CEO discretion and enterprise digital transformation

Abstract

Digital transformation is a new development opportunity available to enterprises in the digital economy and represents a comprehensive application of digital technology in production, operation, and management. This study examined the impact of Chief Executive Officer (CEO) discretion on enterprise digital transformation using data from 2007 to 2022 of A-share listed companies in Shanghai and Shenzhen exchanges in China. The results suggest that CEO discretion enhances digital transformation performance, and the degree of CEO self-confidence plays a moderating role. The greater the degree of CEO self-confidence, the stronger the impact on digital transformation. Sex, age, educational background, and experience in finance affect discretionary powers of CEOs and its effect on digital transformation. The mechanism test revealed that corporate financialisation and technological innovation investment can play a mediating role. Additional analysis revealed that the facilitative effect of CEO discretion on digital transformation is stronger for non-state and large firms. This study enriched and expanded the behavioural theory in finance and related research on the factors influencing corporate digital transformation and provided empirical evidence on how firms can follow the digital development trends and accelerate the process of corporate digital transformation.

1. Introduction

Digital transformation is a disruptive upgrading and transformation process that introduces digital technologies such as big data, cloud computing, blockchain, and artificial intelligence into the production and operation processes of enterprises. Digital transformation promotes the integration of digital technology in enterprises, resulting in changes in various business processes and in the presentation of the results of organization [1]. In the context of the digital economy, digital transformation has become a necessary stage of enterprise development and is regarded by many enterprises as an important tool for overcoming difficulties and navigating changes [[2], [3], [4], [5], [6]]. This study explored how digital transformation efforts could be improved from the company management perspective.

Corporate management makes decisions on enterprise investment, participates in organisational efforts, implements changes, supports managers for digital work, enables organisations' familiarity with digital technology. Thus the management's ability is important in the digital transformation process and specific implementations.

Managers' abilities determine the performance of the enterprise. The CEO, as the top manager of the enterprise after the chairman, has the discretionary power to implement the management policies of the enterprise and can make decisions to strategically position corporate investments, financing, human resources, and so on. Previous studies show that the effect of CEO discretion on executive compensation performance sensitivity is positively correlated; the greater the CEO discretion, the stronger the executive compensation performance sensitivity [7]. Furthermore, managerial discretion has an effective impact on the governance mechanism that determines executive compensation [8]. CEO's decisions also affect capital structure investments [9], and such decisions potentially put shareholder capital at risk [10]. The CEO, as the top manager of the firm, assumes the role of ‘chief helmsman’ in managing operations and has a non-negligible influence on corporate investment decisions and efficiency. Simultaneously, CEO discretion reflects the CEO's control over the entire enterprise and can impact strategic choices and business decisions. Similarly, the implementation of digital transformation inevitably involves investment decisions and corporate strategy changes, which need to be supported and implemented by corporate managers. As the ‘chief helmsman’ of corporate managers, the CEO has a particularly important impact on the digital transformation of the organization. The degree of discretionary power in the hands of the CEO and the CEO's understanding of the role and benefits of digital transformation directly determine the success or failure of digital transformation and the performance of the enterprise. Therefore, it is important and valuable to study CEOs' discretionary power for the digital transformation of the enterprise.

Prior research on the factors influencing digital transformation has revealed some important insights. First, research regarding the external factors—which mainly include the finance supply chain [11], financial technology [12], and digital finance [13]—mainly explore the impact on the digital transformation of enterprises in the digital technology era and in the context of big data. Second, research regarding the internal factors—which mainly include factors such as corporate financialisation [14], tax incentives [15], CEO composite function background [16], and chairman R&D background [17]—show the importance of starting from the perspective of corporate finance and management gains and highlight the impact of CEO capabilities on exploration of corporate digital transformation. However, existing studies have only explored the impact of managers (with an R&D background or composite functional background) on corporate digital transformation from the perspective of CEO personal characteristics; that is, there is no research on the impact of corporate digital transformation from the perspective of economic consequences caused by management capabilities.

In summary, to the best of our knowledge, no research has been conducted regarding the impact of CEO discretion on the digital transformation of companies. Therefore, the contributions of this study are as follows: first, it analyses the relationship between CEO discretion and digital transformation; second, it serves as a reference and adds to the existing literature on digital transformation; third, it expands the current research content on corporate governance structure; and finally, it adds to the existing research on management in the corporate governance structure.

This study aimed to explore the role of CEO discretion with regard to influencing the digital transformation efforts of the firm through multivariate regression analyses of panel data. In addition, sex, age, educational qualifications, and financial background of CEOs were considered as these factors affect their personality traits, thereby determining CEO discretion. The moderating role of CEO overconfidence in the impact of CEO discretion on digital transformation efforts was also explored. Furthermore, the factors through which CEO discretion influences digital transformation and the effect of different types of firms on digital transformation efforts were investigated. The study mainly aimed to examine the role of CEOs’ characteristics with regard to influencing the digital transformation of firms.

Specifically, this study used the degree of CEO discretion as a starting point to analyse the impact of corporate CEO discretion on digital transformation. We found that the greater the degree of discretion held by the CEO, the more conducive it was to digital transformation efforts and quality outcomes. Additionally, we discovered that the stronger the level of CEO confidence, the greater the impact of CEO discretion on digital transformation. CEO discretion influences digital transformation through corporate financialisation and technological innovation investment. Further research showed that firm size and ownership have heterogeneous effects on the relationship between CEO discretion and digital transformation.

This paper has the following sections: Section 1 is the introduction, Section 2 presents the literature review, Section 3 contains the theoretical basis and research hypothesis, Section 4 introduces the methodology, Section 5 includes the empirical discussion, Section 6 contains the discussion of the impact mechanism, Section 7 proposes additional analysis, Section 8 presents the conclusions, and Section 9 has the limitations of this research.

2. Literature review

2.1. Digital transformation

The current literature on digital transformation is mainly focused on influencing factors and economic consequences. This study focused on the internal and external factors influencing digital transformation and explored the factors that affect the outcomes of the digital transformation.

From an intra-firm perspective, Zhu et al. argue that executive cognitive complexity and centrality can facilitate digital transformation in firms [18]. Additionally, the strong institutional pressures brought about by digital policies enhance the role of executive cognitive centrality and cognitive complexity in facilitating the digital transformation of firms. The study by Porfirio et al. on leadership characteristics and digital transformation presents the conditions that facilitate higher stages of technological development, especially in relation to certain firm characteristics such as leadership and management [19]. Gao et al. identify the paradoxes that companies encounter when undergoing digital transformation and the role of digital affordances in overcoming these paradoxes [20]. Danneels et al. explain how companies can overcome digital transformation paradoxes using digital affordances by identifying the paradoxes that traditional companies encounter when undergoing data transformation, such as the paradox of flexibility and stability of organisational structures, the paradox of costs and profits, and the paradox of perceptions between executives and employees [21]. Based on this, the three digital affordances that play a vital role in overcoming the digital transformation paradox are digital decentralisation, digital agility, and digital citizenship.

From the perspective of external factors, Wang et al. argue that configuring different prerequisites can result in high or non-high levels of digital maturity [22]. Technological uncertainty has a significant impact on digital transformation. Synergies between environmental uncertainty and resource coordination can jointly contribute to digital transformation. Cheng and Masron studied the impact of economic-policy uncertainty on firms' digital transformation and found that economic-policy uncertainty had a significant positive impact on firms' digital transformation [23]. Market competition is the main channel through which economic-policy uncertainty affects firms’ digital transformation, and the effect of economic-policy uncertainty is more pronounced among small firms, non-state enterprises, and firms with weak corporate governance.

2.2. CEO discretion

The existing literature on economic consequences of managerial discretion can be classified as studies regarding corporate governance, corporate finance, and individual managers.

From a corporate governance perspective, Cortes and Kiss argue that limitations on managerial discretion may arise from managers’ subjective interpretations of various firm conditions. Specifically, although the flat organisational structure of small firms may allow managers to perceive a larger range of available strategic options, higher resource availability may limit this perception [24]. Conversely, higher resource availability may allow managers to perceive a greater ability to implement strategic actions, but this perception may be limited in firms with flatter organisational structures. Haw et al. find that the contribution of firm value assessment is statistically and economically significant when organisational or industry discretion is high: when organisational discretion is high, each unit increase in enterprise value assessment contributes 7.4 % to value generation and 5.6 % when industry discretion is high [25]. Wolfgang et al. propose that semi-structured, formalised, and decentralised sustainability programs provide the most favourable conditions for managers to use their discretion to advocate for innovative sustainability initiatives [26]. Sun et al. argue that too little or too much managerial discretion may be detrimental to organisational adaptability [27].

From the perspective of individual managers, Youssef and Teng find that cultural values moderate the relationship between cultural practices and managerial discretion in three cultural dimensions: individualism, uncertainty tolerance, and power distance [28]. The more a society values individualism, uncertainty tolerance, and power distance, the weaker the effect of its practices on managerial discretion; this is as per the logic of marginal utility. Baixauli-Soler et al. argue that managerial discretion plays a vital role in the effectiveness of pay as a mechanism for consistent CEO performance [29]. While personal discretion (the freedom to aim) has a negative impact, environmental discretion (the freedom to act) increases effectiveness. When both individual and environmental discretions are high, managerial discretion has a positive effect.

From a corporate finance perspective, Windisch suggests that managerial discretion in accruals has declined [30]. However, findings on the ability of accruals to predict future cash flows and earnings and the contemporaneous link between stock returns and accruals, suggest that accruals' informativeness also decreases after the introduction of a stricter enforcement regime. That is, stricter enforcement has adverse consequences in the form of reduced truthfulness of accrual information. Falkinger and Habib suggest that capital risk, the extent to which management decisions place shareholder capital at risk, plays a central role in imposing constraints rather than granting discretion [31]. Perotti and Windisch argue that future information efficiency increases with the extent to which managers exercise discretionary accruals, with discretionary accruals conveying useful information to investors and facilitating price convergence towards their fundamental value [32]. Magerakis argues that the positive relationship between CEO competence and firm cash holdings is weakened by firm-level managerial discretion, that is, managerial discretion may affect the amount of a firm's cash holdings [33]. Myers et al. argue that increased discretion increases the relevance of reported income but does not reduce faithful representation [34]. Chen et al. examine how managerial discretionary ‘chokehold’ affects firm investment efficiency (IE) [35]. CEO compensation and stock ownership can increase IE but not reduce investment risk (IR). CEO duality and redundant resources neither improve IE nor reduce IR. Organisational inertia can reduce IR but not IE. Capital intensity increases IR, while environmental richness reduces IR, but neither improves IE. Chen and Crossland find that in low-discretionary environments, management forecasts are significantly less credible than in highly discretionary environments, and therefore, the response to these forecasts tends to be much smaller [36]. Management discretion particularly influences analysts' reactions when analysts are typically uncertain about how to interpret management's forecasts. Black et al. find that fair value estimates are less comparable when managers have stronger incentives to introduce discretion and more comparable when there is stronger investor oversight [37].

3. Theoretical basis and research hypothesis

3.1. CEO discretion

The economics and finance literature places managerial discretion within the framework of agency theory [38], where managerial discretion is defined as the degree to which managers can deviate from the goal of maximising shareholders’ interests in pursuit of their own interests, that is, the degree of goal freedom. Shen and Cho describe managerial discretion as a combination of goal freedom and behavioural freedom [39]. Their study classifies managerial decision-making into four dimensions, according to the degree of goal freedom and behavioural freedom, arguing that goal freedom and behavioural freedom portray only one aspect of the environment faced by managers and that only a combination of both can impact firm performance outcomes.

According to agency theory, CEOs are ‘self-interested’ managers seeking to maximise their personal interests. In a modern enterprise system, as managers' compensation is related to their current performance, ‘self-interested’ CEOs have an incentive to manipulate their current performance through surplus management and other means to obtain high-performance pay. When managers have less decision-making autonomy, the board of directors can play a supervisory role in restraining the CEO's management behaviour. High discretionary power weakens the checks and balances of CEO decisions and reduces the effectiveness of board oversight, creating room for CEO mismanagement.

Based on the stewardship theory, CEOs are ‘stewards’ who seek to maximise corporate profits and pay more attention to the long-term development of the company. Seeking to create long-term value, CEOs tend to increase industrial investments, such as fixed assets and R&D investments. When the CEO has high discretionary power, the CEO faces less resistance from major shareholders in making industrial investment decisions, which ensures unity in decision-making and improves the efficiency of industrial investment decisions, thus enhancing corporate investment in industrial assets. However, this limits resources, and the resources available to firms for financial investments are reduced.

According to the different assumptions of agency and stewardship theories on the human nature of management, CEOs must play two roles in the business management process: the self-interested ‘agent role’ and the competent ‘steward role’. The CEO's choice from among these two roles is influenced by his or her internal status, managerial power, organisational perception, organisational loyalty, and other internal factors as well as organisational contextual factors such as corporate goals, distribution of power among organisational members, trust, and culture. With changes in the contextual factors of corporate governance and the CEO's own traits, he or she may shift between the roles of ‘agent’ and ‘steward’.

Behavioural finance theory suggests that senior corporate managers' behavioural preferences have a differential impact on corporate (digital) innovation activities. In practice, senior corporate managers' decisions do not fully follow the rational person assumption; managers are often influenced by their subjective preferences regarding decision outcomes. Thus, managers with higher risk appetite are more willing to take innovation risks, tend to increase the number of R&D investments, expand the scale of corporate investments, invest in high-risk projects, and expand R&D investments, even though the CEO's willingness to take more innovation risks does not mean that the desired results can be obtained. CEOs do not necessarily make the best decisions for the company's digitalisation and may require tax incentives. Compared to other CEOs, those with a background in R&D may have stronger interests and inclinations towards digitalisation but also need the power and freedom to make decisions rather than relying solely on interest and subjective desires. Hence, too much discretionary power may also backfire, and the CEO may be able to make a big splash and show his ambition, while negatively impacting the company and its shareholders because of his own decision-making mistakes. Thus, the company cannot exist more voices, resulting in employees responding negatively, and corporate transformation is blocked. Based on this, the following research hypotheses were proposed:

H1

The discretionary management power of corporate CEOs is related to the performance of digital transformation.

H1a

The greater the discretionary power of the CEO, the higher the degree of digital transformation of the enterprise if the CEO is a ‘housekeeper’ manager.

H1b

The greater the discretionary power of the CEO, the lower the degree of digital transformation of the company if the CEO is a ‘self-interested’ manager.

3.2. Moderating role of CEO overconfidence

In uncertain real-life situations, individuals' overconfidence can lead to irrational psychological biases in their cognitive and behavioural decisions [40]. CEO overconfidence refers to CEOs’ psychological bias caused by the overestimation of personal competence and judgmental precision [41]. A CEO is identified as overconfident if the three conditions of overestimation, over-positioning, and over-precision are present in his or her decision-making process [42].

Overconfident CEOs have strong self-control, risk-taking spirit, and high self-confidence: they are doers who dare to think and act. These traits make CEOs more likely to accept difficult technological innovation activities, long cycle times, large investments, risk of failure, and high potential gain [43]. Hence, overconfident CEOs are often able to withstand difficulties and setbacks and quickly adapt to operational pressures and challenges, which is more conducive to CEO discretion. Thus overconfident CEOs with voting power can play a bigger role in the digital transformation of the enterprise and may obtain better results.

Overconfidence can equally cause managers to overestimate their ability to generate returns, leading to more risks in corporate policy [44]. Overconfident CEOs are also likely to be blindly confident, which can influence managers' judgments of the company's future direction, leading to irrational decisions that can affect the outcomes and performance levels of digital transformation efforts. Based on this, the following research hypotheses were proposed.

H2

Other things being equal, CEO overconfidence plays a moderating role in the relationship between CEO discretion and digital transformation.

H2a

Other things being equal, CEO overconfidence positively moderates the relationship between CEO discretion and digital transformation.

H2b

Other things being equal, CEO overconfidence negatively moderates the relationship between CEO discretion and digital transformation.

3.3. CEOs’ personal characteristics

Considering that the variables related to a company's CEO in CEO research may change over time and that digital transformation of a company entails certain risks, CEOs with different personal characteristics will have different perceptions and practices of digital transformation efforts. For this reason, this section selects different characteristics of CEOs to test the effect of CEO discretion in digital transformation efforts.

3.3.1. The sex of a CEO

Psychologists have found that men and women have different attitudes towards business risk. Men tend to take more risks and accept more challenges [45,46]. Women, on the other hand, are more cautious about taking risks. Therefore, we investigate the role of gender in the impact of CEO discretion on digital transformation. We propose the following hypotheses:

H3

A male CEO is more likely than a female CEO to help an organization achieve digital transformation.

3.3.2. Age of a CEO

The age of the managers is different and he has different experience which affects their decision making behaviour. Senior executives have a lower risk appetite [47,48], they are usually more risk averse and place more importance on the quality of disclosure. By age group, we consider CEOs under 40 years old to be inexperienced. CEOs in their forties and fifties are at the peak of their careers: they are energetic and risk-taking. CFOs over 50 are approaching retirement age and are more risk-averse. We therefore expect that middle-aged CEOs will be more inclined to drive the digital transformation of their organisations than other CEOs, and propose the following hypothesis:

H4

Middle-aged CEOs are more likely than any other age group to drive digital transformation in their organization.

3.3.3. Education level of a CEO

A person's level of education reflects his cognitive abilities and skills, and the level of education is associated with the ability to process information and discriminate between various stimuli [49,50]. People with higher levels of education are better able to acquire and process information and are more likely to assume cross-functional roles. Highly educated cadres are more in line with the requirements of enterprise digital transformation efforts. Therefore, we propose H5 as follows:

H5

A CEO with higher education is more likely to embrace digital transformation than other CEOs.

3.3.4. CEO's background in finance

Existing literature documents that finance expert CEOs actively manage financial policies [51]. Managers with financial experience are better able to understand financial information and capital market operations [52], which helps firms in their digital transformation efforts. Therefore, we propose H6 as follows:

H6

A CEO with a background in finance is more likely to actively drive the digital transformation of the organization.

This study utilizes the research model shown in Fig. 1.

Fig. 1.

Study model.

4. Methodology

4.1. Sample selection and data processing

This study selected data of companies listed in Shanghai and Shenzhen A-shares exchanges during 2007–2022 as the initial research sample. The following criteria were used to exclude companies: financial enterprises; the samples of ST, *ST, and delisted companies in the period; and enterprises with an asset–liability ratio greater than 1. This study only retained those samples with no missing data for at least five consecutive years. To reduce the impact of outliers, this study applied 1 % and 99 % tailoring to all micro-level continuous variables. The raw data were obtained from the China Research Data Service platform and the Guotaian Database. The annual reports of related companies are obtained from the official websites of the Shenzhen Stock Exchange and Shanghai Stock Exchange, and other data are crawled from Python software. Data in the screening and collation process is shown in Table 1.

Table 1.

Sample selection

This table presents the procedure of the sample selection among the main variables in the research.

Constructions of managerial discretion data for firm-year	numbers
Observations with handle managerial discretion data for 2007–2022	33,897
-Listed companies in financial industry	571
-Special treatment listed companies	1009
-Leverage not below 1	1299
-Data listed less than five years	1798
Sample with managerial discretion content data	29,220
Constructions of digital transformation data for firm-year	numbers
Observations with data in CSMAR for 2007–2022	43,462
-Data on enterprises not implementing digital transformation	19,208
-Listed companies in financial industry	530
-Special treatment listed companies	1834
-Leverage not below 1	1929
-Data listed less than five years	2645
Sample with digital transformation data	17,316

4.2. Variable setting

• (1)Explained variable

Enterprise Digital Transformation (EDT). For example, Song uses the ‘0 or 1’ dummy variable representing ‘whether or not the enterprise is undergoing digital transformation’ to measure the digital transformation of enterprises [53]. However, this technical approach is not effective in showing the ‘intensity’ of enterprise digital transformation and may cause misestimations of the extent of enterprise digital transformation. In this study, we argue that digital transformation is a major strategy for the high-quality development of enterprises in the digital era, and such characteristic information is more likely to be reflected in enterprises' annual reports and announcements. Therefore, this study uses the word-frequency statistics for ‘enterprise digital transformation’ in the annual reports of listed enterprises to characterise the degree of transformation. All annual reports and important announcements of A-share listed companies on the Shanghai and Shenzhen stock exchanges were crawled using Python software (crawler function), and all text contents were extracted using the Java PDFbox library, which was used as a data pool for subsequent feature word screening.

• (2)Explanatory variables

CEO Discretion (MD). Jensen and Meckling propose that MD is the degree to which managers deviate from maximising shareholders' interests in pursuit of their own interests, that is, the degree of goal freedom [54]. Shen and Cho suggest that the only way to correctly characterise managerial decision-making is as a combination of goal and behavioural freedom [55]. Drawing on Dong and Gou, we averaged the normal standard values of position, compensation, and operational powers to represent the level of discretionary power [56]. Position power is the legal authority the manager enjoys ensuing from his or her position and is expressed as the inverse of the number of members in the manager's team. Compensation power represents the level of compensation of the manager and is calculated as the logarithmic value of the average of all executive team members' salaries. Operating power refers to the degree of freedom managers have to utilise corporate resources and is the ratio of corporate working capital to operating income.

• (3)Moderating variables

CEO overconfidence (OC). Rational CEOs have a sense of risk diversification [57]. They will adjust their stock holdings based on the expected future performance of their own company compared to other companies. Therefore, whether the CEO has reduced his or her stockholdings during his or her tenure can characterise the CEO's degree of overconfidence to some extent. Considering previous studies and data availability, this study adopted the change in CEO stockholdings as a variable to characterise CEO overconfidence. Change in CEO's stockholdings is when there is a decrease in the percentage of the company's stock in the CEO's holdings during the year. If there is such a decrease, the CEO overconfidence is considered to be 0; if not, it is considered to be 1.

• (4)Control variables

To improve the precision of the study, considering that the degree of CEO discretion may be influenced by factors such as firm characteristics and management governance level, the following control variables were set. Firm characteristics were profitability (measured by return on total assets), firm growth (measured by growth rate of operating income), financial leverage (measured by gearing ratio), and firm size (measured by the natural logarithm of total firm assets). The corporate governance characteristics considered were equity concentration, the shareholding ratio of institutional investors, and shareholding of institutional investors. Time and industry dummy variables are also controlled to further reduce endogenous disturbances.

Specifically, this study uses the following control variables: company size (Size), Book_to_Market (MB), audit opinion type (Audit), the growth rate of gross operating income (Growth), debt ratio (Lev), equity concentration (Shrcr), annual stock turnover rate (VOL), the net profit margin on total assets (ROA), return on net assets (ROE), the shareholding ratio of institutional investors (InsInvestor), board size (Board), and other variables. The interpretations of the specific variables are presented in Table 2.

Table 2.

Variable definition.

Variable Names	Definition
Dependent variables
EDT	Use the frequency of occurrence of key words related to digital transformation in the annual report of the i company on CSMAR.Using the word frequency statistics of ‘enterprise digital transformation’ in the annual reports of listed enterprises to characterise the degree of transformation. All annual reports and important announcements of A-share listed companies on the Shanghai and Shenzhen stock exchanges are crawled using the Python software crawler function, and all text contents are extracted using the Java PDFbox library, which is used as a data pool for subsequent feature word screening.
Independent variables
MD	Calculate the mean of the standard values of the normal distribution of managers' position rights, compensation rights, and operational rights, and then take the logarithm of them.
Position Rights	Position power is the legal authority the manager enjoys that the position should have and is expressed as the inverse of the number of members of the manager's team.
Compensation Rights	Compensation power represents the level of compensation of the manager. It results from calculating the average value of all executive team members' salaries and taking the logarithm, representing the compensation power.
Operational Rights	Operating power refers to the degree of freedom managers have at their disposal of corporate resources. It is reflected in the ratio of corporate working capital to operating income.
L2.MD	The degree of managerial discretion of enterprises lagging behind phase II。
L3.MD	The degree of managerial discretion of enterprises lagging behind three phases。
L4.MD	The degree of managerial discretion of enterprises lagging behind four phases.
Moderator variables
OC	The level of CEO confidence is characterized by whether the CEO has reduced his or her holdings of the company's stock during his or her tenure, 0 if he or she has done so and 1 if he or she has not.
Intermediary variables
FIN	The sum of trading financial assets, derivative financial assets, available-for-sale financial assets, held-to-maturity investments, net loans and advances, and investment real estate divided by total assets.
INNOV	The ratio of R & D investment to operating income of the I enterprise in year t.
Control variables
VOL	The stock turnover rate of enterprise I in year t is calculated based on the number of circulating shares
InsInvestor	Share ownership percentage of institutional investors.
Lev	Leverage computed as total liabilities divided by total assets.
Roa	The net profit divided by average total assets.
Size	The natural logarithm of total assets of the firm at the end of the year.
Book_to_Market	Book value per share and market value per share.
Growth	The growth rate of revenue of firm i in the previous year.
Shrcr	The shareholding concentration of enterprise I in year t is the shareholding ratio of the top ten shareholders.
Board	The size (number) of the board of directors of enterprise I in year t is logarithmicized.
Roe	Ratio of net profit to average net worth of enterprises.
Industry	A dummy variable, if company i belongs to the industry, it will be 1, otherwise it will be 0.
Year	A dummy variable, if year t belongs to the year, it will be 1, otherwise it will be 0.

4.3. Research model

A multiple regression model was constructed in five steps to empirically test the impact of CEO discretion on corporate digital transformation in Chinese A-share listed companies in Shanghai and Shenzhen.

In the first step, the relationship between corporate CEO discretion and digital transformation was verified by not introducing control variables and controlling for time and industry dummy variables. The regression model is shown in Equation (1).

$${\text{EDT}}_{\text{it}}={\mathrm{\theta }}_0+{\mathrm{\theta }}_1{\text{MD}}_{\text{it}}+\mathrm{\Sigma }\text{Year}+\mathrm{\Sigma }\text{Ind}+{\mathrm{\varepsilon }}_{\text{it}}$$ (1)

where i in the equation represents the firm, t represents the year, and ε_it denotes the error term.

In the second step, the degree of influence of the control variables on the digital transformation of the firm was verified without introducing CEO discretion or control variables, while controlling for time and industry dummy variables, as shown in Equation (2).

$${\text{EDT}}_{\text{it}}={\mathrm{\lambda }}_0+{\mathrm{\lambda }}_1{\text{Size}}_{\text{it}}+{\mathrm{\lambda }}_2{\text{Age}}_{\text{it}}+{\mathrm{\lambda }}_3{\text{MB}}_{\text{it}}+{\mathrm{\lambda }}_4{\text{Growth}}_{\text{it}}+{\mathrm{\lambda }}_5{\text{Lev}}_{\text{it}}+{\mathrm{\lambda }}_6{\text{Shrcr}}_{\text{it}}+{\mathrm{\lambda }}_7{\text{VOL}}_{\text{it}}+{\mathrm{\lambda }}_8{\text{ROA}}_{\text{it}}+{\mathrm{\lambda }}_9{\text{InsInvestor}}_{\text{it}}+{\mathrm{\lambda }}_{10}{\text{Board}}_{\text{it}}+{\mathrm{\lambda }}_{11}{\text{Dual}}_{\text{it}}+{\mathrm{\lambda }}_{12}\text{Big}{4}_{\text{it}}+\mathrm{\Sigma }\text{Year}+\mathrm{\Sigma }\text{Ind}+{\mathrm{\varepsilon }}_{\text{it}}$$ (2)

In the third step, the relationship between corporate CEO discretion and digital transformation was verified by introducing the control variables and controlling for time and industry: the regression model is presented in Equation (3). Hypothesis H1a is supported if the coefficient of discretionary power a1 is significantly positive, indicating that corporate CEOs mainly exhibit stewardship attributes; if the coefficient a1 is negative, hypothesis H1b is verified, indicating that corporate CEOs mainly exhibit self-interest attributes, and their discretionary power is negatively related to the level of corporate digital transformation.

EDT_it = a₀+a₁MD_it + a₂Size_it + a₃Age_it + a₄MB_it + a₅Growth_it + a₆Lev_it + a₇Shrcr_it + a₈VOL_it + a₉ROA_it + a₁₀InsInvestor_it + a₁₁Board_it + a₁₂Dual_it + a₁₃Big4_it+ΣYear+ΣInd+ɛ_it (3)

In the fourth step, to verify the effect of CEO overconfidence on digital transformation efforts, the CEO overconfidence variable was introduced based on Model (2): the regression model is presented in Equation (4). If the coefficients of CEO overconfidence and digital transformation are significant, this indicates that CEO overconfidence impacts digital transformation efforts.

EDT_it = b₀+b₁OC_it + b₂Size_it + b₃Age_it + b₄MB_it + b₅Growth_it + b₆Lev_it + b₇Shrcr_it + b₈VOL_it + b₉ROA_it + b₁₀InsInvestor_it + b₁₁Board_it + b₁₂Dual_it + b₁₃Big4_it+ΣYear+ΣInd+ɛ_it (4)

In the fifth step, to verify the effect of CEO overconfidence on the relationship between ‘CEO discretion and digital transformation’, the interaction terms of CEO overconfidence and CEO discretion with overconfidence were introduced based on Model (3). The regression model is shown in Equation (5). Hypothesis 2 was supported. CEO overconfidence plays a moderating role in the influence of CEO discretion on digital transformation if coefficient c3 of the interaction term between CEO discretion and overconfidence is significant.

SYNCH_it = c₀+c₁MD_it + c₂OC_it + c₃MD_OC_it + c₄Size_it + c₅Age_it + c₆MB_it + c₇Growth_it + c₈Lev_it + c₉Shrcr_it + c₁₀VOL_it + c₁₁ROA_it + c₁₂Inshold_it + c₁₃Board_it + c₁₄Dual_it + c₁₅Big4_it+ΣYear+ΣInd+ɛ_it (5)

4.4. Steps in panel data analysis

Panel data were selected for analysis in this study, and the analysis steps experienced were as follows. Firstly, the required panel data was collected by manual and other means. Secondly, test whether the panel dataset are pooled data. Thirdly, descriptive statistical analysis was performed on the data by calculating and analysing the significance represented by the mean, maximum, minimum, standard deviation, and median of each data. Fourth, the correlation between each two data was calculated and analysed to provide a preliminary analysis of the correlation between the variables involved in the research model. Fifth, a multiple regression analysis was conducted following the steps of the research model proposed above to determine whether the research hypotheses were valid through the positive and negative signs and significance of the coefficients of the explanatory variables. Sixth, the research model is tested for robustness and endogeneity through a series of methods to determine whether the research hypotheses are still met.

5. Empirical discussion

5.1. Descriptive statistical analysis

Table 3 presents the descriptive statistics for the 5032 firms with 15,408 non-balanced firm-year observations. Table 3 reports the means, standard deviations, and minimum and maximum values of all the main variables in this study. Variables marked with ‘*’ are scaled to 100 to facilitate the presentation of descriptive statistics. The means and medians for each return metric (EDT) are nearly equal, suggesting fairly symmetric return distributions. The degree of digital transformation is approximately 1.146 for each enterprise, with the maximum value for the degree of digital transformation for each enterprise being 1.5. The maximum, minimum, and median values of CEO discretionary power are 3.349, 0.275, and 1.694, respectively, indicating that the distribution of CEO discretionary power among enterprises is relatively low. However, the CEO's discretionary power is relatively evenly distributed, with no extreme values. The standard deviation of the return on total assets is small, with mean = 0.038 and median = 0.035, indicating that the profitability of the companies was similar. The minimum value of asset size for each company is 19.667, and the maximum value is 26.157, indicating that the sample companies have relatively similar asset sizes. Additionally, the variance inflation factor (VIF) test for each variable shows that the VIF values of each variable were well below 10, suggesting no multicollinearity among the variables.

Table 3.

Descriptive statistics.

Variable	N	Mean	P50	Standard deviation	Max	Min
EDT	15,408	1.146	0.693	1.356	5.056	0.000
MD	15,408	1.698	1.694	0.971	3.349	0.275
OC	15,408	0.112	0.000	0.315	1.000	0.000
VOL	15,408	5.467	4.374	3.890	18.389	0.547
InsInvestor	15,408	45.135	47.061	22.823	90.413	0.642
Lev	15,408	0.458	0.459	0.198	0.916	0.069
Roa	15,408	0.038	0.035	0.063	0.227	−0.239
Size	15,408	22.303	22.147	1.261	26.157	19.667
Book_to_Market	15,408	0.614	0.608	0.252	1.158	0.115
Growth	15,408	0.153	0.095	0.391	2.475	−0.567
Roe	15,408	0.067	0.069	0.125	0.388	−0.550
Shrcr	15,408	0.547	0.548	0.147	0.898	0.221
Board	15,408	2.144	2.197	0.199	2.708	1.609

Note: This table exhibits the descriptive statistics for the variables used in regression model. All variables are as defined in Table 2.

5.2. Correlation analysis

According to the correlation matrix in Table 4, the correlation coefficient r shows a significantly high positive (negative) correlation, high and moderate positive (negative) correlation, and negligible positive (negative) correlation between the two variables. Table 4 reports the correlation coefficients of the main variables of this study. The statistical results show a significant positive correlation between digital transformation and CEO discretion (r = 0.509, P < 0.05); furthermore, it shows a significantly high positive correlation. A significant positive correlation occurs between digital transformation degree and firm size (r = 0.394, P < 0.01), indicating a moderate positive correlation. A significant positive correlation is evident between digital transformation and book value ratio (r = 0.494, p < 0.01), reflecting a high positive correlation. A significant negative correlation exists between digital transformation and institutional investor shareholding ratio (r = −0.183, p < 0.01) and a significant positive correlation between digital transformation and annual stock turnover ratio (r = 0.025, p < 0.01), both of which are negligible correlations. A significant negative correlation occurs between digital transformation degree and gearing ratio (r = −0.141, p < 0.01), a significant negative correlation between the digital transformation degree and size of the board of directors (r = −0.117, p < 0.01), and a significant negative correlation between digital transformation and return on net assets (r = −0.018, p < 0.05); these three correlations are also negligible. To some extent, it is suggested that CEO discretion, annual stock turnover rate, and firm size contribute positively to digital transformation advancement. In contrast, gearing ratio, institutional investor shareholding, and board size hinder the extent of digital transformation. Digital transformation is not correlated with equity concentration and total return on assets ratio.

Table 4.

Correlation matrix.

	EDT	MD	OC	InsInvestor	VOL	Lev	Size	Book_to_Market	Growth
EDT	1.000
MD	0.509**	1.000
OC	0.010	−0.013	1.000
InsInvestor	−0.183***	−0.059***	0.000	1.000
VOL	0.025***	0.023***	−0.016**	−0.355***	1.000
Lev	−0.141***	−0.043***	−0.007	0.212***	−0.074***	1.000
Size	0.394***	−0.053***	0.010	0.396***	−0.346***	0.429***	1.000
Book_to_Market	0.494***	−0.047***	−0.001	0.144***	−0.304***	0.416***	0.563***	1.000
Growth	0.004	−0.007	0.005	0.065***	0.024***	0.056***	0.044***	−0.054***	1.000
Board	−0.117***	−0.019**	−0.007	0.233***	−0.090***	0.154***	0.211***	0.142***	0.010
Shrcr	−0.004	−0.022***	0.012	0.603***	−0.333***	0.014*	0.298***	0.106***	0.093***
Roa	0.004	0.009	0.011	0.158***	−0.110***	−0.317***	0.076***	−0.228***	0.248***
Roe	−0.018**	−0.005	0.008	0.201***	−0.116***	−0.130***	0.166***	−0.113***	0.274***

	Board	Age	Top 1	Shrcr	Roa	Dual	Big 4	Roe	Audit
EDT
MD
OC
InsInvestor
VOL
Lev
Size
Book_to_Market
Growth
Board	1.000
Shrcr	0.063***	−0.149***	0.659***	1.000
Roa	0.045***	−0.080***	0.112***	0.187***	1.000
Roe	0.066***	−0.018**	0.132***	0.189***	0.912***	−0.022***	0.075***	1.000

Note: All variables are as defined in Table 2. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01.

5.3. Testing H1

Panel A of Table 5 reports the core test results of the ‘CEO Discretionary–Digital Transformation’ relationship. A progressive regression strategy is used for baseline regression. Column (1) exclusively controls for time- and industry-fixed effects. The regression coefficient for the EDT indicator is 0.0233, exceeding the 5 % statistical significance test. Column (2) controls for time- and industry-fixed effects and tests the control variables' significance when the dependent variable is digital transformation. In Column (3), the set of control variables is included on top of the dependent variable, and the associated regression coefficient increases (0.0275). However, the significance remains unchanged (p < 0.05); thus implying that the higher the CEO's degree of discretion, the more it facilitates the company's digital transformation efforts, and each showing a significant positive correlation. Thus, H1a is supported by empirical evidence, and the results presented in Panel A of Table 5 strongly support H1a, suggesting that managerial discretion has a significant positive correlation with enterprise digital transformation.

Table 5.

Test of H1-H2.

Panel A
Variables	(1)	(2)	(3)	(4)	(5)
	EDT	EDT	EDT	EDT	EDT
MD	0.0233**		0.0275**		0.0299**
	（1.98）		(2.13)		(2.11)
OC				−0.0106**	−0.0213**
				(-2.33)	(-1.87)
MD_OC					0.0123*
					(1.54)
InsInvestor		−0.0188	−0.0192	−0.0186	−0.0188
		（-1.04）	(-1.06)	(-1.03)	(-1.04)
Lev		−0.0234	−0.0235	−0.0255	−0.0255
		（-1.10）	(-1.11)	(-1.12)	(-1.12)
Size		0.0570***	0.0561***	0.0578***	0.0569***
		（2.93）	(2.92)	(2.88)	(2.87)
Book_to_Market		−0.0222*	−0.0225*	−0.0222*	−0.0227*
		（-1.93）	(-1.94)	(-1.92)	(-1.94)
Growth		−0.0102	−0.0104	−0.0106	−0.0109
		（-1.59）	(-1.61)	(-1.62)	(-1.64)
Board		−0.0270	−0.0267	−0.0270	−0.0266
		（-1.45）	(-1.44)	(-1.45)	(-1.44)
VOL		0.0505**	0.0511**	0.0505**	0.0509**
		（2.51）	(2.52)	(2.49)	(2.50)
Shrcr		−0.0175	−0.0165	−0.0176	−0.0165
		（-0.70）	(-0.66)	(-0.70)	(-0.66)
Roa		−0.0045	−0.0040	−0.0135	−0.0126
		（-0.69）	(-0.61)	(-0.90)	(-0.84)
Roe		0.0090	0.0084	0.0090	0.0085
		(0.65)	(0.60)	(0.65)	(0.61)
Industry	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes
Constant	0.0002	−0.1373***	−0.1397***	−0.1398***	−0.1401***
	(0.01)	(-3.48)	(-3.30)	(-3.38)	(-3.31)
Observations	15,408	15,408	15,408	15,408	15,408
R-sq	0.0005	0.0283	0.0291	0.0285	0.0292
Adjusted R-sq	0.0005	0.0218	0.0225	0.0219	0.0225

Panel B
Variables	(1)	(2)	(3)	(4)	(5)
	EDT_2	EDT	EDT	EDT	EDT
MD	0.0212*	0.0345**
	(1.96)	（2.10）
L2.MD			0.0234*
			（1.71）
L3.MD				0.0268*
				(1.81)
L4.MD					0.0159*
					(1.35)
InsInvestor	−0.0390	−0.0380	−0.0366	−0.0209	−0.0239
	(-1.43)	（-1.59）	(-1.30)	(-0.92)	(-0.82)
Lev	−0.0359*	−0.0338	−0.0416	−0.0473	−0.0597
	(-1.93)	（-1.20）	(-1.34)	(-1.28)	(-1.34)
Size	0.1194***	0.0636***	0.0684***	0.0664**	0.0746**
	(5.88)	（2.62）	(2.63)	(2.28)	(2.21)
Book_to_Market	−0.0408**	−0.0176	−0.0321*	−0.0305	−0.0322
	(-2.47)	（-1.50）	(-1.93)	(-1.57)	(-1.40)
Growth/Growth*	0.0174***	−0.0039	−0.0196*	−0.0197*	−0.0227
	(2.65)	（-1.44）	(-1.92)	(-1.66)	(-1.56)
Board	−0.0136	−0.0298	−0.0293	−0.0262	−0.0314
	(-0.75)	（-1.29）	(-1.14)	(-0.90)	(-0.91)
VOL/VOL*	0.0269**	0.0610*	0.0659**	0.0593*	0.0723*
	(2.25)	（1.91）	(2.09)	(1.69)	(1.71)
Shrcr	−0.0191	−0.0068	−0.0104	−0.0272	−0.0286
	(-0.64)	（-0.26）	(-0.30)	(-0.78)	(-0.70)
Roa/Roa*	−0.0482	0.0001	−0.0147	−0.0146	−0.0221
	(-0.42)	（0.12）	(-0.80)	(-0.71)	(-0.88)
Roe/Roe*	0.0183	−0.0005	0.0071	0.0043	0.0067
	(0.75)	（-0.93）	(0.49)	(0.28)	(0.35)
Industry	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes
Constant	−0.2484***	−0.0781	−0.1228**	−0.1745**	−0.1949**
	(-3.37)	(-1.64)	(-2.13)	(-2.09)	(-1.99)
Observations	15,408	15,408	15,408	15,408	15,408
R-sq	0.2375	0.0280	0.0303	0.0308	0.0330
Adjusted R-sq	0.2323	0.0196	0.0209	0.0198	0.0199

Panel C
Model	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Variables	EDT_2	EDT_2	EDT	EDT	EDT	EDT	EDT
MD		0.0288**		0.0384**
		(2.94)		(2.11)
L2.MD					0.0275*
					(1.81)
L3.MD						0.0215*
						(1.84)
L4.MD							0.0196*
							(1.82)
OC	−0.0122*	−0.0130*	−0.0140**	−0.0318**	−0.0334**	−0.0198*	−0.0278**
	(-1.33)	(-1.76)	(-2.39)	(-2.22)	(-2.49)	(-1.74)	(-2.19)
MD_OC		0.0231*		0.0206*	0.0212**	0.0279*	0.0258*
		(1.88)		(1.69)	(2.16)	(1.92)	(1.93)
InsInvestor	−0.0387	−0.0392	−0.0372	−0.0378	−0.0364	−0.0297	−0.0209
	(-1.41)	(-1.44)	(-1.56)	(-1.58)	(-1.30)	(-1.31)	(-0.92)
Lev	−0.0359*	−0.0358*	−0.0334	−0.0339	−0.0416	−0.0317	−0.0472
	(-1.93)	(-1.93)	(-1.19)	(-1.20)	(-1.34)	(-1.18)	(-1.28)
Size	0.1200***	0.1193***	0.0644***	0.0639***	0.0686***	0.0639***	0.0664**
	(5.92)	(5.88)	(2.63)	(2.63)	(2.63)	(2.80)	(2.28)
Book_to_Market	−0.0405**	−0.0406**	−0.0171	−0.0179	−0.0322*	−0.0283**	−0.0305
	(-2.45)	(-2.46)	(-1.47)	(-1.52)	(-1.93)	(-2.03)	(-1.57)
Growth/Growth*	0.0176***	0.0174***	−0.0027	−0.0042	−0.0195*	−0.0153*	−0.0195*
	(2.69)	(2.66)	(-1.15)	(-1.49)	(-1.92)	(-1.93)	(-1.66)
Board	−0.0138	−0.0137	−0.0302	−0.0298	−0.0294	−0.0296	−0.0263
	(-0.76)	(-0.75)	(-1.30)	(-1.29)	(-1.14)	(-1.34)	(-0.90)
VOL/VOL*	0.0265**	0.0271**	0.0607*	0.0606*	0.0652**	0.0581**	0.0586*
	(2.22)	(2.28)	(1.90)	(1.91)	(2.09)	(2.27)	(1.68)
Shrcr	−0.0198	−0.0192	−0.0079	−0.0070	−0.0105	−0.0127	−0.0274
	(-0.66)	(-0.64)	(-0.30)	(-0.27)	(-0.31)	(-0.44)	(-0.78)
Roa/Roa*	−0.0490	−0.0481	−0.0003	0.0001	−0.0143	−0.0163	−0.0141
	(-1.44)	(-1.41)	(-0.93)	(0.12)	(-0.78)	(-0.95)	(-0.69)
Roe/Roe*	0.0187	0.0182	−0.0005	−0.0005	0.0071	0.0043	0.0067
	(0.77)	(0.75)	(-0.79)	(-0.86)	(0.49)	(0.28)	(0.35)
Industry	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant	−0.2482***	−0.2481***	−0.0799*	−0.0794*	−0.1246**	−0.0497	−0.0303
	(-3.37)	(-3.36)	(-1.75)	(-1.68)	(-2.15)	(-1.13)	(-0.56)
Observations	15,408	15,408	15,408	15,408	15,408	15,408	15,408
R-sq	0.2371	0.2376	0.0273	0.0283	0.0306	0.0294	0.0311
Adjusted R-sq	0.2319	0.2323	0.0188	0.0197	0.0209	0.0212	0.0198

Note: All variables are as defined in Table 2. In this table, numbers in parentheses represent t-values based on standard errors clustered by the company. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01. Variables appearing with an * in the variable columns in this table represent the variables in column (2) where that original variable was replaced with a variable with an * for the robustness test.

5.4. Testing H2

To test H2, we add OC as a moderating variable. As shown in Panel A of Table 6, Column (1) controls for time- and industry-fixed effects and tests the significance when the independent variable is CEO overconfidence, showing a 5 % significance; and Column (2) controls for time- and industry-fixed effects, and the regression coefficient of the cross product of CEO discretion (MD) and CEO overconfidence (OC) is 0.0123. It exceeds the 10 % statistical significance test. Further, MD and the cross-product term of OC (MD_OC) has a regression coefficient of 0.0123, exceeding the 10 % statistical significance test. This indicates that the more confident the CEO is, the more significantly it promotes the influence degree of the CEO's discretion use on digital transformation efforts, thus validating H2.

Table 6.

Test of H3-H6:personality traits of the CEOs

Dependent variable is the enterprise digital transformation (EDT) for firm i in year t. Explanatory variable is firm's CEO discretion (MD).

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	Gender = 0	Gender = 1	Age = 0	Age = 1	Age = 2	Degree = 0	Degree = 1	FinBack = 0	FinBack = 1
	EDT	EDT	EDT	EDT	EDT	EDT	EDT	EDT	EDT
MD	0.0302	0.1425*	0.0523	0.0585*	0.0218	0.1397	0.0718**	0.0319	0.1991*
	(1.45)	(1.94)	(1.00)	(1.76)	(0.22)	(1.04)	(2.33)	(1.05)	(1.94)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Industry	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant	−0.3014***	−0.4899***	−0.3773*	−0.2060*	−0.7766***	−0.1940	−0.3553***	−0.4129***	0.2208
	(-3.19)	(-3.42)	(-1.61)	(-1.84)	(-4.65)	(-0.60)	(-3.07)	(-4.98)	(1.36)
N	4576	10,832	2642	9304	3462	2771	12,637	8300	7108
R-sq	0.1654	0.1316	0.1700	0.1690	0.1681	0.1855	0.1720	0.1642	0.1508
Adj. R-sq	0.1620	0.1309	0.1524	0.1489	0.1542	0.1696	0.1604	0.1518	0.1264

Notes: Gender equal to 0 is female; Gender equal to 1 is male. The age refers to the age of the CEO. Age = 0 is the young group (under 40 years old); Age = 1 is the middle Age group (40–50 years old); Age = 2 is the old group (over 50 years old).The degree in this table refers to the degree of the CEO. Degree equal to 0 is below the undergraduate; Degree equal to 1 is undergraduate or above.FinBack = 0 means CEO has no financial background; Finback = 1 means CEO has a financial background.All variables are as defined in Table 2 t statistics in parentheses. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01.

5.5. Robustness tests

Panels B and C of Table 5 are subjected to robustness tests. First, the explanatory variable measure (EDT) is replaced by counting the word frequency of each digital technology application. This is used to calculate the ratio of digital economy technology intangible assets to total assets. After the replacement, the set of control variables is included on top of the dependent variable in Column (1); the associated regression coefficient is 0.0212, maintaining a 10 % significance, and H1 is tested. Second, after replacing the key control variables, the measures of the key control variables stock annual turnover rate, net profit margin on total assets, operating income growth rate, and return on net assets are replaced. Additionally, the regression results in Column (2) of Panel B in Table 5 show that the explanatory variables maintain a 5 % significance, thus supporting H1.

The explanatory variable digital transformation (EDT) measure remains replaced in the robustness test of the CEO overconfidence moderating effect model. After the replacement, as shown in Columns (1)–(4) of Panel C in Table 5, the results indicate that the CEO discretion variable maintains its significance at 5 %, and the coefficient of the cross-multiplication term MD_OC maintains a 10 % significance, thus supporting H2.

5.6. Endogeneity test

To address the endogeneity problem caused by the possible reverse causality between CEO discretion and digital transformation, considering that the digital transformation of enterprises is a comprehensive application of digital technology when enterprises strategically deploy digital transformation or make related decisions (such as making relevant manufacturing equipment or software investment, system application), it takes some time for its driving effect to have an impact; this study adopts the explanatory variables of the lag period to deal with the endogeneity problem. Panel B of Table 5 shows maintaining a 10 % significance by testing the explanatory variables (MD) with a lag of two to four periods. This is because digital transformation requires a specific duration to react and adapt. Simultaneously, as shown in Panel C of Table 5, using the explanatory variable (MD) with a lag of two to four periods reveals that the cross-product term MD_OC maintains 5 % significance, and the regression coefficient is positive. This indicates that CEO overconfidence plays a positive moderating role in the mechanism of CEO discretionary influence on digital transformation, verifying H2.

5.7. Test of H3–H6：CEOs’ personal characteristics

5.7.1. The sex of a CEO

Table 8 shows the impact of male CEOs on digital transformation. The sample is divided into two groups by gender (male gender = 1, female gender = 0). The results are presented in columns (1)–(2) of Table 6. In the female subsample, the MD coefficient is 0.0302, but it is not statistically significant. In the male subsample, the MD coefficient is 0.1425 and the correlation is significantly positive at the 10% level. These results suggest that male CEOs are able to improve digital transformation and thus support H3 compared to female CEOs.

Table 8.

Additional analysis.

Panel A
Model	Sample group of large scale enterprise		Sample group of small and medium-sized enterprises （4）
Variables	（1）EDT	（2）EDT	（3）EDT	（4）EDT
MD	0.0226**	0.0229**	0.0122	0.0134
	(2.85)	(2.59)	(0.87)	(0.94)
OC		−0.0638*		−0.0033
		(-1.96)		(-0.20)
MD_OC		0.0156*		0.0067
		(1.82) 0		(0.39)
InsInvestor	−0.0871	−0.0876	0.0118	0.0119
	(-0.81)	(-0.80)	(0.59)	(0.60)
Lev	0.1831*	0.1860*	0.0558***	0.0557***
	(1.80)	(1.81)	(2.63)	(2.63)
Size	0.0445	0.0439	−0.0575**	−0.0574**
	(0.59)	(0.58)	(-2.11)	(-2.10)
Book_to_Market	−0.0831	−0.0826	0.0354**	0.0354**
	(-0.91)	(-0.90)	(2.12)	(2.12)
Growth	−0.1252**	−0.1254**	−0.0069	−0.0069
	(-2.33)	(-2.34)	(-0.85)	(-0.85)
Board	0.0569	0.0573	0.0015	0.0013
	(0.68)	(0.68)	(0.09)	(0.08)
VOL	−0.0164	−0.0148	−0.0135	−0.0135
	(-0.29)	(-0.27)	(-0.96)	(-0.96)
Shrcr	0.1120	0.1099	−0.0371**	−0.0371**
	(0.90)	(0.88)	(-1.97)	(-1.97)
Roa	0.2802	0.2859	0.0396	0.0397
	(1.37)	(1.39)	(1.34)	(1.34)
Roe	−0.1755*	−0.1789*	0.0072	0.0071
	(-1.69)	(-1.71)	(0.29)	(0.28)
Industry	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
Constant	−0.8715*	−0.8791*	0.5062**	0.5069**
	(-1.66)	(-1.66)	(2.04)	(2.04)
Observations	6871	6871	8537	8537
R-sq	0.1543	0.1552	0.0451	0.0451
Adjusted R-sq	0.0956	0.0950	0.0372	0.0371

Panel B
Model	Sample group of state-owned enterprise		Sample group of non-state-owned enterprise enterprises （4）
Variables	（1）EDT	（2）EDT	（3）EDT	（4）EDT
MD	0.0013	0.0030	0.0445**	0.0421**
	(0.07)	(0.16)	(2.38)	(2.13)
OC		−0.0347		−0.0271*
		(-0.79)		(-1.74)
MD_OC		0.0336		0.0267**
		(1.04)		(2.08)
InsInvestor	−0.0733	−0.0738	−0.0504**	−0.0508**
	(-1.26)	(-1.28)	(-2.07)	(-2.08)
Lev	−0.0468	−0.0466	−0.0504**	−0.0502**
	(-1.52)	(-1.52)	(-2.27)	(-2.27)
Size	0.1554***	0.1557***	0.1918***	0.1920***
	(3.54)	(3.54)	(7.27)	(7.29)
Book_to_Market	−0.0383	−0.0385	−0.0683***	−0.0686***
	(-1.10)	(-1.10)	(-3.16)	(-3.17)
Growth	0.0085	0.0088	0.0219**	0.0220**
	(0.58)	(0.60)	(2.07)	(2.08)
Board	0.0363	0.0362	0.0193	0.0194
	(1.49)	(1.49)	(1.18)	(1.18)
VOL	0.0070	0.0072	0.0367**	0.0366**
	(0.28)	(0.29)	(2.49)	(2.49)
Shrcr	0.0840	0.0845	0.0104	0.0105
	(1.46)	(1.47)	(0.41)	(0.41)
Roa	−0.0087	−0.0084	−0.0642*	−0.0638*
	(-0.18)	(-0.17)	(-1.95)	(-1.94)
Roe	0.0291	0.0281	0.0330	0.0326
	(0.77)	(0.74)	(1.14)	(1.13)
Industry	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
Constant	−0.5181***	−0.5170***	−0.9711***	−0.9750***
	(-3.14)	(3.10)	(-6.27)	(-6.30)
Observations	6703	6703	8705	8705
R-sq	0.1528	0.1531	0.5452	0.5454
Adjusted R-sq	0.1376	0.1375	0.5385	0.5386

Note: All variables are as defined in Table 2. In this table, numbers in parentheses represent t-values based on standard errors clustered by the company. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01.

Note: All variables are as defined in Table 1. In this table, numbers in parentheses represent t-values based on standard errors clustered by the company. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01.

5.7.2. The age of a CEO

Columns (3)–(5) of Table 6 show the relationship between CEO age and digital transformation. The sample is divided into three groups according to age: age = 0 for the young group (CEOs less than 40 years old); age = 1 for the middle-aged group (CEOs between 40 and 50 years old); and age = 2 for the old-aged group (CEOs over 50 years old). The MD coefficients for the younger and older groups were 0.0523 and 0.0218, respectively, but were not significant. The MD coefficient for the middle-aged group was 0.0585, which was significant at the 10 % level. These results suggest that middle-aged CEOs can improve digital transformation and thus support H4 compared to younger and older CEOs.

5.7.3. The educational level of a CEO

Columns (6)–(7) of Table 6 report the results of the test of the impact of digital transformation by CEOs with different levels of education. The sample is divided into two groups according to education level: Degree = 0 for less than a bachelor's degree; and Degree = 1 for a bachelor's degree or above. In the Degree = 0 subsample, the MD coefficient is 0.1397, but is not significant. In the Degree = 1 sample, the MD coefficient is 0.0718, which is significant at the 5 % level, suggesting that highly educated CEOs improve the digital transformation of their organisations and thus support H5.

5.7.4. The financial background of a CEO

We also examine whether and how CEOs with a financial background influence digital transformation, and the results are presented in columns (8)–(9) of Table 6. The sample is divided into two groups based on the financial background of the CEO: FinBack = 0 indicates that the CEO has no financial background, and FinBack = 1 indicates that the CEO has a financial background. In the subsample with FinBack = 0, the MD coefficient is 0.0319, but not significant. In the subsample with FinBack = 1, the MD coefficient is 0.1991, which is significant at the 5% level. These results suggest that CEOs with a financial background improve the digital transformation of firms and support H6.

6. Discussion of influence mechanism

Upon testing H1, it is found that corporate CEO discretion significantly facilitates digital transformation. The main factors influencing digital transformation are financial asset allocation, stability of the executive team, director linkages, financing methods, R&D expenditure, and firm financialisation [[58], [59], [60], [61], [62], [63]]. Li and Rui argue that the stewardship theory applies to China's real enterprises, and CEO discretionary power shows a significant negative relationship with corporate financialisation [64]. Chen et al. state that CEO power positively affects firms' investments in technological innovation. Further, CEO discretion is related to corporate financialisation and technological innovation investment, and these variables are, in turn, significantly correlated with digital transformation [65]. Therefore, this chapter explores the mechanism of the effect of CEO discretion on digital transformation from both corporate financialisation and technological innovation investment perspectives.

6.1. Corporate financialisation

To verify the mediating role of corporate financialisation in the relationship between CEO discretion and digital transformation, the path ‘CEO discretion → corporate financialisation → digital transformation’ is tested. According to Demir, financial asset allocation measures corporate financialisation [66]. The formula is as follows:

Corporate financialisation (Fin) = (trading financial assets + derivative financial assets + available-for-sale financial assets + held-to-maturity investments + net loans and advances granted + investment properties) / total assets (6)

Panel A of Table 7 presents the corporate financialisation test results. CEO discretion's effect on corporate digital transformation (‘MD→EDT’) is first tested; the regression coefficient of CEO discretion is 0.0275, and the t-value is 2.13, which is significant at the 5 % level, indicating that CEO discretion significantly enhances digital transformation. Moreover, CEO discretion's effect on corporate financialisation (‘MD→FIN’) is tested; the regression coefficient of CEO discretion is −0.0274, and the t-value is −2.60, indicating that discretion increases corporate financialisation, which is significant at the 5 % level. Finally, the CEO discretion and corporate financialisation effects on digital transformation is tested. The regression coefficient of CEO discretion is 0.0275 with a t-value of 2.13, and the regression coefficient of corporate financialisation is −0.0255 with a t-value of −2.10, significant at the 5 % level. Based on the principle of the mediating effect test, corporate financialisation mediates the relationship between CEO discretion and digital transformation.

Table 7.

alysis of influence mechanism.

panel A
Model	(1)	(2)	(3)
Variables	EDT	FIN	EDT
MD	0.0275**	−0.0274**	0.0275**
	(2.13)	(-2.60)	(2.13)
FIN			−0.0255**
			(-2.10)
InsInvestor	−0.0190	0.0637***	−0.0190
	(-1.05)	(2.71)	(-1.04)
Lev	−0.0254	−0.0043	−0.0254
	(-1.12)	(-0.22)	(-1.12)
Size	0.0568***	−0.0108	0.0568***
	(2.87)	(-0.42)	(2.88)
Book_to_Market	−0.0225*	0.0124	−0.0225*
	(-1.93)	(0.60)	(-1.94)
Growth	−0.0109	−0.0063	−0.0109
	(-1.64)	(-0.73)	(-1.64)
Board	−0.0266	0.0010	−0.0266
	(-1.44)	(0.20)	(-1.44)
VOL	0.0514**	−0.0246*	0.0514**
	(2.51)	(-1.68)	(2.52)
Shrcr	−0.0165	−0.0517**	−0.0165
	(-0.66)	(-2.11)	(-0.65)
Roa	−0.0125	0.0299	−0.0125
	(-0.83)	(0.93)	(-0.84)
Roe	0.0084	−0.0079	0.0084
	(0.60)	(-0.28)	(0.61)
Industry	Yes	Yes	Yes
Year	Yes	Yes	Yes
Constant	−0.1397***	0.2990	−0.1397***
	(-3.30)	(1.43)	(-3.31)
Observations	15,408	15,408	15,408
R-sq	0.0291	0.0318	0.0291
Adjusted R-sq	0.0225	0.0252	0.0224

panel B
Model	(1)	(2)	(3)
Variables	EDT	INNOV	EDT
MD	0.0275**	0.2201*	0.0236*
	(2.13)	(1.92)	(1.97)
INNOV			0.4842*
			(1.90)
InsInvestor	−0.0190	0.0005	−0.0208
	(-1.05)	(1.52)	(-1.14)
Lev	−0.0254	0.0002	−0.0237
	(-1.12)	(0.51)	(-0.95)
Size	0.0568***	−0.0003	0.0602***
	(2.87)	(-1.11)	(2.78)
Book_to_Market	−0.0225*	0.0004	−0.0245**
	(-1.93)	(0.94)	(-1.99)
Growth	−0.0109	−0.0001	−0.0133*
	(-1.64)	(-0.35)	(-1.90)
Board	−0.0266	−0.0002	−0.0382
	(-1.44)	(-0.73)	(-1.51)
VOL	0.0514**	0.0010	0.0559**
	(2.51)	(0.04)	(2.57)
Shrcr	−0.0165	0.0002	−0.0143
	(-0.66)	(0.56)	(-0.54)
Roa	−0.0125	−0.0004	−0.0073
	(-0.83)	(-0.86)	(-0.47)
Roe	0.0084	0.0010	0.0060
	(0.68)	(0.12)	(0.42)
Industry	Yes	Yes	Yes
Year	Yes	Yes	Yes
Constant	−0.1397***	−0.0235***	−0.1444***
	(-3.30)	(-12.23)	(-3.04)
Observations	15,408	15,408	15,408
R-sq	0.0291	0.0209	0.0280
Adjusted R-sq	0.0225	0.0137	0.0208

Note: All variables are as defined in Table 2. In this table, numbers in parentheses represent t-values based on standard errors clustered by the company. ∗p < 0.1, ∗∗p < 0.05, ∗∗∗p < 0.01.

6.2. Technological innovation investment

The CEO discretion increases the technological innovation investment, favouring digital transformation performance and positively impacting it, thus testing path of ‘CEO discretion → technological innovation investment → digital transformation’. Panel B of Table 7 presents the technological innovation investment test results. The regression coefficient of CEO discretion on digital transformation (‘MD→EDT’) is 0.0275, and the t-value is 2.13, which is significant at the 5 % level, indicating that discretion significantly enhances digital transformation. To test the effect of CEO discretion on technological innovation investment (‘MD→INNOV’), the regression coefficient of CEO discretion is 0.2201, and the t-value is 1.92, which is significant at a 10 % level, indicating that discretion has a significant effect on technological innovation investment. Finally, to test the effect of CEO discretion and technological innovation input on digital transformation, the regression coefficient of CEO discretion is 0.0236, and the t-value is 1.97. Additionally, the regression coefficient of technological innovation input is 0.4842, and the t-value is 1.90. Technological innovation input significantly mediates between CEO discretion and digital transformation, and CEO discretion expands the firm's technological innovation investment, further promoting digital transformation efforts.

7. Additional analysis

Heterogeneity test is used to investigate whether sample groups are indeed different. We use a group regression to test the heterogeneity of the research model. Enterprises are divided into two groups (large enterprises and small and medium-sized enterprises) according to their size. Further, group regression is conducted to see whether there is a significant change in the regression coefficients of the core research variables in each group. Moreover, the enterprises are divided into state-owned and non-state-owned enterprises according to the nature of property rights, and group regression is performed to judge the heterogeneity of the two groups.

7.1. Firm size

The study sample is divided into large, small, and medium enterprises. Panel A in Table 8 shows the regression results for enterprise size. Firm size is measured by total assets, and firms are classified as small and medium-sized when their total assets are less than the median and as large when their total assets are greater than the median. In Panel A of Table 8, when testing the path ‘CEO discretion → digital transformation’, the regression coefficient of CEO discretion for large firms is 0.0226 with a t-value of 2.85, significant at the 5 % level. The regression coefficient of CEO discretion for small and medium-sized firms is 0.0122, and the t-value is 0.87, which is insignificant. The discretionary power of CEOs is more likely to play a significant role in influencing digital transformation in large firms. In Panel A of Table 8, the regression coefficient of MD_OC in large firms in Column (2) is 0.0156. The t-value is 1.82, which is significant at the 10 % level. In contrast, the regression coefficient of MD_OC in small- and medium-sized firms in Column (4) is 0.0067, and the t-value is 0.39, which is insignificant. This indicates that the effect of CEO discretion on digital transformation is less effective in small and medium-sized enterprises, which may be because of the following reasons: large enterprises are large in size, have mature business strategies, strong economic power and obvious resource advantages, and CEOs can effectively implement discretion in such an environment, thus enhancing the effects of digital transformation.

7.2. Nature of property rights

Considering the special nature of enterprises' property rights in China, differences in the relationship between the CEOs of enterprises with different property rights exercising discretionary power and digital transformation may exist. When the state and government have ownership or control over an enterprise, it is classified as a state-owned enterprise; otherwise, it is considered a non-state-owned enterprise. Panel B of Table 8 presents the test results used to distinguish between the nature of enterprise ownership. In the sample group of state-owned enterprises, the regression coefficient of CEO discretion in Column (1) is 0.0013, and the t-value is 0.07, which is insignificant. The regression coefficient of the cross-product term of CEO discretion and digital transformation in Column (2) is 0.0336, the t-value is 1.04, and the result is insignificant. In the sample group of non-SOEs, the regression coefficient of CEO discretion in Column (3) is 0.0445, and the t-value is 2.38, significant at the 5 % level. The regression coefficient of the cross-product term of CEO discretion and digital transformation in Column (4) is 0.0267, and the t-value is 2.08, which is significant at the 5 % level. The effect of CEO discretion on digital transformation is substantially different between SOEs and non-SOEs. However, compared to SOEs, non-SOE CEOs will have more freedom and range of choice when exercising their discretion, thus providing more constructive opinions on the enterprise's digital transformation work. State-owned enterprises have long approval processes, strict and fixed management systems, slower responses to market changes, inadequate application of digital technologies, and the value-creation function of digital transformation is not highlighted. In contrast, non-SOEs react quickly to market changes, actively seize the historical opportunities of the digital economy to promote the high-quality development of enterprises, and deeply implement digital transformation strategies, making non-SOE CEOs more active in exercising their discretionary power to promote the digital transformation of enterprises.

8. Conclusion

This study empirically examines the effects, mechanisms, and heterogeneity tests of CEO discretionary influence on digital transformation based on data from 2007 to 2022 for Chinese-listed A-share entities in Shenzhen and Shanghai. It is found that corporate CEO discretion is significantly associated with the performance of digital transformation efforts. CEO overconfidence significantly moderates the model of discretion and digital transformation; the level of CEO confidence enhances the effect of discretion on digital transformation efforts. Corporate financialisation plays a mediating role between CEO discretion and digital transformation; that is, CEO discretion reduces the degree of corporate financialisation, financialisation slows, and digital transformation efforts are vigorously promoted. The greater the discretionary power of CEOs, the more investment in technological innovation is given to the company, and the better the digital transformation efforts. Considering heterogeneous characteristics, such as enterprise size and the nature of property rights, the effect of CEO discretion on digital transformation is more significant in non-state and large enterprises.

Based on this conclusion, this study makes the following policy recommendations. First, the enterprises' digital transformation is an important element of the digital economy to promote high-quality economic development. Moreover, the government should vigorously promote the development of the digital economy, promote the industrial digitisation process, and encourage enterprises to promote the deep integration of digital technology with organisational structure, management style, and data processing. Second, within the enterprise, more discretionary power can be given to the management. This helps executives apply their autonomous business power to formulate digital transformation strategies and work content for the enterprise according to specific situations. The company can establish a reward and punishment system to link the performance of executives with their salaries to enhance the confidence level of enterprise management to a greater extent. Third, given that the enterprises’ current digital transformation process face more pain points, such as the high cost of transformation and IT application difficulties, the government should improve digital infrastructure construction, increase policy subsidies and technical support, reduce the threshold of enterprise transformation, effectively improve the penetration rate of information technology, and guide enterprises and digital transformation to adapt to each other and play digital economic dividends according to the characteristics of different enterprises. Fourth, for transitioning enterprises, information technology should remain at the surface application level and promote the deep integration of information technology and enterprises. Moreover, it is necessary to implement digital transformation strategies from the strategic level of enterprises so that digital science and technology can reshape the value chain of enterprises. Fifth, it should combine the fulfilment of corporate social responsibility with the digital transformation goal, strengthening the connection and cooperation between enterprises and various subjects in the social network, accumulating social capital in the fulfilment of social responsibility, acquiring key knowledge and resources required for digital transformation, and creating conditions for the digital transformation of enterprises. Sixth, enterprises should improve their internal control system in the process of promoting digital transformation; ensure the rationality and effective implementation of transformation-related decisions regarding resource utilisation, information communication, and risk control; and formulate reasonable reward and punishment mechanisms according to the quality of internal control and degree of disclosure to promote the continuous improvement of the system of internal control of enterprises.

9. Limitations

Despite its theoretical contributions, this study has some limitations. First, the more common way of measuring CEO discretion was used; however, we did not consider factors such as the CEO's personality. Second, in acquiring digital transformation data, two ways to measure digital transformation were used, and there may be better metrics for digital transformation in the future. Further, more classic literature is cited to support the research content in the future to enrich and solidify the study.

Data availability statement

Data will be made available upon request.

Ethics statement

All participants provided informed consent to participate in the study.

Funding statement

This article is supported by the Key Discipline Consturction Project of Business Administration in Xinhua College of Ningxia University.

Additional information

No additional information is available for this paper.

CRediT authorship contribution statement

Yueyun Wang: Writing - review & editing, Writing - original draft, Methodology, Data curation. Zhenhua He: Writing - review & editing, Funding acquisition, Formal analysis, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.