Enterprise Risk Management and Earnings Management: Accrual-Based and Real Activities Evidence from Chinese Listed Firms

Abstract

Earnings management undermines financial transparency and threatens long-term corporate sustainability, particularly in emerging markets where principal–agent conflicts remain pronounced. In China’s capital market, performance-based incentives may motivate managers to manipulate reported earnings, thereby impairing investor protection and governance quality. Despite growing interest in enterprise risk management (ERM) as a holistic governance mechanism, empirical evidence on its effectiveness in constraining earnings manipulation remains limited. This paper investigates the governance role of ERM in mitigating both accrual-based earnings management (AEM) and real earnings management (REM) among Chinese listed firms over the period 2019–2024. Using panel regression models, this study examines whether higher ERM engagement is associated with lower levels of earnings manipulation. The results indicate that ERM is significantly and negatively related to both AEM and REM. These findings remain robust to alternative variable definitions, different sample period specifications, interaction analyses between accrual-based and real earnings management, alternative constructions of ERM (including PCA-based measures and exclusion of reporting-related components), and endogeneity tests using industry–year average ERM as a proxy. Further heterogeneity analyses reveal that the constraining effect of ERM on REM is more pronounced in firms audited by non-Big Four auditors, while the effect is weaker in Big Four audited firms. Overall, the evidence suggests that ERM functions as an effective internal governance mechanism that enhances financial reporting quality and supports sustainable corporate performance. This paper contributes to the sustainability and corporate governance literature by providing empirical evidence from an emerging market context and offers practical implications for regulators and corporate decision-makers seeking to strengthen risk governance frameworks.

1. Introduction

Earnings management persists in capital markets characterized by pronounced information asymmetry and principal–agent conflicts. Within such environments, managerial discretion over financial reporting and operational decisions creates opportunities to influence reported performance for private benefit. Empirical evidence documents that accrual-based techniques are frequently employed to simultaneously increase book income and reduce taxable income, particularly in emerging economies such as the BRICS countries (Pipatnarapong, 2020). Manipulation of working capital components to avoid loss recognition has also been identified among listed firms (Tran & Duong, 2020). Executive compensation structures further intensify incentives to manage earnings, as performance-linked remuneration and bonus schemes create pressure to achieve short-term targets (Assenso-Okofo et al., 2021). Market participants may misinterpret manipulated earnings signals, resulting in mispricing and potential welfare losses (Hong & Linh, 2020).

Although external governance mechanisms have received substantial scholarly attention, growing recognition has been given to the role of internal governance systems in constraining opportunistic reporting behavior. Evidence indicates that high-quality internal control systems are negatively associated with both accrual-based earnings management (AEM) and real earnings management (REM) (Gao et al., 2020). Accounting reforms have reduced aggregate accrual manipulation, yet have simultaneously induced substitution toward real earnings management (Ho et al., 2015), suggesting that managers may shift between reporting channels when constraints are imposed on specific techniques. These findings underscore the importance of comprehensive governance frameworks capable of addressing multiple dimensions of earnings manipulation. Nevertheless, the effectiveness of particular internal governance arrangements remains insufficiently examined.

ERM represents an integrated framework encompassing risk identification, assessment, monitoring, and internal control processes. By coordinating risk-related information flows and strengthening oversight structures, ERM has been associated with enhanced organizational resilience and improved decision quality (Mottoh & Sutrisno, 2020). Empirical studies in foreign markets suggest that high-quality ERM can restrain real earnings management (Oreshile, 2025) and improve forecast accuracy through more effective processing of internal and external information (Li et al., 2025). Despite this emerging body of evidence, limited research has simultaneously assessed the influence of ERM on both accrual-based and real earnings management within a unified framework, particularly in emerging markets.

The Chinese capital market provides a distinctive institutional setting for examining this relationship. As a developing market, China exhibits comparatively weaker investor protection, evolving regulatory enforcement, and concentrated ownership structures, conditions that may exacerbate information asymmetry and agency conflicts (Yang et al., 2022). Strengthening internal and external governance mechanisms has therefore been regarded as essential for mitigating opportunistic earnings management (Bui, 2024; Y. Wang et al., 2022). However, systematic empirical evidence regarding the governance role of ERM in constraining different forms of earnings manipulation among Chinese listed firms remains limited. Whether ERM operates as a substantive internal governance mechanism or merely as a symbolic compliance arrangement in this context thus constitutes an open empirical question.

Against this backdrop, this study examines the impact of ERM on both AEM and REM among Chinese listed firms. Specifically, the study pursues two research objectives: (1) to assess whether ERM constrains AEM, and (2) to determine whether ERM constrains REM. By simultaneously analyzing both forms of earnings manipulation, the study provides a more comprehensive assessment of ERM’s governance effectiveness and its role in promoting financial reporting quality and corporate sustainability.

The contribution of this research is threefold. First, this paper provides evidence on the governance role of Enterprise Risk Management (ERM) in constraining both accrual-based earnings management (AEM) and real earnings management (REM) within an emerging market context. Unlike prior studies such as (Gao et al., 2020) and (Oreshile, 2025), which primarily examine either internal control systems or real earnings management in isolation, this paper jointly examines accrual-based and real activities manipulation within a unified empirical framework. By jointly examining AEM and REM, the study extends the corporate governance and risk management literature beyond single-dimensional analyses and offers additional evidence on earnings manipulation under ERM mechanisms. In particular, the evidence suggests that ERM operates as a broader governance structure that simultaneously constrains multiple forms of managerial discretion.

Second, by situating the analysis in the Chinese institutional context, this study responds to recent calls for more evidence from emerging economies characterized by weaker investor protection and evolving regulatory environments. The findings enrich the understanding of how ERM functions under institutional conditions where external monitoring mechanisms may be less effective, thereby highlighting the relevance of internal governance systems in mitigating earnings manipulation in these environments.

Third, this paper provides practical implications for regulators and corporate boards seeking to strengthen sustainable corporate governance. The results highlight the importance of substantive ERM implementation over symbolic compliance, in order to improve overall financial reporting quality. For boards of directors, the evidence underscores the value of high-quality ERM as an internal governance mechanism that constrains opportunistic managerial behavior and enhances reporting credibility. Strengthening ERM practices may therefore support more effective risk oversight and enhance earnings quality.

In addition, by explicitly accounting for sample-period robustness and potential COVID-19 shocks, this paper ensures that the documented ERM–earnings management relationship is not driven by temporary institutional disruptions, thereby strengthening the generalizability of the findings. Future research may further explore the mechanisms through which ERM exerts its governance effects and extend the analysis to other emerging markets.

2. Literature Review

2.1. Earnings Management and Governance: Theoretical Foundations and Evidence

Earnings management has been extensively examined within the framework of agency theory, which posits that the separation of ownership and control creates incentives for managers to pursue private benefits at the expense of shareholders (Healy & Wahlen, 1999; Jensen & Meckling, 1976). Information asymmetry between managers and external stakeholders further facilitates opportunistic earnings manipulation, as accounting accruals involve managerial judgment and are less directly observable (Dechow, 1994; Shah & Wan, 2024). When compensation, career prospects, or regulatory thresholds are linked to accounting performance, managers may adjust reported earnings to achieve desired outcomes. In this context, agency theory explains earnings management by emphasizing how weak monitoring mechanisms create conditions for managerial opportunism.

Earnings management generally takes two forms: accrual-based earnings management (AEM) and real earnings management (REM). AEM relies on accounting discretion in accrual estimation, whereas REM involves altering actual business operations, such as production levels, discretionary expenditures, or sales timing (Achleitner et al., 2014). Although accrual manipulation may be more directly observable through audit scrutiny, REM is embedded in operational decisions and may therefore be more difficult to detect. Consequently, managers may substitute toward real activities manipulation when accrual discretion is constrained (Bui, 2024). This distinction implies that effective governance mechanisms must address both accounting discretion and operational opportunism.

However, the relationship between AEM and REM remains theoretically and empirically inconclusive in prior literature. A large body of research supports a substitution perspective, suggesting that managers shift between accrual-based and real activities manipulation depending on relative costs and regulatory constraints (Zang, 2012). In contrast, other studies suggest that firms may jointly adjust accrual-based and real earnings management depending on institutional constraints and managerial incentives, indicating that the two mechanisms are not necessarily mutually exclusive (D. A. Cohen et al., 2008). In addition, evidence from real activities manipulation literature highlights that managers may use operational decisions to achieve earnings targets and signal future performance, although this stream of research does not explicitly examine the interaction between accrual-based and real earnings management (Gunny, 2010). This evidence suggests that the relationship between AEM and REM varies across institutional environments, governance quality, and managerial incentives.

Existing research highlights the role of corporate governance mechanisms, such as board independence, audit quality, internal control effectiveness, and debt monitoring, in mitigating earnings manipulation. Empirical evidence from China shows that firms facing regulatory pressure, delisting risk, or concentrated ownership structures often engage in strategic earnings management to meet short-term objectives (Chi & Gooda, 2024; Jing et al., 2024; Ma & Ma, 2024). At the same time, stronger internal control systems and enhanced monitoring are negatively associated with both AEM and REM (Gao et al., 2020). However, much of the literature focuses on external monitoring mechanisms. In emerging markets characterized by evolving regulatory enforcement and ownership concentration, reliance on external governance alone may be insufficient to constrain managerial discretion.

From the perspective of the resource-based view (RBV), firms obtain sustainable competitive advantages by effectively acquiring, integrating, and deploying valuable internal resources and capabilities (Barney, 1991). RBV emphasizes that organizational capabilities embedded in managerial processes and governance structures can enhance firm efficiency and strategic decision-making. In this context, ERM can be regarded as an important organizational capability that enables firms to systematically identify, assess, and manage risks across the organization (Hayne & Free, 2014; Xu et al., 2024). Importantly, this capability-based perspective complements agency theory. It suggests that ERM not only constrains opportunistic behavior through enhanced monitoring, but also improves firms’ internal information processing and decision-making efficiency, thereby reducing the conditions that give rise to such behavior. By integrating risk information into strategic decision-making and strengthening internal control mechanisms, ERM reduces information asymmetry between managers and stakeholders (Li et al., 2025).

Although prior studies have examined the relationship between ERM and earnings quality, several gaps remain. Existing research primarily focuses on accrual-based earnings management, leaving limited evidence on its impact on real earnings management (REM), which is operationally embedded and potentially more harmful to firm value (Bui, 2024; Oreshile, 2025). Moreover, most studies are conducted in developed economies, providing insufficient insights into ERM’s governance role in emerging markets, where investor protection is weak and ownership is often highly concentrated (Chi & Gooda, 2024; Jing et al., 2024). In addition, the variation in ERM effectiveness across organizational and financial environments, such as firm size, leverage, and risk exposure, remains underexplored (H. Wang et al., 2023; Zinyoro & Aziakpono, 2024). More importantly, prior literature typically adopts a single theoretical perspective, either focusing on monitoring mechanisms under agency theory or emphasizing organizational capabilities under RBV, without explicitly integrating these perspectives to explain how ERM simultaneously constrains managerial opportunism and enhances organizational efficiency. These limitations highlight the necessity for a comprehensive analysis of ERM’s governance function, particularly its ability to constrain both accrual-based and real earnings management in emerging market settings. Accordingly, this study develops a theoretical perspective in which ERM operates through both governance (monitoring) and capability (resource allocation and information integration) channels, thereby providing a more comprehensive perspective on its role in influencing earnings management.

2.2. Enterprise Risk Management as an Internal Governance Mechanism

ERM represents an integrated, firm-wide risk governance framework that coordinates risk identification, assessment, and monitoring across organizational levels (Demidenko & McNutt, 2010; Gupta, 2008). From an integrated theoretical perspective combining agency theory and the resource-based view, ERM serves not only as a governance mechanism but also as a strategic organizational capability. From an agency perspective, ERM enhances transparency and information integration within firms, reduces information asymmetry, and improves information processing capacity and resource allocation efficiency. From an RBV perspective, ERM enhances resource allocation efficiency, allocate resources efficiently, and align operational decisions with long-term strategic objectives. By embedding risk oversight into operational and financial decision-making processes, ERM may limit managerial discretion and reduce incentives for opportunistic reporting behavior.

Importantly, ERM differs from traditional governance mechanisms in several important respects. While conventional governance tools, such as board oversight, audit quality, and internal control systems, primarily operate as ex post monitoring mechanisms that discipline managerial behavior after decisions are made. In contrast, ERM functions as an ex ante, organization-wide system that integrates risk considerations directly into managerial decision-making processes (Crawford & Jabbour, 2024; Monazzam & Crawford, 2024). This forward-looking and integrative nature enables ERM to influence not only financial reporting choices but also underlying operational decisions (J. Cohen et al., 2017; López, 2021). Unlike fragmented governance structures that address specific aspects of oversight, ERM facilitates the integration and dissemination of risk-related information across functional units, thereby reducing information silos (Khan et al., 2016; Saeidi et al., 2019). As a result, ERM is well suited to constrain both accrual-based earnings management, which is primarily accounting-driven, and real earnings management, which is embedded in operational activities. This distinction is particularly important in emerging market contexts, where reliance on external governance may be limited and integrated internal systems play a more substantive role in shaping managerial incentives and behavior.

Prior studies suggest that ERM improves the scope, integration, and aggregation of accounting information systems, thereby enhancing reporting quality and limiting earnings manipulation (Abu Afifa & Saleh, 2021; Oreshile, 2025). International evidence generally supports the governance role of ERM. For example, research in Taiwan finds that firms with stronger ERM frameworks are less likely to engage in opportunistic accrual manipulation, particularly in equity financing settings (T. S. Wang et al., 2018). Evidence from sub-Saharan Africa indicates that high-quality ERM mitigates real earnings management, especially in large or financially constrained firms (Oreshile, 2025). Other studies document improvements in forecast quality and earnings reliability following ERM adoption, particularly under conditions of high volatility or environmental uncertainty (Bamigboye et al., 2024; Li et al., 2025).

Taken together, agency theory suggests that managerial discretion, reinforced by information asymmetry, provides incentives for both accrual-based and real earnings manipulation. Although external governance mechanisms may partially constrain such behavior, their effectiveness may be limited in emerging markets characterized by evolving regulatory enforcement and concentrated ownership structures. As an integrated internal governance framework, ERM enhances transparency, strengthens internal control quality, and embeds risk monitoring into operational and financial decision-making processes. By simultaneously functioning as a monitoring mechanism (agency theory) and a capability-enhancing system (RBV), ERM not only constrains opportunistic accounting choices but also reduces incentives for real activities manipulation by improving operational efficiency. These characteristics indicate that ERM may reduce managerial discretion in both accounting estimation and operational decisions, addressing the research gaps identified above.

3. Research Hypothesis

3.1. The Impact of ERM on Accrual-Based Earnings Management

ERM strengthens internal monitoring and control mechanisms and improves the integration of risk information into financial reporting processes. By enhancing internal control effectiveness and reducing information asymmetry, ERM is expected to constrain managerial discretion in accrual estimation and reporting decisions (Abu Afifa & Saleh, 2021; Demidenko & McNutt, 2010; Gupta, 2008; Oreshile, 2025). Prior studies also suggest that effective risk governance improves financial reporting quality and reduces opportunistic accounting behavior (Farooq et al., 2025; Oreshile, 2025). Therefore, stronger ERM implementation is expected to limit accrual-based earnings manipulation.

H1. 

ERM has a negative effect on accrual-based earnings management.

3.2. The Impact of ERM on Real Earnings Management

Real earnings management involves altering actual operational activities, such as production levels, discretionary expenditures, or sales timing, in order to influence reported earnings. Because these activities are embedded in operational decisions, they are often more difficult for external stakeholders to detect and monitor. As a result, managers may rely on real activities manipulation to achieve short-term performance targets even at the expense of long-term firm value (Lajnef & Ellouz, 2025; Oreshile, 2025). Enterprise Risk Management integrates risk assessment and performance monitoring into managerial decision-making processes and strengthens internal accountability mechanisms (Gupta, 2008; Prewett & Terry, 2018). Through improved risk oversight and operational monitoring, ERM may limit opportunistic operational decisions and reduce incentives for real earnings manipulation.

H2. 

ERM has a negative effect on real earnings management.

4. Research Design

4.1. Data Collection and Sample

The research sample consists of Chinese listed firms over the period from 2019 to 2024. The sample period begins in 2019, following the promulgation of the revised Securities Law of the People’s Republic of China in December 2019, which came into effect in March 2020. As analyzed by C. Wang et al. (2023), the 2019 revision introduced substantial changes to China’s securities regulatory regime, including reforms in securities issuance, trading, information disclosure, and enforcement mechanisms, thereby generating profound implications for market behavior and corporate governance practices. Using 2019 as the starting point allows this study to capture firm behavior under the new regulatory framework. The sample period extends to 2024 to incorporate the most recent available data, allowing for a comprehensive assessment of firm responses to the evolving institutional environment. Firm-level data are obtained from the Wind, CSMAR, and RESSET databases. To mitigate the influence of outliers on the empirical results, all continuous variables are winsorized at the 1st and 99th percentiles. Firms with missing key variables are excluded from the sample. The final dataset comprises 24,636 firm-year observations.

4.2. The Research Model

This paper employs a fixed effects model for empirical estimation. The Hausman test rejects the null hypothesis of random effects, indicating that the fixed effects specification is more appropriate. Accordingly, firm and year fixed effects are included in all regressions to control for unobserved heterogeneity. Equations (1) and (2) examine the impact of enterprise risk management on accrual-based earnings management (AEM) and real earnings management (REM). which serve as the dependent variables. ERM denotes the enterprise risk management index and is the key explanatory variable. AGE is measured as the natural logarithm of the firm’s age. RENI is defined as retained earnings divided by the total assets at the end of the fiscal year. SIZE is proxied by the natural logarithm of total assets. INDDIREC represents the proportion of independent directors on the board. BIG4 is a dummy variable equal to one if the firm is audited by an international B Big Four audit firm and zero otherwise. TOP1 denotes the shareholding ratio of the largest shareholder. ${\alpha }_i$ represents the company-specific fixed effect, which captures unobservable, time-invariant characteristics of each firm that may affect earnings management. ${\delta }_t$ represents the time-specific fixed effect, accounting for common shocks or trends affecting all companies in the same year. ${\epsilon }_{it}$ is the error term that captures unobservable factors and random disturbances at the company-year level.

$${\mathrm{AEM}}_{\mathrm{it}} = {\mathsf{\beta }}_0 + {\mathsf{\beta }}_1{\mathrm{ERM}}_{\mathrm{it}} + {{\mathsf{\beta }}_2{\mathrm{AGE}}_{\mathrm{it}} +{ \mathsf{\beta }}_3\mathrm{RENI}}_{\mathrm{it}} +{{ \mathsf{\beta }}_4\mathrm{SIZE}}_{\mathrm{it}} + {{\mathsf{\beta }}_5\mathrm{INDDIREC}}_{\mathrm{it}} +{{ \mathsf{\beta }}_6\mathrm{BIG}4}_{\mathrm{it}} + {{\mathsf{\beta }}_7\mathrm{TOP}1}_{\mathrm{it}} + {\alpha }_{i }+ {\delta }_t + {\epsilon }_{\mathit{it}}$$

(1)

$${\mathrm{REM}}_{\mathrm{it}} ={ \mathsf{\beta }}_0 + {\mathsf{\beta }}_1{\mathrm{ERM}}_{\mathrm{it} }+ {{\mathsf{\beta }}_2{\mathrm{AGE}}_{\mathrm{it}} + {\mathsf{\beta }}_3\mathrm{RENI}}_{\mathrm{it}} + {{\mathsf{\beta }}_4\mathrm{SIZE}}_{\mathrm{it}} + {{\mathsf{\beta }}_5\mathrm{INDDIREC}}_{\mathrm{it} }+{{ \mathsf{\beta }}_6\mathrm{BIG}4}_{\mathrm{it}} +{{ \mathsf{\beta }}_7\mathrm{TOP}1}_{\mathrm{it}} + {\alpha }_i + {\delta }_t +{ \epsilon }_{\mathit{it} }$$

(2)

4.3. Variable Description

4.3.1. Dependent Variables

Consistent with the earnings management literature, accrual-based earnings management (AEM) is measured using discretionary accruals derived from the modified Jones model (Dechow et al., 1995; Jones, 1991). Total accruals are first estimated using the balance sheet approach, and non-discretionary accruals are modeled as a function of firms’ economic fundamentals.

The modified Jones model is specified as follows:

$${\displaystyle \frac{{TA}_{it}}{A_{i,(t-1)}}}=\alpha {\displaystyle \frac{1}{A_{i,(t-1)}}}+{\beta }_1{\displaystyle \frac{\Delta {REV}_{it}}{A_{i,(t-1)}}}+{\beta }_2{\displaystyle \frac{{PPE}_{it}}{A_{i,(t-1)}}}{+\epsilon }_{it}$$

where ${TA}_{it}$ denotes total accruals for firm $i$ in year $t$, and $A_{i,(t-1)}$ represents total assets at the beginning of the period. $\Delta {REV}_{it}$ is the change in revenues, and ${PPE}_{it}$ denotes gross property, plant, and equipment. All variables are scaled by lagged total assets to control for firm size and mitigate heteroskedasticity. The residual term ε_it captures discretionary accruals, and its absolute value is used as a proxy for the extent of accrual-based earnings management, with larger values indicating higher levels of earnings management. In this study, financial data are obtained from the RESSET database. Following prior literature, the model is estimated cross-sectionally to obtain firm-specific discretionary accruals. The use of standardized financial data from RESSET enhances the reliability and comparability of the empirical results.

Real earnings management (REM) is measured following Roychowdhury (2006), Asmaranti et al. (2024); Kuo et al. (2014), which define REM as deviations from normal operating activities undertaken to influence reported earnings through real economic decisions. Consistent with prior literature, REM is decomposed into three components: abnormal cash flows from operations (ACFO), abnormal discretionary expenses (ADISX), and abnormal production costs (APROD).

ACFO is defined as the residual from the following industry-year regression model:

$${\displaystyle \frac{{CFO}_{it}}{{Assets}_{i,(t-1)}}}=\alpha +{\beta }_1{\displaystyle \frac{1}{{Assets}_{i,(t-1)}}}+{\beta }_2{\displaystyle \frac{{Sales}_{it}}{{Assets}_{i,(t-1)}}}+{\beta }_3{\displaystyle \frac{{∆Sales}_{it}}{{Assets}_{i,(t-1)}}}+{\epsilon }_{it}$$

where ${CFO}_{it}$ denotes cash flows from operations, and ${Assets}_{i,(t-1)}$ represents lagged total assets. ACFO is obtained as the residual from this model.

ADISX is computed as the residual from the following regression:

$${\displaystyle \frac{{DISX}_{it}}{{Assets}_{i,(t-1)}}}=\alpha +{\beta }_1{\displaystyle \frac{1}{{Assets}_{i,(t-1)}}}+{\beta }_2{\displaystyle \frac{{Sales}_{i,(t-1)}}{{Assets}_{i,(t-1)}}}+{\epsilon }_{it}$$

where ${DISX}_{it}$ includes advertising, R&D, and administrative expenses.

APROD is defined as the residual from:

$${\displaystyle \frac{{PROD}_{it}}{{Assets}_{i,(t-1)}}}=\alpha +{\beta }_1{\displaystyle \frac{1}{{Assets}_{i,(t-1)}}}+{\beta }_2{\displaystyle \frac{{Sales}_{it}}{{Assets}_{i,(t-1)}}}+{\beta }_3{\displaystyle \frac{{∆Sales}_{it}}{{Assets}_{i,(t-1)}}}+{\beta }_4{\displaystyle \frac{{∆Sales}_{i,(t-1)}}{{Assets}_{i,(t-1)}}}{+\epsilon }_{it}$$

where production costs are defined as the sum of the cost of goods sold and change in inventory.

The overall real earnings management measure is constructed as:

$${REM}_{it}={(-ACFO}_{it})+({-ADISX}_{it})+{APROD}_{it}$$

A higher value of REM indicates a greater extent of real earnings management.

Consistent with established empirical practice in large-sample accounting research, this paper directly adopts REM measures provided by the RESSET database. These measures are constructed based on the Roychowdhury (2006) framework widely used in the literature. Therefore, the above specifications are reported for transparency and methodological clarity rather than being independently re-estimated in this study.

4.3.2. Independent Variable

This analysis follows Gordon et al. (2009) and employs the ERM index (ERM) to measure the effectiveness of a firm’s risk management. Unlike other ERM proxies that rely on binary measures derived from disclosure statements, this index emphasizes the outcomes of integrated risk management activities embedded within organizational processes, capturing actual ERM effectiveness rather than symbolic adoption (Nasr et al., 2019; Panicker & Hiremath, 2016). Prior research supports employing multidimensional measures to more accurately reflect ERM effectiveness (Pangestuti et al., 2023). The ERM is based on COSO’s (2004) four business objectives, namely strategy, operations, reporting, and compliance, which jointly capture different dimensions of risk management effectiveness. ERM aggregates performance across these four dimensions into a single composite index. Each dimension is measured using two indicators, resulting in a total of eight indicators. The ERM is constructed by summing these indicators across the four dimensions, as follows:

$$\mathrm{ERM} = {\displaystyle \sum _{\mathrm{k}=1}^2}{\mathrm{Strategy}}_{\mathrm{k}}+{\displaystyle \sum _{\mathrm{k}=1}^2}{\mathrm{Operation}}_{\mathrm{k}}+{\displaystyle \sum _{\mathrm{k}=1}^2}{\mathrm{Reporting}}_{\mathrm{k}}+{\displaystyle \sum _{\mathrm{k}=1}^2}{\mathrm{Compliance}}_{\mathrm{k}}$$

The definitions of the indicators for strategy, operations, reporting, and compliance are described as follows.

• Strategy

(1)

When executing its strategy, a firm seeks to develop competitive advantages over other firms in the same industry (Porter, 2008). Such advantages are expected to reduce the overall risk of failure and enhance firm performance and value. Therefore, one measure of whether a firm has a successful strategy is the net sales of firm i minus the average sales in the same industry, divided by the standard deviation of sales of all firms in the same industry.

$${\mathit{Strategy}}_1={\displaystyle \frac{{\mathit{Net} \mathit{Sales} \mathit{of} \mathit{Firm}}_i-\mathit{Average} \mathit{Sales} \mathit{in} \mathit{Industry}}{\mathit{Standard} \mathit{Deviation} \mathit{of} \mathit{Sales} \mathit{in} \mathit{Industry}}}$$

where Net Sales of Firm refers to the sales of the firm i, Average Sales in Industry refers to the average sales of all firms in the same industry, Standard Deviation of Sales in Industry refers to the standard deviation of sales of all firms in the same industry.

(2)

${Strategy}_2$: The major benefit of ERM is to diversify and thus reduce risk by managing a portfolio of risks from all sources (Hoyt & Liebenberg, 2011). The second measure of strategic success is a firm’s ability to reduce its systematic risk compared to other firms in the same industry:

$${\mathit{Strategy}}_2 = {\displaystyle \frac{\Delta {\beta }_i-{\mu }_{\Delta \beta }}{{\sigma }_{\Delta \beta }}}$$

where $∆{\beta }_i$ = −(${\beta }_{i,t}$ − ${\beta }_{i,(t-1)}$); ${\beta }_i$ = firm $i$’s beta; ${\mu }_{∆\beta }$ = average industry $∆{\beta }_i$ in year $t$; and ${\sigma }_{∆\beta }$ = the standard deviation of $\Delta {\beta }_i$ across firms in the same industry.

• Operations

(1)

${Operation}_1$: Higher operating efficiency can mitigate a firm’s overall risk of failure and enhance its performance and value. Operation measures the relationship between a firm’s inputs and outputs. The first indicator measuring operating efficiency (${Operation}_1$) is the turnover of assets, defined as sales divided by year-end total assets.

$${\mathit{Operation}}_1 = {\displaystyle \frac{\mathit{Sales}}{ \mathit{Year-End} \mathit{Total} \mathit{Assets}}}$$

where Sales refers to the total sales of the firm during the period. Year-End Total Assets refer to the total assets of the firm at the end of the fiscal year.

(2)

${Operation}_2$: The other indicator of operating efficiency is the input-output ratio, defined as the log of sales divided by the log of the number of employees.

$${Operation}_2={\displaystyle \frac{log\left(Sales\right)}{ log (Number of Employees)}}$$

where Sales refers to the total sales of the firm during the period. Number of Employees refers to the total number of employees in the firm. log represents the logarithm (usually base ten or natural logarithm).

• Reporting

(1)

${Reporting}_1$: Earnings manipulation, financial restatements, and fraud are commonly associated with poor financial reporting quality (J. R. Cohen et al., 2004). Poor financial reporting should increase a firm’s overall risk of failure and thus reduce its performance. The indicator of poor reporting reliability is a combination of two observable variables, ${Reporting}_1= (Auditor Opinion) + \left(Restatement\right)$.

If the financial reporting of a firm is rated as unqualified in the auditor’s report, Auditor Opinion is set to 0; otherwise, it is set to −1. The restatement is also considered a reduction in a firm’s reporting reliability. If a firm announced a restatement, Restatement is set to −1; otherwise, it is set to 0. The range for ${Reporting}_1$ is therefore from −2 to 0.

(2)

${Reporting}_2$: The absolute value of abnormal accruals can measure a firm’s poor financial reporting quality (Johnson et al., 2002). The second measure of a firm’s reporting reliability is the relative proportion of normal accruals to total accruals divided by the sum of the absolute value of normal and abnormal accruals; the higher the value is, the higher the financial reporting quality is.

$${\mathit{Reporting}}_2 = {\displaystyle \frac{\left|\mathit{Normal} \mathit{Accruals}\right|}{ \left|\mathit{Normal} \mathit{Accruals}\right| + \left|\mathit{Abnormal} \mathit{Accruals}\right|}}$$

where Normal Accruals refers to the expected, regular accruals based on the firm’s financial operations. Abnormal Accruals refers to the accruals that deviate from the norm, potentially signaling poor financial reporting quality. The vertical bars $\left|\right|$ represent absolute value.

• Compliance

(1)

${Compliance}_1$: The enhancement of compliance with laws, regulations, and standards can reduce a firm’s overall risk and improve firm performance. Keefe et al. (1994) find that compliance with Generally Accepted Auditing Standards (GAAS) is positively associated with audit fees. The first measure of compliance is the proportion of auditor fees to total assets.

$${\mathit{Compliance}}_1 = {\displaystyle \frac{\mathit{Auditor} \mathit{Fees}}{ \mathit{Total} \mathit{Assets}}}$$

where Auditor Fees refer to the total fees paid by the firm for auditing services. Total Assets refers to the total assets of the firm.

(2)

${Compliance}_2$: According to Shavell (1982), the reported amount of settlement gains (losses) reflects both the plaintiff’s and the defendant’s agreement on their evaluations. If a firm complies with regulations to a greater extent, it is more likely to act as a plaintiff and achieve higher net gains (or lower net losses) in settlements. Thus, the second measure of compliance used in the study is the settlement net gains (losses) over total assets.

$${\mathit{Compliance}}_2 = {\displaystyle \frac{\mathit{Settlement} \mathit{Net} \mathit{Gain} \left(\mathit{loss}\right)}{\mathit{Total} \mathit{Assets}}}$$

4.3.3. Control Variable Measures

The control variables in this study are adopted based on prior empirical research on firm performance and corporate governance. Following existing studies, six control variables are included: AGE, RENI, SIZE, INDDIREC, BIG4, and TOP1. Variable measurements are defined as follows: AGE is the natural logarithm of the firm’s age, capturing the effects of firm maturity and lifecycle on the dependent variable, as emphasized in prior earnings management and corporate finance research (Lisdiono et al., 2022). RENI is measured as retained earnings at the end of the fiscal year divided by total assets, serving as an indicator of internal financing capacity (Alqam et al., 2022). SIZE is proxied by the natural logarithm of total assets, a commonly used scale measure to control for firm size effects in empirical models (Khuong et al., 2022). INDDIREC represents the percentage of independent directors on the board, capturing board independence as a governance control (Lu et al., 2024). BIG4 is a dummy variable indicating whether the firm is audited by an international Big Four audit firm (1 if audited by Deloitte, PwC, EY, or KPMG; 0 otherwise), a widely used proxy for audit quality in empirical governance studies (Lou et al., 2024). TOP1 denotes the proportion of shares held by the largest shareholder, reflecting ownership concentration. Incorporating these control variables helps mitigate omitted variable bias and improves the reliability of the estimated relationships (Chen & Ye, 2024).

5. Findings and Discussion

5.1. Descriptive Statistics

Table 1. Descriptive statistics.

Variable	N	P50	Mean	Std. Dev.	Min	Max	Skewness	Kurtosis
AEM	24,636	0.568	0.660	0.608	−0.707	2.474	0.644	3.334
REM	24,636	0.045	−0.021	0.514	−1.959	1.518	−0.781	6.139
ERM	24,636	3.791	3.959	1.327	0.876	9.828	1.398	7.463
AGE	24,636	3.091	3.070	0.293	1.386	4.234	−0.629	4.188
RENI	24,636	0.171	0.111	0.335	−1.915	0.555	−3.567	19.562
SIZE	24,636	22.158	22.393	1.533	19.451	27.544	0.878	4.047
INDDIREC	24,636	0.364	0.379	0.053	0.308	0.571	1.066	4.042
BIG4	24,636	0.000	0.070	0.255	0.000	1.000	3.368	12.341
TOP1	24,636	29.980	32.378	14.777	7.810	72.900	0.600	2.772

This table presents descriptive statistics of the variables used in this paper.

presents the descriptive statistics of the variables. All variables are measured following established approaches in prior literature, as detailed in Section 4.3, with specific references provided for each variable definition. The mean (median) values of accrual-based earnings management (AEM) and real earnings management (REM) are 0.660 (0.568) and −0.021 (0.045), respectively, indicating substantial variation in earnings management practices across firms. The ERM index has a mean of 3.959 with a standard deviation of 1.327, suggesting considerable heterogeneity in ERM implementation among Chinese A-share listed companies. Firm age (AGE) and board independence (INDDIREC) exhibit relatively limited dispersion, whereas retained earnings (RENI), firm size (SIZE), and ownership concentration (TOP1) show greater cross-sectional variation. Approximately 7% of the sample firms are audited by Big Four auditors. Overall, the distributions of several variables deviate from normality, but the influence of extreme observations is mitigated through winsorization at the 1st and 99th percentiles. The use of standardized measures enhances the validity and comparability of the empirical results.

5.2. Correlation Analysis

Table 2. Correlation Analysis (AEM as the Dependent Variable).

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
(1) AEM	1.000
(3) ERM	−0.090 *	1.000
(4) AGE	0.012	0.035 *	1.000
(5) RENI	0.017 *	0.037 *	−0.101 *	1.000
(6) SIZE	0.031 *	0.240 *	0.210 *	0.130 *	1.000
(7) INDDIREC	−0.027 *	0.021 *	−0.032 *	−0.025 *	0.001	1.000
(8) BIG4	−0.047 *	0.140 *	0.038 *	0.041 *	0.415 *	0.027 *	1.000
(9) TOP1	0.084 *	0.078 *	−0.070 *	0.213 *	0.132 *	0.032 *	0.103 *	1.000

* p < 0.10.

reports the correlation matrix with AEM as the dependent variable. AEM is negatively correlated with ERM (−0.090, p < 0.10) and BIG4 (−0.047, p < 0.10), but positively associated with REM (0.079, p < 0.10), profitability (RENI) (0.017, p < 0.10), firm size (0.031, p < 0.10), and ownership concentration (TOP1) (0.084, p < 0.10). ERM is positively correlated with firm size (0.240, p < 0.10), BIG4 (0.140, p < 0.10), age (0.035, p < 0.10), profitability (0.037, p < 0.10), board independence (INDDIREC) (0.021, p < 0.10), and ownership concentration (0.078, p < 0.10).

Table 3. Correlation Analysis (REM as the Dependent Variable).

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
(1) REM	1.000
(3) ERM	0.116 *	1.000
(4) AGE	0.022 *	0.035 *	1.000
(5) RENI	−0.188 *	0.037 *	−0.101 *	1.000
(6) SIZE	0.007	0.240 *	0.210 *	0.130 *	1.000
(7) INDDIREC	−0.013 *	0.021 *	−0.032 *	−0.025 *	0.001	1.000
(8) BIG4	−0.020 *	0.140 *	0.038 *	0.041 *	0.415 *	0.027 *	1.000
(9) TOP1	0.006	0.078 *	−0.070 *	0.213 *	0.132 *	0.032 *	0.103 *	1.000

* p < 0.10.

reports the correlation matrix with REM as the dependent variable. REM is positively associated with ERM (0.116, p < 0.10) and age (0.022, p < 0.10), while it is negatively correlated with profitability (−0.188, p < 0.10), INDDIREC (−0.013, p < 0.10), and BIG4 (−0.020, p < 0.10). Across both tables, most coefficients are statistically significant but small in magnitude, with the largest absolute value being 0.415, suggesting that multicollinearity is unlikely to pose a serious concern in subsequent regression analyses.

5.3. Regression Results

Table 4. Regression Results: The Impact of Enterprise Risk Management on Accrual and Real Earnings Management.

Variables	(1)	(2)
	AEM	REM
ERM	−0.0363 ***	−0.0101 ***
	(0.0034)	(0.0030)
AGE	0.2764 **	−0.2180 **
	(0.1182)	(0.1010)
RENI	−0.0464	−0.2417 ***
	(0.0381)	(0.0345)
SIZE	−0.0237	0.0865 ***
	(0.0185)	(0.0166)
INDDIREC	−0.1337	−0.0076
	(0.0864)	(0.0738)
BIG4	0.0268	0.0422
	(0.0280)	(0.0257)
TOP1	−0.0002	−0.0022 ***
	(0.0011)	(0.0008)
Constant	0.5475	−1.1519 **
	(0.5197)	(0.4593)
Observations	24,636	24,636
R-squared	0.762	0.770
Stkcd/year FE	Yes	Yes

This table reports the regression results examining the effect of ERM on accrual-based earnings management (AEM) and real earnings management (REM), *** p < 0.01, ** p < 0.05.

reports the regression results examining the impact of ERM on accrual-based earnings management (AEM) and real earnings management (REM). The results show that ERM is negatively and significantly associated with AEM and REM. Specifically, the coefficient of ERM is −0.036 (p < 0.01) for AEM and −0.010 (p < 0.01) for REM, indicating that firms with more developed ERM systems exhibit lower levels of both accrual-based and real earnings management.

To further assess the economic significance of these effects, the magnitude of the coefficients suggests that a one-unit increase in ERM is associated with a 0.036 decrease in AEM and a 0.010 decrease in REM. Although the magnitude appears economically modest, ERM is a composite index, and variation across firms reflects incremental improvements in risk governance systems rather than discrete policy shocks. Therefore, even relatively small marginal effects may accumulate into meaningful governance impacts.

Overall, the empirical evidence reported in Table 4 is consistent with the theoretical expectations of this paper. The significant negative association between ERM and accrual-based earnings management provides direct support for H1, which posits that ERM has a negative effect on accrual-based earnings management. Likewise, the significantly negative relationship between ERM and real earnings management lends strong empirical support to H2, suggesting that firms with more advanced ERM practices are less inclined to engage in real earnings management.

Beyond these statistical results, the findings suggest several possible mechanisms through which ERM constrains earnings management. First, ERM enhances internal monitoring and facilitates the integration of risk-related information across different organizational levels, thereby reducing information asymmetry and limiting managerial discretion in financial reporting (Jiang et al., 2024; Oreshile, 2025). This channel is particularly relevant for accrual-based earnings management, which depends heavily on accounting estimates and managerial judgment. In addition, ERM may also influence real earnings management by affecting firms’ operational decision-making processes (López, 2021). By incorporating risk assessment into resource allocation and performance evaluation, ERM can reduce incentives for short-term actions such as overproduction or discretionary expense manipulation. These findings suggest that ERM not only constrains accounting-based manipulation but may also reduce opportunistic behavior embedded in real business activities.

From a theoretical perspective, prior studies argue that ERM enhances internal governance by improving risk identification, information integration, and monitoring effectiveness, thereby constraining managerial discretion over financial reporting (Farooq et al., 2025; Gordon et al., 2009; Horvey & Odei-Mensah, 2024). Consistent with this view, existing empirical evidence documents a negative relationship between ERM quality and earnings management across different institutional settings. For example, T. S. Wang et al. (2018) find that firms with more robust ERM frameworks engage less in opportunistic earnings manipulation, while Oreshile (2025) shows that high-quality ERM significantly reduces real earnings management. Compared with prior studies, the findings of this study extend the literature by jointly examining both AEM and REM, thereby providing more comprehensive evidence on the governance role of ERM.

Prior research suggests that in emerging markets such as China, external governance mechanisms may be less effective due to weaker investor protection (Liu et al., 2018). In such settings, internal governance systems, including ERM, may play a more important role in constraining managerial opportunism (T. S. Wang et al., 2018). Therefore, the negative association observed in this paper may reflect the substitutive role of ERM in environments where external monitoring is relatively limited. The association between ERM and REM is also noteworthy, given that real earnings management is generally more difficult to detect and may have long-term consequences for firm performance. This finding suggests that ERM may be effective not only in limiting observable accounting manipulation but also in constraining less transparent forms of earnings management embedded in operational decisions. Taken together, the results support the view that ERM functions as an internal governance mechanism that is associated with lower levels of both accrual-based and real earnings management.

5.4. Robustness Tests

5.4.1. Robustness Tests Using Alternative Measures of the Dependent Variables

Table 5. Robustness Tests: Alternative Measures of the Dependent Variables.

Variables	(1)	(2)
	Re_AEM	Re_REM
ERM	−0.0478 ***	−0.0064 **
	(0.0133)	(0.0031)
AGE	0.6064	−0.0861
	(0.3827)	(0.0558)
RENI	−0.1128	−0.1071 ***
	(0.0963)	(0.0226)
SIZE	−0.0196	0.0290 **
	(0.0642)	(0.0117)
INDDIREC	−0.3780	0.0123
	(0.2308)	(0.0408)
BIG4	−0.1543	0.0506 *
	(0.1324)	(0.0265)
TOP1	0.0063	−0.0015 ***
	(0.0039)	(0.0006)
Constant	−1.3552	−0.3663
	(1.6854)	(0.2885)
Observations	24,047	24,636
R-squared	0.320	0.637
Stkcd/year FE	Yes	Yes

This table presents the robustness test results using alternative measures of the dependent variables to examine the effect of ERM on accrual-based earnings management (AEM) and real earnings management (REM). *** p < 0.01, ** p < 0.05, * p < 0.10.

presents robustness tests based on alternative measures of earnings management. For accrual-based earnings management, the dependent variable is replaced with Re_AEM, which is estimated using an extended modified Jones model based on the balance sheet approach, including an intercept term. Compared with the traditional Jones model, this specification adjusts revenue changes by deducting changes in accounts receivable and includes an intercept term, thereby providing an alternative specification of normal accrual processes (Kothari et al., 2005; Xie et al., 2024).

For real earnings management, REM is replaced with Re_REM, which is constructed following Roychowdhury (2006) using models without intercept terms to estimate abnormal operating cash flows, abnormal production costs, and abnormal discretionary expenses. These components are aggregated into a composite indicator, preserving the directional interpretation of real earnings management, whereby higher values reflect income-increasing manipulation through real operating activities, consistent with Nguyen et al. (2023).

The regression results reported in Table 5 show that, after replacing the dependent variables with Re_AEM and Re_REM, the estimated coefficients on ERM remain negative and statistically significant across both specifications. ERM is significantly negatively associated with Re_AEM at the 1% percent level (coefficient = −0.0478) and with Re_REM at the 5 percent level (coefficient = −0.0064). Overall, these findings are consistent with the baseline results, indicating that the negative effect of ERM on earnings management is robust to alternative measurement approaches.

5.4.2. Robustness Tests Based on Alternative Sample Periods (Year-by-Year Exclusion)

Table 6. Robustness Tests: Alternative Sample Period Specification.

Variables	(1)	(2)
	AEM	REM
ERM	−0.0347 ***	−0.0113 ***
	(0.0038)	(0.0030)
AGE	0.3204 **	−0.1681
	(0.1535)	(0.1200)
RENI	−0.0469	−0.2590 ***
	(0.0426)	(0.0380)
SIZE	−0.0265	0.1180 ***
	(0.0244)	(0.0200)
INDDIREC	−0.1878 **	0.0297
	(0.0948)	(0.0829)
BIG4	0.0308	0.0405
	(0.0310)	(0.0286)
TOP1	−0.0000	−0.0022 **
	(0.0013)	(0.0010)
Constant	0.4786	−2.0106 ***
	(0.6948)	(0.5538)
Observations	21,088	21,088
R-squared	0.773	0.797
Stkcd/year FE	Yes	Yes

This table reports the robustness test results based on alternative sample period specifications to assess the stability of the relationship between ERM and earnings management (accrual-based earnings management (AEM) and real earnings management (REM)). *** p < 0.01, ** p < 0.05.

presents robustness tests based on an alternative sample period. While the baseline regressions are estimated over the 2019 to 2024 period, this robustness analysis excludes each year sequentially from the sample period to examine whether the main results are sensitive to the choice of the time window. Such sample variation tests are commonly employed in empirical research to assess the stability of estimated relationships and to mitigate concerns related to period-specific shocks or structural changes. The results show that the coefficient on ERM remains negative and highly statistically significant for both accrual-based earnings management and real earnings management. Specifically, ERM is negatively associated with AEM and REM at the 1 percent significance level in both specifications. The magnitude and sign of the ERM coefficients are comparable to those reported in the baseline regressions, indicating that the negative effect of enterprise risk management on earnings management persists after adjusting the sample period. These findings further support the stability of the baseline results, indicating that the estimated relationship is not driven by specific time-period effects, thereby enhancing the generalizability of the empirical conclusions.

Overall, these findings suggest that the main conclusions are not driven by any particular year within the original sample window. Consistent with prior methodological recommendations in the empirical literature (Vaithilingam et al., 2024), the exclusion of individual years as a robustness check enhances confidence in the stability and reliability of the estimated effects.

To further address concerns regarding the potential impact of the COVID-19 pandemic,

Table 7. Robustness Tests: Excluding COVID-19 Period and Alternative Sample Specifications.

Variables	(1)	(2)
	AEM	REM
ERM	−0.0331 ***	−0.0157 ***
	(0.0043)	(0.0032)
AGE	−0.1728	−0.3541 *
	(0.2644)	(0.1837)
RENI	0.0962	−0.2339 ***
	(0.0638)	(0.0542)
SIZE	−0.0544	0.1275 ***
	(0.0460)	(0.0348)
INDDIREC	−0.1944 *	−0.0603
	(0.1166)	(0.0985)
BIG4	0.0514 *	0.0081
	(0.0285)	(0.0202)
TOP1	0.0009	−0.0016
	(0.0021)	(0.0017)
Constant	2.6058 **	−1.6096 *
	(1.2400)	(0.9323)
Observations	13,369	13,369
R-squared	0.830	0.868
Stkcd/year FE	Yes	Yes

This table reports robustness test results based on excluding the COVID-19 period (2020–2021) to examine whether the baseline results are driven by pandemic-related shocks. All regressions include firm and year fixed effects. Standard errors are reported in parentheses. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

reports a robustness test excluding the COVID-19 period (2020–2021) from the sample. The results show that the coefficient of ERM remains significantly negative for both AEM (−0.0331, p < 0.01) and REM (−0.0157, p < 0.01). These results indicate that the baseline findings are not driven by COVID-19 period observations and remain robust after excluding pandemic years.

5.4.3. Interaction Between Accrual-Based and Real Earnings Management

To examine the potential interaction between accrual-based and real earnings management, interaction terms between enterprise risk management (ERM) and each form of earnings management are incorporated into the baseline regressions. This specification captures the interdependence and conditional relationships between AEM and REM under the influence of ERM. A statistically significant interaction term indicates that ERM may affect not only the overall level of earnings management but also its composition.

Table 8. Interaction Effects between ERM, Accrual-Based Earnings Management, and Real Earnings Management.

Variables	(1)	(2)
	AEM	REM
ERM	−0.0345 ***	−0.0008
	(0.0033)	(0.0042)
AEM		0.1358 ***
		(0.0195)
ERM_AEM		−0.0095 **
		(0.0040)
AGE	0.3197 ***	−0.2461 **
	(0.1179)	(0.1007)
RENI	−0.0112	−0.2365 ***
	(0.0377)	(0.0338)
SIZE	−0.0360 *	0.0881 ***
	(0.0187)	(0.0166)
INDDIREC	−0.1384	0.0058
	(0.0864)	(0.0739)
BIG4	0.0231	0.0398
	(0.0277)	(0.0255)
TOP1	0.0001	−0.0022 ***
	(0.0011)	(0.0008)
REM	0.2583 ***
	(0.0350)
ERM_REM	−0.0280 ***
	(0.0081)
Constant	0.6749	−1.2093 ***
	(0.5238)	(0.4585)
Observations	24,636	24,636
R-squared	0.765	0.773
Stkcd/year FE	Yes	Yes

This table presents the results of the interaction analysis examining the interdependence between enterprise risk management (ERM), accrual-based earnings management (AEM), and real earnings management (REM). The interaction terms ERM × AEM and ERM × REM are included to capture potential conditional relationships between different forms of earnings management. *** p < 0.01, ** p < 0.05, * p < 0.10.

reports the regression results with REM as the dependent variable. The coefficient on AEM is positively associated with REM (0.1358, p < 0.01), indicating a strong positive association between the two forms of earnings management. This suggests that firms engaging in accrual-based earnings management are also more likely to engage in real earnings management, implying complementarity rather than substitution, suggesting that AEM and REM are not purely substitutes but may exhibit complementary behavior in practice. More importantly, the interaction term ERM × AEM is negative and statistically significant at the 5% level (−0.0095). This result indicates that the constraining effect of ERM on REM becomes stronger when the level of AEM is higher. In other words, under conditions of intensified earnings manipulation incentives, ERM appears to more effectively limit real earnings management. This finding suggests that ERM influences not only the level of REM but also its relationship with AEM. Table 8 presents the results using AEM as the dependent variable. The coefficient on REM is positive and statistically significant at the 1% level (0.2583), confirming a positive association between the two forms of earnings management. In addition, the interaction term ERM × REM (reported in Column 2 of Table 7) is negative and statistically significant at the 1% level (−0.0280).

Taken together, the results indicate that enterprise risk management (ERM) exerts a negative moderating effect on both accrual-based and real earnings management. The significantly negative interaction terms suggest that ERM strengthens its disciplinary role when the alternative form of earnings management is more pronounced, indicating state-dependent governance effects across different manipulation channels. The positive association between accrual-based and real earnings management further suggests that the two forms of earnings manipulation tend to co-move rather than exhibit a clear substitution pattern. This evidence is differs from the substitution mechanism documented in prior literature (Zang, 2012), which suggests a trade-off between accrual-based and real activities manipulation. Overall, the findings suggest that ERM constrains both forms of earnings management simultaneously and weakens the potential substitution between them, rather than inducing a simple shift from one channel to another. In this sense, ERM functions as a governance mechanism that reshapes managerial discretion across multiple earnings management channels in a more coordinated manner.

5.4.4. Robustness Tests Based on Alternative ERM Construction

To ensure robustness, enterprise risk management (ERM) is re-measured using a principal component analysis (PCA) to construct a composite index. The dependent variables include accrual-based earnings management (AEM) and real earnings management (REM). All regressions control for firm characteristics such as age, profitability, size, ownership concentration, and audit quality, and include firm and year fixed effects.

Table 9. Robustness Tests Using a PCA-Based Measure of Enterprise Risk Management (ERM).

Variables	(1)	(2)
	AEM_PCA	REM_PCA
ERM_PCA	−0.0447 ***	−0.0068
	(0.0150)	(0.0076)
AGE	0.3269 ***	−0.2081 **
	(0.1194)	(0.1016)
RENI	−0.0393	−0.2404 ***
	(0.0381)	(0.0343)
SIZE	−0.0071	0.0885 ***
	(0.0198)	(0.0167)
INDDIREC	−0.1245	−0.0021
	(0.0865)	(0.0739)
BIG4	0.0266	0.0409
	(0.0285)	(0.0257)
TOP1	−0.0004	−0.0022 ***
	(0.0011)	(0.0008)
Constant	−0.1223	−1.2677 ***
	(0.5575)	(0.4656)
Observations	24,630	24,630
R-squared	0.760	0.770
Stkcd/year FE	Yes	Yes

This table reports the regression results examining the relationship between enterprise risk management (ERM) and earnings management, where ERM is measured using a principal component analysis (PCA)-based index. Columns (1) and (2) present the results for accrual-based earnings management (AEM) and real earnings management (REM), respectively. Firm fixed effects and year fixed effects are included in all regressions. Standard errors are clustered at the firm level. *** p < 0.01, ** p < 0.05.

shows that the PCA-based ERM index is significantly negatively associated with AEM at the 1% level, indicating that firms with stronger risk management practices are less likely to engage in accrual-based earnings manipulation. In contrast, the coefficient of ERM on REM is negative but statistically insignificant, suggesting that the mitigating effect of ERM on real earnings management is not significant in this specification.

Overall, the findings based on the PCA-based ERM measure are consistent with the baseline results for AEM, while the evidence for REM remains statistically insignificant. This supports the view that ERM enhances monitoring and internal control effectiveness, thereby reducing accounting-based earnings distortion. However, the insignificant coefficient for REM suggests that the effect of ERM on real earnings management is sensitive to the choice of measurement approach. These results are consistent with the substitution perspective in the earnings management literature and highlight the importance of distinguishing between different types of earnings management when evaluating the effectiveness of ERM systems.

5.4.5. ERM Excluding Reporting Components

To address the concern that the ERM index may mechanically overlap with accrual-based earnings management (AEM), an alternative ERM measure is constructed by excluding all reporting-related components. In particular, variables capturing financial reporting quality and abnormal accruals are removed, as these elements are closely related to AEM and may introduce mechanical correlations. Prior literature has raised concerns that the inclusion of accounting-based components in composite indices may create mechanical correlations when similar measures are employed as dependent variables, thereby biasing empirical estimates (Kothari et al., 2005). Therefore, excluding reporting-related components provides a cleaner identification of the governance role of ERM. The regression results are reported in

Table 10. Robustness Test Using ERM Excluding Reporting-Related Components.

Variables	(1)	(2)
	AEM	REM
ERM_excl_reporting	−0.0326 ***	−0.0097 ***
	(0.0032)	(0.0027)
AGE	0.2591 **	−0.2235 **
	(0.1184)	(0.1010)
RENI	−0.0477	−0.2422 ***
	(0.0381)	(0.0344)
SIZE	−0.0242	0.0864 ***
	(0.0185)	(0.0166)
INDDIREC	−0.1315	−0.0072
	(0.0865)	(0.0738)
BIG4	0.0230	0.0413
	(0.0279)	(0.0257)
TOP1	−0.0004	−0.0022 ***
	(0.0011)	(0.0008)
Constant	0.5882	−1.1376 **
	(0.5211)	(0.4594)
Observations	24,636	24,636
R-squared	0.762	0.770
Stkcd/year FE	Yes	Yes

This table reports robustness tests using an alternative ERM measure that excludes reporting-related components. Specifically, variables related to financial reporting quality and abnormal accruals are removed to mitigate potential mechanical overlap with accrual-based earnings management (AEM). The dependent variables are accrual-based earnings management (AEM) and real earnings management (REM). All regressions include firm and year fixed effects, and standard errors are clustered at the firm level. *** p < 0.01, ** p < 0.05.

. The coefficient on ERM_excl_reporting remains negative and statistically significant at the 1% level for both AEM and REM. The magnitude and significance of the coefficients are comparable to the baseline results, indicating that the negative association between ERM and earnings management is not driven by measurement overlap. These findings alleviate concerns regarding internal validity and suggest that the governance effect of ERM on both accrual-based and real earnings management is robust to alternative constructions of the ERM index.

6. Additional Analyses

6.1. Endogeneity Analysis

6.1.1. Endogeneity Test Using Industry-Year Average ERM (Peer Effect Approach)

To address potential endogeneity between ERM and earnings management, this study replaces firm level ERM with the industry year average ERM excluding the focal firm. as a peer-based proxy This peer-based measure reduces concerns related to reverse causality and omitted variable bias (Du & Shen, 2018), although it does not constitute a formal instrumental variable approach. Regressions using this alternative ERM proxy yield results consistent with the baseline findings for both accrual based and real earnings management, indicating that the negative association between ERM and earnings management is robust to endogeneity concerns.

Table 11. Endogeneity Analysis Using Industry-Year Average ERM.

Variables	(1)	(2)
	AEM	REM
ERM_industry	−0.1745 ***
	(0.0277)
AGE	0.3047 **	−0.1965 *
	(0.1191)	(0.1012)
RENI	−0.0437	−0.2428 ***
	(0.0380)	(0.0344)
SIZE	−0.0254	0.0875 ***
	(0.0187)	(0.0168)
INDDIREC	−0.1215	−0.0051
	(0.0867)	(0.0738)
BIG4	0.0178	0.0394
	(0.0285)	(0.0258)
TOP1	−0.0005	−0.0023 ***
	(0.0011)	(0.0008)
Sum_ERM		−0.0001 *
		(0.0001)
Constant	1.0493 **	−1.2237 ***
	(0.5351)	(0.4612)
Observations	24,623	24,636
R-squared	0.760	0.770
Stkcd/year FE	Yes	Yes

This table presents the results of the endogeneity analysis using industry-year average ERM as an instrumental variable to examine the impact of ERM on earnings management (accrual-based earnings management (AEM) and real earnings management (REM)). Sum_ERM is employed as an alternative measure of industry-level ERM, defined as the sum of ERM levels across all peer firms within the same industry-year, capturing the overall intensity of risk management practices in the industry. This alternative specification is used to verify that the results are not sensitive to the construction of the peer-based ERM proxy. *** p < 0.01, ** p < 0.05, * p < 0.10.

reports the results using the industry-year average ERM as an alternative proxy to address endogeneity concerns. The coefficient on ERM_industry is negative and statistically significant at the one percent level in the accrual-based earnings management regression, indicating that higher industry-level ERM is associated with lower firm level accrual manipulation. Consistent patterns are also observed in the real earnings management specification, where the related industry-level ERM measure exhibits a negative and significant association with REM. The control variables largely maintain their expected signs and significance, and the explanatory power of the models remains high. Taken together with the baseline and robustness results, the evidence suggests that the findings are less likely to be driven by reverse causality or omitted variable bias. This strengthens the overall credibility of the empirical findings.

6.1.2. Dynamic Effects of ERM on Accrual-Based Earnings Management (AEM)

Table 12. Dynamic Effects of ERM on Accrual-Based Earnings Management (AEM): Lead–Lag Specification.

Variables	(1)	(2)	(3)
	AEM_L1	AEM_F1	AEM_L2
L.ERM	0.0365 ***
	(0.0038)
F.ERM		−0.0258 ***
		(0.0043)
L2.ERM			0.0162 ***
			(0.0041)
L.AEM	−0.0216 *	−0.1179 ***	−0.1016 ***
	(0.0129)	(0.0146)	(0.0152)
AGE	0.2925 *	0.5327 **	0.0565
	(0.1651)	(0.2329)	(0.2254)
RENI	−0.0470	−0.0709	0.0034
	(0.0429)	(0.0582)	(0.0524)
SIZE	−0.0397	−0.0778 **	−0.0263
	(0.0252)	(0.0313)	(0.0370)
INDDIREC	−0.1592	−0.1098	−0.2760 **
	(0.0968)	(0.1186)	(0.1084)
BIG4	0.0091	−0.0319	0.0201
	(0.0322)	(0.0590)	(0.0281)
TOP1	0.0001	0.0008	0.0013
	(0.0013)	(0.0016)	(0.0018)
Constant	0.5775	0.9564	1.1405
	(0.7246)	(0.9668)	(1.0202)
Observations	19,884	15,277	15,277
R-squared	0.784	0.796	0.811
Stkcd/year FE	Yes	Yes	Yes

This table reports the results from a dynamic specification examining the relationship between enterprise risk management (ERM) and accrual-based earnings management (AEM). The models incorporate lagged, lead, and second-order lag terms of ERM (L.ERM, F.ERM, and L2.ERM) to capture the temporal structure of the relationship. The lagged dependent variable (L.AEM) is included to control for the persistence of earnings management behavior. All regressions control for firm characteristics and include firm and year fixed effects. Standard errors are clustered at the firm level. *** p < 0.01, ** p < 0.05, * p < 0.10.

presents the results from the dynamic specification analyzing the relationship between ERM and accrual-based earnings management (AEM). The analysis incorporates lagged (L.ERM), lead (F.ERM), and second-order lag (L2.ERM) variables of ERM to examine the dynamic association between ERM and AEM. In addition, the lagged dependent variable (L.AEM) is included to account for the persistence of accrual-based earnings management. The number of observations varies across specifications due to the construction of lead and higher-order lag variables. All models include firm and year fixed effects, with standard errors clustered at the firm level.

The results indicate that lagged ERM (L.ERM) is positively associated with AEM and statistically significant at the 1% level, suggesting that the influence of ERM extends beyond the contemporaneous period. In contrast, the lead term (F.ERM) is negative and statistically significant, indicating that higher future ERM is associated with lower current accrual-based earnings management. This pattern provides some evidence against reverse causality concerns, as the effect of future ERM does not mirror the contemporaneous relationship. The second-order lag (L2.ERM) is positive and statistically significant, suggesting that past ERM may have persistent effects on AEM. Overall, the results suggest that the relationship between ERM and AEM exhibits a dynamic pattern over time. The inclusion of lag and lead terms helps reduce concerns about contemporaneous estimation bias.

6.1.3. Dynamic Effects of ERM on Real Earnings Management (REM)

Table 13. Dynamic Effects of ERM on Accrual-Based Earnings Management (AEM): Lead–Lag Analysis.

Variables	(1)	(2)	(3)
	REM_L1	REM_F1	REM_L2
L.ERM	0.0166 ***
	(0.0032)
F.ERM		−0.0064 *
		(0.0033)
L2.ERM			0.0063 **
			(0.0031)
L.REM	0.0228 *	−0.0638 ***	−0.0267 *
	(0.0123)	(0.0146)	(0.0142)
AGE	−0.0379	−0.0943	0.1878
	(0.1290)	(0.1818)	(0.1850)
RENI	−0.2414 ***	−0.3382 ***	−0.2207 ***
	(0.0382)	(0.0576)	(0.0488)
SIZE	0.1159 ***	0.1474 ***	0.1470 ***
	(0.0206)	(0.0300)	(0.0289)
INDDIREC	0.0021	0.1077	−0.0600
	(0.0830)	(0.1006)	(0.0998)
BIG4	0.0279	0.0332	0.0130
	(0.0285)	(0.0602)	(0.0252)
TOP1	−0.0019 **	−0.0025 **	−0.0016
	(0.0010)	(0.0012)	(0.0013)
Constant	−2.4873 ***	−2.9515 ***	−3.8566 ***
	(0.5939)	(0.8285)	(0.8583)
Observations	19,884	15,277	15,277
R-squared	0.808	0.821	0.842
Stkcd/year FE	Yes	Yes	Yes

This table reports the results from a dynamic specification examining the relationship between enterprise risk management (ERM) and real earnings management (REM). The models incorporate lagged, lead, and second-order lag terms of ERM (L.ERM, F.ERM, and L2.ERM) to capture the temporal structure of the relationship. The lagged dependent variable (L.REM) is included to control for the persistence of earnings management behavior. All regressions control for firm characteristics and include firm and year fixed effects. Standard errors are clustered at the firm level. *** p < 0.01, ** p < 0.05, * p < 0.10.

presents the results from the dynamic specification analyzing the impact of ERM on real earnings management (REM). The analysis incorporates lagged (L.ERM), lead (F.ERM), and second-order lag (L2.ERM) terms of ERM to capture the temporal dynamics of the relationship. In addition, the lagged dependent variable (L.REM) is included to account for the persistence of earnings management behavior over time. The number of observations varies across specifications due to the construction of lead and higher-order lag variables. All models include firm and year fixed effects, and standard errors are clustered at the firm level to address potential serial correlation and heteroskedasticity.

The results indicate that lagged ERM (L.ERM) is positively associated with REM and statistically significant at the 1% level, suggesting that the influence of ERM extends beyond the contemporaneous period. In contrast, the lead term (F.ERM) is negative and only weakly significant, indicating limited evidence that future ERM predicts current REM. This pattern provides limited evidence of reverse causality concerns, as the lead effect is weak and not fully consistent. The second-order lag (L2.ERM) also shows a positive and statistically significant coefficient, further indicating that the relationship between ERM and REM exhibits persistence over time.

Overall, the findings suggest that the relationship between ERM and REM is not purely contemporaneous but unfolds over multiple periods. The inclusion of lagged and lead terms, along with the persistence captured by L.REM, provides a more comprehensive depiction of the underlying dynamics. Importantly, the relatively weak and inconsistent effect of the lead variable alleviates concerns that the baseline results are driven by reverse causality. Taken together, the dynamic specification offers additional support that the observed relationship between ERM and earnings management is unlikely to be solely driven by endogeneity concerns.

6.2. Heterogeneity Analysis Based on Auditor Type

Table 14. Heterogeneity Analysis by Auditor Type.

Variables	(1)	(2)
	Big4	NoBig4
	REM	REM
ERM	−0.0029	−0.0100 ***
	(0.0074)	(0.0032)
AGE	−0.1603	−0.2165 **
	(0.3061)	(0.1092)
RENI	−0.5170 ***	−0.2377 ***
	(0.1792)	(0.0352)
SIZE	0.1077 ***	0.0860 ***
	(0.0411)	(0.0177)
INDDIREC	−0.0866	−0.0010
	(0.1908)	(0.0795)
TOP1	−0.0015	−0.0022 **
	(0.0025)	(0.0009)
Constant	−2.0341	−1.1290 **
	(1.3685)	(0.4849)
Observations	1696	22,831
R-squared	0.842	0.767
Stkcd/year FE	Yes	Yes

This table reports the heterogeneity analysis based on auditor type (Big Four versus non-Big Four), examining the association between ERM and real earnings management (REM). *** p < 0.01, ** p < 0.05.

indicates a clear difference in the effect of ERM on real earnings management (REM) between firms audited by Big Four and non-Big Four audit firms. In the subsample of firms audited by non-Big Four auditors, ERM exhibits a significantly negative effect on REM (coefficient = −0.0100, significant at the 1% level), whereas in the Big Four subsample, the effect is negative but not statistically significant (coefficient = −0.002943). This heterogeneity may reflect differences in audit quality and external monitoring mechanisms. In Big Four-audited firms, the higher external audit quality already imposes discipline on managerial behavior (Almuzaiqer et al., 2025), limiting the incremental effect of ERM and resulting in an insignificant coefficient.

Other control variables are generally consistent with prior expectations. Retained earnings (RENI) consistently exert a strong negative effect on REM in both subsamples, highlighting the reliance of real earnings management on internal financing resources. Firm size (SIZE) shows a positive association with REM, while Age and TOP1 exhibit weaker or mixed effects across auditor types. These findings Overall, these findings underscore that the effectiveness of ERM in constraining real earnings management is contingent upon external monitoring quality, illustrating significant heterogeneity across auditor types.

To further examine the robustness of the baseline findings,

Table 15. Heterogeneity Analysis: The Moderating Role of Audit Quality (Big4).

Variables	(1)
	REM
ERM	−0.0112 ***
	(0.0031)
BIG4	−0.0105
	(0.0358)
ERM × BIG4	0.0122 *
	(0.0068)
AGE	−0.2182 **
	(0.1011)
RENI	−0.2421 ***
	(0.0345)
SIZE	0.0866 ***
	(0.0167)
INDDIREC	−0.0076
	(0.0738)
TOP1	−0.0022 ***
	(0.0008)
Constant	−1.1482 **
	(0.4597)
Observations	24,636
R-squared	0.770
Stkcd/year FE	Yes

This table reports the heterogeneity analysis based on audit quality using interaction terms between ERM and a Big4 indicator. Big4 is a dummy variable equal to 1 if the firm is audited by a Big Four audit firm, and 0 otherwise. The interaction term (ERM × Big4) captures whether the association between ERM and earnings management differs across audit quality levels of audit quality. All regressions include firm and year fixed effects, and standard errors are clustered at the firm level. *** p < 0.01, ** p < 0.05, * p < 0.10.

further conducts a heterogeneity analysis by introducing interaction terms between ERM and audit quality (Big4). This approach allows for a more rigorous examination of whether the association between ERM and earnings management varies across firms with different levels of external monitoring. The results show that the interaction term between ERM and Big4 is positive and statistically significant at the 10% level. This indicates that the association between ERM and earnings management differs across auditor types. Specifically, the negative association between ERM and real earnings management is more pronounced among firms audited by non-Big4 auditors, while the effect is weaker among firms audited by Big4 auditors. These findings are consistent with a substitution relationship between internal governance (ERM) and external audit quality, suggesting that ERM may play a relatively more important role in constraining earnings management when external monitoring is relatively weak. The interaction-based evidence provides additional support for the governance role of ERM and highlights heterogeneity in its effect across audit environments.

7. Conclusions

Earnings management remains a persistent concern in corporate governance, particularly in institutional environments characterized by information asymmetry and agency conflicts between managers and shareholders. Within the framework of agency theory, managers may have incentives to manipulate accounting outcomes to pursue private benefits or short-term performance targets (Jensen & Meckling, 1976). Against this background, this paper examines whether ERM functions as an internal governance mechanism capable of constraining opportunistic reporting behavior. Using a sample of Chinese listed firms, the empirical results indicate that ERM is negatively and significantly associated with both accrual-based earnings management and real earnings management. The evidence suggests that firms with more developed risk management systems tend to exhibit lower levels of earnings manipulation, implying that comprehensive risk governance structures may improve the credibility of financial reporting and reduce managerial discretion in reporting decisions.

A series of robustness checks further support the reliability of the baseline results. Alternative measures of earnings management and variations in the sample period yield consistent findings. In addition, endogeneity analyses employing industry-year average ERM as an instrumental proxy provide further support for the conclusion that the negative association between ERM and earnings management is unlikely to be driven by reverse causality or omitted variable bias. Furthermore, interaction analyses between accrual-based and real earnings management reveal that ERM exerts a negative moderating effect across different manipulation channels, indicating that its governance role becomes stronger under higher levels of earnings management intensity. Additional robustness tests based on alternative ERM constructions, including PCA-based measures and specifications that exclude reporting-related components, also confirm the overall stability of the results, although the effect on real earnings management remains sensitive in certain specifications. These results collectively reinforce the view that ERM operates as an effective internal governance mechanism that constrains opportunistic managerial behavior.

Further analyses reveal that the governance effectiveness of ERM is not uniform across firms but varies with firm-specific characteristics and external monitoring conditions. Specifically, the inhibitory effect of ERM on real earnings management is significantly stronger in firms audited by non-Big Four auditors, while the effect is weaker and statistically insignificant in Big Four audited firms, indicating a substitution between internal and external governance mechanisms. Moreover, the analysis based on auditor type indicates that the inhibitory effect of ERM is stronger in firms audited by non-Big Four auditors, highlighting the interaction between internal governance mechanisms and external monitoring environments.

These findings also contribute to the literature on corporate governance and financial reporting by extending the application of agency theory and the resource-based view to the context of enterprise risk management. From an agency perspective, effective governance mechanisms can limit managerial opportunism by strengthening oversight and disciplining managerial decision-making (Healy & Wahlen, 1999). The evidence provided in this study demonstrates that ERM plays such a role by simultaneously constraining both accounting-based and operational forms of earnings manipulation. From the perspective of the resource-based view, firms achieve sustainable advantages through organizational capabilities embedded in managerial processes and governance structures (Barney, 1991). In this regard, ERM can be viewed as a governance-related capability that enables firms to structure risk oversight and strengthen internal control practices across organizational activities (Hayne & Free, 2014). Consequently, firms with more mature ERM systems appear better equipped to limit opportunistic reporting behavior and maintain higher-quality financial reporting.

From a practical perspective, the findings provide implications for different stakeholders. For corporate boards, strengthening ERM effectiveness requires moving beyond formal adoption toward substantive implementation. This can be achieved by establishing dedicated risk management committees, enhancing the frequency and depth of risk reporting, and improving coordination between ERM systems and internal audit functions. For risk managers, the results suggest that ERM should be embedded into operational decision-making processes. In particular, incorporating risk assessment into resource allocation, performance evaluation, and strategic planning may reduce incentives for short-term earnings manipulation and improve decision quality. For policymakers and regulators, especially in emerging markets, the findings highlight the importance of promoting standardized ERM disclosure frameworks and strengthening supervision of firms’ risk management practices. Regulatory initiatives that encourage transparency in risk governance and integrate ERM evaluation into corporate governance assessments may help reduce symbolic compliance and enhance the effectiveness of internal governance mechanisms.

The results also carry implications for policymakers and regulatory authorities, particularly in emerging market contexts where external governance mechanisms may still be evolving. Strengthening institutional guidance on risk governance disclosure and improving consistency in ERM reporting standards may help enhance comparability across firms and improve the effectiveness of internal governance mechanisms. Overall, the findings suggest that ERM plays a meaningful governance role in reducing earnings manipulation and improving the reliability of financial reporting in emerging market contexts.

8. Limitations and Future Research

Despite the contributions of this study, several limitations should be acknowledged. First, the empirical analysis is based on archival data from Chinese listed firms, which may limit the generalizability of the findings to other institutional contexts. Given that governance structures, regulatory environments, and market development differ across countries, the effectiveness of enterprise risk management (ERM) may vary in other settings. In particular, the findings are derived from the Chinese A-share market over the 2019–2024 period, which reflects a specific regulatory and institutional transition context. Second, this paper relies on secondary data obtained from established databases, which may be subject to measurement constraints. Although widely used proxies for earnings management (AEM and REM) and ERM are employed, these measures may not fully capture the underlying constructs, potentially introducing measurement error. Third, as with most empirical studies using observational data, the possibility of omitted variables cannot be entirely ruled out, even though multiple control variables, fixed effects, robustness tests, and endogeneity analyses are incorporated.

Future research may extend this study in several directions. First, cross-country comparative analyses could be conducted to examine whether the governance role of ERM differs across institutional environments, particularly between emerging and developed markets. Such studies could further test whether the substitutive relationship between internal governance (ERM) and external monitoring (e.g., audit quality) observed in this paper is generalizable across different institutional settings. Second, future studies may employ alternative data sources, such as survey data or case studies, to better capture the internal processes and organizational dynamics of ERM implementation. Third, further research could explore more granular mechanisms through which ERM influences managerial behavior, such as risk culture, internal control integration, and information processing within firms, consistent with the governance and monitoring role evidenced in this study. Finally, building on the heterogeneity findings related to audit quality (Big Four vs. non-Big Four), future research may examine the interaction between ERM and other governance mechanisms, including board effectiveness and external auditing, to provide a comprehensive understanding of how internal and external governance jointly shape earnings management behavior.