Does Digital Asset Allocation Improve Corporate ESG Performance? Evidence from China

Abstract

Against the backdrop of the deep integration of the “Dual Carbon” goals and the digital economy, whether digital asset allocation can improve corporate Environmental, Social, and Governance (ESG) performance has become an important topic of academic concern. Taking all Chinese A-share listed firms from 2013 to 2024 as the research population, this study obtains a final panel sample of 29,329 firm-year observations after excluding financial and insurance firms, ST/*ST firms, newly listed firms, and observations with missing key variables. Digital asset allocation is measured by the proportion of digital technology-related intangible assets to total intangible assets. The study employs a two-way fixed-effects panel model, firm-clustered robust standard errors, IV-2SLS estimation, robustness tests based on alternative measurements and sample restrictions, and Bootstrap sequential mediation analysis. The findings reveal that digital asset allocation significantly enhances corporate ESG performance. Mechanism tests indicate that digital asset allocation improves corporate ESG performance through internal control quality, green technological innovation, and the sequential pathway from internal control quality to green technological innovation. Further moderation analysis shows that the promotion effect is more pronounced in heavily polluting industries, while heterogeneity analysis indicates stronger effects among firms in the growth and decline stages, non-state-owned enterprises, and firms with lower financing constraints. This study provides empirical evidence and policy implications for optimizing corporate digital resource allocation, improving internal governance mechanisms, and advancing classified ESG.

1. Introduction

In the historical trajectory of China’s transition toward high-quality economic development, the establishment of “dual-carbon” strategic goals and the profound evolution of the digital economy have emerged as two mutually embedded systemic forces. On one hand, since President Xi Jinping announced China’s carbon peaking and carbon neutrality targets at the United Nations General Assembly in 2020, the concept of Environmental [1], Social, and Governance (ESG) has rapidly permeated China’s capital markets, evolving from a framework for voluntary disclosure into a cornerstone of institutionalized regulation. In 2018, the China Securities Regulatory Commission revised the Code of Corporate Governance for Listed Companies to integrate environmental protection and social responsibility requirements; subsequently, in 2020, the Shenzhen Stock Exchange pioneered the inclusion of proactive ESG disclosure within its annual assessment system, thereby establishing the fundamental institutional architecture for ESG reporting among Chinese listed firms. On the other hand, a nascent generation of digital technologies, epitomized by artificial intelligence, big data, cloud computing, and blockchain, has increasingly integrated with the real economy [2,3]. China’s 14th Five-Year Plan for Digital Economy Development explicitly underscores the deepening of enterprise digital transformation, as the allocation of digital resources has progressively become a critical determinant shaping strategic decision-making, resource configurations, and governance arrangements [4]. Nevertheless, a pivotal practical paradox remains unresolved regarding whether the structural allocation of digital resources can systematically enhance the ESG performance of listed firms under the dual pressures of accelerating digitalization and intensifying ESG expectations. Furthermore, the internal governance mechanisms through which this effect operates and the contextual conditions that cause its efficacy to vary remain insufficiently explored. Addressing these central issues carries significant academic weight and pressing policy relevance, necessitating a rigorous empirical investigation.

Regarding the extant literature, scholars have established a substantial body of empirical evidence concerning the nexus between digital transformation and ESG performance. Utilizing a sample of Chinese A-share listed firms, Lu et al. (2024) demonstrate that corporate digital transformation exerts a significantly positive influence on ESG performance, identifying enhanced information transparency and refined resource allocation efficiency as the primary transmission channels [5]. Expanding on this direct relationship, Wu et al. (2024) reveal that digital transformation not only directly fosters ESG improvements but also amplifies these outcomes through the mediating role of green technological innovation [6]. Similarly, Chen and Ren (2024) employ a sustainable development framework to illustrate that advancements in a firm’s comprehensive digitalization level systematically bolster all three dimensions of ESG [7]. Concurrently, a growing body of research, represented by Li et al. (2023), Luo et al. (2024), and Liu et al. (2024), has extended the discourse toward moderating mechanisms and boundary conditions, elucidating how institutional environments, market competitive pressures, and environmental regulatory intensity differentially shape these effects [8,9,10]. In terms of transmission pathways, Feng and Nie (2024) and Tang et al. (2023) provide systematic empirical support for the localized mechanism wherein digital transformation facilitates green innovation [11,12]. Furthermore, Xu et al. (2024) confirm that the strategic embedding of digital resources can substantively catalyze a firm’s green technological research and development activities [13].

Despite these meaningful contributions, the existing literature still suffers from three notable limitations that require further exploration. First, in terms of the measurement of the core explanatory variable, most mainstream studies adopt the conventional textual frequency method. Although widely applied, this approach mainly captures managerial discourse and rhetorical emphasis, rather than the substantive depth of corporate digital asset allocation embedded in factor structures. This leads to a key ambiguity: it remains unclear whether improvements in ESG performance are driven by strategic narrative intentions or by measurable capital accumulation reflected in balance sheet data. Accordingly, the asset-based empirical foundation of current research is insufficiently anchored [14,15]. Second, in terms of mechanism identification, to the best of our knowledge, research has yet to provide a comprehensive empirical validation of the sequential transmission chain involving digitalization, internal control, green innovation, and ESG performance. Although internal control constitutes a vital managerial link between digital investment and sustainable governance outcomes, it has not been rigorously examined as a mediating variable within a unified framework. Furthermore, the absence of validation through strict Bootstrap procedures has left the hierarchical logic and causal progression of this sequential pathway inadequately clarified. Third, from the perspective of heterogeneity analysis, relatively few studies have explored how the ESG-enabling effect of digital asset allocation varies across different stages of the corporate life cycle. Integrating developmental stages, ownership structures, and financing constraints into a cohesive heterogeneity framework remains an important, yet under-researched, avenue. Addressing these deficiencies is essential for providing a more nuanced understanding of how digital resources translate into sustainable corporate outcomes.

To address these research lacunae, this study utilizes a panel dataset of Chinese A-share listed companies spanning 2013 to 2024. The research population consists of all A-share listed firms during the sample period, and a purposive screening procedure is applied to exclude financial and insurance firms, ST/*ST firms, newly listed firms, and observations with missing key variables, yielding 29,329 firm-year observations. By constructing a structural indicator of digital resource allocation based on the proportion of digital technology-related intangible assets, we employ a two-way fixed-effects model to systematically evaluate the impact of digital asset allocation on corporate ESG performance, alongside its internal governance mechanisms and relevant boundary conditions. The empirical strategy further combines IV-2SLS estimation, robustness checks, Bootstrap sequential mediation, moderation analysis, heterogeneity tests, and multicollinearity diagnostics.

The main empirical results are summarized as follows. First, the baseline results reveal that digital asset allocation exerts a highly significant positive influence on ESG performance. This relationship remains robust to alternative measures of the dependent and independent variables, the inclusion of lagged explanatory variables, alternative winsorization thresholds, and the exclusion of the financial and real estate sectors. Second, the IV-2SLS analysis provides additional evidence that the baseline relationship is unlikely to be driven solely by reverse causality or omitted variable bias. Third, mechanism analysis demonstrates that digital asset allocation bolsters ESG performance through internal control quality, green technological innovation, and the sequential pathway linking internal control quality to green technological innovation. Finally, heterogeneity analyses uncover a distinctive pattern across the corporate life cycle, with more pronounced effects in the growth and recession stages compared to the maturity stage. Moreover, the ESG-enhancing effect of digitalization is significantly stronger among non-state-owned enterprises and firms with lower financing constraints.

The marginal contributions of this study are manifested in three primary dimensions. First, regarding variable measurement, this research diverges from the “strategic intention-oriented” paradigm reliant on textual frequency analysis, opting instead for the proportion of digital intangible assets as a balance sheet-anchored structural indicator. By capturing the tangible capital accumulation of digital resources through the lens of factor allocation, this approach provides a more rigorous identification strategy for elucidating the causal link between digital investment and ESG performance, thereby refining the methodological framework for assessing the consequences of digitalization. Second, in terms of mechanism identification, this study systematically constructs and empirically validates a sequential transmission pathway wherein digital asset allocation enhances internal control quality, which subsequently elevates ESG performance. Through Bootstrap mediation analysis, we provide robust micro-level evidence for this “managerial empowerment” mechanism. Consequently, this study addresses a notable empirical void regarding how digital intangible assets improve ESG performance through internal control. Specifically, digital assets such as software systems, data platforms, management systems, and intelligent monitoring tools strengthen process standardization, real-time information sharing, risk identification, and accountability tracing, thereby improving internal control quality and supporting ESG. This study also addresses the mediating role of internal control and integrates corporate governance optimization into the discourse on digitalization and sustainable development, offering new theoretical insights into how digital investments translate into ESG outcomes. Third, from a heterogeneity perspective, this study is among the first to integrate the corporate life-cycle classification, based on Dickinson’s (2011) cash-flow methodology, into the digitalization-ESG research context [16]. The findings reveal a distinctive U-shaped pattern, with more potent digital empowerment effects occurring during the early and late stages of firm development than in the maturity stage. Furthermore, by providing systematic evidence of boundary conditions across ownership structures, financing constraints, and industry pollution intensity, this study offers nuanced policy implications for fostering ESG performance across diverse organizational profiles.

2. Theoretical Analysis and Research Hypotheses

2.1. Digital Asset Allocation and Corporate ESG Performance

The Resource-Based View posits that strategic resources defined by their rarity, value, and inimitability—constitute the bedrock of a firm’s sustained competitive advantage [17]. Within the context of a burgeoning digital economy, the strategic prioritization of intangible assets toward digitalization represents a fundamental reconfiguration of a firm’s underlying strategic elements. This pivot signifies more than a marginal technological upgrade; it denotes a transformative reorganization of the corporate resource portfolio. Tan and Zhu (2022) contend that digital transformation fundamentally reshapes organizational structures, production modalities, and business models through deep digital integration [18]. This process not only enhances operational efficiency but also optimizes organizational workflows, thereby influencing ESG performance through two primary channels: the promotion of green innovation and the elevation of information disclosure quality [18,19]. Consequently, the integration of digital assets injects vital new resource endowments into the pillars of environmental stewardship, social responsibility, and governance optimization [15]. Complementing this perspective, signaling theory and information asymmetry theory provide integrated lenses for elucidating the ESG-enabling effects of digital asset allocation. On one hand, the deployment of digital assets fortifies internal information integration and sharing capabilities. This empowers management to monitor cross-segment operational status with heightened precision and timeliness, effectively narrowing the scope for information asymmetry within the principal-agent chain and substantially mitigating agency costs [20,21]. On the other hand, firms that proactively advance digital transformation transmit potent signals of technological agility to capital markets, cultivating favorable expectations and legitimacy among investors and external stakeholders. Empirical evidence from Alkaraan et al. (2022) confirms that digital transformation significantly enhances stock liquidity by mitigating information gaps, suggesting a pronounced “signal premium” inherent in digital asset allocation [22]. In sum, digital asset allocation systematically elevates corporate performance across Environmental (E), Social (S), and Governance (G) dimensions through the synergistic mechanisms of resource restructuring, strategic signaling, and the compression of agency costs. Accordingly, this study proposes:

Hypothesis H1. 

Digital asset allocation significantly enhances corporate ESG performance.

2.2. The Sequential Mediation Mechanism of Internal Control and Green Technology Innovation

The influence of digital asset allocation on ESG performance does not constitute an immediate leap to terminal outcomes; rather, it is realized through the staged transmission of internal governance mechanisms. The inaugural link in this causal chain is the “managerial empowerment” effect exerted by digital elements on the quality of internal control. Drawing on data from listed manufacturing firms, Wang et al. (2023) observe that digital transformation significantly bolsters internal control quality [23]. The core mechanism resides in the augmentation of data collection, analytical, and disclosure capabilities, which facilitates the architecture of a data-driven supervisory apparatus. This, in turn, fortifies internal control by compressing agency costs and optimizing human capital structures. Specifically, management can synthesize digital technologies with business operations, leveraging information systems to program and automate workflows, thereby intensifying oversight across all operational phases. Simultaneously, digital tools mitigate information asymmetry across disparate departments and organizational hierarchies, ensuring a more synchronous, accurate, and comprehensive information flow. Liu et al. (2023) further contend that embedding precise, intelligent digital modules into internal control frameworks diminishes information gaps between firms and investment institutions, thus optimizing the internal governance milieu [24].

In this sense, digital intangible assets differ from a general digitalization discourse because they represent capitalized and organizationally embedded digital capabilities. Software systems, intelligent platforms, and management information systems can be repeatedly used in procurement, production, environmental monitoring, financial reporting, and compliance processes. Their embedded use improves traceability, reduces manual intervention, standardizes approval procedures, and strengthens early-warning functions, thereby providing a concrete internal-control channel through which digital asset allocation can translate into better ESG performance.

Accordingly, the following hypothesis is proposed:

Hypothesis H2a. 

Digital asset allocation significantly improves the quality of corporate internal control.

The second stage in this sequential chain involves the catalytic effect of enhanced internal control on green technological innovation. Green innovation is inherently characterized by protracted cycles, high capital intensity, and substantial uncertainty. Consequently, firms often encounter an “adverse selection” dilemma in green R&D, wherein risk-averse management may opt to curtail high-risk green investments [10]. Luo et al. (2024) demonstrate that internal control serves as a pivotal mediator in fostering green innovation; high-quality internal control systems eliminate internal information asymmetries and enhance managerial coordination, thereby reducing decision-making friction and facilitating sustainable development strategies [9]. Li and Zhao (2024) further elucidate these micro-mechanisms: rigorous internal control mitigates agency problems through risk assessment and supervisory protocols, enhancing the rationality of green innovation decisions while simultaneously deploying control measures to attenuate process risks [25]. In essence, a robust internal control system standardizes R&D protocols and clarifies accountability, resolving the persistent “adverse selection” problem in green technology. Thus, Accordingly, the following hypothesis is proposed:

Hypothesis H2b. 

Improvements in internal control quality significantly promote corporate green technology innovation.

Synthesizing these stages yields a coherent sequential transmission pathway: digital asset allocation first optimizes internal control quality through “managerial empowerment,” which subsequently activates green technological innovation by standardizing R&D management and mitigating innovation risks, ultimately culminating in systematic ESG improvements. This logic aligns with the “Innovation Compensation Hypothesis” (Porter & van der Linde, 1995), which suggests that appropriate institutional constraints and governance optimization can “induce” substantive innovation, where the resulting “compensatory gains” outweigh compliance costs to achieve a synergy between competitiveness and sustainability [26]. Tang et al. (2023) validate this logic within the context of Chinese listed firms, finding that digital transformation catalyzes substantive green innovation, which drives aggregate ESG performance through the systematic accumulation of technical capabilities [12]. Consequently, digital asset allocation does not directly “bypass” organizational processes to reach ESG outcomes but operates through the sequential mediation of internal control optimization and the subsequent activation of green innovation. Accordingly, the following hypothesis is proposed:

Hypothesis H2c. 

Internal control quality and green technology innovation play a sequential mediating role in the relationship between digital asset allocation and ESG performance.

2.3. The Moderating Role of Industry Pollution Intensity

While the preceding analysis elucidates the internal transmission mechanisms, Legitimacy Theory suggests that the impact of digital asset allocation is not uniform across all organizational contexts. Corporate ESG behavior is fundamentally shaped by a nexus of formal and informal external pressures; to acquire, maintain, or restore environmental legitimacy, firms often proactively enhance their ESG profiles. This imperative is particularly salient for enterprises operating in heavy-pollution industries. As noted by Berrone et al. (2013), stringent environmental pressures, such as emission charges, can “induce” green innovation as firms strive to mitigate penalties and secure institutional support [17]. Consistent with this logic, the ESG-enabling effect of digital asset allocation is expected to be more pronounced in carbon-intensive sectors for two primary reasons: Heightened Marginal Utility of Green Compensation: Firms in heavy-pollution industries face more rigorous regulatory oversight and intense public scrutiny. Consequently, the marginal utility of deploying digital technologies for precise emission monitoring, energy optimization, and environmental compliance is significantly higher than in “clean” industries. This creates a more robust “green compensation” effect. Song et al. (2025) confirm that the heterogeneous distribution of regulatory pressure dictates the trajectory of ESG improvements in the digital economy, with high-pollution firms exhibiting a stronger motivation to translate digitalization into tangible ESG gains [27]. Signal Distinctiveness and Reputation Premiums: Drawing on Signaling Theory, digital transformation in heavy-pollution sectors carries greater “signal distinctiveness.” Given that the market often anticipates environmental deficiencies in these firms, the proactive deployment of digital assets to bolster ESG performance serves as a potent positive signal, yielding a superior reputation premium. Furthermore, Wang (2025) demonstrates that under the aegis of environmental regulation, the correlation between digital transformation and ESG performance is significantly amplified in high-pollution sectors, suggesting a synergistic reinforcement between regulatory frameworks and digital integration [28]. In summary, industry pollution intensity serves as a critical boundary condition that moderates the relationship between digital investments and sustainability outcomes. Accordingly, this study proposes the following hypothesis:

Hypothesis H3. 

Industry pollution intensity positively moderates the impact of digital asset allocation on corporate ESG performance; specifically, this empowering effect is more significant in heavy-pollution industries.

3. Sample Selection and Data Sources

3.1. Sample Selection and Data Sources

This study utilizes Chinese A-share listed companies from 2013 to 2024 as the initial research sample. Data were obtained from several authoritative databases. The dependent variable, ESG performance, is derived from the Huazheng ESG rating system. This system covers three dimensions—Environmental (E), Social (S), and Governance (G)—and classifies firms into nine grades: C, CC, CCC, B, BB, BBB, A, AA, and AAA. Following Jin et al. [29] and Zhu et al. [30],we assign numerical values from 1 to 9 to these grades in ascending order and use the annual average of a firm’s quarterly ratings as the proxy for its ESG performance.

The primary independent variable, Digital Asset Allocation (dig), is measured following the approach of Song et al. (2026) [31]. By manually extracting and categorizing intangible asset details from the footnotes of annual reports, we identify items and patents containing keywords such as “software,” “network,” “client terminal,” “management system,” and “intelligent platform” as “digital technology intangible assets.” The sum of these items is divided by the total intangible assets of the year to represent the degree of digital transformation [31]. Other financial control variables are retrieved from the CSMAR database. Internal control quality data are sourced from the DIB Internal Control Index of Chinese Listed Companies. Data on green technology innovation (green patents) are obtained from the CNRDS and the CSMAR patent module, identified based on the WIPO Green Patent IPC categorization [30].

In terms of sample screening, the initial sample was cleaned step by step according to the following four criteria. First, listed firms in the financial and insurance industries were excluded, as such firms possess unique asset structures and exhibit ESG performance logics that differ substantially from those of non-financial real-sector firms; their direct inclusion would therefore introduce heterogeneity bias into the sample. Second, firms subject to special treatment, such as ST and *ST status, were excluded because the financial conditions and ESG of these operationally distressed firms are unlikely to reflect a normal business state. Third, observations with severely missing values for key variables were removed, while the remaining scattered missing values were handled through listwise deletion to ensure the validity of the regression sample. Fourth, firms with a listing history of less than one year were excluded to ensure the completeness and comparability of financial data. In addition, to mitigate the influence of extreme outliers on the regression results, all continuous variables were winsorized at the 1st and 99th percentiles. After this series of screening procedures, the final sample consisted of an unbalanced panel covering the period from 2013 to 2024, with 29,329 firm-year observations. As an additional sensitivity test for outlier treatment, we further applied a more stringent winsorization threshold at the 2.5th and 97.5th percentiles to all continuous variables and re-estimated the baseline model. The coefficient of digital asset allocation remains positive and statistically significant (coefficient = 0.0019, t = 4.325), indicating that the main conclusion is not sensitive to a more conservative outlier-treatment threshold.

3.2. Variable Definition and Measurement

(1)

Dependent variable: ESG performance (esg)

This study measures the ESG performance of listed firms using the Huazheng ESG rating system. ESG ratings are converted into numerical scores ranging from 1 to 9, in ascending order from the lowest to the highest grade, and the annual mean of a firm’s quarterly ESG ratings is used as a composite proxy for its ESG performance in a given year so as to smooth out random fluctuations associated with any single rating observation. We acknowledge that ESG ratings are ordinal categories rather than naturally interval-scaled variables. Therefore, the 1–9 conversion used in the main regression should be understood as an ordered scoring proxy that preserves the ranking information of ESG grades and improves interpretability in panel regressions, rather than as a strict assumption that the distance between adjacent rating categories is exactly equal. This treatment is consistent with the empirical practice adopted in prior ESG studies using rating-grade transformations, such as Jin et al. [29] and Zhu et al. [30]. To further alleviate concerns about this measurement treatment, we conduct an additional robustness test by converting the ESG score into a high/low ESG dummy variable. Specifically, firms with ESG scores above the sample median are coded as 1, and the remaining firms are coded as 0. The estimated coefficient of digital asset allocation remains positive and statistically significant (coefficient = 0.0009, t = 3.634), indicating that the main conclusion is not driven by the interval-score transformation of ESG ratings.

(2)

Primary independent variable: digital asset allocation (dig)

This study uses the ratio of digital-related intangible assets to total intangible assets as a proxy for the extent of a firm’s digital asset allocation. Specifically, the measure is constructed through manual collection of detailed intangible asset items disclosed in the notes to the financial statements of listed firms’ annual reports. When a particular item contains digital technology-related keywords—such as software, network, client terminal, management system, or intelligent platform—as well as relevant digital technology patents, it is classified as a digital intangible asset. For each firm-year observation, all such digital intangible assets are aggregated, and their proportion in total intangible assets for the corresponding year is calculated to obtain dig. Compared with the absolute scale of investment, this ratio-based measure more effectively captures the underlying restructuring of factor allocation and strategic resource reorientation at the firm level—that is, the firm’s deliberate tendency to concentrate its limited stock of intangible assets toward digitalization. In doing so, it largely removes the scale effect arising from differences in firm size and thereby enhances the cross-sectional comparability of digitalization levels across firms of different sizes.

(3)

Mediating variables: internal control quality and green technology innovation

Internal Control Quality (lnic) is measured using the DIB Internal Control Index, with the natural logarithm of the original index applied to correct for right-skewness and improve its statistical properties. Developed on the basis of the Committee of Sponsoring Organizations of the Treadway Commission (COSO) framework and China’s internal control regulatory guidelines [31], the DIB index has been widely adopted in empirical research as a standard proxy for firms’ internal control quality.

Green Technology Innovation (LnGreen_Inv) is measured by the natural logarithm of the number of green invention patent applications plus one filed independently or jointly by the firm in a given year. Invention patents are preferred to utility model patents in order to exclude strategic or low-quality patent filings and more accurately capture substantive green innovation. In addition, patent applications, rather than patent grants, are used because they more directly reflect a firm’s immediate innovation intention under the influence of digital asset allocation, while avoiding the time-lag problem inherent in the patent examination and granting process.

(4)

Moderating variable: Industry pollution intensity: Industry Pollution Intensity (Ind_Pollute) is operationalized as a dummy variable. In accordance with the Industry Classification Management Catalog for Environmental Protection Verification of Listed Companies issued by the former Ministry of Environmental Protection, firms operating in heavily polluting industries, such as thermal power, steel, cement, electrolytic aluminum, and chemicals, are assigned a value of 1, whereas firms in all other industries are assigned a value of 0.

(5)

Control Variables: To account for other potential determinants of ESG performance, we include a series of control variables spanning financial characteristics and corporate governance. The detailed definitions are presented in

Table 1. Definitions of Control Variables.

Variable Category	Variable Name	Symbol	Measurement
Financial Characteristics	Firm Size	size	Natural logarithm of total assets at the end of the year
	Leverage	lev	Total liabilities divided by total assets at the end of the year
	Profitability	roa	Net income divided by average total assets
	Growth	growth	Growth rate of operating income, calculated as the increase in operating income in the current year divided by operating income in the previous year
	Tobin’s Q	tobinq	(Market value of equity + total liabilities) divided by total assets at book value
Corporate Governance	Board Size	board	Natural logarithm of the total number of directors on the board
	Independent Director Ratio	indep	Percentage of independent directors on the board (%)
	Duality	dual	A dummy variable = 1 if the chairman and CEO are the same person, and 0 otherwise
	Ownership Concentration	top1	Shareholding ratio of the largest shareholder (%)

Note: All data for the control variables above are obtained from the China Stock Market & Accounting Research (CSMAR) Database. Board size, independent director ratio, duality, and ownership concentration are classified as corporate governance variables rather than financial characteristics.

.

3.3. Model Specification

To investigate the impact of digital asset allocation on corporate ESG performance, this study develops a baseline panel regression model incorporating two-way fixed effects:

$${ESG}_{i,t}= {\alpha }_0 + {\alpha }_1·{Dig}_{i,t} + {\sum }_k{\beta }_k·{Controls}_{i,t} +{ Industry}_i + Y{ear }_t+ {\epsilon }_{i,t}$$

(1)

where $i$ and $t$ denote firm and year, respectively. The dependent variable, $ES{G}_{i,t}$, measures the ESG performance of firm $i$ in year $t$. The key independent variable, $Di{g}_{i,t}$, captures the level of digital asset allocation. $Control{s}_{i,t}$ is a vector of firm-level control variables, including financial indicators and corporate governance characteristics. ${\mu }_i$ denotes firm fixed effects, which control for time-invariant unobservable firm characteristics. ${\lambda }_t$ represents year fixed effects, capturing macroeconomic trends and policy shocks that are common to all firms in a given year. ${\epsilon }_{i,t}$ is the idiosyncratic error term. Standard errors are clustered at the firm level to account for potential heteroscedasticity and serial correlation. This paper further examines the mechanism through which digital asset allocation affects corporate ESG performance, focusing on internal control quality, green technological innovation, and the sequential transmission channel of “internal control promoting green innovation”. Preliminary analysis shows that internal control quality exerts a significantly positive impact on green innovation. Accordingly, this study adopts a sequential mediation test with Bootstrap resampling (1000 replications) and firm-clustered standard errors.

Before estimating the regression models, we conducted Pearson correlation diagnostics and variance inflation factor (VIF) tests to assess potential multicollinearity among digital asset allocation and the control variables. The maximum absolute pairwise correlation coefficient is 0.574, which is below the commonly used threshold of 0.60. The mean VIF is 1.35 and the maximum VIF is 1.69, both far below the conventional warning threshold of 10. These diagnostics indicate that multicollinearity is unlikely to materially affect the stability of the estimated coefficients.

Bootstrap decomposition further verifies three types of indirect effects. These findings provide empirical evidence supporting the innovation compensation hypothesis, indicating that strategic digital investment can improve corporate governance efficiency and systematically enhance corporate ESG performance by stimulating green technological innovation. The specific models are as follows:

First, we examine the total effect of digital asset allocation on corporate ESG performance:

$${ESG }_{i,t}= {\alpha }_0 +{ \alpha }_1{Dig }_{i,t}+ \Sigma \beta k{Controls }_{i,t}+ {Industry}_i +{ Year}_t +{ \epsilon }_{i,t}$$

(2)

Second, we examine the effect of digital asset allocation on the mediator variable, green technological innovation:

$${LnGreen\_Inv}_{i,t}= {\gamma }_0 + {\gamma }_1{Dig}_{i,t} + \Sigma \beta k{Controls}_{i,t} + {Industry}_i + {Year}_t + {\epsilon }_{i,t}$$

(3)

Third, we include both digital asset allocation and green technological innovation in the ESG regression to test the mediating effect:

$${ESG}_{i,t}= {\delta }_0 + {\delta }_1{Dig}_{i,t} + \delta 2Ln{Green\_Inv}_{i,t} + \Sigma \beta k{Controls}_{i,t} + {Industry }_i+{ Year }_t+ {\epsilon }_{i,t}$$

(4)

In the above models, $ES{G}_{i,t}$ denotes the ESG performance of firm i in year t; ${Dig}_{i,t}$ represents the level of digital asset allocation; and $LnGreen\_In{v}_{i,t}$ refers to the natural logarithm of green technological innovation. $Control{s}_{i,t}$ is a vector of firm-level control variables. $Industr{y}_i$ and $Yea{r}_t$ denote industry and year fixed effects, respectively, while ${\epsilon }_{i,t}$ is the error term. ${\alpha }_1$ captures the total effect of digital asset allocation on ESG performance, ${\gamma }_1$ measures the effect of digital asset allocation on green technological innovation, ${\delta }_1$ reflects the direct effect of digital asset allocation on ESG performance after controlling for the mediator, and ${\delta }_2$ captures the effect of green technological innovation on ESG performance.

A mediating effect is supported if: (1) ${\alpha }_1$ in Equation (2) is statistically significant; (2) ${\gamma }_1$ in Equation (3) is statistically significant; and (3) ${\delta }_2$ in Equation (4) is statistically significant.

If the coefficient on $Dig{i}_{i,t}$ decreases in magnitude after the mediator is included (i.e., $\mid {\delta }_1\mid < \mid {\alpha }_1\mid$), this suggests a mediating effect. If ${\delta }_1$ becomes statistically insignificant, the mediation is considered full mediation; otherwise, it is partial mediation.

3.4. Limitations and Countermeasures

Potential data limitations exist in this study, including missing values, extreme outliers of sample data and subjectivity in the manual collection of digital asset data. Corresponding solutions are adopted: strict sample screening, winsorization of all continuous variables at the 1% and 99% percentiles, and unified collection criteria combined with double cross-checks to ensure data quality and empirical model reliability.

3.5. Descriptive Statistics and Correlation Analysis

Descriptive statistics show that ESG performance has a mean of 4.121 and a standard deviation of 0.905, reflecting a generally medium-to-high level of ESG among listed companies with obvious individual differences. The core explanatory variable dig has a mean of 17.184% and a much higher standard deviation of 25.890, a feature of high dispersion that indicates significant cross-sectional heterogeneity in digital intangible asset allocation among listed companies, providing sufficient statistical variation for subsequent effect identification (

Table 2. Descriptive Statistics of Main Variables.

Variable Category	N	Mean	SD	Min	Median	Max
esg	29,329	4.121	0.905	1.75	4	6.5
dig	29,329	17.184	25.890	0.04	6.18	100
lnic	29,329	6.272	1.104	0	6.48	6.69
size	29,329	22.27	1.276	20.02	22.08	26.11
lev	29,329	0.411	0.203	0.06	0.4	0.89
roa	29,329	0.04	0.066	−0.21	0.04	0.21
growth	29,316	0.139	0.343	−0.52	0.09	1.74
tobinq	29,329	1.995	1.197	0.83	1.61	7.34
board	29,322	2.103	0.195	1.61	2.2	2.56
indep	29,322	37.858	5.327	33.33	36.36	57.14
dual	29,329	0.317	0.465	0	0	1
top1	29,329	0.334	0.146	0.09	0.31	0.72

Note: All continuous variables have been winsorized at the 1% and 99% percentiles.

).

Table 2 presents the descriptive statistical results of the main variables, while

Table 3. Correlation and Multicollinearity Diagnostics.

Diagnostic Item	Reported Value	Criterion	Interpretation
corr(dig, esg)	0.046	Pairwise correlations below 0.60	Positive preliminary association; no high pairwise collinearity
corr(lnic, esg)	0.176	Pairwise correlations below 0.60	Supports the internal-control mechanism at the univariate level
Maximum absolute pairwise correlation	0.574	<0.60	No severe correlation-based multicollinearity
Mean VIF	1.35	<10	Within the conventional acceptable range
Maximum VIF	1.69	<10	No serious VIF-based multicollinearity

reports the correlation diagnostics for the core variables and the VIF-based multicollinearity diagnostics for the baseline regressors.

As shown in Table 3, the Pearson correlation coefficient between dig and esg is 0.046 and positive, preliminarily supporting the core hypothesis at the univariate level. The coefficient between internal control quality (lnic) and esg is 0.176, initially supporting the “management empowerment” transmission logic. Moreover, the maximum absolute pairwise correlation coefficient is 0.574, below the commonly used threshold of 0.60. The VIF diagnostics further show a mean VIF of 1.35 and a maximum VIF of 1.69, both far below the conventional threshold of 10, suggesting that multicollinearity is unlikely to distort the subsequent regression estimates.

4. Empirical Results and Discussion

4.1. Baseline Regression: The Main Effect of Digital Asset Allocation on ESG Performance

Table 4. Benchmark Regression.

	(1)	(2)
Variable	ESG	ESG
size	0.258 ***	0.262 ***
	(28.390)	(28.862)
lev	−0.977 ***	−0.982 ***
	(−18.462)	(−18.620)
roa	2.767 ***	2.763 ***
	(22.440)	(22.501)
growth	−0.038 **	−0.038 **
	(−2.276)	(−2.265)
tobinq	−0.014 **	−0.015 **
	(−2.027)	(−2.230)
board	0.154 ***	0.153 ***
	(2.944)	(2.919)
indep	0.010 ***	0.010 ***
	(6.010)	(5.983)
dual	0.048 ***	0.047 ***
	(3.055)	(2.988)
top1	0.304 ***	0.303 ***
	(5.121)	(5.113)
dig		0.002 ***
		(5.166)
_cons	−2.140 ***	−2.257 ***
	(−9.350)	(−9.831)
N	29309	29309
Adj. R2	0.240	0.242
F	272.395	248.718

Note: t-statistics are reported in parentheses; standard errors are clustered at the firm level; *** p < 0.01, ** p < 0.05; industry and year fixed effects are controlled.

reports the benchmark regression results of digital asset allocation on corporate ESG performance. A stepwise regression approach is adopted: Column (1) only includes control variables, and Column (2) further incorporates the core explanatory variable dig to examine the stability of the main effect coefficient and rule out the interference of omitted variables. All models control for industry and year fixed effects and employ firm-clustered robust standard errors.

Results for control variables show that firm size and profitability yield significantly positive coefficients, indicating that firms with stronger resource endowments have greater investment capacity and stakeholder response incentives for ESG. The significantly negative coefficient of leverage reflects that highly leveraged firms are constrained by debt repayment pressure, leading to limited free cash flow for ESG activities. These findings are highly consistent with existing literature.

Results in Column (2) show that the coefficient of dig is 0.002, significantly positive at the 1% level, which robustly supports Hypothesis H1. Economically, a 1-percentage-point increase in the proportion of digital intangible assets raises the corporate ESG score by an average of 0.002 points; scaled by one standard deviation, the actual impact is about 0.052 points, with non-negligible economic significance. This suggests that allocating a higher proportion of intangible assets to digitalization effectively enhances firms’ internal information integration capacity, reduces agency costs, and thus systematically improves ESG performance through the governance empowerment mechanism. The Adjusted R² of Column (2) is 0.242, indicating the model has good overall explanatory power.

4.2. Endogeneity Treatment: Instrumental Variable Approach (IV-2SLS)

To address potential bidirectional causality—where firms with higher ESG ratings may possess more resources and stronger incentives for digital investment—and omitted variable bias, the Two-Stage Least Squares (2SLS) method with dual instrumental variables is employed.

The first instrument (iv), following Jiang et al., is the interaction between the number of post offices in each city in 1984 and the previous year’s total national internet users (log-transformed). Historical postal infrastructure shaped local IT diffusion via path dependence, establishing a macro-foundation for corporate digital investment (relevance), while the 1984 distribution predates modern ESG frameworks, minimizing direct influence on ESG decisions (exogeneity) [21].

The second instrument (dige) is the one-period lagged provincial digital economy development index, constructed via PCA from five indicators: internet development (broadband subscribers, computer services employment, telecom volume, mobile users) and digital financial inclusion (Peking University Index). Provincial policy and infrastructure levels drive firm digitalization (relevance), and the lag reduces simultaneity with macro shocks while avoiding direct impact on ESG scores.

Instrument validity is confirmed (

Table 5. Endogeneity Treatment: Regression Results of IV-2SLS Instrumental Variable Method.

Variables	ESG (IV-2SLS)
dig	0.016 ** −0.008
size	0.317 *** −0.024
lev	−1.088 *** −0.071
roa	2.531 *** −0.14
growth	−0.049 *** −0.018
tobinq	−0.018 * −0.009
board	0.143 ** −0.066
indep	0.010 *** −0.002
dual	0.041 ** −0.021
top1	0.315 *** −0.075
Observations	24,497
LM p-value (Underidentification Test)	0.001
Cragg-Donald Wald F	41.32
Kleibergen-Paap Wald F	7.54
Hansen J p-value	0.104
Controls/Year/Industry FE	YES

Note: Robust standard errors are reported in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1. The two instrumental variables are: (1) the number of post offices in 1984 × the national number of internet users (iv); and (2) the one-period lagged provincial digital economy composite index (dige).

). The Kleibergen-Paap rk LM test rejects under-identification, the Cragg-Donald F-statistic and Kleibergen-Paap Wald F exceed Stock-Yogo thresholds, excluding weak instrument concerns, and the Hansen J test supports exogeneity. Second-stage 2SLS results show dig positively affects ESG performance, consistent with baseline regressions, indicating effective correction of endogeneity. The economic interpretation of the significant second-stage coefficients is also consistent with the baseline results. Digital asset allocation remains significantly positive, indicating that after accounting for potential endogeneity, firms with a higher proportion of digital intangible assets still exhibit better ESG performance. Firm size and profitability are positively associated with ESG performance, suggesting that resource endowment and operating performance strengthen firms’ capacity to invest in sustainability governance. Leverage is negatively associated with ESG performance, implying that debt pressure may crowd out long-term ESG investment and managerial attention.

4.3. Robustness Tests

To further verify the robustness of the baseline results, multidimensional tests were conducted from four perspectives: alternative variable measurements, sample adjustments, ESG rating recoding, and outlier-treatment sensitivity (

Table 6. Robustness Test: Variable Measurement Replacement and Special Industry Exclusion.

Variables	(1) Replace Dependent Variable ESG_med	(2) Replace the Independent Variable: Substitute DCG with ESG	(3) Exclude Samples from the Finance and Real Estate Industries, and Replace DIG with ESG
dig/dcg/dig	0.0019 *** 0	0.05 *** −0.007	0.0019 *** 0
size	0.288 ***	0.250 ***	0.268 ***
lev	−1.048 ***	−0.970 ***	−1.017 ***
roa	2.953 ***	2.800 ***	2.748 ***
observed value	29,309	29,308	26,060
Adj. R2	0.231	0.246	0.239
Industry/Annual Forecast Estimate	YES	YES	YES

Note: Column (1) replaces the dependent variable with the median Environmental, Social, and Governance performance score; Column (2) replaces the independent variable with the digitalization index constructed by the textual word frequency method; Column (3) excludes firms in the financial and real estate industries. Robust standard errors are reported in parentheses; *** p < 0.01.

and

Table 7. Robustness Test: Lagged Explanatory Variable (L.dig).

Variables	(1) ESG (L.dig)
L.dig	0.0022 *** 0
size	0.2793 *** −0.011
lev	−1.0278 *** −0.062
roa	2.6826 *** −0.146
observed value	20,889
Adj. R2	0.248
Industry/Annual Forecast Estimate	YES

Note: Robust standard errors are reported in parentheses; *** p < 0.01; industry and year fixed effects are controlled.

). (1) Alternative Dependent Variable. Replacing the original Huazheng ESG rating with the median ESG score by “Year-Industry” yields a dig coefficient of 0.0019, significant at the 1% level, consistent with the baseline, indicating independence from the specific ESG framework. (2) Alternative Independent Variable. Using a text-based digitalization index derived from the frequency of 76 keywords across five dimensions in the MD&A sections (log-transformed), the coefficient is 0.05, significant at 1%, supporting the baseline measure. (3) Sample Adjustments. Excluding financial, insurance (Industry I), and real estate (Industry K) firms, which differ in accounting standards and ESG weighting, the dig coefficient remains 0.0019, confirming the effect across real-sector enterprises. (4) ESG Rating Recoding. To address the ordinal nature of ESG ratings, the ESG score is transformed into a high/low ESG dummy variable based on the sample median. The coefficient of dig remains positive and statistically significant (coefficient = 0.0009, t = 3.634), suggesting that the baseline conclusion is not dependent on the 1–9 interval-score transformation. (5) Lagged Independent Variable. Using the one-period lag of digital asset allocation (L.dig) to address potential reverse causality and capture cumulative effects, the coefficient is 0.0022, significant at 1%, demonstrating intertemporal stability. Overall, across alternative measurements, sample restrictions, a more stringent 2.5%/97.5% winsorization threshold, and time-lagged specifications, results remain consistent, validating the reliability of the baseline regression.

4.4. Mechanism Test

This subsection examines the mediating mechanisms through which digital asset allocation shapes corporate ESG performance. Consistent with the theoretical analysis, internal control quality and green technological innovation are treated as mediators rather than moderators. Specifically, the study tests whether digital asset allocation improves ESG performance through internal control quality, through green technological innovation, and through the sequential pathway in which improved internal control quality further promotes green technological innovation. To verify these indirect effects, a serial mediation model with Bootstrap resampling (1000 iterations) and firm-clustered standard errors is adopted.

The stepwise regression results reported in

Table 8. Sequential Mediation Test: Regression Results and Bootstrap Indirect Effect Decomposition.

Variables	(0) ESG	Baseline (1) Internal Control Quality	Lnic (2) Green Innovation	LnGreen_Inv (3) ESG
dig	0.0019 *** −4.396	0.0009 ** −1.994	0.0028 *** −5.033	0.0016 *** −3.815
lnic	—	—	0.0193 *** −3.398	0.0703 *** −12.231
LnGreen_Inv	—	—	—	0.0757 *** −7.816
size	0.2737 ***	0.0509 ***	0.4591 ***	0.2353 ***
lev	−1.0280 ***	−0.3695 ***	0.0158	−1.0026 ***
roa	2.7138 ***	3.5111 ***	−0.3246 **	2.4864 ***
observed value	25,648	25,648	25,648	25,648
Adj. R2	0.238	0.077	0.392	0.249
Industry/Year FE	YES	YES	YES	YES
Indirect Effect 1 [95% CI]	—	—	—	0.0000645 ** [2 × 10−6, 1.3 × 10−4]
Indirect Effect 2 [95% CI]	—	—	—	0.0002097 *** [1.2 × 10−4, 3.2 × 10−4]
Indirect Effect 3 [95% CI]	—	—	—	1.34 × 10−6 * [2.6 × 10−8, 3.1 × 10−6]

*** p < 0.01, ** p < 0.05, * p < 0.1.

show that digital asset allocation significantly improves internal control quality. Digital asset allocation and internal control quality both significantly promote green technological innovation. After internal control quality and green technological innovation are included in the ESG regression, the coefficient of digital asset allocation remains positive and significant but decreases in magnitude, indicating partial mediation.

The Bootstrap decomposition further confirms three statistically significant indirect effects. First, the indirect effect through internal control quality is significant at the 5% level, with a 95% confidence interval excluding zero. Second, the indirect effect through green technological innovation is significant at the 1% level. Third, the sequential indirect effect through internal control quality and subsequent green technological innovation is significant at the 10% level. These results indicate that green technological innovation is the primary mediating channel, while internal control quality also enables the sequential transmission from digital asset allocation to ESG performance.

Overall, the mediation evidence supports the innovation compensation hypothesis: strategic digital investment improves corporate governance efficiency and stimulates green technological innovation, thereby systematically enhancing ESG performance.

4.5. Moderating Effect Test

To further examine the boundary conditions of digital asset allocation’s impact on ESG performance, this subsection introduces industry pollution intensity (Ind_Pollute) as a moderator and constructs an interaction term between digital asset allocation and industry pollution intensity.

Table 9. Moderating Effect of Industry Pollution Intensity on the Relationship Between Digital Asset Allocation and ESG Performance.

Variables	ESG
Dig	0.003 ***
	(0.000)
Ind_Pollute = 1, omitted	-
0b.Ind_Pollute#co.dig	0.00
	(0.000)
1.Ind_Pollute#c.dig	0.00 *
	(0.001)
Size	0.27 ***
	(0.010)
Lev	−1.02 ***
	(0.057)
ROA	2.75 ***
	(0.134)
Growth	−0.04 **
	(0.018)
TobinQ	−0.01 *
	(0.007)
Board	0.16 ***
	(0.055)
Indep	0.01 ***
	(0.002)
Dual	0.05 ***
	(0.017)
Top1	0.29 ***
	(0.063)
Constant	−2.42 ***
	(0.243)
Observations	26,060
R-squared	0.239
Control Variables	YES
Year FE	YES
Industry FE	YES

(I) Robust standard errors in parentheses; (II) *** p < 0.01, ** p < 0.05, * p < 0.1.

reports the corresponding moderation regression results.

The regression results show that the coefficient of the interaction term between digital asset allocation and industry pollution intensity is 0.0025 and is significantly positive at the 10% level. This indicates that industry pollution intensity positively moderates the relationship between digital asset allocation and ESG performance: the ESG-enhancing effect of digital asset allocation is stronger among heavily polluting firms. Therefore, Hypothesis H3 is supported.

The theoretical mechanisms underlying this moderating effect are twofold. First, differentiated regulatory pressure amplifies the marginal governance benefits of digitalization. Heavily polluting firms face stricter emission standards and more intensive public scrutiny. Digital assets such as industrial internet platforms and real-time emission monitoring systems enable more accurate and timely control of pollution sources, reduce non-compliance risks, and generate a stronger green compensation effect.

Second, the signaling mechanism produces a stronger value premium in heavily polluting industries. Because these firms often have lower historical ESG baselines, ESG improvements driven by digital asset allocation are more visible to investors, regulators, and other stakeholders. Digital asset allocation therefore becomes an effective tool for reshaping corporate ESG images in pollution-intensive industries. Taken together, these findings confirm that industry pollution intensity is a moderator, rather than a mediating mechanism, in the relationship between digital asset allocation and ESG performance.

4.6. Heterogeneity Analysis

A firm’s resource endowment, institutional background, and life-cycle characteristics may systematically influence the efficiency and depth of the transmission from digital asset allocation to ESG performance. To explore these dynamics, this study conducts heterogeneous group tests across three dimensions: corporate life cycle, nature of ownership, and financing constraints, aiming to uncover the structural differences in the ESG-enabling effects of digitalization.

4.6.1. Heterogeneity Based on Corporate Life Cycle

Drawing on Dickinson’s (2011) cash flow combination classification method [16], the sample is divided into growth, mature and decline stages, with the group regression results reported in

Table 10. Heterogeneity Test Based on Enterprise Life Cycle.

	(1)	(2)	(3)
Variables	ESG	ESG	ESG
dig	0.002 ***	0.001 *	0.002 ***
	(4.520)	(1.906)	(3.302)
size	0.280 ***	0.232 ***	0.243 ***
	(22.781)	(11.095)	(12.473)
lev	−1.016 ***	−1.000 ***	−1.243 ***
	(−13.275)	(−8.886)	(−12.550)
roa	2.755 ***	2.011 ***	2.410 ***
	(13.674)	(7.011)	(10.225)
growth	−0.023	0.005	0.062
	(−0.750)	(0.136)	(1.398)
tobinq	−0.007	−0.041 ***	−0.012
	(−0.724)	(−2.642)	(−0.812)
board	0.182 **	0.111	0.094
	(2.555)	(0.882)	(1.044)
indep	0.011 ***	0.009 **	0.008 **
	(4.643)	(2.162)	(2.563)
dual	0.055 **	0.056	−0.007
	(2.357)	(1.636)	(−0.205)
top1	0.323 ***	0.238 *	0.530 ***
	(4.007)	(1.834)	(4.523)
_cons	−2.768 ***	−1.309 **	−1.649 ***
	(−9.029)	(−2.420)	(−3.522)
N	10430	3235	3880
Adj. R2	0.256	0.255	0.275

Note: Columns 1–3 report results for samples in the growth, mature, and decline stages, respectively; *** p < 0.01, ** p < 0.05, * p < 0.1.

. The coefficient of dig is significantly positive across all three stages, presenting a distinct U-shaped pattern of being stronger in the early and late stages but weaker in the mature stage: the coefficient stands at 0.002 for the growth stage, 0.001 for the mature stage, and 0.002 for the decline stage.

This pattern aligns with the logic of corporate strategic behavior. Firms in the growth stage leverage digital intangible assets to rapidly build core competencies, and integrating ESG practices into their digital layout helps attract institutional capital and reduce financing frictions, resulting in the most sufficient driving force for digital empowerment. Firms in the decline stage are under pressure to break through development bottlenecks; synergizing digital transformation with ESG improvement aids in rebuilding stakeholder trust and securing policy support, leading to equally strong marginal empowerment effects. In contrast, mature firms feature rigid management structures, and organizational inertia creates greater internal resistance to the introduction of digital assets, thus restricting the conversion efficiency of digitalization into ESG outcomes.

4.6.2. Heterogeneity Test Based on Ownership Type

Group regression results are presented in

Table 11. Heterogeneity Test Based on Ownership Type.

	(1)	(2)
Variables	ESG	ESG
dig	0.002 ***	0.002 ***
	(3.063)	(4.168)
size	0.298 ***	0.241 ***
	(19.044)	(19.975)
lev	−1.127 ***	−0.918 ***
	(−11.491)	(−14.773)
roa	2.719 ***	2.778 ***
	(10.979)	(19.693)
growth	−0.045	−0.024
	(−1.430)	(−1.199)
tobinq	−0.024 *	−0.011
	(−1.655)	(−1.387)
board	0.174 *	0.111 *
	(1.818)	(1.801)
indep	0.010 ***	0.009 ***
	(3.282)	(4.294)
dual	−0.002	0.059 ***
	(−0.054)	(3.399)
top1	0.180	0.254 ***
	(1.602)	(3.606)
_cons	−2.970 ***	−1.645 ***
	(−7.708)	(−5.364)
N	9318	19,989

Robust standard errors in parentheses; *** p < 0.01, * p < 0.1.

. The coefficient of dig is highly significant and positive at 0.002 for state-owned enterprises and 0.002 for non-state-owned enterprises, indicating that digital asset allocation driving ESG improvement is a universal law transcending ownership boundaries. Nevertheless, the higher t-value for non-state-owned enterprises reveals their relatively superior efficiency in converting digitalization into ESG performance amid market competition pressure. This disparity stems from fundamental differences in institutional contexts: non-state-owned enterprises are directly exposed to the pressures of product market competition and ESG rating screening in the capital market, resulting in a stronger profit-driven mechanism to rapidly translate digital investment into ESG performance. In contrast, the behavioral logic of state-owned enterprises is intertwined with policy implementation orientation and reliance on administrative resources, leading to a relatively longer market-oriented conversion chain and a certain degree of conversion frictions.

4.6.3. Heterogeneity Test Based on Financing Constraints

The SA index is adopted to measure the degree of corporate financing constraints, and the sample is divided into high and low financing constraint groups by the annual sample median, with the regression results reported in

Table 12. Results of Grouped Test for Financing Constraints.

	(1)	(2)
Variables	High Financing Constraints Group	Low Financing Constraints Group
dig	0.001 ***	0.002 ***
	(0.001)	(0.001)
size	0.279 ***	0.248 ***
	(0.012)	(0.014)
lev	−1.054 ***	−0.956 ***
	(0.075)	(0.074)
roa	2.323 ***	3.175 ***
	(0.179)	(0.182)
growth	−0.034	−0.058 **
	(0.026)	(0.024)
tobinq	−0.007	−0.022 **
	(0.010)	(0.009)
board	0.093	0.250 ***
	(0.071)	(0.071)
indep	0.010 ***	0.011 ***
	(0.002)	(0.002)
dual	0.034	0.039 *
	(0.022)	(0.022)
top1	0.181 **	0.321 ***
	(0.080)	(0.082)
Constant	−2.384 ***	−2.262 ***
	(0.299)	(0.344)
Observations	13,028	13,031
R-squared	0.250	0.240
Control Variables	YES	YES
Year FE	YES	YES
Industry FE	YES	YES

Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

.

The coefficient of dig is 0.002 for the low financing constraint group and 0.001 for the high financing constraint group. The asymmetric coefficient gap between the two groups verifies the core logic of the resource slack theory. Both digital asset accumulation and ESG responsibility fulfillment are characterized by large investment scales and long payback periods, forming a highly complementary, long-cycle and capital-intensive strategic portfolio. Firms with low financing constraints, endowed with abundant resource slack, can overcome the initial pains of digital transformation and systematically translate technological accumulation into long-term ESG outcomes. In contrast, firms with high financing constraints tend to adopt short-term resource allocation strategies due to survival pressure, making it difficult for them to fully channel limited digital dividends into ESG initiatives, which weakens the long-term enabling effect of digitalization in the dimension of sustainable development.

5. Conclusions and Policy Implications

5.1. Main Conclusions

Using a sample of Chinese A-share listed firms from 2013 to 2024, this study examines the impact of digital asset allocation, measured as the ratio of digital intangible assets to total intangible assets, on corporate ESG performance and its underlying mechanisms. Five key findings emerge. First, digital asset allocation significantly enhances ESG performance. Baseline regressions with two-way fixed effects and clustered standard errors show a positive and robust effect, which remains consistent when using alternative ESG measures, text-based digitalization indices, lagged variables, or excluding financial and real estate firms. Instrumental variable analysis confirms the causal impact and mitigates endogeneity. Second, internal control quality and green technological innovation serve as mediating channels. Digital asset allocation improves ESG performance through internal control quality, through green technological innovation, and through the sequential pathway from internal control quality to green technological innovation. Third, industry pollution intensity plays a positive moderating role: the ESG-enhancing effect of digital asset allocation is stronger in pollution-intensive industries, suggesting that regulatory pressure and public scrutiny amplify the green benefits of digital investment. Fourth, the impact varies across firm life-cycle stages and ownership. Firms in growth and decline stages benefit more than mature firms, and non-state-owned firms show slightly stronger ESG gains, reflecting higher efficiency in converting digital assets into ESG outcomes under competitive conditions. Fifth, financing constraints limit the effectiveness of digitalization. Firms with greater resource slack achieve stronger ESG improvements, while constrained firms tend to adopt short-term resource allocation strategies, weakening the long-term benefits.

5.2. Policy Implications

Based on the findings, this study offers policy recommendations for governments, financial institutions, and firms.

First, incentives for digital intangible assets should be strengthened to support digital factor allocation. Beyond existing R&D tax deductions, investments in digital assets such as software systems and intelligent platforms should be eligible for targeted tax incentives. Simultaneously, accounting recognition and valuation frameworks for digital intangible assets should be improved to reduce institutional frictions, encouraging firms to expand digital asset holdings and transform digital technologies from operational tools into strategic resources.

Second, digital transformation should be integrated with ESG disclosure requirements, with enhanced guidance for pollution-intensive industries. Regulators should establish coordinated disclosure frameworks linking digitalization and ESG, requiring firms to report on the deployment and effectiveness of digital technologies in environmental monitoring and energy management. Prioritizing policy support for heavily polluting sectors can amplify the ESG benefits of digital investment.

Third, differentiated support mechanisms should address firm heterogeneity across life-cycle stages and financing constraints. Development-oriented financial institutions and government-backed funds should tailor support to firms’ life-cycle positions and resource limitations. For declining or highly constrained firms, instruments such as digital transformation loan guarantees or ESG-linked interest rates can directly tie financing costs to ESG outcomes, alleviating resource bottlenecks and facilitating the translation of digitalization into sustainable performance gains.

6. Limitations and Future Research

Despite its contributions, this study has several limitations that warrant further investigation.

First, the measure of digital asset allocation relies on the ratio of digital intangible assets to total intangible assets, extracted manually from annual report notes. This approach is limited in standardization and granularity. Future research could integrate IT expenditure data, large-scale text analysis, and alternative measurement methods to improve precision and coverage, capturing the structural dimensions of digitalization more comprehensively. Second, although internal control quality is identified as a mediating mechanism, digital asset allocation may also influence ESG performance through supply chain transparency, stakeholder management, and data governance. Future studies could incorporate a broader set of mediators and apply structural equation modeling or machine learning to systematically assess the relative importance of multiple pathways, deepening understanding of how digitalization translates into ESG outcomes. Third, the focus on listed firms may limit external validity. Unlisted small and medium-sized enterprises differ in digital asset intensity, financing constraints, and ESG reporting practices. Expanding analysis to SMEs using micro-level administrative data, such as business registration and tax records, could provide broader empirical support for inclusive ESG policies.