Effects of the Digital Economy on Reducing Carbon Emissions in China’s Energy-Intensive Manufacturing Enterprises

Abstract

The energy-intensive manufacturing industry is a significant contributor to carbon emissions, necessitating urgent measures to reduce its carbon footprint. The advent of the digital economy has engendered a milieu conducive to the decarbonization of energy-intensive manufacturing enterprises. This paper utilizes panel data from A-share listed energy-intensive manufacturing enterprises in China from 2012 to 2024 to empirically analyze the impact of the digital economy on carbon abatement performance. The findings reveal the following: First, the digital economy has a significant effect on enhancing the carbon emission reduction performance of energy-intensive manufacturing enterprises. Second, total factor productivity, R&D investment, and technology innovation have partial mediating effects in this relationship. Third, heterogeneous effects exist across enterprises of six major energy-intensive industries. Fourth, the carbon reduction effect of the digital economy is more significant in central and western regions compared to eastern regions in China. These results underscore the importance of accelerating digital transformation and formulating diversified policies predicated on industries and regions to facilitate the realization of China’s “dual carbon” goals.

1. Introduction

Promoting the green transition of the manufacturing industry is a key lever for China to achieve its “dual-carbon goals “ (carbon peaking and carbon neutrality) [1,2]. However, energy-intensive industries still dominate the manufacturing industry, making emission reductions increasingly urgent [3]. Traditional decarbonization strategies focus primarily on improving energy efficiency and promoting technological innovation. Nevertheless, the rapid development of the digital economy has revealed substantial potential for digital technologies to optimize production processes, improve energy utilization efficiency, and reduce carbon emissions [4]. Furthermore, proactive policies have accelerated the deep integration of the digital economy with the real economy, injecting new impetus into the green transformation of energy-intensive manufacturing industry [5]. Consequently, investigating the impact of the digital economy on the carbon abatement performance of energy-intensive manufacturing enterprises is critical for sustainable development in this industry.

Existing studies on the nexus between digital economic development and carbon mitigation have predominantly emphasized macro-level investigations at the national [6], provincial [7], and municipal scales [8]. While some studies have expanded to strategic regional contexts such as the Yangtze River Economic Belt and the Yellow River Basin [9,10], a knowledge gap remains with respect to micro-level, firm-specific analyses. Present studies mainly examine the facilitating role of the digital economy in green innovation [11], productivity optimization [12], and the adoption of ESG (Environmental, Social, Governance) paradigms [13]. Meanwhile, empirical research on the drivers of enterprise decarbonization predominantly focused on regulatory mechanisms, carbon disclosure protocols [14], and green financing instruments [15]. Notably, the intrinsic carbon abatement potential embedded in digital economic transformation remains insufficient, particularly in operational mechanisms at the enterprise operational level.

Micro-level enterprise research can provide more detailed and specific data for decision-making, especially when challenges such as market competition and unsustainability of traditional business models come to enterprises. In light of this, it is crucial to effectively harness the industrial integration effects of the digital economy to achieve sustainable development [16]. Existing studies have shown that the digitalization of enterprises contributes to the carbon emissions reduction [17,18]. However, since the digital economy takes data as a production factor, its essential mechanism is to promote the deep integration of digital technologies with enterprise operations, including R&D and supply chain management, thereby realizing lean management and improving overall performance [19,20]. In addition, the digital economy inherently makes green value [21], which can reduce resource misallocation of resources and the carbon emissions. Research on the digital economy enables enterprises to deepen technology applications based on digital foundations, thereby demonstrating significant advantages in achieving carbon reduction targets and promoting green transformation [22].

Energy-intensive manufacturing industry is a critical component of the national economy while facing the pressure of carbon emission control in the era of digital economy [23], this paper takes Chinese A-share listed enterprises in Energy-intensive manufacturing industry from 2012 to 2021 as the research sample to empirically examine how digital economy takes effect on carbon emission reduction performance of Energy-intensive manufacturing enterprises.

This study makes threefold contributions.

First, presenting a novel analytical framework that integrates three critical mediating factors, including Total Factor Productivity (TFP), R&D Investment (IIN), and technology innovation (TEI), into a unified model to elucidate the mechanisms by which the digital economy enhances corporate carbon emission reduction performance (CERP). By combining these factors into a unified model, the study offers a comprehensive analysis of their concurrent and interactive effects, addressing a gap in the literature that often examines them separately.

Second, employing a robust, multi-layered empirical strategy to rigorously test the proposed framework, which is central to its methodological contribution. Building on established methods, it advances beyond standard mediation analysis by (1) using bootstrap tests for mediating effects to ensure the statistical robustness of identified indirect pathways; (2) explicitly modeling and testing the interaction between TFP and R&D investment as a mediating mechanism, thereby capturing the synergistic effect of productivity gains and innovation investment on the impact of the digital economy; and (3) conducting comprehensive heterogeneity analyses across industries, regions, and firm sizes, with rigorous statistical tests to formally validate coefficient differences rather than relying on simple subsample regressions. In summary, this approach surpasses traditional mediation analyses by formally validating the observed heterogeneities through advanced empirical techniques

Third, offering comprehensive, evidence-based insights for policy and practice. The results confirm a positive effect and identify the specific contexts and populations in which it is pronounced. Observations of regional and industrial heterogeneity provide a rigorous foundation for policymakers to develop targeted, rather than generic, strategies to support the digital sustainable transition.

2. Literature Review and Hypothesis Development

2.1. The Digital Economy and Carbon Emission Reduction in Energy-Intensive Manufacturing Enterprises

The digital economy can facilitate the reduction of carbon emissions in energy-intensive manufacturing enterprises at several levels [24]. First, the integration of the digital economy with the real economy can improve the production efficiency of enterprises that integrate and analyze data through digital production, promote data visualization, and optimize the production process through big data models [25]. This process can facilitate accurate decision-making and refined production management [26,27]. Levinson et al. compared innovate digital technologies to traditional production technologies, the application of digital technologies enables companies to increase productivity with less energy consumption while reducing environmental pollution [8]. Wang et al. examined digital applications that could improve the productivity of industrial enterprises using text-based big data [28]. Similarly, Guo et al. developed two systematic indicator systems to explain how the digital economy promotes green total factor productivity [29]. Another study considered AI as a mediating factor within a governance framework to identify how the digital economy drives technological progress in energy industries, thereby inducing an enhancement in proclivities [30]. Through the deep integration of digital technologies into the production process, enterprises are able to monitor and analyze energy flow and consumption in real time. This facilitates the development of sustainable management strategies, optimizes the production chain, improves resource utilization, and consequently reduces carbon emissions [31,32].

Second, the emergence of the digital economy has substantially increased information acquisition and alleviated information asymmetry [33]. For instance, Sun & Guo suggested that increased market transparency makes it easier for enterprises to address information asymmetries with downstream entities [34], enabling them to formulate production plans based on the balance of supply and demand more accurately. Thus, the digital economy has helped enterprises improve the efficiency of resource use, promote the reduction of carbon emissions, and increase profits [19,35]. In addition, the digital economy has increased the transparency of business-related information, including organizational structure, operational status, and credit ratings [36]. Such information transparency fosters greater trust among enterprises [37], allowing investors to make optimal investment decisions based on comprehensive data, thereby providing financial support for technological innovation and green transformation in high energy-intensive manufacturing enterprises [11,38]. Increased information transparency also mitigates conflicts between shareholders and management, encouraging enterprises to prioritize long-term profits and adopt low-carbon production strategies [39]. Therefore, this study proposes the following:

Hypothesis 1.

The digital economy can effectively enhance the carbon emission reduction performance of energy-intensive manufacturing enterprises.

2.2. The Mechanism of Total Factor Productivity Between the Digital Economy and Carbon Emission Reduction in Energy-Intensive Manufacturing Enterprises

As a new factor of production, data is increasingly integrated into the production processes of energy-intensive manufacturing enterprises, which has a profound impact on enterprise productivity [40]. Gao et al. employed a fixed-effect model to demonstrate that digital transformation significantly enhances total factor productivity (TFP) [41]. digital economy facilitates the integration of digital technologies with traditional production, thereby transforming the production model and relations to establish a data-driven efficient production system [42,43]. Wang et al. employed a DID model to demonstrate that the implementation of digital applications in enterprises significantly enhances firm-level total factor productivity (TFP) [44]. Yang et al. further asserted that this transformation contributes to the improvement of TFP [45], enabling energy-intensive enterprises to achieve greater economic benefits while adhering to predetermined carbon emission constraints [46], thereby improving their carbon emission reduction performance [47].

The enhancement of TFP facilitates the transition to renewable energy [48], generating positive spillover effects on the carbon emission reduction performance of energy-intensive manufacturing enterprises. Zhang et al. used data from 171 countries over the period 1990–2019 to identify that total factor productivity (TFP) can facilitate the clean energy transition [18]. By optimizing TFP, these enterprises can gain insights into the consumption patterns of capital, labor, and resources during production processes, thereby pursuing cost-minimization strategies to maximize economic returns [40,49]. This mechanism helps reduce energy waste across operational cycles, consequently improving carbon mitigation performance [14]. Furthermore, the digital economy drives the intelligent transformation of production systems, enabling highly efficient coordination and operations among interconnected production modules [50]. Such intelligence not only enhances overall production efficiency but also reinforces TFP growth through optimized resource allocation, creating synergistic pathways for improved carbon reduction [49]. Therefore, this study proposes the following:

Hypothesis 2

. The digital economy enhances the carbon emission reduction performance of energy-intensive manufacturing enterprises by increasing total factor productivity which has a mediating effect.

2.3. Mechanisms of R&D Investment in Carbon Emission Reduction Between the Digital Economy and Energy-Intensive Manufacturing Enterprises

As a key determinant of business competitiveness, research and development (R&D) intensity influences the entire business lifecycle [51]. In the context of the digital economy, the strategic importance of R&D becomes even more pronounced. Digital technologies have become essential components of industrial production ecosystems, compelling energy-intensive manufacturing companies to escalate their R&D investments. This trend injects modernization momentum into traditional manufacturing sectors, accelerating the digital and intelligent transformation of production workflows while improving operational efficiency [52]. Furthermore, the digital economy streamlines companies’ innovation processes by mobilizing financial resources for R&D initiatives, thereby institutionalizing mechanisms for sustainable technological innovation [53]. These dual mechanisms not only drive technological progress but also optimize resource utilization, ultimately contributing to the reduction in industrial carbon emissions reduction.

The digital economy promotes carbon emission reduction in energy-intensive manufacturing enterprises by increasing R&D investment [54], which is manifested as a substitution effect. Some studies have demonstrated that R&D-supported digital transition in manufacturing enterprises can lead to a reduction in carbon emissions [55,56]. Driven by the digital economy, enterprises can acquire more funds for R&D [10], not only reducing marginal costs and increasing profits, but also continuously reinvesting the funds in the application of digital technologies through continuous capital accumulation, reasonably substituting the proportion of traditional factors and gradually reducing marginal returns [57]. The increase in R&D investment also enables enterprises to continuously accumulate production data and knowledge [33]. Through the processing and analysis of big data algorithms, enterprises can form unique competitive advantages [58], thereby providing impetus for the sustainable development of carbon emission reduction [59]. Therefore, this study proposes the following:

Hypothesis 3.

The digital economy enhances the carbon emission reduction performance of energy-intensive manufacturing enterprises by increasing R&D investment, which has a mediating effect.

2.4. Mechanisms of Enterprises Technological Innovation in Carbon Emission Reduction Between the Digital Economy and Carbon Emission Reduction in Energy-Intensive Manufacturing Enterprises

The digital economy equips enterprises with a wide range of technological tools and platform-based capabilities. The widespread adoption of cloud computing and big data analytics facilitates information collection, data storage, and intelligence analysis, thereby enhancing institutional capabilities for information sharing and knowledge management systems [60,61]. Green transformation through green technology innovation refers to leveraging digital economy advancements to achieve carbon reduction in manufacturing enterprises [62]. The integration of these digital solutions has led to dual accelerations in information fluidity and analytical precision, while simultaneously improving market responsiveness through data-driven decision frameworks [63]. This technological empowerment accelerates innovation diffusion cycles, prompting energy-intensive manufacturing enterprises to prioritize the strategic development of green and low-carbon technologies. Such R&D commitments systematically reduce energy intensity by optimizing energy utilization efficiency [64], thereby achieving measurable progress in mitigating carbon footprints.

The digital economy facilitates the reduction of carbon emissions in energy-intensive manufacturing enterprises through technological innovation [65]. Energy-intensive manufacturing enterprises typically have high energy consumption, large equipment, and poor energy utilization. Although reducing energy consumption is essential for reducing carbon emissions, it often conflicts with the company’s profit goals [66]. However, the use of advanced technologies and tools can improve energy efficiency and minimize energy loss during the production process [67]. Through information integration and efficient resource utilization, technological innovation permeates traditional production processes and raw material use, thereby improving energy use efficiency, accelerating the transformation of production stages, enhancing process coordination, and ultimately reducing overall energy consumption [27,36,68]. As a result, the digital economy optimizes enterprise operations and enhances energy efficiency through technological innovation, thus improving the performance of carbon emission reduction [65]. Therefore, this study proposes the following:

Hypothesis 4.

The digital economy enhances the carbon emission reduction performance of Energy-intensive manufacturing enterprises by improving technological innovation. And enterprise technological innovation has a mediating effect.

3. Methodology

This section consists of three parts. The first introduces the models estimated to identify the carbon emission reduction impact of digital economy on energy-intensive manufacturing enterprises. The second explains variable selection and description, while the third provides data sources.

3.1. The Model

To verify whether the digital economy can facilitate carbon emission reduction in energy-intensive manufacturing enterprises, this paper constructs the following model.

$${CERP}_{it}={\alpha }_0+{\alpha }_1{DE}_{it}+{\alpha }_2{X}_{it}+{\mu }_{it}+{\nu }_{it}+{\epsilon }_{it}$$

(1)

where i and t are enterprises and years, respectively; explanatory variable $CERP$ represents carbon emission reduction performance of energy-intensive manufacturing enterprises; explanatory variable $DE$ represents the development level of digital economy in enterprises; X is a set of control variables; ${\mu }_{it}$ and ${\nu }_{it}$ are industry- and time-fixed effects, respectively; ${\epsilon }_{it}$ is random error term.

Based on the above theoretical analysis, it is obvious that the development of digital economy may affect the carbon emission performance of enterprises through total factor productivity, R&D investment and enterprise technological innovation. In order to test the above hypotheses, the following mediating effect model is constructed based on model (1) according to Lu et al. [69]:

$$M_{it}={\beta }_0+{\beta }_1{DE}_{it}+{\beta }_2{X}_{it}+{\mu }_{it}+{\nu }_{it}+{\epsilon }_{it}$$

(2)

$${CERP}_{it}={\lambda }_0+{\lambda }_1{DE}_{it}+{{\lambda }_{2}M}_{it}+{\lambda }_3{X}_{it}+{\mu }_{it}+{\nu }_{it}+{\epsilon }_{it}$$

(3)

where M is the mediating variable representing total factor productivity (TFP), R&D investment (IIN) and technological innovation (TEI); the rest of the variables are defined as in the baseline model.

3.2. Variable Selection and Description

3.2.1. Explained Variables

Carbon emission reduction performance (CERP) is selected as the dependent variable. Given the limited availability of disclosed enterprise carbon emissions data, this study adopts an indirect estimation approach. First, following Yan et al. [7], enterprise carbon emissions are estimated through industry-level benchmark. Specifically, an enterprise’s carbon emissions are calculated as the product of its operating cost to industry operating cost ratio multiplied by the total carbon emissions of its corresponding industry. Then, based on the methodology developed by Li and Li [70], carbon emission reduction performance is measured by the ratio of a enterprise’s operating income to its carbon emissions. This metric reflects the revenue generated per unit of carbon emissions, with a higher value indicating superior carbon emission reduction performance. The CEPR will be measured as follows:

$$CEPR={\displaystyle \frac{\mathrm{E}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{p}\mathrm{r}\mathrm{i}\mathrm{s}\mathrm{e} \mathrm{o}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{n}\mathrm{g} \mathrm{i}\mathrm{n}\mathrm{c}\mathrm{o}\mathrm{m}\mathrm{e}}{\frac{Enterprise operating costs}{Industry operating costs}\times \mathrm{C}\mathrm{a}\mathrm{r}\mathrm{b}\mathrm{o}\mathrm{n} \mathrm{e}\mathrm{m}\mathrm{i}\mathrm{s}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{s} \mathrm{o}\mathrm{f} \mathrm{t}\mathrm{h}\mathrm{e} \mathrm{i}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}}}$$

(4)

3.2.2. Core Explanatory Variables

The core explanatory variable is digital economy (DE). Given the limited data at the enterprise level, this study constructs a digital terminology dictionary through textual analysis following Yuan et al. [31], and quantifies DE for energy-intensive manufacturing enterprises by calculating the ratio of digital economy keywords extracted from annual reports to the total words in management discussions. Furthermore, following the methodology of Zhao et al. [71], we compute a composite index using the entropy method across four dimensions: digital technology application, Internet business models, intelligent manufacturing, and information systems. Experts scoring (3/2/1/0 points) is implemented on the basis of two points. Whether the scored company takes digitalization as a primary investment direction and how the company integrates digital economy with production and operation activities. Finally, digital economy (DE_1) is derived by equally weighting textual and expert scoring metrics (50% each) for robustness checks.

3.2.3. Mediating Variables

For total factor productivity (TFP), we integrate the benchmarking approach of Lu and Lian [72] with the system GMM estimator developed by Blundell and Bond [73]. R&D investment (IIN) is calculated as the ratio of R&D expenditure to operating income, following the methodology of Xuan and Zhang [74]. Technological innovation (TEI) is quantified by the total number of patents granted during the sample period, using the metric proposed by Chen and Wang [75].

3.2.4. Control Variables

To control for potential confounding factors affecting carbon emissions, this study selects control variables based on the established literature, which include: (1) Proportion of R&D Employees (PRD), measured as the ratio of R&D employees to total employees. (2) Enterprise Size (ENS), represented by total assets. (3) Working Capital Ratio (WCR), calculated as the difference between current assets and current liabilities divided by total assets. (4) Intangible Assets Ratio (IAR), defined as net intangible assets divided by total assets. (5) Reporting Standardization (refGRI): coded as 1 if the company’s reports follow the GRI Sustainability Reporting Standards; otherwise, coded as 2. (6) Total Debt Ratio (TDR), calculated as total liabilities divided by total assets.

3.2.5. Data Sources

The core explanatory variable (digital economy data) is extracted from the annual reports of listed enterprises. Data on the dependent variable (carbon emission reduction performance) are obtained from the China Statistical Yearbook and CSMAR database. Mediating variables, including total factor productivity, R&D investment, enterprise technological innovation, and control variables are obtained from CSMAR Database. Our sample consists of 4375 observations from 619 A-share listed energy-intensive manufacturing firms, forming a balanced panel dataset spanning from 2012 to 2024.

Table 1. Results of descriptive statistics.

Variable	Average Value	(Statistics) Standard Deviation	Minimum Value	Maximum Values
CERP	0.964	0.599	0.082	6.195
DE	0.321	0.215	0.000	2.414
TFP-GMM	5.928	0.791	3.987	8.725
IIN	0.0240	0.022	0.000	0.253
TEI	15.600	54.390	0.000	890.000
PRD	10.140	7.938	0.000	88.200
ENS	243.400	529.400	0.965	5868.000
WCR	0.074	0.261	−2.702	0.970
IAR	0.050	0.054	0.000	0.677
refGRI	1.827	0.379	1.000	2.000
TDR	0.474	0.215	0.013	3.262

listed descriptive statistics of these variables.

4. Results

4.1. Baseline Regression

The results of the baseline regressions are presented in column (1) of

Table 2. Benchmark regression results.

	(1)	(2)
	CERP	CERP
DE	0.132 ***	0.108 ***
	(5.48)	(4.59)
PRD		0.001 *
		(1.77)
ENS		0.000 ***
		(2.91)
WCR		0.100 ***
		(2.72)
IAR		0.954 ***
		(3.64)
refGRI		−0.005
		(−0.31)
TDR		−0.185 ***
		(−3.87)
constant term (math.)	0.288 ***	0.335 ***
	(8.26)	(6.55)
control variable	yes	yes
industry-fixed effect	yes	yes
Year-fixed effects	0.674	0.690
R2	4357.000	4357.000

Note: *, *** denote significance at the 10%, and 1% levels, respectively; numbers in parentheses are t-statistics; same below.

. The digital economy (DE) is the sole explanatory variable included, with its coefficient indicating a positive and statistically significant association with the carbon emission reduction performance (CERP) of manufacturing enterprises at the 1% significance level. Specifically, the coefficient of 0.132 suggests that a one-unit increase in the digital economy index corresponds to a 0.132-unit increase in an enterprise’s CERP. Considering that the standard deviation of CERP is 0.599 (as reported in Table 1), this effect represents a substantial economic impact. Moreover, based on mean values, a 1% increase in the digital economy index is associated with an approximate 0.041% improvement in CERP at the sample mean. This quantifiable effect provides clear implications for policymakers and managers, suggesting that targeted investments in digital infrastructure and capabilities can yield measurable improvements in carbon efficiency. The relationship remains robust when controlling for various factors, including industry-fixed effects, as indicated by the corresponding regression results. These findings corroborate the positive influence of the digital economy on carbon emission reduction performance, thereby providing empirical support for Hypothesis 1. Unlike prior macro-level analyses of the carbon mitigation potential of the digital economy, this micro-level empirical investigation systematically validates the carbon mitigation benefits of digital economy adoption within enterprise operations. This evidence highlights that digital transformation functions not only as a national decarbonization strategy but also as a critical enabler of enterprise-level emissions reduction, positioning it as an essential accelerator for achieving sustainable development objectives.

Departing from traditional analyses that treat all energy-intensive manufacturing enterprises as a homogeneous group when examining the emission reduction effects of digital transformation [76], this study pioneers a novel focus on the specificity of individual enterprises, generating new insights. Moreover, our results indicate that applications of the digital economy extend beyond conventional digital transformation paradigms, which are typically limited to internal process optimization. By leveraging digital technologies, the digital economy systematically enhances resource utilization efficiency across internal operations, thereby improving carbon mitigation performance. Crucially, the integration of big data analytics and artificial intelligence within digital economy frameworks fosters intelligent production ecosystems [51], resulting in significant decarbonization effects that are especially pronounced in the energy-intensive manufacturing sector.

4.2. Robustness Check

Notwithstanding statistically significant baseline regression results, comprehensive robustness checks are conducted through placebo testing, explanatory variable substitution, and one-period lagged variable analysis to ensure empirical validity.

4.2.1. Placebo Test

To ensure the baseline regression results were not due to randomness, we conducted placebo tests using systematic subsampling. We randomly selected 300 observations (48% of the total sample) and repeated this process over 500 times. The scatterplot and distribution of estimated coefficients are shown in Figure 1. The pseudo-estimates ranged from −0.1 to 0.1, differing from the original coefficient (0.108), with only a few reaching statistical significance (p < 0.1). These test results confirm the robustness of the baseline regression findings, empirically validating that the digital economy has a significant impact on carbon mitigation in energy-intensive manufacturing enterprises.

4.2.2. Substitution of Explanatory Variables

Following Zhao et al. [71], we reconstruct the digital economy development index (DE_1) for energy-intensive manufacturing enterprises through integrated textual analysis and expert scoring, substituting the economic metric we used. Columns (1)–(2) of

Table 3. Robustness test results.

	(1)	(2)	(3)	(4)
	CERP	CERP	CERP	CERP
DE_1	0.001 ***	0.001 ***
	(4.21)	(1.36)
L.DE			0.122 ***	0.101 ***
			(4.81)	(4.19)
constant term (math.)	(2.04)	(2.18)	(9.81)	(7.61)
control variable	no	yes	no	yes
industry-fixed effect	yes	yes	yes	yes
Year-fixed effects	yes	yes	yes	yes
R2	0.156	0.132	0.678	0.697
observed value	4357	4357	4357	4357

Note: *** denote significance at the 1% levels, respectively; numbers in parentheses are t-statistics; same below.

show that DE_1 retains a statistically significant positive coefficient (β = 0.001, p < 0.01) on carbon emission reduction performance (CERP), with significance levels fully preserved. This result confirms the robustness of our main findings regarding the carbon reduction effects of the digital economy.

4.2.3. One-Period Lagged Explanatory Variables

The digital economy is a kind of economic activity in the daily operation of enterprises, there may be some delay in its influence on the carbon emission reduction performance of high-energy-consuming manufacturing enterprises. Therefore, we take the one-period lag of digital economy and subject to regression tests. The results, as presented in columns (3) and (4) of Table 3, indicate that the digital economy with a one-period lag has a significant positive impact on the carbon emission reduction performance. The regression coefficient of L.DE is 0.101 and is significant at the 1% level, suggesting that there exists a certain lag effect of the digital economy on the carbon emission reduction performance of high-energy-consuming manufacturing enterprises. The empirical results of this paper are robust.

5. Discussion

5.1. Mechanism Analysis

In order to explore how the digital economy enhances the carbon emission reduction performance of energy-intensive manufacturing enterprises, this study conducted a regression test of Hypotheses 2–4 using the mediation effect model, and the results are shown in

Table 4. Mediating effect results.

	(1)	(2)	(3)	(4)	(5)	(6)
	TFP_GMM	CERP	IIN	CERP	TEI	CERP
TFP_GMM		0.045 ***
		(3.42)
IIN				1.721 ***
				(5.93)
TEI						0.000 ***
						(2.83)
DE	0.138 **	0.102 ***	0.012 ***	0.087 ***	8.681 **	0.107 ***
	(2.50)	(4.34)	(5.79)	(3.78)	(2.46)	(4.53)
constant term (math.)	5.791 ***	0.072	0.004 *	0.328 ***	32.400 ***	0.329 ***
	(64.58)	(0.72)	(1.82)	(6.47)	(4.20)	(6.41)
control variable	yes	yes	yes	yes	yes	yes
industry-fixed effect	yes	yes	yes	yes	yes	yes
Year-fixed effects	yes	yes	yes	yes	yes	yes
R2	0.459	0.692	0.363	0.693	0.261	0.690
observed value	4357.000	4357.000	4357.000	4357.000	4357.000	4357.000

Note: *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively; numbers in parentheses are t-statistics; same below.

.

Columns (1)–(2) of Table 4 show partial mediating effects of digital economy on carbon emission reduction performance (CERP) through total factor productivity (TFP). Specifically, TFP has a positive and statistically significant coefficient of 0.045 (p < 0.01) on CERP, while the coefficient of digital economy on TFP reaches 0.138 (p < 0.01). These two results confirm Hypothesis 2: the digital economy enhances carbon reduction performance by systematically improving production efficiency.

These findings are also aligned with the results from Wang et al. [77], that the digital economy drives innovation enhancement performance through TFP. While both studies identify total factor productivity (TFP) as a critical mediating pathway linking digital economy to enterprises performance, they diverge in the key reason. This paper illuminates the digital economy-driven environmental performance through TFP, whereas Wang et al. [77] points to innovation-driven performance. Crucially, these findings collectively underscore the mediator of TFP under digital transformation, which simultaneously stimulates green transition and technological advancement.

Columns (3)–(4) of Table 4 reveal a positive and statistically significant coefficient of R&D investment (β = 1.721, p < 0.01) on carbon emission reduction performance (CERP), demonstrating that targeted environmental R&D expenditures in energy-intensive manufacturing enterprises effectively drives decarbonization. This linear relationship contrasts with Li et al.’s [59] inverted U-shaped relationship derived from a sample of all A-shares, highlighting industry specificity: R&D in carbon-intensive industries prioritizes emission reduction technologies, generating direct abatement returns that are absent in diversified samples. Furthermore, the regression coefficient of digital economy on R&D investment is also significantly positive. This indicates that the digital economy improves the technological level of enterprises and reduces energy waste by increasing R&D investment, thereby improving the carbon emission reduction performance, which verifies Hypothesis 3.

Whereas Ma et al. [78] identified a moderating effect of R&D investment in the digital economy-carbon mitigation nexus at the provincial level, our enterprise-level analysis of energy-intensive manufacturers reveals a distinct mediating mechanism. This divergence stems from the industrial imperative that energy-intensive enterprises, confronted with stringent decarbonization requirements, strategically channel digital-driven R&D investments (β = 1.721) into evolving emissions mitigation technologies [68], translating innovation inputs into direct carbon performance improvements rather than merely moderating existing relationships.

Column (6) reveals a statistically significant positive coefficient of technological innovation (p < 0.01) on carbon emission reduction performance (CERP). Furthermore, Column (5) demonstrates that the digital economy (DE) significantly enhances technological innovation (β = 8.681, p < 0.05). This dual-channel evidence confirms Hypothesis 4: digital transformation drives carbon mitigation by fostering innovation of resource and energy utilization efficiency, thereby systematically reducing carbon emissions.

While Li and Zhou [12] identified the carbon mitigation effects of the digital economy through technological innovation at the city level, this enterprise-level analysis specifically validates the tripartite relationship within energy-intensive enterprise. This micro foundational perspective is theoretically grounded the primacy of enterprises as innovation agents [69], which is further reinforced by China’s dual-carbon policy. Meanwhile, technological innovation of enterprises is prioritized to reduce carbon emissions [46]. Digital transformation empowers enterprises to radically enhance information transmission efficiency (β = 8.681 innovation boost), which accelerates breakthroughs in low-carbon technologies and clean energy systems [21]. This innovation catalysis effect provides empirical evidence for synchronizing digital progress with technological upgrading to achieve climate goals.

To further investigate the interaction of multiple mediating effects, this study incorporates the interaction term between total factor productivity (TFP) and R&D investment into the analysis as a mediating mechanism. Specifically, building upon the previously described mechanism, the term TFP * IIN is added as a mediator to demonstrate the partial mediating effects of the digital economy on carbon emission reduction performance (CERP) through both TFP and R&D investment. The results are shown in

Table 5. Mediating effect results (interaction term included).

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	TFP_GMM	CERP	IIN	CERP	TEI	CERP	TFP *IIN	CERP
TFP_GMM		0.045 ***
		(3.42)
IIN				1.721 ***
				(5.93)
TEI						0.000 ***
						(2.83)
TFP *IIN								0.363 ***
								(6.53)
DE	0.138 **	0.102 ***	0.012 ***	0.087 ***	8.681 **	0.107 ***	0.071 ***	0.082 ***
	(2.50)	(4.34)	(5.79)	(3.78)	(2.46)	(4.53)	(6.29)	(3.60)
constant term (math.)	5.791 ***	0.072	0.004 *	0.328 ***	32.400 ***	0.329 ***	0.019	0.328 ***
	(64.58)	(0.72)	(1.82)	(6.47)	(4.20)	(6.41)	(1.52)	(6.48)
control variable	yes	yes	yes	yes	yes	yes	yes	yes
industry-fixed effect	yes	yes	yes	yes	yes	yes	yes	yes
Year-fixed effects	yes	yes	yes	yes	yes	yes	yes	yes
R2	0.459	0.692	0.363	0.693	0.261	0.690	0.388	0.693
observed value	4357.000	4357.000	4357.000	4357.000	4357.000	4357.000	4357.000	4357.000

Note: *, **, *** denote significance at the 10%, 5% and 1% levels, respectively; numbers in parentheses are t-statistics; same below.

.

Columns (7)–(8) of Table 5 confirm a positive and statistically significant coefficient of the interaction term of TFP and R&D investment (β = 0.363, p < 0.01) on carbon emission reduction performance (CERP), demonstrating that the improvement of total factor productivity and targeted environmental R&D expenditures in energy-intensive manufacturing enterprises significantly drives carbon emissions reduction.

To ensure the robustness of the estimated mediating effects of total factor productivity (TFP), R&D investment, and technological innovation, this study conducts a bootstrap test with 1000 replications at the 95% confidence level. The results are presented in

Table 6. Bootstrap test.

		Observed Coefficient	BootstrapStd.err	z	p > z	Normal-Based [95%canf. Interval]
TFP_GMM	Indirect	0.059 ***	0.011	5.48	0.000	0.038	0.080
	direct	0.247 ***	0.041	6.06	0.000	0.167	0.327
	Total Eff	0.306 ***	0.039	7.76	0.000	0.229	0.383
IIN	Indirect	0.078 ***	0.011	7.17	0.000	0.057	0.100
	direct	0.228 ***	0.040	5.76	0.000	0.150	0.305
	Total Eff	0.306 ***	0.041	7.45	0.000	0.226	0.387
TEI	Indirect	0.022 ***	0.011	6.30	0.000	0.015	0.029
	direct	0.308 ***	0.041	7.54	0.000	0.228	0.388
	Total Eff	0.331 ***	0.041	7.53	0.000	0.226	0.386
TFP *IIN	Indirect	0.104 ***	0.013	8.08	0.000	0.078	0.129
	direct	0.203 ***	0.038	5.30	0.000	0.128	0.277
	Total Eff	0.306 ***	0.040	7.59	0.000	0.227	0.385

Note: replication is 1000 times at the 95% confidence level. *, *** denote significance at the 10%, and 1% levels, respectively; numbers in parentheses are t-statistics; same below.

.

The bootstrap test validates the hypothesized multiple mediation model. It reveals that the influence of the digital economy on CERP is partially transmitted through the individual channels of TFP R&D investment and technology innovation with the strongest effect occurring through the interaction and mutual reinforcement of these three factors.

Specifically, the bootstrap results confirm the significant mediating roles of TFP IIN and TEI individually. The indirect effects are positive and statistically significant (β = 0.059, p < 0.01 for TFP; β = 0.078, p < 0.01 for IIN;β = 0.022, p < 0.01 for TEI), as their 95% confidence intervals do not contain zero. These findings support Hypothesis 2 to Hypothesis 4, indicating partial mediation.

Crucially, the interaction term TFP*IIN demonstrates a highly significant and substantial indirect effect (β = 0.104, p < 0.01; 95% CI [0.078, 0.129]). This finding strongly suggests that the digital economy enhances CERP not only through TFP and IIN individually but, more importantly, through their synergistic interaction. The effect size of this interactive pathway is the largest among all mediators, underscoring its importance.

5.2. Heterogeneity Analysis

Industrial characteristics and geographic disparities significantly moderate the carbon mitigation performance of energy-intensive manufacturing enterprises in the context of the digital economy. This study employs a meticulous classification of enterprises according to industry categories and geographic disparities, followed by a comparative analysis of the carbon emission reduction performance of different enterprise groups.

5.2.1. Sub-Industry Analysis

To dissect the heterogeneous impact of the digital economy (DE) on Carbon Emission Reduction Performance (CERP), this study conducts a sub-industry analysis based on the official classification of the six major energy-intensive industries in China [25]. The regression results, systematically presented in

Table 7. Results of industry heterogeneity analysis.

	(1)	(2)	(3)	(4)	(5)	(6)
	C25	C26	C30	C31	C32	D44
	CERP	CERP	CERP	CERP	CERP	CERP
DE	0.118	0.116 ***	−0.035	0.012	0.360 ***	0.118 **
	(0.65)	(3.31)	(−0.88)	(1.06)	(4.68)	(2.33)
constant term (math.)	0.107	1.101 ***	1.031 ***	0.381 ***	1.887 ***	0.251 ***
	(0.42)	(12.27)	(18.65)	(28.01)	(10.58)	(4.58)
control variable	yes	yes	yes	yes	yes	yes
industry-fixed effect	yes	yes	yes	yes	yes	yes
Year-fixed effects	yes	yes	yes	yes	yes	yes
control variable	yes	yes	yes	yes	yes	yes
R2	0.179	0.218	0.508	0.751	0.512	0.054
observed value	169.000	1859.000	569.000	350.000	679.000	731.000

Note: **, *** denote significance at the 5%, and 1% levels, respectively; numbers in parentheses are t-statistics; same below.

, reveal significant disparities in the digital economic influence across these sectors.

The empirical findings indicate that the effectiveness of the digital economy in promoting CERP is not uniform across industries. Specifically, the digital economy coefficients for the petroleum processing, coking, and nuclear fuel processing industry (C25), as well as the production and supply of electric and heat power (D44), are statistically insignificant. This lack of significant effect can be explained from two primary perspectives. First, from a resource-based view, the prohibitively high capital investment required for digital equipment upgrades in these capital-intensive sectors creates a substantial barrier, discouraging enterprises from undertaking deep digital transformation. Second, from an organizational inertia perspective, the complex and entrenched production processes in these industries may generate greater resistance to change, resulting in slower and less effective adoption of digital technologies.

In stark contrast, the digital economy exhibits statistically significant and substantially larger positive coefficients in the Chemical Raw Materials and Chemical Products Manufacturing (C26), Smelting and Rolling of Non-Ferrous Metals (C32), and D44 industries compared to the baseline regression. The particularly strong effect observed in C32 (coefficient = 0.360) indicates a more pronounced promotional impact. This heterogeneity can be explained through the technology-organization-environment framework and innovation diffusion theory. The relative affordability and compatibility of digital solutions in these sectors likely lower the adoption threshold, thereby increasing enterprises’ willingness to implement technological improvements. The digital economy acts as a catalyst by enhancing energy efficiency through smart process control and optimizing internal production logistics, ultimately leading to more substantial carbon reduction gains.

A noteworthy example is the ferrous metal smelting and rolling processing industry (C31), where the DE coefficient is positive and consistent with the baseline results, although it is not the largest. This suggests that despite facing high equipment replacement costs similar to those in C25, regulatory pressure and the strategic imperative for sustainability may motivate firms in this highly polluting sector to continue incremental technological optimization efforts. This observation aligns with institutional theory, which posits that firms undertake environmental innovations, including digital ones, to gain legitimacy and comply with stringent regulatory requirements, even when the direct economic returns are not immediately evident.

In conclusion, this sub-industry analysis demonstrates that disparities in digital technology adoption and assimilation capabilities among the six energy-intensive industries are key determinants of their varying carbon emission reduction performances. The findings move beyond a monolithic view of the digital economy’s impact, highlighting how industry-specific structural factors, resource endowments, and institutional pressures interact to shape the pathways and effectiveness of digital decarbonization.

5.2.2. Sub-Regional Analysis

Due to regional differences in population, economic development, and other factors across China’s eastern, central, and western regions, enterprises in different areas have different sizes, development trajectories, business types, and levels of digital transformation [79].

Table 8. Results of the analysis of regional heterogeneity.

	(1)	(2)
	The East	Midwest
	CERP	CERP
DE	0.117 ***	0.108 ***
	(4.18)	(2.98)
constant term (math.)	0.416 ***	0.246 ***
	(6.02)	(3.48)
control variable	yes	yes
industry-fixed effect	yes	yes
Year-fixed effects	yes	yes
control variable	0.712	0.679
R2	2257.000	2100.000
observed value	0.103 ***
Coefficients Test of Difference Between Groups p-value	0.117 ***	0.108 ***

Note: *** denote significance at the 1% levels, respectively; numbers in parentheses are t-statistics; same below.

presents empirical results on the impact of the digital economy on the carbon emission reduction performance of energy-intensive manufacturing enterprises in these regions.

The estimated coefficient for DE is 0.117 (p < 0.01) in the eastern region, compared to 0.108 (p < 0.01) in the central-western region. Crucially, the test confirms that the difference between these coefficients is statistically significant at the 1% level. This validates that the regional disparity is not due to random chance, allowing for a meaningful theoretical interpretation of why the effect is more pronounced in the central and western regions.

This finding can be explained by combining regional development and technology diffusion theories. The concepts of the “Latecomer Advantage” and technology leapfrogging are particularly relevant: firms in the central and western regions are not burdened by legacy systems or sunk costs associated with older, carbon-intensive technologies that are more common in the industrialized eastern region. This freedom enables them to adopt the latest and most efficient digital and green technologies directly. Furthermore, from a knowledge spillover perspective, these regions can readily absorb and implement advanced production and management knowledge developed and refined in the eastern region, thereby accelerating their CERP improvements and narrowing the regional development gap.

Institutional Pressure and Legitimacy-Driven Change: The findings align with institutional theory. Enterprises in the central and western regions, which often depend on resource-intensive and heavily polluting industries, face increasing regulatory and social pressures to enhance their environmental legitimacy. The digital economy offers a viable and measurable pathway to achieve this compliance. Consequently, adopting digital technologies in these regions is not merely an efficiency-driven activity but also a strategic response to institutional demands, thereby amplifying its perceived impact on carbon reduction.

Resource Reallocation and the “Sailing Ship Effect” in Inland Development: The stronger effect observed inland reflects a strategic geographic shift in digital infrastructure and policy support, as noted in prior research [3]. This shift fosters a regional innovation system that increasingly facilitates digital transformation. For local firms, this external “push” alleviates traditional resource constraints. The significant coefficient indicates that, for these firms, marginal investments in digitalization yield higher returns in CERP by effectively addressing fundamental inefficiencies in their production processes. This phenomenon resembles the “sailing ship effect,” where late-stage entrants leverage new technologies to overcome core disadvantages.

In conclusion, the regional heterogeneity analysis reveals that the digital economy acts as a powerful equalizing force in China’s green development landscape. While it benefits all regions, its impact is most pronounced in the less developed central and western areas due to a combination of latecomer advantages, stronger institutional pressures for environmental legitimacy, and targeted regional digitalization policies. This underscores the importance of considering regional contextual factors when designing and implementing national digital decarbonization strategies.

5.2.3. Sub-Enterprise Size Analysis

To further investigate the differential impact of the digital economy (DE) on Carbon Emission Reduction Performance (CERP) across firms with varying resource endowments and capacities, this study stratifies the sample into small, medium, and large enterprises according to the classification standards of the National Bureau of Statistics [80]. The regression results, presented in

Table 9. Results of the analysis of firm size heterogeneity.

	(1)	(2)	(3)
	Small	Medium	Big
	CERP	CERP	CERP
DE	0.224 ***	0.019	0.108 **
	(5.20)	(0.60)	(2.36)
constant term (math.)	0.291 ***	0.280 ***	0.362 ***
	(2.73)	(2.89)	(4.85)
control variable	yes	yes	yes
industry-fixed effect	yes	yes	yes
Year-fixed effects	yes	yes	yes
control variable	0.641	0.649	0.806
R2	1089.000	2179.000	1089.000

Note: **, *** denote significance at the 5%, and 1% levels, respectively; numbers in parentheses are t-statistics; same below.

, reveal a striking and non-monotonic relationship between firm size and the efficacy of the digital economy in promoting carbon reduction.

The heterogeneity analysis reveals a nonlinear relationship between the digital economy and carbon emission reduction performance (CERP). The results indicate that the digital economy’s positive effect on CERP is strongest and most significant for small enterprises (β = 0.224, p < 0.01), significant but weaker for large enterprises (β = 0.108, p < 0.05), and statistically insignificant for medium-sized enterprises (β = 0.019).

This divergence can be theoretically explained by differences in firms’ resource endowments and organizational characteristics. Small enterprises leverage their organizational agility to rapidly adopt digital technologies, enabling “technological leapfrogging” and yielding higher marginal returns in emission reduction. Medium-sized enterprises may encounter a trap, where their resources are insufficient for comprehensive digital transformation, yet they have lost the flexibility characteristic of smaller firms [81]. Although large enterprises possess substantial resources, their structural inertia leads to a more incremental approach to digital transformation, thereby limiting the marginal effects.

These findings highlight the differential impact of the digital economy on enabling carbon reduction across firms of varying sizes, offering a theoretical foundation for developing tailored, size-specific policy measures.

6. Conclusions

This study systematically examines the impact of the digital economy on the carbon emission reduction performance (CERP) of China’s energy-intensive manufacturing enterprises, identifying key transmission mechanisms and heterogeneous effects. The main findings confirm that the digital economy significantly enhances CERP primarily by increasing total factor productivity (TFP), stimulating R&D investment (IIN), and promoting technological innovation. Furthermore, these effects are notably stronger in specific subsectors, such as the chemical (C26), non-metallic mineral (C30), and non-ferrous metal (C32) industries, and in the central and western regions compared to the eastern region.

Beyond these empirical validations, our findings provide concrete guidance for corporate strategy and public policy. For enterprise managers, the results emphasize that digital transformation is not merely an operational upgrade but a strategic lever for decarbonization. The identified mechanisms suggest that investments should be strategically directed toward digital tools that simultaneously enhance total factor productivity (TFP), research and development (R&D) investment and technology innovation outcomes. For example, implementing industrial Internet of Things (IoT) and AI-driven analytics can optimize resource allocation (thereby improving TFP) while generating data that accelerate innovation cycles. The pronounced effect observed in small and medium-sized enterprises (from heterogeneity analysis) indicates that smaller firms should view digitalization as a competitive advantage in the green transition, potentially leveraging it to close the sustainability gap with larger incumbents.

For policymakers, this research provides micro-level evidence to design more precise and effective interventions. The heterogeneous effects underscore the need to move away from one-size-fits-all approaches. The stronger regional impact observed in the central and western regions justifies prioritizing investments in digital infrastructure and fiscal incentives in these areas, as they yield higher marginal returns in carbon reduction. Similarly, policy support can be tailored to specific industries. For sectors such as C26 and C32, where the digital dividend is evident, policies could facilitate access to cloud computing and smart energy management systems. Conversely, for sectors with insignificant effects, the initial focus should be on addressing fundamental barriers, such as high retrofitting costs or a lack of technical expertise.

In conclusion, the digital economy serves as a critical enabler of the green transformation in energy-intensive industries. However, its effectiveness depends on a synergistic alignment between corporate digital strategies and nuanced public policies. Future research could explore cost–benefit analyses of specific digital technologies and investigate the roles of digital finance and data governance in supporting this transition.

This study is limited by the lack of enterprise-level emissions and energy data. Future research should adopt more precise metrics by integrating high-resolution data from mandatory reports, carbon registries, or IoT systems with detailed operational and supply chain energy data. The application of big data analytics and machine learning could then reveal deeper causal patterns. Furthermore, the analytical framework confined to pathways such as total factor productivity, R&D, and technological innovation, should be expanded to incorporate mechanisms like digital platforms for the circular economy, consumer behavior influences, smart logistics, and digital policy tools. These enhancements would improve the framework’s rigor and practical relevance.