The effect of supplier digitalization on customer firm emissions

Abstract

The application of digital technology is a crucial means to achieve low-carbon development goals, with supplier firms’ digital transformation driving supply chain optimization as a key pathway to this goal. This study systematically examines the forward spillover effects of supplier firms’ digital transformation. Using panel data from A-share listed companies in China over the period 2007–2022, we focus on vertical supply chain linkages. The results indicate that supplier firms’ digital transformation significantly reduces customer firms’ environmental pollution emissions, a conclusion that remains robust across various tests. Moreover, we find that this effect is asymmetric: it is significantly stronger when suppliers’ digitalization levels exceed those of their customers. Heterogeneity analysis shows that the pollution-reduction effect is stronger when customer firms are located in regions with lower resource endowments, have higher supply-chain dependence, operate in non-heavy-pollution industries, or are non-state-owned enterprises. Mechanism analysis demonstrates that supplier digitalization achieves these gains by enhancing green technological innovation and increasing external stakeholder attention. The findings provide a foundation for fully leveraging the environmental potential of digital transformation within supply chains and offer important insights into advancing supply chain modernization and promoting green, high-quality development.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-24812-z.

Against the backdrop of worsening global ecological and environmental issues, advancing green and low-carbon development has become a critical priority. The Report of the 20th National Congress of the Communist Party of China emphasizes the need to accelerate the green transformation of development models and promote harmony between humanity and nature. In this process, firms—as market entities responsible for promoting green development—are obligated to balance profit maximization with building sustainable and efficient business models, while actively adopting green and low-carbon practices. While policy efforts over the past decade have aimed to curb emissions growth, aggregate trends suggest that corporate pollution has continued to rise modestly, as illustrated in Fig. 1, which shows an increase from 1.13 to 2.25 in the pollution emissions index of Chinese listed companies between 2007 and 2022. Pollution emissions permeate various stages of the supply chain, including energy use, production, and logistics; isolated green initiatives by individual firms cannot effectively achieve overall low-carbon transformation¹. Thus, driving green and low-carbon transformation throughout the entire supply chain is essential for achieving the “dual carbon” goals and promoting high-quality economic development in China. Meanwhile, China’s “14th Five-Year Plan for Industrial Green Development” underscores the need to “coordinate development with green and low-carbon transformation, implement green manufacturing comprehensively,” and “advance the construction of a green, intelligent, and ecological civilization while promoting integration of digitalization and greening.” Digital transformation is gradually emerging as a new driver of enterprise development², enabling companies to overcome path dependence on traditional models. Against this backdrop of a deepening technological and industrial revolution, examining how enterprise digital transformation drives emission reduction and generates spillover effects across supply chains holds significant practical importance.

Fig. 1.

Pollution emissions of Chinese listed companies (2007–2022) Note: Data are indexed measures of aggregate pollution emissions for Chinese listed companies; actual values are calculated based on the data in the CSMAR Database.

Existing studies have found that digital transformation can significantly reduce firm pollution emissions and empower enterprises in their green transformation efforts³. For instance, Zhang et al.⁴ demonstrate that IoT-driven data analytics can reduce energy waste by 15–20% in manufacturing sectors, while Lin⁵ highlights how blockchain technology enhances transparency in carbon footprint tracking. Most current studies focus on the direct impact of enterprises’ digital transformation on their environmental performance, with limited attention paid to the interlinkages between upstream and downstream firms in the supply chain. Enterprises do not operate in isolation within market environments but are embedded within supply chain networks. These complex networks foster interactions between firms, leading to extensive factor flows and information sharing. This effect propagates to customer firms through a “ripple effect” significantly impacting the achievement of overall pollution reduction and carbon neutrality goals⁶. However, research on supply chain spillover effects has predominantly focused on how customer firms influence supplier firms’ productivity⁷ or information spillovers⁸, with relatively less attention paid to the leading and driving roles of supplier firms within supply chains. As the initial nodes in supply chains, supplier firms, due to their potential monopolistic attributes and market power, can propagate their behavioral decisions downstream to customer firms through supply chain diffusion effects. For instance, supplier firms can directly influence customer firms’ production and operational decisions through pricing, technology, and resource allocation, while indirectly guiding customer firms’ strategic adjustments through long-term partnerships and signal transmission⁹. Customer firms serve as critical intermediaries bridging upstream suppliers and end markets. Their environmental behaviors are shaped not only by internal governance but also by supplier-driven technological standards and contractual norms. Therefore, studying how supplier firms influence customer firms’ decision-making through supply chain spillover effects is critical for understanding the dynamic interactions between firms and advancing collaborative supply chain development.

Although some studies have explored the influence of supplier firms on customer firms’ decision-making, they often analyze supplier firms’ behaviors, such as upstream foreign investment openness, at the industry level¹⁰. Systematic research on how supplier firms’ behaviors are transmitted to customer firms-particularly how supplier firms’ digital transformation impacts customer firms’ environmental pollution emissions through supply chain diffusion-remains insufficiently explored. Research on green supply chains¹¹ rarely addresses vertical interdependencies, despite evidence that supplier activities account for over 40% of industrial carbon emissions in China’s manufacturing sector. This limitation is critical, as firms operate within interconnected networks where supplier decisions propagate downstream through pricing, technology diffusion, and contractual norms. Specifically, existing studies mainly examine the mechanisms by which digital transformation affects firms’ internal corporate governance¹², labor income share¹³, and green innovation¹⁴. It can be observed that while existing literature touches on the impact of supply chain firms’ digital transformation on focal firms’ behavioral performance, research remains insufficient on how supplier firms’ digital transformation influences customer firms’ environmental pollution emissions through supply chain spillover effects. For example, while studies like Fan¹⁵ emphasize digital tools’ role in energy efficiency, they neglect the moderating effects of regional resource endowments—a key factor in China’s spatially heterogeneous industrial landscape. Furthermore, prior literature has predominantly focused on internal governance mechanisms, with insufficient exploration of how external stakeholders amplify digital transformation’s environmental impact¹⁶.

Therefore, this paper adopts a vertical linkage perspective of upstream and downstream enterprises in the supply chain, focusing on three core questions: First, How does supplier firms’ digital transformation affect customer firms’ environmental pollution emissions through supply chain spillover effects? Second, What are the specific mechanisms through which supplier firms’ digital transformation influences customer firms’ environmental pollution emissions? Third, the heterogeneity in the effects of supplier digital transformation on customer firms’ environmental pollution emissions—across individual firms, industries, and geographic regions—remains underexplored, despite its relevance to China’s diverse economic landscape. In-depth research on this topic can help uncover the interconnected effects of digital transformation within supply chains at the micro level, providing critical insights for achieving the dual transformation goals of digitalization and greening while continuously optimizing overall supply chain efficiency.

The potential marginal contributions of this study are as follows: First, it examines the impact of digital transformation on customer firms’ environmental pollution emissions from the perspective of the vertical diffusion of supplier firms’ behavior. This reveals the optimization role of supplier firms in reducing customer firms’ pollution emissions within supply chain relationships, enriching and expanding existing research on the economic outcomes of supplier firms’ digital transformation and the determinants of customer firms’ environmental pollution emissions. Furthermore, in our large-sample empirical analysis, we find that supplier digital transformation yields a significant reduction in customer firms’ pollution emissions only when suppliers’ digitalization levels exceed those of their customers. Second, it explores the mechanisms through which supplier firms’ digital transformation reduces customer firms’ environmental pollution emissions by enhancing green technological innovation and external stakeholder attention from a supply chain spillover perspective. This provides micro-level evidence and theoretical explanations for the vertical spillover of supplier firms’ digital transformation to customer firms along the supply chain. This study not only clarifies the logic behind the forward spillover effects of supplier firms’ digital transformation on pollution reduction but also offers insights into improving supply chain coordination mechanisms and promoting firm green transformation. Third, it systematically examines and tests the specificities and heterogeneities in how supplier firms’ digital transformation affects customer firms’ environmental pollution emissions, considering factors such as resource endowments, supply chain dependence, industry characteristics, and ownership structure. This contributes to clarifying the complex effects of pollution reduction from the supply chain perspective of supplier firms’ digital transformation.

The remainder of this paper is organized as follows. Section 2 constructs the theoretical framework and hypotheses, focusing on supply chain spillover mechanisms. Section 3 details the empirical strategy, including data sources, variable definitions, and model specifications. Section 4 then presents baseline regression results and rigorously addresses endogeneity through instrumental variable approaches and Heckman selection models. Section 5 conducts robustness checks by altering variable measurements and sample scopes. Section 6 examines heterogeneous effects across firm attributes, industries, and institutional contexts. Finally, Sect. 7 concludes with policy implications and outlines future research directions.

Theoretical analysis and research hypotheses

The digital transformation of supplier firms and the environmental pollution emissions of customer firms

Customer firms serve as critical in direct transactions with upstream suppliers within vertically integrated supply chains. Their operational decisions, including production planning and technology adoption, as well as environmental outcomes, are significantly influenced by supplier-driven contractual terms and digital spillovers. As a critical approach to enhancing overall supply chain efficiency and promoting sustainable development, the digital transformation of supplier firms plays a significant role in reducing the environmental pollution emissions of customer firms. This effect can be analyzed from two perspectives: the resource effect and the synergy effect.

Resource Effect. Supplier digital transformation optimizes resource allocation, cutting customer-firm pollution by improving information flow and reducing waste. Real-time data sharing and intelligent algorithms enable suppliers to deliver precise demand forecasts and manage resource inputs efficiently, allowing customers to maintain output with fewer materials and energy, thus lowering emissions and minimizing upstream extraction and transport impacts¹⁷.

Synergy Effect. Digital platforms enhance supply-chain collaboration and automated governance. Real-time analytics and data-sharing tools help suppliers provide customers with actionable usage insights, improving production planning and cutting waste. Smart contracts and blockchain establish clear environmental responsibilities and enforce incentives, aligning supplier-customer goals and ensuring compliance. Immutable records boost credibility, prompting customers to adopt cleaner technologies and management practices, which further reduces pollution¹⁸.

In addition to these forward pathways, we recognize that customer firms’ environmental performance may feed back to influence supplier digitalization decisions. Superior pollution-control outcomes at the customer level can increase market and regulatory pressure, compelling upstream suppliers to accelerate their own digital transformation to meet higher green standards and secure continued business. Accordingly, to mitigate potential estimation bias from reverse causality in our baseline regressions, we employ both an instrumental variables approach and a Heckman two-stage estimation procedure in the empirical analysis.

Based on the above analysis, the following research hypothesis is proposed:

H1a

The digital transformation of supplier firms, on the whole, facilitates reductions in environmental pollution emissions by their customer firms.

In addition, we contend that supplier digitalization reduces customer firms’ pollution emissions only when suppliers occupy a relative position of digital leadership within the buyer–supplier relationship. In other words, suppliers’ digital investments generate downstream environmental spillovers only if suppliers’ digital capabilities exceed those of their customers.

First, from a power-asymmetry perspective, supplier firms that possess superior digital platforms and interoperable architectures command greater bargaining power in the relationship. This bargaining power allows digitally leading suppliers to set technical standards and design contractual incentives that make it optimal for downstream firms to adopt cleaner inputs and production routines¹⁹. When supplier firms lack this relative advantage, they are unlikely to convert upstream digital investments into downstream behavioral change. Absent the ability to influence standards and procurement rules, supplier firms’ innovations tend to remain underutilized by customers²⁰. Second, digital maturity operates through threshold effects. Meaningful operational and environmental gains accrue only after firms cross capability thresholds that enable seamless integration and automated control across organizational boundaries²¹. Supplier firms that have not reached such thresholds cannot reliably deliver turnkey solutions that bridge customer firms’ capability gaps; by contrast, supplier firms that exceed these thresholds can provide technologies that customer firms can feasibly adopt, thereby unlocking efficiency gains and removing emission-intensive bottlenecks²². Overall, only when suppliers’ digital capabilities surpass those of their customer firms will interoperable technologies and actionable data combine to produce coordinated operational changes that lower pollution. We therefore hypothesize:

H1b

Supplier firms’ digital transformation yields a significant reduction in customer firms’ pollution emissions only when suppliers’ digitalization levels surpass those of their customers.

The mechanism of the impact of supplier firms’ digital transformation on customer firms’ environmental pollution emissions

Digital transformation boosts firms’ green innovation by streamlining R&D processes and enabling real-time progress tracking. In supply chains, highly digitalized suppliers not only optimize their own operations but also elevate downstream green innovation through knowledge and technology transfer. First, digital tools enhance suppliers’ information networks, slashing search costs and enabling customers to quickly adopt green technologies via shared data streams. This rapid access to production and R&D insights improves innovation efficiency, thereby lowering pollution intensity²³. Second, suppliers leverage digital platforms and contractual clauses to mandate customer adoption of low-carbon processes and provide on-demand technical support, accelerating clients’ green technology upgrades at minimal cost²⁴. Finally, suppliers’ own advancements in green tech diffuse through interconnected innovation networks, prompting customers to emulate and refine these solutions. Consequently, customer firms enhance their processes and further reduce per-unit emissions²⁵. Based on the above analysis, this paper proposes the following hypothesis:

H2

The digital transformation of supplier firms can reduce customer firms’ environmental pollution emissions by enhancing their green technological innovation levels.

Digital transformation not only strengthens firms’ green innovation but also enhances environmental governance via capital markets, media scrutiny, and policy signals. First, IoT and big-data platforms link supply-chain actors to analysts and investors, making customer firms’ environmental performance more transparent and increasing the risk of public exposure for violations. This scrutiny deters short-term decision-making and spurs green innovation²⁶. Second, peer effect forces amplify digital adoption downstream, attracting green-oriented investors who integrate environmental criteria into funding decisions, further incentivizing customer firms to bolster their sustainability practices²⁷. Third, networked media can rapidly disseminate information on environmental practices, creating a “spotlight effect” that raises reputational stakes and compels firms to adopt cleaner technologies. Positive coverage, in turn, enhances green reputational capital and accelerates customer firms’ environmental improvements²⁸. Based on the above analysis, this paper proposes the following hypothesis:

H3

The digital transformation of supplier firms can reduce customer firms’ environmental pollution emissions by increasing external stakeholders’ attention.

Building on the foregoing theoretical analysis, we further develop our study’s theoretical model, as illustrated in Fig. 2.

Fig. 2.

Theoretical framework of supplier digitalization impact on customer firms’ environmental pollution emissions.

Research design

Model construction

To examine the impact of supplier firms’ digital transformation on customer firms’ environmental pollution emissions, the following econometric model is constructed:

1

where represents the dependent variable, indicating the pollution emissions of firm i in year, is the independent variable, denoting the level of digital transformation of supplier firmi in year , represents control variables, including firm-specific characteristics, governance variables, and supplier firm characteristics, and denote firm and year fixed effects, respectively, and ​ is the random disturbance term.

Variable selection

Pollution emissions of customer firms

Customer firms’ pollution emissions primarily include water and air pollution²⁹. Water pollution emissions consist mainly of chemical oxygen demand and ammonia nitrogen emissions, while air pollution emissions include sulfur dioxide, ammonia nitrogen compounds, and soot and dust emissions. Due to unit inconsistencies among different pollutants, this study employs a composite index method to construct a comprehensive measure of firm pollution emissions. First, the original data for the five pollutants are standardized. It should be noted that the primary source for the statistical data on various pollutant emissions is the CSMAR database; however, the sample size disclosed in this database is relatively limited. To further expand the sample size, this study supplemented the pollutant emissions data of firms by extracting information from annual reports, firm social responsibility reports, and sustainability reports of publicly listed companies, thereby enhancing both the number of entities and the temporal coverage of the dataset.

2

where represents the emissions of pollutant k by firm in year, and denote the maximum and minimum emissions of pollutant kacross all years in the sample, respectively.

Next, the adjustment coefficient for pollutant k by firm is calculated:

3

where represents the average emissions of pollutant k across the sample.

Finally, combining Eqs. (2) and (3), the comprehensive index of firm pollutant emissions is obtained ():

4

A higher value indicates greater pollution emissions by the firm.

Digital transformation of supplier firms

The China Securities Regulatory Commission (CSRC) mandates listed companies to provide outlooks on industry trends and firm strategies. Therefore, if a firm adopts a digital transformation strategy, it is usually elaborated in the “Management Discussion and Analysis” (MD&A) section. Generally, a higher proportion of relevant digital transformation keywords indicates stronger motivation for implementing digital strategies. Drawing on Yuan et al.³⁰ here, and Zhen et al.³¹, this study compiles digital transformation keywords from three dimensions: technology classification, organizational enablement, and digital application. It should be noted that the keywords for digital transformation are detailed in Appendix Ⅰ. The MD&A sections of annual reports are analyzed using text analysis methods to extract the frequency of relevant keywords, and their proportion is calculated to measure the degree of digital transformation. The specific steps are as follows:

First, Python is used to scrape annual reports of listed companies from 2007 to 2022 and convert the MD&A sections into text format, forming a data pool for keyword statistics. Second, digital transformation keywords are added to the Jieba Chinese word segmentation library, and sample annual reports are segmented to count the frequency of specified keywords in the MD&A section for each dimension. These frequencies are categorized, aggregated, and summed. Third, the total frequency of digital transformation keywords is divided by the total word frequency in the MD&A section to measure the degree of digital transformation of supplier firms. For convenience, the value of this indicator is multiplied by 100. A higher value indicates a higher level of digital transformation of supplier firms.

Control variables

To accurately estimate the effect of supplier firms’ digital transformation on customer firms’ pollution reduction, this study includes a series of control variables, following the research of Tang et al.³², Chen et al.³³. These variables include customer firms’ leverage ratio (), customer firm’ size (), customer firms’ return on equity (), customer firms’ listing age (), customer firms’ growth (), customer firms’ capital intensity (), customer firms’ book-to-market ratio (), customer firms’ industry concentration (), customer firms’ ownership concentration (), supplier firm’ size (), supplier firm’ fixed asset ratio (), and supplier firm’ return on equity (​). Specifically, is included to account for financial flexibility, as high debt levels may constrain firms’ ability to invest in green technologies³⁴. reflects resource availability, with larger firms often possessing greater capacity to adopt digital and green innovations, though organizational complexity may slow implementation³⁵. serves as a proxy for profitability, which is linked to sustainability investments, whiledistinguishes between established firms with stronger institutional commitments to environmental regulations and younger firms prioritizing growth³⁶. Customer firms’ growth () addresses the potential trade-off between rapid expansion and pollution reduction efforts, particularly in manufacturing sectors³⁷. highlights the dual role of capital-intensive industries, which face higher emissions but benefit more from digital efficiency tools³⁸. captures firms’ growth potential, with low BM firms more likely to prioritize green innovation³⁹. reflects competitive dynamics, where fragmented markets may drive firms to adopt digital tools for differentiation, while monopolistic industries exhibit weaker incentives⁴⁰. explores governance structures, as concentrated ownership may accelerate sustainability decisions but also prioritize short-term profits over long-term environmental goals⁴¹. Supplier firm characteristics (, ,​) are included to capture upstream influences, which determine their ability to comply with downstream digital and environmental requirements⁴². Table 1 provides the definitions and descriptions of these variables.

Table 1.

Definition of core and control variables.

Type of variable	Name of variable	Definition of Variable
Explained variable	Pollution_cust	Comprehensive Index Method (logarithm taken)
Explaining variable	Updigi_supp	Text Analysis Method - Word Frequency Statistics
Control variable	Lev_cust	Total assets / Total liabilities of customer firms
	Emp_cust	Customer firms’ number of employees (logarithm)
	ROE_cust	Return on equity of customer firms
	Age_cust	Listing age of customer firms
	Growth_cust	Revenue growth rate of customer firms
	Capital_cust	Capital intensity of customer firms
	BM_cust	Book-to-market ratio of customer firms
	Top_cust	Shareholding ratio of the largest shareholder in customer firms
	HHI_cust	Customer firms’ main business income / Industry main business income (sum of squares of ratios)
	Emp_supp	Number of employees in Supplier firm (logarithm)
	Fa_supp	Fixed assets / Total assets of supplier firm
	ROE_supp	Return on equity of supplier firm

Data description

This study uses data from China’s Shanghai and Shenzhen A-share listed companies from 2007 to 2022. Since a supplier firm (Supp) in a given year (e.g., 2022) may correspond to multiple downstream customer firms (Cust1, Cust2, Cust3), we construct observations such as Supp—Cust1—2022, Supp—Cust2—2022, and Supp—Cust3—2022, forming a supplier firm—customer firm—year dataset. The sample was further processed as follows: retained samples where both upstream suppliers and downstream customers are listed firms, excluded samples from the financial industry, excluded ST firms and those delisted or with missing data during the observation period, excluded outliers and samples with missing data, and excluded firms that underwent IPOs during the sample period. This resulted in 887 observations of supplier firm—customer firm—year data. It should be noted that during the construction of supplier firms–customer firms–annual observation values, samples where customer firms were joint ventures, associates, or subsidiaries of a listed company were excluded. Only samples where the customer firms were the listed companies themselves were retained. The data mainly come from the CSMAR and Wind databases. Table 2 presents descriptive statistics for the main variables. The results show that the mean customer firm pollution emissions are 1.9086, with a maximum value of 2.4093, a minimum value of 1.0534, and a standard deviation of 0.228, indicating significant differences in pollution emissions among firms. The mean supplier firm digital transformation score is 0.0965, with a standard deviation of 0.4195. Control variable data, such as leverage ratio, return on equity, and ownership concentration, are all within reasonable ranges.

Table 2.

Descriptive statistics of main variables.

Variable	Obs	Mean	SD	Min	Max
Pollution_cust	887	1.9086	0.228	1.0534	2.4093
Updigi_supp	887	0.0965	0.4195	0.0000	8.1006
Lev_cust	887	0.4682	0.2031	0.0327	0.9763
Emp_cust	887	7.8666	1.2051	3.6636	12.5914
ROE_cust	887	0.0234	0.9505	− 27.5949	4.2476
Age_cust	887	10.7711	7.2318	1.0000	28.0000
Growth_cust	887	0.1751	0.8835	− 0.7652	23.3152
Capital_cust	887	2.4395	2.2919	0.1396	45.4294
BM_cust	887	0.6626	0.2566	0.0534	1.3354
Top_cust	887	0.3461	0.1451	0.0431	0.8504
HHI_cust	887	0.1876	0.1548	0.0380	1.0000
Emp_supp	887	23.0633	1.8515	18.9183	28.7183
Fa_supp	887	0.2798	0.1758	0.0013	0.8098
ROE_supp	887	0.0700	0.2914	− 7.2203	0.6298

Empirical analysis

Baseline regression

Table 3 Reports the baseline regression results on the impact of supplier firms’ digital transformation on the pollution emissions of downstream customer firms. Column (1) controls only for firm fixed effects, with the coefficient of the core explanatory variable, supplier firms’ digital transformation, being − 0.0198 and significant at the 1% level. Column (2) simultaneously controls for both time and firm fixed effects, with the coefficient for supplier firms’ digital transformation being − 0.0195, and its significance level remains unchanged. This indicates that supplier firms’ digital transformation significantly reduces customer firms’ pollution emissions, demonstrating a significant negative correlation between the two. Hypothesis 1a is thus verified.

Table 3.

Benchmark regression results

Variable	Pollution_cust
	(1)	(2)	(3)	(4)
Updigi_supp	− 0.0198***	− 0.0195***	− 0.0281***	0.2221
	(0.0052)	(0.0062)	(0.0091)	(0.3918)
Lev_cust	0.1382*	0.1012	− 0.1350	0.1277
	(0.0777)	(0.0747)	(0.1444)	(0.1991)
Emp_cust	0.0394	0.0090	− 0.1200**	0.0946
	(0.0247)	(0.0237)	(0.0557)	(0.0655)
ROE_cust	− 0.0051***	− 0.0046**	− 0.0072**	− 0.1244
	(0.0016)	(0.0018)	(0.0032)	(0.2197)
Age_cust	0.0626***	0.0718***	0.0927***	0.0313***
	(0.0029)	(0.0052)	(0.0136)	(0.0113)
Growth_cust	− 0.0087***	− 0.0080***	− 0.0088	− 0.0608*
	(0.0022)	(0.0022)	(0.0401)	(0.0315)
Capital_cust	− 0.0036	− 0.0058	0.0117	0.0183
	(0.0070)	(0.0067)	(0.0226)	(0.0358)
BM_cust	− 0.0511	− 0.0562	0.0840	− 0.1950*
	(0.0311)	(0.0418)	(0.0964)	(0.1028)
Top_cust	0.0516	0.1305	0.5819*	0.1211
	(0.0906)	(0.0974)	(0.3025)	(0.2715)
HHI_cust	− 0.0522	− 0.0275	0.0253	0.5355*
	(0.0707)	(0.0661)	(0.1479)	(0.3011)
Emp_supp	0.0224	0.0183	0.0860	− 0.0019
	(0.0201)	(0.0232)	(0.0534)	(0.0442)
Fa_supp	0.1152	0.0276	0.2316	− 0.1124
	(0.0888)	(0.0948)	(0.1989)	(0.2406)
ROE_supp	0.0164**	0.0211***	− 0.0879	− 0.0650
	(0.0066)	(0.0066)	(0.0866)	(0.0485)
_cons	0.3470	0.4343	− 0.7528	0.9425
	(0.4561)	(0.5230)	(1.2096)	(1.1738)
Firm FE	YES	YES	YES	YES
Year FE	NO	YES	YES	YES
R2	0.6959	0.7223	0.6844	0.6529
Obs	887	887	301	298

*, **, and ***indicate passing the significance test at 10%, 5%, and 1% levels, respectively. The brackets below the coefficients indicate the standard error of firm clustering robustness. The same applies to the tables below.

Testing the asymmetry of supplier-firm digitalization empowerment

We further investigate whether our baseline estimates remain robust when suppliers and customers differ in their relative digitalization intensity. Following the approach of Houston and Shan⁴³, we split the sample into two subsamples: one in which suppliers’ digitalization exceeds that of their customers, and another in which suppliers lag behind. We then re-estimate model (1) for each group. As shown in column (3) of Table 3, when suppliers are more digitally advanced than their customers, supplier digital transformation significantly reduces customer firms’ pollution emissions. By contrast, column (4) indicates that this effect becomes statistically insignificant when suppliers exhibit lower digitalization than their customers. These results reveal a clear asymmetry: only when suppliers possess a digitalization advantage can their digital initiatives effectively empower downstream firms to achieve greener outcomes. Hypothesis 1b is thus verified.

Addressing endogeneity

Instrumental variable approach

The conclusions of the baseline regression may be affected by endogeneity issues. On one hand, supplier firms’ digital transformation facilitates improvements in customer firms’ green innovation capabilities, thereby reducing environmental pollution emissions. On the other hand, customer firms’ pollution emissions may create backward spillover effects along the supply chain. Specifically, customer firms’ demands for pollution reduction exhibit clear directional characteristics, with terminal market demands for pollution reduction transmitted upstream through the supply chain, driving the digital transformation of upstream related firms. In other words, reductions in customer firms’ pollution emissions may influence the digital transformation level of supplier firms through supply chain traction, leading to reverse causality issues. To mitigate the endogenous impact of this reverse causality on the estimation conclusions, this paper employs two-stage least squares (2SLS) regression with appropriate instrumental variables. Following the approach of Yuan et al.³⁰ and Tao et al.⁷, we use the average digitalization of firms in the same industry and city as an instrument. A valid instrumental variable must satisfy the assumptions of “relevance” and “exogeneity.” First, the degree of digitalization of a firm is influenced by the information technology application environment within its regional industry, creating a positive correlation between an individual firm’s digitalization level and the digitalization level of its industry peers in the same region. Second, the overall digitalization status at the industry level within the same city is determined by the external environment faced by firms. Furthermore, unless multiple firms act collectively, it is unlikely for a single firm’s digital transformation to significantly elevate the digitalization level of a particular industry within a city. Given the random selection of sample firms, the likelihood of such collective actions is minimal, thereby satisfying the exogeneity condition.

Table 4 Presents the results of the two-stage least squares regression. Column (1) shows the first-stage regression results, indicating a significant positive correlation between the instrumental variable and the endogenous variable, thereby satisfying the relevance assumption. Column (2) displays the second-stage regression results, showing that the coefficient of firms’ digital transformation is negative and significant at the 1% level, consistent with the baseline regression results. Additionally, the underidentification test results indicate that the Kleibergen-Paap Rk LM statistic rejects the null hypothesis of underidentification at the 5% significance level. Meanwhile, the weak instrument test results, based on the Kleibergen-Paap Rk Wald F and Cragg-Donald Wald F statistics, indicate that when only one endogenous variable is present, both statistics exceed the critical values corresponding to a 10% tolerance distortion, as suggested by stock and Yogo (2005), confirming the absence of weak instrument issues.

Table 4.

Instrumental variable regression results

Variable	First-stage regression	Second-stage regression
	(1)	(2)
Updigi_supp		− 0.0152**
		(0.0065)
IV	0.999***
	(0.0254)
_cons	0.0608	0.9646***
	(0.1145)	(0.0831)
Controls	YES	YES
Firm fixed	YES	YES
Year fixed	YES	YES
Obs	887
Kleibergen-Paap rk LM statistic	3.9**
Kleibergen-Paap rk Wald F statistic	1550.07
Cragg-Donald Wald F statistic	10708.95***
	[16.38]
R2		0.8501

[ ] represents the critical value at the 10% level in the Stock-Yogo test, with control variables consistent with Model (1).

Heckman two-stage regression

Since listed companies disclose only major customers and do not provide information on customers with relatively small sales shares, this may lead to a “selection bias” in the sample. Referring to Yang Jinyu et al.⁴⁴, this paper adopts the Heckman two-stage regression to address potential selection bias in the sample. In the first-stage selection equation, the dependent variable is “whether the supplier firm has undergone digital transformation”. A dummy variable is assigned a value of 1 if digital transformation has occurred and 0 otherwise. The explanatory variables include the number of shares held by supplier firm management (MS_supp), the shareholding ratio of the largest shareholder (Top_supp), and the control variables from Model (1), which are used to conduct a Probit regression. In the second stage, the inverse Mills ratio (IMR) estimated in the first stage is included in the regression model to correct for potential self-selection bias in the sample. In addition, to assess the validity of the exclusion restrictions—MS_supp and Top_supp—we conduct a placebo test. Specifically, we use prior-period environmental pollution emissions as a placebo outcome and regress this outcome on MS_supp and Top_supp; statistically insignificant coefficients would support their exogeneity and thus their suitability as exclusion restrictions. Column (1) of Table 5 reports the first-stage of the Heckman two-step procedure and shows that the excluded variables satisfy the relevance condition in the selection equation. Column (2) presents the placebo regression, where the coefficients on MS_supp and Top_supp are statistically insignificant, supporting the validity of the exclusion restrictions. Column (3) reports the second-stage results. After correcting for sample-selection bias, the coefficient on supplier digitalization remains significantly negative, consistent with the baseline estimates.

Table 5.

Heckman two-stage regression results.

Variable	(1)	(2)	(3)
Updigi_supp			− 0.0191***
			( 0.0072 )
MS_supp	0.0577**	0.0008
	(0.0236)	(0.0032)
Top_supp	-0.0161	0.0029
	(0.0108)	(0.0020)
IMR			− 0.0408*
			(0.0216)
_cons	3.4658	1.8815	0.1220
	(2.3281)	(0.0894)	(0.7373)
Controls	YES	NO	YES
Firm fixed	YES	YES	YES
Year fixed	YES	YES	YES
R2		0.0118	0.8779
Obs	691	218	431

Robustness checks

Changing the measurement method for supplier firms’ digital transformation

Measuring the degree of firms’ digital transformation remains a challenge. The measurement of the digital transformation index for supplier firms in the baseline regression may involve errors. To alleviate estimation bias caused by measurement errors, the proportion of supplier firms’ digital intangible assets and digital technology innovation are separately used to reassess the digital transformation indicators. The notes in the financial reports of listed companies disclose detailed items of year-end intangible assets. Referring to Li et al.²² and He et al.⁴⁵, this paper uses the ratio of supplier firms’ digital intangible assets to the total intangible assets to measure digital transformation. Specifically, if the detailed items of intangible assets include keywords such as “software,” “network,” and “customer,” which are related to digital transformation, the item is considered a digital intangible asset. It should be noted that the keywords of digital intangible assets are detailed in Appendix II. The total of multiple digital intangible assets for the same firm in the same year is calculated, and its ratio to the firm’s total intangible assets for the year is used as a proxy variable for digital transformation. The regression for Model (1) was re-estimated. Column (1) of Table 6 shows that the coefficient for supplier firms’ digital transformation is significantly negative at the 1% level. Additionally, given that digital technology innovation is also a key indicator for measuring firms’ digital transformation, this paper uses the number of digital economy patent applications by supplier firms to measure their digital technology innovation level and as a proxy variable for digital transformation. Column (2) of Table 6 indicates that the coefficient for supplier firms’ digital transformation remains significantly negative. The above regression results suggest that, after controlling for the potential impact of measurement errors, the results remain consistent with the baseline regression.

Table 6.

Robustness test results: excluding other factors and replacing explanatory variables.

Variable	(1)	(2)	(3)	(4)
Updigi_supp	− 0.2134***	− 0.0170*	− 0.0193***	− 0.0192***
	(0.0775)	(0.0092)	(0.0036)	(0.0041)
_cons	0.3062	0.2492	0.8423***	0.8202***
	(0.4792)	(0.4543)	(0.1685)	(0.1783)
Controls	YES	YES	YES	YES
Firm fixed	YES	YES	YES	YES
Year fixed	YES	YES	YES	YES
R2	0.7021	0.6941	0.7567	0.7479
Obs	874	915	688	786

Excluding confounding factors

The spillover effect of supplier firms’ digital transformation on customer firms’ pollution emissions may not only propagate along the supply chain vertically but also through other channels. First, the spillover effect may originate from competitive effects within the industry in which the firms operate. The reduction in environmental pollution emissions resulting from firms’ digital transformation can be easily imitated by other firms within the same industry. Thus, the core conclusion of this paper may be driven by the fact that supplier and customer firms belong to the same industry. To address this, the sample in which upstream suppliers and downstream customers belong to the same industry was excluded, and the regression was re-estimated. Column (3) of Table 6 shows that the regression coefficient for supplier firms’ digital transformation remains significantly negative, indicating that the conclusion of supplier firms’ digital transformation significantly reducing customer firms’ pollution emissions still holds after excluding the influence of intra-industry competition effects. Second, the inter-firm spillover effects of digital transformation may result from the geographical proximity between supplier firms and customer firms. Geographical proximity is an important channel for traditional knowledge spillovers, as the closer proximity between upstream suppliers and downstream customers facilitates the dissemination of tacit knowledge within the production network. To mitigate the spillover effects arising from geographical proximity, the sample where supplier firms and customer firms are located in the same city was excluded, and the regression was re-estimated. Column (4) of Table 6 shows that the coefficient of the core explanatory variable remains significantly negative, indicating that the baseline regression conclusion remains valid even after further eliminating the interference caused by geographical proximity.

Inverse probability weighting and expanded sample

In the baseline regressions, some control variables are missing for a subset of observations, which reduces the effective sample size. To assess whether this sample reduction drives our results, we conduct two robustness exercises. (1) Inverse-probability weighting (IPW). We estimate a logit model for inclusion on the key explanatory variables and baseline controls, predict each firm’s inclusion probability, construct firm-level weights as the inverse of the mean predicted probability, and truncate extreme weights at 10. Re-estimating the baseline model produces a coefficient on supplier digitalization of − 0.0194, significant at the 1% level, as shown in Table 7, column (1). (2) Expanded-sample specification. We drop four control variables with relatively high rates of missingness—BM, HHI, ROE, Emp—to enlarge the estimation sample and re-estimate the baseline fixed-effects model. The coefficient on supplier digitalization remains negative and statistically significant in this specification, as shown in Table 7, column (2). Overall, these complementary checks indicate that our core finding—that supplier digitalization substantially reduces customer firms’ pollution emissions—is robust to plausible concerns about sample selection and missing controls.

Table 7.

Additional robustness tests.

Variable	(1)	(2)	(3)	(4)
Updigi_supp	− 0.0194***	− 0.0166**	− 0.0183***	− 0.0195***
	(0.0062)	(0.0079)	(0.0064)	(0.0064)
_cons	0.4255	0.7199*	1.2670*	1.3792**
	(0.5248)	(0.4156)	(0.7044)	(0.5536)
Controls	YES	YES	YES	YES
Firm fixed	YES	YES	YES	YES
Year fixed	YES	YES	YES	YES
R2	0.7220	0.7007	0.8718	0.8781
Obs	887	1021	390	578

Other robustness tests

The spillover effects of supplier firms’ digital transformation on customer firms’ pollution reduction were re-estimated using the following methods:

Retaining only manufacturing firm samples

The spillover effects of supplier firms’ digital transformation on customer firms’ pollution reduction may be influenced by the industry to which they belong. Considering that the observed sample includes firms from various industries, and firms with relatively high levels of pollution emissions are primarily in the manufacturing industry, the regression was performed after retaining only manufacturing firm samples. The results in Column (3) of Table 7 indicate that the core conclusion remains supported.

Adjusting the clustering level

When clustering the variance-covariance structure of the disturbance term, different levels of clustering imply different assumptions. As the clustering level increases, the assumptions become less restrictive, and the standard error estimates become more robust. To further enhance the robustness of the results, the clustering level was adjusted to the industry-time dimension, and the regression was re-estimated. The results, shown in Column (4) of Table 7, indicate that the coefficient of supplier firms’ digital transformation remains significantly negative, robustly supporting the basic conclusion that supplier firms’ digital transformation significantly promotes the reduction of customer firms’ pollution emissions. The above regression results suggest that, after controlling for the impact of industry affiliation and clustering on the estimation results, the baseline regression conclusion remains valid.

Mechanism testing

Based on the theoretical analysis above, the digital transformation of supplier firms can effectively enhance customer firms’ green technological innovation, thereby reducing environmental pollution emissions. To verify the mechanism operating through customers’ green innovation, we examine two dimensions—innovation quantity (Innovnum) and innovation quality (Innovqua). Specifically, we measure quantity by the total number of green patent applications and quality by the number of citations to green patents⁴⁶. Because patent grant data are delayed, we use patent application data. We take the natural logarithm of one plus each count to construct the indicators. Guided by Jiang Ting⁴⁷, who notes that stepwise mediation models are vulnerable to mediator endogeneity in the third-stage outcome-on-mediator regression and recommends validating mechanisms by first identifying the causal effect of X on a theoretically salient mediator M and then relying on accumulated evidence for the effect of M on Y, we treat customers’ green technological innovation as the mediator of interest. In the regressions, given the lagged nature of innovation behavior, we lag the innovation-quantity and innovation-quality variables by one period. Columns (1) and (2) of Table 8 report the mechanism-test results for internal green technological innovation. The coefficients on supplier firms’ digital transformation are significantly positive at the 5% and 1% levels, respectively, indicating that supplier firms’ digital transformation significantly enhances both the quantity and quality of customers’ green innovation. Furthermore, improvements in green technological innovation levels can optimize production processes, enhance resource utilization efficiency, and reduce pollution emissions per unit of output, effectively driving firms toward achieving pollution reduction targets. Meanwhile, the development and application of green technologies can not only reduce resource waste during production but also mitigate regulatory penalty risks associated with environmental violations, thereby enhancing firms’ market competitiveness and long-term sustainability⁴⁸. Moreover, the application of green technologies can strengthen firms’ environmentally friendly image, making them more attractive to investors and consumers. Crucially, these effects transfer from suppliers to customers through supplier-customer digital linkages: suppliers expose real-time process data and design specifications via interoperable platforms, embed low-carbon process requirements in procurement contracts, and provide on-demand technical assistance; customers then absorb and adapt these solutions, thereby upgrading their own stock of green innovation and the quality of implementation⁴⁹. The infusion of these external resources further enhances firms’ environmental management capabilities, creating a virtuous cycle that continuously improves environmental performance. Overall, supplier firms’ digital transformation can promote the enhancement of green technological innovation in customer firms, thereby reducing their environmental pollution emissions. Hypothesis 2 is thus validated.

Table 8.

Mechanism tests: internal green technological innovation and external stakeholder attention.

Variable	Innovnum	Innovqua	Analys	Investor	OA
	(1)	(2)	(3)	(4)	(5)
Updigi_supp	0.0903**	0.3166***	0.2879***	0.0637*	0.0589***
	(0.0454)	(0.0325)	(0.0490)	(0.0362)	(0.0215)
_cons	− 14.1298**	− 4.6072	− 11.8750**	12.4565***	9.1300**
	(6.0871)	(5.0503)	(4.6071)	(3.3836)	(3.5589)
Controls	YES	YES	YES	YES	YES
Firm fixed	YES	YES	YES	YES	YES
Year fixed	YES	YES	YES	YES	YES
R2	0.7386	0.8437	0.8471	0.8954	0.8426
Obs	251	251	251	251	251

The level of green development in firms relies not only on endogenous driving forces but is also significantly influenced by external supervisory mechanisms¹⁶. This paper focuses on three key stakeholders—analysts, investors, and online networks—and explores the mechanism through which external market attention impacts the green development of customer firms from a multidimensional perspective. Specifically, analyst attention (Analys) is measured by the number of securities analysts covering the same listed company. Investor attention (Investor) is measured by the total number of reads for posts in stock forums related to the listed company on a given day. In the process of constructing the indicators, web crawling technology was employed to collect data on investor discussions about listed companies from the stock forum section of the East Money website. To ensure the representativeness of the sample and the reliability of the data, only discussion data labeled as “hot posts” were used as the research sample. Online attention (OA) is measured by the total number of news reports in online media. For these metrics, we take the natural logarithm of one plus the number of securities analysts, one plus the total number of post reads, and one plus the total number of news reports. Because attention transmission involves a time lag, we lag analyst attention, investor attention, and online attention by one period to capture this dynamic relationship more accurately. Columns (3), (4), and (5) of Table 8 report the regression results of supplier firms’ digital transformation on customer firms’ analyst attention, investor attention, and online attention, respectively.

The coefficients are significantly positive across specifications, indicating that supplier firms’ digital transformation significantly increases external market attention—by analysts, investors, and online networks—toward customer firms. Based on stakeholder theory, firms must not only satisfy shareholders’ economic interests but also actively respond to the environmental expectations of various stakeholders, including investors, consumers, and the general public⁵⁰. The heightened attention from external market participants increases supervision and constraint pressure on firms. To protect their reputation and meet the green demands of external stakeholders, firms often proactively fulfill environmental responsibilities and adopt green production practices, thereby effectively reducing environmental pollution emissions. Importantly, attention initiated at the supplier level diffuses to connected customers through sell-side analysts’ coverage overlap and media co-mentions on supply-chain traceability platforms; moreover, suppliers’ ESG disclosures routinely reference key customers, thereby increasing public visibility and regulatory scrutiny of those customers. This transferred attention raises the expected costs of non-compliance and increases the reputational returns to cleaner production for customers, strengthening their incentives to reduce emissions intensity⁵¹. Overall, supplier firms’ digital transformation can enhance external market attention to customer firms, thereby reducing their environmental pollution emissions. Hypothesis 3 is thus validated.

Heterogeneity analysis

Heterogeneity in resource endowments

Differences in resource endowments across regions where customer firms are located may lead to varying effects of supplier firms’ digital transformation on their pollution reduction. Following the approach of Guo et al.⁵², who classified regions based on natural resource reserves and economic development levels, this study categorizes the regions where customer firms are located into high-resource-endowment and low-resource-endowment areas to examine the relationship between supplier firms’ digital transformation and the environmental pollution of firms in these regions. The high-resource-endowment regions include Ningxia Hui Autonomous Region, Hunan Province, Liaoning Province, Xinjiang Uygur Autonomous Region, Yunnan Province, Sichuan Province, Hebei Province, Heilongjiang Province, Anhui Province, Guizhou Province, Shandong Province, Henan Province, Shaanxi Province, Inner Mongolia Autonomous Region, and Shanxi Province. The low-resource-endowment regions include Tianjin Municipality, Shanghai Municipality, Hainan Province, Zhejiang Province, Guangdong Province, Guangxi Zhuang Autonomous Region, Beijing Municipality, Tibet Autonomous Region, Hubei Province, Qinghai Province, Fujian Province, Jiangsu Province, Jiangxi Province, Chongqing Municipality, Jilin Province, and Gansu Province. Columns (1) and (2) of Table 9 show that the digital transformation of listed companies significantly promotes pollution reduction in downstream firms in low-resource-endowment regions, but has no significant effect on firms in high-resource-endowment regions. This disparity aligns with Li et al.⁵³, who found that firms in resource-constrained regions exhibit stronger incentives to adopt digital technologies due to competitive pressures, thereby improving resource efficiency and reducing emissions. This difference may stem from the fact that firms in low-resource-endowment regions, constrained by limited resources, have a more urgent need to improve resource utilization efficiency and advance green transformation. Constrained by resource endowments, firms may be more inclined to absorb and transform the digital knowledge spillovers of upstream firms, effectively integrating digital technologies into production processes and management practices to enhance production efficiency and reduce environmental pollution emissions. In contrast, firms in resource-rich regions tend to prioritize the development of resource-intensive industries, which often results in resource dependency, lower resource utilization efficiency, and insufficient innovation incentives. Consequently, these firms are less affected by the digital spillover effects of supplier firms.

Table 9.

Heterogeneity analysis: discussions based on resource endowment and supply chain dependency intensity.

Variable	High Resource Endowment Regions	Low Resource Endowment Regions	Strong Supply Chain Dependence	Weak Supply Chain Dependence
	(1)	(2)	(3)	(4)
Updigi_supp	− 0.0035	− 0.0268***	− 0.0182***	0.0128
	(0.0248)	(0.0064)	(0.0069)	(0.0563)
_cons	0.1206	0.5452	0.7826	− 0.1605
	(0.7695)	(0.5978)	(0.6438)	(1.2401)
Controls	YES	YES	YES	YES
Firm fixed	YES	YES	YES	YES
Year fixed	YES	YES	YES	YES
R2	0.7285	0.7759	0.7338	0.7543
Obs	381	506	549	338

Heterogeneity in supply chain dependence

Upstream and downstream firms in the supply chain network are interdependent, collaborative, and resource-complementary stakeholders. Differences in the degree of dependence may lead to varying environmental spillover effects of supplier firms’ digital transformation on customer firms. Specifically, customer firms with high supply chain dependence rely heavily on supplier firms for production and resource management. As a result, the demand for supplier firms’ digital transformation is more likely to propagate along the supply chain to these firms. To ensure the stability of long-term cooperative relationships, these firms are more inclined to actively promote digital transformation and adopt supplier firms’ digital environmental protection technologies and governance standards, thereby effectively improving environmental pollution conditions. In contrast, customer firms with low supply chain dependence are less reliant on supplier firms in production decisions. Their green transformation needs are more autonomous, based on their development strategies and organizational management models, making them less influenced by supplier firms’ digital transformation. This study uses the proportion of a firm’s sales to its customers relative to its annual sales as a measure of supply chain dependence and divides the sample into two groups based on the mean value: high and low supply chain dependence. This measure aligns with mainstream research practice on supply chain concentration, which captures buyer or supplier reliance intensity through transaction ratios, especially in manufacturing and logistics contexts⁵⁴. Columns (3) and (4) of Table 9 show that the digital transformation of listed companies significantly promotes pollution reduction for customer firms with high supply chain dependence, but has no significant effect on those with low supply chain dependence. This suggests that supplier firms’ digital transformation has a more pronounced pollution reduction effect on customer firms with high supply chain dependence. This pattern is consistent with evidence that higher supply chain dependence enhances the coordination, visibility, and environmental responsiveness of digital initiatives within interfirm networks, reinforcing their carbon reduction outcomes⁵⁵.

Heterogeneity in industry pollution intensity

Significant differences in pollution intensity and energy consumption across industries may lead to varying effects of supplier firms’ digital transformation on the environmental pollution emissions of customer firms in different industries. Following Pan Ailing et al. ⁵⁶, this study categorizes customer firms into two groups—heavy-pollution industries and non-heavy-pollution industries—based on industry pollution intensity to analyze the heterogeneity in the impact of supplier firms’ digital transformation on environmental pollution emissions across industries. This paper classifies the following industries as heavily polluting industries: coal mining and washing, oil and natural gas extraction, petroleum processing, coking and nuclear fuel processing, electricity and heat production and supply, ferrous metal mining, ferrous metal smelting and rolling processing, non-ferrous metal mining, non-ferrous metal smelting and rolling processing, chemical raw materials and chemical products manufacturing, chemical fiber manufacturing, paper and paper product manufacturing, textile manufacturing, leather, fur, feather, related products, and footwear manufacturing, non-metallic mineral products manufacturing, and rubber and plastic products manufacturing. All other industries are classified as non-heavily polluting industries. According to the regression results in columns (1) and (2) of Table 10, the coefficient of supplier firms’ digital transformation is significantly negative in non-heavy-pollution industries but not significant in heavy-pollution industries. This indicates that the digital transformation of supplier firms has a more pronounced effect on promoting pollution reduction in customer firms within non-heavy-pollution industries. This phenomenon may be attributed to the higher environmental sensitivity and stronger transformation willingness of customer firms in non-heavy-pollution industries⁵⁷. These firms are more willing to absorb and effectively apply supplier firms’ digital technology outcomes to optimize production processes and improve environmental performance. In heavy-pollution industries, customer firms’ motivations for pollution reduction and green development are primarily tied to substantial improvements in green production technology. However, these firms face higher cost pressures in technology upgrades and pollution control, which slows their response to supplier firms’ digital transformation⁵⁸. Consequently, the environmental spillover effects of supplier firms’ digital transformation are not evident in the short term.

Table 10.

Heterogeneity analysis: discussions based on industry pollution intensity and firm ownership structure.

Variable	High Pollution Industry Firms	Non-High Pollution Industry Firms	State-Owned Enterprises	Non-State-Owned Enterprises
	(1)	(2)	(3)	(4)
Updigi_supp	0.1136	− 0.0194***	− 0.0623	− 0.0271***
	(0.1178)	(0.0065)	(0.0626)	(0.0066)
_cons	1.0770	0.6911	1.1804*	0.3229
	(0.8145)	(0.7031)	(0.6967)	(0.7389)
Controls	YES	YES	YES	YES
Firm fixed	YES	YES	YES	YES
Year fixed	YES	YES	YES	YES
R2	0.8095	0.7194	0.8193	0.6634
Obs	240	647	413	474

Heterogeneity in ownership structure

Significant differences in goal orientation and governance structure between state-owned enterprises (SOEs) and non-state-owned enterprises (non-SOEs) necessitate a detailed examination of the heterogeneous effects of supplier firms’ digital transformation on the environmental pollution emissions of customer firms with different ownership structures. Accordingly, this study divides customer firms into two categories—SOEs and non-SOEs—based on ownership structure to examine the differential impact of supplier firms’ digital transformation on the pollution emissions of these firms. The regression results in columns (3) and (4) of Table 10 indicate that the coefficient of supplier firms’ digital transformation is significantly negative for non-SOEs, while it is negative but not significant for SOEs. This implies that the forward spillover effects of supplier firms’ digital transformation have a more significant impact on promoting pollution reduction in non-SOEs. A possible explanation is as follows. First, non-SOEs are typically more profit-oriented and subject to market discipline, and therefore have stronger incentives to adopt supplier-driven process or input improvements that lower costs and reduce emissions⁵⁹. By contrast, SOE managers often operate under multi-pronged mandates that include administrative objectives in addition to commercial targets, and these multiple mandates can dilute managerial incentives and weaken the rapid adoption of supplier innovations⁶⁰. Second, SOEs frequently face more complex procurement and compliance procedures—longer procurement chains, stricter internal approval requirements, and formal supplier selection rules—which can delay or blunt the transmission of technological and material spillovers from suppliers. Non-SOEs, with greater managerial autonomy and simpler procurement arrangements, are better positioned to integrate new supplier technologies or cleaner inputs into production processes on shorter notice⁶¹. Third, non-SOEs face tighter budget constraints and clearer cost-benefit signals, which make investments in adopting supplier innovations that generate both cost savings and emissions reductions more attractive. SOEs, by contrast, often operate under softer budgetary constraints and benefit from regulatory protections, reducing the urgency to internalize supplier innovations that have longer-term or more uncertain payoffs⁶².

Conclusions and future perspectives

Key findings and policy implications

Digital transformation is a “must-answer question” for firms in the era of the digital economy and a crucial lever for driving supply chain network innovation and accelerating the construction of a modern industrial system. The strong penetrability and high synergy of digital technologies not only enhance the connectivity of upstream and downstream firms in the supply chain but also create favorable conditions for digital transformation to generate external spillovers through supply chain linkages. Whether supplier firms’ digital transformation influences the green behavior decisions of customer firms is the primary research question of this paper. Based on this premise, this study examines the impact of supplier firms’ digital transformation on the environmental pollution emissions of customer firms and explores the underlying mechanisms from the perspective of vertical supply chain linkages. Our findings are threefold.

First, supplier digital transformation significantly suppresses customer firms’ environmental pollution emissions, with the effect being asymmetric: it is notably stronger when suppliers’ digitalization levels exceed those of their customers. This conditional reduction mechanism reconciles conflicting perspectives in the literature²⁴.

Second, mechanism tests reveal that supplier digitalization reduces pollution emissions through dual channels: (1) enhancing green technological innovation and (2) amplifying external stakeholder oversight. By integrating internal supply-chain dynamics with external governance forces, our study unpacks the “black box” of digital-driven sustainability²⁷.

Third, heterogeneity analysis identifies four boundary conditions under which the pollution-reduction effect is most pronounced: low-resource regions, high supply-chain dependence, non-heavy-pollution industries, and non-state-owned enterprises. These findings expand the theoretical and empirical dimensions of digitalization’s environmental benefits. The findings of this study offer the following policy implications:

First, optimize pollution-reduction pathways by emphasizing both endogenous green governance capabilities and supplier collaboration as two focal points. On the one hand, firms should enhance their internal environmental governance capacity by increasing investments in green technology R&D, adopting advanced digital environmental-protection technologies and equipment, and optimizing production processes to reduce pollution emissions at the source. On the other hand, regulators can design explicit incentives for digital “leaders” to share technologies with downstream partners—for example, by offering preferential treatment in public procurement for suppliers that commit to providing interoperable solutions to customers and by supporting pooled licensing arrangements that lower adoption costs for downstream firms. Customer firms should also promote pollution reduction through collaborative supply chain models and strengthen digital linkages with suppliers. Leveraging intelligent supply-chain management platforms, customer firms should actively learn from supplier firms’ digital resources and green technologies, dynamically adjust strategies, and comprehensively optimize pollution reduction-pathways.

Second, advance differentiated implementation of supply chain digital-linkage policies aligned with the objectives of China’s carbon-pricing architecture. This study finds that the impact of supplier firms’ digital transformation on customer firms’ environmental pollution emissions varies significantly by ownership structure and supply-chain dependence. Policy responses should therefore be tailored to firm type, region, and industry. For example, non-SOEs can receive grants to support adoption, whereas SOEs may benefit more from governance reforms that accelerate internal modernization. In regions with low resource endowments, policy support should be intensified to facilitate the introduction and implementation of digital technologies. For highly dependent supply chains, policymakers can promote collaborative-governance platforms to encourage joint investments in shared digital infrastructure. Policymakers can also pilot programs linking participation in digital sharing to preferential treatment under China’s ETS. For instance, regulators could offer temporary compliance-cost relief to firms that demonstrably lower supply chain emission intensity through interoperable digital solutions.

Third, fully leverage the supervisory and incentive roles of external market actors to promote collaborative green transformation of supply chains. Establish a more comprehensive green finance support system and a standardized environmental information disclosure mechanism to strengthen the role of external market actors in driving supply chain greening and to provide robust support for firms’ sustainable development. For example, regulators and market actors can expand instruments such as green bonds earmarked for investments in interoperable digital platforms and clean technologies, as well as concessional green credit lines that lower borrowing costs for customer firms adopting supplier-shared technologies. Meanwhile, firms should accelerate the deep integration of digital technologies with environmental protection, improve their environmental governance information systems, proactively disclose environmental governance measures and outcomes, respond actively to market supervision and public concerns, and enhance ecological collaboration within the supply-chain network to secure long-term competitive advantages through green competitiveness.

Limitations and future research directions

As an initial exploration of the environmental spillover effects of supplier digitalization in supply chains, this study prioritizes static mechanisms and a single institutional context—China—to ensure analytical tractability. Future research could enrich this line of inquiry in two key directions. First, dynamic frameworks such as computable general equilibrium models may rigorously quantify how digital transformation reconfigures market dynamics, including price elasticity adjustments and profit redistribution, and environmental outcomes over time. Second, extending the empirical scope to cross-country settings, particularly in regions with divergent carbon pricing regimes such as the EU’ s Carbon Border Adjustment Mechanism and Southeast Asia’ s fragmented systems, could uncover how institutional heterogeneity shapes the efficacy of digital-environmental synergies. Such efforts would strengthen the policy relevance of findings and contribute to global strategies for aligning Industry 4.0 transitions with climate governance goals.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

J.L. and H.C. designed the study, developed the methodology, and registered the initial protocol. S.S. managed the data collection and statistical analysis and prepared the first draft of the study. X.X. supervised the project. All authors contributed to the preparation of the manuscript, revised it critically, and approved the final draft.

Funding

This work was supported by National Natural Science Fund of China (Grant Numbers: 72472090).

Data availability

The datasets used or analysed during the current study available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.