Digitalization and Supply Chain Carbon Performance: The Role of Focal Firms

Abstract

This study explores how digitalization by focal firms affects carbon performance across the entire supply chain, advancing the literature by adopting a holistic supply chain perspective rather than a single-firm lens. We further draw on dynamic capability theory to explain the mechanisms through which digitalization enhances supply chain carbon performance. Based on an unbalanced panel dataset of Chinese listed firms from 2008 to 2022, we construct a three-tier supply chain panel linking upstream, focal, and downstream firms. The benchmark regression results show that focal firm digitalization significantly enhances overall supply chain carbon performance. Mechanism analyses identify two critical transmission channels: (1) optimizing supply chain resource allocation efficiency, through improved inventory turnover and strengthened supply chain finance; and (2) enabling collaborative technological upgrading, by enhancing the total factor productivity of upstream and downstream partners. Further heterogeneity analysis reveals that the effect of digitalization on improving carbon performance is more pronounced in regions with stronger environmental regulation and in non-regulated industries. In addition, we analyze the moderating role of the Supply-Chain Pilot-Cities Program. The findings provide practical insights for firm managers globally seeking to leverage digitalization for supply chain decarbonization and for policymakers across jurisdictions aiming to design supportive mechanisms that facilitate digital and green integration.

1. Introduction

Under the global “dual carbon” goals of carbon peaking and carbon neutrality, improving supply chain carbon performance and promoting green, low-carbon supply chains are crucial. These efforts are not only important steps toward achieving climate targets but also key components of the broader green transition of the economy and society. A supply chain involves all activities, information, resources, and financial flows among organizations, from raw materials to end consumers. It includes suppliers, manufacturers, distributors, retailers, and final customers; it emphasizes coordination coordination, collaboration, and value co-creation among firms [1]. The rapid rise of e-commerce and digital trade has made modern supply chains increasingly complex, interconnected, and data-intensive. Firms now operate within multilayered, dynamic networks, making carbon emissions a network-level issue rather than a firm-level one. Increasingly, carbon regulations and trade mechanisms require full supply chain carbon disclosure as a condition for market access and competitiveness [2]. This shift compels firms to not only manage their own emissions but also collaborate with upstream and downstream partners in joint decarbonization efforts.

Carbon reduction at the individual firm level often has limited effectiveness. On the one hand, a single firm’s emissions usually represent only a small part of the full product life cycle. Efforts that focus solely on internal emissions may ignore upstream and downstream sources, or even shift carbon-intensive activities elsewhere, which fails to reduce total emissions [3,4]. On the other hand, low-carbon technology investments by individual firms can be expensive and lack economies of scale.

In contrast, coordinated emission reduction across the supply chain offers clear advantages. It enables full-process carbon optimization—from sourcing and production to logistics and consumption—avoiding the problem of local optimization. It also supports resource sharing and joint technology efforts. Focal firms can integrate carbon data, green technologies, and financial resources across upstream and downstream partners. This can help small and medium-sized enterprises (SMEs) participate and reduce their costs.

Existing studies have widely confirmed the positive impact of digitalization at the firm level, including improvements in production efficiency [5], innovation capability, and corporate social responsibility [6]. Research has also extended to environmental outcomes, such as reductions in carbon emissions and the promotion of green technological innovation [7]. Meanwhile, other studies have examined how policies such as urbanization strategies [8], manufacturing agglomeration [9], carbon tax [10], and carbon emissions trading schemes affect regional carbon performance [11]. However, the majority of existing research remains confined to the firm or regional level, lacking a systematic identification of the supply chain as a structurally integrated and interdependent source of carbon emissions [12].

This study adopts dynamic capability theory as the theoretical foundation to examine how focal firms leverage digitalization to sense environmental transition pressures, seize digital transformation opportunities, and reconfigure resource allocation and technological paths across the supply chain to enhance carbon performance [13]. As central hubs within supply chains, focal firms exhibit notable efficiency advantages in digital transformation. They often control the critical flows of information, materials, and capital, enabling them to more rapidly perceive internal and external changes and respond through operational adjustments [14].

The mechanisms through which focal firm digitalization affects supply chain carbon performance have yet to be systematically identified. In particular, there exists a clear theoretical gap and lack of empirical validation regarding meso-level pathways such as improvements in resource allocation efficiency and collaborative technological upgrading. Meanwhile, the role of the policy environment—an important external determinant of firms’ green transformation—has not been empirically examined as a moderator in the digitalization–carbon performance relationship. Against the backdrop of China’s accelerated rollout of supply chain innovation pilots and supply chain finance reforms, identifying how digitalization performs under varying policy contexts holds substantial practical relevance.

To examine how focal firm digitalization affects the overall carbon performance of the supply chain, this study further explores the mediating mechanisms and the moderating effects of policy heterogeneity. Taking a holistic supply chain perspective and drawing on the dynamic capabilities theory, the analysis centers on the digitalization of focal firms and draws on data from Chinese listed companies and their major suppliers and customers from 2008 to 2022. A three-tier supply chain dataset was constructed by matching firm-year observations based on year and stock code, resulting in “upstream–focal firm–downstream–year” linkages. Using this dataset, this study conducts an empirical analysis based on an unbalanced panel fixed effects model to address the following three research questions:

Q1: What is the effect of focal firm digitalization on supply chain carbon performance?

Q2: To what extent do resource allocation optimization and collaborative technological upgrading mediate the relationship between focal firm digitalization and supply chain carbon performance?

Q3: To what extent does the external policy environment moderate the relationship between focal firm digitalization and supply chain carbon performance?

This study aims to provide theoretical support for understanding the mediating mechanisms through which digitalization improves supply chain carbon performance. It also offers empirical evidence to inform government initiatives on supply chain digitalization, digital empowerment of small and medium-sized enterprises (SMEs), and carbon governance policies. Compared with existing literature, this paper makes several theoretical contributions:

First, this study shifts the focus from individual firms to the supply chain. It addresses the limitation of prior research that could not capture the systemic and interdependent nature of supply chain carbon emissions. Focal firms, as hubs controlling flows of information and resources, are theoretically best placed to drive collaborative carbon reduction. To capture this, we construct a three-tier dataset (upstream–focal–downstream) to provide empirical evidence that focal firms’ digitalization significantly improves supply chain carbon performance.

Second, this study innovatively introduces dynamic capability theory. Unlike prior studies that stayed at the firm level, this study moves beyond to examine how focal firms’ digitalization builds supply chain-wide capabilities for improving carbon performance. Two key mechanisms are identified: (1) improving resource allocation efficiency across the supply chain and (2) enabling collaborative technological upgrading with upstream and downstream partners. In doing so, this study extends supply chain management research by framing digitalization as a capability-building process.

Third, this study incorporates external policy contexts, including environmental regulation, supply-chain pilot programs, and industry regulatory characteristics, into the empirical framework. It reveals significant heterogeneity and moderating effects in the relationship between digitalization and carbon performance, enriching theoretical explanations of how firms interact with policy environments.

On a practical level, this study provides empirical support for firms to enhance supply chain carbon performance through digitalization, offering guidance on developing dynamic capabilities and improving inter-organizational coordination for green transition. It also informs policymakers seeking to integrate digitalization with green governance and to strengthen low-carbon supply chain mechanisms.

2. Literature Review and Theoretical Analysis

2.1. Carbon Performance

As global climate challenges intensify, reducing carbon emissions has become a widely shared international commitment. Carbon performance, which captures the relationship between carbon emissions and economic activities, has received extensive attention in the academic literature. Improving carbon performance is viewed as a critical approach to addressing the current carbon challenge. Existing studies have primarily examined the determinants of carbon performance from either the micro (firm-level) or macro (regional or city-level) perspectives [15]. A substantial body of research focuses on how firms can improve their carbon performance. Technological innovation is widely recognized as a core driver [13], while optimizing production processes and management practices has also been shown to significantly enhance firm-level carbon performance [16]. Other studies emphasize the role of policy instruments, such as carbon trading schemes and carbon taxes, in improving regional carbon performance. Factors such as green finance development [17] and industrial structure upgrading have also been found to contribute to regional-level improvements [18].

However, most of these studies adopt a single-entity perspective, either focusing on firms or local governments, without considering the broader networked nature of carbon emissions. With the emergence of supply chain perspectives, carbon emissions are increasingly recognized as a systemic issue embedded across interconnected production and logistics networks [19]. A firm’s carbon performance is not only shaped by its own operations, but also by its upstream and downstream partners, coordination mechanisms, and supply chain-wide data transparency. Some researchers note that existing research rarely adopts a supply chain-wide perspective to systematically identify the drivers of carbon performance [20]. In particular, the mechanisms through which supply chain digitalization affects carbon performance remain underexplored, leaving a significant gap in both theory and empirical evidence. This study addresses this gap by examining how focal firm digitalization affects overall supply chain carbon performance, thus contributing to a more holistic understanding of digital green transformation.

2.2. Firm Digitalization

Digitalization has become a key driver of organizational transformation and business process reengineering. As an essential trend in industrial upgrading, firm-level digitalization has been extensively studied, leading to a relatively mature analytical framework concerning its internal effects.

A large body of empirical research has shown that digitalization significantly enhances a firm’s innovation capability [21,22], operational management [23], production performance [24], energy efficiency [25], and market competitiveness [26]. In recent years, emerging studies have begun to explore the externalities of firm digitalization. Some scholars argue that digitalization may influence upstream and downstream firms through mechanisms such as technology diffusion and platform-based coordination.

Leveraging such spillover effects, the digitalization of focal firms—through the adoption of supply chain big data platforms, digital supply chain management systems, and other technologies—enables real-time data access and feedback along the supply chain. These capabilities allow focal firms to monitor changes in demand, inventory, and logistics in real time, thereby mitigating the bullwhip effect [27], reducing energy waste caused by information delays, and ultimately improving supply chain carbon performance. In addition, digitalization enhances the focal firms’ innovation capacity. This improvement can spill over to upstream and downstream partners via innovation diffusion, information sharing, competitive pressure, and structural incentives [28], thereby elevating the overall technological level of the supply chain and improving its carbon performance at a more fundamental level.

Based on this, the following hypothesis is proposed:

H1: 

Focal firm digitalization improves supply chain carbon performance.

2.3. Dynamic Capabilities Theory

The theory of dynamic capabilities was proposed by David Teece, Gary Pisano, and Amy Shuen in 1997 and has since become highly influential in the field of strategic management [29]. It refers to a firm’s ability to integrate, build, and reconfigure internal and external competencies to address rapidly changing environments and maintain competitive advantage. This framework provides a coherent framework for understanding how firms dynamically manage transformation.

Relevant literature in this domain can be broadly categorized into two streams: one focuses on the intrinsic composition and mechanisms of dynamic capabilities, while the other applies this theory to solve concrete managerial problems. For example, A part of scholars emphasize the interaction between dynamic capabilities, creativity, and innovation [30]. Another part of scholars through case studies in crisis settings, demonstrate the central role of corporate headquarters in adaptive responses to external shocks [31]. Some researchers developed a multi-dimensional measure of supply chain resilience based on the dynamic capabilities framework, showing how firms enhance adaptability and performance in the face of disruptions [32]. More recently, the literature has extended the scope of dynamic capabilities into networked contexts, emphasizing their interorganizational manifestations. Dynamic capabilities are increasingly understood as mechanisms that allow firms to leverage interfirm relationships, promote knowledge sharing, and build modular platforms that reduce governance frictions and enable coordinated adaptation across organizational boundaries. This perspective highlights that dynamic capabilities are not confined within firm boundaries but can be mobilized in networked supply chain settings to enhance collective adaptability and performance [33].

Against the backdrop of rapid digital advancement and tightening carbon constraints today, firms operate in an increasingly dynamic and complex external environment. They must leverage digitalization to transform operational models and improve carbon performance across the entire supply chain in order to stay competitive and compliant. Enhancing supply chain carbon performance is an ongoing, dynamic process. Digital technologies not only enhance firms’ sensing capabilities—allowing timely identification of carbon-related pressures and green transition opportunities—but also provide the enabling foundation for seizing and reconfiguring resources [29].

Recent studies have increasingly integrated the dynamic capabilities perspective with digital transformation. Researchers argue that firms must develop sensing and reconfiguring capabilities to drive deep changes in organizational structures and business models. Other studies also link dynamic capabilities to corporate sustainability. Research has found that dynamic capabilities facilitate environmentally friendly innovation, enhance corporate social responsibility, and support green competitiveness [34].

Building on this theoretical foundation, this study views the digital transformation of focal firms—and their continuous process of sensing, seizing, and reconfiguring resources to improve supply chain carbon performance—as a manifestation of dynamic capability development and deployment. Through the application of digital technologies and platform-based management, focal firms continuously reconfigure resource allocation and upgrade technological capabilities in response to changing internal and external environments, thereby driving sustainable carbon performance improvement across the supply chain and securing long-term competitive advantage. Based on this, we propose two mediating mechanisms: optimization of resource allocation and collaborative technological upgrading. Furthermore, as dynamic capabilities are inherently shaped by external environments, we also examine how different institutional settings influence the impact of focal firm digitalization on supply chain carbon performance.

2.4. Internal Mechanisms

2.4.1. Digitalization and Resource Allocation

Resource allocation efficiency acts as a crucial channel through which firm digitalization influences carbon performance, particularly in the context of multi-tier supply chain coordination. Prior research has shown that digital tools—by enabling process integration, real-time monitoring, and data-driven forecasting—can effectively alleviate problems such as information asymmetry and response delays, thereby improving the efficiency of factor flows and overall system-level resource allocation [35]. Higher levels of digitalization facilitate faster information exchange and responsiveness, which in turn enhance sustainable operational performance [36].

Empirical evidence further confirms the impact of improved resource allocation on carbon emissions. At the macro level, misallocation of capital and energy across Chinese regions has been found to significantly hinder carbon efficiency. Enhancing allocation efficiency not only improves productivity but also reduces resource waste and redundant output, thereby lowering carbon emissions per unit of output [37]. At the firm level, mismatches in internal energy use and excess capital accumulation are often major drivers of carbon intensity [38]. Digital management and supply chain coordination systems can help identify low-marginal-efficiency production segments and reallocate resources accordingly, resulting in substantial reductions in carbon intensity.

In supply chain environments, misallocation exists not only within firms but also across upstream and downstream partners. Many supply chains lack flexibility due to rigid logistics networks, fixed costs, and structural inertia. These constraints limit capacity adjustments in response to demand shocks or external disruptions, leading to energy waste and increased carbon emissions during production and transportation. As a result, coordinated optimization of resource allocation across the supply chain is particularly critical [39].

From the financial perspective, firm digitalization can enhance resource access for upstream and downstream firms by supporting the development of supply chain finance (SCF) platforms and dynamic credit models [40]. These tools help address traditional mismatches in capital allocation. Empirical findings show that digitalization improves firms’ energy efficiency and green transformation capabilities by promoting green technology innovation and optimized resource scheduling.

Recent studies have highlighted that supply chain finance has become a vital channel for optimizing supply chain-level resource allocation. It improves capital flow efficiency and provides crucial financial support for low-carbon upgrades among supply chain partners. For instance, SCF has been shown to significantly reduce firm-level carbon emissions by easing financing constraints and enabling investment in emission reduction technologies [41]. A game-theoretic study finds that, under carbon cap and capital constraints, hybrid financing schemes—combining bank loans and accounts receivable financing—can simultaneously enhance emission reduction efficiency and improve supply chain profits, thus supporting the implementation of green projects [42].

Nevertheless, the literature seldom investigates this as an integrated causal chain linking focal firm digitalization, resource allocation reconfiguration, and supply chain-wide decarbonization—a gap this study aims to address. Therefore, this study proposes the following hypothesis:

H2: 

Focal firm digitalization improves supply chain carbon performance by optimizing supply chain resource allocation.

2.4.2. Digitalization and Collaborative Technological Upgrading

Technological upgrading is a core driver of carbon reduction and high-quality green development. Total factor productivity (TFP) is widely regarded in the economics literature as a technology-related component of performance, reflecting efficiency gains from technology adoption and innovation [43]. Accordingly, TFP serves as an appropriate indicator of technological upgrading in the supply chain context. In particular, improvements in total factor productivity (TFP) can significantly reduce carbon emissions per unit of output [44]. A higher TFP typically reflects better resource utilization and lower energy consumption, thus contributing to lower carbon intensity [45]. Additional studies have shown that productivity gains are often accompanied by clean technology adoption and process innovation, which not only reduce carbon emissions but also help control other pollutants [46], especially in energy-intensive manufacturing sectors. According to the green technological change framework, productivity improvements—when guided by supportive policy—can induce directed green innovation, enabling compatibility between economic growth and environmental goals [47]. These findings emphasize the importance of productivity-driven innovation in achieving low-carbon outcomes.

A growing body of literature has also highlighted the role of digitalization in enhancing firms’ technological upgrading and TFP. On one hand, digital tools such as smart equipment and data systems improve production efficiency and managerial capabilities [48]. On the other hand, digitalization is often accompanied by organizational restructuring and strategic innovation, which promote product development and technological advancement [49,50].

Moreover, the construction of digital platforms has strengthened firms’ capacity to absorb external knowledge and created synergistic effects within innovation networks [51]. In this process, digitalization not only improves a firm’s own productivity but also imposes digital upgrading pressure on upstream and downstream partners through platform interfaces, data standards, and coordination mechanisms—resulting in technology spillovers across the supply chain [52]. Research shows that focal firms, while advancing their own digitalization, often transfer advanced managerial practices and technological capabilities to their supply chain partners via platforms, data systems, and collaborative tools [53]. Especially in platform-led ecosystems, digital platforms have become critical infrastructure for enabling collaborative technological upgrading and efficiency improvement throughout the value chain. Empirical studies further demonstrate that these externalities can increase supplier patenting and TFP, thereby strengthening the technological base of the entire supply chain.

Recent research has begun to empirically capture this chain-based upgrading mechanism. However, a systematic identification of the mediating effect model and the quantitative pathway linking digitalization, collaborative technological upgrading among upstream and downstream firms, and carbon performance remains largely absent in the current literature. This gap provides important research space for the mechanism framework developed in this study.

Accordingly, we propose the following hypothesis:

H3: 

Focal firm digitalization enhances supply chain carbon performance by promoting collaborative technological upgrading across the supply chain.

2.5. External Environment

Dynamic capabilities theory emphasizes a firm’s ability to adapt to changing external environments. Likewise, carbon performance is not only driven by internal factors such as resource allocation and technological upgrading, but also strongly shaped by external policy contexts [54]. Therefore, it is crucial to examine how different external environments condition the dynamic capability process through which focal firm digitalization enhances supply chain carbon performance. A wide range of macro-level policy instruments—such as environmental regulations, carbon trading markets, green transition pilot programs, and coal city transformation policies—have been empirically shown to exert significant influence on firm-level carbon outcomes. Moderately stringent environmental regulations can incentivize firms to engage in green technological innovation, thereby improving environmental performance [55]. As a market-based regulatory tool, the carbon emissions trading system (ETS) has been found to promote emission reductions by encouraging energy structure optimization and low-carbon technology adoption among firms [56]. In addition, carbon-related pilot policies—such as low-carbon city initiatives and supply chain innovation application pilots—help reduce the cost of green transitions and enhance firms’ willingness to adopt green technologies, ultimately improving regional carbon performance [57].

Furthermore, macroeconomic policies such as urbanization [58], green finance initiatives, and industrial structure adjustments are increasingly recognized as important drivers of emission reduction and environmental outcomes.

While these studies have greatly enriched our understanding of carbon performance from multiple policy dimensions, few have systematically explored how digitalization performs under different external policy environments across the supply chain. To address this gap, this study empirically investigates how institutional variations—including the stringency of local environmental regulation, industry regulatory status, and participation in supply-chain pilot policies—influence the effectiveness of focal firm digitalization in improving supply chain carbon outcomes.

Accordingly, the following hypotheses are proposed:

H4: 

The effect of focal firm digitalization on supply chain carbon performance varies across regions with different levels of environmental regulation stringency.

H5: 

The effect of focal firm digitalization on supply chain carbon performance differs between regulated and non-regulated industries.

H6: 

The effect of focal firm digitalization on supply chain carbon performance is strengthened by the implementation of supply chain innovation pilot policies.

The theoretical mechanism model of this study is illustrated in Figure 1.

3. Research Design

3.1. Sample Selection and Data Processing

This study uses A-share-listed firms in China from 2008 to 2022 as the research sample. Following the method of Chu (2018) [59], we match each listed firm with its top five suppliers and top five customers based on firm names and stock codes provided in the CSMAR database. We exclude non-listed firms and subsidiaries of listed firms among the suppliers and customers. We also remove observations for ST and *ST firms, companies in financial or insurance industries (which lack carbon emissions data), and firms with missing or clearly erroneous values in key financial indicators such as operating revenue, operating cost, founding year, and total profit.

Using firm-year information based on stock codes, we construct a three-tier supply chain dataset with “upstream–focal firm–downstream” linkages across years. We integrate four main data sources into the constructed supply chain sample. (1) Financial data (2008–2022): this includes total assets, operating revenue, operating costs, operating profit, cash flow, and other financial indicators from the CSMAR database. (2) Patent data (2008–2022): We collect invention patent information for listed firms, including patent numbers, IPC classifications, textual descriptions, and application dates. These data are obtained from the China National Intellectual Property Administration and the China Research Data Service Platform. (3) The China Statistical Yearbook (2008–2022), which is used to supplement regional-level economic and environmental variables. (4) Annual report data (2008–2022): this includes listed firms’ operational information and full-text annual reports, sourced from publicly disclosed documents.

Finally, all datasets are merged based on the firm-year identifiers for upstream firms, focal firms, and downstream firms. The final panel dataset contains 3070 annual observations of complete three-tier supply chain linkages.

3.2. Model Specification and Variable Definitions

To examine the impact of focal firm digitalization on supply chain carbon performance, this study conducts empirical analysis using an unbalanced panel fixed effects model. The benchmark regression model is specified as Equation (1):

$${CP}_{i,t}+{\alpha }_0+{\alpha }_1·{Digital}_{j,t}+{\alpha }_2·{Control}_{jt}+{Firm}_j+{Year}_t+\epsilon$$

(1)

where $i$ denotes the supply chain, $j$ the focal firm, and $t$ the year. The dependent variable, $CP,$ represents supply chain carbon performance; the core explanatory variable, $digital,$ measures the degree of digitalization of the focal firm; $control$ denotes a set of control variables. $Firm$ and $year$ are firm and year fixed effects, respectively, and $\epsilon$ is the error term. The coefficient ${\alpha }_1$ is of primary interest. A significantly positive ${\alpha }_1$ indicates that focal firm digitalization improves supply chain carbon performance, supporting the paper’s theoretical expectation. Robust standard errors are clustered at the firm level.

(1) Dependent variable: carbon performance

The dependent variable is supply chain carbon performance (CP). Following Clarkson (2010) [60], we measure carbon performance as the amount of operating revenue generated per 10,000 tons of carbon emissions:

$$\mathrm{Carbon} \mathrm{Performance} = \mathrm{Operating} \mathrm{Revenue}/\mathrm{Total} \mathrm{Carbon} \mathrm{Emissions}$$

Supply chain carbon performance is constructed by aggregating the carbon performance of upstream, focal, and downstream firms. The total supply chain carbon emissions and total revenue are calculated by summing the respective emissions and revenues of these three tiers. When a focal firm is linked to multiple upstream (or downstream) firms, the upstream (or downstream) carbon emissions are calculated as the average across all associated firms. This approach also applies to the calculation of other upstream and downstream variables. The specific calculation formula is shown in Equation (2):

$${CP}_{i,t}={\displaystyle \frac{\frac{\sum _{k=1}^n{Revenue}_{ikt}}{n}+{Revenue}_{j,t}+\frac{\sum _{l=1}^n{Revenue}_{ilt}}{n}}{\frac{\sum _{k=1}^n{Carbon}_{ikt}}{n}+{Carbon}_{j,t}+\frac{\sum _{l=1}^n{Carbon}_{ilt}}{n}}}$$

(2)

where ${Carbon}_{ikt}$ is the carbon emission of upstream supplier $k$ in chain $i$, and ${Carbon}_{ilt}$ is the carbon emission of downstream customer $l$. (The subsequent analysis applies the same computational method to data from multiple upstream and downstream firms.)

Firm-level carbon emissions are estimated following Shen (2019) [61]: The carbon emissions of a firm are estimated by allocating total industry energy consumption proportionally based on the firm’s share of industry operating costs. Specifically, firm-level carbon emissions are calculated as the ratio of the firm’s operating cost to the total operating cost of its industry, multiplied by the industry’s total energy consumption and the corresponding CO₂ emission conversion factor (2.493). (The industry-level energy consumption data from China Statistical Yearbook have already been converted into standardized coal equivalents (measured in 10,000 tons of standard coal, or 10,000 tce) based on the national standard GB/T 2589-2020, and both firm-level and industry-level operating costs are measured in CNY 10,000.)

In addition, this study identifies a subsample of supply chain linkages where upstream, focal, and downstream firms have all disclosed detailed environmental information—such as direct carbon emissions, consumption of various fossil fuels, electricity use, and heat consumption—in their corporate social responsibility (CSR) reports, sustainability reports, or environmental disclosures. Using this enriched dataset, we construct a small sample with more precise carbon emissions data and conduct robustness checks to validate the reliability of the main findings.

(2) Independent variable digitalization

We measure the digitalization level of the focal firm using the natural logarithm of the number of digital technology patent applications plus one. Unlike most existing studies that rely on text-based frequency analysis of digitalization-related keywords in annual reports, we use patent-based indicators to address two issues:

First, keyword-based methods often rely on inconsistent keyword sets, resulting in incomplete identification of digital technologies. Second, keyword mentions do not necessarily reflect actual implementation of digital technologies.

Following Huang (2023) et al. [62], we construct our proxy using the Classification of Digital Economy and Its Core Industries (2021) by the National Bureau of Statistics and the Patent Classification System for Key Digital Technologies (2023) by the China National Intellectual Property Administration. We extract relevant patents based on text and IPC classification matching, then aggregate them at the firm-year level.

To ensure robustness, we also test alternative digitalization measures used in prior literature, including keyword-based indicators, and rerun regressions using these as substitutes.

(3) Control variables

To account for other factors that may influence the effect of digital transformation on supply chain carbon performance, we include both firm-level and region-level control variables. Firm Size; Lev; ROA; TobinQ; Firm Age; AgC2s; Cap1; Cap2; PGDP; ER. In addition, we include firm fixed effects and year fixed effects to control for unobserved heterogeneity and time trends. Detailed definitions are provided in

Table 1. Variable definition.

Type	Variable	Definition
Dependent Variable	Digital	ln (digital technology patent applications + 1)
Independent Variable	CP	Operating Revenue/Total Carbon Emissions
Control Variable	Size	ln (Total Assets)
	Lev	Total Liabilities/Total Assets
	ROA	Return on Assets = Net Profit/Total Assets
	TobinQ	Market Value of Equity + Debt Value)/ Asset Replacement Cost
	FirmAge	ln (Current Year − Establishment Year + 1)
	AgC2	The second type of agency cost
	Cap1	Capital intensity = ln (Total Assets/Number of Employees)
	Cap2	Alternative capital intensity = Total Assets/Operating Revenue
	PGDP	Provincial Nominal GDP per Capita
Mediating Variables	SCE	Supply Chain Efficiency
	SCF	Supply Chain Finance
	TFPLP up	TFP of upstream by LP method
	TFPOP up	TFP of upstream by OP method
	TFPLP down	TFP of downstream by LP method
	TFP OP down	TFP of downstream by OP method

.

(4) Mediating variables

Based on the theoretical framework, we explore two mediating mechanisms: resource allocation optimization and collaborative technological upgrading. To test the resource allocation mechanism, we focus on product resource allocation and financial resource allocation, proxied by the following:

Supply chain efficiency (SCE), measured as the natural logarithm of inventory turnover days of the focal firm, capturing how efficiently inventory resources are utilized. Supply chain finance (SCF), measured using keyword frequency analysis of supply chain finance–related terms in the annual reports of upstream and downstream firms. Following Zhou and Wu (2022) [63], the total frequency count is log-transformed (ln(count + 1)) to construct the SCF index.

To test the collaborative technological upgrading mechanism, we use the total factor productivity (TFP) of upstream and downstream firms as proxies for their technological levels.

Table 2. Descriptive statistics.

Variable	Obs	Mean	Std. Dev.	Min	Max
Digital	3070	3.044	1.973	0	7.791
Digital_2	3070	2.715	1.557	0	6.865
Digital_3	3070	0.049	0.065	0	0.511
CP	3070	4.7137	3.36	0.334	32.944
Size	3070	22.093	1.331	19.277	28.109
Lev	3070	0.443	0.36	0.033	8.612
ROA	3070	0.032	0.101	−1.859	0.37
TobinQ	3070	2.011	1.752	0.744	24.495
FirmAge	3070	2.923	0.34	1.099	3.784
AgC2	3070	0.016	0.031	0	0.337
Cap1	3070	14.504	0.863	12.035	17.854
Cap2	3070	2.584	8.063	0.095	310.094
PGDP	3070	57,272.851	20,201.968	19,858	190,313
SCF	3070	0.339	0.869	0	7.765
SCE	3070	4.334	1.4201	−4.820	8.632
TFP LP up	3070	9.565	1.224	5.913	12.986
TFP OP up	3070	8.782	1.111	5.716	12.242
TFP LP down	3070	9.913	1.08	6.166	12.853
TFP OP down	3070	9.073	0.96	5.672	12.172

reports the descriptive statistics for all key variables used in this study.

4. Results and Analysis

4.1. Benchmark Regression Results

The benchmark regression results are reported in

Table 3. Benchmark regression results.

	(1)	(2)	(3)	(4)	(5)
Variables	CP	CP	CP	CP Mid	CP Up and Down
Digital	0.474 ***	0.475 ***	0.473 ***	0.396 ***	0.493 ***
	(0.049)	(0.049)	(0.049)	(0.088)	(0.051)
Size		−1.140 ***	−1.154 ***	−1.953 ***	−1.130 ***
		(0.265)	(0.265)	(0.479)	(0.279)
Lev		−0.200	−0.198	0.823	−0.207
		(0.309)	(0.310)	(0.561)	(0.327)
ROA		1.306	1.301	7.033 ***	1.315
		(1.044)	(1.048)	(1.899)	(1.106)
FirmAge		0.225	0.209	−1.554	0.285
		(1.206)	(1.212)	(2.197)	(1.279)
AgC2		−1.201	−1.551	−5.416	−1.082
		(2.298)	(2.303)	(4.174)	(2.430)
Cap1		0.120	0.124	0.703	0.157
		(0.248)	(0.248)	(0.450)	(0.262)
Cap2		0.016	0.019	0.167	−0.018
		(0.072)	(0.072)	(0.131)	(0.076)
TobinQ		−0.381 ***	−0.382 ***	−0.554 ***	−0.397 ***
		(0.065)	(0.065)	(0.118)	(0.069)
ER			−8.353	−42.698	−9.068
			(72.066)	(130.580)	(76.038)
PGDP			−0.000 *	−0.000	−0.000 *
			(0.000)	(0.000)	(0.000)
Constant	2.417 ***	24.167 ***	25.111 ***	38.394 ***	24.693 ***
	(0.871)	(5.941)	(5.992)	(10.858)	(6.323)
Observations	3070	3070	3070	3070	3070
R-squared	0.109	0.142	0.144	0.121	0.143
Firm FE	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes

Notes: * p < 0.05, *** p < 0.001. Values in parentheses are t-statistics. CP is carbon performance; digital is firm digitalization.

. Column (1) shows that under a two-way fixed effects model, focal firm digitalization has a significantly positive effect on supply chain carbon performance (α₁ = 0.474, p < 0.01). This coefficient implies that a one-unit increase in digitalization leads to an average improvement of 0.474 units in supply chain carbon performance. Given that the mean value of carbon performance in the sample is 3.36, this increase accounts for approximately 14% of the average level, indicating a substantial practical effect.

Columns (2) and (3) incorporate a full set of firm-level and regional-level control variables. The coefficients remain significantly positive at the 1% level, confirming the robustness of the results and supporting Hypothesis 1 (H1).

Furthermore, to explore whether the impact of focal firm digitalization arises from improvements in its own carbon performance or that of its supply chain partners, we decompose supply chain carbon performance into two components: the carbon performance of the focal firm and the combined carbon performance of upstream and downstream firms. As shown in Columns (4) and (5), digitalization significantly enhances both components, with a stronger marginal effect observed for upstream and downstream firms. This suggests that digital transformation in focal firms plays a systemic role in driving supply chain-wide carbon improvements.

4.2. Robustness Checks

4.2.1. Alternative Measures of the Core Explanatory Variable

In the benchmark model, the level of digitalization of focal firms is measured by the number of digital technology patent applications. To test the robustness of the results, this study adopts several alternative measures commonly used in the literature. First, following Wu et al. (2021) [64], we calculate the frequency of digitalization-related keywords in annual reports, take the sum, add one, and apply a natural logarithm transformation to construct a proxy for firm digitalization. Second, drawing on Zhang (2022) [65], we measure focal firm digitalization as the ratio of digital intangible assets to the net value of total intangible assets. Third, we introduce digital asset investment as another proxy for firm level digitalization that reflects actual deployment rather than invention disclosure. Specifically, we screened firms’ annual reports in the fixed assets and intangible assets sections to identify digital-related items. For each firm-year, we then summed the year-end balances of identified digital fixed assets and digital intangible assets to construct the firm’s total digital asset investment. In addition, we also aggregated the number of digital technology patent applications across all firms within each supply chain to capture the overall level of supply chain digitalization. Considering the potential time lag in the effects of digital transformation, we further used the one-period lagged digitalization level in regression models for robustness testing.

The regression results in

Table 4. Robustness check results.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Variables	CP	CP	CP	CP	CP	CP	CP	CP2	CP
Digital						0.460 ***	0.505 ***	2.183 ***	0.480 ***
						(0.064)	(0.087)	(0.394)	(0.049)
Digital total	0.010 ***
	(0.001)
Digital 2		0.288 ***
		(0.072)
Digital 3			8.304 ***
			(1.906)
Digital 4				0.193 ***
				(0.019)
Digital lag					0.179 **
					(0.071)
Constant	26.543 ***	29.364 ***	30.074 ***	32.716 ***	31.219 ***	13.253 *	35.875 ***	−8.040	30.175 ***
	(5.919)	(6.155)	(6.136)	(5.947)	(6.175)	(6.798)	(12.154)	(16.660)	(7.127)
Obs	3070	3070	3070	3070	3070	1535	1535	1081	3070
R-squared	0.160	0.093	0.095	0.148	0.082	0.145	0.196	0.139	0.173
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Industry FE	No	No	No	No	No	No	No	No	Yes

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001. Values in parentheses are t-statistics. CP is carbon performance; digital is firm digitalization.

Columns (1), (2), (3), and (4) indicate that all three alternative measures of digitalization exhibit significantly positive effects on supply chain carbon performance (p < 0.01). This demonstrates that the positive impact of digitalization on supply chain carbon performance remains robust regardless of the specific measurement approach employed. Moreover, the regression using the one-period lagged digitalization variable also yields a statistically significant result. Notably, the coefficient of the lagged digitalization variable is substantially smaller than that of the contemporaneous measure, suggesting that the effect of firm digitalization on supply chain carbon performance is more pronounced in the same period.

4.2.2. Robustness Check Through Group Regressions by Capital Intensity

A potential concern in our estimation approach is that differences in cost structures across firms may introduce systematic biases in the allocation-based emission estimates. To examine the robustness of our findings, we conducted group regressions by capital–labor intensity. We measured capital intensity as the ratio of total assets to the number of employees and divided the sample into high and low capital-intensity groups. The baseline model was then re-estimated separately for the two groups. Table 4, Columns (6) and (7) report the regression results for the low capital-intensity group and the high capital-intensity group, respectively. The results indicate that digitalization continues to exert a significant positive effect on supply chain carbon performance in both subsamples. This evidence suggests that even if measurement errors exist due to heterogeneity in cost structures, our main conclusion regarding the positive role of digitalization remains robust.

4.2.3. Alternative Measures of the Dependent Variable

In the benchmark regression, firm-level carbon emissions were estimated indirectly based on industry-level total energy consumption, weighted by the firm’s share of industry operating costs. While this approach enables large-scale estimation, it may introduce measurement error. According to the internationally recognized Greenhouse Gas Protocol, corporate carbon emissions are typically categorized into three scopes: direct GHG emissions (Scope 1), indirect emissions from purchased electricity and heat (Scope 2), and other indirect emissions (Scope 3).

To enhance the robustness of the findings, this study identifies a subset of supply chain samples where all upstream, focal, and downstream firms disclosed quantitative information on direct carbon emissions, fossil fuel consumption, electricity use, and heat consumption in their corporate social responsibility (CSR), sustainability, or environmental reports. Based on the official “Corporate GHG Emissions Accounting and Reporting Guidelines” issued by the National Development and Reform Commission (NDRC) of China for different industries, both direct (Scope 1) and indirect (Scope 2) emissions are calculated and summed to obtain total carbon emissions for each firm.

This yielded a refined subsample of 1081 supply chain linkages. We compared the emissions calculated using our industry cost-share allocation method with the directly disclosed firm-level data. The correlation coefficient between the two measures is 0.784, showing a strong positive relationship. Using this more accurate emissions data, we reconstructed the dependent variable and re-estimated the benchmark regression. As shown in Column (8) of Table 4, focal firm digitalization still exhibits a significantly positive effect on supply chain carbon performance, demonstrating the robustness of the main conclusion under alternative, higher-quality carbon emission measures.

4.2.4. Including Industry Fixed Effects

To further rule out the influence of industry-specific unobserved factors, we add industry fixed effects to the regression model. The results in Column (9) of Table 4 indicate that the sign and significance of the digitalization coefficient remain stable, suggesting that industry-level heterogeneity does not bias the estimated relationship between digitalization and supply chain carbon performance.

4.2.5. Addressing Endogeneity: Instrumental Variable Approach

Although the benchmark model includes a comprehensive set of control variables and fixed effects, potential endogeneity issues—such as omitted variable bias and reverse causality—may still persist in the relationship between focal firm digitalization and supply chain carbon performance. To address this concern, we employ an instrumental variable (IV) approach using a two-stage least squares (2SLS) estimation. Following prior studies, we construct an instrument: the interaction between the lagged industry-level digitalization (excluding the focal firm) and the provincial-level broadband internet access per capita.

The industry-average digitalization level in the previous year captures intra-industry learning effects and technology spillovers, while regional broadband access reflects the local digital infrastructure, which significantly influences a firm’s adoption of digital technologies. Their interaction captures the notion that in regions with better digital infrastructure, external industry-level digitalization may more strongly facilitate a firm’s digital transformation—thus ensuring a strong correlation with the explanatory variable.

On the exogeneity front, the average industry digitalization is an aggregate, sector-level measure that, once lagged and purged of the focal firm’s own value, avoids mechanical correlation with the firm’s own digitalization and reduces the risk of direct influence on carbon performance. Likewise, provincial broadband access is primarily driven by regional policy and infrastructure investments, making it unlikely to be determined by any individual firm’s revenue or carbon emissions. After controlling for firm, year, and industry level fixed effects, as well as key firm and regional covariates, the instruments satisfy the core requirements of relevance, exogeneity, and exclusion.

Table 5. SLS regression results.

	(1)	(2)
Variables	1st Stage	2nd Stage
IV	0.024 ***
	(0.010)
Digital		0.902 ***
		(0.089)
Constant	13.589 ***	11.866 ***
	(4.117)	(1.369)
Observations	3070	3070
R-squared	0.139	0.208
KP rk Wald F		386.8
Firm Year FE	Yes	Yes
Industry FE	Yes	Yes

Notes: *** p < 0.001. Values in parentheses are t-statistics.

reports the 2SLS estimation results. In the first-stage regression, the instrument exhibits a strong and significant positive relationship with the explanatory variable (F = 386.8), indicating no concern of weak instruments. In the second-stage regression, the coefficient on focal firm digitalization remains significantly positive at the 1% level, reaffirming the robustness of our main conclusion after addressing endogeneity concerns.

4.2.6. IV-OLS Decomposition Test

To further validate the identification strategy, we adopted the econometric decomposition framework proposed by Ishimaru (2024), which decomposes the gap between IV and OLS estimates into three interpretable components [66]. The intuition is that both IV and OLS coefficients can be expressed as weighted averages of treatment marginal effects, but they differ in how weights are assigned and how endogeneity is handled. Formally, the IV–OLS gap can be decomposed as follows:

$${\beta }_{IV}-{\beta }_{OLS}=∆CW+∆TW+∆ME$$

(3)

$∆CW$ (covariate weight difference) captures how IV and OLS place different weights on observable covariates. In practice, IV tends to overweight groups for which the instrument strongly shifts treatment, while OLS puts more weight on groups with higher variance in the treatment variable.

$∆TW$ (treatment-level weight difference) captures how IV and OLS assign different weights across treatment intensities (e.g., low vs. high digitalization levels). For example, if the instrument disproportionately affects firms at certain treatment margins, the IV coefficients will be more sensitive to those margins, whereas OLS reflects the conditional distribution of treatment in the sample.

$∆ME$ (marginal effect difference/endogeneity bias) compares the causal marginal effects identified by IV with the conditional mean slopes identified by OLS, under IV weighting. The difference arises because digitalization $X$ is correlated with the unobserved error term, $U$. Thus, $∆ME$ isolates the part of the IV–OLS gap that reflects correction of endogeneity bias. As Ishimaru (2024) rigorously shows, $∆ME$ equals the difference between the IV-identified marginal effect $\partial g(x,w)/\partial x$ and the OLS-identified conditional slope $\partial m(x,w)/\partial x$, weighted by the IV weight function. In other words,$\Delta ME$ is the “endogeneity bias component.” More concretely, $\Delta ME$ measures the gap between IV’s causal marginal effect and OLS’s conditional slope under IV weights; their divergence stems precisely from the correlation between treatment $X$ and unobservables U. Therefore, $\Delta ME$ is interpreted as the correction of endogeneity bias.

Our empirical results (Figure 2) show that the increase in the IV coefficient relative to the OLS estimate is almost entirely attributable to $\Delta ME$, while $\Delta CW$ and $\Delta TW$are negligible. This indicates that the IV–OLS gap mainly reflects correction of endogeneity rather than instrument invalidity, thereby reinforcing the robustness of our identification strategy.

4.2.7. Kinky Least Squares (KLS) Test

Traditional instrumental variable methods rely on the exclusion restriction, namely that instruments are strictly exogenous and uncorrelated with the disturbance term. However, this assumption is difficult to verify in econometric practice. To better address potential endogeneity concerns, we adopted the Kinky Least Squares (KLS) method proposed by Kiviet (2020) [67].

The main contribution of the KLS method is that it abandons the strong assumption of perfect exogeneity and instead allows for bounded correlations between endogenous regressors and the disturbance term. Researchers can specify a correlation interval and conduct inference within that interval.

The key idea of KLS is to abandon the strong assumption of perfect exogeneity and instead allow bounded correlations between endogenous regressors and the disturbance term. By introducing a correlation vector ${\rho }_{xu}$, the KLS method explicitly adjusts the OLS estimator for possible endogeneity bias. The method essentially imposes an “endogeneity penalty”: if results remain significant even when moderate correlations between regressors and disturbances are allowed, then the empirical findings can be considered robust.

Formally, the KLS estimator modifies the OLS coefficient as follows:

$$\widehat{\beta }(r,{\widehat{\sigma }}_u)={\widehat{\beta }}_{OLS}-{\widehat{\sigma }}_u{\left(n^{-1}{X}^{\prime }X\right)}^{-1}{S}_{x}r$$

(4)

where r is the assumed correlation vector between regressors and the disturbance, $S_x$ is the sample standard deviation matrix of regressors, and ${\widehat{\sigma }}_u$ is the adjusted disturbance variance estimator. If r equals the true correlation, then $\widehat{\beta }(r,{\widehat{\sigma }}_u)$ is consistent.

The KLS method provides a sensitivity analysis framework based on interval assumptions. Define $C_0(Q,q,\alpha )$ as the set of correlation values under which the null hypothesis $H_0: Q\beta =q$ cannot be rejected at significance level $\alpha$, and $C_1(Q,q,\alpha )$ as its complement. If $C_0$ is wide, conclusions are robust to endogeneity assumptions; if $C_0$ is narrow or empty, inference critically depends on specific assumptions.

Our KLS test results show that even when allowing for correlations between endogenous regressors and the disturbance term, the estimated positive effect of digitalization on supply chain carbon performance remains statistically significant at the 1% level. Figure 3 plots the KLS coefficients and their 95% confidence intervals under varying assumptions about the correlation between endogenous variables and disturbances. The results indicate that even when moderate endogeneity bias is introduced, the estimated effect remains positive and statistically significant.

4.3. Mechanism Analysis

To further uncover the mechanisms through which focal firm digitalization affects overall supply chain carbon performance, we construct mediation models based on the two theoretical pathways derived from the dynamic capabilities perspective: optimizing resource allocation and upgrading technologies through supply chain collaboration.

4.3.1. Mechanism of Optimizing Resource Allocation

Resource misallocation is a common issue in traditional supply chain operations and management, manifested as inefficient material flow, rigid production systems, and financial mismatch. These distortions in resource allocation not only increase operational costs but also contribute to excessive carbon emissions. Digitalization provides a promising solution to these structural bottlenecks by enhancing data acquisition, breaking down information silos, and improving the efficiency of capital flows. These improvements can facilitate better supply–demand matching, coordinated processes, and financing efficiency, ultimately lowering carbon intensity.

We therefore explore two mediation pathways under this mechanism: whether digitalization improves operational efficiency and thereby optimizes physical resource allocation along the supply chain and whether it enhances supply chain finance and mitigates capital misallocation issues among firms.

(1) Improvement in Supply Chain Efficiency

Traditional supply chains suffer from rigidities in information transmission, production response, and logistics coordination—leading to delays, excessive inventories, and redundant transportation. These inefficiencies exacerbate resource waste and increase carbon emissions. Digital technologies can alleviate these rigidities through real-time monitoring, predictive scheduling, and dynamic coordination. Improved inventory management, accelerated response to demand fluctuations, and streamlined logistics reduce both material and energy waste.

We proxy supply chain efficiency (SCE) using the natural logarithm of the focal firm’s inventory turnover days, which reflect the average number of days required to convert inventory into sales. Lower turnover days indicate faster inventory conversion, better management efficiency, and reduced capital lock-up, reflecting greater supply chain agility and responsiveness. Our empirical results show that focal firm digitalization significantly improves SCE. When SCE is included as a mediator, the coefficient on digitalization in the carbon performance regression declines but remains significant, suggesting partial mediation. This supports the hypothesis that digitalization improves carbon performance indirectly by enhancing supply chain efficiency and flexibility.

(2) Optimization of Capital Allocation

Firms, especially SMEs in supply chains, often face financing constraints that hinder investment in green technologies and cleaner production equipment. Digitalization enables the development of supply chain finance (SCF) by establishing data-driven credit profiles, increasing transaction transparency, and facilitating innovative financial services. By reducing credit evaluation costs and improving trade-based financing efficiency, digital tools enable focal firms to extend credit to upstream and downstream partners, facilitating green project financing and addressing capital mismatches.

Following Zhou and Wu (2022) [63], we apply text mining techniques to firm annual reports and compute the frequency of supply chain finance-related keywords. The log-transformed count (plus one) serves as a proxy for SCF. Three-step mediation regressions reveal that digitalization significantly enhances SCF; when SCF is added to the regression, the coefficient on digitalization decreases slightly but remains significant, consistent with a mediation effect. This indicates that focal firm digitalization improves supply chain carbon performance by facilitating access to capital and promoting green investment along the chain. H2 is supported through the comprehensive analysis of the results presented in

Table 6. Mediation mechanism 1: supply chain efficiency (SCE).

	(1)	(2)	(3)
Variables	Step 1:CP	Step 2: SCE	Step 3: CP
Digital	0.473 ***	0.0339 **	0.462 ***
	(0.0485)	(0.0135)	(0.0485)
SCE			0.319 ***
			(0.100)
Constant	25.11 ***	−2.903 *	26.04 ***
	(5.992)	(1.663)	(5.978)
Observations	3070	3070	3070
R-squared	0.144	0.044	0.151
Controls	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes
Year FE	Yes	Yes	Yes

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001. Values in parentheses are t-statistics.

and

Table 7. Mediation mechanism 2: supply chain finance (SCF).

	(1)	(2)	(3)
Variables	Step 1:CP	Step 2: SCF	Step 3: CP
Digital	0.473 ***	0.0433 ***	0.465 ***
	(0.0485)	(0.0166)	(0.0486)
SCF			0.174 **
			(0.0814)
Constant	25.11 ***	7.668 ***	23.78 ***
	(5.992)	(2.052)	(6.017)
Observations	3070	3070	3070
R-squared	0.144	0.088	0.147
Controls	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes
Year FE	Yes	Yes	Yes

Notes: ** p < 0.01; *** p < 0.001. Values in parentheses are t-statistics.

.

At the resource allocation level, digitalization enables firms to more accurately forecast demand, optimize inventory turnover, and streamline production schedules, thereby reducing redundant output and energy waste. Additionally, improved access to and precision of supply chain finance—enabled by digital platforms—helps alleviate financial constraints for SMEs, enhancing capital flow efficiency and avoiding carbon-intensive production induced by underinvestment.

4.3.2. Mechanism of Collaborative Technological Upgrading

Beyond optimizing resource and capital flows, improvements in carbon performance fundamentally rely on technological progress and productivity growth throughout the supply chain. On the one hand, digital tools reduce information barriers and enable rapid knowledge sharing; on the other hand, firms can connect with research institutes and service providers to enhance the specialization and coordination of green innovation. Moreover, emerging digital technologies such as blockchain and smart contracts help establish trust and IP protection in collaborative innovation, reducing moral hazard and opportunism, and thereby incentivizing deeper engagement in green technology development. Total factor productivity (TFP) has been widely recognized as a key determinant of both economic growth and carbon reduction. Micro-level and macro-level studies consistently show that higher TFP contributes positively to carbon performance via some channels. Efficiency improvement: firms with higher TFP utilize fewer resources and energy to produce the same output, reducing carbon intensity [45]. Technological substitution: high-TFP firms are more likely to adopt energy-efficient equipment and clean technologies. Structural optimization: TFP growth is often accompanied by production reorganization and modern management systems that enhance the green transition [47].

Against this backdrop, we examine whether focal firm digitalization enhances the TFP of upstream and downstream firms through data sharing, technological diffusion, and platform coordination—thereby forming a “chain-based upgrading” mechanism that improves carbon performance.

To rigorously estimate TFP, we adopt OP and LP approaches, which help correct for endogeneity in production function estimation due to input selection being influenced by unobserved productivity. We assume a Cobb–Douglas functional form for the production function. The OP method uses investment as a proxy to model productivity as a dynamic process; the LP method uses intermediate inputs (e.g., energy, raw materials) as proxies, suitable for samples with zero or lumpy investment. To mitigate concerns related to structural heterogeneity and time-varying shocks, all regression models control for firm fixed effects, year fixed effects, and industry fixed effects.

The results in

Table 8. Mediation mechanism 3: collaborative technological upgrading 1.

	(1)	(2)	(3)	(4)
Variables	TFPLP Up	TFPOP Up	TFPLP Down	TFPOP Down
Digital	0.116 ***	0.076 ***	0.250 ***	0.188 ***
	(0.019)	(0.018)	(0.016)	(0.015)
Constant	−5.299 **	−4.725 **	8.654 ***	7.996 ***
	(2.346)	(2.184)	(2.016)	(1.820)
Observations	3070	3070	3070	3070
R-squared	0.147	0.132	0.227	0.205
Controls	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes

Notes: ** p < 0.01; *** p < 0.001. Values in parentheses are t-statistics.

shows that focal firm digitalization significantly improves the TFP of both upstream and downstream firms, with stronger effects observed for downstream partners.

Table 9. Mediation mechanism 3: collaborative technological upgrading 2.

	(1)	(2)	(3)	(4)
Variables	CP	CP	CP	CP
Digital	0.454 ***	0.457 ***	0.421 ***	0.435 ***
	(0.0497)	(0.0493)	(0.0530)	(0.0518)
TFPLP up	0.120 *
	(0.0708)
TFPOP up		0.135 *
		(0.0760)
TFPLP down			0.197 **
			(0.0824)
TFPOP down				0.190 **
				(0.0917)
Constant	25.63 ***	25.61 ***	23.35 ***	23.58 ***
	(5.996)	(5.994)	(6.027)	(6.031)
Observations	3070	3070	3070	3070
R-squared	0.146	0.137	0.148	0.147
Controls	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001. Values in parentheses are t-statistics.

reveals that TFP has a significantly positive effect on supply chain carbon performance. These findings support the chain-based upgrading mechanism and provide empirical confirmation for H3.

4.3.3. Robustness Check

To further strengthen the credibility of the mediating mechanisms, we re-estimated all mediation models using the instrumental variable (IV) approach. This addresses potential endogeneity in the mediators that may arise from simultaneity, omitted variables, or measurement errors. Specifically, we applied the IV strategy to the three identified mediators: supply chain efficiency (SCE), supply chain finance (SCF), and upstream/downstream total factor productivity (TFP).

The IV results are reported in

Table 10. Robustness check: IV estimation of mediators.

	(1)	(2)	(3)	(4)
Variables	2nd Stage: SCE	2nd Stage: SCF	2nd Stage: TFPOP	2nd Stage: TFPOP
Digital	0.091 *	0.045 *	0.129 ***	0.154 ***
	(0.047)	(0.028)	(0.029)	(0.037)
Constant	8.664 ***	7.237 ***	7.286 ***	6.022 ***
	(1.000)	(0.452)	(0.495)	(0.612)
Observations	3070	3070	3070	3070
R-squared	0.089	0.030	0.148	0.134
Controls	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes

Notes: * p < 0.05; *** p < 0.001. Values in parentheses are t-statistics.

. We find that digitalization continues to exert a statistically significant positive effect on all three mediators under the IV estimation framework. These results are consistent with the OLS estimates. This robustness check provides strong support for the validity of the proposed mediating mechanisms. By addressing the potential endogeneity of mediators, we demonstrate that the pathways through which digitalization improves supply chain carbon performance remain reliable.

4.4. Further Analysis

To further explore whether the impact of digitalization on supply chain carbon performance is affected by external environmental factors, we conduct subgroup regressions along two dimensions: regional environmental regulation intensity and industry characteristics. In addition, to assess the moderating effect of the 2018 supply-chain pilot-city policy, we construct a difference-in-differences (DID) specification by interacting a treatment group indicator (pilot cities) with a post-policy dummy.

4.4.1. Heterogeneity in Regional Environmental Regulation

Following the official data published by the National Bureau of Statistics, we calculate the intensity of environmental regulation as follows:

Environmental Regulation Intensity = (Investment in Industrial Pollution Control)/(Value Added of Secondary Industry)/10,000.

We then match this index to each supply chain based on the province in which the focal firm is located. The sample is divided into high- and low-environmental-regulation groups using the median value, and regressions are run separately.

The results in

Table 11. Heterogeneity analysis: environmental regulation, pilot policy, and industry type.

	(1)	(2)	(3)	(4)
Variables	ER High	ER Low	Regulated	Unregulated
Digital	0.441 ***	0.313 ***	0.456 ***	0.430 ***
	(0.060)	(0.088)	(0.033)	(0.078)
Constant	0.677	49.139 ***	3.792	50.431 ***
	(9.266)	(12.055)	(4.274)	(10.848)
Observations	1533	1537	1197	1873
R-squared	0.170	0.237	0.380	0.167
Controls	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes

Notes: *** p < 0.001. Values in parentheses are t-statistics.

(Columns (1) and (2)) show that the digitalization level of focal firms is positively associated with supply chain carbon performance in both groups. However, the effect is more pronounced in regions with stronger environmental regulation. This finding suggests that in high-regulation regions, firms face stricter carbon emission constraints and thus have stronger incentives and necessity to use digital tools to enhance information transparency, resource allocation, and upstream–downstream coordination—leading to more substantial improvements in carbon performance. It supports H4.

4.4.2. Heterogeneity Between Regulated and Unregulated Industries

Considering that firms in different types of industries face varying levels of environmental supervision and carbon reduction pressures, we divide the sample into regulated and unregulated industries based on official classifications and conduct separate regressions. (In China, regulated industries refer to sectors that are subject to special government supervision and control based on objectives such as national security, public interest, economic strategy, and resource conservation. Regulation is implemented through instruments such as laws and regulations, market entry restrictions, price controls, and technology export restrictions. These industries are generally categorized into three types: (1) economic regulation, which applies to sectors such as energy and mineral resources, which are vital to the national economy; (2) social regulation, covering areas such as pharmaceuticals and finance, which have a significant impact on public welfare; and (3) antitrust regulation, which targets natural monopolies such as electric power and railways.)

The results in Table 11 show that digitalization significantly improves carbon performance in both groups. However, the effect is stronger in unregulated industries than in regulated industries. This may indicate that digitalization more effectively unlocks operational constraints and legacy practices in unregulated industries, enabling better integration of information and logistics across the supply chain. In contrast, regulated industries tend to be more constrained by rigid compliance procedures and capital-intensive structures, which may limit the potential gains from digitalization. The heterogeneity in regression coefficients can be observed from Figure 4.

4.4.3. Moderating Role of the Supply-Chain Pilot-City Policy

In 2018, China’s National Development and Reform Commission, together with seven other ministries, launched the “Pilot Cities for Supply Chain Innovation and Application” program, aiming to promote supply chain digital transformation at the city level. Based on this policy context, we construct a difference-in-differences (DID) policy indicator: observations are assigned a value of 1 if the focal firm is located in a pilot city and the year is 2018 or later and 0 otherwise. This policy dummy is then interacted with the digitalization variable to examine whether the policy amplifies the effect of focal firm digitalization on supply chain carbon performance.

In addition, we divide the sample into two subsamples based on whether the firm is located in a pilot city (SD city = 1) or not (SD city = 0) to conduct subgroup regressions.

The results are shown in

Table 12. Moderating role of the supply-chain pilot-city policy.

	(1)	(2)	(3)	(4)
Variables	Benchmark	Interaction	SD City = 1	SD City = 0
Digital	0.473 ***	0.416 ***	0.577 ***	0.363 ***
	(0.049)	(0.067)	(0.099)	(0.054)
DID_ Digital		0.211 *
		(0.160)
DID		−1.635 **
		(0.780)
Constant	25.111 ***	25.978 ***	39.750 *	15.606 **
	(5.992)	(9.466)	(20.418)	(6.324)
Observations	3070	3070	859	2211
R-squared	0.144	0.157	0.180	0.113
Controls	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes

Notes: * p < 0.05; ** p < 0.01; *** p < 0.001. Values in parentheses are t-statistics.

. Column (2) includes the policy interaction term. The main effect of digitalization remains significantly positive, though the magnitude is slightly reduced. The interaction term is also positive and marginally significant, indicating that the pilot city policy strengthens the positive effect of digitalization on carbon performance. Interestingly, the DID term itself is negative and significant, suggesting that—absent digital transformation—firms in pilot cities may face greater carbon-performance pressure, thereby highlighting the importance of focal firm digitalization. The moderating effect of the supply-chain pilot-city policy is illustrated in Figure 5.

Furthermore, Columns (3) and (4) report subgroup regressions. In pilot cities, the coefficient of digitalization is 0.577, significantly higher than the 0.363 observed in non-pilot cities, confirming that stronger institutional support and digital infrastructure amplify the carbon-reduction benefits of digital transformation. These findings provide empirical support for H5.

While this study is grounded in China’s institutional and regulatory context, the underlying mechanisms identified—such as resource allocation efficiency and policy-modulated digital adoption—may also have broader international relevance. For instance, digitalization in European and US supply chains has been closely tied to decarbonization efforts, though often shaped by different institutional logics [68]. Western contexts tend to emphasize market-based mechanisms, voluntary ESG initiatives, and private-sector digital platforms, such as those supporting Scope 3 carbon disclosure and green finance integration. In contrast, China’s policy-led digital governance emphasizes top-down coordination and supply-chain pilot-city interventions.

5. Conclusions and Implications

5.1. Conclusions

This study investigates how focal firm digitalization affects supply chain carbon performance from a holistic supply chain perspective, grounded in dynamic capability theory. Both theoretical modeling and empirical analysis demonstrate that improvements in digitalization significantly enhance carbon performance at the supply chain level. Two key mediating mechanisms are identified: (1) digitalization optimizes resource allocation efficiency across the supply chain, thereby improving operational efficiency and reducing emissions intensity; (2) digitalization promotes upstream and downstream technological upgrading, leading to higher total factor productivity and generating a coordinated decarbonization effect across the supply chain.

Further heterogeneity analyses reveal that the impact of digitalization is significantly affected by policy environments (e.g., environmental regulation intensity and supply chain innovation pilots) and industry characteristics (e.g., whether the industry is regulated).

This research contributes to the literature in three main aspects.

First, this study shifts the analytical lens from individual firms to the entire supply chain. By constructing a three-tier panel dataset linking upstream, focal, and downstream firms, it provides systematic evidence that focal firm digitalization can significantly improve overall supply chain carbon performance. This expands the analytical boundaries of digital green transition research and integrates supply chain-wide considerations into existing firm-centric frameworks.

Second, drawing on the dynamic capabilities theory, this study identifies and empirically validates two meso-level mechanisms through which digitalization enhances carbon performance: resource allocation optimization and collaborative technological upgrading. This provides a clearer theoretical explanation of how digitalization affects carbon outcomes by improving supply chain efficiency and enabling joint innovation across upstream and downstream partners. This study enriches the theoretical framework for understanding the environmental effects of digitalization.

Third, this study incorporates external policy contexts—including environmental regulation, supply-chain pilot programs and industry-specific institutional settings—into the empirical framework. It reveals significant heterogeneity and moderating effects in the relationship between digitalization and carbon performance, enriching theoretical explanations of how institutional environments shape digitalization outcomes.

5.2. Implications

For firms, the findings confirm that digitalization significantly enhances supply chain carbon performance. This suggests that corporate managers should further increase investments in digital technologies and strengthen capability building—particularly to facilitate coordinated digital transformation across upstream and downstream partners.

Specifically, firms should align their digitalization efforts with two key mechanisms: optimizing resource allocation across the supply chain and enabling collaborative technological upgrading. Digitalization should be reframed not merely as a tool but as a core carrier of dynamic capabilities. Focal firms, in particular, should move beyond a firm-centric approach and leverage their network position to act as hubs for digital collaboration. They are encouraged to build data-sharing platforms across the supply chain, enabling real-time monitoring of resource consumption and emissions. Through advanced analytics, they can optimize procurement, production, and logistics processes, thereby reallocating resources toward low-carbon, high-efficiency nodes.

For small and medium-sized enterprises (SMEs), which typically face limited digital capabilities, it is essential to integrate into digital supply chains via industrial collaboration platforms to strengthen coordination with focal firms. On one hand, SMEs should actively respond to digital transformation initiatives led by focal firms and governments by accelerating digital infrastructure and adopting cost-effective, adaptable solutions. On the other hand, they should leverage public policy and financial support to embed themselves into digital ecosystems early and avoid marginalization due to the widening digital divide.

For policymakers, the results underscore the importance of institutional environments in shaping the carbon impacts of digitalization. Governments should further enhance the guiding role of supply chain innovation pilot programs. First, digital empowerment programs should be expanded to support SME integration into digital supply chains through scalable, cost-effective digital solutions, thereby promoting the widespread adoption of digital technologies in supply chain operations. Simultaneously, the government should invest in resilient logistics and transportation infrastructure to reduce operating costs, improve supply chain efficiency, and enhance overall carbon performance. Finally, supply chain finance (SCF) policies should be refined. This includes establishing regulatory frameworks, offering risk-mitigation instruments, and encouraging focal firms to adopt receivables confirmation mechanisms. These efforts will ease financing constraints—especially for SMEs—and promote capital flows toward green transformation initiatives across the supply chain.

5.3. Limitations and Future Research

Although this study has achieved substantial progress in both theoretical development and empirical testing, several limitations remain, offering opportunities for future research.

First, due to commercial confidentiality, inconsistent internal data systems, and the absence of mandatory carbon disclosure standards, few firms publicly disclose complete carbon emissions data. To overcome this limitation and expand the sample size, this study estimates firm-level carbon emissions by combining industry-level energy consumption data from China Statistical Yearbook with each firm’s cost share of total industry operating revenue. This method effectively mitigates sample selection bias and enhances data coverage. Nevertheless, it inevitably neglects the heterogeneity in cost structures among firms within the same industry, which may introduce estimation errors. Future research can take advantage of the ongoing improvement in China’s carbon disclosure system and the increasing deployment of carbon monitoring technologies to obtain more accurate firm-level emission data. In addition, researchers could use government administrative datasets, voluntary disclosure platforms, or emerging digital carbon accounting tools to validate and refine emission estimates.

Second, this study measures firm digitalization primarily through digital patent counts, supplemented by text-based digitalization word frequency and digital asset investment indicators for robustness checks. Although these proxies are widely adopted in existing literature, they cannot fully capture the multifaceted nature of digital transformation or the specific implementation of digital initiatives within firms. As digital transformation encompasses strategic, technological, and organizational dimensions, future research could construct more comprehensive and multidimensional measures by combining survey-based assessments, digital project disclosures, and internal digital infrastructure indicators. Such approaches would allow scholars to better capture both the intensity and depth of digital transformation across different firms and industries.

Third, the empirical analysis in this study is based on publicly listed Chinese firms, which typically possess stronger digital capabilities and more transparent disclosure practices. However, this focus inevitably limits the generalizability of the findings to non-listed small and medium-sized enterprises (SMEs), which represent a large proportion of supply chain participants. In addition, due to the limited availability of corporate disclosure data, the definition of supply chain boundaries in this study does not yet capture secondary suppliers or downstream clients. Future research could address these limitations through more extensive enterprise surveys and field investigations, thereby constructing a more comprehensive picture of supply chain relationships and their carbon performance dynamics.

Finally, while this study identifies two core mediating mechanisms—resource allocation optimization and collaborative technological upgrading—digitalization may affect carbon performance through additional pathways in real-world supply chains. For example, digital technologies may reshape supply chain governance, foster real-time carbon monitoring, or enhance green innovation spillovers. These mechanisms have yet to be systematically verified. Moreover, because the time span of our study is relatively short, some long-term dynamic mechanisms embedded in supply chain evolution may not have been fully captured or identified. Future research could adopt mixed-method approaches, such as case studies, expert interviews, or structural equation modeling, to explore the nonlinear and synergistic effects among multiple mechanisms. Longitudinal analyses over extended periods would also help uncover how digital capabilities evolve and influence supply chain carbon performance over time.