Can Digital Transformation Reduce Enterprise Carbon Intensity? An Empirical Analysis of Chinese Manufacturers

Abstract

Reducing the carbon intensity of enterprises will help reduce greenhouse gas emissions, mitigate the negative impacts of global climate change, and protect the ecological environment. To this end, based on the data of A-share listed companies in China’s manufacturing industry from 2012 to 2022, the paper calculates the carbon emission intensity of enterprises, and at the same time, with the help of crawler technology, it crawls the keywords of digital transformation (DT) in the annual reports of the listed companies, portrays the intensity of DT of enterprises, and examines the impact of the level of digitization of enterprises on the carbon intensity of enterprises, along with the heterogeneous role and the mechanism of the role of the level of digitization of enterprises. The results of the study show that the digital revolution notably reduces the carbon emission intensity of enterprises. The inhibitory effect of DT is more significant for non-state-owned firms, industries with high market concentration, and regions with low environmental regulations. The findings of the mechanism test show that DT decreases the intensity of corporate carbon emissions by encouraging green innovation, with substantive green innovation being the main transmission channel behind strategic green innovation.

1. Introduction

In the last several years, global warming caused by CO₂ and other greenhouse gas emissions has become a typical climate problem confronted by countries and regions globally. The Sustainable Development Goals Report 2023: Special Edition, released by the United Nations, states that the speed and scope of the current efforts to combat climate change are entirely inadequate to properly handle climate variability and that extreme weather events are becoming more common and powerful. In addition, with an increase in the frequency and severity of extreme weather occurrences affecting every area of the planet, countries must take tougher measures to reduce emissions than those currently committed to in the Paris Agreement, or the world will warm by 2.5 to 2.9 degrees Celsius by 2030. Global carbon emissions reached a record 37.4 billion tons in 2023, up 1.1% from 2022, according to the International Energy Agency (IEA). Among them, China’s emissions will grow by about 565 million tons, by far the largest increase in the world, with emissions mainly coming from fossil fuel combustion, cement manufacturing, and other industrial sectors. Therefore, energy saving and carbon reduction in China’s manufacturing sector are particularly important for global climate governance [1]. As a conscientious emerging nation, China actively contributes to global climate governance. Compared with 2005, China’s administration has pledged to reduce carbon emissions per GDP unit by 60–65% by 2030, and the total emissions of carbon will peak by 2030, with efforts to achieve carbon neutrality by 2060. This implies that China will succeed in achieving the world’s highest rate of carbon reduction emission intensity in the quickest period of time in recorded history, and ensuring China’s transition to a low-carbon economy in the manufacturing sector and fulfilling the objectives of carbon neutrality and carbon peaking have become significant problems that need to be immediately researched and solved.

While carrying out the objectives of carbon neutrality and carbon peak, digital technology is key to promoting eco-friendly and low-carbon technology research [2]. Recently, thanks to the improvement and upgrading of China’s digital infrastructure, digital transformation with digital technology as the main content has penetrated and been integrated into the manufacturing industry in a wider field and at a deeper level, advancing the transformation in the manufacturing sector to smart manufacturing [3]. DT is the process of continuous improvement and optimization of business processes, management models, and customer experience by enterprises using digital technology, and it mainly includes a series of measures, such as reviewing the existing business model, introducing technological innovations, and adjusting the organizational structure and corporate culture. DT breaks through the institutional and institutional barriers that hinder the unrestricted movement of elements of production with the efficient flow of data, giving enterprises the possibility of using new productivity, which brings about the generation of new production methods and the innovation of business models. These undoubtedly create fresh chances for businesses to lessen the amount of carbon dioxide output [4]. The Research Report on the Development of China’s Digital Economy pointed out that 305 pilot demonstration projects on intelligent manufacturing and 420 projects on the application of new models have been implemented, and more than 700 intelligent factories and digitized workshops have been constructed. Among them, the digital transformation of the financial processing and smelting industry, the paper and printing products industry, and the special equipment manufacturing industry has made remarkable achievements. Specifically, Shandong Huifeng Petroleum Chemical Co., Ltd. (Zibo, China) is building a green chemical plant based on 5G, realizing the in-depth integration of 5G, AI, IoT, and other technologies with the chemical industry, and realizing the low-carbon and green development of the production process; Zhejiang Mizuda Textile Printing & Dyeing Technology Co., Ltd. (Huzhou, China) has established the Degree of Energy AIoT Energy Carbon Number Intelligent Platform and the Kaiwu Industrial Internet Platform to open up the data silos, realize energy monitoring, and greatly reduce energy consumption; and Guangzhou Techtech Textile Printing & Dyeing Technology Co. Ltd. (Guangzhou, China) relies on 5G technology to create multiple 5G application scenarios to improve operational efficiency and promote the digitalization and upgrading of the industry’s overall 5G application. However, in the process of China’s vigorous promotion of DT, many enterprises do not have a clear understanding of DT and even fall into the misconception that DT is only a technological challenge. However, while carrying out DT, technology is only a tool, and the system, processes, management elements, and operational mechanism are the core. Only by solving the above problems can enterprises carry out DT better and empower low-emission construction. Therefore, in the Chinese context, can digital transformation effectively curb businesses’ carbon intensity? Through what mechanism does it have an impact on businesses’ carbon intensity? The discussion of these questions is certainly one of great theoretical and practical implications.

Most previous studies have examined the elements impacting the carbon intensity of firms. Among the many factors, it is widely recognized in the academic community that green innovation is crucial for curbing corporate carbon emission intensity. As a major measure for coping with climate change, green innovation provides new support for energy conservation and lower emissions [5]. Chen et al. [6] considered green innovation to be the key to lowering carbon emissions. Compared with traditional technological innovation, it greatly lowers the intensity of carbon emissions of enterprises through end-processing technology upgrades, modernizing industrial structures and enhancing human capital [7]. In addition, green innovation can reduce carbon abatement costs and promote carbon productivity [8].

In recent years, given the rapid increase in digital technology, academics have gradually begun to observe the connection between DT and the intensity of carbon emissions, but no consistent conclusion has been reached. Most scholars are optimistic about the development of DT and believe that DT helps to curb corporate carbon emission intensity. Relying on data elements [9,10,11], digital technology possesses the qualities of high permeability, convergence, and innovativeness and can be used to suppress corporate carbon emission intensity by reducing incremental quantity and suppressing stock [12,13], two paths to suppressing corporate carbon emission intensity [14]. However, some scholars have questioned this, arguing that digital transformation may not suppress corporate carbon intensity. The “rebound effect” of digital technology, which is dependent on electricity consumption [15], is more likely to offset the environmentally friendly impacts of digital transformation [16], which is not advantageous for decreasing the intensity of carbon emissions. In addition, a few academics believe that there is a nonlinear relationship between digital transformation and the intensity of carbon dioxide emissions. Pre-digital transformation requires the consumption of a large amount of energy due to its own development, which will cause carbon emission intensity to increase in a short period of time, but further integration with the production process and the enhancement of its empowering effect on the economy can promote the overall transition to digitalization, intelligence intensification, and low carbon intensity [17,18,19].

This above literature has inspired this paper to some extent, but there are still the following areas that need to be expanded: first, the literature has examined the outcome of the digital economy in terms of the degree of carbon emissions on a large scale [20,21], and very little research has been conducted at the micro level. Although there are individual papers that have explored the link between DT and corporate carbon dioxide output [22,23], no study has focused on the carbon content perspective of manufacturing companies. As the largest source of carbon dioxide output in China, it is essential to conduct targeted research with the manufacturing sector as the subject of the study. Second, research on the mechanism by which DT impacts businesses’ carbon emission intensity is not complete in the literature, and it is mostly limited to the perspectives of technical breakthroughs, human resources, and industrial structure [2]. In fact, green innovation has a more significant role in the connection between DT and a company’s carbon footprint, and although the literature has explored the mechanism of green innovation, no study has focused on corporate green innovation motives or further classified them into two types—strategic and substantive—to study the mechanism and differential impact of different motives for green innovation. Third, due to the different manufacturing industry characteristics, the geographic location of the city where the businesses are located, and the different organic material endowments [24,25], the impacts of DT on the carbon intensity of firms may be quite different. However, these heterogeneous impacts may be overlooked in current research.

The potential breakthroughs and contributions of this essay are as follows. First, this paper focuses on the emission intensity of carbon from manufacturing companies and explores the effect of DT on the emission intensity of carbon from firms, which is a useful complement to the literature at the macro level. Second, according to the current literature, based on the perspective of innovation motivation, substantial and strategic green innovation are the two categories of green innovation, and the effect of DT on enterprise carbon dioxide output intensity is examined through different green innovation motivations, which offers factual proof for enriching the intrinsic logic of the influence of DT on enterprise carbon emission intensity. Third, the differential influence of DT on enterprise carbon emission intensity is further examined from the perspectives of enterprises, industries, and regions, providing strong theoretical support for more targeted digital transformation.

The body of the document is organized as follows: Section 2 provides the theoretical analysis and research theories; Section 3 presents the model and variables; Section 4 presents the actual outcomes and analysis; and Section 5 presents the conclusions and policy suggestions.

2. Theoretical Analysis and Study Hypothesis

2.1. DT and Enterprise Carbon Intensity

DT is favorable for changing businesses from a traditional production mode to a green production mode and it inhibits businesses’ carbon emission intensity. First, enterprises can use digital technology to process unstructured, nonstandard, and massive information data from within the enterprise [26], competitors, and markets to reduce the cost of information acquisition and management while carrying out production and operation [27], thereby helping enterprises integrate resources, improve production efficiency, and curb businesses’ carbon dioxide emission intensity. Second, firms can use digital docking platforms to achieve real-time integration of the entire process and digital technologies to monitor, collect, and evaluate the production and manufacturing of big data in energy-intensive industries [28,29]. At the same time, through the feedback function of data information, limited human, financial, material, and other resources in production activities are reconfigured to optimize the allocation and utilization of production factor resources [30,31], which reduces resource misallocation during the manufacturing process, thereby curbing the intensity of corporate carbon emissions. Finally, the increasing enhancement of digital infrastructure is also favorable for the introduction of new digital equipment by manufacturing enterprises, which not only eliminates old equipment with serious environmental pollution but also comprehensively transforms production and pollution control at the terminal links of firms through digital technology to bring forth the green metamorphosis of the whole chain of products and processes [32], thereby lowering the intensity of carbon emissions of enterprises. Then, the following theories are proposed:

Hypothesis 1.

DT helps curb corporate carbon intensity.

2.2. The Mechanism through Which DT Inhibits Enterprises’ Carbon Emission Intensity

DT can curb the carbon emission intensity of enterprises, mainly through increasing the level of green innovation [33]. DT continuously optimizes the green innovation process through data analytics and provides a digital joint R&D platform to stimulate collaborative innovation within and outside the enterprise and accelerate green technology breakthroughs [34]. First, DT continuously optimizes the innovation process by analyzing internal and external data and information and it drives a new paradigm of green innovation and R&D for enterprises. With the assistance of modern technologies, such as big data, artificial intelligence, and the internet, firms can absorb important information, including the external macroeconomy, market requisites, technological frontiers, and new opportunities for low-carbon transformation, in a comprehensive and timely manner, effectively reducing the sunk cost and failure risk of green innovation [35]. In addition, enterprises can use digital technology to assemble and evaluate internal environmental information, and to monitor and analyze data on energy use, carbon emissions, etc., in real time, exploring new ways to develop green technologies [36] and encouraging the growth of green business innovation. Moreover, DT provides a system, platform, and tools for the joint research and growth of firms’ green innovation and accelerates collaborative green innovation [37]. By building a digital joint R&D platform, the company has effectively gathered, integrated, and allocated innovative elements and resources, including capital, skills, technology, and data; promoted information-sharing and collaboration within the enterprise; and stimulated the vitality of green innovation. In addition, DT has given rise to a networked collaborative innovation ecosystem, accelerated the aggregation of different green R&D technologies among enterprises, and promoted breakthroughs in complex green key technologies. Many studies’ authors believe that successful catch-up often occurs when major breakthrough innovations emerge. Breakthroughs are often associated with the generation of new knowledge and new technological tracks or paradigms [18,38], which have caused the rise and decline of industries, the production of new products, and the process of producing new markets. Within the framework of DT, DT significantly lowers the price of gathering data and effectively increases the cash reserves of firms by strengthening the data circulation and incorporation of the supply chain and value chain [39]. When enterprises have sufficient capital flows, green substantive innovation activities with high technology content and high potential benefits will become the primary concern, and enterprises may further increase their investment in substantive green innovation to lock in market share and form their own technical barriers. However, tactical green innovation, similar to utility design, has comparatively low R&D capability requirements and a relatively small effect on output outcomes, which cannot significantly contribute to green and superior business development. Due to the openness of DT, the cost of firms searching for resources and knowledge through digital platforms will decrease, and the construction of a platform ecosystem will bring opportunities for enterprises that were difficult to integrate into the industrial revolution in the past. Therefore, Hypothesis 2 is proposed:

Hypothesis 2.

DT can help promote corporate green innovation, and its impact on strategic green innovation is greater than that on substantive green innovation.

3. Models and Variables

To investigate the effects of DT on the intensity of carbon emissions of firms, the following panel measurement model is constructed:

$${CEI}_{it}={\alpha }_0+{\alpha }_{1}DT+{\alpha }_{2}controls+{\nu }_i+{\gamma }_t+{\epsilon }_{it}$$

(1)

In Equation (1), i signifies the company and t signifies the time. CEI_it represents the dependent variable emission intensity of carbon; DT_it represents the independent variable digital transformation; controls is a run of control variables, including management expenses, fixed asset ratio, capital intensity, profitability level, proportion of independent directors, and enterprise size; and α₀ is an intercept term. α₁ is the regression coefficient of the primary explanatory variable. α₂ is the regression coefficient of the control variable.${\nu }_i$ and ${\gamma }_t$ denote the controlling firm and the annual effect, respectively. ${\epsilon }_{it}$ is a random error term.

3.1. Selection of Variables

3.1.1. Dependent Variable: Corporate Emission Intensity of Carbon

CEI is characterized by the proportion of a company’s carbon dioxide output to its operating income. Regarding the measurement of corporate carbon emissions, according to the practices [40], the overall carbon footprints of businesses are separated into carbon emissions directly and carbon emissions indirectly. Direct carbon emissions include combustion and energy emissions from the production process, solid waste incineration emissions, and fuel; and emissions caused by sewage treatment and greenhouse gases generated by land use conversion emissions. Such carbon emission data can be obtained from reports on sustainable development, and environmental and social responsibility. Indirect carbon emissions include emissions of greenhouse gases generated by emissions from raw material mining, and indirect emissions from the inflow and outflow of electricity, which cannot be directly obtained and need to be calculated from the usage of heat consumption, electricity consumption, and fossil fuels. The computation technique is as follows: $E=AD\times EF$, where AD is the activity level information on fossil fuels, which is acquired by multiplying the amount of fuel used by the typical low calorific value, and EF is the fossil fuel’s emission factor.

3.1.2. Independent Variable: DT

DT involves not only the application of digital technology but also the development of digital business philosophies and change strategies. Considering the number of sample companies and the availability of data, the frequency of words related to DTs within yearly reports can better reflect the extent of DTs in firms. Using the findings of the study [41], a comprehensive digital transformation index system was constructed by applying the text analysis technique to gage the extent of the DT of enterprises. The following are the steps: from the two dimensions of the application of underlying technologies and practical technical implementation, feature words were extracted from the text of the mentioned companies’ annual reports, and the frequency was searched, matched, and counted based on the feature words. Afterward, the frequency of key technology directions was categorized and collected, and the ultimate overall word frequency was created to construct an index system for the DTs of enterprises. The underlying technology application dimension includes artificial intelligence, blockchain, cloud computing, and big data, and the technology practice application dimension includes digital technology application. DTs include artificial intelligence, blockchain, cloud computing, and big data [42], which constitute the core underlying technology architecture of firm digitalization. Using digital technologies effectively measures the degree of implementation of DT and provides a foundation for the construction of digital words.

3.1.3. Control Variables

With reference to the research [43,44,45], the management expenses (Mange), fixed asset ratio (FA), capital intensity (CI), profitability level (Pro), proportion of independent directors (Indep), and enterprise size (Size) were selected as control variables to regulate the elements impacting businesses’ intensity of carbon emissions at the firm level as much as possible. Management expenses are represented by the assets invested in management affairs in a year, and the fixed assets ratio is characterized by the ratio of the firm’s net fixed assets to the enterprise’s total assets. Profitability is characterized by the ratio of the enterprise’s year-end total profit to its operating revenue. The proportion of independent directors is the proportion between the number of directors and the number of independent directors. The scale of the enterprise is represented by the usual number of workers when the year first started and the number of employees at the end of the year.

3.2. Data Sources

This study adopts the A-share manufacturing firms mentioned in Shenzhen and Shanghai as the research sample from 2012 to 2022. The manufacturing sector classification is based on the sector classification criteria of the China Securities Regulatory Commission, and all the original information comes from the CSMAR and Wind databases. See Appendix A for specific classification criteria for manufacturing industries. To ensure that the samples are representative, the study case was conducted in accordance with the following guidelines: (1) a sample with missing variable data was eliminated; (2) excluding the observation samples with abnormal financial data during the sample period, i.e., firms warned by PT, ST, and *ST, and excluding the serious samples with missing values, we finally obtained 12,242 observations for 1681 firms. Among all the samples, the average carbon emission was 626.0 thousand tons. (3) To avoid the effect of radical ideals, each continuous variable was reduced by 1% while excluding enterprises with more missing data. In addition, all variables were logarithmically treated to eliminate heteroskedasticity, and the results from the descriptive statistics for every variable are reported in

Table 1. Results of the descriptive statistics.

Variable	N	Mean	Sd	Min	Max
CEI	12,242	0.001	1.000	−3.187	0.906
DT	12,242	−0.003	0.991	−1.116	2.538
Mange	12,242	18.977	1.109	15.401	23.967
FA	12,242	0.194	0.099	0.020	0.455
CI	12,242	1.060	0.323	0.398	2.096
Pro	12,242	0.089	0.146	−0.626	0.483
Indep	12,242	0.318	0.038	0.134	0.588
Size	12,242	7.904	1.147	3.892	13.253
GP 1	12,242	0.548	0.979	0.000	7.062
GIP 2	12,242	0.378	0.814	0.000	6.594
GUP 3	12,242	0.337	0.717	0.000	6.080

¹ Green Innovation. ² Substantive green innovation. ³ Strategic Green Innovation.

.

4. Analyses and Empirical Findings

4.1. Results of Baseline Regression and Analysis

To investigate the effects of DT on the intensity of carbon emissions of firms, a fixed effect model is used for regression, and the results are displayed in

Table 2. Results of the benchmark regression.

Variables	(1)	(2)
DT	−0.011 ***1	−0.010 ***
	(−3.450)	(−3.096)
Mange		−0.020
		(−1.153)
FA		−0.031
		(−0.595)
CI		0.027
		(1.345)
Pro		−0.060 **
		(−2.398)
Indep		−0.004
		(−0.032)
Size		−0.002
		(−0.239)
Constant	0.000 ***	0.388
	(15.267)	(1.273)
Company FE	Yes	Yes
Year FE	Yes	Yes
Observations	12,242	12,242
R-squared	0.933	0.933

¹ Note: **, and *** denote significance at the 5%, and 1% levels, respectively, and robust standard errors clustered at the industry level are in parentheses.

. Column (1) displays the outcome of the univariate regression, and column (2) displays the inclusion of control variables at the firm level. In terms of statistical significance, the regression coefficients of digital transformation are all remarkably favorable at the 1% level, demonstrating that DT has a positive influence on the intensity of carbon emissions of enterprises. In terms of economic significance, when enterprise DT increases by 1 standard deviation, firm carbon emission intensity decreases by 0.011 and 0.010 standard deviations, respectively. This shows that DT helps to suppress the carbon emission intensity of enterprises, both in terms of statistical significance and economic significance, and Hypothesis 1 is verified. First, DT promotes the in-depth use of digital technology in connection with the manufacturing process and management style of firms, and it increases the effectiveness of resource factor utilization, and thus decreases carbon emissions [26]. In contrast, DT can help firms expand new business models, accumulate new kinetic energy for enterprises to achieve efficiency improvement and scale growth, and increase the total output level of enterprises.

4.2. Endogenous Processing

Although an array of control factors has been added in this work to minimize the negative impact of endogeneity problems, this paper may still suffer from endogeneity issues, such as reverse omitted and causation variables, which make the parameter estimates biased and inconsistent. On the one hand, the government’s “dual control” regarding the overall quantity and strength of carbon emissions may force high-carbon-emitting firms to undergo DT. In contrast, whether firms execute DT depends on the awareness of enterprise managers and the recognition of the need for change, so it is certainly challenging to avoid the influence of human factors in DT. In addition, various factors, such as the US–China trade dispute and the Russia–Ukraine war, have exacerbated the uncertainty of economic development to a certain extent, which may also have an effect on the DT of enterprises. Therefore, to address the possible endogeneity issue of the model, it is necessary to find appropriate instrumental variables.

The choice of instrumental variables must follow the fundamentals of correlation and exogeneity. Specifically, instrumental variables should be highly connected with internal explanatory factors but not associated with phrases of disruption. In light of this, this study references Cheng et al. [46] and Dai and Yang [47], who adopted how many post offices there were per million persons in 1984 (Iv1) and the average value of DT of other companies in the same sector in an identical year (Iv2) as instrumental variables. On the one hand, post offices as an infrastructure can be used to communicate and transfer information while carrying out DT, and the historical telecommunication infrastructure in each region will affect the DT of firms in the subsequent stage through various factors, such as the level of technology and the use of habits. The average amount of DT of other firms in the same sector during the same year can show the extent of information technology needs of the firms in the industry, which will have a direct effect on the firms’ decision-making over DT and satisfy the requirement of relevance for this choice of instrumental variables. The relevance requirement for the selection of instrumental variables is met. The historical telecommunication infrastructure in each region does not directly influence the intensity of carbon emissions of its current firms, and the average value of DT of other firms will not have an immediate effect on the carbon emission intensity of a certain firm, which meets the exclusivity requirement of the instrumental variable. Therefore, the choice of these is theoretically appropriate.

It should be noted that since the instrumental variable for the number of post offices per 100 people in 1984 chosen for this paper is cross-sectional, it cannot be used directly in the econometric study of panel statistics. For this reason, the approach for this problem introduces a temporal fluctuations variable to construct the panel instrumental variable. In particular, the instrumental variable for DT is constructed by constructing a word of interaction between the number of post offices per million people in 1984 and the level of digital transformation in the previous year for each city.

The regression results of the instrumental variables approach are shown in

Table 3. Regression findings for instrumental variables.

Variables	(1)	(2)
	DT	CEI
Indep	−0.545 ***	0.089
	(0.293)	(0.143)
Size	0.099 ***	−0.009
	(0.021)	(0.011)
KP-LM statistic	7.841 **
Cragg–Donald Wald F statistic	104.046
Hansen J statistic p value	0.012
Company FE	Yes	Yes
Year FE	Yes	Yes

** and *** denote significance at the 5% and 1% levels, respectively, and robust standard errors clustered at the industry level are in parentheses.

. KP-LM is noteworthy at the 5% level, denying the initial theory of inadequate recognition of instrumental factors, and the CDF is more than the Stock–Yogo weak instrumental variables based on an F test at the 10% significance level of the crucial importance, denying the initial hypothesis of weak instrumental variables, which shows that the instrumental variables selected in this paper are reasonable. Column (1) shows these regression outcomes for the first stage, and the coefficient of the instrumental variable is remarkably favorable at the 1% level. Column (2) indicates the regression outcomes for the second stage. After mitigating the possible endogeneity problem, the coefficient of the enterprise’s DT is significantly negative at the 10% level, which indicates that DT indeed reduces the enterprise’s carbon intensity. The regression baseline results of this paper are valid after mitigating endogeneity using the instrumental variable approach.

4.3. Robustness Tests

To guarantee the reliability of the findings, robustness assessments were carried out by swapping the explanatory variables, swapping out the explanatory factors, excluding the sample of municipalities, and changing the study period. (1) Replacement of the dependent variable. Using the technique of Peng and Li [48], the adjustment coefficient W, i.e., the weight of carbon emissions in each province, is calculated first: $W=(P_i/\sum P_I)/(O_i/\sum O_i)$, where $P_i$ is the amount of carbon released by province $i$; $\sum P_i$ is the total national carbon emission; $O_i$ is the overall manufacturing output value of province $i$; and $\sum O_i$ is the national gross manufacturing resultant value. Then, the weighted-adjusted carbon emissions of the province are calculated as $=W\times Y_i$, where $Y_i$ is the original carbon emissions of province $i$. Then, we calculate the carbon emissions of firm $k$ in province $em={em}_i(Q_k/\sum Q_k)$, where $Q_k$ is the output of firm $k$, and $\sum Q_k$ is the overall result of the province in which company $k$ is situated. The proportion of business revenue to carbon emissions is calculated in this way as a variable to replace the explanatory variables, and the estimation results are displayed in column (1) in Table 3. (2) Substitution of independent variables. Drawing on Wang et al. [49], this research arranges and identifies year-end intangible asset line items according to the information disclosed by listed firms’ financial reports’ notes to determine the portions that are related to digital technology. The percentage of these relevant components to all intangible assets is calculated as a proxy indicator of the extent of DT of the firm (DTB). The regression is also rerun with this replacement of the results of the estimation, and the explanatory variables are shown in column (2) in Table 3. (3) Fixed-time and provincial shock. Adverse financial events can cause firms to have illiquid stocks [50], which ultimately may result in impediments to DT. Considering the presence of financial shocks and major health and other event shocks during the study period of this document, to avoid endogeneity problems caused by such major factors, the fixed-time and province cross-multiplier terms are estimated, as shown in column (3) in Table 3. (4) Municipalities are excluded from the sample. Considering that municipalities are provincial administrative units immediately beneath the management of the national authority, their economic foundation and development conditions are different from those of other cities; four cities, namely, Beijing, Shanghai, Chongqing, and Tianjin, were removed from the sample and regressed once more, and the estimation outcomes are shown in column (4) of Table 3. As observed from

Table 4. Robustness test results.

Variables	(1)	(2)	(3)	(4)
	Replacement of the Dependent Variable	Replacement of Independent Variables	Fixed-Time and Provincial Shock	Excluding Municipalities
DT	−8.691 ***		−0.009 ***	−0.012 ***
	(−4.550)		(−3.036)	(−3.679)
DTB		−1.258 *
		(−1.751)
Mange	−0.009	−0.020	−0.017	−0.009
	(−0.477)	(−1.125)	(−1.081)	(−0.477)
FA	−0.013	−0.035	−0.022	−0.013
	(−0.211)	(−0.619)	(−0.438)	(−0.211)
CI	0.039 *	0.023	0.030	0.039 *
	(1.929)	(1.145)	(1.529)	(1.929)
Pro	−0.050	−0.062 **	−0.061 **	−0.050
	(−1.544)	(−2.411)	(−2.629)	(−1.544)
Indep	0.015	−0.013	−0.017	0.015
	(0.106)	(−0.106)	(−0.150)	(0.106)
Size	−0.010	−0.003	−0.005	−0.010
	(−0.904)	(−0.327)	(−0.611)	(−0.904)
Constant	0.222	0.410	0.354	0.222
	(0.717)	(1.322)	(1.295)	(0.717)
Company FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Year × Province FE	No	No	Yes	No
Observations	12,242	12,242	12,242	9940
R-squared	0.933	0.933	0.935	0.933

*, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively, and robust standard errors clustered at the industry level are in parentheses.

, DT significantly suppresses corporate carbon intensity at least at the 10% level, either by changing the variables that explain the pattern, swapping out the explanatory factors, excluding the sample of municipalities, or changing the study period. Although the magnitude of the regression coefficients is slightly different from that of the previous estimation, the sign of the coefficients is exactly the same, which suggests that the strength of the benchmark regression results is good.

4.4. Analysis of Heterogeneity

4.4.1. Firm Heterogeneity

There are significant differences in the business objectives, incentives, and forms of supervision between both state-owned and privately held businesses due to different ownership systems [51]. Does the effect of DT vary depending on the intensity of carbon emissions of enterprises with different ownership systems? This study splits the research sample into non-state-owned and state-owned businesses and then analyzes the impact of the character of enterprise ownership on the outcomes of the regression, which are displayed in columns (1) and (2) of

Table 5. Heterogeneity regression results.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	State-Owned Enterprises	Non-State-Owned Enterprises	Low Concentration Ratio	High Concentration Ratio	Low Environmental Regulation	High Environmental Regulation
DT	−0.007	−0.011 ***	−0.000	−0.014 *	−0.021 ***	0.001
	(−0.670)	(-3.378)	(−0.032)	(−1.746)	(−4.490)	(0.169)
Mange	0.013	−0.034 *	0.008	−0.042	−0.002	−0.034
	(0.362)	(−1.799)	(0.629)	(−1.217)	(−0.144)	(−1.693)
FA	−0.090	0.008	−0.131 *	−0.019	0.061	−0.156 **
	(−0.825)	(0.143)	(−1.829)	(−0.214)	(0.717)	(−2.070)
CI	0.002	0.034	0.047	0.003	0.011	0.044*
	(0.039)	(1.525)	(1.660)	(0.122)	(0.470)	(1.799)
Pro	−0.027	−0.071 ***	−0.041	−0.089 **	−0.002	−0.111 **
	(−0.264)	(−2.905)	(−1.010)	(−2.557)	(−0.055)	(−2.628)
Indep	−0.080	0.090	−0.025	0.094	0.057	−0.138
	(−0.295)	(0.680)	(−0.169)	(0.507)	(0.392)	(−0.940)
Size	−0.003	0.004	−0.043 ***	0.019	−0.018 *	0.014
	(−0.221)	(0.331)	(−2.972)	(0.786)	(−1.767)	(1.159)
Constant	−0.125	0.534 *	0.195	0.659	0.155	0.554
	(−0.175)	(1.726)	(0.758)	(1.292)	(0.691)	(1.462)
Company FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Observations	2532	9666	5883	5916	6049	6118
R-squared	0.917	0.937	0.942	0.934	0.934	0.935

*, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively, and robust standard errors clustered at the industry level are in parentheses.

. According to columns (1) and (2) in Table 5, the digital transformation coefficients estimated for non-state-owned businesses are greater and considerably negative at the 1% level, and the estimated coefficients of DT for state-owned enterprises are not significant. This may be because non-state-owned businesses are profit-oriented and distribute benefits based on contributions, encouraging personnel to enhance efficiency, improve technology, and promote innovation [52]. Moreover, a few privately held businesses have a certain foundation in the field of digitization, and they can use digitization to stimulate their own potential and contribute to the process of suppressing businesses’ carbon intensity. Conversely, while state-owned manufacturing firms are not entirely profit-oriented and their own survival pressure is small, compared with non-state-owned manufacturing businesses, their innovation power is insufficient, and their vitality is insufficient [53]. Thus, it is difficult for digital transformation to reduce businesses’ carbon emission intensity.

4.4.2. Industrial Heterogeneity

Industry concentration will cause agglomeration effects, the concentration of production factors and labor markets, further input sharing and knowledge spillovers, and scale effects. However, as industry concentration increases, enterprises face limitations in terms of market capacity, geographic space, and resources, creating a crowding effect that inhibits improvements in financial effectiveness. Therefore, within the framework of different industry concentrations, the effects of the shift to digital technology on the carbon intensity of enterprises may be different. According to previous studies [54], total assets are utilized to calculate the degree of industry concentration using the Herfindahl index, and the formula is shown below:

$$HHI=\sum _{t=1}^n{\left(\frac{X_i}{X}\right)}^2$$

(2)

$$X=\sum _{t=1}^n{X}_i$$

(3)

$X_i$ is the annual assets of firm $i$ in the industry, and $n$ is the quantity of businesses in the sector. When a firm’s HHI value is above the overall HHI median, it falls within the category of a low industry concentration firm; otherwise, it is categorized as a high industry concentration firm. The outcomes are shown in columns (3) and (4) of Table 5. According to columns (3) and (4) of Table 5, the estimated coefficients of DT for firms with high industry concentration are not significant, and the calculated digital transformation coefficients for firms with high industry concentration are larger and notably negative at the 10% level. This may be because enterprises can more conveniently and quickly exchange and transfer input elements, market information, industrial technology, and other content through the construction of cooperation mechanisms [55], accelerate the speed of enterprise digital transformation through the continuous improvement of digital technologies, and inhibit the intensity of enterprise carbon emissions. Enterprises with low industry concentration have less information and data in the same industry and cannot fully utilize digital platforms for communication and upgrading [56], so there is little impact from reducing businesses’ carbon emission intensity.

4.4.3. Variability in the Level of Environmental Control

To protect the environment, several actions that damage public space are regulated under environmental regulation. With the implementation of the new development concept in recent years, the effects of environmental regulations on the mode and path of suppressing carbon emission intensity have attracted the attention of academics. Therefore, to investigate the effects of DT on the carbon emission intensity of enterprises under different environmental regulatory pressures in greater detail, with reference to the study of He and Luo [57], the degree of environmental control at the provincial level is gauged by the Completed Manufacturing Pollution Control Investment for every 1000 yuan of industrial added value. When the business’s environmental regulation intensity is valued in the province or city above the median, the enterprise will be classified as having a high level of environmental control; otherwise, the enterprise will be classified as having a low degree of environmental regulation. The outcomes are displayed in columns (5) and (6) of Table 5. According to columns (5) and (6) of Table 5, the estimated coefficients on digital transformation for firms with low environmental regulation are considerably negative at the 1% level, while the estimated coefficients on DT for firms with high environmental regulation are not significant. This could be because high environmental regulation causes firms to increase their environmental governance activities, cost expenditures to increase significantly [58], and the carbon intensity of firms to increase in the short term. In regions with low environmental regulation, enterprises face less pressure from environmental governance [59], and their cost expenditures are relatively more stable, which is favorable for businesses’ long-term green development; therefore, a major part of reducing firms’ carbon emission intensity is digital transformation.

5. Further Discussion: Mechanism Analysis

5.1. Mechanism Modeling

The results show that DT significantly suppresses a company’s carbon footprint. Then, through what channels does DT inhibit enterprises’ carbon emission intensity? To further clarify the connection between the two, this paper further examines the channels through which DT suppresses a company’s carbon footprint intensity from the standpoint of what drives green innovation, on the basis of the theoretical analysis in the previous paper, and it constructs the following model [60]:

$$M_{it}={\beta }_0+{\beta }_1{DT}_{it}+{\beta }_{2}controls+{\vartheta }_i+{\gamma }_t+{\epsilon }_{it}$$

(4)

$M_{it}$ in Equation (4) is the mechanism variable, which includes green innovation, green substance innovation, and green strategy innovation.${ \beta }_0$ is the intercept term.${ \beta }_1$ is the explanatory regression coefficient variable. ${\beta }_2$ is the control regression coefficient variable. The remaining variables are defined as in Equation (1).

This study uses listed firms’ green patent application data as a mechanism variable of green innovation. The main reason for using it as a mechanism variable is that compared to R&D inputs, green patents most naturally represent the results of businesses’ green technological innovation activities, and patents have a clear technological categorization, reflecting that the majority of green patents have distinct value implications and that they actively represent the contributions of innovation activities [61]. To further explore the processes underlying the different types of green innovation, this paper refers to Du and Guo [62] and Lou et al. [63] and refers to “high-quality” innovation behaviors that aim to promote technological advancement and gain competitive advantages for the purpose of substantive innovation; innovation behaviors that aim to seek other benefits by pursuing “quantity” of innovation; and innovation behaviors that seek “quantity” of innovation by pursuing other benefits for the purpose of “quantity” of innovation. The innovation strategy of pursuing “quantity” and “speed” of innovation to satisfy regulation and the government is referred to as strategic innovation [64]. How businesses apply for green invention patents is recognized as green substantive innovation, and the act of firms requesting patents for green utility models is recognized as green strategic innovation.

5.2. Mechanism Analysis

The regression outcomes of the effect of DT on green technology innovation mechanisms are presented in

Table 6. Mechanistic regression results.

Variables	(1)	(2)	(3)
	Overall Green Innovation	Substantive Green Innovation	Strategic Green Innovation
DT	0.061 ***	0.054 ***	0.034 *
	(3.715)	(4.924)	(1.991)
Mange	0.005	−0.006	0.027
	(0.194)	(−0.240)	(1.352)
FA	0.020	0.046	0.069
	(0.098)	(0.318)	(0.402)
CI	−0.065	−0.036	−0.043
	(−1.507)	(−0.942)	(−1.380)
Pro	0.055	0.052	0.044
	(1.646)	(1.449)	(1.304)
Indep	0.494 **	0.510 **	0.100
	(2.208)	(2.746)	(0.522)
Size	0.081 ***	0.067 ***	0.039 **
	(3.955)	(4.126)	(2.499)
Constant	−0.296	−0.173	−0.493
	(−0.518)	(−0.317)	(−1.061)
Company FE	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
Observations	12,242	12,242	12,242
R-squared	0.753	0.747	0.695

*, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively, and robust standard errors clustered at the industry level are in parentheses.

. The variables listed in columns (1), (2), and (3) are the green innovation level, strategic green innovation, and substantive green innovation, respectively. The results in columns (1) and (2) indicate that the calculated coefficients of DT are all substantially positive at the 1% level, demonstrating that DT is able to inhibit the carbon emission intensity of businesses by increasing the total amount of green inventions and substantive green inventions. The primary justification is that in light of digitization, enterprises can utilize fully digital platforms to fully collect green innovation-related information and can quickly distribute green innovation information within and between enterprises through the real-time characteristics of data elements [65], which can enrich firms’ green invention resources and information and improve overall green innovation in enterprises. As a result, manufacturing companies’ carbon emission intensity is lower. However, compared with strategic green innovation, substantive green innovation can bring more production and development advantages to enterprises. Digital platforms provide enterprises with communication and learning opportunities with research institutes, universities, and other research and incubation centers, and enterprises choose to take advantage of learning opportunities to improve their substantive green innovation capabilities [66].

6. Conclusions and Recommendations

6.1. Conclusions

On the basis of exploring the effect of DT on the carbon emission intensity of businesses, the mechanism and effect of DT on the carbon emission intensity of firms are empirically examined using manufacturing sector A-share-listed firm data from Shenzhen and Shanghai, China, from 2012 to 2022. DT significantly reduces the intensity of carbon emissions of firms; it reduces the carbon intensity of firms mainly by lowering their carbon emissions and increasing their total output, and there is enterprise, industry, and regional heterogeneity in this promotion. The possible endogeneity problem still holds and passes a series of robustness tests. In non-state-owned businesses, the effect of DT on the carbon footprint intensity of the corporation is noteworthy for enterprises with high industry concentration and a high level of environmental control, while it does not matter for state-owned businesses, businesses with a minimal degree of environmental regulation, or businesses with little industry concentration. The results of the mechanism analysis suggest that DT decreases the carbon intensity of corporations by encouraging green inventiveness, with substantive green inventiveness being the main transmission channel compared to strategic green innovation. This also means that businesses can leverage digital platforms, because DT provides enterprises with communication and learning opportunities with research institutes, universities and other research and incubation centers, and enterprises can choose to take advantage of learning opportunities to improve their substantive green innovation capabilities.

This paper considers the mechanism of green technological innovation according to the different purposes of technological innovation, and divides green technological innovation into different types, which provides scholars with new perspectives for in-depth research on the mechanism of technological innovation and expands the theoretical research on green innovation. Secondly, most scholars explore the impact of digital transformation on corporate carbon emissions from the enterprise level, or study the impact of digital transformation on the intensity of corporate carbon emissions from the macro level [67]. This paper explores the impact of digital transformation on the carbon emission intensity of enterprises from the perspective of microenterprises, and enriches the relevant theories. Finally, this paper analyzes the impact of digital transformation on carbon emission intensity from the perspective of input and output, and analyzes the theoretical logic of digital transformation to reduce the carbon emission intensity of enterprises.

The practical significance of this paper lies in the following: (1) this paper discovers the role mechanism of substantive green innovation in digital transformation to reduce the carbon emission intensity of enterprises through the study, which provides a realistic basis for the government and enterprises to reduce the carbon emission intensity of enterprises, and enables the government and enterprises to improve the level of innovation in a targeted manner; (2) the heterogeneity analysis in this paper helps enterprises to utilize digital transformation means to reduce the carbon emission intensity of enterprises according to the nature of their own property rights, the current situation of industry development, and the regional environment; and (3) it provides help for China’s realization of dual-carbon goals and sustainable development.

6.2. Management Insights

The results of this study are extremely important for understanding and assessing DT, formulating a rational plan for reducing the intensity of carbon emissions of enterprises from this perspective of combining the government and the market, effectively combining the degree of DT, and promoting the DT and upgrading of traditional enterprises, as well as realizing China’s “dual-carbon” goals. In light of the aforementioned conclusions, the following legislative suggestions are put forth to advance DT and reduce businesses’ carbon emission intensity:

First, the extent of DT should be accelerated. China is currently undergoing a crucial transition from a “manufacturing power” to a “manufacturing power”. Namely, in the past few decades, China has become one of the largest manufacturing countries in the world through massive investment, technology introduction, and labor advantages. The products produced in China cover a wide range of fields from low-end to high-end and are exported to all over the world. However, despite the large scale of China’s manufacturing industry, there is still a gap between its technological innovation capacity and that of developed countries at this stage, and its consumption of resources and environmental impact are also greater. Thus, in order to realize the transformation from a “manufacturing power” to a “manufacturing power”, China has become one of the largest manufacturing countries in the world. China’s excessive pollution and energy consumption problems need to be urgently addressed in order to solve the problem of excessive energy use and pollution caused by the traditional industrial development model as early as possible, and to encourage economic excellence in growth and low-carbon development. The government should first introduce appropriate policy guidance to assist DT in significant research and development projects, build a platform for enterprises to exchange information on DT, help enterprises solve the practical difficulties encountered in DT and upgrading, and concurrently curb the intensity of companies’ carbon footprints. Second, the R&D investment in DT-related technologies should be increased; the number of DT-related subject projects should be increased; the subject mode should be strengthened to increase the DT direction of industry, academia, and research projects; the level of DT of enterprises should be improved; and the intensity of a company’s carbon footprint should be inhibited. Finally, reducing the cost of introducing digital infrastructure to enterprises, guiding enterprises to eliminate old equipment with serious environmental pollution, and carrying out comprehensive transformation of enterprise production and end-to-end pollution control and other links through digital technology, for example, by means of policy subsidies, fee reductions, and other incentives to guide businesses, should quicken the pace of DT and thus curb the amount of carbon emissions from enterprises.

Second, the effect of DT on lowering the intensity of carbon emissions should be considered “according to local conditions”. Non-SOEs should continue to promote scientific and technical advancement, accelerate the DT process, and fully utilize digital technology and other digital means to stimulate their own green potential and help curb the intensity of their carbon emissions. State-owned enterprises can make targeted adjustments to their digital strategy according to the specific conditions of their assets, businesses, and organizations; simplify their internal systems; and fully communicate and collaborate with their business and technical departments to promote DT and reduce the intensity of carbon emissions. For firms with low industry concentration, the government should use preferential policies to attract enterprises to enter, mainly by reducing industry barriers, controlling the minimum price of products, and increasing the government’s purchase of products in a variety of ways to improve industry concentration. This in turn is conducive to exchanging and transmitting input elements, market information, industrial technology, and other content between enterprises through digital platforms and accelerating the DT of enterprises through iterative upgrading of digital technology speed and curbing businesses’ carbon emission intensity. For enterprises with a high degree of industry concentration, enterprises are guided to utilize cloud computing, big data, the internet, and other techniques to reengineer existing processes, promote industrial transformation and upgrading, and further support the “scale effect” and “knowledge spillover effect” brought about by industry concentration to improve the level of DT and curb the intensity of carbon emission carbon emissions. For firms located in areas with strict environmental regulations, a better assessment and evaluation system and regulation oversight mechanism for the environment should be developed, and the requirements of rules pertaining to environment enforcement should be met to avoid softening of the environmental regulation system and to ensure that the effect of DT on reducing the intensity of carbon emissions will remain the same. Firms located in regions where there are few environmental regulations increase the intensity of environmental regulations in the region so that environmental regulations become a push for firms to improve to a certain level of DT and enhance the ability to inhibit the intensity of corporate carbon emissions. For enterprises in violation of policies for environmental protection, the government can continue to increase pollution charges and increase the penalty for noncompliance, and the use of DT to inhibit the intensity of carbon emissions. Other eco-innovations should be carried out, recognized, and incentivized.

Third, enterprises should promote substantive green innovation. The government should be patient enough to subsidize enterprise innovation, especially substantive green innovation, and use subsidy funds to direct operating firms to shift to substantive green innovation. Local governments correct the “innovation concept”, abandon the “innovation face project”, insist that it is better to have fewer than more, and lead businesses from pursuing strategic green innovation to pursuing substantive green innovation. Firms’ investment in R&D should focus on strengthening substantive green innovation with DT; utilizing digital technologies, digital platforms, and digital management methods; forming a large number of substantive green patented technologies; and then mastering a number of key green technologies with core competitiveness and independent intellectual property rights to strengthen the ability of sustained development, contribute to the realization of China’s dual-carbon aim, and seize the commanding heights of industry development in the future.