Economic Policy Uncertainty: Does It Truly Matter?—Evidence from Corporate Fraudulent Behaviors in Chinese Capital Market

Abstract

A macro policy environment affects managers’ decision-making behaviors. When there is greater uncertainty in economic policy, will they engage in extreme violations? This paper explores the economic consequences of policy uncertainty at the firm level from the perspective of corporate fraud. We focus on the fraudulent behaviors of listed companies in the Chinese capital market and conduct our empirical research through the multiple mediation model. The results show that economic policy uncertainty not only has a direct effect on corporate fraud but it also has a mediating effect that can be explained by four mediating variables. Increased economic policy uncertainty will increase the likelihood of company fraud, and this direct effect will vary due to corporate heterogeneity. Companies with a lower risk tolerance and productive capacity will be more affected and have more significant violation motives. From the results of the mediating effect test, we determine that increased economic policy uncertainty increases mergers and acquisitions, decreases cash holdings, increases stock price volatility, and decreases institutional investors’ shareholdings, which will increase the possibility of corporate fraud.

1. Introduction

Since economic fluctuations cannot be accurately predicted and judged, economic policies related to these fluctuations also show varying degrees of uncertainty. In recent years, uncertainty about economic policy has been widespread around the world [1], particularly during periods of economic downturn [2], when government actions to prevent an economic recession increase policy-related economic uncertainty [3]. After the 2008 financial crisis, governments adjusted their economic policies to revive their countries’ economies: examples include the Fed’s (the Federal Reserve System) four rounds of quantitative easing (QE) and the Chinese government’s “four trillion rescue” economic stimulus package. Although the formulation and implementation of these policies help alleviate the economic growth dilemma (at least to a certain extent), these policy reform initiatives are unprecedented, which makes their long-term impacts and economic consequences difficult to assess, inevitably increasing policy-related uncertainty and the potential risks of economic development.

At present, the Chinese economy is in an important stage of comprehensively deepening reforms, and it is also a crucial period for both realizing an economic transition and upgrading the economy. Compared with other mature and developed market economies, Chinese economic policy has experienced significant uncertainty [4]. The Chinese government uses fiscal, monetary, and regulatory policy instruments to regulate macroeconomic goals, deepen “supply-side structural reforms”, stimulate the competition mechanism of market participants, and promote high-quality economic development. In this process, the original intention of policy implementation was to improve economic conditions, but it has potentially increased the uncertainty of economic development and increased the risk of adverse economic consequences.

Researchers currently have opposing views on the economic impact of economic policy uncertainty: on the one hand, the uncertainty in economic policy will affect the strategic development of enterprises. Frequent changes in economic policies make the business environment of enterprises complicated, changeable, and unpredictable, leaving market players in a vague business environment with poor anticipation. This greatly limits managers’ confidence in policy expectations and makes it difficult to accurately judge the time when policies are introduced and the potential impact after a policy is introduced; thus, they are unable to make timely adjustments and responses. Benhabib, Liu, and Wang believed that the financial uncertainty of the capital market and macroeconomic uncertainty are linked [2]. As a major part of the capital market and the market economy, listed companies are directly affected by macroeconomic policies mainly in three aspects: uncertainty caused by the company’s pre-policy expectations and adjustments; uncertainty caused by policy changes; and uncertainty caused by policy implementation. An increase in economic policy uncertainty means that a company’s external environmental risks are increasing, stock price volatility is increasing, and earnings growth is stagnant [5]. Additionally, the company will face higher decision-making costs and decision-making risks. Youssef et al. focused on eight countries where COVID-19 was the most widespread (China, Italy, France, Germany, Spain, Russia, the US, and the UK) and found that stock markets were highly connected during the entire period, but the dynamic spillovers reached unprecedented heights during the COVID-19 pandemic in the first quarter of 2020.

On the other hand, economic policy uncertainty may create new markets and investment opportunities, boosting the development of enterprises. How to deal with the uncertainty of economic policy becomes the key issue that determines the survival and development of enterprises. Hoang et al. pointed out the positive association between EPU and corporate diversification [6]. Moreover, economic policy uncertainty may also be conductive to innovation and sustainable development in some countries. Alandejani and Al-Shaer investigated the companies located in the USA, China, and the UK during the period of 2013–2020, and their results showed that during times of economic uncertainty, companies were more likely to engage in ESG activities, including establishing emission reduction targets [7]. Shabbir et al. analyzed how trade policy uncertainty and sustainable development policies affected investment in medical innovation in developing nations [8]. They found that developing countries have had little effect on the long-term ramifications of sectoral innovation patterns, political shifts, and imported technology. Adedoyin et al. investigated the role of economic policy uncertainty in the energy–growth–emissions nexus for 32 countries in Sub-Saharan Africa over the period from 1996 to 2014. Although economic policy uncertainty also propels high levels of emissions in the region, its moderating effect on the impact of both renewable and non-renewable energy generation leads to a reduction in the emissions level in the region [9].

In order to avoid the negative impact of economic policy uncertainty, enterprises need to take effective countermeasures based on predictions of future policy trends and different possible situations. However, in this process, enterprises are often faced with the asymmetric problem of cost and benefit under the expectation of policy uncertainty: higher adjustment costs and lower expected income are generated in order to cope with the uncertainty impact. This is embodied in the following three aspects, the first of which is the uncontrollability of sunk costs. It is mainly due to the deviation of the policy expectations of management that an enterprise does not develop along the correct track and thus faces higher sunk costs and failure risks. The second is the complexity of opportunity costs. In anticipation of policy uncertainty, enterprises plan and deploy according to different possible situations. The adjustments and countermeasures made by enterprises under different strategies will inevitably cost a lot of internal and external resources. Such diversified and complex opportunity cost sources lead to the liquidity dilemma of enterprises, resulting in difficult business development. The third is the fuzziness of unanticipated earnings. The adjustment and transformation of enterprises also carry hidden risks, resulting in great uncertainty in future expected earnings. Enterprises cannot clearly judge whether their operation and development can meet expectations, thus increasing the difficulty of decision-making.

Therefore, this paper focuses on fraudulent corporate behaviors under the uncertainty of a macro policy environment. Through a theoretical analysis and empirical tests, it clarifies the influence path of economic policy uncertainty on corporate fraud and draws policy-oriented conclusions. Specifically, our research makes four main contributions.

First, our research makes a connection between macroeconomic policies and the micro behavior of enterprises and discusses the economic consequence of economic policy uncertainty at the firm level from the perspective of corporate fraud. Because of the different levels of corporate risk tolerance and productive capacity, we further discuss the heterogeneity of corporate fraud under the impact of economic policy uncertainty. The results show that economic policy uncertainty is associated with the possibility of corporate fraud, and companies with a lower risk tolerance and lower productive capacity will be more affected and show more significant fraud motivation.

Second, a small number of existing papers have considered the impact of economic policy uncertainty on corporate profit manipulation, earnings management, and other irregularities [4,10], but research is lacking on the direct relationship and influence path of policy uncertainty and corporate fraud. Our paper closely examines corporate fraudulent behaviors and explores the impact mechanism of economic policy uncertainty on corporate fraud based on a theoretical and empirical study. Using firm-level data, we find four mediating variables between economic policy uncertainty and corporate fraud and confirm that increased economic policy uncertainty results in increases in mergers and acquisitions, decrease in cash holdings, the expansion of stock price volatility, and a reduction in institutional investors’ shareholdings, which will, in turn, increase the probability of corporate fraud.

Third, we develop our empirical analysis by constructing a multiple mediation model with reference to the methods of the psychology and social science fields. In this model, the main explanatory variable exerts a direct effect on the dependent variable while indirectly exerting a mediating effect on the dependent variable through the mediating variables. We sort out the empirical test procedure of the mediating effect and propose a mediating effect test for binary dependent variables. More specifically, we summarize the outcomes of mediating effect regressions and make our judgement strictly according to the procedure. In the robustness section, various methods are used to test the robustness of the empirical results.

Fourth, we further explore the different consequences of economic policy uncertainty on corporate fraudulent behaviors from the perspective of the regional development level and the industry competition level, concluding that companies located in undeveloped regions and monopolized industries have a greater tendency to commit corporate fraud under the impact of external economic policy uncertainty.

The rest of the paper is organized as follows: Section 2 describes the theoretical analysis and the research hypothesis; Section 3 presents the variable selection and model setting; Section 4 discusses the hypothesis test and results analysis of the direct effect and the mediation effect of economic policy uncertainty on corporate fraud; Section 5 presents the robustness test; Section 6 provides a further discussion from the perspective of regional economy and industry competition; and Section 7 concludes the paper and proposes policy recommendations.

2. Theoretical Analysis and Research Hypotheses

2.1. The Relationship between Economic Policy Uncertainty and Corporate Fraud

A series of economic policies formulated and issued by the government will directly affect the macroeconomic environment in which a company operates. Policy uncertainty often causes serious economic consequences, such as difficulty in accurately determining the timing of policy introduction and adjusting to the potential impact of policy formulation [1,3,11]. Changes in a macro environment can affect the micro behavior of companies and executives [12], which in turn can stimulate fraudulent executive behaviors in the decision-making process. This is because uncertainty in the external environment increases the volatility of the company’s performance and increases the difficulty of the company’s business operations and management. The increased risk of drastic fluctuations in corporate business performance sends an unfavorable signal to the capital market and investors and increases corporate financing costs and credit risks [10]. When a company fails to meet its financial goals through normal operations, executives are more likely to manage earnings and violate regulations [13]. Managers treat these violations as an effective way to hedge external risks, through which they can smooth company performance volatility, achieve corporate earnings targets, meet the requirements of financial indicators, and obtain substantial personal benefits [14].

Based on the assumption of the rational economic man, executives’ fraudulent behavior is determined by the cost and benefit of violation [10]. The benefit of violation is derived from the fact that senior executives help achieve corporate business “goals” by means of illegal operations and profit falsification to fool investors and bid up the stock price, thus avoiding corporate financing constraints and lowering the risk of being delisted. At the same time, the senior executives themselves can receive compensation incentives. The cost of violation is reflected in the credibility crisis faced by executives and the disastrous impact on the company’s operation after the violation is investigated and exposed. When the benefit of a violation is greater than its cost, executives show a greater willingness to engage in fraudulent behaviors. Therefore, the relationship between economic policy uncertainty and corporate violations can be analyzed from a cost–benefit perspective. On the one hand, when economic policy uncertainty is high, there are potential benefits for corporate fraud. Since the compensation system for executives in a company is directly affected by company performance, corporate violations, such as covering up negative information and embellishing the company’s performance, result in fewer financing constraints, ensuring the company’s stable business earnings in the future. Executives can also receive a generous salary compensation when they commit violations [15,16]. On the other hand, a high level of economic policy uncertainty can decrease corporate risks and costs for two main reasons. First, the deterioration of the macroeconomic environment caused by economic policy uncertainty will affect companies in various industries, which means that all of the surviving participants in the capital market will be affected to varying degrees. Companies will make adjustments and changes in their business operations, R&D innovations, and financial management. Therefore, it is difficult to identify abnormal fluctuations in the operating and financial indicators displayed by illegal enterprises, which increases the difficulty of detection by regulatory departments and reduces the probability of the company being audited and the risk of corporate fraud. Second, economic policy uncertainty aggravates the problem of information asymmetry. When the corporate business and financing environment remains relatively stable, information is more easily known and understood by external stakeholders. If policies change frequently, the corporate operating environment becomes more complicated and corporate business performance becomes more unpredictable, making it difficult for investors to assess a company’s future profitability and credit default risk. The degree of information asymmetry increases while the cost of violations decreases and the motive for corporate fraud becomes even stronger. Based on the above analysis, economic policy uncertainty increases the potential benefit of corporate violations while reducing their risks and costs. Executives tend to commit fraudulent behaviors when making corporate strategic decisions. Therefore, we propose our first research hypothesis:

H1. 

Ceteris paribus, the probability of corporate fraud will increase when economic policy uncertainty is high.

Macroeconomic environment uncertainty caused by policy changes will affect all participants in the capital market. However, due to individual differences in production technology, business models, management philosophy, financial situations, and other aspects, there are also differences in participants’ ability to respond to policy changes. Companies with different individual characteristics are affected by the impact of uncertainty to different degrees, and their motivations to engage in fraudulent behaviors will not be the same. Therefore, it is necessary to explore the relationship between corporate fraud and economic policy uncertainty from the perspective of corporate heterogeneity. Referring to the existing literature’s research methods [12,17], we will focus on corporate risk tolerance and corporate productive capability to conduct an in-depth analysis of the corporate heterogeneity problem.

We define the state of corporate ownership and the corporate debt leverage (LEV) as a measure of corporate risk tolerance. Compared with non-state-owned companies, state-owned companies mostly belong to the national pillar industry and control the lifeline of overall economic development, which is strongly supported by policy makers [17]. In the face of external environmental impacts, there are differences between the responses of non-state-owned companies and the responses of state-owned companies. For the debt leverage indicator, companies with low debt leverage have good commercial credit and a low default risk, while companies with higher debt leverage have a higher risk of debt crises and financing constraints as well as a lower risk tolerance. In summary, we believe that state-owned companies have both lower debt leverage and a higher risk tolerance, which are less affected by the negative impact of policy changes. On the one hand, state-owned enterprises can survive a crisis by relying on good commercial credit, a relaxed financing environment, and sustained government subsidies [17,18]. On the other hand, when the external business environment deteriorates, companies with high debt will see their market share shrink significantly because of financing constraints and decreasing market demand, while companies with low debt leverage have some buffer space to deal with financing difficulties [12]. Above all, when economic policy uncertainty increases, companies with a higher risk tolerance are less likely to face financial distress and operational crises and often have relatively weak incentives to commit fraud [19].

For corporate productive capability, we select return on assets (ROA) and company size (Size) as the measurement indicators. Companies with a higher ROA indicate higher production and operation capabilities. Different-scale companies have different market shares and output levels. It is undeniable that large-scale companies can achieve economy of scale at a higher production output, but large-scale companies also have higher policy costs [4]. When the external policy environment changes, large-scale companies need a large amount of capital input to adjust their factor allocation and transform their production technology. Moreover, it will take a long time to reach the state of the scale economy again, and small-scale companies are more likely to make flexible adjustments and transformations without a high cost in terms of time and money [20]. In summary, we believe that small-scale companies have both a higher ROA and higher productive capability, which are less affected by the negative impact of policy changes. On the one hand, when the operating environment deteriorates, high-profit companies have the ability to cope with the impact of uncertainty by maintaining stable performance and achieving financial goals. On the other hand, to survive in a fierce market, small-scale companies must be sensitive to the market environment and make rapid adjustments to business management and production. In addition, the cost of production adjustment and technology transformation for small-scale companies is low when the macroeconomic policy changes, that is, the policy cost is low [4]. Accordingly, small-scale companies are more sensitive to changes in the external policy environment and are less affected. Overall, when economic policy uncertainty increases, companies with a higher productive capability are less likely to face financial difficulties and performance losses and often have relatively weak incentives to commit fraud [19]. Based on the above analysis, economic policy uncertainty has a stronger impact on companies with a lower risk tolerance and lower productive capability. The second and third research hypotheses of this paper are as follows:

H2. 

Ceteris paribus, companies with a lower risk tolerance have a higher probability of committing fraud when economic policy uncertainty is high.

H3. 

Ceteris paribus, companies with a lower productive capability have a higher probability of committing fraud when economic policy uncertainty is high.

2.2. The Influence Path of Economic Policy Uncertainty on Corporate Fraud

2.2.1. M&A

Mergers and acquisitions (M&A) are not only a corporate long-term development strategy but also important investment and financing activities. In China, the total transaction value of M&A by listed companies each year accounts for a large proportion of total capital expenditures. Faced with the impact of economic policy uncertainty and a changing economic environment, corporate operation performance is subject to large fluctuations, and M&A activities will also be affected. Specifically, from the perspective of the “uncertainty diversification hypothesis”, Cao, Li, and Liu explain the phenomenon in which cross-border M&A investment activities have increased significantly in the face of higher uncertainty [21]. Cross-border M&A can help companies diversify their operating income, which will reduce the risk of large fluctuations in business performance. If a company is engaged in both domestic and foreign operations, the risk caused by domestic policy uncertainty can be dispersed through cross-border M&A. From the perspective of corporate risk management, Garfinkel and Hankins believe that increased cash flow uncertainty will prompt corporate restructuring and integration and may also cause a wave of M&A in the capital market [22]. Companies regard the M&A strategy as a risk management tool to hedge against the impact of future uncertain cash inflow and reduce the inefficient cost of fixed assets. In addition, for companies attempting to expand in scale, a more uncertain economic environment will not only lead to limited investor attention and a reduction in the price of collective bidding, but also will be more conducive to M&A activities [23]. Therefore, corporate M&A activities increase when economic policy uncertainty is high.

With respect to the relationship between M&A and corporate fraud, Erickson, Heitzman, and Zhang as well as Lo, Ramos, and Rogo found a positive relationship between M&A activities and corporate earnings management [24,25]. After a M&A transaction, the accounting information of consolidated financial statements becomes complicated. This provides an opportunity for senior managers to engage in profit manipulation and illegal disclosure as the detection of financial auditing and fraudulent behaviors becomes increasingly difficult. Moreover, a company’s acquisition strategy can provide benefits to managers who have already engaged in earnings management and profit manipulation. A fraudulent company’s managers can use an acquisition to conceal the company’s true business performance, improve the company’s book value profitability, and appear rich. Further, Wang and Harjoto pointed out that corporate M&A activities are positively correlated with the probability of committing fraudulent behaviors [26,27]. Companies engaged in mergers and acquisitions hope to obtain considerable valuation and earnings. In the process of seeking trading targets, they will attract market attention and improve their bargaining power through corporate fraud, such as window-dressing business performance and or falsified financial indicators. Therefore, companies that have engaged in M&A activities show more significant motivations for violations. In summary, the fourth research hypothesis of this paper is proposed as follows:

H4. 

Ceteris paribus, higher economic policy uncertainty is associated with increased M&A activities, which will increase the possibility of corporate fraud.

2.2.2. Cash Holdings

Analyzing the impact of policy uncertainty on corporate cash holdings is the primary focus of corporate financing and investment. From the perspective of corporate financing, an uncertain environment leads to an unpredictable future business situation and cash flow, which will aggravate the information asymmetry problem between companies and their investors. Worrying about the higher risk of corporate credit default, investors become more cautious when making investment decisions, leading to higher-cost corporate refinancing. Companies will find themselves trapped in financial distress [28]. Furthermore, banks will also be reluctant to lend, leading to a decline in credit size [29]. From the perspective of corporate investment, Knight pointed out that high uncertainty gives firms opportunities to grow and develop, which is the fundamental source of their profits [30]. Barllan and Strange proposed the growth option theory, which argues that investing under economic policy uncertainty is similar to exercising a call option [31]. Because there is a time lag between corporate investment and income realization, the intrinsic value does not equal the current value of the investment project but depends on the growth opportunities and value-added benefits that the project may generate in the future. This increase in uncertainty will bring potential excess returns to investors [32] while potential investment losses are limited, especially for the Internet industry, emerging technologies, and R&D investments [17]. When bad results appear, companies can choose to terminate their investments and stop losses immediately. Once an investment is successful, the future benefits are considerable. Therefore, when economic policy uncertainty increases, companies will face financing constraints and there are still many choices for growth-oriented investment, which will decrease cash holdings.

Corporate cash holding is one factor that affects the probability of corporate fraud. Schrand and Zechman compared the cash flow levels of fraudulent companies and non-fraudulent companies in the study and found that fraudulent companies have lower cash flows [33]. The relationship between corporate cash holding level and corporate fraud can be analyzed from two aspects. On the one hand, when corporate cash and cash equivalents are insufficient, it is difficult to maintain business operations and production, not to mention technology investment in research and development. Managers may attract investors’ attention by whitewashing performance, falsifying disclosures, and engaging in other illegal means of obtaining financial support to protect a company’s normal business activities. On the other hand, when a company’s cash assets are small, indicating that capital expenditure is relatively large or excessive investment has occurred, there are hidden dangers of large fluctuations in future capital flows [27], which will trigger shareholders’ interests. Relevant doubts and dissatisfaction and increased negative expectations for the company’s performance all create the risk of a crash in the company’s stock price. Managers will be more inclined to cover up and hide the actual situation of corporate capital investment. In summary, the fifth research hypothesis of this paper is proposed as follows:

H5. 

Ceteris paribus, higher economic policy uncertainty is associated with lower corporate cash holdings, which will increase the possibility of corporate fraud.

2.2.3. Stock Price Volatility

According to the risk premium theory, investors are rational and risk-averse. Faced with higher risks and uncertainties, investors will seek more income compensation. Since higher economic policy uncertainty brings about a larger fluctuation in the company’s future operating performance, it will be more difficult for investors to make decisions based on historical financial data and current information. The risks of a stock price crash and corporate default increase, and risk-averse investors expect higher returns as compensation for risk taking, resulting in sharp stock price volatility. Anderson, Ghysels, and Juergens constructed an uncertainty analysis model and studied policy uncertainty as an important asset pricing factor, pointing out that investors show more significant risk aversion characteristics to seek a greater premium when policy uncertainty is relatively high [34]. Therefore, rising economic policy uncertainty will increase the volatility of stock prices [1].

Stock issue plays an important role in corporate financing for business production and development, and the stability of stock prices can guarantee continuous and stable financial support for a company’s future activities. Large fluctuations in stock prices will also lead to large fluctuations in the future financing situation of the company, which is not conducive to the company’s normal development and strategic planning. For this reason, we cannot neglect the depressive effect of stock price volatility on corporate output performance. Wang classifies investment failure and operating loss as the triggers for corporate fraudulent behaviors, which means that stock price volatility is positively correlated with corporate fraud [26]. In contrast, according to the efficient markets hypothesis (EMH), stock price information is a direct reflection of a company’s operating conditions. Abnormal fluctuations in stock prices can be measured by the company’s stock volatility, reflecting the company’s possible litigation risks, and a sharp fall in stock prices may be the result of insider trading [35]. Wang, Winton, and Yu found that shareholders may even prosecute companies that have experienced large negative returns and high volatility due to dissatisfaction with their investment losses [36]. In addition, Lo, Ramos, and Rogo believe that stock price volatility is positively correlated with earnings management [25]. Kuang and Lee and Harjoto also verified that stock price volatility increases the likelihood of violations [27,37]. In summary, the sixth research hypothesis of this paper is proposed as follows:

H6. 

Ceteris paribus, higher economic policy uncertainty is associated with higher stock price volatility, which will increase the possibility of corporate fraud.

2.2.4. Institutional Shareholdings

According to Francis, Hasan, and Zhu, both the uncertainty of future economic policy and the heterogeneity of corporate investment and capital allocation will affect the shareholding strategies of institutional investors [38]. The turmoil caused by political uncertainty leads to constraints on future operating cash flow and financing cash flow. Institutional investors are more experienced in acquiring and processing information than individual investors and are more sensitive to macroeconomic policy uncertainty and business performance fluctuations. To reduce the negative effects of policy uncertainty on investment income, institutional investors will rebalance their investment portfolios and reduce their equity holdings to a moderate level [39]. Therefore, when economic policy uncertainty rises, institutional equity holdings decline.

Institutional investors are the major shareholders who hold a large proportion of shares in a company. Company performance and stock price volatility are closely related to their economic interests [40]. Institutional investors are more willing to learn about corporate strategies, supervise managers’ behaviors, and strengthen the internal and external mechanisms of corporate governance [41]. Based on these facts, Agrawal and Mandelker proposed the “active monitoring hypothesis” of institutional investors [42]. Compared with average small and medium shareholders, institutional investors can better exercise the rights of major shareholders and play a significant role in corporate internal governance and supervision. They actively participate in the company’s business management and planning decisions and urge management to disclose business and financial information in an accurate, timely, and fair manner. Hartzell and Starks provide strong empirical evidence that institutional investors play a supervisory role in executive compensation contracts [43]. Institutional investors can directly influence corporate business activities by exercising the rights of major shareholders and exerting pressure on company management under the pretext of the mass selling of stocks, thus achieving an indirect impact on management [44]. As a result, companies with a large proportion of institutional holdings can effectively avoid the abuse of power by senior executives [45]. Hartzell and Starks proved that institutional investors effectively play a supervisory role in easing agency problems between shareholders and top managers [43]. The disclosure of earnings information is timelier and more transparent, which reduces the degree of accounting earnings management and profit manipulation. Lu, Zhu, and Hu pointed out that the supervision of institutional investors can improve the performance of companies, reduce the possibility of losses, and inhibit the occurrence of corporate fraud [19]. Institutional holding is negatively related to a company’s motives for violations. In summary, the seventh research hypothesis of this paper is proposed as follows:

H7. 

Ceteris paribus, higher economic policy uncertainty is associated with fewer shareholdings of institutional investors, which will increase the possibility of corporate fraud.

3. Research Methodology

3.1. Construction of Empirical Models

The research design of this paper is mainly divided into two parts to test the above research hypothesis—the direct effect of economic policy uncertainty on corporate fraud and the mediating effect of economic policy uncertainty on corporate fraud—through four channels (M&A, cash holdings, stock price volatility, and institutional shareholdings). The hypothesis test model consists of a main explanatory variable and four mediation variables belonging to the multiple mediation model [46]. This model includes both the direct effect of the independent variable on the dependent variable and the mediating effect on the dependent variable through the mediating variable. We judge the existence of four influence paths according to the mediation effect test process proposed by Wen and Ye [47]. First, we construct the total effect equation of economic policy uncertainty affecting the corporate fraud tendency, which explains the relationship between the main explanatory variable and the dependent variable. Second, we construct the dynamic adjustment equation of corporate M&A, cash holdings, stock price volatility, and institutional shareholdings under the impact of economic policy uncertainty, which explains the relationship between the main explanatory variable and the mediating variables. Third, we construct the function to explore the effect of economic policy uncertainty on corporate fraud, with the consideration of the possible influence of mediating variables (M&A, cash holdings, stock price volatility, and institutional shareholdings) at the same time. The specific forms of models are as follows:

3.1.1. Direct Effect Model

In order to test the direct effect, we establish the model as followed:

$$\mathit{P}\left({\mathit{Fruad}}_{\mathit{i},\mathit{t}}=1\right)={\gamma }_0+{\gamma }_1{\mathit{EPU}}_{\mathit{i},\mathit{t}-1}+{\displaystyle \sum _{\mathit{j}=2}^{\mathit{n}}{\gamma }_{\mathit{j}}}{\mathit{Control}\mathit{Variables}}_{\mathit{i},\mathit{t}-1}+{\mu }_{i,industry}+{\mu }_{i,area}+{\epsilon }_{\mathrm{i},t-1}$$

(1)

When investigating the occurrence of corporate fraud, we consider the impact of ex ante factors so that all explanatory variables and control variables are treated with one lag phase. ${\mu }_{i,industry},{\mu }_{i,area}$ represent fixed effects at the industry and province level, respectively.

It is worth noting that the explained variable corporate fraud (Fraud) is a binary dummy variable. We often choose the linear probability model (LPM), logit model, or probit model for the estimation of binary dependent variables. Wooldridge studied the binary selection problem, pointing out that regardless of the ordinary least square estimate (OLS) of the LPM method or the maximum likelihood estimate (MLE) of the multivariate probability ratio regression model (logit and probit), the conclusion is consistent [48]. The coefficients of the same variable in different models are identical and statistically significant. Furthermore, Mood pointed out that LPM method has significant advantages and persuasiveness when comparing coefficients among nested models and complex forms [49]. Therefore, in our empirical analysis, the linear probability model is used to compare the coefficients and then the probit model is used to test again for supplementation and robustness.

3.1.2. Mediating Effect Model

In order to investigate the mediating effect, we construct the following three-step regression method and equations are as followed:

$${\mathit{Fruad}}_{i,t}={\alpha }_0+{\beta }_1{\mathit{EPU}}_{i,t-1}+{\displaystyle \sum _{j=2}^n{\alpha }_j}{\mathit{Control}\mathit{Variables}}_{i,t-1}+{\mu }_{i,industry}+{\mu }_{i,area}+{\epsilon }_{1,i,t-1}$$

(2)

$$M_{i,t-1}={\alpha }_0+{\beta }_2{\mathit{EPU}}_{i,t-1}+{\displaystyle \sum _{j=2}^n{\alpha }_j}{\mathit{Control}\mathit{Variables}}_{i,t-1}+{\mu }_{i,industry}+{\mu }_{i,area}+{\epsilon }_{2,i,t-1}$$

(3)

$${\mathit{Fruad}}_{i,t}={\alpha }_0+{\beta }_3{\mathit{EPU}}_{i,t-1}+{\beta }_4{M}_{i,t-1}+{\displaystyle \sum _{j=2}^n{\alpha }_j}{\mathit{Control}\mathit{Variables}}_{i,t-1}+{\mu }_{i,industry}+{\mu }_{i,area}+{\epsilon }_{3,i,t-1}$$

(4)

EPU (Economic Policy Uncertainty Index) is the main explanatory variable. M represents mediator variables, including corporate mergers and acquisitions (M&A), cash holdings (Cash), stock price volatility (VOL), and institutional shareholdings (In_Investor). Fraud is a dependent variable that indicates that the listed company has committed fraud and has been detected. We test the mediating effect by examining the significance of the coefficients ${\beta }_1,{\beta }_2,{\beta }_3,{\beta }_4$. If the mediation effect hypothesis is established, it is necessary to ensure that the coefficients ${\beta }_1,{\beta }_2,{\beta }_4$ are significant. Moreover, the more rigorous requirement is to ensure that there is no adjustment effect between the main explanatory variables and the mediating variables.

It should be noted that the test of the mediating effect is generally directed to a linear model, and the independent variables, moderating variables, and dependent variables are continuous variables. In our study, the independent variable (Fraud) is a dummy variable, while the independent variable (EPU) and the mediating variables (M&A, Cash, VOL, In_Investor) are continuous variables. If we perform the nonlinear model in Equation (2) and the linear model in Equation (3), the coefficient estimation methods of the linear models and the nonlinear models are different, and it is difficult to compare them directly [47]. To ensure that the coefficients are persuasive and easily understood, we follow the method of Chen and Fan [50]. First, we perform an OLS regression on the dependent variable (Fraud) through the LPM method, then we perform an OLS regression in Equations (3) and (4), and finally, we determine the interpretation power of mediating variables according to the mediation effect test procedure. In addition, we use the probit model for the dummy variable (Fraud) regression in Equations (2) and (4) to ensure the accuracy and robustness of the mediating effect test results.

3.2. Variable Selection and Indicator Construction

3.2.1. Independent Variable

China’s Economic Policy Uncertainty Index (EPU) was measured by Baker, Bloom, and Davis [51], who constructed a scaled frequency count of articles about policy-related economic uncertainty in the South China Morning Post (SCMP), Hong Kong’s most influential English-language newspaper. As China’s Economic Policy Uncertainty Index is measured monthly, in our empirical analysis, we convert it from monthly to annually by calculating the annual average.

3.2.2. Characteristic Indicator

The test of corporate heterogeneity is carried out through the two aspects of corporate risk tolerance and corporate productive capability. Combined with the theoretical analysis in the previous section, we define four corporate characteristic indicators for the heterogeneity test. Indicators representing corporate risk tolerance are measured by state of ownership (State) and debt leverage (LEV). Indicators representing corporate productive ability are measured by return on assets (ROA) and company total asset (Size).

3.2.3. Control Variables

Referring to the relevant study, we choose the following factors that influence the possibility of corporate fraud as control variables. To begin with, state-owned companies (State) and large-scale companies (Size) are able to implement a strong internal governance system, which will reduce the risk of fraudulent behaviors. Companies with higher leverage ratio (LEV) and lower profitability ratio (ROA) are more easily trapped into financial difficulties and poor business conditions. As a result, they have the higher risk of fraud [40]. The longer a company goes public (Firmage), the more exposed it is to risk, so the risk of fraud is relatively high [25]. When a CEO also serves as the company general manager (Duality), the power of decision making, strategy execution, and supervision are all given to one person. Since the excessive power of management will weaken the corporate internal control mechanism, the abuse of power or even illegal behaviors by CEO may occur [37]. Next, we select the company research and development expenditure (R&D) and corporate social responsibility (CSR) to measure other types of corporate investment [27,52,53]. Meanwhile, we control for the auditors’ reputation using a dummy variable equal to 1 if the auditor is one of the Big Four auditors (PwC, DTT, KPMG, and EY) and for the type of audit opinion to reflect auditors with any unqualified opinions with unnecessary language or adverse opinions (Au_Opinion). Lower-level auditor and negative auditor opinions are always corelated higher probability of fraud. Additionally, we use the market value of a company divided by its assets’ replacement cost to calculate Tobin’s Q value. Companies with Tobin’s Q value are supposed to have higher tendency of fraud [54]. The definitions and source of each variable are listed in

Table 1. Data Description and Sources.

Variable	Description and Sources
State	The ownership of the listed company. State-owned company is defined as equal to one and non-state-owned company as equal to zero.
LEV	Debt leverage ratio defined as the ratio of total debts divided by total assets.
ROA	Return on asset defined as the company’s EBIT to total assets.
Size	Natural log value of the company’s total assets.
Firmage	The number of years since the company’s IPO.
Duality	Indicator that equals one if the company’s chairman also serves as the general manager and zero otherwise.
R&D	Corporate R&D expenditure divided by total assets.
CSR	Indicator that equals one if the company discloses its CSR report and zero otherwise.
Big4	A dummy variable that equals 1 if the auditor is one of the Big Four and zero otherwise.
Au_Opinion	A dummy variable that equals 1 if the auditor issued an unqualified opinion with explanatory language or an adverse opinion and zero otherwise.
TobinsQ	The market value of a company divided by its asset replacement cost.

.

3.3. Statistical Analysis of Sample Data

We conduct the empirical analysis at the firm level based on the corporate violation database of the China Stock Market and Accounting Research Data Services (CSMAR). We select companies with Shanghai and Shenzhen A-shares from 2007 to 2022 as our research objects. To ensure the consistency and reliability of the financial data of each sample, we remove the samples of ST stocks and the delisted stocks and exclude the stocks of banks, security companies, and any other financial institutions whose financial indicators do not reveal their business conditions. With respect to limiting extreme values in the statistical data to reduce the effect of possibly spurious outliers, we conduct 99% winsorization on the continuous variables. Finally, we manually match data from CSMAR and corporate financial data from the Wind database. As a result, we construct a sample of the 33,943 annual observations, which consists of 5740 fraud cases. We perform the regression analyses in STATA 16.0.

3.3.1. Descriptive Statistical Analysis

Table 2. Annual and Regional Distribution of Corporate Fraud.

Year	Amount	Proportion
2008	89	8.01%
2009	102	8.13%
2010	115	8.73%
2011	117	8.42%
2012	195	11.34%
2013	332	16.27%
2014	385	17.63%
2015	335	15.73%
2016	364	16.47%
2017	477	19.54%
2018	545	20.55%
2019	614	19.41%
2020	669	20.80%
2021	617	18.46%
2022	784	20.75%

summarizes the distribution of fraud cases in different regions and different years.

Table 3. Corporate Fraud in each Province.

Province	Amount	Province	Amount	Province	Amount
Guangdong	1006	Henan	147	Jiangxi	74
Zhejiang	649	Liaoning	116	Neimeng	59
Jiangsu	539	Chongqing	108	Yunnan	55
Beijing	407	Tianjin	99	Gansu	55
shanghai	337	Hebei	98	Heilongjiang	52
Shandong	332	Guangxi	88	Guizhou	39
Fujian	232	Shanxi	87	Ningxia	34
Sichuan	215	Jilin	86	Xizang	33
Hubei	194	Xinjiang	80	Qinghai	25
Hunan	171	Shanxi	80
Anhui	168	Hainan	75

reports the distribution of fraud cases in provinces in China. It can be seen that the number of corporate violations is increasing year by year, and there are significant differences between regions and provinces. The number of violations in the eastern region is the highest, and the number of violations in the central region is relatively small. The number of violations is relatively high in economically developed provinces and relatively low in economically underdeveloped, remote provinces.

Referring to the “Securities Law” and listing rules of the Shanghai Stock Exchange and the Shenzhen Stock Exchange, corporate fraud can be subdivided into nine categories: fictitious profits, false records, postponed disclosures, major omissions, false disclosures, company assets misappropriation, insider trading, illegal trading of stocks, and noncompliance guarantees. The proportion of each type of fraud is shown in

Table 4. Types of Corporate Fraud.

Types of Fraud	Proportion
Fictitious Profits	0.28%
False Records	26.86%
Postponed Disclosures	48.56%
Major Omissions	29.91%
False Disclosure	5.30%
Fraudulent Listing	0.03%
Violation of Capital Contribution	0.00%
Unauthorized Change in Use of Capital	2.09%
Company Assets Misappropriation	6.40%
Insider trading	3.54%
Illegal Trading of Stocks	34.07%
Stock Price Manipulation	0.26%
Noncompliance Guarantees	3.15%
General Accounting Mishandling	10.67%

. In general, companies with false records, postponed disclosures, major omissions, and illegal trading of stocks accounted for more than 25% of companies, which indicates that there are many forms of fraud in listed companies in China. However, we cannot find a dominant fraud type, and a fraudulent case often involves multiple types of fraud, making supervision and investigation quite difficult. The results in Table 4 also account for the reasons that there are so many cases of fraud among listed companies and that fraud cannot be completely eradicated in the Chinese capital market.

In

Table 5. Descriptive Statistical Analysis.

Variable	N	Mean	Median	SD	Min	Max	Sample Mean Difference
Fraud	33,943	0.169	0	0.375	0	1	Fraud = 0	Fraud = 1	Mean Diff
EPU	33,943	0.202	0.129	0.109	0.050	0.390	0.198	0.223	−0.025 *
State	33,943	0.406	0	0.491	0	1	0.432	0.283	0.149 *
LEV	33,943	0.438	0.435	0.205	0.057	0.895	0.435	0.456	−0.022 *
ROA	33,943	0.036	0.036	0.063	−0.267	0.195	0.04	0.016	0.024 *
Size	33,943	22.180	22.000	1.295	19.810	26.190	22.207	22.079	0.128 *
Firmage	33,943	10.330	9	7.082	1	29	10.263	10.641	−0.378 *
Duality	33,943	0.256	0	0.437	0	1	0.247	0.299	−0.052 *
R&D	33,943	0.017	0.013	0.020	0	0.098	0.017	0.019	−0.002 *
CSR	33,943	0.262	0	0.440	0	1	0.272	0.214	0.058 *
Big4	33,943	0.062	0	0.242	0	1	0.068	0.033	0.036 *
Au_Opinion	33,943	0.029	0	0.169	0	1	0.019	0.082	−0.064 *
TobinsQ	33,943	2.069	1.635	1.346	0.860	8.783	2.046	2.181	−0.135 *

* indicates at least 10% significance level.

, a descriptive statistical analysis is performed on each variable in the empirical analysis. We also conduct a sample mean difference test between the violation group (Fraud = 1) and the non-violation group (Fraud = 0). The results show that there are significant differences in the Economic Policy Uncertainty Index between the violation group and the non-violation group, and the value of economic policy uncertainty is higher in the fraud commission group. In addition, each control variable also shows significant mean value differences between the two groups at the 10% significance level.

3.3.2. Correlation Analysis

Before the empirical analysis, the Spearman correlation test and the Pearson correlation test between the variables are also needed. The results are shown in

Table 6. Correlation Coefficients.

	Fraud	EPU	State	LEV	ROA	Size	R&D	CSR	Firmage	Dual	Big4	Au Opinion	TobinsQ
Fraud	1
EPU	0.085 *	1
State	−0.114 *	−0.159 *	1
LEV	0.040 *	−0.066 *	0.279 *	1
ROA	−0.141 *	−0.037 *	−0.067 *	−0.357 *	1
Size	−0.037 *	0.129 *	0.306 *	0.459 *	0.026 *	1
Firmage	0.020 *	0.100 *	0.412 *	0.307 *	−0.151 *	0.355 *	1
Duality	0.044 *	0.098 *	−0.303 *	−0.137 *	0.025 *	−0.145 *	−0.215 *
RD	0.035 *	0.254 *	−0.276 *	−0.263 *	0.116 *	−0.162 *	−0.256 *	0.172 *	1
CSR	−0.049 *	0.051 *	0.201 *	0.130 *	0.067 *	0.453 *	0.213 *	−0.078 *	−0.016 *	1
Big4	−0.055 *	0.001	0.128 *	0.096 *	0.044 *	0.348 *	0.064 *	−0.056 *	−0.041 *	0.206 *	1
Au_Opinion	0.142 *	0.028 *	−0.041 *	0.121 *	−0.291 *	−0.054 *	0.077 *	−0.003	−0.039 *	−0.047 *	−0.020 *	1
TobinsQ	0.037 *	−0.070 *	−0.158 *	−0.272 *	0.144 *	−0.384 *	−0.064 *	0.079 *	0.213 *	−0.088 *	−0.082 *	0.079 *	1

* indicates at least 10% significance level.

. It can be seen that economic policy uncertainty is positively associated with corporate fraud, which is consistent with our theoretical analysis. In addition, the correlation coefficient between explanatory variables is relatively small, and the variance inflation factor (Mean VIF) is 1.39, much less than 10, indicating that there is no severe multicollinearity problem between explanatory variables, which is in line with our preliminary expectations.

4. Empirical Results

4.1. The Direct Effects of EPU on Corporate Fraud

4.1.1. Benchmark Regression

Column (1) in

Table 7. Heterogeneity Test based on Corporate Risk Tolerance.

	LPM	LPM	LPM	LPM	LPM	Probit	Probit	Probit	Probit	Probit
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
	ALL	State = 1	State = 0	High LEV	Low LEV	ALL	State = 1	State = 0	High LEV	Low LEV
EPU	0.185 ***	0.158 ***	0.200 ***	0.230 ***	0.134 ***	0.748 ***	0.713 ***	0.819 ***	0.876 ***	0.603 ***
	(0.020)	(0.041)	(0.018)	(0.033)	(0.028)	(0.066)	(0.066)	(0.188)	(0.102)	(0.109)
State	−0.088 ***			−0.082 ***	−0.092 ***	−0.376 ***			−0.326 ***	−0.434 ***
	(0.006)			(0.010)	(0.007)	(0.024)			(0.036)	(0.032)
LEV	0.069 ***	0.092 ***	0.061 ***			0.309 ***	0.493 ***	0.232 ***
	(0.012)	(0.021)	(0.019)			(0.050)	(0.116)	(0.071)
ROA	−0.622 ***	−0.333 ***	−0.730 ***	−0.721 ***	−0.499 ***	−2.112 ***	−1.403 ***	−2.338 ***	−2.232 ***	−1.902 ***
	(0.060)	(0.062)	(0.072)	(0.060)	(0.087)	(0.209)	(0.269)	(0.241)	(0.182)	(0.335)
Size	0.004	0.000	0.009 ***	0.003	0.005 *	0.013	−0.005	0.034 ***	0.006	0.015
	(0.003)	(0.003)	(0.003)	(0.004)	(0.002)	(0.011)	(0.017)	(0.011)	(0.016)	(0.011)
Firmage	0.003 ***	0.001 **	0.003 ***	0.001	0.004 ***	0.011 ***	0.008 **	0.012 ***	0.006 *	0.017 ***
	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.002)	(0.003)	(0.003)	(0.003)	(0.003)
Duality	0.011 **	0.032 **	0.006	0.025 ***	0.002	0.046 **	0.154 ***	0.024	0.094 ***	0.009
	(0.005)	(0.012)	(0.005)	(0.008)	(0.009)	(0.020)	(0.053)	(0.020)	(0.026)	(0.040)
RD	0.131	0.658 **	−0.084	0.290	0.050	0.726	3.632 **	−0.326	1.293	0.445
	(0.165)	(0.288)	(0.168)	(0.332)	(0.129)	(0.707)	(1.468)	(0.664)	(1.258)	(0.609)
CSR	−0.022 ***	−0.011	−0.030 ***	−0.017 **	−0.028 ***	−0.094 ***	−0.054	−0.115 ***	−0.066 **	−0.130 ***
	(0.004)	(0.007)	(0.005)	(0.007)	(0.005)	(0.021)	(0.038)	(0.016)	(0.031)	(0.026)
Big4	−0.045 ***	−0.041 ***	−0.044 ***	−0.050 ***	−0.036 ***	−0.263 ***	−0.322 ***	−0.171 ***	−0.264 ***	−0.244 ***
	(0.007)	(0.010)	(0.014)	(0.009)	(0.011)	(0.035)	(0.080)	(0.058)	(0.050)	(0.068)
Au_Opinion	0.202 ***	0.140 ***	0.225 ***	0.201 ***	0.193 ***	0.549 ***	0.451 ***	0.602 ***	0.550 ***	0.560 ***
	(0.016)	(0.040)	(0.012)	(0.016)	(0.017)	(0.045)	(0.119)	(0.034)	(0.043)	(0.052)
TobinsQ	0.009 ***	0.011 ***	0.008 ***	0.009 ***	0.008 ***	0.031 ***	0.046 ***	0.026 ***	0.027 ***	0.032 ***
	(0.001)	(0.003)	(0.001)	(0.002)	(0.001)	(0.005)	(0.013)	(0.005)	(0.008)	(0.006)
_cons	0.056	0.017	−0.021	0.098	0.045	−1.414 ***	−1.707 ***	−1.736 ***	−1.189 ***	−1.448 ***
	(0.052)	(0.065)	(0.054)	(0.075)	(0.046)	(0.218)	(0.341)	(0.195)	(0.301)	(0.209)
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	33,943	13,795	20,148	16,891	17,052	33,943	13,795	20,148	16,891	17,052
Pseudo R2						0.062	0.051	0.053	0.070	0.052
Adj. R2	0.059	0.034	0.057	0.069	0.042

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. The specification contains a full set of industry (${\mu }_{i,industry}$) and province (${\mu }_{i,area}$) fixed effects. Standard errors (in parentheses) are clustered at the industry level. *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Columns (1)–(5) present the group regression results of the LPM, and columns (6)–(10) present the group regression results of the probit model.

gives the full sample regression results of the LPM model estimation. The coefficient of the EPU is 0.185, and the significance reaches the 1% level. The probability of corporate fraud increases significantly under the impact of economic policy uncertainty, which confirms that our research hypothesis H1 is rational. Meanwhile, the estimation results of the control variables are consistent with the results of prior literatures [19,37,54]. State-owned companies are less likely to violate the rules and the coefficient is −0.088. The coefficient of debt leverage (LEV) is 0.069, indicating that a company is prone to operational dilemmas and more inclined to commit fraud when its total liabilities account for a higher proportion of its total assets. Company size and the time to market since IPO are positively related to the possibility of corporate fraud. Duality reflects the situation in which the chairman also serves as the general manager so that the power of decision making, strategy execution, and supervision are all given to one person. The coefficient of duality is 0.011, which implies that it is easy for a corporate manager with excessive power to engage in misconduct or fraud. We control for the auditors’ reputation whether the auditor is one of the Big Four (Big 4) audit units (Deloitte, Ernst & Young (EY), KPMG, and PricewaterhouseCoopers (PwC)), and we also control for the type of audit opinion to reflect any unqualified opinions with explanatory language or adverse opinions from the auditors (Au_Opinion) and obtain the same results as Hu et al. (2019) [53]. In addition, consistent with Harjoto [27] and Wang [26], the return on assets (ROA) is significantly negatively correlated with the probability of fraud, with coefficients of −0.622. The coefficient of corporate social responsibility (CSR) is −0.022, indicating that companies that actively assume corporate social responsibility have lower incentives to commit fraud. In column (6), we provide the full sample regression result of the probit model based on the MLE estimation. The main explanatory variable (EPU) has a coefficient of 0.748 and reaches a significance level of 1% without reducing the explanatory power of the control variables, supporting the H1 hypothesis.

4.1.2. Heterogeneity Test Based on Corporate Risk Tolerance

The discrepancy in corporate risk tolerance among different companies is mainly reflected in corporate ownership and the corporate leverage ratio. In general, state-owned companies and companies with low debt leverage have a relatively high-risk resistance capacity, while non-state-owned companies and companies with high debt leverage have a relatively low-risk resistance capacity. This character determines the difference in corporate risk-averse behaviors under the impact of economic policy uncertainty. In Table 7, we divide the overall sample into four groups with different risk tolerance levels according to their ownership and debt leverage and conduct the benchmark regression accordingly.

We report the regression result of state-owned companies in column (2), and the EPU coefficient is 0.158. The regression coefficient of the EPU in the non-state-owned sample is 0.200 (column (3) of Table 7), indicating that non-state-owned companies have a higher likelihood of committing fraud under the impact of policy uncertainty. The regression results of different leverage ratios are shown in columns (4) and (5); the EPU coefficient in the high-leverage sample group is 0.230, while the EPU coefficient in the low-leverage sample group is 0.134, indicating that companies with a high debt ratio have a higher likelihood of committing fraud under the impact of policy uncertainty. From the above analysis, we can confirm that companies with a lower risk tolerance have a higher probability of violations when economic policy uncertainty is high, and hypothesis H2 is proven. In addition, the group regression results of the probit model (columns (7)–(10) of Table 7) are consistent with the results in columns (2)–(5).

4.1.3. Heterogeneity Test Based on Productive Capability

The discrepancy in corporate productive capability among different companies is mainly reflected in the profitability ratio (ROA) and asset size of each company. Small-scale, high-ROA companies have a relatively high productive capability, and large-scale, low-ROA companies have a relatively low productive capability. This character determines the difference in corporate risk-averse behaviors under the impact of economic policy uncertainty. In

Table 8. Heterogeneity Test Based on Productive Capability.

	LPM	LPM	LPM	LPM	LPM	Probit	Probit	Probit	Probit	Probit
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
	All	High ROA	Low ROA	Large Scale	Small Scale	All	High ROA	Low ROA	Large Scale	Small Scale
EPU	0.185 ***	0.088 ***	0.282 ***	0.195 ***	0.170 ***	0.748 ***	0.423 ***	1.039 ***	0.832 ***	0.647 ***
	(0.020)	(0.019)	(0.037)	(0.030)	(0.035)	(0.066)	(0.084)	(0.114)	(0.109)	(0.130)
State	−0.088 ***	−0.082 ***	−0.092 ***	−0.084 ***	−0.086 ***	−0.376 ***	−0.408 ***	−0.353 ***	−0.369 ***	−0.352 ***
	(0.006)	(0.006)	(0.010)	(0.011)	(0.007)	(0.024)	(0.029)	(0.034)	(0.044)	(0.030)
LEV	0.069 ***	0.074 ***	0.073 ***	0.049 **	0.078 ***	0.309 ***	0.350 ***	0.271 ***	0.281 ***	0.313 ***
	(0.012)	(0.015)	(0.018)	(0.020)	(0.016)	(0.050)	(0.068)	(0.065)	(0.087)	(0.064)
ROA	−0.622 ***			−0.616 ***	−0.623 ***	−2.112 ***			−2.160 ***	−2.086 ***
	(0.060)			(0.052)	(0.070)	(0.209)			(0.219)	(0.228)
Size	0.004	−0.000	0.009 *			0.013	−0.006	0.031 *
	(0.003)	(0.003)	(0.005)			(0.011)	(0.013)	(0.018)
Firmage	0.003 ***	0.003 ***	0.002 *	0.001	0.003 ***	0.011 ***	0.014 ***	0.007 *	0.007 *	0.013 ***
	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.002)	(0.002)	(0.003)	(0.004)	(0.002)
Duality	0.011 **	0.006	0.018 **	0.023 ***	0.004	0.046 **	0.025	0.068 ***	0.091 ***	0.019
	(0.005)	(0.007)	(0.007)	(0.008)	(0.004)	(0.020)	(0.031)	(0.024)	(0.031)	(0.016)
RD	0.131	0.178	0.206	0.248	0.067	0.726	0.948 *	0.955	1.361	0.286
	(0.165)	(0.120)	(0.341)	(0.359)	(0.105)	(0.707)	(0.547)	(1.236)	(1.499)	(0.425)
CSR	−0.022 ***	−0.011 **	−0.034 ***	−0.020 ***	−0.020 ***	−0.094 ***	−0.045 **	−0.136 ***	−0.086 ***	−0.079 ***
	(0.004)	(0.004)	(0.008)	(0.007)	(0.006)	(0.021)	(0.019)	(0.032)	(0.030)	(0.027)
Big4	−0.045 ***	−0.029 **	−0.069 ***	−0.033 ***	−0.071 ***	−0.263 ***	−0.181 ***	−0.346 ***	−0.201 ***	−0.375 ***
	(0.007)	(0.010)	(0.010)	(0.006)	(0.021)	(0.035)	(0.064)	(0.048)	(0.038)	(0.114)
Au_Opinion	0.202 ***	0.192 ***	0.195 ***	0.213 ***	0.189 ***	0.549 ***	0.622 ***	0.543 ***	0.593 ***	0.511 ***
	(0.016)	(0.035)	(0.019)	(0.012)	(0.022)	(0.045)	(0.100)	(0.049)	(0.042)	(0.059)
TobinsQ	0.009 ***	0.006 ***	0.011 ***	0.004	0.012 ***	0.031 ***	0.024 ***	0.037 ***	0.012	0.042 ***
	(0.001)	(0.001)	(0.004)	(0.003)	(0.002)	(0.005)	(0.005)	(0.013)	(0.012)	(0.006)
_cons	0.056	0.140 **	−0.049	0.087	−0.218 *	−1.414 ***	−1.063 ***	−1.786 ***	−1.337 ***	−2.596 ***
	(0.052)	(0.054)	(0.091)	(0.060)	(0.111)	(0.218)	(0.266)	(0.337)	(0.294)	(0.440)
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	33,943	16,950	16,974	17,055	16,872	33,943	16,950	16,974	17,055	16,872
Pseudo R2						0.062	0.036	0.074	0.069	0.061
Adj. R2	0.059	0.025	0.075	0.059	0.060

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. The specification contains a full set of industry (${\mu }_{i,industry}$) and province (${\mu }_{i,area}$) fixed effects. Standard errors (in parentheses) are clustered at the industry level. *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Columns (1)–(5) present the group regression results of the LPM, and columns (6)–(10) present the group regression results of the probit model.

, we divide the overall sample into four groups with different productive capabilities according to their ROA and total asset size and conduct the benchmark regression accordingly. In columns (2) and (3), the EPU coefficient in the sample group with a higher return on assets is 0.088, while the EPU coefficient in the sample group with a lower return on assets is 0.282. Under this impact, companies with lower profitability tend to be more likely to commit fraud than companies with higher profitability.

In columns (4) and (5), the EPU coefficient in the sample group of large-scale companies is 0.195, while the EPU coefficient in the sample group of small-scale companies is 0.170, indicating that under the impact of policy uncertainty, large-scale companies’ tendency to operate illegally is higher than that of small-scale companies. Due to the large scale of production and operation, large-scale companies often passively make technical factor adjustments and management mode changes when economic policies change, and they need to spend money. In contrast, small-scale companies appear more flexible and proactive in the face of uncertain economic policy shocks. This explains the larger company size and the positive impact of deterministic economic policies on the company’s violating motives. In summary, the regression results of the LPM model in columns (2) to (4) can explain how hypothesis H3 is established. In addition, the probit model grouping regression results according to columns (7)–(10) are consistent with the results of columns (2)–(4), indicating that the company’s production and management capabilities are lower when the positive relationship between economic policy uncertainty and company violation probability is stronger.

From the above analysis, we can confirm that companies with a lower productive capability have a higher probability of violations when economic policy uncertainty is high, and hypothesis H3 is proven. In addition, the group regression results of the probit model in columns (7)–(10) of Table 8 are consistent with the results in columns (2)–(5).

4.2. The Mediating Effects of EPU on Corporate Fraud

4.2.1. M&A

A dummy variable that equals 1 if the company announced the merger and acquisition event from the current year and that equals zero otherwise. We present the results of the mediating effect test among economic policy, M&A, and corporate fraud in

Table 9. The Relationship among EPU, M&A, and Corporate Fraud.

	LPM	OLS	LPM	Probit	OLS	Probit
	(1)	(2)	(3)	(4)	(5)	(6)
	Fraud	M&A	Fraud	Fraud	M&A	Fraud
EPU	0.185 ***	0.417 ***	0.180 ***	0.748 ***	1.068 ***	0.726 ***
	(0.020)	(0.045)	(0.019)	(0.066)	(0.115)	(0.065)
M&A			0.013 ***			0.054 ***
			(0.004)			(0.016)
_cons	0.056	0.338 ***	0.052	−1.414 ***	−0.417 ***	−1.434 ***
	(0.052)	(0.051)	(0.052)	(0.218)	(0.133)	(0.218)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes	Yes	Yes
N	33,943	33,943	33,943	33,943	33,943	33,943
Pseudo R2				0.062	0.018	0.062
Adj. R2	0.059	0.022	0.059

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. The specification contains a full set of industry (${\mu }_{i,industry}$) and province (${\mu }_{i,area}$) fixed effects. Standard errors (in parentheses) are clustered at the industry level. *** denote statistical significance at the 1% level. Columns (1), (3) show the regression results of the LPM for the dependent variable (Fraud). Columns (4), (6) show the regression results of the probit model for the dependent variable (Fraud).

. To make a more intuitive comparison with the subsequent test results, we continue to present the results of the direct effect that economic policy uncertainty has on the possibility of committing fraud in column (1) of Table 9. Column (2) shows the regression result of the main explanatory variable (EPU) for the mediating variable (M&A). The EPU coefficient is 0.417 and reaches a significance level of 1%, indicating that corporate M&A activities frequently occur when economic policy uncertainty increases. Next, we consider the dependent variable (Fraud), the independent variable (EPU), and the mediating variable (M&A) simultaneously and present the regression results in Column (3). It can be seen that the coefficient of M&A is significantly positive at the 1% significance level, indicating the significant positive relationship between M&A and the possibility of corporate fraud. The estimated coefficient of economic policy uncertainty is still significant and the absolute value is decreased. Accordingly, we can draw the conclusion that M&A is an influence path for economic policy uncertainty affecting corporate fraud. Meanwhile, according to the regression results of the probit model for the dependent variable (Fraud) in columns (4)–(6) of Table 9, we can still reach a consistent conclusion. Higher economic policy uncertainty is associated with more frequent M&A activities, which will raise the possibility of violations. The results shown below in Table 9 confirm hypothesis H4.

4.2.2. Cash Holdings

Cash holdings are measured by total cash and cash equivalents divided by total assets. We present the results of the mediating effect test on economic policy, cash holdings, and corporate fraud in

Table 10. The Relationship among EPU, Cash Holdings, and Corporate Fraud.

	LPM	OLS	LPM	Probit	OLS	Probit
	(1)	(2)	(3)	(4)	(5)	(6)
	Fraud	Cash	Fraud	Fraud	Cash	Fraud
EPU	0.185 ***	−0.064 **	0.176 ***	0.748 ***	−0.064 **	0.707 ***
	(0.020)	(0.023)	(0.019)	(0.066)	(0.023)	(0.067)
Cash			−0.141 ***			−0.636 ***
			(0.023)			(0.118)
_cons	0.056	0.343 ***	0.105 *	−1.414 ***	0.343 ***	−1.200 ***
	(0.052)	(0.049)	(0.052)	(0.218)	(0.049)	(0.214)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes	Yes	Yes
N	33,943	33,943	33,943	33,943	33,946	33,943
Pseudo R2				0.062		0.063
Adj. R2	0.059	0.246	0.060		0.246

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. The specification contains a full set of industry (${\mu }_{i,industry}$) and province (${\mu }_{i,area}$) fixed effects. Standard errors (in parentheses) are clustered at the industry level. *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Columns (1),(3) show the regression results of the LPM for the dependent variable (Fraud). Columns (4), (6) show the regression results of the probit model for the dependent variable (Fraud).

. The result of the basic regression of Equation (2) is shown in column (1) for further comparison. Column (2) shows the regression result of the main explanatory variable (EPU) for the mediating variable (Cash). The EPU coefficient is −0.064 and reaches a significance level of 5%, indicating that the corporate cash holding level decreases when economic policy uncertainty increases. Next, we consider the dependent variable (Fraud), independent variable (EPU), and mediating variable (Cash) simultaneously and present the regression results in column (3). It can be seen that the coefficient of cash is significantly negative at the 1% significance level, indicating that companies with a low level of cash holdings are associated with a higher likelihood of corporate fraud. The EPU is still significantly positive and the absolute value decreases, so we can conclude that corporate cash holdings are an influence path for economic policy uncertainty affecting corporate fraud. According to the regression results of the probit model for the dependent variable (Fraud) in columns (4)–(6) of Table 10, we can still reach a consistent conclusion. Higher economic policy uncertainty is associated with a lower corporate cash holding level, which will create the possibility of violations. Hypothesis H5 is proven.

4.2.3. Stock Price Volatility

Stock price volatility is measured by the standard deviation of daily stock returns of the year. We test hypothesis H6 and present the results of the relationship among economic policy, stock price volatility, and corporate fraud in

Table 11. The Relationship among EPU, Stock Price Volativity, and Corporate Fraud.

	LPM	OLS	LPM	Probit	OLS	Probit
	(1)	(2)	(3)	(4)	(5)	(6)
	Fraud	VOL	Fraud	Fraud	VOL	Fraud
EPU	0.185 ***	0.092 **	0.184 ***	0.748 ***	0.092 **	0.746 ***
	(0.020)	(0.040)	(0.020)	(0.066)	(0.040)	(0.067)
VOL			0.011 **			0.042 **
			(0.004)			(0.018)
_cons	0.056	−0.693 ***	0.064	−1.414 ***	−0.693 ***	−1.384 ***
	(0.052)	(0.085)	(0.051)	(0.218)	(0.085)	(0.217)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes	Yes	Yes
N	33,943	33,943	33,943	33,943	33,946	33,943
Pseudo R2				0.062		0.062
Adj. R2	0.059	0.013	0.059		0.013

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. The specification contains a full set of industry (${\mu }_{i,industry}$) and province (${\mu }_{i,area}$) fixed effects. Standard errors (in parentheses) are clustered at the industry level. ** and *** denote statistical significance at the 5% and 1% levels, respectively. Columns (1),(3) show the regression results of the LPM for the dependent variable (Fraud). Columns (4), (6) show the regression results of the probit model for the dependent variable (Fraud).

. The result of the basic regression of Equation (2) is still shown in column (1) for further comparison. Column (2) shows the regression result of the main explanatory variable (EPU) to the mediating variable (VOL). The EPU coefficient is 0.092 and reaches a significance level of 5%, which shows that increased economic policy uncertainty increases the volatility of corporate stock prices. Next, we consider the dependent variable (Fraud), independent variable (EPU), and mediating variable (VOL) simultaneously and present the regression results in column (3). It can be seen that the coefficient of VOL is significantly positive at the 5% significance level, indicating that corporate stock price volatility is positively associated with a higher likelihood of fraudulent behaviors. The estimated value of the EPU is still significantly positive, so we can conclude that corporate stock price volatility is an influence path for economic policy uncertainty affecting corporate fraud. Meanwhile, according to the regression results of the probit model for the dependent variable (Fraud) in columns (4)–(6) of Table 11, we can still reach a consistent conclusion. Higher economic policy uncertainty is associated with higher stock price volatility, which will create the possibility of violations. Hypothesis H6 is proven.

4.2.4. Institutional Shareholdings

Institutional shareholdings are measured by the proportion of total equity held by institutional investors. Similarly, we test hypothesis H8 and present the results of the relationship among economic policy, institutional shareholdings, and corporate fraud in Table 11. The result of the basic regression of Equation (2) is shown in column (1) for further comparison. Column (2) shows the regression result of the main explanatory variable (EPU) on the mediating variable (In_Investor). The EPU coefficient is −0.197 and reaches a significance level of 1%, indicating the negative impact of economic policy uncertainty on institutional equity holdings. Institutional investors can be better aware of changes in the macro economic environment and company’s business performance so that they are able to adjust their equity holding level timely and effectively. Next, we consider the dependent variable (Fraud), independent variable (EPU), and mediating variable (In_Investor) simultaneously and present the regression results in column (3). It can be seen that the coefficient of the mediating variable (In_Investor) is significantly negative at the 1% significance level, which confirms the negative relationship between institutional shareholdings and the likelihood of fraudulent behaviors. The estimated value of the EPU is still significantly positive, so we can conclude that institutional equity holding is an influence path for economic policy uncertainty affecting corporate fraud. Furthermore, according to the regression results of the probit model for the dependent variable (Fraud) in columns (4)–(6) of

Table 12. The Relationship among EPU, Institutional Shareholdings, and Corporate Fraud.

	LPM	OLS	LPM	Probit	OLS	Probit
	(1)	(2)	(3)	(4)	(5)	(6)
	Fraud	In_Investor	Fraud	Fraud	In_Investor	Fraud
EPU	0.185 ***	−0.197 ***	0.169 ***	0.748 ***	−0.197 ***	0.671 ***
	(0.020)	(0.024)	(0.018)	(0.066)	(0.024)	(0.063)
In_Investor			−0.084 ***			−0.364 ***
			(0.006)			(0.024)
_cons	0.056	−0.973 ***	−0.026	−1.414 ***	−0.973 ***	−1.778 ***
	(0.052)	(0.051)	(0.050)	(0.218)	(0.051)	(0.212)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes	Yes	Yes
N	33,943	33,943	33,943	33,943	33,946	33,943
Pseudo R2				0.062		0.064
Adj. R2	0.059	0.348	0.061		0.348

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. The specification contains a full set of industry (${\mu }_{i,industry}$) and province (${\mu }_{i,area}$) fixed effects. Standard errors (in parentheses) are clustered at the industry level. *** denote statistical significance at the 1% level. Columns (1), (3) show the regression results of the LPM for the dependent variable (Fraud). Columns (4), (6) show the regression results of the probit model for the dependent variable (Fraud).

, we can still reach a consistent conclusion. Higher economic policy uncertainty is associated with lower institutional shareholdings, which will create the possibility of violations. The seventh hypothesis of our research, H7, is confirmed.

4.2.5. Mediating Effect Judgement

Finally, according to the mediating effect test procedure [47,55], we infer the existence of the mediating effects of the four mediating variables analyzed above. The summary results in

Table 13. Mediating Effect Judgment.

Step 1	Step 2	Step 3	Step 4
Test the direct effect of the main explanatory variable (EPU) on the dependent variable (Fraud) β1.	Test the effect of the main explanatory variable (EPU) on the mediator variable (M) β2, and the effect of the mediator variable (M) on the dependent variable (Fraud) β4, after controlling the influence of EPU.	After controlling the mediation variable (M), test the effect of the main explanatory variable (EPU) on the dependent variable (Fraud) β3.	Test whether β2 × β4 and β3 have the same sign, and draw a conclusion.
Empirical Results of Hypothesis H4: EPU, M&A, and Corporate Fraud
β1 = 0.185 *** There is a significant, direct effect of EPU.	β2 = 0.417 *** β4 = 0.013 ** Both β2 and β4 are significant, indicating that the indirect relationship among EPU, M&A, and Corporate Fraud is significant.	β3 = 0.180 *** There is a significant mediating effect and there may be another mediating variable.	β2 × β4 is the same sign as β3, and it can be inferred that M&A is one of the mediating variables accounting for the influence of EPU on Fraud. The degree of influence is β2 × β4/β3 = 0.030.
Empirical Results of Hypothesis H5: EPU, Cash Holdings, and Corporate Fraud
β1 = 0.185 *** There is a significant, direct effect of EPU.	β2 = −0.064 ** β4 = −0.141 ** Both β2 and β4 are significant, indicating that the indirect relationship among EPU, Cash Holdings, and Corporate Fraud is significant.	β3 = 0.176 *** There is a significant mediating effect and there may be another mediating variable.	β2 × β4 is the same sign as β3, and it can be inferred that Cash is one of the mediating variables accounting for the influence of EPU on Fraud. The degree of influence is β2 × β4/β3 = 0.051.
Empirical Results of Hypothesis H6: EPU, Stock Price Volatility, and Corporate Fraud
β1 = 0.185 *** There is a significant, direct effect of EPU.	β2 = 0.092 *** β4 = 0.011 *** Both β2 and β4 are significant, indicating that the indirect relationship among EPU, Stock Price Volatility, and Corporate Fraud is significant.	β3 = 0.184 *** There is a significant mediating effect and there may be another mediating variable.	β2 × β4 is the same sign as β3, and it can be inferred that VOL is one of the mediating variables accounting for the influence of EPU on Fraud. The degree of influence is β2 × β4/β3 = 0.0055.
Empirical Results of Hypothesis H7: EPU, Institutional shareholdings, and Corporate Fraud
β1 = 0.185 *** There is a significant, direct effect of EPU.	β2 = −0.197 *** β4 = −0.084 *** Both β2 and β4 are significant, indicating that the indirect relationship among EPU, Institutional Shareholdings, and Corporate Fraud is significant.	β3 = 0.169 *** There is a significant mediating effect and there may be another mediating variable.	β2 × β4 is the same sign as β3, and it can be inferred that In_Investor is one of the mediating variables accounting for the influence of EPU on Fraud. The degree of influence is β2 × β4/β3 = 0.098.

**, *** denote statistical significance at the 5% and 1% levels respectively.

show that the increase in economic policy uncertainty significantly increases the possibility of corporate fraud, and the influence channel contains four mediation variables: M&A, cash holdings, stock price volatility, and institutional shareholdings. The results in Table 13 still support hypotheses H4–H7. In addition, we can explain the result of Table 13 in another way. Because of the existence of the four mediating variables, we are able to account for the reasons why economic policy uncertainty leads to a higher likelihood of corporate fraud based on four aspects. When the macroeconomic policy environment develops in a manner that is not clear or consistent, the business and financial status of the companies will be affected. To avoid the potential risks caused by economic policy uncertainty, companies and investors make corresponding dynamic adjustments. As policy uncertainty worsens, corporate M&A increase and cash holdings decrease. Meanwhile, when investors reduce their holdings and seek higher risk returns, resulting in stock price volatility, institutional shareholdings decline. Changes in the company’s operating and financial indicators directly lead to an increased risk of corporate violations, creating a large hidden danger for the capital market.

5. Robust Test

5.1. Endogenous Discussion

Considering that there may be endogeneity problems caused by missing variables and two-way causality in the empirical analysis process, we construct three instrumental variables to address the estimation bias caused by endogeneity. The first instrumental variable (EPU1) is measured by the one-phase lag value of the EPU index [17], which can effectively eliminate the reverse causality problem. The second instrumental variable (EPU2) is replaced by a dummy variable, which is based on the mean value of the EPU index during the year. The index above the standard is recorded as 1 and the index below the standard is recorded as 0 [12]. In addition, with reference to the research methods of Wang, Zhang, and Bao [56], there is a significant spillover effect of US government policy, which affects economic policy uncertainty and spread to other countries. For the third instrumental variable (EPU3), we use the United States’ Economic Policy Uncertainty Index to measure China’s economic policy uncertainty. These three instrumental variables (EPU1, EPU2, and EPU3) are used to replace the independent variable (EPU) in the benchmark regression, and the results are shown in

Table 14. Instrumental Variable Estimation.

	(1)	(2)	(3)	(4)	(5)	(6)
	LPM	LPM	LPM	Probit	Probit	Probit
EPU1	0.188 ***			0.748 ***
	(0.026)			(0.105)
EPU2		0.195 ***			0.800 ***
		(0.021)			(0.087)
EPU3			0.064 **			0.250 **
			(0.028)			(0.113)
_cons	0.085	0.059	0.063	−1.310 ***	−1.402 ***	−1.403 ***
	(0.057)	(0.052)	(0.057)	(0.237)	(0.221)	(0.237)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes	Yes	Yes
N	33,943	33,943	33,943	33,943	33,943	33,943
Pseudo R2				0.075	0.075	0.075
Adj. R2	0.059	0.059	0.058

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. The specification contains a full set of industry (${\mu }_{i,industry}$) and province (${\mu }_{i,area}$) fixed effects. Standard errors (in parentheses) are clustered at the industry level. ** and *** denote statistical significance at the 5% and 1% levels, respectively. Columns (1) to (3) show the 2SLS regression results of the dependent variable (Fraud). Columns (4) to (6) show the ivprobit regression results of the dependent variable (Fraud).

. The results in columns (1)–(3) show that no matter which instrumental variable is used, the coefficient of economic policy uncertainty is significantly positive, consistent with the outcomes of Table 7.

Furthermore, we select EPU3 to replace the EPU and conduct the robustness test of the corporate heterogeneity analysis and the mediation effect analysis. From the results shown in

Table 15. Heterogeneity Analysis Based on EPU3 index.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	ALL	State = 1	State = 0	High LEV	Low LEV	High ROA	Low ROA	Large Scale	Small Scale
EPU3	0.808 ***	0.794 ***	0.810 ***	0.944 ***	0.614 ***	0.399 ***	1.301 ***	1.062 ***	0.630 ***
	(0.082)	(0.073)	(0.235)	(0.113)	(0.152)	(0.103)	(0.146)	(0.153)	(0.130)
State	−0.384 ***			−0.330 ***	−0.439 ***	−0.400 ***	−0.375 ***	−0.373 ***	−0.369 ***
	(0.024)			(0.035)	(0.032)	(0.029)	(0.034)	(0.025)	(0.037)
LEV	0.293 ***	0.461 ***	0.223 ***				0.408 ***	0.356 ***	0.246 ***
	(0.050)	(0.115)	(0.073)				(0.070)	(0.067)	(0.079)
ROA	−2.143 ***	−1.449 ***	−2.365 ***	−2.391 ***	−1.935 ***			−1.918 ***	−2.330 ***
	(0.211)	(0.343)	(0.241)	(0.172)	(0.342)			(0.218)	(0.223)
Size	0.017	0.000	0.038 ***	0.015	0.021 **	0.014	0.014
	(0.011)	(0.005)	(0.010)	(0.015)	(0.010)	(0.013)	(0.019)
_cons	−1.491 ***	−1.818 *	−1.809 ***	−1.270 ***	−1.545 ***	−1.430 ***	−1.521 ***	−1.289 ***	−0.995 ***
	(0.217)	(0.950)	(0.192)	(0.296)	(0.194)	(0.275)	(0.369)	(0.041)	(0.079)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	33,943	16,950	20,148	16,891	17,052	16,950	16,974	17,055	16,872
Pseudo R2	0.061	0.049	0.053	0.069	0.051	0.034	0.067	0.063	0.067

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. The specification contains a full set of industry (${\mu }_{i,industry}$) and province (${\mu }_{i,area}$) fixed effects. Standard errors (in parentheses) are clustered at industry level. ** and *** denote statistical significance at the 5% and 1% levels, respectively. Columns (1) to (9) show the regression results of the probit model for the dependent variable (Fraud).

and

Table 16. Mediating Effect Analysis Based on EPU3.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	Fraud	M&A	Fraud	Cash	Fraud	VOL	Fraud	In_Investor	Fraud
EPU3	0.808 ***	0.683 ***	0.772 ***	−0.053 *	0.771 ***	0.054	0.808 ***	−0.240 ***	0.711 ***
	(0.082)	(0.047)	(0.081)	(0.027)	(0.079)	(0.050)	(0.082)	(0.029)	(0.077)
M&A			0.055 ***
			(0.016)
Cash					−0.657 ***
					(0.121)
VOL							0.044 **
							(0.018)
In_Investor									−0.373 ***
									(0.024)
_cons	−1.491 ***	0.317 ***	−1.509 ***	0.352 ***	−1.264 ***	−0.707 ***	−1.458 ***	−0.956 ***	−1.856 ***
	(0.217)	(0.053)	(0.216)	(0.049)	(0.213)	(0.084)	(0.216)	(0.051)	(0.210)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	33,943	33,943	33,943	33,943	33,943	33,943	33,943	33,943	33,943
Pseudo R2	0.061		0.061		0.063		0.061		0.063
Adj. R2		0.026		0.244		0.013		0.347

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. The specification contains a full set of industry (${\mu }_{i,industry}$) and province (${\mu }_{i,area}$) fixed effects. Standard errors (in parentheses) are clustered at the industry level. *, **, and *** denote statistical significance at the 10%, 5% and 1% levels, respectively. The regression results of the dependent variable (Fraud) are estimated by probit model.

, we find that when economic policy uncertainty increases, companies with a lower risk tolerance and lower productive capacity have a stronger incentive to commit violations. More specifically, the increased EPU leads to higher motives for corporate M&A, decreases corporate cash holdings, increases the volatility of stock prices, and reduces institutional shareholdings consistent with the empirical results in Table 7, Table 8, Table 9, Table 10, Table 11 and Table 12. In column (6) of Table 16, the coefficient of influence of the EPU on stock price volatility does not reach the significance level of 10%, but it can still be seen that when uncertainty increases, stock price volatility increases. Thus, we retain the results of column (6) of Table 16 and continue our robust analysis of the four mediating effects in the next section.

5.2. Re-Calculate EPU Index with Different Method

When studying fraudulent cases of listed companies, we select the annual financial data at the firm level for an empirical analysis. In the Chinese capital market, corporate annual data are more scientific and accurate than quarterly data because the annual financial reports of listed companies need to undergo formal audits to ensure fairness and compliance. Since the China Economic Policy Uncertainty Index is constructed monthly [51], we need to transform the monthly EPU index and obtain the annual EPU data applicable to our empirical test. To reduce the estimation bias of data construction, we re-calculate the EPU index with three different methods for a robust analysis. First, we define EPU4 as the geometric average of the monthly Economic Policy Uncertainty Index to obtain the annual EPU data [12]. Second, we define the annual median value of the EPU index as EPU5 [17]. Third, considering the fact that a higher degree of policy uncertainty has a large impact range and will cause worse economic consequences, we define the annual maximum value of the EPU index as EPU6. The benchmark regression outcomes of three re-measured indicators are summarized in

Table 17. Re-calculate EPU Index with Different Method.

	(1)	(2)	(3)	(4)	(5)	(6)
	Fraud	Fraud	Fraud	Fraud	Fraud	Fraud
EPU4	0.086 ***			0.353 ***
	(0.009)			(0.030)
EPU5		0.163 ***			0.653 ***
		(0.019)			(0.063)
EPU6			0.089 ***			0.354 ***
			(0.011)			(0.036)
_cons	0.074	0.050	0.053	−1.338 ***	−1.440 ***	−1.427 ***
	(0.051)	(0.052)	(0.052)	(0.217)	(0.217)	(0.218)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes	Yes	Yes
N	33,943	33,943	33,943	33,943	33,943	33,943
Pseudo R2				0.062	0.061	0.061
Adj. R2	0.059	0.058	0.058

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. The specification contains a full set of industry (${\mu }_{i,industry}$) and province (${\mu }_{i,area}$) fixed effects. Standard errors (in parentheses) are clustered at the industry level. *** denote statistical significance at the 1% level. Columns (1) to (3) show the regression results of the LPM for the dependent variable (Fraud). Columns (4) to (6) show the regression results of the probit model for the dependent variable (Fraud).

. It is clear that the coefficients of the three EPU indicators are significantly positive, consistent with the conclusions of Table 7.

Next, we use EPU6 as the main explanatory variable to conduct a robustness test for a corporate heterogeneity analysis and mediation effect analysis, and the outcomes are summarized in

Table 18. Heterogeneity analysis based on EPU6.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	ALL	State = 1	State = 0	High LEV	Low LEV	High ROA	Low ROA	Large Scale	Small Scale
EPU6	0.354 ***	0.344 ***	0.368 ***	0.395 ***	0.290 ***	0.160 ***	0.583 ***	0.415 ***	0.335 ***
	(0.036)	(0.035)	(0.101)	(0.053)	(0.059)	(0.045)	(0.054)	(0.059)	(0.069)
State	−0.381 ***			−0.327 ***	−0.436 ***	−0.400 ***	−0.370 ***	−0.374 ***	−0.356 ***
	(0.024)			(0.036)	(0.031)	(0.029)	(0.035)	(0.044)	(0.029)
LEV	0.299 ***	0.475 ***	0.225 ***				0.418 ***	0.294 ***	0.348 ***
	(0.050)	(0.117)	(0.072)				(0.070)	(0.088)	(0.060)
ROA	−2.132 ***	−1.431 ***	−2.356 ***	−2.383 ***	−1.926 ***			−2.145 ***	−2.012 ***
	(0.208)	(0.272)	(0.238)	(0.171)	(0.341)			(0.214)	(0.232)
Size	0.015	−0.002	0.036 ***	0.014	0.019 **	0.014	0.011
	(0.011)	(0.017)	(0.011)	(0.015)	(0.010)	(0.013)	(0.020)
_cons	−1.427 ***	−1.739 ***	−1.748 ***	−1.202 ***	−1.488 ***	−1.408 ***	−1.411 ***	−1.055 ***	−1.167 ***
	(0.218)	(0.340)	(0.194)	(0.298)	(0.194)	(0.275)	(0.370)	(0.111)	(0.032)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	33,943	16,950	20,148	16,891	17,052	16,950	16,974	17,055	16,872
Pseudo R2	0.061	0.050	0.053	0.069	0.051	0.034	0.068	0.068	0.060

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. The specification contains a full set of industry (${\mu }_{i,industry}$) and province (${\mu }_{i,area}$) fixed effects. Standard errors (in parentheses) are clustered at the industry level. ** and *** denote statistical significance at the 5% and 1% levels. Columns (1) to (9) show the regression results of the probit model for the dependent variable (Fraud).

and

Table 19. Mediating Effect Analysis Based on EPU6.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	Fraud	M&A	Fraud	Cash	Fraud	VOL	Fraud	In_Investor	Fraud
EPU6	0.354 ***	0.327 ***	0.337 ***	−0.038 ***	0.329 ***	0.042 *	0.353 ***	−0.100 ***	0.315 ***
	(0.036)	(0.028)	(0.036)	(0.012)	(0.036)	(0.021)	(0.036)	(0.012)	(0.035)
M&A			0.051 ***
			(0.016)
Cash					−0.639 ***
					(0.120)
VOL							0.043 **
							(0.018)
In_Investor									−0.371 ***
									(0.024)
_cons	−1.427 ***	0.384 ***	−1.447 ***	0.341 ***	−1.213 ***	−0.696 ***	−1.396 ***	−0.973 ***	−1.796 ***
	(0.218)	(0.052)	(0.217)	(0.049)	(0.214)	(0.084)	(0.217)	(0.051)	(0.211)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	33,943	33,943	33,943	33,943	33,943	33,943	33,943	33,943	33,943
Pseudo R2	0.061		0.061		0.063		0.061		0.063
Adj. R2		0.030		0.247		0.013		0.348

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. The specification contains a full set of industry (${\mu }_{i,industry}$) and province (${\mu }_{i,area}$) fixed effects. Standard errors (in parentheses) are clustered at the industry level. *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. The regression results of the dependent variable (Fraud) are estimated by probit model.

. From the regression results of the LPM model and the probit model, we can draw the same conclusion in accordance with the research results of Table 7, Table 8, Table 9, Table 10, Table 11 and Table 12 and once again prove hypotheses H2–H7.

5.3. Mediating Effect Test with Different Method

In this paper, we test the four mediation hypotheses with binary dummy variables in the empirical part. We mainly refer to the method of Chen and Fan which deals with the adjustment effect and mediating effect of the binary selection problem [50]. Imai, Keele, and Tingley [57] and Hicks and Tingley [58] proposed a method for identifying and analyzing the mediating causal effects based on the medeff and medsens commands in the STATA software package when the mediating variable or dependent variable is a binary dummy variable. Here, we follow this method to conduct the mediation effect test again. In addition, according to Wooldridge [48], we can also perform “cluster correlation” processing to correct standard errors and eliminate correlation problems within cluster sample groups. As we can see from the outcomes summarized in

Table 20. Mediating Effect Test with Different Method.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	M&A	Fraud	Cash	Fraud	VOL	Fraud	In_Investor	Fraud
EPU	1.072 ***	4.560 ***	−0.024	4.454 ***	0.454 ***	4.462 ***	−0.605 ***	4.281 ***
	(0.099)	(0.533)	(0.058)	(0.521)	(0.018)	(0.549)	(0.072)	(0.539)
M&A		0.084 ***
		(0.018)
Cash				−0.598 ***
				(0.103)
VOL						0.043 **
						(0.020)
In_Investor								−0.278 ***
								(0.023)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	33,943	33,943	33,943	33,943	33,943	33,943	33,943	33,943
Pseudo R2		0.076		0.077		0.076		0.077
Adj. R2	0.150		0.262		0.046		0.364

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. ** and *** denote statistical significance at the 5% and 1% levels, respectively. Standard errors (in parentheses) are clustered at the province and the industry level.

, when economic policy uncertainty increases, corporate M&A activities frequently occur, the level of cash holdings decreases, the stock price fluctuates more severely, and the proportion of institutional shareholdings declines, resulting in a higher likelihood of corporate violations. This indicates that there are four feasible and robust influence paths between economic policy uncertainty and corporate fraud.

6. Further Discussion

Judging from the geographical distribution of listed companies in the Chinese capital market, there are significant differences in the number of listed companies in each province. Listed companies are mainly located in the eastern region, whereas there are relatively few listed companies in the central and western regions. Moreover, because of China’s vast territory, the resource endowments and market development levels of different provinces are different, which results in differences in the level of economic development and the consequence of policy uncertainty. Therefore, it is necessary to further investigate the impact of economic policy uncertainty on corporate fraud considering corporations’ geographic locations and the corresponding economic levels of their locations. We divide the overall sample into two groups: the eastern region and the central and western regions, and we regress the baseline model separately. The results of columns (1)–(2) in

Table 21. Regional Difference in the Impact of EPU on Corporate Fraud.

	(1)	(2)	(3)	(4)
	Eastern Region	Central & Western	High-GDP	Low-GDP
EPU	0.713 ***	0.813 ***	0.488 ***	0.625 ***
	(0.104)	(0.142)	(0.090)	(0.159)
_cons	−1.116 ***	−1.619 ***	−0.699 *	−2.081 ***
	(0.290)	(0.371)	(0.388)	(0.279)
Control Variables	Yes	Yes	Yes	Yes
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes
N	21,169	12,765	16,996	16,947
Pseudo R2	0.069	0.059	0.060	0.076

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. The specification contains a full set of industry (${\mu }_{i,industry}$) and province (${\mu }_{i,area}$) fixed effects. Standard errors (in parentheses) are clustered at the industry level. * and *** denote statistical significance at the 10% and 1% levels, respectively. Columns (1) and (2) are grouped based on geographical distribution, and columns (3) and (4) are grouped based on economic output level.

show that the EPU coefficient is significantly positive in both groups. However, the EPU coefficient of the central and western regions is larger than that of the eastern region, which indicates that the companies in central and western regions have a stronger tendency to commit fraud under the impact of economic policy uncertainty. The economic development level of the central and western regions is still relatively low and thus, companies in those regions lack the ability to adjust their output supply, optimize their resource allocation, and change their business strategies to avoid the risks of policy uncertainty. In addition, most provinces in the central and western regions are relatively remote and lack an effective supervision and inspection mechanism, which provides an opportunity for executives to engage in illegal operations. Put differently, we define the provinces at the top 20% of the total output value (GDP) as the high-GDP group, and the remaining provinces are defined as the low-GDP group. In Table 21, columns (3) and (4) show the benchmark results of two groups, indicating that the EPU coefficient in the low-GDP group is much higher than that in the high-GDP group, confirming the conclusions of this section.

Next, we focus on the industry competition to which the company belongs. We construct the Herfindahl–Hirschman Index (HHI) of each industry according to company operating income and total assets. The construction equation for the HHI index is as follows:

$${\mathit{HHIIndex}}_{i,t}={{\displaystyle \sum _{\mathrm{f}\in {\mathsf{\Omega }}_i}\left(s_{ft}÷S_{it}\right)}}^2={{\displaystyle \sum _{\mathrm{f}\in {\mathsf{\Omega }}_i}shar{e}_{ft}}}^2$$

(5)

where $s_{ft}$ represents operating income (total assets), $S_{it}$ represents the total operating income (total assets) of the whole industry, and $shar{e}_{ft}$ represents the market share of each company. If the value of the HHI index is small, the degree of industry competition is high and vice versa. The industries with an above-average HHI value are classified as the monopolized-industry group, and those with a below-average HHI value are classified as the competitive-industry group. The benchmark regression results for each group are shown in

Table 22. Industrial Difference in the Impact of EPU on Corporate Fraud.

	(1)	(2)	(3)	(4)
	HHI-ASSET	HHI-ASSET	HHI-SALE	HHI-SALE
	Monopolized Industry	Competitive Industry	Monopolized Industry	Competitive Industry
EPU	0.771 ***	0.712 ***	0.820 ***	0.663 ***
	(0.054)	(0.101)	(0.095)	(0.108)
_cons	−2.015 ***	−1.206 ***	−1.862 ***	−1.291 ***
	(0.303)	(0.392)	(0.116)	(0.362)
Control Variables	Yes	Yes	Yes	Yes
${\mu }_{i,industry}$	Yes	Yes	Yes	Yes
${\mu }_{i,area}$	Yes	Yes	Yes	Yes
N	13,289	20,654	15,320	18,623
Pseudo R2	0.055	0.069	0.055	0.070

Each variable corresponds to two rows of results: the first row is the coefficient value, and the second row is the robust standard error. The specification contains a full set of industry (${\mu }_{i,industry}$) and province (${\mu }_{i,area}$) fixed effects. Standard errors (in parentheses) are clustered at the industry level. *** denote statistical significance at the 1% levels. Columns (1) and (2) are grouped based on total assets, and columns (3) and (4) are grouped based on operating income.

, where columns (1) and (2) are grouped based on total assets and columns (3) and (4) are grouped based on operating income. The EPU coefficient of monopolized industry is larger than that of competitive industry, indicating that in the less competitive market, the impact of economic policy uncertainty on increasing corporate violations is stronger. Monopolized companies in the industry occupy the dominant position and have greater market shares. When the economic policy changes, these monopoly companies will bear greater market risks and suffer greater losses. In addition, monopoly companies have strong market power and bargaining power. When faced with policy shocks, companies are prone to use unfair violations to manipulate the market and eliminate a crisis.

In contrast, companies in highly competitive industries are more sensitive to changes in macroeconomic policy. The market competition mechanism of a highly competitive industry encourages companies to actively respond to the impact of uncertainty and flexibly adjust to avoid risk. Thus, when economic policy uncertainty increases, companies in monopolized industries are more likely to commit violations.

7. Conclusions

This paper focuses on listed companies in the Chinese capital market and discusses the relationship between macroeconomic policy uncertainty and corporate fraudulent behaviors. This paper argues that there are both direct and mediating effects of economic policy and constructs multiple mediation models to conduct a regression. The results of this study show that economic policy uncertainty is positively associated with corporate fraud, and this direct effect differs due to corporate heterogeneity. Companies with a lower risk tolerance and productive capacity have a higher likelihood of committing corporate fraud. This is in line with the research conclusions of Shen [10] as well as Gulen and Ion [11], who stated that uncertainty has a relatively negative impact on companies and capital markets. In terms of mediating, we confirm that four mediating variables can explain the relationship between economic policy uncertainty and corporate fraud and clarify that increased economic policy uncertainty results in more M&A activities [22], a decline in corporate cash holdings [28,29], an expansion in stock price volatility [1], and a reduction in institutional shareholdings [38], which increase the probability of corporate fraud. Furthermore, we discuss the impact of economic policy uncertainty on corporate violations from the perspective of regional development and industry competition. Companies located in undeveloped regions and within a monopolized industry are more affected by the impact of macro policy changes, showing a higher possibility of corporate fraud.

We find that only a small number of existing papers consider the impact of economic policy uncertainty on corporate profit manipulation, earnings management, and other irregularities [4,10], but research is lacking on the direct relationship and influence path of policy uncertainty and corporate fraud. Therefore, based on the conclusions of this paper, we can provide policy recommendations for effectively monitoring and preventing corporate violations and correctly guiding companies to cope with crises and avoid risks. The government should publicize and explain the process of introducing new policies so that the market can fully understand its policy direction and the information asymmetry between the government and enterprises is reduced. Enterprises must accurately understand and grasp the direction of policy development and change and minimize the risks of macro policy uncertainty. At the same time, companies should improve their risk tolerance and productive capability, for example, by reducing debt leverage, improving profitability, adapting to economic development trends, adjusting production, and transforming technology. For regulators, when the macro policy environment changes, they should pay close attention to companies in slow-growing regions and monopolized industries, focusing on their M&A activities, cash holdings, stock price volatility, and institutional shareholdings. They should also detect and curb corporate illegal behaviors as soon as possible.

As we all know, in the context of a transition economy, China’s financial market has a relatively high degree of regulation, and the operation and management decisions of market players are highly dependent on national economic policies, which means that the strategic decisions of companies are highly sensitive to the uncertainties of economic policies. Therefore, we naturally think of the following question as one of the important factors affecting enterprise operation: what is the relationship between the external uncertain environment and management behaviors?

This paper specifically answers whether the uncertainty of the economic environment caused by economic policy uncertainty will have an impact on a company’s fraudulent behaviors; through what channels economic policy uncertainty affects corporate fraudulent behaviors; and what policy makers should do to avoid these negative effects. It is not only helpful to reasonably assess the impact and consequences of economic policy uncertainty on business operation and development, but also helpful to analyze and understand corporate fraud from a new perspective, so as to provide empirical evidence for the regulatory authorities to strengthen the regulatory measures, improve the market environment, and promote the healthy development of the capital market. Moreover, the theoretical and empirical evidence obtained from the transition of China’s economy is also applicable to other developing countries. In particular, for countries in the process of economic transformation, unreasonable policy formulation should be avoided to stable companys’ expectations, thus avoiding greater negative impacts. Therefore, it is of great theoretical and practical value to investigate the relationship between economic policy uncertainty and corporate fraud in China’s specific institutional system and economic development stage.

Objectively speaking, this study also has some limitations. First, the perspective of this study only focuses on the economic policy uncertainty in the macro environment, which may ignore the influence of other external uncertainties, such as wars, epidemics, natural disasters, and so on. Secondly, this paper does not describe the considerable heterogeneity of enterprises in a more detailed way, so it cannot explain the difference influences of economic policy uncertainty at the firm level precisely.

Therefore, this study can be further expanded in the following directions. First, one could improve the construction of macro uncertainty indicators, take more external factors affecting enterprise operation and development into account, and obtain more comprehensive evaluation indicators. Additionally, more advanced deep learning algorithms could be used to provide accuracy in the construction of uncertainty indicators. Moreover, one could also conduct a comparative analysis for the different types of uncertainty factors. Lastly, one could construct the index of enterprise uncertainty, so as to discuss the economic correlation and influence mechanism between macro uncertainty and enterprise uncertainty, and further carry out a detailed comparative analysis around the heterogeneity characteristics of enterprises.