indicate significance levels of 5% and 1%; “Data” refers to Data Elements.

.

2.

Re-estimated the results after excluding samples from the four municipalities of Beijing, Shanghai, Tianjin, and Chongqing. The main reason is that as the leading area of national strategy, municipalities directly under the central government enjoy special preferential treatment in data element market construction, new energy vehicle industry policy and innovation ecology. Their distinct data resource endowments and corporate innovation capabilities may introduce outlier observations that distort estimation results. By excluding these samples, we test whether the promotional effect of data elements on innovation capacity holds under more general conditions. The results are presented in Column (2) of Table 4.

The results consistently show significantly positive effects, confirming the robustness of our conclusion that data elements have a significantly positive impact on the innovation capability of NEV enterprises.

4.3.2. Endogeneity Treatment

Given potential bidirectional causality and omitted variable bias, the baseline regression results may suffer from endogeneity issues. To mitigate this concern, we employ the instrumental variable (IV) approach for robustness testing. Following the methodology of Tian and Liu et al. [56], we construct an instrumental variable 1 as the interaction term between the spherical distance to the nearest node city of the “Eight Vertical and Eight Horizontal” optical fiber backbone network and the provincial postal service volume (time-varying). The rationale is twofold: First, the “Eight Vertical and Eight Horizontal network“ serves as critical infrastructure for digital communication, where proximity to node cities typically implies better network conditions and higher efficiency in data acquisition and transmission. Second, provincial postal service volume reflects the development level of regional postal services and the intensity of information flow. Higher postal activity indicates more frequent information exchange and material circulation, facilitating data element mobility and sharing. The test results are shown in

Table 5. Endogeneity Test Results.

	Instrument Variable 1		Instrument Variable 2
	(1)	(2)	(1)	(2)
Variables	First Stage	Second Stage	First Stage	Second Stage
Data		0.499 *		1.176 *
		(1.88)		(1.91)
IV1	0.006 ***
	(6.03)
IV2			0.013 ***
			(3.17)
K-PaapWald rk F-statistic	21.278		21.278
K-Paap rk LM statistic	6.146 **		6.146 **
Control Variables	YES	YES	YES	YES
Year/Individual FE	YES	YES	YES	YES

Notes: *, **, and *** indicate significance levels of 10%, 5%, and 1%; “Data” refers to Data Elements; “IV1” and “IV2” represent Instrumental variable 1 and Instrumental variable 2, respectively.

under Instrument Variable 1.

Furthermore, post offices serve as critical infrastructure for traditional information and material circulation. The post office density in 1984 reflects the foundational conditions of regional communication networks in earlier periods. Regions with higher post office density in the early stages often gained first-mover advantages in subsequent informatization development, thereby influencing current data element accumulation and circulation efficiency. Mobile phones, as core carriers of modern information transmission, have a penetration rate that directly reflects the convenience of regional information exchange and the activity level of data interaction. Higher mobile phone penetration implies lower costs for enterprises to acquire and transmit data, leading to more frequent and extensive utilization of data elements. Therefore, following the approach of Jin and Cai [57], we select the interaction term between the number of post offices per 10,000 people in 1984 and the mobile phone penetration rate of the previous year as the instrumental variable 2. The test results are shown in Table 5 under Instrument Variable 2.

The first-stage regression results show that the coefficients of the instrumental variables are 0.006 and 0.013, respectively, both significantly positive at the 1% level, indicating a strong correlation between the instrumental variables and the core explanatory variable. The second-stage regression results reveal that the coefficients of Data are 0.499 and 1.176, respectively, both significant at the 10% level. The Kleibergen–Paap rk LM statistic rejects the null hypothesis of underidentification at the 1% significance level, and the Kleibergen–Paap Wald rk F statistic exceeds the critical value of the Stock-Yogo weak identification test at the 10% level, suggesting that weak instrument issues can be largely ruled out. Therefore, the selected instrumental variables are appropriate, further confirming the robustness of our earlier conclusions.

4.4. Moderating Effect Analysis of Human Resources

The moderating effect results are shown in

Table 6. Moderating Effect Results.

	INN	INN
	(1)	(2)
Data	0.130 ***	0.149 ***
	(2.87)	(3.24)
Data × Rga		0.971 ** (2.43)
_cons	−9.919 ***	−10.038 ***
	(−6.95)	(−7.04)
Control Variables	YES	YES
Year/Individual FE	YES	YES
N	1384	1384
R2	0.226	0.230

Notes: ** and *** indicate significance levels of 5% and 1%; “INN” Innovation Capability; “Data” refers to Data Elements; “Rga” refers to human resources.

. The coefficient for data elements is 0.149, significant at the 1% level, indicating a slightly enhanced main effect of data elements. This suggests that after accounting for the moderating role of human resources, the positive impact of data elements on innovation capability remains significant and is somewhat strengthened. The coefficient of the interaction term is 0.971, significant at the 5% level. Thus, H2 has been confirmed that Human resources play a positive moderating role in the relationship between data elements and the innovation capabilities of NEV enterprises. Specifically, the more abundant the human resources, the stronger the promoting effect of data elements on innovation capability. This implies that NEV enterprises must invest in both data elements and sufficient human resources to fully unlock the innovative potential of data elements.

4.5. Heterogeneity Analysis

4.5.1. Heterogeneity in Capital Sources

Innovation capabilities vary across enterprises with different capital sources. Therefore, the research sample was grouped into foreign-funded enterprises and non-foreign-funded enterprises for grouped regression analysis. Columns (1) and (2) in

Table 7. Test Results on the Heterogeneity of Corporate Capital Sources.

	Foreign-Funded Enterprises	Non-Foreign-Funded Enterprises
	(1)	(2)
Data	0.090	0.144 **
	(1.04)	(2.42)
_cons	−6.094 **	10.447 ***
	(−2.23)	(−4.35)
Control Variables	YES	YES
Year/Individual FE	YES	YES
N	540	844
R2	0.235	0.182

Notes: ** and *** indicate significance levels of 5% and 1%; “Data” refers to Data Elements; (1) and (2), respectively, represent the results of grouped regression based on the sources of corporate capital.

present the grouped regression results showing the impact of data elements on the innovation capabilities of NEV enterprises under different capital sources.

Column (1) shows a coefficient of 0.090 for Data, which is not significant. Column (2) reveals a coefficient of 0.144 for Data, significant at the 5% level. This indicates that within non-foreign-funded enterprises, data elements exert a more pronounced positive influence on the innovation capabilities of NEV companies. Foreign-funded enterprises, constrained by cross-border data flow restrictions, rely more heavily on their internal R&D systems for innovation due to their global technological accumulation and brand advantages. Their inherent technological strengths reduce dependence on external data, resulting in the insignificant role of data elements. Non-foreign-funded enterprises, however, rely more on data elements to compensate for technological shortcomings. They need to integrate external resources through data elements, and domestic data policies are better suited to local enterprises, leading to higher data utilization efficiency.

4.5.2. Regional Heterogeneity

The development of China’s eastern, central, and western regions exhibits significant imbalances, with marked disparities across areas in data infrastructure development, industrial policy support, technical talent reserves, and marketization levels. This regional heterogeneity may lead to divergent effects of data elements on corporate innovation. Therefore, following the standard practice of the National Bureau of Statistics and regional economic research in China [58,59], the entire sample was divided into three regional subsamples—Eastern, Central, and Western—based on the registered location of enterprises, and regression analysis was conducted separately for each.

Columns (1), (2), and (3) in

Table 8. Results of Regional Heterogeneity Test.

	Eastern Region	Central Region	Western Region
	(1)	(2)	(3)
Data	0.090 *	0.240 *	0.407
	(1.81)	(1.88)	(1.65)
_cons	10.047 ***	15.772 ***	0.582
	(−6.80)	(−2.82)	(0.07)
Control Variables	YES	YES	YES
Year/Individual FE	YES	YES	YES
N	1044	260	80
R2	0.214	0.355	0.331

Notes: * and *** indicate significance levels of 10% and 1%; “Data” refers to Data Elements; (1), (2), (3) represent the test results for regression grouped by region.

present the grouped regression results for the impact of data elements on the innovation capabilities of new energy vehicle enterprises across different regions. In the eastern region, the Data coefficient is 0.090, significant at the 10% level, indicating that data elements exert a significant positive effect on the innovation capabilities of new energy vehicle enterprises in the eastern region. In the central region, the Data coefficient is 0.240, also significant at the 10% level. The positive impact of data elements is slightly stronger than in the eastern region, but the level of significance is the same. In the western region, the Data coefficient is 0.407 but fails the significance test, indicating that the promotional effect of data elements on innovation for enterprises in the western region is not significant.

The reasons for this disparity may lie in the fact that eastern regions possess well-developed data infrastructure and ample digital talent, enabling effective integration of data elements into R&D processes. Central regions, which have taken over industrial transfers from the east, are gradually expanding data application scenarios, but their data industry chains remain under development, thus showing a significant impact with marginal effects slightly higher than in the east. In contrast, western regions suffer from weak data infrastructure and a shortage of digital talent, making it difficult to convert data elements into innovation outcomes, resulting in negligible impact. Differences in sample size also reflect the regional development pattern of China’s new energy vehicle industry, characterized by “strong east, weak west.” Eastern regions represent the earliest starting point for China’s new energy vehicle industry and the area with the most concentrated policy support. They host leading enterprises such as BYD, NIO, and XPeng, along with a large number of supporting industrial chain enterprises. The industrial ecosystem is mature, and the number of enterprises is far greater than in the central and western regions. In contrast, central and western regions entered the new energy vehicle sector later, with traditional manufacturing or resource-based industries dominating. New energy vehicle enterprises are sparsely distributed, particularly in remote western areas where weak industrial foundations naturally limit the number of companies.

5. Conclusions and Recommendations

5.1. Conclusions

Using 173 new energy vehicle enterprises in China from 2016 to 2023 as the research sample, this empirical analysis examines the impact of data elements on the innovation capabilities of new energy vehicle enterprises, yielding the following conclusions:

• Data elements exhibit a significant positive correlation with the innovation capabilities of new energy vehicle enterprises. That is, data elements can enhance the innovation capabilities of these enterprises, and this conclusion remains valid after undergoing a series of robustness and endogeneity tests.
• Moderation analysis reveals that human resources exert a significant positive moderating effect on the relationship between data elements and innovation capacity. This indicates that the promotion effect of data elements on innovation capacity intensifies as human resources become more abundant.
• The impact of data elements on innovation capabilities exhibits heterogeneity. From the perspective of capital sources, data elements significantly boost innovation capabilities more effectively in non-foreign-invested new energy vehicle enterprises than in foreign-invested ones. Regionally, the innovation-driving effect of data elements is markedly stronger in eastern and central regions than in western regions.

5.2. Policy Recommendations

5.2.1. Strengthen the Foundational Role of Data Elements and Enhance Human Resource Synergies

To fully unleash the innovation-driving potential of data elements in NEV enterprises, coordinated efforts are needed in both data element provision and human resource adaptation.

Regarding data element provision, accelerating the establishment of industry-wide data sharing mechanisms is crucial. First, government-led NEV sector data trading platforms should be developed to clarify property rights and transaction rules for core data categories including vehicle operation data, battery lifecycle data, and charging network data. This will reduce corporate data acquisition costs and break down data silos among automakers, component suppliers, and mobility platforms, facilitating cross-entity, cross-process data flows. Second, increased investment in data infrastructure is essential, with resource allocation prioritizing central and western regions to narrow regional disparities, while also ensuring the continued development of eastern China’s data centers.

Concerning human resource synergies, both quantitative expansion and qualitative improvement are required. To address the shortage of interdisciplinary talent, universities and enterprises should jointly establish specialized programs cultivating professionals with dual competencies in technology and data analytics. Concurrently, a “targeted talent subsidy” system should be implemented, offering tax reductions or social security subsidies for high-quality talent recruitment, particularly incentivizing eastern experts to support central/western enterprises through technical consulting or short-term assignments. Furthermore, corporate internal talent development mechanisms should be enhanced by motivating R&D personnel to upgrade data tool application capabilities, forming an integrated “attraction-cultivation-utilization-retention” talent strategy.

5.2.2. Implement Differentiated Policies to Promote Balanced Innovation Capability Enhancement

Given the heterogeneous characteristics of corporate capital sources and regional development, targeted policies should be implemented.

At the enterprise level, for domestic enterprises, priority should be given to supporting the establishment of data-driven R&D systems, encouraging collaborations between enterprises and research institutions to build NEV data innovation laboratories. Simultaneously, preferential policies should be provided for data procurement, storage, and analysis expenditures to enhance their motivation for data investment. For foreign-funded enterprises, it is necessary to optimize cross-border data flow management, allowing them to utilize local market user data for technological iteration in their Chinese R&D centers while ensuring data security, and guiding them to establish data-sharing chains with domestic enterprises to foster complementary advantages in data technologies between foreign and domestic firms.

At the regional level, a tiered development framework of “eastern leadership, central catch-up, and western foundation-building” should be established. The eastern region should focus on high-end applications of data elements, leveraging industrial clusters in the Yangtze River Delta and Pearl River Delta to promote deep integration between data elements and critical “bottleneck” technologies. The central region can undertake the transfer of data industry chains from the east, developing segments such as data labeling and processing, while accumulating technical experience by providing services like data cleaning and feature extraction for eastern enterprises. The western region needs to strengthen foundational data application capabilities, prioritizing support for local automakers’ data management systems and subsidizing the purchase of data collection equipment to gradually cultivate data application competencies.

5.2.3. Strengthening Policy Support Systems and Optimizing the Data Innovation Ecosystem

The innovation-driven development powered by data elements requires robust institutional safeguards. First, governments should accelerate the improvement of market-oriented regulatory frameworks for data elements to provide clear guidance for corporate data applications. Concurrently, through coordinated efforts between industry associations and judicial authorities, issues such as data theft and misuse should be promptly addressed to mitigate legal risks in corporate data utilization. Second, a dynamic monitoring and evaluation mechanism should be established, employing a corporate data innovation capability index to regularly track industry progress and enable timely policy adjustments. For instance, when persistent lagging data element adoption is identified in western regions, additional investments in local data training facilities could be allocated; when domestic enterprises demonstrate deficiencies in data security technologies, targeted data protection training programs should be organized to ensure continuous policy relevance to innovation needs. Third, policy instrument coordination must be enhanced. Special funds should be created, adopting a “performance-based reward” approach to support corporate data technology R&D, with particular emphasis on domestic enterprises and central/western regions. Furthermore, industry-academia-research-application collaboration should be strengthened by engaging automakers, universities, and data service providers to regularly publish industry data demand catalogs, thereby facilitating precision matching of innovation resources.

Although this study encompasses key factors such as human resources, corporate property rights, and regional location, it may still fail to capture all influencing elements fully. Future research could employ methods such as field surveys and expert interviews to more comprehensively identify and incorporate additional factors, thereby constructing a more comprehensive and holistic analytical framework. Furthermore, subsequent studies may expand their scope to include non-listed enterprises, entities across various segments of industrial chains, or conduct cross-industry comparisons to enhance the universality and specificity of their conclusions.