Keyword co-occurrence analysis.

A keyword co-occurrence network for Chinese literature was constructed based on the CNKI database (Fig 2). The network comprises 156 nodes and 157 links, with a network density of 0.013. The largest connected subgraph contains 101 nodes, accounting for 64% of the entire network, indicating that domestic research has formed a relatively clear structure of thematic interconnections. Keywords such as “green innovation,” “financing constraints,” “digital economy,” and “corporate innovation” occupy key positions in the network. Among these, “green innovation” embodies the technical logic behind the transformation of digital resources into green outcomes, while “financing constraints” reflect a critical mechanism through which digital transformation impacts enterprises’ ability to allocate resources. Meanwhile, the presence of keywords such as “media attention,” “corporate governance,” “information disclosure,” and “social responsibility” indicates that some studies have begun to focus on the governance effects of digital transformation; however, related research still primarily relies on traditional governance perspectives.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. CNKI keyword co-occurrence map.

https://doi.org/10.1371/journal.pone.0343345.g002

A co-occurrence network of keywords from English-language literature was constructed based on the Web of Science Core Collection (Fig 3). This network comprises 515 nodes and 1,979 links, with a network density of 0.015. The largest connected subgraph contains 450 nodes, accounting for 87% of the total, indicating that the international research network exhibits stronger interconnectivity and thematic expansiveness. Keywords such as “performance,” “technology,” “innovation,” and “sustainable development” occupy the core of the network, while digital technology themes such as “big data” and “the Internet” are closely associated with them. This suggests that international research also focuses on the application of digital technologies, the enhancement of innovation capabilities, and the improvement of sustainable performance. Although governance themes such as “corporate social responsibility” and “information disclosure” are touched upon, systematic research centered on the governance functions of digital information has not yet emerged.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. WOS keyword co-occurrence map.

https://doi.org/10.1371/journal.pone.0343345.g003

Overall, existing research hotspots are concentrated on the application of digital technologies, the enhancement of innovation capabilities, and the improvement of corporate sustainable performance. Although governance-related themes are gradually attracting attention, they have not yet become a core research direction in this field.

Cluster analysis.

Building on the co-occurrence analysis, cluster analysis was further conducted to identify the thematic structure of this field. The clustering results for Chinese literature (Fig 4) show a network modularity Q = 0.8286 and a weighted average silhouette coefficient S = 0.9701, indicating that the cluster structure possesses good stability and internal consistency. The analysis identified thematic clusters primarily centered on green innovation, financing constraints, media attention, internal governance, information disclosure, and corporate innovation. In terms of thematic relationships, domestic research has established a value realization logic centered on technological innovation, extending to mechanisms such as resource allocation and corporate governance. Among these, green innovation and corporate innovation reflect the transformation of digital resources into technological achievements; financing constraints reflect the impact of resource acquisition and allocation; and information disclosure and internal governance pertain to governance effects. However, there remains a lack of effective linkage between the themes of technological innovation and governance mechanisms, and the ways in which digitalization achieves ESG value enhancement through information mechanisms have not yet been fully explained.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. CNKI Chinese literature keyword clustering map.

https://doi.org/10.1371/journal.pone.0343345.g004

The clustering results for the English-language literature (Fig 5) show a network modularity of Q = 0.5397 and a weighted average silhouette coefficient of S = 0.8143, indicating reliable clustering results. The analysis primarily formed thematic clusters such as “internet,” “performance,” “corporate social responsibility,” “management,” “innovation,” “CO₂ emissions,” and “renewable energy,” covering areas including the application of digital technologies, corporate social responsibility, environmental performance, and energy transition. Although the research topics span multiple fields—including innovation, responsibility, and environmental governance—their common logic remains the view of digitalization as a tool for enhancing corporate capabilities and improving sustainability performance. The existing thematic framework has not yet fully revealed how digital resources achieve ESG value enhancement by transforming the corporate information environment and external governance relationships.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. WOS English literature keyword clustering map.

https://doi.org/10.1371/journal.pone.0343345.g005

Literature review and the focus of this study.

The above analysis of the knowledge structure indicates that existing research has established a relatively clear framework oriented toward capability enhancement, namely, that digital resources primarily achieve corporate sustainability by promoting innovation capabilities, driving green innovation, and improving environmental performance. However, there remains room for expansion in two areas.

First, existing research primarily focuses on the technological transformation process through which digital transformation promotes green innovation, while paying insufficient attention to how digital resources create value in different industrial contexts. The production systems of energy-intensive enterprises are characterized by significant capital investment and path dependence; therefore, whether digital transformation can effectively improve ESG performance through traditional green innovation pathways requires further analysis in the context of specific industries.

Second, while existing research addresses topics such as information disclosure and corporate governance, it has not fully elucidated how digitalization influences ESG performance by improving the external information environment. Digital technologies not only transform internal production and innovation processes but also enhance information processing efficiency, improve information transmission mechanisms, and influence stakeholder perceptions; yet, this information governance mechanism remains under-researched.

In summary, the impact of digital transformation on ESG performance is not merely a single process of technological innovation, but rather the result of embedding digital resources into a firm's value creation system. This paper introduces the value-embedding model perspective, dividing the value realization process through which digitalization influences ESG performance into asset-embedding and information-embedding pathways. It then compares the mechanisms of these two pathways in the context of energy-intensive enterprises to reveal the intrinsic logic by which digital transformation drives improvements in corporate ESG performance.

General theoretical foundation: The Resource-based view and the value of digital resources.

The resource-based view posits that a firm's sustainable competitive advantage stems from scarce, difficult-to-imitate, and irreplaceable strategic resources, while the realization of resource value depends on effective resource allocation and the transformation of capabilities [24]. With the deepening development of the digital economy, intangible assets such as data resources, digital technologies, and digital management capabilities have gradually become new strategic resources for enterprises [25–27], providing a new resource foundation for the green transition and sustainable development of energy-intensive industries.

Specifically in the context of energy-intensive industries, the value of digital resources does not stem from technological investment itself, but rather from their ability to provide differentiated pathways to overcome structural constraints in the industry's green transition: First, resources for precise energy consumption control rely on real-time data collection and intelligent analysis to improve energy efficiency and the controllability of production processes, thereby alleviating industry constraints such as high energy consumption intensity and high environmental governance costs [28]; Second, resources for collaborative resource allocation reduce coordination costs in the green transition of industrial chains by promoting information sharing and upstream-downstream optimization within supply chains, thereby alleviating the practical constraints faced by energy-intensive industries where emissions reductions rely on multi-stakeholder collaboration [29,30]; third, resources for information governance reduce stakeholders’ information acquisition costs by enhancing the standardization, visualization, and transmission efficiency of ESG information, thereby alleviating information asymmetry issues during environmental governance [31].

However, the resource-based view does not hold that strategic resources can automatically translate into competitive advantage; the realization of their value still depends on how they are integrated, allocated, and utilized. Furthermore, the resource orchestration perspective points out that the same resource base may lead to different outcomes in value creation depending on the mode of embedding and the organizational context. For digital resources, the mode of value embedding is determined by their target applications and the vehicles through which value is realized: digital resources aimed at optimizing production processes, technological upgrades, and enhancing industrial synergy require value transformation through adjustments to the enterprise's physical production system, forming an asset-embedded pathway; whereas digital resources aimed at enhancing information generation, transmission, and governance capabilities primarily achieve value creation by improving information relationships between the enterprise and external stakeholders, forming an information-embedded pathway.

Accordingly, there are at least two modes of value transformation for digital resources: one relies on resource integration and capability restructuring to foster green innovation and achieve technological empowerment; the other relies on the generation, transmission, and communication of information to improve the external information environment and achieve governance empowerment. Since these two pathways rely on adjustments to the physical production system and improvements in information governance capabilities, respectively—and thus have different foundations for value realization—the asset rigidity characteristic of energy-intensive industries may subject them to varying degrees of transformation constraints. This constitutes the core logical starting point for the comparison of the two pathways in this paper. The following sections will provide a detailed explanation of the intrinsic transformation mechanisms of these two pathways, drawing on dynamic capabilities theory and stakeholder theory, respectively.

Mechanisms of technology transfer: The theory of dynamic capabilities and pathways to green innovation.

The theory of dynamic capabilities posits that a firm's sustained competitive advantage stems not only from the possession of strategic resources but also depends on its ability to continuously integrate, reconfigure, and allocate resources to adapt to changes in the external environment [32,33]. This theory provides a core perspective for explaining the logic behind the transformation of digital resources into green innovation: the data resources, technological capabilities, and managerial capabilities generated by digital transformation do not automatically lead to green innovation; rather, they require a process of internal resource integration, capability restructuring, and innovative transformation within the enterprise to further yield substantive green innovation outcomes.

In the context of the green transition of energy-intensive enterprises, this dynamic capability is specifically manifested as “green dynamic capability”—that is, an enterprise's unique ability to integrate internal and external green resources, respond to environmental protection demands, and achieve green innovation [34]. Digital resources can enhance a company's environmental data processing capabilities, technology absorption capabilities, and R&D collaboration efficiency; reduce the costs of acquiring and applying green knowledge; promote the development of green products and the improvement of green processes; and ultimately systematically boost the company's green innovation output. Therefore, green innovation constitutes a key technological transformation pathway through which digital transformation influences ESG performance, and it is also the mechanism that has received the most attention in existing research.

However, it is important to note that dynamic capabilities theory primarily explains how firms develop and reconfigure green innovation capabilities; it does not guarantee that such capabilities will necessarily translate into improved ESG performance. For energy-intensive firms, green innovation ultimately requires value realization through tangible adjustments, including equipment upgrades, process optimization, and production system reforms, and therefore follows an asset-embedded value realization pathway. Given the structural characteristics of energy-intensive industries, such as capital intensity and long equipment replacement cycles, the transformation of green innovation outcomes into ESG value may involve substantial implementation and transformation costs, thereby limiting the effectiveness of this pathway under industry-specific constraints. Consequently, dynamic capabilities explain the underlying logic behind the formation of the technology-enabled pathway, while industry structural conditions define the boundaries of value realization through this pathway.

Governance-oriented transformation mechanism: Stakeholder theory and the external information environment pathway.

Stakeholder theory suggests that corporate sustainable development depends not only on economic performance but also on firms’ ability to respond to the diverse demands of stakeholders, including governments, investors, employees, consumers, and the broader public [35,36]. Enhancing organizational transparency and legitimacy through the continuous provision of accurate and comprehensive information constitutes an important foundation for firms to gain stakeholder trust and achieve long-term sustainable development.

For energy-intensive firms, the highly specialized, complex, and often opaque nature of environmental governance makes it difficult for external stakeholders to promptly and accurately assess firms’ actual environmental responsibility performance, thereby intensifying information asymmetry. Digital transformation improves the authenticity, completeness, and timeliness of environmental information through comprehensive data collection, real-time monitoring, and standardized information management. Meanwhile, digital platforms broaden communication channels between firms and stakeholders, reduce the costs associated with accessing ESG-related information, improve the external information environment.

Once the external information environment improves, its governance effects are primarily realized through two channels: On the one hand, a more transparent information environment enhances the efficiency of government regulation, capital market oversight, and public supervision, thereby strengthening corporate environmental responsibility and corporate governance practices and establishing a governance constraint mechanism; on the other hand, the improvement in the external information environment reduces the cost for stakeholders to identify companies’ actual responsibility-fulfillment behaviors, enabling ESG rating agencies, investors, and the general public to evaluate corporate sustainability performance more objectively, thereby reducing undervaluation caused by information asymmetry and forming an information identification mechanism [37]. Therefore, unlike green innovation pathways that rely on the restructuring of physical assets, the external information environment manifests as information-embedded value transformation, whose value realization depends primarily on improvements in information generation, transmission, and governance capabilities, rather than adjustments to the production system. Since this process involves minimal equipment upgrades and process modifications, the realization of its value is relatively less affected by the asset rigidity characteristic of energy-intensive industries.

Digital transformation and corporate ESG performance.

The preceding analysis indicates that digital transformation may influence ESG performance by changing resource allocation and value realization. Therefore, this study examines the overall relationship between digital transformation and the ESG performance of energy-intensive firms.

Accordingly, the following research hypothesis is proposed:

H1: There is a positive relationship between digital transformation and the ESG performance of energy-intensive firms.

Green innovation as an asset-embedded path.

Digital transformation provides the foundational capabilities for corporate green innovation by enhancing the efficiency of resource integration, promoting the absorption of green knowledge, and strengthening R&D collaboration [38]. Green innovation has thus become a key technological pathway for digital resources to realize ESG value. However, green innovation is a typical asset-embedded value conversion pathway. Given the rigid constraints on assets in energy-intensive industries, converting digital resources into green innovation outcomes entails high transaction costs, and the ESG value of the technology-enabled pathway may be difficult to realize. Based on this, the following research hypotheses are proposed:

Based on this, the following research hypotheses are proposed:

H2: Green innovation is a key value-transformation pathway through which digital transformation influences the ESG performance of energy-intensive enterprises, but its effectiveness is constrained by the characteristics of assets in energy-intensive industries.

External information environment as an information-embedded path.

Unlike green innovation, which relies on adjustments to physical assets to achieve value transformation, the external information environment manifests as an information-embedded governance-driven empowerment pathway. As discussed in Section 3.1.3, digital transformation improves a company's external information environment and reduces stakeholders’ information identification costs, making the company's ESG performance easier for capital markets and other stakeholders to identify and evaluate. Since this pathway relies primarily on information-processing capabilities and governance coordination efficiency—rather than on adjustments to physical assets—its value realization is less constrained by the rigidity of assets in energy-intensive industries.

Based on this, the following research hypothesis is proposed:

H3: The external information environment mediates the relationship between digital transformation and the ESG performance of energy-intensive enterprises.

Differential effects of two value embedding paths.

Based on the theoretical analysis presented above, it can be observed that both pathways are important means by which digital resources realize ESG value, but the resource vehicles upon which the realization of value depends differ between the two. Accordingly, this paper proposes the analytical perspective of “Value Embedding Mode,” arguing that digital resources do not directly create ESG value on their own; rather, their role depends on the specific manner in which they are embedded into a company's value creation process. Specifically, green innovation falls under the “asset-embedded” pathway, in which digital resources are embedded within the production system to realize value; the realization of this value relies on the accumulation of physical capabilities such as green R&D, equipment upgrades, and production process improvements. The external information environment, on the other hand, falls under the “information-embedded” pathway, in which digital resources are embedded within the corporate governance process; the realization of its value primarily depends on improvements in information processing efficiency, information transparency, and governance coordination capabilities. Given the differences in the resource bases upon which these two types of value embedding rely, they may exhibit different levels of statistical stability in energy-intensive industries.

Based on this, the following research hypothesis is proposed:

H4: In the context of energy-intensive industries, the two transmission mechanisms through which digital transformation affects ESG performance exhibit different levels of value-conversion stability, with the information-embedded pathway being more stable than the asset-embedded pathway.

Baseline regression model.

To test the positive relationship between digital transformation and the ESG performance of energy-intensive enterprises (H1), we employ mixed OLS regression, controlling for year and industry fixed effects, with standard errors clustered at the firm level. The full model includes all control variables; the baseline model (1) is specified as follows:

(1)

In the model, denotes the dependent variable (ESG performance) for firm i in period t. The core independent variable represents the one-period lagged digital transformation index. includes a set of control variables: firm size (Size), leverage ratio (Lev), profitability (ROA), growth (Growth), and board independence (Board_Indep). and indicate year fixed effects and industry fixed effects, respectively, and is the random error term. If the estimated coefficient β₁ is significantly positive, Hypothesis 1 is supported.

Robustness test models.

To verify the reliability of the aforementioned regression results, this study designed the following robustness test models based on the empirical findings, including a model with individual fixed effects, nonlinearity tests, and endogeneity controls.

1. 1. Individual Fixed-Effects Model

(6)

1. denotes the firm fixed effect, which captures time-invariant firm-level heterogeneity; represents the year fixed effect. is consistent with the baseline model; standard errors are clustered at the firm level.
2. 2. Nonlinearity Test (Quadratic Term Specification)
3. To test the potential nonlinear relationship, we add the quadratic term of to the baseline model:

(7)

1. denotes the quadratic term of the one-period lagged digital transformation index. includes all control variables adopted in the baseline regression. We control for year fixed effects and industry fixed effects , with standard errors clustered at the firm level.
2. To mitigate multicollinearity, the quadratic term is constructed based on the centered (denoted as ), and the estimation results remain equivalent.
3. 3. Addressing Endogeneity
4. To alleviate potential endogeneity issues such as reverse causality and omitted variable bias between the one-period lagged digital transformation () and ESG performance (), we employ the instrumental variable approach, namely two-stage least squares (2SLS) estimation. Following existing studies (e.g., Fisman & Svensson [43]), we adopt the average value of for other firms within the same industry and year as the instrumental variable ().

The relevance of this instrument is based on the peer effect of digital transformation within the same industry, as firms’ digitalization processes may influence the digital transformation decisions of their industry peers. Considering that industry-level common factors may simultaneously affect firms’ ESG performance, this study further controls for industry and year fixed effects to mitigate the influence of industry-wide shocks. The instrumental variable results are therefore treated as supplementary evidence to address potential endogeneity concerns.

The model specifications are set out as follows:

First stage:

(9)

Second stage:

(9)

where denotes the instrumental variable (the average of other firms in the same industry); represents firm fixed effects; denotes year fixed effects. Standard errors are clustered at the firm level.

Descriptive statistics and correlation analysis.

The descriptive analysis (Table 2) systematically presents the basic distribution characteristics of the sample variables. The mean ESG score for the sample companies was 72.24, indicating that the overall ESG performance of energy-intensive enterprises has reached a certain level; however, the standard deviation of 5.972 suggests that significant differences in sustainable governance capabilities still exist among these companies. The mean Digital Transformation Index was 30.75, with a standard deviation of 6.194, reflecting significant variations in the degree of digital transformation among enterprises, as the accumulation of digital resources and the pace of transformation differ across companies. As two types of mechanism variables, green innovation had a mean of 0.418 and a standard deviation of 0.777, with a median of 0, indicating strong heterogeneity in the green innovation activities of energy-intensive enterprises; the external information environment had a mean of 5.362 and a standard deviation of 7.578, with the high degree of dispersion indicating significant differences in the extent of improvement in the information environment across enterprises.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Descriptive analysis table.

https://doi.org/10.1371/journal.pone.0343345.t002

Table 3 reports the correlation analysis among the main variables. The one-period lagged digital transformation variable (DT_Index_lag) is significantly and positively correlated with ESG performance (r = 0.144, p < 0.05), providing preliminary evidence for the subsequent regression analysis. Regarding the control variables, firm size, profitability, information transparency, and green innovation are positively associated with ESG performance, whereas firm growth and leverage exhibit negative correlations. Furthermore, both green innovation and information transparency are significantly correlated with digital transformation and ESG performance, providing preliminary support for the theoretical expectation that digital transformation influences ESG performance through technological embedding and governance embedding pathways. In addition, all variance inflation factor (VIF) values are below the critical threshold, indicating that there is no serious multicollinearity problem among the variables.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Correlation analysis (n = 3,399).

https://doi.org/10.1371/journal.pone.0343345.t003

Benchmark regression results.

To test the relationship between digital transformation and the ESG performance of energy-intensive firms (H1), Table 4 presents the results of the baseline regression. After controlling for year fixed effects, industry fixed effects, and firm-level control variables, the coefficient for the digital transformation index (one period lagged) is 0.0467 (t = 1.50); the direction is positive but does not pass the conventional significance test; In a simplified model controlling only for year and industry fixed effects, the coefficient for digital transformation was 0.1349 and was significantly positive at the 1% level. These two sets of results indicate that digital transformation is generally positively associated with ESG performance, but this association weakens significantly when firm-level characteristics are included, suggesting that the ESG value generated by digital transformation may not be realized through a stable direct effect but rather relies on further transmission mechanisms.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Regression results on the impact of digital transformation on the ESG performance of energy-intensive enterprises.

https://doi.org/10.1371/journal.pone.0343345.t004

The results for the control variables were generally in line with expectations. Firm size significantly promotes ESG performance, indicating that larger firms possess a more robust resource base for implementing ESG practices; the debt-to-equity ratio was significantly negative, suggesting that higher financial constraints may limit a firm's capacity for sustainable investment; the effects of profitability and growth were not significant, indicating that short-term operating performance has limited explanatory power for ESG differences among energy-intensive firms.

Test results for the two transmission mechanisms.

Table 5 presents the results of the mediation analysis for the two types of transmission mechanisms. Based on the analytical framework proposed in Chapter 3, this study examines the mechanisms through which the technology pathway and the governance pathway influence the impact of digital transformation on ESG performance.

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Mediation effect estimation results.

https://doi.org/10.1371/journal.pone.0343345.t005

Regarding the technology-enabled mechanism, the estimated indirect effect of digital transformation through green innovation is positive but does not exhibit statistical robustness. Although Path a (digital transformation → green innovation) is marginally significant at the 10% level and Path b (green innovation → ESG performance) shows a significant positive effect, the confidence interval of the indirect effect includes zero. Therefore, based on the indirect-effect criterion, green innovation cannot yet be identified as a statistically robust transmission mechanism through which digital transformation affects ESG performance in the full sample. The bootstrap test further confirms this result: the confidence interval of the indirect effect still includes zero (Table 6), suggesting that the green innovation pathway does not constitute a statistically robust mediation channel in the full sample. These results suggest that digital transformation possesses the potential to promote green innovation, but the further translation of this potential into ESG value remains subject to industry-specific conditions. Specifically, the outcomes of green innovation must undergo adjustments to production systems and the integration of organizational resources before they can ultimately manifest as improvements in ESG performance; factors such as capital-intensive characteristics and technology upgrade cycles may increase the difficulty of achieving this transformation.

Download:

• PNG
  larger image
• TIFF
  original image

Table 6. Bootstrap test of indirect effects for the two transmission mechanisms.

https://doi.org/10.1371/journal.pone.0343345.t006

In contrast, the governance-empowerment pathway exhibits more stable statistical characteristics. Digital transformation significantly improves a firm's external information environment and further promotes enhanced ESG performance, with the mediating effect accounting for 32.6% of the total effect. Bootstrap resampling results further show that the Bootstrap confidence interval for the indirect effect of this pathway does not cross zero, indicating that improvements in the external information environment constitute a key transmission channel through which digital transformation influences ESG performance. These results indicate that digital transformation not only enhances a firm's technological capabilities but also facilitates the realization of governance value by improving the processes of information generation, transmission, and external communication. Since this pathway primarily relies on improvements in information processing capabilities and governance coordination efficiency—and involves relatively few adjustments to physical assets—it demonstrates strong adaptability in the context of energy-intensive industries.

Overall, the two transmission mechanisms exhibit distinct statistical stability characteristics: the green innovation pathway is subject to stronger constraints associated with physical production-system adjustments and does not exhibit a statistically robust transmission effect; in contrast, the external information environment pathway generates a significant mediating effect and demonstrates greater stability. The above conclusions remain valid under bootstrap resampling tests, further enhancing the reliability of the mechanism identification results. This finding supports the analytical framework of “value embedding modes” proposed in this paper, namely, that the ESG value of digital resources depends on the specific manner in which they are embedded in the firm's value creation process.

Robustness tests.

1. Firm-Specific Fixed Effects Model

As shown in column (6) of Table 7, after controlling for firm-specific fixed effects, the coefficient for digital transformation is 0.0333. Although it remains positive, it does not pass the significance test. This result indicates that, after further controlling for firm-specific characteristics that do not change over time, the direct impact of digital transformation on ESG performance still does not exhibit a consistent statistical effect, consistent with the conclusion from the baseline regression that “digital transformation unleashes ESG value through internal transmission mechanisms.”

Download:

• PNG
  larger image
• TIFF
  original image

Table 7. Summary of Robustness Test Results.

https://doi.org/10.1371/journal.pone.0343345.t007

1. Nonlinear (Squared Term) Test

To further examine whether a nonlinear relationship exists between digital transformation and ESG performance, this study incorporates a squared term for digital transformation (constructed based on the centered variable) into the baseline model. As shown in column (7) of Table 6, the coefficient of the squared term for digital transformation is 0.0065 (t = 1.87), which is marginally significant at the 10% level, indicating that the impact of digital transformation on ESG performance is not entirely linear.

Further analysis of the marginal effect results in Fig 6 reveals that the marginal effect of digital transformation on ESG performance remains consistently positive across the sample range and exhibits an increasing trend as the level of digitalization rises. This implies that there is no distinct inflection point where digital transformation shifts from a negative constraint to a positive driver; rather, the release of its ESG value gradually intensifies as digital resources accumulate and organizational capabilities deepen. In other words, the impact of digital transformation does not depend on a single technological investment, but rather on whether digital resources can be continuously embedded into corporate governance and operational processes to generate cumulative value creation.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Marginal effects of digital transformation on ESG performance.

https://doi.org/10.1371/journal.pone.0343345.g006

This finding further complements the results of the benchmark regression and mechanism tests: although digital transformation does not exhibit a stable direct effect on ESG, its value is not nonexistent; rather, it needs to be gradually unleashed through the restructuring of internal capabilities and the refinement of governance mechanisms. It should be noted that, since the squared term reached only marginal significance and the marginal effects plot reflects trends within the sample range, this study does not interpret these findings as a strict U-shaped relationship or a clear threshold effect, but rather regards them as supporting evidence for the gradual release of digital value.

1. Treatment of Endogeneity

Given the potential for a reverse causal relationship between digital transformation and corporate ESG performance, this paper employs the instrumental variables method to test for endogeneity (Table 8). Specifically, we select the mean digital transformation level of other firms in the same industry as the instrumental variable (IV_IndustryMean) to reflect the influence of the digital diffusion environment within the firm's industry on its own digital transformation decisions, while controlling for industry and year fixed effects.

Download:

• PNG
  larger image
• TIFF
  original image

Table 8. Impact of being one phase behind in digital transformation on ESG (N = 3,390).

https://doi.org/10.1371/journal.pone.0343345.t008

The results of the first stage show that the instrument coefficient is 0.377, which is significantly positive at the 1% level, indicating that the digitalization level of other firms in the same industry effectively explains the firm's own degree of digital transformation. The Kleibergen-Paap rk LM statistic is 17.98 and is significant at the 1% level; the F-statistic for Stage 1 is 18.80, which is higher than the Stock-Yogo 10% critical value of 16.38, indicating that there is no significant weak instrument problem.

The results of the second stage show that the coefficient for digital transformation remains positive but does not reach statistical significance. This indicates that, after correcting for potential endogeneity biases using the instrumental variables method, the direct impact of digital transformation on ESG performance still does not exhibit stable and significant characteristics, consistent with the findings from the baseline regression that the direct effect of digital transformation is weak. Further analysis of the mediation effect test results suggests that the impact of digital transformation on ESG performance may rely more on transmission mechanisms such as green innovation and the external information environment, rather than manifesting as a stable direct effect.

Key research conclusions.

The core conclusion of this paper can be summarized as follows: The ESG value of digital resources is not directly determined by the scale of investment, but rather depends on the specific manner in which value is embedded; the asset rigidity in energy-intensive industries imposes asymmetric constraints on the asset-embedded (green innovation) and information-embedded (external information environment) pathways, ultimately resulting in a divergent pattern characterized by “technological pathways constrained by asset rigidity, while governance pathways exhibit greater stability.”

Specifically, there is a positive association between digital transformation and the ESG performance of energy-intensive enterprises, but no stable direct effect. Digital resources cannot automatically translate into ESG performance; the realization of their value must rely on specific transmission mechanisms. Further analysis reveals that the asset-embedded pathway corresponding to green innovation has not established a stable transmission mechanism. Constrained by structural characteristics such as capital intensity and high asset specificity, the efficiency of converting digital resources into green innovation is weakened by asset rigidity, making it difficult to effectively unlock the value of technological empowerment.

In contrast, the information-embedded pathway—corresponding to the external information environment—serves as a robust transmission channel. Digital technology reduces information asymmetry by optimizing mechanisms for information generation, transmission, and oversight, thereby driving ESG improvements through governance-enabled approaches; compared to the green innovation pathway constrained by asset rigidity, the realization of its value does not depend on adjustments to physical assets, resulting in greater transmission stability. Thus, the core difference between the two pathways lies not in the magnitude of their effects, but in their statistical stability under highly constrained conditions. The information-embedded pathway exhibits greater resistance to constraints and represents a more reliable means of realizing digital ESG value for energy-intensive enterprises.

Theoretical contributions.

First, this paper proposes an analytical framework of “value embedding modes,” thereby expanding the explanation of the mechanisms through which digital transformation influences ESG performance. Existing research largely remains at the level of assessing effects, failing to open the “black box” of value transformation from digital resources; moreover, mechanism studies generally focus on the magnitude of effects while neglecting contextual stability. This paper distinguishes between two value-conversion models—asset embedding and information embedding—and demonstrates that they exhibit divergent performance under asset rigidity constraints. This not only provides a new explanatory tool for the heterogeneity of digital sustainable value but also broadens the perspective of mechanism research from “effect strength” to “path stability.”

Second, this paper reexamines the applicability of green innovation as the core mechanism through which digitalization empowers ESG. While existing literature, largely based on cross-industry samples, assumes green innovation to be the core transmission mechanism, this study—using a sample from energy-intensive industries—demonstrates that asset rigidity significantly weakens the conversion efficiency of technological pathways, indicating that technology-enabled pathways are not consistently stable or effective across all industrial contexts.

Policy implications.

At the corporate level, energy-intensive enterprises should select a digital ESG transformation path suited to the inflexible nature of their assets. For traditional manufacturing enterprises with highly specialized assets and long retrofitting cycles for existing equipment, the short-term priority should be to leverage digital technologies to enhance energy consumption monitoring, environmental data collection, and ESG disclosure capabilities. By empowering governance, they can unlock the value of digitalization and avoid blindly pursuing high-cost technological upgrades when conditions are not yet ripe. For enterprises with significant room for asset renewal and the conditions for production capacity upgrades, digital technologies can be further integrated into green R&D and production processes in conjunction with equipment upgrades and process optimization, thereby achieving synergistic progress in both technological empowerment and governance enhancement.

At the industrial policy level, support mechanisms for green innovation should be refined to address the transformation constraints faced by industries with high asset rigidity. For industries such as steel, cement, and coking—where equipment retrofitting costs are high—policy support should not adopt a one-size-fits-all digital assessment model. Instead, it should reduce the costs of adjusting physical assets through targeted support for digital green technological upgrades, financing facilitation, and tax incentives, thereby increasing the feasibility of converting digital resources into green innovation. At the same time, avoid imposing technological transformation requirements on existing enterprises that exceed their transformation capabilities to prevent a mismatch between policy objectives and the actual capacity of enterprises.

At the ESG regulatory level, a tiered disclosure mechanism aligned with enterprises’ information governance foundations should be further refined. For enterprises with strong digital foundations and favorable external information environments, verifiable digital ESG disclosure requirements should be implemented first, mandating the integration of operational condition monitoring and emissions network data to amplify the effect of governance constraints. For small and medium-sized enterprises (SMEs) with weak information foundations, support should be provided through industry-standardized disclosure templates and lightweight energy-and-carbon management SaaS tools to gradually improve information governance levels and unlock the inclusive value of digital governance.

Research limitations and prospects.

First, in terms of variable measurement, this paper still has certain limitations. Constrained by data availability, this paper uses analyst coverage and patent counts to measure the external information environment and green innovation level, respectively. Although these proxies can, to some extent, characterize the relevant mechanisms, they still cannot fully capture the dynamic changes in firms’ information governance capabilities and innovation quality. Future research could combine ESG disclosure text analysis, patent quality evaluation, and data on firms’ digital application scenarios to further improve the precision of variable measurement.

Second, in terms of causal identification, there remains room for further deepening. Although this paper enhances the reliability of its conclusions through multiple empirical methods, endogeneity issues may still exist between digital transformation and firms’ ESG performance. Future research could leverage quasi-natural experimental designs such as digital economy pilot programs and mandatory ESG disclosure policies to obtain stronger causal identification evidence.

Finally, in terms of expanding the theoretical framework, the analysis of this paper still has certain boundaries. This paper primarily focuses on two typical value realization pathways—asset embeddedness and information embeddedness—and has not yet covered other possible ways through which digital resources affect ESG performance. Subsequent research could explore more models such as supply chain embeddedness and organizational embeddedness, and conduct cross-industry comparisons based on the asset rigidity characteristics of different industries, in order to refine the theoretical explanatory framework for how digital transformation promotes sustainable development.