The Spillover Effect of Customer Data Assets on Suppliers’ Green Innovation

Abstract

Green innovation is important for environmental sustainability and long-term ecological balance. Using 1129 observations of Chinese listed firms spanning 2014–2024, combined with text mining method to quantify data assets, this paper empirically examines the impact of customer data assets on suppliers’ green innovation. Our model is integrated with fixed effects for both industry and year. We find that there is a significant improvement in suppliers’ green innovation when customers have more data assets, with a one-notch improvement in the customer data assets of a customer firm. This results in an overall 0.06 increase in supplier green innovation output. Specifically, the spillover effect is more pronounced when there is a shorter geographic distance between suppliers and customers, as well as higher customer concentration. After conducting a variety of endogeneity tests, our results are robust. The mechanism analysis shows that customer data assets facilitate supplier digital transformation and improve supplier operational capacity. The heterogeneity analysis also reveals stronger effects when (1) customers are located in eastern regions, (2) customers belong to technology-intensive industries, (3) suppliers are state-owned enterprises (SOEs), and (4) suppliers face lower financial constraints. Further analysis suggests that customers with more data assets also increase suppliers’ R&D investment and improve green innovation quality. Our research contributes to understanding the spillover effect of customer data assets along the supply chain.

1. Introduction

In the future, global climate change and environmental deterioration will be great challenges for human beings. Green innovation, as a critical strategy, effectively balances economic and environmental performance [1]. According to Chen et al. (2018) and Zhang and Zhu (2019), green innovation refers to the development of green technologies, services, processes and eco-friendly products as well as generation of new ideas or behaviors, which reduce environmental burden or achieve ecological-specified sustainability [2,3]. As key participants in high-quality economic development, firms actively engage in green innovation. Existing studies evidence that green innovation can enhance resource utilization and reduce costs, thereby boosting firms’ financial performance and competitive advantage [3,4,5]. In addition, Chen et al. (2023) find that corporate green technology innovation could improve corporate reputations [6]. Therefore, how to enhance firms’ green innovation has been attracting increasing research and practice attention. Previous studies examined the influencing factors of firms’ green innovation from internal and external perspectives, such as corporate governance, leadership, technological progress, executives’ gender, governmental regulations, market competition, stakeholder pressures and financial resources availability [2,3,4,6,7,8,9,10,11,12,13]. With the development of data, firms collect and generate massive amounts of data through daily business activities or related transactions, gaining huge economic benefits through data production and data application [14]. Current studies explore the impact of digital transformation and big data on green innovation [15,16,17,18]; however, the influence of data assets on green innovation remains unexplored.

In the digital era, data serves as a critical resource in firms’ operation processes, monitoring ability, and decision-making effectiveness, and is regarded as an essential firm asset [19,20]. According to The White Paper on Data Asset Management Practice (Version 4.0), data assets refer to the data resources owned or controlled by firms that have the potential to generate future economic benefits. With the popularization of the concept of data assets, a strand of theoretical research focuses on the conceptualization, valuation, and accounting treatment of data assets [14,19,21,22,23,24]. Specifically, Tambe (2014) [22] and Perrons and Jensen (2015) [19] explore the valuation of corporate data resources and validate the legitimacy of recognizing data resources as assets. Xu et al. (2024) [23] emphasize the necessity of incorporating data assets into the national accounting framework by the cost method. Additionally, with a sample of Chinese companies, several studies empirically examined the effect of data assets from the perspective of firm value, operational efficiency, and sustainable development, as well as firm cost stickiness [25,26,27,28]. Since firms that process data assets are encouraged to reveal information regarding their attributes, data asset information becomes increasingly useful in alleviating the information gap and supplementing traditional financial information disclosure [29]. Therefore, some studies empirically examine the effect of data asset disclosure, and evidence the positive impacts on firms’ nonprofessional investor judgment, market efficiency, and bank lending acquisition [30,31].

However, there has been limited research dedicated to understanding the spillover effect of data assets along supply chains. To fill this gap, this study investigates the effect of customer data assets. Customers significantly influence suppliers’ financial policies and operating performance through their purchase of goods and services, thereby playing a vital role in the supply chain [32]. Specifically, customers’ growth prospects and business stability can enhance suppliers’ expected returns from supply chain collaboration [33]. Conversely, Munir et al. (2020) [34] and Baghersad and Zobel (2021) [35] find that if the important customer is exposed to high risk, suppliers may face great sales uncertainty. Therefore, suppliers are highly sensitive to the operations and activities of their customers. Prior studies further confirm that customer firms’ economic performance transfers along the supply chain and significantly impacts their suppliers’ capital market performance [36,37,38,39].

With the growing emphasis on sustainable development, it is important to explore the impact on green innovation from a supply chain perspective concerning the data assets utilization of customers, and China provides ideal research contexts for exploring such spillover effects. As the world’s largest developing country and a major emerging market, China faces challenges related to environmental pollution and sustainable development. The Chinese government has recognized that the importance of sustainable development and has enacted environmental policies encouraging green practices [40]. Accelerating green innovation is regarded as the key to realizing national “double carbon” goals and achieving long-term sustainable growth. In recent years, data has become a key factor of production and a strategic resource [41]. The Chinese government provides comprehensive support for firm data asset deployment and application. Data assets information disclosure in Chinese firms serve as a valuable model for data asset information provision in the digital era. Moreover, Li et al. (2021) [42] suggest that Chinese firms rely heavily on relationship-based transactions, indicating that information, resources, and strategic orientations are likely to transmit along supply chains through close customer–supplier relationship. Therefore, customer data assets may impact suppliers’ resource allocation and incentives to undertake green innovation. However, the relationship between customer data assets and the spillover of green innovation along the supply chain is unclear. To fill this gap, this study investigates whether, and how, customer data assets influence suppliers’ green innovation in the context of Chinese listed firms.

We argue that customer data assets have a positive effect on suppliers’ green innovation. Specifically, there are two possible channels. On the one hand, data assets, as unique and valuable intangible resources, improve customers’ own production, operational efficiency and risk response capabilities [43]. Due to supply chain spillover effects, customers with stronger performance generate more stable and sustainable demand [44], reducing suppliers’ risk exposure and improving their capital circulation and liquidity [45], which in turn boosts suppliers’ operational efficiency and market value. According to slack resource theory [46,47], better operational efficiency allows suppliers to devote more resources toward green innovation, while the enhanced capabilities help them to overcome barriers like high initial costs and technological uncertainties [48], ultimately promoting green innovation performance.

On the other hand, customers’ data assets facilitate suppliers’ digital transformation and data assets, thus improving their green innovation. Specifically, suggested by stakeholder theory [49], customers’ superior data assets drive suppliers to accelerate digital transformation and enhance their own data asset levels to maintain cooperative relationships and business competitive advantages [50]. This transformation improves suppliers’ green innovation performance. Firstly, abundant data use enables suppliers to have better understanding about business operations and market positions, thereby bringing targeted innovation [25,28]. Secondly, enhanced digital capabilities convey positive signals to investors, not only reducing information asymmetry but also alleviating green innovation financing constraints [29,30].

We assess how the green innovation of Chinese listed firms from 2014 to 2024 are affected by their customers’ data assets. We analyze publicly disclosed information about suppliers’ top five customers and find that customer data assets play a significant role in improving suppliers’ green innovation. And our results are robust after conducting a variety of endogeneity tests. We further examine the role of geographic distance and customer concentration, finding that the impact of customer data assets on supplier green innovation is more pronounced when there is a lower geographic distance between suppliers and customers and a higher customer concentration. In additional tests, we then perform analyses to validate the channels by which customers’ data assets influence suppliers’ green innovation. We find that customers’ data assets facilitate supplier digital transformation captured by the digital transformation index and improve supplier operational performance measured by firm total factor productivity through supply chain transmission mechanisms. The cross-sectional analysis suggests that the spillover effects are more significant when customers are located in eastern regions, and are technology intensive firms, as well as when suppliers are SOEs and have lower financial constraints. Furthermore, this study also finds that customers with higher data assets enhance suppliers’ green innovation quality and spurs supplier R&D investment.

Our study contributes to the literature as follows. Firstly, this study enriches the literature on the economic implications of data assets and provides new evidence along the supply chain. Most studies explore the economic consequences of data assets disclosure, such as [29]. And the empirical studies related to data assets mainly examine the effects of data assets on firm itself, such as firm operational efficiency, innovation, and ESG performance [25,27,28]. As customer–supplier relationships are economically important, this study extends the current literature by identifying that customers’ data assets positively drive suppliers’ green innovation performance. Secondly, it contributes to the literature on factors influencing green innovation by identifying the role of data assets from the supply chain perspective. Previous research indicates that firm-specific characteristics and external environmental factors can improve firm green innovation [1,3,8,9]. Our study spans corporate boundaries and investigates how the data assets of customers affect suppliers’ green innovation along the supply chain.

The remainder of this study is organized as follows. Section 2 presents the literature review and hypothesis development. Section 3 describes the sample selection and research design. And Section 4 reports empirical results. Section 5 concludes the study.

2. Literature Review and Hypothesis Development

2.1. Theoretical Background

To investigate the impact of customer data assets on suppliers’ green innovation, this study develops a theoretical framework based on a resource-based view, stakeholder theory, and slack resources theory.

According to the resource-based theory, the competitive advantages of firms’ sustainable competitive advantage lies in possessing strategic resources that are valuable, rare, inimitable, and non-substitutable [50,51]. In this study, data assets, as scarce and inimitable critical resources, serve as core strategic resources for firms and enhance their responsiveness in dynamic environments [27,52]. Specifically, data assets are crucial for improving firms’ innovation capabilities, operational efficiency, and high-quality development [26,28]. As a form of knowledge asset, data assets also hold significant potential in facilitating informed decision-making, enhancing governance, and promoting sustainable development [53]. In terms of supply chain, customer data assets can generate spillover effects, enabling suppliers to leverage these valuable resources to enhance their green innovation performance.

Slack resources theory answers how operational efficiencies are converted into substantive green innovation performance. Slack resources theory argues that firms are more likely to undertake such long-term and risky innovation activities when they possess sufficient slack and buffer resources [27,47]. In this study, customer data assets enhance supply chain information sharing and coordination, which helps suppliers reduce demand uncertainty, and improve capacity utilization [54,55,56]. These operational improvements generate financial and organizational slack, like cost savings from inventory reduction or human resources reallocated from crisis management, that can be used in green innovation.

Finally, the stakeholder theory further addresses the question of how customer resources and expectations shape suppliers’ behaviors. Stakeholder theory suggests that customers are one of the most important stakeholders [49] and have great impacts on firms’ strategic decisions, such as environmental and innovation strategies [4,57]. Accordingly, when customers have more data assets, suppliers face both incentive and pressure to maintain close, stable cooperative relationships. To align with customers’ digitalized operations and meet their sustainability expectations, suppliers tend to pursue digital transformation and enhance their own level of data assets.

2.2. Literature Review

Climate change and environmental issues are of broad concern. The Chinese government has recognized the importance of sustainable development and enacted environmental policies encouraging green practices [40]. Green innovation refers to the development of green technologies, services, processes, and eco-friendly products, as well as the generation of new ideas and practices, bringing both environmental and economic benefits to firms [2,3]. To maintain competitive advantages and increase corporate reputation [58], firms are motivated to implement green innovation. As environmental concerns intensify, research on driving factors of green innovation has gained attention.

Current studies have identified multiple internal and external factors that affect green innovation. Internally, scholars have examined the role of corporate governance, leadership, technological progress, executives’ gender, non-financial disclosure, and executive traits [1,5,8,9,40,59,60]. With the development of big data, studies evidence that digital transformation and big data strengthen firms’ absorptive capacity for green knowledge and optimize innovation processes [15,16,17,18,61,62], ultimately improving green innovation performance. Moreover, some studies argue that governmental regulations, stakeholder pressures, market competition, financial resources availability, government support, and public policies are crucial external factors for green innovation [2,3,63,64,65,66,67]. In the digital economy, data assets as a core strategic resource in the digital economy can directly and indirectly promote firm green innovation [20]. And few studies investigate the role of data assets on green innovation.

The concept of data assets evolves from data [23]. Data are the fundamental elements that record objective facts and are represented in the form of numbers, text, images and other media. Usually, data is unstructured and in raw form, with little economic value [26]. Data resources refer to the raw data and derivatives that have been identified, collected, processed, stored, and managed, holding potential economic value [7]. Data assets are data resources owned or controlled by firms recorded in physical or electronic form that have the potential to generate future economic benefits [23,30].

Data assets are regarded as a firm’s strategic assets and become key drivers of firms’ future growth and innovation [29,42]. Since they lack accurate methods for quantifying data assets, companies generally choose to disclose textual information about data assets in annual reports to convey the scale and development prospects of their data assets to the public [28]. Therefore, some scholars focus on the effects of data asset disclosure; for example, Li et al. (2021), Sun and Du (2024), Wei et al. (2025), and Qian et al. (2025) [29,30,31,42] evidenced that data asset information can improve nonfinancial investor evaluation judgment, enhance market efficiency, and strengthen bank lending. In addition, other studies primarily focus on the indirect effects of data assets on firm performance and decision-making effectiveness, as well as firm innovation from the perspective of digital transformation and big data deployment [53,68,69,70,71]. Only some empirical studies evidence that data assets can improve firm operating efficiency, facilitate firm decision-making, reduce cost stickiness as well as improve ESG performance [25,26,27,72]. However, the spillover effects of data assets, particularly along upstream and downstream supply chain relationships, remain underexplored. To fill this gap, this study investigates the impact of data asset from supply chain perspective.

In supply chains, suppliers and customers maintain close financial credit relationships and goods transactions. Customers, as an important and ongoing source of revenue, are considered to be the most direct party of suppliers [73]. A rich body of literature has documented the significant influence of customers on suppliers’ outcomes through information spillover effects. For example, customer earning announcements, risk information, and bankruptcy announcements [37,38,74,75,76]. Furthermore, customer information also signals the operational status and prospects of the supply chain and provides useful information to external entities, such as auditors, analysts, and investors [36,77,78]. And some studies examine the role of customer characteristics on suppliers’ behaviors, such as financial decision-making, cost of capital, management earnings and sales forecasts, and tax avoidance [34,35,79,80]. Munir et al. (2020) and Baghersad and Zobel (2021) [34,35] evidence that customer growth prospects and stability enhance suppliers’ collaborative returns, while customer risk increases suppliers’ sales uncertainty. Building on these, we further explore how customer data assets influence suppliers’ green innovation.

2.3. Hypothesis Development

We propose that the spillover effect of customers’ data assets influences suppliers’ operational efficiency and digital transformation, and thus their green innovation. Firstly, customer data assets generate a spillover effect that improves supplier operational efficiency and thereby facilitates their green innovation (customer data assets → suppliers operational efficiency → suppliers green innovation). According to the resource-based view, firms gain sustainable competitive advantages when possessing valuable, scarce, and inimitable resources [51]. Data assets, as unique and heterogeneous resources, are critical drivers for firms’ productivity, growth, and innovation [26,42]. Specifically, customer data assets are helpful to break down information barriers and facilitate supply chain communication [81]. Through information sharing, suppliers gain data-driven insights and obtain accurate operational information, thereby enabling precise demand forecasting, real-time order tracking, and optimized inventory management and production planning [25,54,55,56]. This facilitates optimal resource allocation for supplier firms. In addition, data assets deepen the collaboration between upstream and downstream firms. Customers can share and utilize their data and innovation resources to suppliers and jointly develop production processes, improve product quality, and enhance overall production efficiency [25,26,52]. These effects not only boost suppliers’ operational efficiency but also elevate the supply chain’s overall efficiency and responsiveness by ensuring the smooth flow of information across the network.

In addition, customer data assets improve suppliers’ operational efficiency indirectly by enhancing customers’ own operational capabilities, profitability, and market competitiveness. Specifically, the accumulation and use of data assets reshape customers’ production process, organizational structures and innovation models [66,69], while enhancing their information processing and decision-making capabilities [27,28]. Consequently, customers with more data assets have improved operational capabilities and strong earnings growth. Given the phenomenon of ‘‘prosperity for all, loss for all” along supply chains, suppliers heavily rely on customers’ orders and business operations. When customers have strong earnings growth, their procurement demands are more stable and sustainable [26,27]. This reduces suppliers’ exposure to customer-related shocks and demand uncertainty, helps suppliers to maintain smooth capital circulation, sound liquidity, and optimized operational processes [44,45], thereby enhancing suppliers’ operational efficiency.

Drawing on slack resource theory, improved efficiency and performance generate organizational slack that can be devoted to green innovation [47]. Green innovation typically requires substantial upfront investment, longer payback periods, and higher technological uncertainty [48]. Suppliers who benefit from customer data-driven operational improvements are better positioned to overcome these key barriers associated with green innovation. Therefore, customer data assets improve suppliers’ operational efficiency and increase suppliers’ resource slack, thereby promoting green innovation performance.

Secondly, we argue that customer data assets promote suppliers’ digital transformation, which in turn enhances suppliers’ green innovation capabilities and improve green innovation performance (customers data assets → suppliers digital transformation → suppliers green innovation). According to stakeholder theory, firms should respond to the expectations and pressures of their main stakeholders [4,49]. In supply chains, customers are considered as the most important stakeholders [57]. Wang et al. (2024) [82] suggest that customers could shape suppliers’ strategic orientation, technology adoption, and management practices. As a result, when customers demonstrate superior digital capabilities and higher level of data assets, to maintain the cooperative relationship and their own competitive advantages, suppliers have strong motivation to adjust their business strategies and accelerate digital transformation [50]. Furthermore, data assets and digitalization reshape corporate supply chain management and impose higher requirements on demand responsiveness and diversified services [52]. In order to enhance their own digitalization capabilities, suppliers tend to learn and emulate their customers’ digital technologies as well as data assets development. In addition, data assets facilitate more opportunities for resource sharing and knowledge transfer between suppliers and customers [25,52]. Suppliers can thus access customers’ experience and technical knowledge in digital transformation and data management through knowledge transfer, thereby ultimately enhancing their own digital transformation. Overall, customers’ data assets promote suppliers’ digital transformation.

And the digital transformation, in turn, provides robust foundation for suppliers’ green innovation. Specifically, on the one hand, with abundant use of data, suppliers are more capable in data collection, integration, and management [25,28], and thereby gain precise understanding of internal operational status and external market dynamics (e.g., green technology trends and policy requirements). Consequently, suppliers are able to identify and capture green innovation opportunities and eventually improve green innovation performance. On the other hand, suppliers’ digital transformation and data assets demonstrate their strength in digitalization and convey a more positive signal to external investors [26,30,52]. This can help to reduce the degree of information asymmetry, thereby alleviating the financing constraints on enterprises’ green innovation [28].

Based on the above analysis, this paper proposes the following hypothesis:

H1. 

Customers’ data assets improve suppliers’ green innovation.

Further, we consider the role of geographical distance. The geographical distance between customers and suppliers reflects the degree of connection between them from a spatial perspective [83]. Previous studies suggest that supplier–customer geographic proximity affects supplier decisions, such as innovation, stock price crash risk and risk-taking [84,85]. Specifically, a shorter distance fosters closer economic connections and brings lower information transformation costs [86,87]. This proximity facilitates superior demand forecasting and inventory coordination, leading to improved operational performance [84], which ultimately enhances the supplier’s green innovation. On the other hand, a shorter distance facilitates knowledge spillovers [83]. Suppliers can better obtain data resources through the active learning and imitation of their customers’ digital technologies and data assets. And this process improves suppliers’ data assets and digital transformation, consequently improving green innovation performance. Thus, we posit the following:

H2. 

When there is a shorter geographic distance between customers and suppliers, the spillover effects of customers’ data assets on suppliers’ green innovation are more pronounced.

In addition, we examine the role of customer concentration. Customer concentration reflects a supplier’s dependence on their customers and therefore signals customers’ bargaining power in business relationships [88]. On the one hand, the higher customer concentration indicates a stable cooperative relationship between customers and suppliers [1,80], which facilitates in-depth information exchange and provides favorable environments for suppliers’ green innovation. On the other hand, when customer concentration is higher, suppliers are likely to be more motivated to adjust their own resources to meet the demands brought about by customer data assets.

Thus, we posit the following:

H3. 

When customers have greater bargaining power, the spillover effects of customers’ data assets on suppliers’ green innovation are more pronounced.

3. Data and Research Design

3.1. Data and Sample

We used data of Chinese listed firms from 2014 to 2024. Financial data were obtained from the CSMAR database, while annual reports were collected from the official websites of Shenzhen and Shanghai Stock Exchanges. Following the prior study, we obtained the top five customers data from the CSMAR database, and we excluded the samples without specific customer names, valid names, or non-public firms [84,88,89], since some information about customers’ identity is missing in the database; for example, some customer codes are disclosed in formats such as “JD” and “ALBB”, which are insufficient to identify. In addition, many of the customers are not listed firms. Therefore, we only kept the observations with full disclosure of names and sales figures.

Further, samples with missing variables, observations of financial firms, firms marked ST or ST*, and insolvent firms were excluded. After these screening procedures, our final sample consisted of 1129 unique supplier-customer-year pairs. And our sample size was comparable to those in customer–supplier relationship research based on similar settings in China [83,87,88,89,90]. And all continuous variables were winsorized at the first and 99th percentiles.

3.2. Measures of Customer Data Assets

In this paper, the level of data assets deployment of firms is measured by the logarithm value of the frequency statistics of data asset-related vocabulary from annual reports. Due to lack of quantitative methods for data assets, firms generally disclose data assets information within annual reports [30]. For example, Midea Group (Stock code: SZ.000333) disclosed in its report, to better integrate online and offline data resources, that the firm established a user-oriented operating platform based on users’ data assets, bringing CNY 5.5 billion revenue. Therefore, this study attempts to quantify the level of data assets using text analysis approaches. However, using keyword frequency to measure data asset deployment still faces potential limitations. For example, whether the use of digital terminology genuinely reflects substantive data asset development is uncertain. Therefore, to ensure the vocabulary can effectively capture data-asset-related information, we reviewed the relevant vocabulary descriptions in the policies related to data assets and referred to previous studies [28,29,30,31]. The term “data asset” is designed as the seed word. Sets of similar words are derived based on seed words by employing the word2vec neural network model and deep learning techniques. Specifically, word2vec simplifies the processing of text content into vector operations of vector space, and the similarity of vector space is used to measure the semantic similarity of text [91,92]. By using the local context of words, word2vec not only captures richer semantic information but also maintains computational efficiency. Previous studies on data assets measurements have predominantly employed word2vec models, and referring to previous studies [26,52] we adopt the word2vec approach in our research, as the cosine similarity was adopted as the similarity metric for word2vec [93]. On the one hand, prior studies demonstrate that a threshold of 0.5 serves as a natural cutoff point across various text similarity distributions in textual analysis [94]. On the other hand, aligning with established empirical practices related to data assets, a 0.5 threshold is widely employed, such as Chen et al. (2025) and Zhu et al. (2025) [95,96]. As a result, this study retains those terms with a similarity score above 0.5. Specific similar words are reported in Table 1. Eventually, this paper uses the logarithm value of the frequency of words related to data assets in the annual reports of firms (CusDA) as a proxy variable for firm data assets.

Table 1. Keywords of Data Assets.

Keywords

Data asset; Data resource; Data source; Big Data; Mass data; Data analysis system; Data sharing; Data platform; Data mining; Information resource; Knowledge base

Following Wang et al. (2025) [97], green innovation is measured by taking the natural logarithm of the sum of the number of green invention patents and utility model patents applied for by the enterprise plus one (SupGreenInv).

3.3. Research Design

To examine the impact of customers’ data assets on suppliers’ green innovation, we construct the following regression model:

$${SupGreenInv}_{i,t}={\beta }_0+{\beta }_1{CusDA}_{j,t}+Controls+f_{Year}+f_{Industry}+{\epsilon }_{i,j,t}$$

where ${SupGreenInv}_{i,t}$ is the green innovation performance of suppliers measured by the natural logarithm of the number of green patent applications (including green invention patents and green utility model patents) plus one. The higher the value, the greater the level of a firm’s green innovation. ${CusDA}_{j,t}$ represents the customer data asset variables measured by the logarithm value of the frequency of words related to data assets in the annual reports. Controls is the battery of variables related to supplier green innovation and customer companies. Specifically, following Wang et al. (2025) [98] and Sun et al. (2023) [88], we first include the characteristics of the supplier including firm size (Size), leverage ratio (Lev), return on assets (ROA), whether the supplier is in a loss position (Loss), suppliers’ board size (Board), the ownership of largest shareholder (Top1), supplier’s age of establishment (FirmAge), and institutional investor ownership (INST). We further control customer characteristics, including customer firm size (CusSize), customer leverage ratio (CusLev), customer return on assets (CusROA), whether the customer is in a loss position (CusLoss), customer board size (CusBoard), ownership of largest shareholder of customer (CusTop1), customer’s age of establishment (CusFirmAge), and institutional investor ownership of customer firm (CusINST). In addition, the sales ratio (Salesratio) and customer sales ranking (Rank) are included. The detailed definitions of the main variables are presented in Table 2, and $f_{Year}$ and $f_{Industry }$ represent the year and industry fixed effects. Some industries use more data asset information; we control for both customer and supplier industries. We also adjust regression results for robust standard errors. Table 2 shows the detailed variables’ definitions.

Table 2. Definitions of Main Variables.

Variable	Variable Definitions
SupGreenInv	The natural logarithm of one plus the number of green patent applications (including green invention patents and green utility model patents) for supplier firm
CusDA	The natural logarithm of frequency of data-assets-related vocabulary in annual report of customer firm
Size	Natural logarithm of total assets of supplier firm
Lev	The book value of total debts divided by the book value of total assets of supplier firm
ROA	The ratio of net profits to total assets of supplier firm
Loss	Equals to 1 if supplier firm is profitable, and 0 otherwise
Board	Natural logarithm of the number of directors on the board of supplier firm
Top1	The shareholding ratio of the largest shareholder of supplier firm
Firmage	Natural logarithm of firm age of supplier firm
INST	The percentage of shares held by institutional investors of supplier firm
CusSize	Natural logarithm of total assets of customer firm
CusLev	The book value of total debts divided by the book value of total assets for customer firm
CusROA	The ratio of net profits to total assets for customer firm
CusLoss	Equals to 1 if customer firm is profitable, and 0 otherwise
CusBoard	Natural logarithm of the number of directors on the board for customer firm
CusTop1	The shareholding ratio of the largest shareholder of customer firm.
CusFirmage	Natural logarithm of firm age for customer firm
CusINST	The percentage of shares held by institutional investors for customers
Rank	Customer sales ranking
Salesratio	The purchasing share of customers
SupGreenInv2	The natural logarithm of one plus the number of green invention patents for supplier firm
DA2	The frequency of disclosure of data assets information in the annual reports measured by Formula (2)
DA3	Equals to ln (Market value − Fixed assets − Financial assets − Intangible assets)
DA4	The natural logarithm of frequency of customer data assets measured by alternative dictionaries
IV	The third power of the difference between the means of the customer data asset indicator by industry code and province
Digital	Digital transformation index which is acquired from CSMAR.
TFP	The total factor productivity of suppliers measured by OP method.
Distance	The spatial distance is calculated based on the registered addresses of supplier and customer firm
East	Equals to 1 if customer firm is located in the eastern region, and 0 otherwise
TechIntensive	Equals to 1 if customer firm belongs to the technology intensive industry, and 0 otherwise
SOE	Equals to 1 if supplier is state-owned and 0 otherwise
FC	Financial constraints measured by FC index
Application	The number of the supplier’s green innovation patent applied
Grant	The number of the supplier’s green innovation patent grants
R&D	Natural logarithm of firm R&D expenses for supplier firm

3.4. Descriptive Statistics

Table 3 presents the descriptive statistics for the main variables. The mean value of supplier green innovation level (SupGreenInv) is 0.272, the minimum value is 0, the maximum value is 3.045, and the standard deviation is 0.627. This indicates significant heterogeneity in green innovation capacity across suppliers, as over half of the sample (p50 = 0) has not applied for green patents. The mean of the customer data asset level (CusDA) is 1.035, the minimum value is 0, the maximum value is 3.761, and the standard deviation is 1.035. This suggests that data assets utilization level varies widely among customers, and the overall level still remains low, as reflected by the median of 0.693. These variations in the variables provide a good data basis for the study of the data assets and green innovation provides a good data basis for study.

Table 3. Descriptive Statistics.

Variables	N	Mean	Sd	Min	p25	p50	p75	Max
SupGreenInv	1129	0.272	0.627	0	0	0	0	3.045
CusDA	1129	1.006	1.035	0	0	0.693	1.792	3.761
Size	1129	21.95	1.367	19.56	20.81	21.81	22.86	25.19
Lev	1129	0.408	0.210	0.054	0.233	0.400	0.549	0.947
ROA	1129	0.032	0.069	−0.262	0.01	0.037	0.068	0.169
Loss	1129	0.170	0.376	0	0	0	0	1
Board	1129	2.116	0.176	1.609	1.946	2.197	2.197	2.565
Top1	1129	0.346	0.145	0.104	0.240	0.319	0.442	0.719
FirmAge	1129	2.997	0.300	2.079	2.833	3.045	3.178	3.584
INST	1129	0.406	0.253	0.00616	0.176	0.413	0.623	0.903
CusSize	1129	24.48	1.915	20.84	23.01	24.29	25.81	28.61
CusLev	1129	0.561	0.170	0.122	0.444	0.588	0.689	0.887
CusROA	1129	0.042	0.048	−0.092	0.015	0.035	0.0667	0.201
CusLoss	1129	0.094	0.292	0	0	0	0	1
CusBoard	1129	2.187	0.209	1.609	2.079	2.197	2.303	2.708
CusTop1	1129	0.391	0.178	0.0903	0.247	0.372	0.514	0.826
CusFirmAge	1129	3.071	0.283	2.197	2.890	3.091	3.258	3.689
CusINST	1129	0.601	0.242	0.031	0.455	0.627	0.784	0.982
Rank	1129	3.013	1.414	1	2	3	4	5
Salesratio	1129	8.641	11.43	0.440	2.690	4.800	9.040	68.66

4. Empirical Results

4.1. Baseline Regression Results

This paper uses the regression model (1) to investigate the impact of customer data assets on supplier green innovation. The regression results are shown in Table 4. Column (1) reports regression results without adding control variables; the coefficient on CusDA, 0.039, is positive and significant at the 10% level (t-statistic = 1.7). Column (2) is the results after adding control variables, demonstrating that the coefficient on CusDA, 0.06, is positive and significant at the 1% level (t-statistic = 2.74). In terms of economic significance, the findings reported in column (2) indicate that a one-notch improvement in the data assets of a customer firm leads to a 0.06 increase in supplier green innovation output. The results indicate that CusDA is positively and significantly associated with supplier green innovation across model specifications, regardless of whether control variables are included, thus confirming Hypothesis 1. And the finding further suggests that the larger the scale of customer data assets, the more significant the promotion effect on supplier green innovation.

Table 4. Baseline regression.

	(1)	(2)
Variables	SupGreenInv	SupGreenInv
CusDA	0.039 *	0.060 ***
	(1.70)	(2.74)
Size		0.122 ***
		(5.79)
ROA		0.346
		(1.14)
Lev		0.050
		(0.46)
Loss		−0.089
		(−1.64)
Board		0.264 **
		(2.57)
Top1		0.910 ***
		(5.47)
FirmAge		−0.099
		(−1.50)
INST		0.029
		(0.31)
CusSize		−0.033 *
		(−1.83)
CusROA		−0.336
		(−0.74)
CusLev		0.102
		(0.76)
CusLoss		−0.081
		(−1.32)
CusBoard		−0.111
		(−1.13)
CusTop1		−0.010
		(−0.08)
CusFirmAge		−0.022
		(−0.31)
CusINST		0.199 *
		(1.86)
Salesratio		0.004 *
		(1.68)
Rank		−0.002
		(−0.16)
Year FE	Yes	Yes
Industry FE	Yes	Yes
Constant	−0.052	−2.209 ***
	(−0.56)	(−3.30)
Observations	1129	1129
R-squared	0.143	0.274

Notes: This table reports the results of main regression. T-statistics are reported in parentheses. Detailed variable definitions are summarized in Table 2. ***, **, and * indicate significance levels of 1%, 5%, and 10%, respectively.

In addition, control variables also show statistical significance in Table 4 Column 2. For example, supplier size (Size) has a significantly positive impact on green innovation at the 1% level. This aligns with the resource-based view that the larger suppliers possess more abundant financial resources, R&D personnel, and technological infrastructure to invest in green innovation [51]. Supplier Board Size (Board) exhibits a strongly positive effect on green innovation at the 5% level (Coefficient = 0.264, t = 2.57), suggesting that larger boards bring diverse expertise, broader networks, and enhanced monitoring capabilities, ultimately facilitating suppliers’ green innovation. In addition, the coefficient of supplier ownership concentration (Top1) is 0.910 (t = 5.47), significant at the 1% level, reflecting the higher ownership concentration of the largest shareholder facilitates firms to focus on long-term strategies, such as green innovation.

For customers, customer size (CusSize) has a marginally negative impact on supplier green innovation at the 10% level. And the coefficient of CusINST is 0.199 (t = 1.86), significant at the 10% level, indicating that customer-side institutional investors pay more attention to sustainable development goals and promote suppliers’ green innovation through supply chain collaboration. Further, the Salesratio is also positive at the 10% level. A higher sales ratio indicates stronger dependence of the supplier on the customer, incentivizing the supplier to align its strategies with the customer’s needs. When customers possess valuable data assets, suppliers with greater revenue dependence are more motivated to invest in green innovation to maintain the relationship. And this is consistent with our Hypothesis 2.

4.2. Testing the Role of Geographic Distance

In this study, we use the spatial distance between the two firms registered addresses to measure the geographical distance. The sample was divided into close (ShortDistance) and far (LongDistance) groups based on the median distance. The results are reported in Column (1) and (2) of Table 5, demonstrating that there is a stronger positive effect when customers and suppliers have closer distance. It suggests that closer geographic proximity facilitates better communication and more effective implementation of sustainable initiatives, enhancing their impact on suppliers’ green innovation.

Table 5. Role of geographic distance and customer concentration.

	(1)	(2)	(3)	(4)
	ShortDistance	LongDistance	High Sales Ratio	Low Sales Ratio
Variables	SupGreenInv	SupGreenInv	SupGreenInv	SupGreenInv
CusDA	0.098 ***	0.036	0.091 ***	0.042
	(2.92)	(1.18)	(2.82)	(1.37)
Size	0.122 ***	0.145 ***	0.176 ***	0.097 ***
	(3.90)	(4.34)	(5.51)	(3.15)
ROA	−0.510	1.003 **	0.303	0.783
	(−1.10)	(2.20)	(0.75)	(1.51)
Lev	0.183	0.023	−0.004	0.121
	(1.09)	(0.13)	(−0.02)	(0.75)
Loss	−0.205 ***	0.022	−0.173 **	0.037
	(−2.61)	(0.25)	(−2.27)	(0.44)
Board	0.647 ***	0.100	0.034	0.432 ***
	(4.07)	(0.73)	(0.21)	(3.17)
Top1	1.261 ***	0.786 ***	0.866 ***	0.948 ***
	(5.17)	(3.33)	(3.67)	(4.10)
FirmAge	−0.154	−0.102	0.005	−0.120
	(−1.39)	(−1.19)	(0.05)	(−1.24)
INST	−0.185	0.069	−0.162	0.268 **
	(−1.39)	(0.53)	(−1.12)	(2.23)
CusSize	−0.079 ***	0.002	−0.041	−0.030
	(−3.25)	(0.06)	(−1.45)	(−1.17)
CusROA	−0.878	0.118	−0.212	−0.708
	(−1.45)	(0.16)	(−0.28)	(−1.31)
CusLev	0.182	0.013	−0.065	0.293
	(0.96)	(0.06)	(−0.31)	(1.61)
CusLoss	−0.114	−0.085	−0.034	−0.142
	(−1.31)	(−0.90)	(−0.37)	(−1.47)
CusBoard	0.054	−0.228	−0.164	0.027
	(0.43)	(−1.43)	(−1.17)	(0.19)
CusTop1	−0.303	0.144	−0.064	0.050
	(−1.51)	(0.69)	(−0.32)	(0.25)
CusFirmAge	−0.167	0.146	−0.070	−0.095
	(−1.55)	(1.40)	(−0.59)	(−0.97)
CusINST	0.437 ***	0.040	0.338 **	0.103
	(2.73)	(0.27)	(1.99)	(0.70)
Salesratio	−0.000	0.008 **	0.004	0.013
	(−0.13)	(2.25)	(1.48)	(0.50)
Rank	−0.011	−0.001	0.003	−0.011
	(−0.57)	(−0.04)	(0.13)	(−0.50)
YearFE	Yes	Yes	Yes	Yes
IndustryFE	Yes	Yes	Yes	Yes
Constant	−3.570 ***	−1.697	−2.396 **	−2.227 **
	(−3.62)	(−1.61)	(−2.56)	(−2.04)
Observations	565	564	556	573
R-squared	0.296	0.391	0.350	0.319

Notes: This table reports the results of testing the role of geographic distance and customer concentration. t-statistics are reported in parentheses. Detailed variable definitions are summarized in Table 2. *** and ** indicate significance levels of 1% and 5% respectively.

4.3. Testing the Role of Customer Concentration

In this study, we measure customer concentration (Concentration) by the sum of the square of the top five customer sales to total sales. And we portioned the sample into two groups by the median (the low-concentration group with the value of customer concentration lower than sample median and the high-concentration group with the value of customer concentration greater than the sample median). We find that the relation is significant in the high customer concentration group. Table 5 Columns (3) and (4) shows the results.

4.4. Robustness Tests

The robustness tests are conducted in this section, including the replacement of dependent variables, replacement of independent variables, and exclusion of specific observations.

Firstly, this paper uses the natural logarithm of one plus the number of green invention patents for supplier firms (SupGreenInv2) as an alternative measurement for dependent variables. The results are shown in column (1) of Table 5, which shows that the baseline results of this paper are robust (β = 0.048, t = 2.77). Next, different measurements of data assets are used. To capture the heterogeneity in the intensity of data assets to firms, DA2 is alternatively introduced, which is the frequency of seed words and similar words in each firm’s annual reports. The specific estimation model is shown as follows:

$$DA{2}_{\mathrm{i},\mathrm{t}} ={\displaystyle \frac{{\sum }_j^n{Fre}_{i,t,j}}{{TotalFre}_{i,t}}}\times 100$$

(1)

where $D$A2_i,t is the data asset deployment level variable, ${\sum }_j^n{Fre}_{i,t,j}$ is the data-asset-related keyword frequency of the j-th term in the set of similar words in the annual report of company i in year t, and ${TotalFre}_{i,t}$ is the total word frequency in the annual report. The results are shown in column (2) of Table 6, which shows that the baseline results of this paper are robust (β = 3.787, t = 1.81). In addition, DA3 is measured as ln [Market Value − (Fixed Assets + Financial Assets + Intangible Assets)]. Specifically, market value equals the book value of total liabilities plus the market value of equity. Financial assets are calculated as the sum of trading financial assets, derivative financial assets, net loans and advances, net available-for-sale financial assets, net held-to-maturity investments, and net investment properties. And the intangible assets follow the narrow accounting definitions. The results are shown in column (3) of Table 6, which shows that the baseline results of this paper are robust (β = 0.081, t = 2.43). Lastly, following He et al. (2024) [98] this paper uses alternative dictionaries to measure data assets (DA4). Specifically, the term “digital”, “data”, “information”, and “network” are used as seed words. And sets of similar words are derived based on seed words by employing the word2vec neural network model and deep learning techniques. We compile data-asset-related keywords and use the natural logarithm to calculate their term frequency in the annual reports of sample firms. The specific keywords are shown in Appendix A. And the regression results are shown in column (4) of Table 6, suggesting that the baseline results of this paper are robust (β = 0.066, t = 1.87). In all regressions in columns (1)–(4) in Table 6, customers’ data assets are significantly related to suppliers’ green innovation.

Table 6. Robustness test: Alternative measurements.

	(1)	(2)	(3)	(4)
Variables	SupGreenInv2	SupGreenInv	SupGreenInv	SupGreenInv
CusDA	0.048 ***
	(2.77)
DA2		3.787 *
		(1.81)
DA3			0.081 **
			(2.43)
DA4				0.066 *
				(1.87)
Size	0.099 ***	0.120 ***	0.105 ***	0.121 ***
	(5.68)	(5.71)	(4.71)	(5.71)
ROA	0.322	0.354	0.455	0.308
	(1.35)	(1.16)	(1.46)	(1.01)
Lev	−0.028	0.054	0.151	0.051
	(−0.32)	(0.49)	(1.32)	(0.47)
Loss	−0.038	−0.092 *	−0.087	−0.093 *
	(−0.89)	(−1.69)	(−1.50)	(−1.71)
Board	0.190 **	0.265 ***	0.373 ***	0.266 ***
	(2.37)	(2.59)	(3.60)	(2.59)
Top1	0.743 ***	0.922 ***	0.998 ***	0.923 ***
	(5.43)	(5.51)	(5.60)	(5.50)
FirmAge	−0.095 *	−0.099	−0.102	−0.104
	(−1.74)	(−1.51)	(−1.49)	(−1.57)
INST	0.090	0.027	0.010	0.017
	(1.24)	(0.29)	(0.10)	(0.19)
CusSize	−0.022	−0.029	−0.095 ***	−0.031 *
	(−1.55)	(−1.59)	(−2.64)	(−1.69)
CusROA	−0.291	−0.359	−1.193 **	−0.290
	(−0.87)	(−0.79)	(−2.39)	(−0.64)
CusLev	0.022	0.098	0.071	0.112
	(0.21)	(0.72)	(0.51)	(0.83)
CusLoss	−0.061	−0.085	−0.128 **	−0.085
	(−1.22)	(−1.37)	(−1.96)	(−1.37)
CusBoard	−0.109	−0.103	−0.033	−0.093
	(−1.34)	(−1.05)	(−0.33)	(−0.95)
CusTop1	0.001	−0.025	0.074	−0.035
	(0.01)	(−0.19)	(0.49)	(−0.26)
CusFirmAge	0.013	−0.030	−0.034	−0.021
	(0.23)	(−0.42)	(−0.43)	(−0.29)
CusINST	0.135 *	0.206 *	0.160	0.209 *
	(1.65)	(1.91)	(1.34)	(1.93)
Salesratio	0.002	0.004	0.000	0.003
	(1.15)	(1.63)	(0.15)	(1.52)
Rank	0.000	−0.001	−0.012	−0.002
	(0.02)	(−0.10)	(−0.79)	(−0.12)
YearFE	Yes	Yes	Yes	Yes
IndustryFE	Yes	Yes	Yes	Yes
Constant	−1.841 ***	−2.225 ***	−2.589 ***	−2.474 ***
	(−3.31)	(−3.31)	(−3.48)	(−3.57)
Observations	1129	1129	1047	1128
R-squared	0.276	0.271	0.277	0.272

Notes: This table reports the results of the robustness test with the alternative measure of suppliers’ green innovation and the alternative measures of customers’ data assets. t-statistics are reported in parentheses. Detailed variable definitions are summarized in Table 2. ***, **, and * indicate significance levels of 1%, 5%, and 10%, respectively.

Furthermore, the effect of customer data assets on supplier green innovation is re-estimated by using alternative samples. First, customers in the high technology industry are excluded. For firms in high technology industries, data assets are their main products, profit sources and strategic priorities. Additionally, suppliers paired with such high-tech customers are more likely to operate in high-value-added sectors, where green innovation is easier to implement. To reduce such industry-specific self-selection bias, we exclude the sample of high technology customers. The regression results in Table 7 column (1) show that customer data assets still have a significantly positive effect on supplier green innovation (β = 0.044, t = 1.76), further confirming the robustness of our findings. Secondly, we excluded samples in the year 2020 to eliminate the macro-economic shock of the pandemic, as the COVID-19 outbreak may have had a negative impact on firms’ production, business conditions, as well as supply chain operations. After the shock, customers may have focused on maintaining normal operations and the level of data assets of the enterprise may have been affected. Meanwhile, suppliers may have temporarily reduced green R&D and shifted resources away from long-term green innovation projects toward short-term survival. Therefore, to eliminate the potential impacts of the COVID-19 pandemic, we excluded the 2020 samples to test robustness. Column (2) of Table 7 shows that our results remain qualitatively similar (β = 0.06, t = 2.67).

Table 7. Robustness test: Alternative samples.

	(1)	(2)
Variables	SupGreenInv	SupGreenInv
CusDA	0.044 *	0.060 ***
	(1.76)	(2.67)
Size	0.130 ***	0.128 ***
	(5.46)	(5.76)
ROA	0.575 *	0.401
	(1.69)	(1.23)
Lev	−0.030	0.091
	(−0.24)	(0.79)
Loss	−0.055	−0.083
	(−0.94)	(−1.44)
Board	0.216 *	0.195 *
	(1.96)	(1.81)
Top1	1.030 ***	0.885 ***
	(5.35)	(5.17)
FirmAge	−0.090	−0.090
	(−1.17)	(−1.32)
INST	−0.097	0.053
	(−0.94)	(0.55)
CusSize	−0.039 *	−0.031
	(−1.88)	(−1.63)
CusROA	−0.261	−0.458
	(−0.53)	(−0.94)
CusLev	−0.003	0.074
	(−0.02)	(0.53)
CusLoss	−0.069	−0.131 **
	(−1.04)	(−2.10)
CusBoard	−0.116	−0.110
	(−1.14)	(−1.08)
CusTop1	0.089	0.045
	(0.59)	(0.32)
CusFirmAge	−0.058	−0.024
	(−0.72)	(−0.34)
CusINST	0.221 *	0.167
	(1.71)	(1.50)
Salesratio	0.004 *	0.004 *
	(1.70)	(1.83)
Rank	0.007	−0.001
	(0.44)	(−0.05)
YearFE	Yes	Yes
IndustryFE	Yes	Yes
Constant	−1.994 ***	−2.313 ***
	(−2.62)	(−3.25)
Observations	921	1058
R-squared	0.304	0.287

Notes: This table reports the results of the robustness test with alternative samples. t-statistics are reported in parentheses. Detailed variable definitions are summarized in Table 2. ***, **, and * indicate significance levels of 1%, 5%, and 10%, respectively.

4.5. Endogeneity Tests

To mitigate the endogeneity concerns, we use propensity score matching (PSM) and the Heckman two-stage analysis. Firstly, this study adopts the PSM method for testing to address the endogeneity problem due to systematic differences in supply chain samples. Specifically, one-to-one nearest-neighbor matching is used, requiring a caliper of 0.01. We introduced a dummy variable (DADummy) in this test. DADummy equals to 1 if customers’ data assets are above the median value of the industry and 0 otherwise. The matching variables include all control variables in the main regression. Column (1) of Table 8 displays the regression results with matched samples. The regression results in a significant positive effect of customer data assets on supplier green innovation (β = 0.056, t = 1.72). This finding suggests that the positive effect of customer data assets on supplier green innovation is not driven by systematic differences between samples. Panel B of Table 8 presents the results of the balance tests after PSM. After PSM, there is no significant difference between the control and treatment groups (p-values > 0.1), suggesting that the PSM is effective.

Table 8. Endogeneity tests. Panel (A): PSM and Heckman Two-Stage Method Results. Panel (B): PSM Balance Test.

Panel (A)
			(1)		(2)	(2)
Variables			PSM		Heckman Two-Stage Model
			SupGreenInv		DataAssets	SupGreenInv
CusDA			0.056 * (1.72)			0.059 ***
						(2.71)
DADummy
IV					0.116 ***
					(3.43)
IMR						0.053
						(0.21)
Size			0.149 ***		−0.184 ***	0.117 ***
			(4.43)		(−4.45)	(3.69)
ROA			−0.307		−1.053	0.315
			(−0.67)		(−1.26)	(0.89)
Lev			0.141		−0.001	0.049
			(0.86)		(−0.00)	(0.45)
Loss			−0.255 ***		−0.214	−0.096
			(−3.19)		(−1.50)	(−1.48)
Board			0.142		−0.212	0.258 **
			(0.97)		(−0.85)	(2.44)
Top1			0.956 ***		−0.061	0.908 ***
			(3.64)		(−0.19)	(5.44)
FirmAge			−0.117		0.363 ***	−0.088
			(−1.05)		(2.58)	(−1.03)
INST			−0.109		0.032	0.028
			(−0.79)		(0.15)	(0.30)
CusSize			−0.055 *		0.072 **	−0.031
			(−1.85)		(2.33)	(−1.57)
CusROA			−1.384 **		−0.979	−0.361
			(−2.02)		(−0.89)	(−0.75)
CusLev			0.225		0.252	0.110
			(0.94)		(0.82)	(0.80)
CusLoss			−0.197 *		−0.184	−0.087
			(−1.94)		(−1.15)	(−1.26)
CusBoard			−0.021		−0.525 ***	−0.125
			(−0.14)		(−2.67)	(−1.16)
CusTop1			0.069		−1.525 ***	−0.056
			(0.32)		(−4.82)	(−0.24)
CusFirmAge			−0.069		0.418 ***	−0.010
			(−0.63)		(2.87)	(−0.11)
CusINST			0.228		0.691 ***	0.220
			(1.40)		(2.75)	(1.49)
Salesratio			−0.001		0.004	0.004 *
			(−0.27)		(0.86)	(1.70)
Rank			−0.002		0.090 ***	0.0003
			(−0.08)		(2.77)	(0.02)
YearFE			Yes		Yes	Yes
IndustryFE			Yes		Yes	Yes
Constant			−2.459 ***		1.686	−2.204 ***
			(−2.91)		(1.40)	(−3.29)
Observations			519		1129	1129
R-squared			0.311			0.274
Panel (B)
	Mean				Difference in Treated and Control
Variables	Treated	Control	%Bias	t-Value	p-Value
Size	21.809	21.848	−2.9	−0.41	0.681
ROA	0.0332	0.0348	−2.2	−0.32	0.747
Lev	0.404	0.403	0.2	0.03	0.975
Loss	0.153	0.134	5.1	0.73	0.465
Board	2.108	2.116	−4.1	−0.54	0.589
Top1	0.346	0.345	0.7	0.1	0.924
FirmAge	2.999	3.001	−0.6	−0.08	0.932
INST	0.399	0.419	−7.7	−1.07	0.283
CusSize	24.41	24.362	2.6	0.38	0.707
CusROA	0.041	0.039	3.4	0.45	0.655
CusLev	0.573	0.572	0.3	0.04	0.972
CusLoss	0.099	0.121	−7.3	−0.94	0.350
CusBoard	2.196	2.2	−3.2	−0.44	0.662
CusTop1	0.364	0.37	−3.3	−0.48	0.634
CusINST	0.588	0.595	−3.0	−0.43	0.669
CusFirmAge	3.06	3.08	−7	−0.95	0.343
Salesratio	7.25	7.17	0.7	0.12	0.903
Rank	3.194	3.27	−5.8	−0.79	0.43

Notes: This table provides the results of the endogeneity test, the PSM method and the Heckman two-stage analysis. And panel B reports the balancing tests results after PSM process. Detailed variable definitions are summarized in Table 2. t-statistics are reported in parentheses. ***, **, and * indicate significance levels of 1%, 5%, and 10%, respectively.

Secondly, we use the Heckman two-stage analysis to address the potential endogeneity problems. This study uses the third power of the difference between the means of the customer data asset indicator by industry code and province as the instrumental variable (IV). The level of utilization of data assets could be similar within an industry and CusDA is correlated with data assets in other firms within the industry. Also, Column (2) shows the first-stage regression result that the IV coefficient is 0.116 and significant at the 1% level (t = 3.43), satisfying the relevance assumption. However, while the industry-province average data assets do not directly affect individual suppliers’ green innovation, the IV satisfies the exclusion assumption, confirming the causal relationship. As a result, the IV satisfies both the relevance and exclusion assumption of a valid instrumental variable. Column (2) of Table 7 shows the results. The coefficient of CusDA is still highly significant and positive (β = 0.059, t = 2.71), which is consistent with the main results, indicating that our results are not subjected to endogeneity.

4.6. Mechanism Tests

We found a robust relationship between customer data assets and supplier green innovation. This section examines two potential paths by which customer data assets may affect supplier green innovation. We use the causal steps approach to examine the proposed channels of operational efficiency and digital transformation.

Specifically, the models are shown below:

$${TFP}_{j,t}/{Digital}_{j,t}={\beta }_0+{\beta }_1{CusDA}_{j,t}+Controls+f_{Year}+f_{Industry}+{\epsilon }_{i,j,t}$$

(2)

$${SupGreenInv}_{i,t}={\beta }_0+{\beta }_1{TFP}_{j,t}/{Digital}_{j,t}+Controls+f_{Year}+f_{Industry}+{\epsilon }_{i,j,t}$$

(3)

First, we expect that customer data assets can increase suppliers’ operational efficiency, thereby improving supplier corporate capacity of green innovation implementation. The total factor productivity (TFP) was chosen as the proxy for operation efficiency. Empirical results are reported in Table 9. Column (1) of Table 9 shows that customer data assets have a significant and positive effect on TFP (coefficient = 0.031, t = 1.84). In column 2, the coefficient of TFP is 0.062 (t = 1.69), significant at the 10% level and indicating that higher operational efficiency provides suppliers with more abundant resources (e.g., cost savings, technical capacity) to support green innovation. Overall, the first mechanism is valid, that customer data assets indirectly boost suppliers’ green innovation investment by improving their operational efficiency.

Table 9. Mechanism analysis.

	(1)	(2)	(3)	(4)
Variables	TFP	SupGreenInv	Digital	SupGreenInv
CusDA	0.035 *		1.030 ***
	(1.88)		(2.93)
TFP		0.062 *
		(1.69)
Digital				0.011 ***
				(3.79)
Size	0.598 ***	0.055 **	1.303 ***	0.099 ***
	(35.12)	(2.04)	(4.33)	(4.15)
ROA	2.343 ***	0.099	1.085	0.287
	(7.70)	(0.32)	(0.21)	(0.80)
Lev	0.843 ***	0.024	−1.880	0.027
	(8.65)	(0.23)	(−1.15)	(0.23)
Loss	−0.016	−0.097 *	0.434	−0.116 **
	(−0.28)	(−1.75)	(0.50)	(−2.04)
Board	0.017	0.248 **	4.973 ***	0.279 **
	(0.18)	(2.41)	(2.86)	(2.40)
Top1	0.008	0.773 ***	−3.696	1.070 ***
	(0.06)	(4.77)	(−1.62)	(5.02)
FirmAge	−0.046	−0.232 ***	2.171 **	−0.076
	(−0.87)	(−3.66)	(2.18)	(−0.97)
INST	−0.236 ***	0.045	−4.075 ***	0.034
	(−2.99)	(0.48)	(−2.75)	(0.31)
CusSize	0.021	−0.036 **	−0.370	−0.032
	(1.48)	(−1.99)	(−1.41)	(−1.46)
CusROA	−0.438	−0.193	8.267	−0.302
	(−0.89)	(−0.42)	(1.14)	(−0.60)
CusLev	−0.179	0.173	1.139	0.242
	(−1.34)	(1.24)	(0.51)	(1.58)
CusLoss	−0.104	−0.096	−0.451	−0.089
	(−1.62)	(−1.57)	(−0.48)	(−1.34)
CusBoard	0.022	−0.036	0.879	−0.137
	(0.29)	(−0.34)	(0.60)	(−1.27)
CusTop1	0.035	−0.011	−2.354	0.005
	(0.28)	(−0.08)	(−1.07)	(0.03)
CusFirmAge	0.017	−0.047	−1.888	−0.032
	(0.24)	(−0.64)	(−1.61)	(−0.43)
CusINST	−0.080	0.115	0.585	0.127
	(−0.78)	(1.06)	(0.35)	(1.03)
Salesratio	−0.001	0.004 *	−0.009	0.004 *
	(−0.62)	(1.68)	(−0.27)	(1.80)
Rank	−0.019	0.005	−0.061	0.001
	(−1.45)	(0.35)	(−0.28)	(0.05)
YearFE	Yes	Yes	Yes	Yes
IndustryFE	Yes	Yes	Yes	Yes
Constant	−5.660 ***	−0.762	−3.850	−2.198 ***
	(−8.82)	(−1.10)	(−0.38)	(−2.82)
Observations	1095	1095	935	935
R-squared	0.771	0.200	0.418	0.299

Notes: This table provides the results of mechanism analysis. t-statistics are reported in parentheses. Detailed variable definitions are summarized in Table 2. ***, **, and * indicate significance levels of 1%, 5%, and 10%, respectively.

Second, we expect that customers’ data assets facilitate supplier digital transformation captured by the digital transformation index. The regression results listed in Table 9 Column (3) suggest that CusDA has a strong positive and highly significant effect on suppliers’ digitalization level (coefficient = 1.030, t-statistic = 2.93). The results evidence that when customers have superior digital capabilities, suppliers adjust their business strategies and accelerate digital transformation to maintain relationships. In Column 4, the coefficient of Digital is 0.011 (t = 3.79) and significant at the 1% level. It is suggested that digital transformation could provide robust foundation for suppliers’ green innovation. Overall, the second mechanism is valid.

4.7. Cross-Sectional Analysis

4.7.1. Customers’ Characteristics

This section examines the regional and industrial heterogeneity. Since there are significant regional differences in China regarding economic growth, infrastructure, and innovation resource endowments, in this study we portioned the samples into an East group and a Non-East group to examine the locational heterogeneity. The results are listed in Panel B Columns (1) and (2) of Table 10. In the East group (Column 1), the CusDA coefficient is 0.061 and significant at the 5% level (t = 2.14), while the Non-East group (Column 2) is insignificant (β = 0.063, t = 1.52). The results show that the spillover effect of customer data assets on supplier green innovation is more pronounced for customers in the eastern region. This may be attributed to the fact that the eastern region has significant advantages in infrastructure and human resources, which facilitates the effective transmission and sharing of data and information, thereby promoting the spillover effect of firm data assets [71].

Table 10. Cross-sectional analysis. Panel (A): Customer characteristics. Panel (B): Supplier characteristics.

Panel (A)
	(1)	(2)	(3)	(4)
	East	Non-East	Tech Intensive	Non-Tech Intensive
Variables	SupGreenInv	SupGreenInv	SupGreenInv	SupGreenInv
CusDA	0.061 **	0.063	0.083 **	−0.003
	(2.14)	(1.52)	(2.38)	(−0.10)
Controls	Included	Included	Included	Included
YearFE	Yes	Yes	Yes	Yes
IndustryFE	Yes	Yes	Yes	Yes
Constant	−1.313	−4.555 ***	−2.946 ***	−2.992 ***
	(−1.57)	(−3.00)	(−3.05)	(−3.24)
Observations	785	344	468	661
R-squared	0.267	0.462	0.271	0.400
Panel (B)
	(1)	(2)	(3)	(4)
	SOE	Non-SOE	Higher FC	Lower FC
Variables	SupGreenInv	SupGreenInv	SupGreenInv	SupGreenInv
CusDA	0.115 ***	0.028	0.023	0.102 ***
	(2.73)	(1.12)	(0.93)	(2.69)
Controls	Included	Included	Included	Included
YearFE	Yes	Yes	Yes	Yes
IndustryFE	Yes	Yes	Yes	Yes
Constant	−3.233 *	−1.208	−2.681 ***	−0.134
	(−1.95)	(−1.60)	(−2.89)	(−0.09)
Observations	404	725	700	429
R-squared	0.495	0.236	0.248	0.512

Notes: This table presents the results of cross-sectional analyses. Panel (A) reports the results in regard to customer characteristics. Panel (B) reports the results in regard to supplier characteristics. t-statistics are reported in parentheses. Detailed variable definitions are summarized in Table 2. ***, **, and * indicate significance levels of 1%, 5%, and 10%, respectively.

Moreover, we examine the role of technological intensity by classifying the samples into Non-tech Intensive (Tech-intensive = 0) and Tech Intensive (Tech-intensive = 1). The results are shown in Table 10 Panel B Columns (3) and (4). For technology-intensive customers (Column 3), the CusDA coefficient is 0.083 and significant at the 5% level (t = 2.38) but insignificant for non-technology customers (Column 4, β = −0.003, t = −0.10), suggesting that technology-intensive firms pay more attention to scientific and technological innovation and have a richer accumulation of knowledge, enabling them to utilize innovative resources more effectively and fully leverage the advantages of data assets and digital technologies. As a result, the spillover effect of firm data assets on supplier green innovation is promoted when customers are technology-intensive firms.

4.7.2. Suppliers’ Characteristics

Due to China’s special enterprise system background, state-owned enterprises and non-state-owned enterprises have significant differences in resource endowment, social responsibilities, and policy support. These differences may influence their perspectives on data assets, and capabilities and motivations for green innovation. We expect that state-owned enterprises (SOEs) have a more significant effect than non-state-owned enterprises (Non-SOEs). The results are shown in Columns (1) and (2) of Table 10. For state-owned suppliers (Column 1), the CusDA coefficient is 0.115 and significant at the 1% level (t = 2.73), while Non-SOE (Column 2) are insignificant (β = 0.028, t = 1.12). These results suggest that compared to non-SOEs, SOEs typically possess well-developed infrastructure, mature R&D systems, and production technologies, creating a good green innovation environment. In addition, under the policy drive of the “dual carbon” goals, SOEs have greater responsibilities for environmental governance. This institutional pressure prompts SOEs to take the initiative in carrying out green innovation activities. As a result, the spillover effect of customer data assets on suppliers’ green innovation is more pronounced when suppliers are SOEs.

Financing constraints may restrict suppliers from obtaining the financial support they need for green innovation. This study uses the FC index to measure the degree of financing constraints. The larger the FC index, the more severe the financing constraints faced by the enterprise. We divide the sample into two subsets by the median, Higher FC and Lower FC. The results are shown in Columns (3) and (4) of Table 10. In the lower financial constraint group (Column 4), the CusDA coefficient is 0.089 and significant at the 5% level (t = 2.35), whereas the higher constraint group (Column 3) is insignificant (β = 0.031, t = 1.23). This verifies our expectation and suggests that when suppliers are confronted with financing constraints, their pressure on capital turnover increases, making it difficult for them to provide sufficient funds for green innovation.

4.8. Further Analysis

We further examine whether data assets can also improve green innovation quality. Following Qi et al. (2021) [99], we use the number of the firm’s green innovation patent applications (Application) and grants (Granted) as proxy measurements. The regression results are reported in Table 11. In column (1), the impact of CusDA on Application is positive and significant at the 1% level (coefficient = 0.289, t-statistic = 3.10), suggesting that customer data assets also contribute to the volume of supplier green patents. Moreover, in column (2), CusDA exerts a positive and significant effect on Granted (coefficient = 0.136, t-statistic = 2.11), indicating that CusDA also facilitates the success of applications (e.g., patent grants).

Table 11. Further analysis.

	(1)	(2)	(3)
Variables	Application	Granted	R&D
CusDA	0.289 ***	0.136 **	0.089 *
	(3.10)	(2.11)	(1.91)
Controls	Included	Included	Included
Constant	−10.397 ***	−5.876 **	1.838
	(−2.99)	(−2.41)	(1.38)
Year FE	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes
Observations	1129	1129	839
R-squared	0.247	0.263	0.565

Notes: This table provides the results of further analysis. The t-statistics are reported in parentheses. Detailed variable definitions are summarized in Table 2. ***, **, and * indicate significance levels of 1%, 5%, and 10%, respectively.

In addition, we also find that customer data assets spur the R&D investment of suppliers. In this research, the natural logarithm of R&D spending is used as a proxy variable. The empirical results are listed in Column (3) of Table 10, which verifies our expectation with a positive and significant coefficient (β = 0.089, t = 1.91).

5. Conclusions and Discussion

Using data from Chinese listed firms between 2014 and 2024, this study investigates how customers’ data assets influence suppliers’ green innovation. We show that customers’ data assets significantly promote suppliers’ green innovation and that this spillover effect is more pronounced when geographic distance between customers and suppliers is shorter, and when customer concentration is higher. Furthermore, mechanism tests indicate that customers’ data assets enhance suppliers’ green innovation primarily by facilitating their digital transformation and improving operational efficiency, thereby strengthening suppliers’ capacity to undertake and sustain green innovation. Our heterogeneity analysis reveals that the spillover effects vary systematically with both customer and supplier characteristics, including firm location, industry type, ownership structure, and financial constraints. Further analysis demonstrates that customers with more substantial data assets not only stimulate suppliers’ green innovation but also drive greater R&D investment intensity and improve the quality of suppliers’ green innovation outputs, as measured by green patent granted. Taken together, these findings underscore the pivotal role of customer data assets in driving green innovation along the supply chain and highlight the importance of fostering data-enabled collaboration between upstream and downstream firms in the digital economy.

This study offers several theoretical contributions. On the one hand, we extend the literature on green innovation by identifying data assets as a key driving factor. As the important outcome of digital transformation, the value creation potential of data assets still needs to be verified for its role in green innovation. According to the resource-based view, data assets are scarce and inimitable resources that are essential for firms’ sustained competitive advantages and long-term success [51]. Based on this, this study integrates data assets and green innovation into a theoretical framework, revealing the intrinsic mechanism through which data assets, as strategic resources, impact green innovation performance.

On the other hand, we extend the literature on the economic implications of data assets by uncovering supply-chain-level effects on green innovation. Prior studies evidence that data assets hold immense potential for driving informed decision-making and fostering governance efficiency, thereby enhancing ESG performance and promoting green innovation [26,28]. However, these studies are primarily from the perspective of the firms themselves but fail to explore the spillover effects along the supply chain. Specifically, a critical gap remains: whether and how data assets owned by customer firms (as core stakeholders in supply chains) influence suppliers’ green innovation has not been examined. To address this gap, we incorporate stakeholder theory and slack resources theory into our analytical framework. According to the stakeholder theory, customers motivate suppliers to align their strategic behaviors (e.g., digital transformation) with customers’ demands to maintain cooperative relationships. Meanwhile, slack resources theory suggests that the improved operational capacity of suppliers generates organizational slack, which mitigates the resource constraints and risk uncertainties inherent in green innovation. Based on these theories, we find that customers’ data assets promote suppliers’ green innovation through facilitating supplier digital transformation and improving supplier operational capacity.

Based on prior empirical results, this study offers following practical implications. For customers, firstly, customers should recognize the importance of strengthening data asset utilization and information disclosure. On the one hand, since the effective utilization of data assets enhances firm competitiveness, firms should proactively recognize the strategic value of data assets and leverage them to build long-term competitive advantages. On the other hand, data asset disclosure is not only a future trend in information dissemination but also a strategic initiative to enhance financing capabilities. Therefore, firms should disclose information related to data assets more actively and transparently, establish clear data asset disclosure policies and integrate these policies into their corporate governance frameworks. Secondly, customers should promote firm green innovation with digital technologies and data assets. As green innovation becomes an optimal means of balancing environmental and developmental goals, customers should realize the importance of promoting green innovation. Therefore, customers should fully leverage digital technologies to promote the integration and dissemination of green knowledge, achieve cross-departmental collaborative innovation, and provide continuous impetus for green innovation. For example, customers may use their data assets to conduct market demand analysis as well as accurately identify green innovation directions.

For suppliers, firstly, suppliers could leverage partnerships with customers within their supply chain to enhance their green innovation performance. Since green innovation involves relatively high R&D risks and requires sustained efforts, the willingness of firms to participate is not strong. However, this study finds that the spillover effect of customer data assets can promote supplier green innovation. Therefore, suppliers should seek collaboration within supply chain networks, such as technology sharing, resource integration, and information interoperability to improve firms’ green innovation performance. Secondly, suppliers should continuously improve their own digital capabilities. Since data assets become firms’ core intangible assets and are helpful for firms to generate novel value and gain competitive advantages, suppliers should establish a standardized data collection and integration system, increase investment in digital infrastructure, as well as fully leverage digital technology to accelerate firms’ digital transformation.

For governments, firstly, to promote data value realization, governments should establish and improve data asset ownership confirmation and protection systems, clarify the definition, classification, and valuation rules of data assets, and accelerate the integration and development of data assets. Further, governments should establish incentive schemes for data sharing and strengthen platform supervision to foster more active data collaboration across firms. Secondly, since green and low-carbon development has become a key priority for high-quality growth in the new era, governments should provide financial support for green innovation and promote supply chain green innovation collaboration. On the one hand, governments should expand green financing channels, such as green bonds, and offer subsidies for green R&D. On the other hand, governments should recognize the positive impact of the supply chain on corporate green innovation and introduce or optimize supportive policies to facilitate collaboration within supply networks.

However, this study still has potential limitations. Firstly, the measurements of data assets are still in the exploratory stage. Considering the intangible nature of data assets, this paper uses text analysis methods to construct a measure of data assets, which may not fully capture the true value and heterogeneity of data assets. Future research could explore different types of data assets, such as internal or customer-derived data, or proprietary data assets and transactional data assets. Further, with the gradual integration of data elements into accounting systems, some qualitative methods may be employed to accurately measure data assets. Secondly, the sample size is constrained by customer disclosure availability, and the small- and medium-sized firms are excluded. This limits the external validity of the research conclusions to specific firms’ contexts. Future research could incorporate unlisted firm data to expand coverage to examine the effects of data assets on green innovation in small and micro-enterprises. Thirdly, the Chinese setting limits generalizability, while the cross-country comparisons with different digitalization levels would be valuable. Further research may conduct cross-country comparative studies to assess institutional boundary conditions, which can enhance the external validity and generalizability of findings. Furthermore, the measurement of green innovation may not fully capture the innovation quality or commercial impact. This study uses the natural logarithm of the number of green patent applications to measure green innovation. In further research, we employ green patent grants as the proxy for supplier green innovation to partially capture innovation quality. However, these measurements still fail to reflect whether these green patents are eventually implemented or generate substantive environmental and economic benefits. Future research could address this gap by integrating multiple indicators to measure green innovation more comprehensively, such as incorporating patent citation rates, technology licensing revenue, or green product performance. Lastly, this study considers a static perspective on supplier–customer pairs, while future research may consider the dynamic changes in supply chain relationships, such as customer changes or supplier changes.