The Impact of Enterprise Digital Transformation on Audit Fees—An Intermediary Role Based on Information Asymmetry

Abstract

This study investigates the impact of enterprise digital transformation through information and communication technology (ICT) on auditing fees. Based on data from publicly listed companies in China and employing information asymmetry theory, the research finds that the adoption of three factors associated with digital transformation—artificial intelligence (AI), cloud computing (CC), and big data technologies (BD)—exhibits a significant inverted U-shaped effect on auditing fees. Further analysis reveals that this effect is moderated by the quality of internal controls, the level of corporate governance, and discretionary accruals. These findings underscore the necessity for a nuanced understanding of the relationship between technology and auditing, as well as the importance for audit organizations to integrate new technologies into their practices to effectively respond to the rapid adoption of digital technologies by enterprises.

1. Introduction

Information and communication technology (ICT) has tremendously influenced different industries [1]. It has profoundly impacted the economy, industries and even social operations with the development of a new generation of ICT, such as artificial intelligence (AI), big data, cloud computing, blockchain, etc. [2]. ICT most directly impacts companies, which can reflect two aspects. One aspect is that new emerging technology can promote many new industries; the other is that the traditional operation mode has been reshaped by ICT. Theoretically, in view of the operational business process as a system, the essence of business operation is to integrate the cognition of the company’s current situation and external information and then output the decision-making. ICT reshapes business operations through digital transformation, reflected by information collection, processing, and decision-making [3]. The most significant impact is that the digital transformation driven by ICT has changed companies’ business models and operations, establishing innovative revenue methods for companies in terms of custom relationships, competition, data, value, etc. [4]. Thus, in the context of emerging ICT in different areas, it is critical to understand the effect of digital transformation on the economy, industry, and business operations from theoretical and practical perspectives.

The impact of digital reform on enterprises has become a prominent topic of interest among scholars. Numerous researchers have attempted to comprehend the effects of the new digital transformation from different perspectives. Some scholars have posited that digitization enhances internationalization strategies and increases the likelihood of successful internationalization for enterprises [5]. Furthermore, some academics have posited that the implementation of digital reforms has the potential to enable newly established firms to disrupt the monopoly of established players in regional industries by altering the means and spatial–temporal boundaries through which corporate value advantages are acquired [6].

Numerous scholars have focused their research on the relationship between digital reform and firm performance and have found a significant positive correlation between the two. Some scholars have studied the impact of digital transformation on companies’ performance and found that digital transformation has greatly improved corporate performance by stimulating and encouraging innovation, reducing costs, increasing revenue, and increasing efficiency [7]. Some research focuses on the impact of digital transformation on principal–agent conflict. Ivaninskiy and Ivashkovskaya (2022) empirically conclude that companies with a higher degree of digital transformation show a low level of principal–agent conflict, which indicates a mitigating effect of digital transformation [8]. In addition, some other studies show that digital transformation has effectively improved corporate social performance, and this impact is enhanced by the increase in the independence of the board of directors and the level of institutional ownership [9]. Previous studies provide a deep understanding of digital transformation as a concept and its role. However, limited research investigates the relationship with auditing. Auditing is essential to reduce the information and expectation gap between management and shareholders, which impacts and changes companies’ business processing and operating models. According to previous discussion of the effect of digital transformation on companies’ business processing, this study will investigate the nature of the relationship between auditing and digital transformation and whether digital transformation can profoundly change auditing.

The financial market is dependent on corporate information disclosure, and the quality of information can significantly affect corporate investment [10]. Thus, it is important to ensure a faithful representation of financial information. The audit can promote better corporate governance by minimizing the information asymmetry between shareholders and managers and managers’ discretionary powers. The capital market has an increasing demand for high-quality auditing services to control information asymmetry and excessive earnings management. In addition, environmental, social, and governance (ESG) and auditing are also a hot topic [11]. Listed companies are required to conduct annual financial audits, so the relationship between auditing services and auditing fees has been a concern [12]. Digital transformation has a profound impact on traditional companies, significantly altering their operations and likely affecting audit functions as well. This influence can lead to two outcomes. On the positive side, ICT-driven digital transformation enhances companies’ information collection and processing capabilities, potentially lowering costs and simplifying the financial audit process, which may reduce auditing fees. Conversely, it may also empower companies to conceal their operational status through fraudulent information, complicating the audit process and increasing auditing fees.

To sum up, research on the relationship between digital transformation and auditing is still limited. It raises two questions: Does digital transformation ultimately impact auditing fees by influencing companies’ operational functions? Moreover, how do these impact factors work, and what is the mechanism of influence? According to the perspective of information asymmetry, this paper explores the relationship between the digital transformation of enterprises and audit fees based on empirical research. This paper establishes a mediated stepwise model for intermediary effects and uses internal control quality, corporate governance level, and discretionary accruals as indicators to measure financial information asymmetry, governance information asymmetry, and earnings information asymmetry.

The contributions of this paper are primarily reflected in several key aspects. Firstly, it emphasizes the changes in auditing practices within the context of digital transformation and explores how this transformation influences auditing fees. Additionally, the paper identifies intermediary variables based on the operating environment and management processes of enterprises, enhancing the applicability and effectiveness of information asymmetry theory in corporate research. This approach provides a new perspective for understanding the impact of digital transformation on auditing fees and the mechanisms involved. Finally, the paper offers insights into digital transformation strategies and the resulting changes in audit practices by analyzing the significant effects of digital transformation on audit fees and their underlying mechanisms. Furthermore, it provides guidance for audit institutions on accurately assessing and planning audit fee levels, which is practically beneficial for advancing the digital evolution of auditing.

2. Literature Review and Hypothesis Development

2.1. Auditing in the Context of Digital Transformation

With the development of traditional audit activities towards new digitization, big data and AI are more widely adopted, among many digital technologies, by public accounting firms to improve their processes and expand service products [13]. Auditors can evaluate almost all audited clients’ data using big data digital tools. Meanwhile, auditors can also obtain the latest data to improve audit quality by analyzing data indicators such as current sales levels and order quantities, reducing opportunistic behavior of managers, and improving corporate governance [14]. Auditors need to provide digital solutions through technological innovation to remain competitive [15]. Digital technologies can help auditors come up with predictive solutions and optimize operational processes by automating processes to increase productivity and efficiency. Zhang et al. (2015) state that big data analytics allow auditors to address data gaps, enhancing the effectiveness of auditors’ judgment and decision-making behavior and assisting auditors with automatic data correction [16]. Thus, digitization in auditing can reduce disruptions, save time and focus on data analysis rather than data collection, which new technologies can automate. As such, digitization is changing the auditing field and will completely change the practice of auditing in the future.

In the context of digital auditing, the advancement and application of big data technology significantly enhance the auditing process. First, as enterprises increasingly rely on digitization, electronic information systems facilitate off-site audits. This transition allows auditors to mitigate risks through modeling and data analysis, reducing the errors associated with traditional manual auditing. Second, traditional audits often face limitations in scope and information access, leading to higher audit risks. In contrast, the digital audit model expands the scope of comparable information, enabling comprehensive data analysis and reducing sampling risks. Additionally, digital audits shift evidence collection from paper-based methods to electronic formats, streamlining data analysis and improving the presentation of audit findings. This transition allows auditors to generate more refined reports efficiently. Ultimately, digital transformation simplifies, standardizes, and enhances the flexibility of the auditing process, moving from on-site sampling to comprehensive data analysis, which improves decision-making effectiveness and efficiency [17,18,19,20].

2.2. Asymmetry Information Theory

Asymmetry information theory is a core theory of information economics, which was first proposed by American economists Akerlof, Spence and Stiglitz [21]. In this theory, each economic subject holds the amount of information differently in the operation of market economic activities, and information owners (usually the sellers) have far more information than information receivers (usually the buyers). The information owners with sufficient high-value information have a relatively favorable dominant advantage, while the information receivers are in a passive disadvantaged position. Information asymmetry exists in various economic activities, and pure information symmetry does not exist [22]. The reasons for information asymmetry can be summarized into two categories. One is that economic entities subjectively have different information acquisition capabilities due to environmental and capital limitations. The other is that with the refinement of the social labor division and the enhancement of professional levels in the industry, the information gap between professionals and non-professionals becomes more asymmetric, and the information distribution is more unbalanced.

The premise of decision-making is information. Based on rational prediction assumptions, people tend to pursue the maximization of their interests under the motivation of self-interest. From the perspective of the information disadvantage, obtaining additional information resources costs more due to information asymmetry. However, information owners usually hide information possessed or actions taken to maintain their information advantage and further increase information gaps. Depending on when the conflict occurs, information asymmetry can be categorized as ex-ante and ex-post information asymmetry, which leads to adverse selection and moral hazard, respectively. The former is reflected in information suppression and interest encroachment before signing a contract. The latter refers to behaviors that damage the client’s interests, such as bad money driving out good money after signing the contract [23].

According to the theory of asymmetric information in accounting and auditing, there is also a severe information asymmetry between enterprises and external information users. External actors usually cannot participate in the operation and management activities of the enterprises and can only understand the operation and financial status via the information disclosed by the enterprises. For external information users, including auditors, enterprises are likely to use the information gap to manipulate and beautify their financial statements or fail to disclose their operating conditions in a timely and accurate manner to achieve the purpose of increasing their stock price and attracting investment or avoiding delisting, etc. These enterprises typically have poor internal control quality [24,25], complex internal organizational structures and a lower level of voluntariness and quality of information disclosure [26]. As such, auditors, as an independent third-party agency, play an intermediary role and need to effectively and efficiently obtain various types of information and provide reasonable assurance of accounting information that can meet the information users’ requirements for information sufficiency, authenticity and reliability [27].

2.3. Quality of Internal Control, Corporate Governance, and Discretionary Accruals

Internal control was first defined by the American Institute of Certified Public Accountants (AICPA) in 1949 and summarized as a process effected by an entity’s oversight body, management, and other personnel that provides reasonable assurance that the objectives of an entity will be achieved. To make the definition of internal control more specific and narrow the scope of responsibility of CPAs, the AICPA made amendments in 1953, dividing internal control into two parts: accounting control and management control, and further standardizing the scope of internal control evaluation and inspection. In the 1990s, COSO was established to investigate internal control issues and solve scandals involving financial fraud, and in September 1992, it released an internal control integration framework. COSO believes that internal control is a long-term process that should be implemented, including five elements: control environment, risk assessment, control activities, information and communication, and monitoring Subsequently, COSO issued a number of guidance documents successively, and released a revised version of the internal control framework in 2013, which put more emphasis on practical and principal orientation based on maintaining the views of the original framework. The non-financial report is included in the internal control evaluation’s scope, and IT to corporate strategy is emphasized. So, it is regarded as an upgraded version of the 1992 framework. The COSO integrated framework has laid a solid foundation for developing internal control theory and practice and has become the industry’s most authoritative and highly recognized theoretical system [28].

Corporate governance arose with the emergence of corporate enterprise organizations, marked by the establishment of the East India Company, and the practice of corporate governance has a history of more than 200 years. In 1776, Adam Smith proposed in The Wealth of Nations that the separation of ownership and management raised corporate governance issues, and a set of practical methods and systems needed to be established to solve this conflict [23]. The concept of early corporate governance originated in the United States in the 1930s, introduced the shareholding structure, explained the conflict between shareholders and professional managers in companies and provided solutions. The decentralization of the ownership structure provides convenient conditions for managers to abuse their control rights by taking advantage. The role of corporate governance is to ease the tension between shareholders and management, making their interests tend to be consistent.

Corporate profit is equal to the sum of operating cash flows and various receivables and payables, the latter of which can be further subdivided into discretionary accruals and non-manipulation accruals. The operating cash flows and non-discretionary accruals are generally difficult to adjust, while discretionary accruals are established by professional judgment and are less difficult to manipulate, the amounts of which are often considered subject to earnings management. The higher the absolute value of discretionary accruals, the more earnings management activities in companies, and the lower the reliability of the earnings management data disclosed in the annual report.

Since the promulgation of the COSO framework, the importance of internal control has been realized, so it has become a means of corporate governance to reduce risks and ensure the faithful representation and completeness of the financial information to maintain the efficiency of production and operation processes. According to the effect of internal control on listed companies’ accounting information, auditors need to change auditing risk decisions based on the quality of internal control [29], which impacts the behavior of auditors and the auditing fees [30]. Secondly, corporate governance is composed of many elements involving all aspects of enterprise operation. The level of corporate governance has a continuous impact on the audit work. Auditors consider the quality of corporate governance when making audit planning decisions [31]. Operating status, including management capabilities, changes in the number of employees, new product development, etc., may affect auditing judgments in addressing the ongoing concern, which eventually is reflected in auditing fees [32,33]. Finally, manipulable accruals can represent the status of corporate earnings management. The studies found that debtors may artificially manipulate corporate financial information through earnings management behaviors in order to reduce the risk of default, which reduces the quality of accounting information [34], increasing the risk of earnings management, and thereby increasing audit time and increasing audit fees [35]. The above three elements are essentially within the enterprise and are affected by the overall changes of the enterprise. The impact of digital transformation embedded in all aspects of the enterprise [36] affects the internal control of the enterprise [37], changes the organizational structure, management model and governance method of traditional corporate governance [38,39], and significantly influences the level of earnings management by changing the levels of information asymmetry [40].

2.4. Hypotheses

2.4.1. The Digital Transformation and Auditing Fees

At present, digital technologies such as big data, cloud computing, blockchain, and artificial intelligence are steadily being integrated into various fields of social development. On the one hand, digital technology has a profound impact on auditing. When digitization is high in audited companies, the audit information and auditing environment tend to be complicated [41]. Accountants and auditors may not be able to properly integrate information technology with their professional judgments [42], causing additional audit risks and increasing the external auditing difficulty, which results in additional auditing procedures and increasing auditing fees. On the other hand, effective digital transformation can improve the efficiency of corporate governance, reducing auditing risk and fees. For companies with good internal control and a high level of informatization, external auditing usually reduces the auditing fees [43], because the improvement of management level heavily depends on digital informatization application [44]. Dorantes et al. (2013) argued that implementing digital information management models has significantly improved information transparency and reduced auditing fees [45]. In addition, companies with a higher degree of digital transformation are more able to show the outside world that their operating conditions are stable and well profitable, thereby releasing a signal of low auditing risk and reducing auditing fees.

Therefore, in the early stage of enterprise digital transformation, the application of digital technology makes the carrier of audit evidence more diversified and complicated, which increases the auditing difficulty and costs. When the degree of digital transformation exceeds a specific critical value, it will not lead to a further increase in audit fees. The negative correlation shows between digital transformation and auditing fees because of the improvement in information transparency and audit efficiency. Therefore, the first hypothesis is

H1. 

The effect of enterprise digital transformation on auditing fees is in an inverted U-shape.

2.4.2. Digital Transformation, Financial Information Asymmetry and Auditing Fees

The application of digital technology has changed the form and timeliness of information generation, enabling enterprises to use digital technology in almost all aspects to obtain unstructured data such as audio, pictures, and videos. It can extract financial information and other information from that, using data analysis to grasp the complete operating status of the enterprise and strengthen the supervision of important information and critical positions through tracking and monitoring methods, and ultimately improve the reliability of information and internal control [46,47]. In addition, enterprises with a higher degree of digital transformation have significantly better internal control than those with a lower degree of digital transformation. Thus, according to the signaling theory, we can predict that such enterprises are more likely to falsely transmit a signal of reasonable internal control to the outside, which exacerbates information asymmetry.

There is a close relationship between the quality of internal control and auditing fees. Firstly, auditors charge higher auditing fees to companies with lower internal control quality [48]. When internal control deficiencies in terms of risk and profitability or the internal control risk are assessed at a high level, the auditing fees are high because auditors need to spend extra time and incur extra costs to deal with auditing testing and changing audit procedures, discussing relevant issues with management, separately thinking about the materiality and importance of a defect and classifying, etc. [49]. Due to the low quality of internal controls, auditors increase working hours and charge higher fees to clients with internal control deficiencies [50]. The magnitude of this factor increases as the underlying control problem intensifies [51]. It shows that auditors respond to a higher control risk level by increasing audit fees [50]. There is a correlation between low-quality internal control and auditing fees. Elder and Yebba (2020) found that both the level of auditing fees and the lag in auditing reports increased significantly after investigating auditing in the education industry [52].

However, some scholars hold different views. Felix believed that there was no significant correlation between internal control quality and that auditing fees and the factors such as costs, risk and workload of external auditors would not be affected by changes in internal control quality [53,54]. Blankley found that corporate financial restatement was negatively correlated with auditing fees, and internal control quality was positively correlated with the corporate financial restatement [55]. Thus, internal control quality was negatively correlated with auditing fees. Hoitash found that even if some companies had serious internal control deficiencies, their auditing fees would not be significantly higher than the industry average [56]. On the contrary, if these enterprises could correct the internal control deficiencies disclosed in previous years, the auditing fees would also decrease in the current year.

Therefore, the digital transformation affects the quality of internal control, and the quality of internal control will influence auditing fees. We expect internal control quality to mediate between corporate digital transformation and auditing fees. The second hypothesis is

H2. 

The quality of internal control has a mediating effect between digital transformation and auditing fees.

2.4.3. Digital Transformation, Governance Information Asymmetry and Auditing Fees

According to the agency theory, the separation of ownership and management leads to inconsistency of interests between shareholders and managers, causing principal–agent problems [57]. The auditor, as an independent third party, enhances financial statements’ reliability. On the one hand, digital transformation can improve corporate governance. Manita stated that digital development expands the external audit business, making it more dependent on analyzing clients’ data to improve audit quality, which significantly improves corporate governance [14]. The results showed that the audit committee’s independence and the committee members’ professional competence were positively correlated with auditing fees. On the other hand, digital transformation may bring risks to corporate governance, involving multiple aspects such as strategy, operation, and finance and increasing the difficulty of corporate governance. Firstly, digital transformation requires continuous investment of resources, capital, and human resources, which will bring a certain degree of financial risk to the enterprise. Secondly, the timeliness of digital transformation and the flat information acquisition method have dramatically impacted the original top-down management method. The conflict and iteration of the old and new systems lead to the low operating efficiency of enterprises and aggravate information asymmetry, resulting in operational risk. Moreover, due to the high degree of uncertainty in the data information and digital environment, corporate strategy updates lag the ever-changing external environment, and strategic risk is obviously manifested in the lag of strategic transformation [58].

From the perspective of internal governance, the enterprise’s size and the business’s complexity are the most critical factors affecting auditing fees [59]. Further studies have shown that corporate governance, internal control, financial risk, operational risk, and control risk have an impact on auditing fees [52,60]. In addition, auditing fees may also be affected by stock prices, legal proceedings, profitability, and organizational structures [61]. Companies with a high degree of digital transformation have significant advantages in information transmission speed and quality, which is conducive to reducing corporate governance issues [62]. The agency theory holds that a company’s size significantly affects the principal’s supervision costs and agency costs. Most of the time, the agency cost of large companies is higher than that of small and medium companies. With a higher level of governance, the company can more reasonably and effectively control its agency issues, such as reducing information asymmetry and risk by disclosing more information [63]. Auditors can obtain useful or higher-quality information at a lower cost and reduce auditing fees. In addition, Yin found that executives who receive a higher level of remuneration would be more cautious in the choice of accounting policies, leading to auditing risk reduction accordingly, and auditing fees would be reduced [64]. Cassell found that in a company with more employees with a bachelor’s degree or higher degrees, the auditing fees were lower [65]. Jizi and Nehme investigated 794 US-listed banks from 2009 to 2015 and found a positive correlation between the auditing fees and the proportion of independent directors [66]. The above research reveals the internal influence of mechanisms among digital transformation, corporate governance, and auditing fees. Thus, the third hypothesis is

H3. 

The level of corporate governance has a mediating effect between digital transformation and auditing fees.

2.4.4. Digital Transformation, Income Information Asymmetry and Auditing Fees

Discretionary accruals are an important indicator of corporate earnings management, and earnings management is regarded as a high inherent risk, and it is also an essential consideration for auditors to assess auditing risk [67]. The high level of digitization and informatization of the audited companies may provide a new way for the management’s earnings management activities, making it more concealed and increasing the difficulty of auditing. The strength of the internal supervision level and the level of digitization impact the motivation of earnings management. When internal supervision is low and digitization is high, companies are more likely to carry out excessive earnings management and increase the uncertainty of accounting information. The decline in information transparency leads to more severe information asymmetry [68]. The studies found that auditors charged higher auditing fees for companies with significant tax differences [69,70,71]. On the one hand, earnings management with a tendency to excessive tax avoidance and the resulting aggressive tax operation motives increase auditing risk [72]. The vague transactions based on this motive reduce information transparency and exacerbate information asymmetry. On the other hand, the persistence of the cash flow surplus of companies with significant taxation differences is low, and the possibility of a potential financial crisis is relatively high [73], which increases the possibility of material misstatement in financial statements. Auditors have to adopt more stringent and complicated audit procedures, increasing audit fees. Healy and Economics also concluded that audit fees were positively correlated with corporate accrual earnings management [74]. As such, one can see a logical chain of a high degree of digitization—strong motivation for earnings management—high auditing fees.

Some scholars have specifically studied the impact of enterprise resource planning (ERP) systems. On the one hand, the ERP system, as an information technology, can realize real-time search and release of information, speed up accounting information processing, improve efficiency and information-analytical capability, help to break information transmission barriers, and reduce the distortion of accounting information caused by transmission and effectively improves the quality of earnings. On the other hand, implementing an ERP system increases the ability of the management to manipulate information, making the information easy to be tampered with, causing information loss or damage, and reducing the quality of earnings [75,76,77]. In addition, Morris and Laksmana believe that enterprises using ERP systems were less likely to conduct excessive earnings management, and auditors charged lower auditing fees [78].

To sum up, the level of digital transformation impacts discretionary accruals, significantly affecting auditing fees. Thus, the fourth hypothesis is proposed:

H4. 

Discretionary accruals mediate between digital transformation and auditing fees.

3. Research Design

3.1. Sample Selection

We selected a preliminary sample of 4170 A-share companies listed on the Shanghai Stock Exchange and the Shenzhen Stock Exchange from 2010 to 2020. To ensure the reliability of the testing results, the original samples were sorted as follows: (1) According to the 2012 Industry Classification of the China Securities Regulatory Commission, financial institutions are excluded, because the financial industry may have a better understanding of the audit process that affects the conclusions of the study. (2) ST companies (companies that have incurred losses for two consecutive accounting years) and *ST companies (companies that have incurred losses for three consecutive years and are at risk of delisting) are excluded, because their abnormal operating conditions may affect the robustness of our research results. (3) Companies are excluded if missing data. The original financial data are collected from CSMAR database. The data related to the internal control are collected from DIB database.

3.2. Research Model

3.2.1. Digital Transformation and Audit Fees

To test hypothesis H1, we construct the following fixed effect regression model:

$$FE{E}_{\mathrm{it}}={\alpha }_0+{\alpha }_{1}DC{G}_{it}+{\alpha }_{2}DC{G}_{it}^2+{\alpha }_{3}contro{l}_{it}+{\mu }_s+{\lambda }_t+{\epsilon }_{it}$$

(1)

$FE{E}_{\mathrm{it}}$ represents the audit fee, which is measured by the natural logarithm of the total audit fee of the company in that year; $DC{G}_{it}$ represents the digital transformation. Considering the data availability and consistency of the measurement, the digital transformation information is all obtained from the CSMAR database. In the samples’ 2010–2020 annual reports, the keywords, namely artificial intelligence (AI), cloud computing (CC), big data (BD) and blockchain (BC), were manually collected. The digital transformation is measured by the natural logarithm of the frequency of these keywords [79].

Referring to the existing literature, we first control the size of listed companies and the year of listing [34]. The former is represented as SIZE, which is measured by the natural logarithm of the total assets. The latter is denoted as AGE, which is calculated by subtracting the company’s first year of listing from the sample year. Second, following the literature [80], companies’ performance, growth and capital are controlled by ROA, revenue growth (RG), and leverage (LEV). Finally, the ownership structure has been considered to have a significant impact on the corporate decision-making process [81]. It is represented as TOPHOLD and is measured by the shareholding ratio of the top ten shareholders. ${\mu }_s$ represents the time-fixed effect, controlling the difference among sample years, ${\lambda }_t$ represents the industry-fixed effect, controlling the difference among industries. ${\epsilon }_{it}$ is the random error.

3.2.2. Mediating Effect

To verify the mediating effects of internal control, corporate governance, and discretionary accruals on the digital transformation and audit fees, we construct the following three models to test H2, H3, and H4.

$$FE{E}_{\mathrm{it}}={\beta }_0+{\beta }_{1}DC{G}_{it}+{\beta }_{2}DC{G}_{it}^2+{\beta }_{3}contro{l}_{it}+{\mu }_s+{\lambda }_t+{\epsilon }_{it}$$

(2)

$$M_{\mathrm{it}}={\gamma }_0+{\gamma }_{1}DC{G}_{it}+{\gamma }_{2}DC{G}_{it}^2+{\gamma }_{3}contro{l}_{it}+{\mu }_s+{\lambda }_t+{\epsilon }_{it}$$

(3)

$$FE{E}_{\mathrm{it}}={\delta }_0+{\delta }_{1}DC{G}_{it}+{\delta }_{2}DC{G}_{it}^2+{\delta }_3{M}_{it}+{\delta }_{4}contro{l}_{it}+{\mu }_s+{\lambda }_t+{\epsilon }_{it}$$

(4)

$M_{\mathrm{it}}$ represents the mediation variable, that is, the mediating effect of the information asymmetry. Model (2) tests the correlation between the digital transformation and audit fees. Model (3) tests the significance of digital transformation on mediator variables. Model (4) investigates the impacts of the digital transformation on audit fees with mediating variables.

Based on the internal control theoretical framework outlined in the Internal Control Implementation Guide, which is part of the COSO framework (2013), we use the natural logarithm of the Dibo internal control index to measure a company’s internal control quality (IC). We use the principal component analysis method to construct a company’s corporate governance index. Specifically, we collect seven major governance data from CSMAR database including: executive compensation, executive shareholding, institutional investor shareholding, the number of independent directors, the board size, ownership balance degree (shareholding ratio of the second to fifth largest shareholders/controlling shareholder’s shareholding) and whether the chairman and the general manager are the due role. They are then used as the inputs in the principal component analysis. The first principal component (the score) obtained from the component analysis is used to measure a company’s corporate governance (CG). The higher the score, the better the level of governance. The discretionary accruals (DA) are derived from the CSMAR database.

4. Results

4.1. Descriptive Statistics

Table 1. Descriptive Statistics.

	Obs.	Mean	Std. Dev.	Min.	Max.
FEEit	12,279	1,423,226	1,544,383	300,000	1.10 × 107
AIit	12,279	2.245	9.377	0	256
CCit	12,279	4.823	13.629	0	229
BDit	12,279	0.575	3.790	0	157
BCit	12,279	0.059	0.564	0	19
DCGit	12,279	7.947	16.714	0	422
ICit	12,279	658.092	78.957	8.97	835.8
CGit	11,767	0.0999	1.042	−1.568	2.930
DAit	12,279	0.002	0.138	−3.642	2.341
ROAit	12,279	0.039	0.062	−0.320	0.205
SIZEit	12,279	22.284	1.269	19.835	26.087
RGit	12,279	0.307	4.637	−0.952	363.068
AGEit	12,279	10.963	7.213	2	31
LEVit	12,279	0.412	0.195	0.051	0.884
TOPHOLDit	12,279	0.586	0.144	0.241	0.902

shows the descriptive statistics. The maximum value of audit fees ($FE{E}_{\mathrm{it}}$) is 110,000,000, the minimum value is 300,000, the mean is 1,423,226, and the standard deviation is 1,544,383, indicating that there are differences in audit fees among the sample companies. The minimum values of the indicators measuring digital transformation are all 0, indicating that some sample companies have not yet begun the digital transformation process. In addition, the standard deviation of artificial intelligence (AI_it) is 9.377, and its average is 2.245. The standard deviation of cloud computing (CC_it) is 13.623, and its average is 4.823. The standard deviation of big data (BD_it) is 3.790. Its average is 0.575. The standard deviation of the blockchain (BC_it) is 0.564, and its average is 0.059. Based on these statistics, cloud computing (CC_it) is the most widely used form of the digital transformation. There is a large scale of differences in the digitization. The digital transformation among sample companies is in the developing stage.

Among the three mediating variables, the standard deviation of the quality of internal control (IC_it) is 78.957, and its average is 658.092, indicating that the internal control quality is significantly different among sample companies. The maximum value of the corporate governance (CG_it) is 2.930, and its minimum value is −1.568, indicating that the governance level in the context of the digital transformation is not consistent. It can be caused by the size of the sample company and its industry. The standard deviation of discretionary accruals (DA_it) is 0.138, and its mean is 0.002, its maximum value is 2.341, and the minimum value is −3.642, indicating that the scale of earnings management among samples is different. There are large dimensional differences among samples, so all continuous variables were standardized.

4.2. Base Regression

The results of the xtgee tests are shown in

Table 2. Results: Base Regression.

	M1 FEEit (1)	M2 FEEit (2)	M3 FEEit (3)	M4 FEEit (4)
Testing Variables
AIit	0.116 ***	0.099 ***	0.088 ***	0.088 ***
	(0.021)	(0.021)	(0.022)	(0.022)
AIit2	−0.011 ***	−0.009 **	−0.009 **	−0.009 **
	(0.004)	(0.004)	(0.004)	(0.004)
CCit		0.078 ***	0.072 ***	0.072 ***
		(0.021)	(0.021)	(0.021)
CCit2		−0.014 ***	−0.013 ***	−0.013 ***
		(0.004)	(0.004)	(0.004)
BDit			0.074 ***	0.073 ***
			(0.022)	(0.022)
BDit2			−0.008 **	−0.008 **
			(0.004)	(0.004)
BCit				0.006
				(0.016)
BCit2				−0.001
				(0.001)
Controls
ROAit	0.030 ***	0.030 ***	0.030 ***	0.030 ***
	(0.008)	(0.008)	(0.008)	(0.008)
SIZEit	−2.630 ***	−2.606 ***	−2.613 ***	−2.614 ***
	(0.610)	(0.610)	(0.609)	(0.609)
RGit	0.014 *	0.013 *	0.013 *	0.013 *
	(0.008)	(0.008)	(0.008)	(0.008)
AGEit	0.187 ***	0.187 ***	0.188 ***	0.188 ***
	(0.010)	(0.009)	(0.009)	(0.009)
LEVit	2.903 ***	2.878 ***	2.884 ***	2.885 ***
	(0.610)	(0.610)	(0.609)	(0.609)
TOPHOLDit	0.217 ***	0.217 ***	0.218 ***	0.218 ***
	(0.009)	(0.009)	(0.009)	(0.009)
Period-fixed	Y	Y	Y	Y
Industry-fixed	Y	Y	Y	Y
C	−0.239 *	−0.210 *	−0.185	−0.184
	(0.127)	(0.127)	(0.127)	(0.127)
Obs.	12,279	12,279	12,279	12,279
Wald Chi2	4805.02 ***	4825.03 ***	4849.40 ***	4850.72 ***

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

. According to columns (1), (2), and (3), the signs of the coefficients of AI_it, CC_it, and BD_it are significantly positive, and the coefficients of their quadratic terms are significantly negative. These results indicate that there is an inverted U-shaped relationship between the digital transformation and audit fees. That is, audit fees will increase at the initial stage of the digital transformation, and when the degree of transformation exceeds a certain level, audit fees will decrease. Hypothesis H1 is supported. It is argued that the impacts of the digital transformation on audit fees are different at the different company development stages. It is a nonlinear relationship between the digital transformation and companies’ audit fees. In column (4), the coefficient of BC_it is 0.006, and the coefficient of its quadratic term is −0.001. Results are not significant.

4.3. Results: Mediating Effects

The mechanism of how the digital transformation affects audit fees is investigated in this section. From Table 2, the relationship between the blockchain and audit fees is not significant. Therefore, this relationship will not be discussed further.

4.3.1. Internal Control

We use the mediated stepwise model to verify whether the quality of internal control acts as a mediating variable. The results are shown in

Table 3. Mediator: Internal Control.

	M1 FEEit (1)	M2 FEEit (2)	M3 ICit (3)	M4 ICit (4)	M5 ICit (5)
Testing Variables
AIit	0.088 ***	0.078 ***	0.073 ***
	(0.022)	(0.021)	(0.022)
AIit2	−0.009 **	−0.008 **	−0.008 **
	(0.004)	(0.004)	(0.004)
CCit	0.072 ***	0.065 ***		0.062 ***
	(0.021)	(0.021)		(0.021)
CCit2	−0.013 ***	−0.012 ***		−0.008 *
	(0.004)	(0.004)		(0.004)
BDit	0.074 ***	0.079 ***			−0.013
	(0.022)	(0.021)			(0.022)
BDit2	−0.008 **	−0.009 **			0.003
	(0.004)	(0.004)			(0.004)
ICit		0.142 ***
		(0.009)
Controls
ROAit	0.030 ***	−0.024 ***	0.383 ***	0.383 ***	0.383 ***
	(0.008)	(0.009)	(0.009)	(0.009)	(0.009)
SIZEit	−2.613 ***	−2.172 ***	−3.117 ***	−3.108 ***	−3.141 ***
	(0.609)	(0.603)	(0.634)	(0.635)	(0.635)
RGit	0.013 *	0.011	0.018 **	0.018 **	0.018 **
	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)
AGEit	0.188 ***	0.184 ***	0.025 **	0.024 **	0.024 **
	(0.009)	(0.009)	(0.010)	(0.010)	(0.010)
LEVit	2.884 ***	2.426 ***	3.238 ***	3.228 ***	3.263 ***
	(0.609)	(0.603)	(0.634)	(0.635)	(0.635)
TOPHOLDit	0.218 ***	0.205 ***	0.093 ***	0.093 ***	0.092 ***
	(0.009)	(0.009)	(0.009)	(0.009)	(0.009)
Period-fixed	Y	Y	Y	Y	Y
Industry-fixed	Y	Y	Y	Y	Y
C	−0.185	−0.260 **	0.525 ***	0.513 ***	0.475 ***
	(0.127)	(0.126)	(0.132)	(0.132)	(0.132)
Obs.	12,279	12,279	12,279	12,279	12,279
Wald Chi2	5232.75 ***	4849.40 ***	3514.47 ***	3508.23 ***	3492.91 ***

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

. Column (1) presents the test results for the impact of digital transformation on audit fees. After controlling for other factors, the coefficient of AI_it is significant and positive, while the coefficient of its quadratic term is also significant but negative, indicating a mediating effect. Column (2) shows the results regarding whether digital transformation affects the quality of internal control, revealing a significant relationship between digital transformation and the quality of internal control. Columns (3), (4), and (5) display the results for the impact of digital transformation on audit fees after including the quality of internal control as a mediating variable. As shown in Table 3, except for big data (BDit) in the last column, all variables in the models of the other columns are significant. The coefficient for the effect of the quality of internal control on audit fees is 0.142 (Column (2)), which is significant at the 1% level. This demonstrates that the quality of internal control plays a mediating role between digital transformation and audit fees, indicating that financial information asymmetry serves as a mediating variable between digital transformation and audit fees. Hypothesis H2 is supported.

4.3.2. Corporate Governance

In Table 2, it has been verified that the digital transformation has significant impacts on audit fees. Column (2) of

Table 4. Mediator: Corporate Governance.

	M1 FEEit (1)	M2 FEEit (2)	M3 CGit (3)	M4 CGit (4)	M5 CGit (5)
Testing Variables
AIit	0.088 ***	0.094 ***	0.061 ***
	(0.022)	(0.021)	(0.022)
AIit2	−0.009 **	−0.010 **	−0.008 **
	(0.004)	(0.004)	(0.004)
CCit	0.072 ***	0.075 ***		0.030
	(0.021)	(0.021)		(0.021)
CCit2	−0.013 ***	−0.013 ***		−0.000
	(0.004)	(0.004)		(0.004)
BDit	0.074 ***	0.064 ***			−0.036 *
	(0.022)	(0.022)			(0.022)
BDit2	−0.008 **	−0.006 *			0.006
	(0.004)	(0.004)			(0.004)
CGit		−0.045 ***
		(0.009)
Controls
ROAit	0.030 ***	0.026 ***	−0.066 ***	−0.066 ***	−0.067 ***
	(0.008)	(0.008)	(0.009)	(0.009)	(0.009)
SIZEit	−2.613 ***	−2.562 ***	−0.047	−0.045	−0.069
	(0.609)	(0.626)	(0.650)	(0.650)	(0.651)
RGit	0.013 *	0.048 ***	0.015	0.015	0.015
	(0.008)	(0.013)	(0.013)	(0.013)	(0.013)
AGEit	0.188 ***	0.176 ***	−0.369 ***	−0.370 ***	−0.370 ***
	(0.009)	(0.010)	(0.010)	(0.010)	(0.010)
LEVit	2.884 ***	2.820 ***	−0.093	−0.096	−0.070
	(0.609)	(0.626)	(0.650)	(0.650)	(0.650)
TOPHOLDit	0.218 ***	0.200 ***	−0.078 ***	−0.078 ***	−0.079 ***
	(0.009)	(0.009)	(0.009)	(0.009)	(0.009)
Period-fixed	Y	Y	Y	Y	Y
Industry-fixed	Y	Y	Y	Y	Y
C	−0.185	−0.249 *	−0.774 ***	−0.792 ***	−0.827 ***
	(0.127)	(0.132)	(0.137)	(0.136)	(0.136)
Obs.	12,279	11,767	11,767	11,767	11,767
Wald Chi2	4849.40 ***	4741.08 ***	4160.97 ***	4161.50 ***	4151.18 ***

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

shows the results for whether the digital transformation affects corporate governance. All coefficients are significant, indicating a strong association between them. Columns (3), (4) and (5) show the results for the impacts of the digital transformation on audit fees after adding corporate governance as the intermediating factor. As shown in Table 4, the coefficients of AI_it and BD_it are significant, and the coefficient of AI_it² is also significant. The coefficient of the level of corporate governance on audit fees is −0.045 (Column (2)), and it is significant at the 1% level. These results prove that corporate governance plays a mediating role, that is, governance information asymmetry is a mediating factor between the digital transformation and audit fees, supporting Hypothesis H3.

4.3.3. Discretionary Accruals

The results in Column (2) of

Table 5. Mediator: Discretionary Accruals.

	M1 FEEit (1)	M2 FEEit (2)	M3 DAit (3)	M4 DAit (4)	M5 DAit (5)
Testing Variables
AIit	0.088 ***	0.087 ***	−0.010
	(0.022)	(0.022)	(0.022)
AIit2	−0.009 **	−0.009 **	0.007 *
	(0.004)	(0.004)	(0.004)
CCit	0.072 ***	0.072 ***		0.020
	(0.021)	(0.021)		(0.021)
CCit2	−0.013 ***	−0.013 ***		−0.004
	(0.004)	(0.004)		(0.004)
BDit	0.074 ***	0.076 ***			0.074 ***
	(0.022)	(0.022)			(0.022)
BDit2	−0.008 **	−0.008 **			−0.014 ***
	(0.004)	(0.004)			(0.004)
DAit		−0.032 ***
		(0.009)
Controls
ROAit	0.030 ***	0.042 ***	0.383 ***	0.383 ***	0.383 ***
	(0.008)	(0.009)	(0.009)	(0.009)	(0.009)
SIZEit	−2.613 ***	−2.603 ***	0.284	0.299	0.309
	(0.609)	(0.609)	(0.628)	(0.628)	(0.628)
RGit	0.013 *	0.014 *	0.023 ***	0.023 ***	0.023 ***
	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)
AGEit	0.188 ***	0.188 ***	0.010	0.010	0.010
	(0.009)	(0.009)	(0.010)	(0.010)	(0.010)
LEVit	2.884 ***	2.875 ***	−0.260	−0.276	−0.287
	(0.609)	(0.609)	(0.628)	(0.628)	(0.628)
TOPHOLDit	0.218 ***	0.217 ***	−0.024 ***	−0.025 ***	−0.024 ***
	(0.009)	(0.009)	(0.009)	(0.009)	(0.009)
Period-fixed	Y	Y	Y	Y	Y
Industry-fixed	Y	Y	Y	Y	Y
C	−0.185	−0.169	0.482 ***	0.490 ***	0.510 ***
	(0.127)	(0.127)	(0.130)	(0.130)	(0.130)
Obs.	12,279	12,279	12,279	12,279	12,279
Wald Chi2	4849.40 ***	4867.75 ***	3847.18 ***	3833.48 ***	3850.14 ***

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

are used to verify whether the digital transformation affects discretionary accruals. The coefficients are all significant, indicating the digital transformation is associated with discretionary accruals. Columns (3), (4) and (5) in Table 5 show the results for the mediating effects of discretionary accruals. After controlling other factors, the coefficients indexing artificial intelligence, cloud computing and big data are all different from zero. The coefficients of AI_it², BD_it and BD_it² are all significant. The coefficient of DA_it is −0.032 (Column (2)) and is significant at the 1% level. These results prove that discretionary accruals play a mediating role. Income information asymmetry is a mediator between the digital transformation and audit fees. Hypothesis H4 is supported.

4.4. Bootstrap Test

This paper uses the bootstrap sampling method to validate the robustness of testing the mediating effect. Under the 95% confidence interval, all testing samples are randomly selected 1000 times. The results are shown in

Table 6. Bootstrap Test.

Mediating Relationship	Effect	Coefficient (Standard Errors)	95% Confidence Interval
			Lower Limit	Upper Limit
AI-IC-FEE	Indirect Effect	0.018 ** (0.0008)	0.00016	0.0035
	Direct Effect	0.687 *** (0.008)	0.052	0.846
CC-IC-FEE	Indirect Effect	0.020 *** (0.0004)	0.001	0.002
	Direct Effect	0.0002 *** (0.00007)	0.0001	0.0004
AI-CC-FEE	Indirect Effect	−0.004 *** (0.0007)	−0.0055	−0.0025
	Direct Effect	0.070 *** (0.007)	0.0548	0.0858
BD-DA-FEE	Indirect Effect	0.0013 *** (0.0004)	0.0004	0.002
	Direct Effect	0.050 *** (0.006)	0.037	0.063

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

. Under the 95% confidence interval, all interaction terms of coefficients exclude zero, strongly indicating the existence of mediating effects. Bootstrap testing results support our findings.

4.5. Robustness Check

To verify the reliability of the empirical results and correct potential endogenous issues, we conduct the following robustness tests.

In order to enhance the robustness of the findings, the digital transformation is re-measured. This paper uses the overall frequency of keywords counted to measure the digital transformation (DCG) and re-tests all the models. The results are shown in

Table 7. Replacement of Measurement.

	M1 FEEit (1)	M2 FEEit (2)	M3 FEEit (3)
Testing Variables
DCGit		0.044 ***	0.084 ***
		(0.006)	(0.012)
DCGit2			−0.011 ***
			(0.003)
Controls
ROAit	0.007	0.007	0.006
	(0.005)	(0.005)	(0.005)
SIZEit	1.393 ***	1.422 ***	1.442 ***
	(0.368)	(0.366)	(0.366)
RGit	0.007	0.006	0.006
	(0.004)	(0.004)	(0.004)
AGEit	0.185 ***	0.188 ***	0.188 ***
	(0.016)	(0.016)	(0.016)
LEVit	−1.260 ***	−1.291 ***	−1.310 ***
	(0.368)	(0.366)	(0.366)
TOPHOLDit	0.118 ***	0.123 ***	0.123 ***
	(0.008)	(0.008)	(0.008)
Period-fixed	Y	Y	Y
Industry-fixed	Y	Y	Y
C	−0.304 **	−0.271 *	−0.249 *
	(0.149)	(0.148)	(0.148)
Obs.	12,279	12,279	12,279
Wald Chi2	4212.55 ***	4297.95 ***	4319.05 ***

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

. It can be found that after changing the measurement method, the coefficients are still significant, indicating the findings are robust.

To alleviate the endogeneity issue that may be caused by reciprocal causality between the digital transformation and audit fees, we conduct a two-stage regression (2SLS) using the digital technology application (Digitalusage) as an instrumental variable. Digitalusage is a continuous variable and is collected from the CSMAR database. The weak instrumental variable test and the over-identification test are both carried out to check the instrumental variable. When the digital usage is greater than average, it is denoted as 1, otherwise it is denoted as 0. The Sargon value in the regression result is 0, indicating that the instrumental variable is exogenous. The Cragg–Donald Wald F values are respectively 198.98, 386.66 and 105.99, verifying the effectiveness and rationality of the instrumental variable selected. As shown in

Table 8. Results: 2SLS.

	Second Stage			First Stage
	M1 FEEit (1)	M2 FEEit (2)	M3 FEEit (3)	AIit (4)	CCit (5)	BDit (6)
Testing Variables
Digitalusage				0.368 ***	0.309 ***	0.251 ***
				(0.020)	(0.019)	(0.021)
AIit	0.527 ***
	(0.168)
AIit2	−0.080 ***
	(0.028)
CCit		0.423 ***
		(0.117)
CCit2		−0.076 ***
		(0.022)
BDit			0.798 ***
			(0.237)
BDit2			−0.119 ***
			(0.037)
Controls
ROAit	0.032 ***	0.027 ***	0.032 ***	−0.001	−0.004	−0.021 **
	(0.009)	(0.009)	(0.009)	(0.009)	(0.008)	(0.009)
SIZEit	−2.494 ***	−2.509 ***	−2.571 ***	0.174	−0.396	0.595
	(0.622)	(0.618)	(0.637)	(0.632)	(0.615)	(0.668)
RGit	0.014 *	0.011	0.012	0.003	0.008	0.004
	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)
AGEit	0.189 ***	0.185 ***	0.193 ***	−0.018 *	−0.004	−0.026 **
	(0.010)	(0.010)	(0.010)	(0.010)	(0.010)	(0.010)
LEVit	2.767 ***	2.774 ***	2.833 ***	−0.181	0.435	−0.593
	(0.622)	(0.618)	(0.638)	(0.632)	(0.614)	(0.668)
TOPHOLDit	0.218 ***	0.216 ***	0.224 ***	−0.047 ***	−0.044 ***	−0.038 ***
	(0.009)	(0.009)	(0.009)	(0.009)	(0.009)	(0.010)
Period-fixed	Y	Y	Y	Y	Y	Y
Industry-fixed	Y	Y	Y	Y	Y	Y
C	0.039	−0.103	−0.009	−0.755 ***	−0.659 ***	−0.628 ***
	(0.167)	(0.140)	(0.159)	(0.131)	(0.127)	(0.138)
Obs.	12,279	12,279	12,279	12,279	12,279	12,279
CD-W F test	198.95	386.66	105.99
Sargon	0.00	0.00	0.00
Wald Chi2				3626.17 ***	4568.49 ***	1977.19 ***

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

, after controlling endogeneities, the regression results are all consistent with the previous testing results. The findings are still robust and reliable.

Considering the self-selection issue, we further conducted a Heckman two-step test. Digitalusagebool is a binary variable. If a company applies digital technology, it is denoted as 1; otherwise, it is denoted as 0. The results are shown in

Table 9. Results: Heckman Two-Stage.

Step One: Selection Model	M1 Digitalusagebool	M2 Digitalusagebool	M3 Digitalusagebool
ROAit	−0.020	−0.020	−0.020
	(0.016)	(0.016)	(0.016)
SIZEit	−1.070	−1.070	−1.070
	(1.148)	(1.148)	(1.148)
RGit	0.159 *	0.159 *	0.159 *
	(0.082)	(0.082)	(0.082)
AGEit	0.017	0.017	0.017
	(0.018)	(0.018)	(0.018)
LEVit	1.110	1.110	1.110
	(1.147)	(1.147)	(1.147)
TOPHOLDit	0.003	0.003	0.003
	(0.016)	(0.016)	(0.016)
Period-fixed	Y	Y	Y
Industry-fixed	Y	Y	Y
C	1.157	1.157	1.157
	(0.237)	(0.237)	(0.237)
Step Two: Response Model	FEEit	FEEit	FEEit
AIit	0.110 ***	0.091 ***	0.079 ***
	(0.022)	(0.023)	(0.023)
AIit2	−0.010 ***	−0.008 **	−0.007 *
	(0.004)	(0.004)	(0.004)
CCit		0.080 ***	0.073 ***
		(0.022)	(0.022)
CCit2		−0.014 ***	−0.013 ***
		(0.004)	(0.005)
BDit			0.079 ***
			(0.024)
BDit2			−0.009 **
			(0.004)
Controls
ROAit	0.028 ***	0.027 ***	0.028 ***
	(0.010)	(0.010)	(0.010)
SIZEit	−2.236 ***	−2.203 ***	−2.200 ***
	(0.761)	(0.763)	(0.764)
RGit	0.010	0.010	0.009
	(0.009)	(0.009)	(0.009)
AGEit	0.192 ***	0.192 ***	0.192 ***
	(0.012)	(0.012)	(0.012)
LEVit	2.496 ***	2.461 ***	2.458 ***
	(0.763)	(0.764)	(0.765)
TOPHOLDit	0.219 ***	0.219 ***	0.220 ***
	(0.011)	(0.011)	(0.011)
Period-fixed	Y	Y	Y
Industry-fixed	Y	Y	Y
C	−0.070	−0.031	−0.003
	(0.190)	(0.191)	(0.191)
Obs.	12,279	12,279	12,279
Selected Obs.	10,434	10,434	10,434
Non-selected Obs.	1845	1845	1845
lambda	−0.890	−0.903	−0.911
	(0.477)	(0.478)	(0.478)
Wald Chi2	3418.77 ***	3416.80 ***	3424.94 ***

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

. By controlling the self-selection, the results are all consistent and support the findings.

5. Conclusions

This study uses the data of A-share listed companies in China’s Shanghai and Shenzhen Stock markets from 2010 to 2020 as a sample by selecting internal control quality, corporate governance level, and discretionary accruals as mediating variables to analyze the relationships between the digital transformation and audit fees. It is found that the impact of a company’s digital transformation on audit fees is in an inverted U-shape, indicating that they are positively correlated within a specific range, but when the digital transformation reaches a certain level, they are negatively correlated. During the early stages of the digital transformation, the internal and external auditing environments would change to varying degrees, particularly in the complexity of information and the slowness of staff in adapting to the new development, making auditing tasks more challenging and ineffective. With the digital transformation gradually maturing, companies can reduce information asymmetry and simplify information interaction by improving their business performance, corporate information quality, transparency, and analysis processes to reduce auditing fees. In addition, the internal control quality, corporate governance levels, and discretionary accruals play a mediating role between the digital transformation and audit fees. The digital transformation affects the accuracy and timeliness of the corporate control environment and internal control risk identified. While promoting the improvement of corporate governance and the quality of corporate earnings, it changes auditing costs and ultimately affects audit fees. The findings are supported by the bootstrap sampling and robustness checks.

Based on the above conclusions, this study has the following implementations. For listed companies, creating a solid digital transformation development strategy, combining corporate goals, understanding the pace and severity of the change, and dealing sensibly with any short-term negative effects are all important. Secondly, enterprises should hasten the development of the digital management system, promptly identify the modifications brought by the digital transformation of auditing, ensure the smooth operation of internal control, and improve corporate governance. According to the development stage of a company, a good earnings management strategy needs to be appropriately carried out to mitigate the negative impact brought on by the fluctuation in auditing fees in the early stages of digitization. Accounting firms should strike a balance between the degree of digitization and auditing fees, accelerate their digital development, build new digital platforms, such as information-sharing systems and pricing decision-making systems, promote and adapt to the audit digital transformation, and ensure the rationality of auditing fees, taking into account customer data, audit risk, and auditing efficiency. In addition, they should focus on recruiting and cultivating digital compound talents and building a modern audit team. When recruiting employees, it is important to evaluate their professional ability and their capacity to assimilate new information and enhance their theoretical knowledge through digital actual combat simulation and case summary teaching to guide subsequent practical work; for existing employees, retention will enhance their willingness and ability to use digital technology to conduct audit work.