Can Digital Transformation Restrain Corporate ESG Greenwashing—A Test Based on Internal and External Joint Perspectives

Abstract

Digital technology has the function of information governance, and digital transformation of enterprises may be the key way to identify and restrain ESG greenwashing. Based on the theory of digital empowerment, this study analyzes the influence and mechanism of digital transformation on restraining corporate green washing behavior from the perspective of internal and external factors. This study takes A-share listed companies in 2012–2022 as research samples and tests the effectiveness of digital transformation. Research has found that (1) digital transformation can significantly suppress corporate greenwashing behavior, and this conclusion still holds after a series of endogeneity and robustness tests. (2) In the context of high environmental awareness among executives, the inhibitory effect of digital transformation on corporate ESG greenwashing is more pronounced. (3) Mechanism analysis shows that digital transformation has inhibited the company’s greenwashing behavior by increasing the attention of investors. (4) Heterogeneity analysis shows that in state-owned enterprises, non-heavily polluting industries, high-tech industries, and enterprises located in the eastern region digital transformation has a more effective inhibitory effect on corporate greenwashing behavior. This study examines the impact of digital transformation on corporate ESG greenwashing, expands the research on the non-economic effects of digital transformation, and may provide empirical evidence for improving the quality of ESG information disclosure and sustainable development of enterprises.

1. Introduction

Environmental, social, and governance (ESG) is a concept of sustainable development that considers the coordinated progress of the environment, society, and corporate governance. It aligns closely with the idea of green development. ESG has become a crucial indicator for evaluating a company’s environmental performance, its ability to communicate with stakeholders, and the effectiveness of its corporate governance (Martiny, 2024) [1]. To implement sustainable development, an increasing number of enterprises are integrating environmental, social, and governance factors into their business decisions. In September, 2015, China officially issued the Overall Plan for the Reform of the Ecological Civilization System, requiring the gradual establishment of a compulsory disclosure system of environmental information of listed companies. Guided by this plan, ESG information disclosure by Chinese enterprises has developed rapidly in recent years, with increasing initiatives for ESG information disclosure each year. However, a “catch me if you can” phenomenon often occurs between enterprises and the government. The overall effect of corporate ESG disclosure practices is not ideal. Increasingly, companies resort to greenwashing to gain fraudulent environmental benefits and resist environmental disclosure systems (Tuyen, 2023) [2]. This has become a significant obstacle to green development.

The concept of greenwashing was first put forward by American environmentalists in 1986. Initially, it refers to the act of disguising economic benefits as environmental protection. Nowadays, green cleaning has evolved into a tool for enterprises to establish an image of environmental responsibility through false, untrue, and inaccurate means (Du, 2015; Delmas and Montes-Sancho, 2011) [3,4]. As ESG practices in China lag behind those in the West, the corporate environmental information disclosure system is not yet fully developed, creating opportunities for greenwashing behavior. With increasing environmental protection pressures, more enterprises attempt to create an environmental image through greenwashing, aiming for high economic returns while minimizing environmental protection costs. If greenwashing becomes widespread, it will seriously undermine the effectiveness and enforcement of environmental laws and regulations, particularly eroding external trust in the formal system. Currently, most investors consider ESG information as a crucial indicator of a company’s development and closely examine the ESG system of target companies when making investment decisions (Albuquerque et al. 2013) [5]. The falsification of ESG information easily misleads investors’ judgment, disrupts the normal supply of green credit and green investment, and disrupts the normal order of the securities market (Delmas and Burbano, 2011) [6], resulting in incalculable economic consequences. In addition, greenwashing has deepened consumers’ green doubts about products and companies and further weakened their green purchase intention (Leonidou and Skarmeas, 2017) [7]. This shows that enterprises in greenwashing will further erode the green consumer market. Therefore, the “exaggerated” ESG disclosure and commitment of enterprises is the focus of academic circles, and improving greenwashing behavior is also a difficult problem needing to be solved urgently in green development.

In addition to the study of its economic consequences, the academic circles have also analyzed the motivation of “greenwashing” from different angles. From the regulatory point of view, due to the relatively loose regulation of ESG and the lack of uniform and strict standards for data disclosure, there is a great space for enterprises to perform greenwashing (Kim and Lyon, 2015) [8]. From the perspective of information transmission, the information obtained by consumers and investors is biased and asymmetric, which is conducive to the “greenwashing” behavior of enterprises (Hartman and Ibanna, 2006) [9]. From the perspective of the company’s own capabilities, when organizations with shortcomings in resources and capabilities are unable to achieve standard-compliant goals in the face of regulatory requirements, they are more inclined to passively avoid and adopt symbolic behaviors to divert the attention of stakeholders to cover up their own unethical business practices (Marquis and Toffe, 2012) [10], such as selective disclosure (Matejek and Gössling, 2014) [11]. For instance, many companies seeking government incentives and benefits may opt for greenwashing to secure unfair advantages (Delmas and Burbano, 2011) [6]. Furthermore, since ESG reports may contain crucial data about a company’s operations, the necessity to protect trade secrets could give rise to the moral hazard of “greenwashing” (Lyubov et al. 2023) [12].

It is evident that the majority of existing research on “greenwashing” is centered on the motivation and economic implications of such practices, yet the issue of how to mitigate greenwashing remains unaddressed.

As new digital technologies—such as the Internet, cloud computing, big data, and artificial intelligence—rapidly evolve, their integration with traditional industries is increasing. Consequently, more enterprises are joining the wave of digital transformation. Simultaneously, the phenomenon of digitalization has gradually attracted significant attention from the academic community. Can the digital transformation of enterprises serve as a tool to curb greenwashing?

First of all, digital transformation can empower economic development and management efficiency, and many scholars interpret the role of digital transformation from the perspective of economic management: for example, it can promote the upgrading of industrial structure, reduce some specific economic costs, and improve the efficiency of enterprise labor division and financial performance (Wang and Li, 2024, Goldfarb and Tucker, 2019, Commander et al., 2011, Jia and Wang, 2022) [13,14,15,16]. In addition, digital transformation also plays an important role in the process of greening enterprises. Digitalization can support the green innovation of enterprises by optimizing resource allocation (Tang et al., 2023) [17], and effectively promoting the green transformation of enterprises (Cheng et al., 2022) [18]. On the other hand, it can also reduce the pollution emissions per unit output value of enterprises and promote enterprises to improve carbon performance (Lv and Wu, 2024) [19]. Digital technology is also effective in the information governance of enterprises. Enterprises that undergo digital transformation are able to inhibit the management of earnings management, and this improves the comparability of accounting information (Yang etc. 2024) [20]. Digital transformation can break the information isolated island within the enterprise and help to improve the efficiency of information collection and communication within the enterprise (Kalthoff and Mike, 2015) [21]. In terms of regulation, the digital transformation of enterprises can also curb the behavior of infringement of information disclosure (Ma Defang, 2023) [22].

Based on the analysis of the above-mentioned literature, it can be seen that digitalization can help enterprises achieve corresponding results in corporate governance, business ethics, and information governance. ESG greenwashing is obviously inseparable from corporate management efficiency, green ethics, and information disclosure. Then, there may be a subtle connection between digital transformation and an enterprise’s greenwashing behavior.

Based on the aforementioned context, we are ready to explore and solve the following two research problems: (1) What impact does digital transformation have on corporate greenwashing behavior? (2) Through what mechanisms does digital transformation inhibit corporate greenwashing behavior?

The purpose of this article is to analyze the role of digital transformation in corporate ESG greenwashing behavior to reveal its impact on the quality of ESG information disclosure. Through empirical research, the specific mechanisms by which digital transformation suppresses corporate greenwashing behavior will be determined, providing references for corporate management practices and policy formulation.

Thus, starting from the theory of digital empowerment, this paper selects A-share listed companies from 2012 to 2022 as the research sample to empirically test the impact and mechanism of digital transformation on corporate greenwashing behavior.

The subsequent structure of this paper is arranged as follows: the second part is theoretical analysis and research hypothesis; the third part is the study design; the fourth part is empirical analysis; the fifth part is to further examine mechanisms and heterogeneity; and the final part is the conclusion and discussion of this study.

2. Theoretical Analysis and Research Hypothesis

2.1. Digital Transformation and ESG Greenwashing

As a new way to integrate emerging production factors and resources, digital transformation has attracted numerous enterprises to engage in its reformative wave. In addition to solving the problems of survival and growth, the enhancement of digitalization is expected to bring a qualitative leap in corporate ethics and business ethics. DeGiovanni (2021) proposed the concept of responsible digitization, advocating for digital transformation from the perspective of corporate social responsibility [23]. Responsible digitization can guide technology towards good, promoting green development while achieving digital transformation (Cardinali and DeGiovanni, 2022) [24]. Sun et al. (2021)’s empirical results revealed a significant role of enterprise digitalization in promoting social responsibility [25]. Additionally, relevant studies have demonstrated that digital transformation can significantly enhance the ESG performance of both listed companies and small and medium-sized enterprises (Wang and José, 2023; Zhao and Cai, 2023) [26,27]. Obviously, digital transformation is closely related to business ethics, and the quality of ESG information disclosure is an important aspect of business ethics. Therefore, this paper will explore the relationship between digital transformation and the quality of ESG information disclosure from the following three perspectives.

In terms of information governance capability, digital technology provides technical support for enterprise technological innovation and professional division of labor, and effectively improves the data processing and application capabilities. On the one hand, in the process of production and operation, digital transformation will informationize and digitize the business data of enterprises and generate structured and standardized digital information (Wu et al. 2021) [28]. On the other hand, the application of digital technology has produced a certain information spillover effect, which has strengthened the communication between enterprises and the external market. The application of digital tools such as blockchain and big data will enable market players to reduce the loss in the process of information transmission and reduce the degree of information asymmetry in information production, thus improving the speed and quality of information processing (Zhang Jiruo, 2017) [29]. Many empirical results have also confirmed that digital transformation of enterprises can improve the quality of information disclosure (Wang He, 2023; LU and Li, 2023) [30,31]. In addition, digital technology optimizes the traceability and safety of information. The traceability of blockchain technology can help to record and track the process of information disclosure, and it can effectively audit and supervise information. The application of encryption technology ensures the security of data logical link and improves the authenticity and integrity of information disclosure (Chen et al. 2024) [32].

From the perspective of enterprise behavior, the application of digital technology drives enterprises to transform into service-oriented services, and at the same time, it further deepens their service consciousness (Li et al., 2018) [33]. Enterprises are willing to take on environmental responsibilities while utilizing digital technology to achieve their own economic benefits and continuously improve their ESG information management level in the production and operation management process, thereby providing better information services for stakeholders.

From the perspective of the motivation behind corporate information disclosure, high-quality ESG information disclosure can help companies establish a good green image, and digital transformation provides an opportunity for this. Specifically, in the process of enterprise digital transformation, as enterprise information continues to be standardized and structured, disclosing effective carbon information will help establish its excellent image, alleviate public opinion pressure from the media and other departments, reduce information risks for investors, enhance investor confidence, and thus, increase the likelihood of obtaining investment and support from stakeholders (Li et al. 2018) [33]. Based on the above analysis, we put forward the following research hypothesis:

H1: 

Under other unchanged conditions, digital transformation can significantly suppress the greenwashing of enterprise ESG information.

2.2. Digital Transformation, Investor Attention, and ESG Greenwashing

Kahneman and Tversky (1972) defined investor attention as the reaction of investment participants to captured market information through their own perception mechanisms and feedback systems [34]. There is a potential connection between investor attention and digital transformation. Digital transformation can drive organizational management change and model innovation (Klöckner, 2022) [35], empowering enterprises to build core competitive advantages and create long-term value (Zhong et al. 2021) [36]. Therefore, investors have widely realized the necessity of digital transformation in the process of enterprise development. This study conducted two aspects of analysis in this section. One is the attractiveness of digital transformation itself to investors, and the other is the analysis of investor behavior from the perspective of information recognition.

From the perspective of signal transmission theory, digital transformation will bring a “spotlight” effect (Chen et al. 2021) [37]. Enterprises undergoing digital transformation can convey their different images to the market, thus making them “stand out” and become the focus of the market. Therefore, enterprises undergoing digital transformation will receive sustained attention from all walks of life and form the effect of joint supervision (Song et al. 2019) [38]. With the increase of attention, the improper behavior of enterprises in their daily operations will be continuously supervised. Then, under the supervision of these external forces, enterprises will be under greater pressure, the cost of managers’ violations will become higher, and the management will be more cautious in disclosing enterprise information to the society, thus inhibiting the “greenwashing” behavior of enterprises.

On the other hand, as far as information communication channels are concerned, digital transformation can enhance the interaction between enterprises and stakeholders by improving the transparency, accessibility, and openness of information (Li and Tian, 2024) [39]. The empirical results of Wang and Chen (2017) show that investors’ attention can alleviate the information asymmetry between enterprises and investors to a certain extent and it is also conducive to the flow of enterprise information among investors, which invisibly brings public pressure to enterprises, forcing enterprises to actively improve the level of information disclosure [40]. From the point of view of information mining, digital technology can help enterprises to empower the production, transmission, and management of internal information, strengthen the ability of enterprises to mine and integrate information, transform non-standardized information into standardized information, and make information more easily recognized by the outside world (Wu et al. 2021) [28]. Therefore, this lower information acquisition cost is also able to easily attract investors’ continuous attention. Investors’ attention is helpful in directly improving the degree and quality of information disclosure of enterprises and increase the cost of information disclosure manipulation (Raimo et al. 2021) [41]. Therefore, investors’ attention, as an informal system, governs the ESG greenwashing behavior of enterprises.

Based on the above analysis, this study infers that digital transformation improves corporate greenwashing behavior by increasing investor attention.

H2: 

Investor attention mediates the impact of digital transformation on ESG’s greenwashing.

2.3. Digital Transformation, Executives’ Environmental Cognition, and ESG Greenwashing

The strategic cognition theory emphasizes that it is the subjective cognition of managers rather than the objective environment that directly affects the behavior and decision-making of enterprises. Executive cognition is the leading variable of enterprise behavior, which can promote enterprise decision-making and strategic implementation (Sharma and Pablo, 1999; Florida and Davison, 2001) [42,43]. Executive environmental cognition is based on their own knowledge structure and values (Sabbir and Taufique, 2023) [44]. Specifically, it refers to the perception and scientific knowledge of resources and the environment formed by enterprise executives through their understanding of resources and the environment, as well as the psychological experience when they undertake the obligation of saving resources and protecting the environment, including the awareness of green competitive advantage, social responsibility, external pressure perception, and other factors. For executives with high environmental awareness, enterprises can fully fulfill their environmental responsibilities, improve their environmental management system, meet the requirements of the market and supervision, and they can also actively reflect on the shortcomings of their own green development (Grossman and Helpman, 2014) [45]. Additionally, based on the upper echelons theory, the traits of managers influence their strategic choices and corporate behavior. Management’s understanding of the environment may serve as a boundary condition that influences corporate behavior and performance. In terms of ESG practices, the stronger the green cognition of executives, the more likely they are to integrate ESG principles into the company’s long-term strategy and daily operations. Accordingly, as an internal contextual characteristic, the degree of executives’ green cognition becomes a boundary condition in the study of the relationship between corporate digitalization and ESG greenwashing.

At the level of corporate green governance, many scholars have verified the positive influence of executive environmental cognition on corporate green development. Zou (2019) pointed out that executive environmental awareness can lead a company’s green behavior and improve green performance [46]. Managers with a high level of green awareness will increase their investment in green innovation (Jiang et al. 2022) [47]. Secondly, the green cognition of executives can serve as a situational variable to explore the boundary mechanism of related relationships: executive environmental cognition significantly positively moderates the relationship between environmental investment and the sustainable development of enterprises (Zhou and Jin, 2023) [48]. In addition, when executives have a high level of green awareness, they have the ability to actively perceive and interpret green development opportunities in the market, pay more attention to environmental issues, and comply with environmental ethics and regulations (Xiao et al. 2024) [49]. From the above inference, it can be seen that there will be significant differences in the internal governance methods of executives with different levels of environmental awareness. Information disclosure is an important part of company governance. The environmental awareness level of executives may have an impact on the quality of ESG information disclosure.

The disclosure of ESG information can be seen as feedback from a team of executives led by the CEO based on their own experience and cognition under environmental regulatory constraints. Therefore, compared to executives with lower levels of environmental awareness, executives with higher levels of awareness attach greater importance to corporate environmental protection, possess awareness of sustainable development strategies and long-term planning, pursue higher environmental performance and quality of environmental information, and are more likely to utilize the information spillover effect of digital transformation to improve the quality of environmental information disclosure and curb greenwashing.

H3: 

Executive environmental cognition has strengthened the inhibition of digital transformation on ESG’s greenwashing.

Overall, the schematic diagram of the research model proposed is presented in Figure 1:

3. Research Design

3.1. Sample Selection and Data Source

In 2012, the Chinese Internet industry began to enter the era of mobile terminals, and the digital economy gradually started. Therefore, this paper selects A-share listed companies in Shanghai and Shenzhen stock markets as samples from 2012 to 2022. To avoid interference from other factors, this study follows existing research by excluding financial industry samples, specially treated stock samples or companies with delisting risk warnings, and samples with missing or abnormal data, resulting in 7489 observations. Additionally, to eliminate the influence of outliers, continuous variables are truncated at 1%. The data for digital transformation are sourced from corporate annual reports, the ESG greenwashing degree is obtained from Bloomberg and Huazheng databases, and investor attention is measured using the Baidu Index website. Other variables are primarily sourced from the China Stock Market & Accounting Research (CSMAR) database.

3.2. Variable Definition

3.2.1. ESG Greenwashing

In essence, “greenwashing” is a manifestation that language transcends actions. Therefore, following Zhang’s (2022) [50] measurement method for “ESG greenwashing”, a relative greenwashing index within the same industry was established to measure the degree of greenwashing. The specific method are as follows: the company’s “greenwashing” index is measured by the difference between its ESG disclosure score and the actual ESG performance score. Bloomberg ESG score is the ESG data disclosed by the company to the public. As a disclosure score, the score ranges from 0 to 100, and the higher the value, the better. The Huazheng ESG score, as a true performance score, also ranges from 0 to 100, with higher scores indicating better performance. Different from Bloomberg’s ESG rating, the Huazheng ESG rating, as a performance rating, has the characteristics of being close to the China market, with wide coverage and strong timeliness. It also incorporates indicators that are suitable for the current development stage, such as information disclosure quality and violations, which are closer to the true performance indicators of the enterprise.

In order to make the comparison between disclosure scores and performance scores meaningful, a “greenwashing” index is established as shown in formula (1): subtract the annual industry average from each score and divide it by the standard deviation to standardize the measurement. The difference between the two parts is the greenwashing part, and a larger positive value indicates a more severe greenwashing; a negative value means that the company reports its environmental effectiveness in a low-key manner.

Specifically, ESGdis i, t represents the ESG information disclosure situation of company i in year t. The larger this indicator, the more ESG-related information that the company discloses to the outside world that is beneficial to the company; ESGperi, t is the ESG performance of enterprise i in year t, which measures the true ESG performance of the enterprise. If the enterprise makes more real contributions in environmental, social, and corporate governance aspects, the larger this indicator is. In addition, $\overline{ES{G}_{{\mathrm{dis} }_{\mathrm{i},\mathrm{t}}}}$ and $\overline{ES{G}_{{\mathrm{per} }_{\mathrm{i},\mathrm{t}}}}$ are the average ESG information disclosure and actual ESG performance of enterprise i’s industry in year t, respectively. σ dis and σ per are the standard deviations of ESG information disclosure and actual ESG performance for the industry of company i in year t.

$$G{\mathrm{reenwashing} }_{\mathrm{i},\mathrm{t}}=({\displaystyle \frac{ES{G}_{{\mathrm{dis}}_{\mathrm{i},\mathrm{t}}}-\overline{ES{G}_{{\mathrm{dis}}_{i,t}}}}{\sigma \mathrm{dis}}})-({\displaystyle \frac{ES{G}_{{\mathrm{per}}_{\mathrm{i},\mathrm{t}}}-\overline{ES{G}_{{\mathrm{per}}_{ \mathrm{i}, \mathrm{t}}}}}{\sigma \mathrm{per}}})$$

(1)

3.2.2. Digital Transformation

At present, research mainly adopts the index method and the text data mining method to measure the degree of digital transformation of enterprises, because the former is an indirect measurement method, and its effectiveness is still doubtful; therefore, this paper uses the measurement method of text data mining.

Some studies have started utilizing text data mining methods to measure the extent of digital transformation in enterprises. (Zhuo and Chen 2023; Fang and Liu 2024) [51,52]. This article draws on Yuan’s measurement method for enterprise transformation and classification criteria [53]. The specific measurement steps are as follows:

The first step involves creating a digital terminology dictionary for enterprises. This article constructs the dictionary based on the national policy semantic system. By searching the websites of government, industry, and information technology departments, 30 significant national-level digital economy-related policy documents released from 2012 were manually selected to extract keywords related to enterprise digitalization. Following Python word segmentation and manual recognition, 197 enterprise digitalization-related terms with a frequency of five or more occurrences were identified, forming the digital terminology dictionary used in this study.

The second step is to conduct text analysis on relevant sections of the annual reports. This article incorporates the 197 terms from the digital terminology dictionary into the “jieba” Chinese word library of the Python software package. Using machine learning methods, text analysis is performed on the “Management Discussion and Analysis” (MD&A) section of the annual reports of listed companies. The frequency of the 197 enterprise digitalization-related terms (see

Table 1. Construction of enterprise digital transformation index and selection of characteristic terms.

Category	Main Characteristic Terms
Artificial Intelligence	Artificial Intelligence (AI), Business Intelligence, Image Understanding, Deep Learning, Investment Decision Support System, Intelligent Data Analysis, Machine Learning, Intelligent Robotics, Semantic Search, Face Recognition, Speech Recognition, Biometric Technology, Natural Language Processing, Identity Verification, AI Technology, Autonomous Driving
Big Data	Big Data, Data Mining, Text Mining, Data Visualization, Heterogeneous Data, Credit Scoring, Augmented Reality, Mixed Reality, Virtual Reality
Cloud Computing	Cloud Computing, Stream Computing, Graph Computing, In-Memory Computing, Multi-Party Secure Computing, Brain-Inspired Computing, Green Computing, Cognitive Computing, Converged Architecture, Billion-Scale Concurrency, EB-Level Storage, Internet of Things, Cyber-Physical System
Application of Digital Technology	Mobile Internet, Industrial Internet, Mobile Connectivity, Internet Healthcare, Electronic Commerce, Mobile Payment, Third-Party Payment, Near Field Communication Payment (NPC Payment), Smart Energy, Business-to-Business (B2B), Business-to-Consumer (B2C), Customer-to-Business (C2B), Online-to-Offline (O2O), Network Connectivity, Smart Wearables, Smart Agriculture, Intelligent Transportation, Smart Healthcare, Smart Customer Service, Smart Home, Blockchain Energy Grid, Smart Investment Advisory, Smart Cultural Tourism, Smart Environmental Protection, Unmanned Retail, Internet Finance, Digital Finance, Fintech, Financial Technology, Quantitative Finance, Open Banking
Blockchain	Blockchain, Digital Currency, Distributed Computing, Privacy-Enhancing Technologies, Smart Financial Contracts

Note: To facilitate reading, this paper classifies digital transformation terms by technological domains and application scenarios. Since the study mainly calculates the frequency of specific terms, this classification standard will not affect data calculation.

) appearing in the annual reports is calculated.

The third step is to construct indicators for the degree of enterprise digitalization. Given the variations in text length in the MD&A sections of the annual reports, this article measures the level of enterprise digitalization (Digital) by dividing the total frequency of the digitalization-related terms by the length of the MD&A section. For convenience, this measure is multiplied by 100. The larger the value, the higher the degree of transformation.

3.2.3. Executive Environmental Cognition

The text analysis method can effectively measure the green cognition of executives and can be used for longitudinal data research. The data needed to measure executives’ cognition come from the annual report of listed companies (Osborne et al. 2001) [54]. Therefore, referring to Xiao et al. (2024) [49], this study selects a series of key words based on three dimensions, green competitive advantage cognition, corporate social responsibility cognition, and external environmental pressure perception and measures the environmental awareness of executives through the frequency of the above words appearing in the annual reports of listed companies in 2012–2022.

3.2.4. Investor Attention

With the advancement of information technology and the Internet, it has become common for individual investors to use search engines to obtain company information. Da et al. (2011) applied web search data to the financial market research field [55]. Subsequently, more scholars have used web search indices to measure attention indices. Internationally, many scholars have used Google search data for theoretical research on investor attention. In China, however, Baidu dominates the search engine market, and most individual investors use Baidu to search for investment-related information. Therefore, the Baidu Index is more representative. The Baidu Index reflects the active search demand of Chinese netizens and is updated daily. Hence, this study adopts the Baidu Index, following the measurement methods of Ying et al. (2017) and Wang et al. (2021), to better represent the real situation of individual investors in the Chinese A-share market [56,57]. Specifically, this paper uses the “stock abbreviation” as the search keyword. We collected daily user attention data for all accessible Chinese listed companies from 2012 to 2022. The total Baidu Index for each calendar year was used. To avoid heteroscedasticity issues, the data were log transformed. This value represents the level of investor attention: the higher the value, the greater the attention.

The details of the variables used in this article are shown in Table 2.

3.2.5. Control Variables

Following the related research of Zhai et al., 2022, and Lu et al., 2023 [58,59], the following indicators were selected as control variables: company size (Size), return on equity (ROE), current ratio (Liquid), financial leverage (LEV), fixed asset ratio (Fixed), firm age (FirmAge), ownership concentration (TOP1), whether audited by the Big Four (Big4), Tobin’s Q ratio (TobinQ), and CEO duality (Dual).

Table 2 provides specific measurements of the relevant control variables.

Table 2. Variable definitions.

Type of Variable	Variable Name	Variable Abbreviation	Variable Measurement	Reference	Data Source
Dependent Variable	ESG Greenwashing	GWS	The result calculated using the formula	Zhang [50]	Bloomberg, Huazheng [60,61]
Independent Variable	Digital Transformation	Digital	The results obtained from Python processing of the company’s annual report	Yuan [53]	Company Annual Report
Mechanism variable	Investor Attention	Lnbaidu	Add 1 to the network search index and take the logarithm	Ying [56], wang [57]	Baidu Index website [62]
Moderating variable	Executive environmental cognition	EEC	Frequency of keywords related to executives’ environmental awareness in annual reports of listed companies	Xiao [49]	Company Annual Report
Control variables	Company Size	Size	The natural logarithm of total assets	Zhai [58], lu [59]	CSMAR database [63]
	Return on Equity	ROE	Net Profit/Average Owner’s Equity Balance
	Current Ratio	Liquid	Current assets/current liabilities
	Financial Leverage	LEV	Total liabilities/total assets
	Fixed Asset Ratio	Fixed	Fixed assets/total assets
	Firm Age	FirmAge	ln(Current year − Year of establishment + 1)
	Ownership Concentration	Top1	Number of shares held by the largest shareholder/total number of shares
	Whether Audited by the Big 4	Big4	The company has been audited by the four major accounting firms as 1, otherwise it is 0
	Tobin’s Q ratio	TobinQ	Stock market value/total assets
	CEO Duality	Dual	The Chairman and General Manager are the same person, with a value of 1; otherwise, it is 0

Note: Big 4 refers to the four world-renowned accounting firms: PricewaterhouseCoopers (PwC), Deloitte (DTT), Klynveld Peat Marwick Goerdeler (KPMG), and Ernst & Young (EY).

Table 2. Variable definitions.

Type of Variable	Variable Name	Variable Abbreviation	Variable Measurement	Reference	Data Source
Dependent Variable	ESG Greenwashing	GWS	The result calculated using the formula	Zhang [50]	Bloomberg, Huazheng [60,61]
Independent Variable	Digital Transformation	Digital	The results obtained from Python processing of the company’s annual report	Yuan [53]	Company Annual Report
Mechanism variable	Investor Attention	Lnbaidu	Add 1 to the network search index and take the logarithm	Ying [56], wang [57]	Baidu Index website [62]
Moderating variable	Executive environmental cognition	EEC	Frequency of keywords related to executives’ environmental awareness in annual reports of listed companies	Xiao [49]	Company Annual Report
Control variables	Company Size	Size	The natural logarithm of total assets	Zhai [58], lu [59]	CSMAR database [63]
	Return on Equity	ROE	Net Profit/Average Owner’s Equity Balance
	Current Ratio	Liquid	Current assets/current liabilities
	Financial Leverage	LEV	Total liabilities/total assets
	Fixed Asset Ratio	Fixed	Fixed assets/total assets
	Firm Age	FirmAge	ln(Current year − Year of establishment + 1)
	Ownership Concentration	Top1	Number of shares held by the largest shareholder/total number of shares
	Whether Audited by the Big 4	Big4	The company has been audited by the four major accounting firms as 1, otherwise it is 0
	Tobin’s Q ratio	TobinQ	Stock market value/total assets
	CEO Duality	Dual	The Chairman and General Manager are the same person, with a value of 1; otherwise, it is 0

Note: Big 4 refers to the four world-renowned accounting firms: PricewaterhouseCoopers (PwC), Deloitte (DTT), Klynveld Peat Marwick Goerdeler (KPMG), and Ernst & Young (EY).

3.3. Model Settings

To examine the impact of digital transformation on corporate ESG greenwashing, we adopt panel data regression and control for industry and year fixed effects. The following regression model is constructed accordingly:

$$GW{S}_{\mathrm{i},\mathrm{t}}={\beta }_0+{\beta }_{1}D{\mathrm{ig}}_{\mathrm{i},\mathrm{t}}+\sum {\mathrm{Controls}}_{\mathrm{i},\mathrm{t}}+I\mathrm{nd}+Y\mathrm{ear}+{\xi }_{\mathrm{i},\mathrm{t}}$$

(2)

Specifically, GWS_i,t represents the level of greenwashing of firm i in year t, Dig_i,t denotes the level of digital transformation of firm i in year t, and EGP_i,t and Lnabidui, t are the moderating and mechanism variables, respectively. Control variables are represented as Controls. Ind denotes industry fixed effects, Year represents year fixed effects; ξ_i,t stands for random disturbances. Here are the moderation and mediation effect models:

$$\begin{array}{l}GW{S}_{\mathrm{i},\mathrm{t}}={\beta }_0+{\beta }_{1}D{\mathrm{ig}}_{\mathrm{i},\mathrm{t}}+{\beta }_{2}EG{P}_{\mathrm{i},\mathrm{t}}+{\beta }_{3}D{\mathrm{ig}}_{\mathrm{i},\mathrm{t}}\ast EG{P}_{\mathrm{i},\mathrm{t}}+\sum {\mathrm{Controls}}_{\mathrm{i},\mathrm{t}}\\ +I\mathrm{nd}+Y\mathrm{ear}+{\xi }_{\mathrm{i},\mathrm{t}}\end{array}$$

(3)

$$L\mathrm{n}B{\mathrm{aidu}}_{\mathrm{i},\mathrm{t}}={\beta }_0+{\beta }_{1}D{\mathrm{ig}}_{\mathrm{i},\mathrm{t}}+\sum {\mathrm{Controls}}_{\mathrm{i},\mathrm{t}}+I\mathrm{nd}+Y\mathrm{ear}+{\xi }_{\mathrm{i},\mathrm{t}}$$

(4)

$$GW{S}_{\mathrm{i},\mathrm{t}}={\beta }_0+{\beta }_{1}D{\mathrm{ig}}_{\mathrm{i},\mathrm{t}}+{\beta }_{2}L{\mathrm{nBaidu}}_{\mathrm{i},\mathrm{t}}+\sum {\mathrm{Controls}}_{\mathrm{i},\mathrm{t}}+I\mathrm{nd}+Y\mathrm{ear}+{\xi }_{\mathrm{i},\mathrm{t}}$$

(5)

The detailed research process and construction of the model are shown in Figure 2.

4. Empirical Test and Result Analysis

4.1. Descriptive Statistic

Table 3. Descriptive statistics.

Variables	(1)	(2)	(3)	(4)	(5)
	N	Mean	SD	Min	Max
GWS	7489	−0.492	1.236	−3.344	2.846
Digital	7489	0.865	0.913	0.0374	4.826
Lnbaidu	7489	13.031	0.731	11.449	15.137
EEC	7489	4.010	4.830	0	21
Size	7489	23.35	1.269	19.630	26.452
Lev	7489	0.480	0.190	0.0349	0.908
ROE	7489	0.0869	0.120	−0.926	0.407
Liquid	7489	1.854	1.843	0.268	25.080
FIXED	7489	0.239	0.176	0.00164	0.719
Dual	7489	0.194	0.395	0	1
Top1	7489	37.633	15.856	8.087	75.779
TobinQ	7489	1.930	1.402	0.802	15.607
FirmAge	7489	2.972	0.310	1.609	3.611
Big4	7489	0.144	0.351	0	1

presents the descriptive statistical results for all variables in our research model. As shown in Table 3, the mean value of GWS is −0.492, with a standard deviation of 1.326, ranging from a minimum of −3.344 to a maximum of 2.846. These findings suggest considerable variability in the degree of ESG “greenwashing” among the sample enterprises in this study, thus implying ample room for improvement in the quality of ESG information disclosure among A-share listed companies in China. On the other hand, the average value of Digital is 0.865, with a standard deviation of 0.913, exceeding the mean, indicating significant differences in the level of digitization among different sample enterprises, which lays the foundation for our subsequent analysis

4.2. Benchmark Regression and Moderation Test

Table 4. Benchmark regression and moderation test results.

Variables	(1)	(2)	(3)
	GWS	GWS	GWS
Digital	−0.068 ***	−0.105 ***	−0.151 ***
	(−4.07)	(−4.76)	(−6.09)
EEC*DIG			−0.019 ***
			(−3.56)
EEC			−0.022 ***
			(−6.59)
Size		0.058 ***	0.067 ***
		(3.56)	(4.13)
ROE		0.311 **	0.346 ***
		(2.51)	(2.80)
Lev		−0.169	−0.156
		(−1.48)	(−1.37)
Liquid		0.020 *	0.020 *
		(1.89)	(1.84)
FIXED		−0.202 *	−0.126
		(−1.95)	(−1.20)
Dual		0.214 ***	0.210 ***
		(5.92)	(5.84)
FirmAge		−0.073	−0.063
		(−1.35)	(−1.16)
TobinQ		0.043 ***	0.035 ***
		(3.71)	(3.01)
Big4		0.584 ***	0.565 ***
		(12.74)	(12.34)
Top1		−0.001	−0.000
		(−0.79)	(−0.49)
Constant	−0.433 ***	−1.723 ***	−1.907 ***
	(−22.01)	(−4.25)	(−4.72)
Ind/year	YES	YES	YES
Observations	7489	7489	7489
R-squared	0.003	0.061	0.068

Note: robust t-statistics in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1(the same below); the interaction items in column (3) are centralized.

reports the empirical results of the benchmark regression and moderating effects. Column (1) displays the regression outcome without including any control variables or fixing industry and time effects, showing a significantly negative coefficient for Digital. Upon adding control variables and fixing annual and industry effects, comparing columns (1) and (2), the absolute value of the Digital coefficient increases and remains significantly negative at the 1% level. This indicates that corporate digital transformation contributes to mitigating ESG greenwashing, thus supporting hypothesis H1.

Based on the above findings, we further examine the moderating role of executives’ environmental awareness in the relationship between digitalization and ESG greenwashing. Column (3) in Table 4 presents the results of introducing executives’ environmental awareness into the moderating effect model. The coefficient of the interaction term between digitalization and the executives’ environmental awareness (−0.019) is significantly negative at the 1% level, and its sign is consistent with the coefficient of the explanatory variable on the explained variable. This suggests that as executives’ environmental awareness increases, the inhibitory effect of digitalization on ESG greenwashing is further strengthened, thereby validating hypothesis H3.

4.3. Endogenous Problem Handling

4.3.1. Instrumental Variables Method

Improving the level of digital transformation can suppress corporate greenwashing behavior, and when corporate greenwashing behavior improves, it may also lead to further digital transformation. There may be a bidirectional causal relationship between the explanatory variable and the dependent variable. Therefore, the instrumental variable method was adopted for testing.

Our research refers to Bartik’s (2009) method of constructing instrumental variables using the Shift-Share method (SSIV). [64]. This method of constructing instrumental variables has been widely used in empirical research related to digitization (Yi Xingjian, 2018; Gilbert et al. 2022) [65,66]. By following the ideas of Goldsmith-Pinkham et al. (2020) and Wenke et al. (2023), the average value of digital transformation of other sample enterprises in the industry to which each sample enterprise belongs is calculated, indicating the initial share of the analysis unit (Share). Then, calculate the average annual growth rate of digital transformation of all sample enterprises, indicating the growth rate of sample population (Shift); Share×Shift is taken as a tool variable for digital transformation [67,68].

In the first stage regression results, the regression coefficient of DIG-IV was significantly positive at the 1% level, confirming the correlation between Digital-IV and Digital. In the “unidentifiable test” in the second stage, the Kleibergen Paaprk LM statistic was significant at the 1% level, rejecting the hypothesis of insufficient identification of instrumental variables. In the second stage of the “weak IV test”, the Kleibergen Paap Wald rk F statistic is 541.55, which is much higher than the 10% level critical value of Stock Yogo, which is 16.38, indicating the absence of weak instrumental variables, indicating that the selected instrumental variables are effective. Finally, according to column (2) of

Table 5. Results of instrumental variables.

Variables	(1)	(2)
	The First Stage	The Second Stage
	Dig	GWS
Digital_IV	2.181 *** (15.72)
Digital		−0.234 *** (−3.03)
Controls	Yes	Yes
Kleibergen-Paap rk LM	195.86 ***	195.86 ***
Kleibergen-Paap Wald rk F	247.00 {16.38}	247.00 {16.38}
Year	Yes	Yes
Industry	Yes	Yes
Observations	5914	5914

Note: *** p < 0.01. The value in the second column of Kleibergen-Paap rk LM statistics {} is the critical value of the 10% level of the Stock-Yogo test; due to the limitation of space, the detailed report of control variables (the same below) is omitted here, and the results are kept on request.

, it can be seen that after excluding endogeneity issues, the regression results of digital transformation are still significantly negative at the 1% level, which further confirms the previous conclusion.

4.3.2. Heckman Two-Stage Model

Due to the fact that many companies in A-shares have not yet disclosed ESG related information and there are a large number of companies whose greenwashing levels cannot be observed, there may be a problem of sample selection bias in our study. In view of this, we refer to Li Sumei’s (2024) approach and construct a dummy variable (Dummy) in the first stage of the Heckman two-stage model to determine whether a company’s level of greenwashing can be observed. The dummy variable (Dummy) assigns a value of 1 to the company’s level of greenwashing that can be observed, while the dummy variable (Dummy) assigns a value of 0 to the company’s level of greenwashing that cannot be observed. We introduce the proportion of companies in the same industry that can observe a level of greenwashing in the same year as the instrumental variable (IV_1) and use the entire A-share sample from 2012 to 2022 for the first stage of probit regression [39].

Table 6. Heckman Two-Stage Model.

Variables	(1) First-Stage	(2) Second-Stage
	Dummy	GWS
Digital		−0.105 ***
		(−4.79)
IV1	0.794 ***
	(11.95)
IMR		0.464 ***
		(4.56)
Controls	YES	YES
Constant	−4.628 ***	−7.013 ***
	(−84.75)	(−5.53)
Ind	Yes	Yes
year	Yes	Yes
Observations	32, 441	7489
R-squared	0.338	0.064

Note: *** p < 0.01.

column (1) reports the regression results of the first stage, followed by the calculation of the Inverse Mills ratio. In the second stage, the IMR calculated from the previous stage is placed in the benchmark model for regression. As shown in column (2) of Table 6, the IMR is positively significant, indicating that there is indeed a problem of self-selection bias in this study. However, after overcoming the sample selection problem, digital transformation is negatively significant at the 1% level, which once again demonstrates the robustness of the previous conclusion.

4.3.3. Entropy Balancing Matching

Considering that the PSM method solely focuses on the propensity score and cannot guarantee the reduction of slight differences in each covariate between the treatment and control groups, this paper refers to the approach of Hainmueller (2012) and further employs entropy balancing matching to mitigate the endogeneity between digital transformation and corporate information disclosure quality. By balancing the covariate distributions between the control group and experimental groups, this method can achieve smaller coefficient deviation and standard error [69]. Drawing on the treatment of digital transformation variables by Wu et al. (2023) [70], this paper defines the experimental and control groups based on the median value of corporate digital transformation each year. Specifically, the group with a digital transformation level higher than the median is defined as the experimental group, with Treatdig taking a value of 1, while the other group serves as the control group with a value of 0 for Treatdig. Then, the control variables are weighted according to Treatdig to make the mean values and distribution characteristics of variables similar between the control and experimental groups, thus promoting variable balance. As shown in

Table 7. Control and experimental groups based on entropy balancing matching.

Variables	Experimental Group			Weighted Precontrol Group			Weighted Control Group
	Mean	Variance	Skewness	Mean	Variance	Skewness	Mean	Variance	Skewness
Size	23.3331	1.4335	0.3320	23.3766	1.7848	0.1983	23.3333	1.4338	0.3319
ROE	0.0981	0.0150	−2.2271	0.0758	0.0134	−1.9256	0.0981	0.0150	−2.2270
Lev	0.4662	0.0317	−0.1249	0.4944	0.0405	−0.1989	0.4662	0.0317	−0.1251
Liquid	1.9301	2.9895	5.1834	1.7771	3.7945	4.6230	1.9298	2.9896	5.1839
FIXED	0.2000	0.0221	0.9917	0.2790	0.0367	0.4763	0.2001	0.0221	0.9921
Dual	0.2465	0.1858	1.1766	0.1413	0.1214	2.0596	0.2464	0.1857	1.1770
FirmAge	2.9932	0.0962	−0.8815	2.9502	0.0953	−1.0829	2.9932	0.0962	−0.8815
TobinQ	2.1257	2.4193	2.7723	1.7351	1.4356	3.3269	2.1254	2.4187	2.7727
Big4	0.1351	0.1169	2.1348	0.1533	0.1298	1.9245	0.1351	0.1169	2.1347
Top1	35.7465	256.2942	0.4131	39.5213	238.8450	0.1572	35.7493	256.3281	0.4128

, the results of entropy balancing matching indicate that the mean, variance, and skewness between the control and experimental groups after entropy balancing matching are largely consistent, thereby avoiding the impact of differences between companies in the control group on the results of this paper.

Table 8. Entropy balancing matching results.

VARIABLES	(1)	(2)	(3)
	Benchmark Regression	Before Entropy Equilibrium	After Entropy Equilibrium
	GWS	GWS	GWS
Digital	−0.105 ***
	(−4.76)
Treatdig		−0.078 **	−0.190 ***
		(−2.42)	(−3.48)
Size	0.058 ***	0.054 ***	0.082 ***
	(3.56)	(3.34)	(3.07)
ROE	0.311 **	0.318 **	0.378
	(2.51)	(2.57)	(1.59)
Lev	−0.169	−0.182	−0.237
	(−1.48)	(−1.59)	(−1.37)
Liquid	0.020 *	0.021 **	0.003
	(1.89)	(1.98)	(0.18)
FIXED	−0.202 *	−0.119	−0.547 ***
	(−1.95)	(−1.16)	(−3.05)
Dual	0.214 ***	0.214 ***	0.187 ***
	(5.92)	(5.93)	(2.83)
FirmAge	−0.073	−0.061	−0.015
	(−1.35)	(−1.13)	(−0.20)
TobinQ	0.043 ***	0.043 ***	0.048 ***
	(3.71)	(3.75)	(3.23)
Big4	0.584 ***	0.593 ***	0.589 ***
	(12.74)	(12.93)	(8.67)
Top1	−0.001	−0.001	0.001
	(−0.79)	(−0.55)	(0.97)
Ind/year	YES	YES	YES
Constant	−1.723 ***	−1.722 ***	−2.121 ***
	(−4.25)	(−4.25)	(−3.44)
Observations	7489	7489	7489
R-squared	0.061	0.059	0.077

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

shows the regression result after entropy balance matching. In the result, the coefficient of Treatdig is significantly negative at 1% level, indicating that digital transformation can significantly inhibit enterprise ESG greenwashing, further verifying the robustness of the conclusions of this paper.

4.3.4. Change Model Method

Finally, we employed alternative model estimation methods to address the issue of omitted variables. To mitigate the impact of differences in digital technology application conditions across provinces on the regression results, this study further controlled for the fixed effects of provinces based on the benchmark regression model with fixed industries and years. The results are listed in column (1) of

Table 9. Results of model replacement.

VARIABLES	(1)	(2)	(3)
	GWS	GWS	GWS
Digital	−0.108 ***	−0.113 ***	−0.110 ***
	(−4.86)	(−4.95)	(−4.81)
Controls	YES	YES	YES
Constant	−1.495 ***	−1.489 ***	−1.395 ***
	(−3.55)	(−3.04)	(−2.92)
Ind	YES	NO	YES
Year	YES	NO	NO
Province	YES	YES	NO
Ind×year	NO	YES	NO
Province×year	NO	NO	YES
Observations	7489	7489	7489
R-squared	0.081	0.088	0.104

Note: *** p < 0.01.

, which shows that the digitization level is still at 1%, which has a negative impact on ESG’s “greenwashing”. In addition, in order to control the unobservable differences at the industry level and the provincial level over time, this paper further adopts interactive fixed effects to reduce the endogenous problems in estimation. Specifically, the interaction terms between industry and year, as well as province and year, were included in columns (2) and (3), respectively. The regression coefficients and significance levels are basically consistent with the main regression results, which further verifies the robustness of our conclusion.

4.4. Robustness Test

4.4.1. Replace Variables

To verify the reliability of the previous conclusions, we draw on the research of Wu et al. (2021) [28]. Using the same variable measurement method as text measurement, we reconstructed the digital transformation level index (DCG) of enterprises to replace the core explanatory variables in the benchmark model and tested the robustness. The specific measurement method of variables are as follows: using Python software to grab the target keywords related to digital transformation from the annual report and then statistically analyzing the total number of word frequency comments; finally, find the logarithm of the total word frequency to reduce right skewness. The regression results are shown in column (1) of

Table 10. Robustness test results.

Variables	(1)	(2)	(3)	(4)	(5)
	Replace Independent Variables		Lag Regression		Expand the Range of Tail Reduction
	GWS	GWS	GWS	GWS	GWS
DIG					−0.100 ***
					(−4.33)
L2.Digital				−0.170 ***
				(−5.71)
L.Digital			−0.147 ***
			(−5.50)
DigCompre		−0.126 ***
		(−7.65)
DCG_w	−0.032 **
	(−2.40)
Controls	YES	YES	YES	YES	YES
Constant	−1.793 ***	−1.603 ***	−2.629 ***	−3.560 ***	−1.673 ***
	(−4.42)	(−3.97)	(−5.75)	(−7.00)	(−4.58)
Ind/year	Yes	Yes	Yes	Yes	Yes
Observations	7489	7489	5934	4930	7489
R-squared	0.059	0.066	0.071	0.077	0.059

Note: *** p < 0.01, ** p < 0.05.

, and the regression coefficient of DCG is −0.032, which still has significant negative correlation at the level of 5%. From this, it can be seen that digital transformation can indeed suppress corporate greenwashing behavior.

To reduce the subjective influence of text mining methods, this article incorporates official second-hand data on digital transformation obtained from the comprehensive evaluation method in the CSMAR-listed company database. The digital transformation degree is calculated as follows: 0.3472 * strategic leadership score + 0.162 * technology-driven score + 0.0969 * organizational empowerment score + 0.0342 * environmental support score + 0.2713 * digital achievement score + 0.0884 * digital application score. The empirical results, as shown in column (2), demonstrate that the coefficient of digital transformation (Digcompre) remains negative at the 1% significance level. This finding aligns with previous conclusions, indicating a high degree of robustness.

4.4.2. Lag Regression

The lag period takes into account the time relationship between explanatory variables and explained variables, which can weaken the bidirectional relationship between them. Therefore, the digital transformations with a lag of one or two cycles are selected for regression. From column (3) and column (4) of Table 10, it can be seen that the regression coefficients of one and two stages of digital transformation to enterprise greening are −0.147 and −0.170, respectively, both of which are significantly negative at the level of 1%. This shows that after considering the endogenous problem the previous research conclusions have not been affected.

4.4.3. Expand the Range of Tail Reduction

To mitigate the impact of outliers within a broader range on the findings of this study, we further expanded the range by applying bilateral 5% winsorization to continuous variables using the winsor2 command in Stata17. As indicated in column (5) of Table 10, the regression coefficient for digital transformation is −0.100, remaining significantly negative at the 1% level. This suggests that enhancing enterprises’ digital transformation levels will incentivize them to reduce greenwashing practices, thus affirming the robustness of our conclusion.

4.4.4. Changing Sample Range

As a special region in China, municipalities directly under the Central Government have certain economic and political particularities (Wu Changqi et al., 2022) [71]. Specifically, because the municipality directly under the Central Government has great economic independence and enjoys the inclination of national policies, it will be affected by the special institutional environment to a certain extent. Therefore, this paper deletes the samples of enterprises in Beijing, Shanghai, Tianjin, and Chongqing for regression, and column (2) in

Table 11. Changing sample range.

Variables	(1)	(2)	(3)
	Excluding COVID-19 Pandemic Years	Excluding Directly Administered Municipalities	Including Years After 2015
	GWS	GWS	GWS
Digital	−0.101 ***	−0.069 ***	−0.124 ***
	(−3.54)	(−2.76)	(−5.22)
Controls	YES	YES	YES
Constant	−0.996 **	−1.257 ***	−2.961 ***
	(−2.10)	(−2.83)	(−6.23)
Ind	Yes	Yes	Yes
year	Yes	Yes	Yes
Observations	5592	5716	5742
R-squared	0.053	0.048	0.071

Note: *** p < 0.01, ** p < 0.05.

is the regression result. In the results, the coefficient of Digital is significantly negative at the level of 1%, which is consistent with the previous conclusion.

In addition, considering the impact of the COVID-19 epidemic, this paper draws lessons from the practice of Liu et al.(2024) and excludes the samples of enterprises in 2020 and 2021 for regression. [72]. Column (3) is the regression result. In the results, the coefficient of Digital is significantly negative at the level of 1%, which indicates that digital transformation can still significantly inhibit ESG greenwashing.

To further examine the reliability of the main regression results, this paper continues to conduct a robustness analysis by reducing the research sample. In 2015, the China Municipal Government put forward the “Internet plus” initiative for the first time in its work report, which marked the shift of China’s big data development from conceptual discussions to substantive implementation in the big data industry. Therefore, 2015 serves as the inaugural year for China’s introduction of big data research planning. In view of this background, this paper adopts the sample of alternative research from 2015 to 2022 for regression analysis. As shown in column (3), the coefficient of digital transformation is −0.124, significant at the 1% level, passing the robustness test.

4.5. Further Analysis

4.5.1. Mechanism Verification

This paper empirically tests the intermediary role of investors’ attention in the digital transformation of enterprises and ESG “greenwashing”. As shown in column (2) of

Table 12. Mechanism test.

VARIABLES	(1)	(2)	(3)
	GWS	lnbaidu	GWS
lnbaidu			−0.104 ***
			(−4.18)
Digital	−0.105 ***	0.073 ***	−0.097 ***
	(−4.76)	(7.41)	(−4.42)
Size	0.058 ***	0.350 ***	0.094 ***
	(3.56)	(47.96)	(5.14)
ROE	0.311 **	−0.268 ***	0.283 **
	(2.51)	(−4.52)	(2.28)
Lev	−0.169	−0.301 ***	−0.200 *
	(−1.48)	(−5.80)	(−1.75)
Liquid	0.020 *	−0.012 ***	0.019 *
	(1.89)	(−2.95)	(1.77)
FIXED	−0.202 *	−0.092 *	−0.211 **
	(−1.95)	(−1.83)	(−2.04)
Dual	0.214 ***	−0.078 ***	0.206 ***
	(5.92)	(−4.56)	(5.70)
FirmAge	−0.073	−0.013	−0.074
	(−1.35)	(−0.53)	(−1.38)
TobinQ	0.043 ***	0.113 ***	0.054 ***
	(3.71)	(16.65)	(4.63)
Big4	0.584 ***	0.112 ***	0.596 ***
	(12.74)	(5.12)	(12.98)
Top1	−0.001	−0.008 ***	−0.002
	(−0.79)	(−17.50)	(−1.60)
Constant	−1.723 ***	5.592 ***	−1.142 ***
	(−4.25)	(30.67)	(−2.66)
Ind/year	YES	YES	YES
Observations	7489	7489	7489
R-squared	0.061	0.446	0.063

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

, the regression coefficient of Digital on investor attention is significantly positive, indicating that corporate digital transformation can enhance investors’ attention to the company.

From the results in column (3), it can be seen that the regression coefficient of investors’ attention to ESG greenwashing is negative and significant at the 1% level. In addition, comparing column (3) with column (1), it can be found that the regression coefficient of Digital is still significantly negative in column (3) and the absolute value of the regression coefficient is slightly lower than that in column (1). This indicates that it plays a partial mediating role.

Furthermore, we supplement the coefficient product to test the mediating effect. In the sobel–goodman test, the Z statistic of investor attention is −3.664 (|z| > 1.96, p < 0.01), which is significant at the level of 1%.

All the above results verify that investors’ attention plays a part of the intermediary role in the above relationship, that is, the conduction path of “digital transformation of enterprises → increasing investors’ attention → restraining greenwashing” is formed, thus verifying hypothesis H2.

4.5.2. Heterogeneity Testing and Analysis

Under the heterogeneous situational factors, the digital transformation of enterprises may have an asymmetric impact on the phenomenon of ESG greenwashing. Therefore, this paper further discusses the differences of three aspects: the nature of enterprise property rights, industry characteristics, and regions.

Table 13. Heterogeneity test results.

VARIABLES	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	SOE	Non-SOE	Heavily Polluting	Non-Heavily Polluting	High Tech	Non-High-Tech	Eastern	Central	Western
	GWS	GWS	GWS	GWS	GWS	GWS	GWS	GWS	GWS
Digital	−0.187 ***	−0.045	0.071	−0.113 ***	−0.097 ***	−0.045	−0.113 ***	−0.103	−0.044
	(−5.79)	(−1.53)	(1.09)	(−4.55)	(−3.64)	(−1.13)	(−4.54)	(−1.12)	(−0.74)
Controls	YES	YES	YES	YES	YES	YES	YES	YES	YES
Constant	−3.633 ***	−0.169	−2.326 ***	−1.201 **	−1.654 ***	−1.996 ***	−2.509 ***	0.350	0.519
	(−6.17)	(−0.28)	(−3.76)	(−2.28)	(−2.97)	(−3.49)	(−4.84)	(0.38)	(0.54)
year	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Ind	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	4055	3434	2789	4700	3939	3550	5101	1083	1305
R-squared	0.081	0.071	0.046	0.081	0.068	0.076	0.086	0.075	0.090

Note: *** p < 0.01, ** p < 0.05.

presents the results of the heterogeneity analysis in this research.

(1) The heterogeneity of enterprise property rights.

There are huge differences between state-owned enterprises and non-state-owned enterprises in resources, goals, and values. State-owned enterprises have more resource endowments and policy inclinations, and it is easier to optimize and upgrade their products, services, processes, and management through digital transformation, and to enhance their economic and non-economic benefits in the process of applying digital technology. Therefore, according to the nature of property rights, the enterprises in the sample are divided into state-owned enterprises and non-state-owned enterprises. When the enterprises are state-owned, the value of the virtual variable SOE is 1, otherwise it is 0. Next, we tested in groups. The results, as shown in column (1) of Table 13, show that the coefficient of Digital is significantly negative at the level of 1%, indicating that the digital transformation of state-owned enterprises has a more obvious effect on improving the greenwashing behavior of enterprises. Column (2) indicates that the digital transformation of non-state-owned enterprises is difficult to effectively suppress the phenomenon of ESG greenwashing.

(2) Heterogeneity of industry attributes of enterprises.

The nature of industrial pollution will affect the green washing behavior of enterprises. According to the Classified Management List of Environmental Verification of Listed Companies issued in 2008, which stipulates 16 heavily polluting industries, we divided them into heavily polluting industries and non-heavily polluting industries. We classify all enterprises into two categories: heavily polluting enterprises and non-heavily polluting enterprises. When an enterprise belongs to a heavily polluting industry (POLLUTE), the dummy variable is assigned a value of 1; otherwise, it is 0. Further group discussions show that, according to the results in columns (3) and (4) of Table 13, only the coefficients of non-heavily polluting enterprises are significantly negative, while the digital transformation of heavily polluting enterprises has no significant impact on ESG greenwashing treatment. The reason for this phenomenon may be that because of its inherent characteristics of high emissions and high pollution, heavily polluting enterprises are seriously affected by greenwashing. Publishing their actual environmental data would undoubtedly result in huge fines and public condemnation. Therefore, in order to protect their own interests, the management of heavily polluting enterprises often has a stronger tendency to wash green and is good at deceiving the outside world by using “whitewashing” data. In view of such obstacles, even if heavily polluting enterprises are equipped with advanced digital technology, it is difficult to curb greenwashing.

In models (5) and (6), we further examine the differences in technological attributes of enterprises. Dummy variables are constructed based on the classification method mentioned in the text, assigning a value of 1 when an enterprise belongs to a high-tech industry and 0 otherwise. The results show that in the high-tech enterprise group, digital transformation has a significant inhibitory effect on greenwashing (with a coefficient of 0.097 and passing the 1% statistical significance test). However, in the non-high-tech enterprise group, digital transformation failed the statistical significance test (t value is only −1.13). This indicates that, compared to non-high-tech enterprises, the effect of digital transformation on optimizing ESG information quality is more pronounced in high-tech enterprises, thus demonstrating a certain degree of differentiation. This paper argues that, on one hand, technological innovation is a crucial orientation for the production and operation of high-tech enterprises. Digital technology innovation and transformation, as the frontier field of the new era, is naturally the focus of attention and investment for high-tech enterprises. On the other hand, digital transformation requires strong innovative foundations, which high-tech enterprises can effectively meet with their advanced technological conditions. They can really embed digital transformation into their organizational structure, decision-making system, and production processes. To sum up, the proactive will and objective basic conditions of high-tech enterprises in digital transformation determine that they are more effective in promoting the digital transformation process, and they are more capable of achieving the inhibition effects on ESG greenwashing. In contrast, the development orientation of non-high-tech enterprises does not emphasize innovation, and their development and decision-making orientations are not sufficiently sensitive to digital transformation. They also do not possess the objective technological foundation for deep digital transformation, making it difficult to truly apply the effects of digital governance to information governance.

(3) Regional heterogeneity.

Furthermore, we divided the sample enterprises into three groups based on their geographical locations, the eastern, central, and western regions, and conducted empirical tests accordingly. According to the results presented in columns (7), (8), and (9) of Table 13, the coefficients of digital transformation in the eastern, central, and western regions are all negative, and only the coefficient in the eastern region is significant. After comparing the coefficients, it is found that the degree of digital transformation in the eastern region has the greatest influence on restraining “greenwashing”, followed by the central and western regions. The reasons for this are due to its geographical advantages: the eastern region boasts a developed economy, a high level of marketization, and inherent advantages in the institutional environment, which lay a solid foundation for digital innovation in enterprises and thus enable them to better suppress greenwashing behaviors.

5. Research Conclusion and Discussion

5.1. Conclusion and Comparison

After the above theoretical analysis and empirical testing, this paper draws the following conclusions.

Firstly, digital transformation is conducive to enhancing the level of ESG disclosure in enterprises, and this conclusion remains valid after a series of endogeneity and robustness tests. This is consistent with the empirical research findings of Lu et al. (2023) [58]. Moreover, the robustness tests in this paper are more rigorous and cautious than similar studies, which further confirms the inhibitory effect of digital transformation on the phenomenon of ESG greenwashing. In addition, we also compared with research samples from outside China. Monteiro (2024) used data from companies across various economic sectors in Portugal and concluded that digital technology can enhance the quality of accounting information [73]. Specifically, the quality of ESG information disclosure is a part of accounting information quality. Therefore, the findings of this study are to some extent consistent with his conclusions. This indicates that the conclusions of this study have a certain degree of universality and, more precisely, highlight the important role of digital technology in improving the quality of ESG information.

Secondly, from the perspective of internal governance, executives’ green cognition has strengthened the relationship between digital transformation and ESG greenwashing. Although the existing literature has not incorporated executives’ environmental cognition as a contextual variable into the research framework of digital transformation and ESG, most of the literature believes that executives’ environmental cognition significantly promotes the green development of enterprises (Zhou and Jin, 2023; Xiao et al. 2024) [48,49]. This is roughly consistent with the deductive logic of this paper and also proves that this paper has further explored the boundary function of environmental cognition of senior executives. Furthermore, the conclusions of this study align with existing theoretical frameworks: this study demonstrates that executives’ green cognition can influence corporate ESG behavior, which fully reflects the core idea of the strategic cognition theory—that corporate behavior is guided by the cognition of its executives. Additionally, this study verifies the basic logic of the upper echelons theory from a green perspective. While previous research mostly examined variables such as executives’ overseas experience, academic background, and financial literacy to validate the relationship between managerial traits and corporate behavior [74,75,76], few studies extended executives’ green traits to corporate behavior. This study fills that gap by validating the upper echelons theory from a green perspective, showing that managers’ green traits influence their environmental strategy choices and corporate environmental behavior. This further expands the applicability of the upper echelons theory.

From an external point of view, digital transformation can inhibit the “greenwashing” behavior of enterprises by increasing the attention of investors. Firstly, digital transformation brings about the spotlight effect, i.e., attracting external attention. This is similar to the conclusion of Chen et al. (2021) that digital transformation can bring about governance effects by attracting external attention [37]. Secondly, the empirical research results of this paper show that with the increase of investors’ attention, the company’s “greenwashing” behavior is inhibited. Raimo et al. (2021) points out that investors’ attention will directly improve the level and quality of company information disclosure, so the empirical results of this paper are consistent with their conclusions [41]. However, these studies have not integrated investors’ attention, digital transformation, and ESG greenwashing into a unified research framework. Distinct from previous research, this paper integrates these three variables into a unified research framework, exploring the mechanism of investors’ attention within it. This further contributes to revealing the “black box” between digital transformation and ESG greenwashing.

Finally, this paper delves into the heterogeneity of the impact of digital transformation on corporate greenwashing, examining various factors such as property rights, industry characteristics, and geographical disparities. The findings reveal that digital transformation is particularly effective in mitigating greenwashing behaviors among state-owned enterprises, non-heavily polluting industries, high-tech sectors, and enterprises situated in the eastern regions of China.

5.2. Marginal Contribution

The marginal contribution of this paper may be reflected in the following aspects: First, in terms of theoretical analysis, digital transformation and corporate greenwashing behavior are included in the same framework for analysis, which expands the research on the non-economic effects of digital transformation and supplements the relevant research on corporate greenwashing. Second, this paper connects the external attention mechanism and the internal governance mechanism in the construction of the model in order to explore the potential mechanism and boundary mechanism of the impact of digital transformation on corporate greenwashing behavior from both internal and external perspectives. Third, it examines the heterogeneity in the impact of digital transformation on corporate greenwashing behavior considering property rights, industry attributes, and regional differences.

5.3. Managerial Implications

First, this paper uses listed companies in China as samples to test the restraining effect of digital transformation on the ESG “greenwashing” phenomenon. Given the relatively underdeveloped ESG market in the Asia–Pacific region, with China as a typical representative, the conclusions of this paper may provide empirical evidence for improving ESG practices and disclosure quality in the region. Therefore, enterprises in the ESG transition phase should accelerate digital transformation and utilize digital technologies to promote high-quality sustainable development, thereby suppressing ESG greenwashing and gaining competitive advantages.

Second, relying solely on the self-awareness of enterprises is not enough to curb greenwashing. The government needs to introduce stronger policies to promote the digital economy, guiding enterprises towards coordinated digital transformation and green development. Simultaneously, more stringent regulatory measures should be implemented to create a favorable institutional environment and support the governance of greenwashing.

Third, based on the results of the heterogeneity analysis, it was found that the impact of digital transformation on greenwashing management is limited for enterprises in the central and western regions, as well as in heavily polluting and non-high-tech industries. The government should prioritize these industries and provide external support. Technical support and training should be enhanced for these enterprises, guiding them to leverage the information governance function of digital technologies to improve their ESG disclosure levels.

5.4. Research Limitations and Future Prospects

Firstly, by selecting samples from domestic listed companies for empirical research, the practical implications of this paper may be more instructive for Asia–Pacific countries and regions. Future research could extend the sample to Western countries to test the adaptability of the results.

Secondly, this paper reveals the mediating role of investor attention in the impact of digital transformation on ESG greenwashing and the moderating role of executive environmental cognition. However, other influencing mechanisms need to be further explored in the future.

Most existing empirical studies on ESG mainly use the ESG data disclosed by companies to measure variables. However, the quality of ESG data varies greatly, which may lead to issues such as errors or inaccuracies in related research results. Therefore, in the subsequent measurement process of ESG performance (or achievement) variables, consideration should be given to the overestimation of data caused by greenwashing, aiming to build a more accurate ESG measurement method to evaluate the company’s true ESG performance.