Voluntary Audits of Nonfinancial Disclosure and Earnings Quality

Abstract

We investigate the association between voluntary assurance of a firm’s corporate social responsibility (CSR) report and earnings management. A concern with CSR reports is they are used to promote a socially responsible image without a meaningful commitment to CSR activities, referred to as “greenwashing”. To credibly signal the CSR report is reliable, a firm can incur the additional costs to voluntarily obtain assurance. Our results show that strong corporate governance plays a crucial role in limiting earnings management. The most consistent improvements in earnings quality occur when firms combine strong governance with CSR assurance from a non-accounting provider (NonACCT). The combination of strong governance and NonACCT assurance appears to be mutually reinforcing, suggesting a symbolic legitimacy strategy that is also substantively effective.

1. Introduction

As firms increasingly disclose their environmental, social, and governance (ESG) activities in standalone corporate social responsibility (CSR) reports, concerns have arisen regarding the credibility of these disclosures. Specifically, some firms may use CSR reports to project a socially responsible image without a meaningful commitment to CSR activities, a practice referred to as “greenwashing” (Huang & Watson, 2015). To enhance the credibility of CSR reports, firms can voluntarily obtain assurance from third-party providers. This study examines whether voluntary CSR assurance is associated with higher financial reporting transparency, measured through earnings management.

Prior research suggests that firms engaged in socially responsible practices tend to constrain earnings management and misconduct, consistent with building a positive corporate reputation and providing more reliable financial information (Kim et al., 2012; Christensen, 2016). However, the effectiveness of voluntary CSR assurance as a signal of financial reporting transparency remains an open question. While some firms may use assurance as a commitment to reliable disclosures, others may procure it to mitigate stakeholder concerns without substantively improving reporting quality. This study seeks to bridge this gap by investigating whether voluntary assurance of CSR reports correlates with reduced earnings management, both through accrual-based methods and real activities manipulation.

Furthermore, we explore whether the type of assurance provider influences earnings quality. Firms may seek assurance from accounting firms (ACCT) or non-accounting firms (NONACCT), such as boutique consulting or engineering firms. Cohen and Simnett (2015) highlight the need for further research on the implications of different assurance providers. If ACCT auditors, perceived as high-quality financial auditors, enhance transparency more effectively than NONACCT providers, we expect firms engaging ACCT assurers to exhibit lower earnings management.

Additionally, we investigate the interaction between governance quality and CSR assurance. Firms with strong governance structures are generally expected to engage in less earnings management (Beasley, 1996; Klein, 2002). It remains unclear whether CSR assurance complements strong governance as an additional signal of reporting quality or serves as a substitute, making assurance redundant for firms with robust governance mechanisms.

This study contributes to the literature in several ways. First, it extends the discussion on voluntary audits by examining their role in nonfinancial disclosures. Second, it builds upon research exploring the relationship between assurance on CSR reports and earnings management (Kim et al., 2012; Chih et al., 2008; Prior et al., 2008). Third, it utilizes a large international sample to provide cross-country insights, as the prevalence of CSR assurance varies significantly (GRI, 2013). Finally, this study is among the first to assess the impact of assurance providers on earnings quality, addressing a key gap in sustainability assurance research.

The structure of this paper is as follows. Section 2 provides a review of relevant literature and formulates hypotheses. Section 3 details our data, sample selection, and research methodology. Our empirical findings are presented in Section 4, followed by the conclusion in Section 5.

2. Literature Review and Hypothesis Development

Legitimacy theory is based on the idea that organizations must align their actions with societal norms, values, and expectations to gain and maintain approval from stakeholders. Legitimacy is not static; it evolves over time as societal norms, values, and stakeholder expectations change. Organizations must continuously adapt to maintain their legitimacy (Suchman, 1995). There are two strategies for legitimacy management, the substantive and the symbolic approach (Ashforth & Gibbs, 1990). Substantive management entails making real changes in policies, processes, or performance to align with societal expectations, whereas symbolic management involves using communication, reporting, or public relations strategies to shape stakeholder perceptions without necessarily making substantial internal changes.

The voluntary and largely unregulated nature of CSR obligations and disclosure allows for subjectivity in how firms construct their reports (Dhaliwal et al., 2011; Marquis & Qian, 2014). Companies have significant discretion in determining the scope of their CSR disclosures, which tend to lead to selective disclosures (Clarkson et al., 2019; García-Sánchez et al., 2022; Tashman et al., 2019).

Legitimacy management, especially when firms adopt the symbolic approach, undermines the reliability and integrity of CSR reports (Boiral et al., 2019) and weakens their credibility (Clarkson et al., 2019; Jauernig & Valentinov, 2019; Michelon et al., 2015). It is therefore imperative to note that merely producing a CSR report (without assurance) does not automatically equate to transparent financial reporting. Prior research on the link between CSR disclosure and earnings management, a common proxy for financial reporting opacity, is mixed at best.

2.1. Assurance of CSR Reports

Voluntary third-party assurance of corporate social responsibility (CSR) reports has gained traction as a mechanism to enhance the credibility of sustainability disclosures. This trend is largely driven by growing concerns over “greenwashing”, wherein companies may publish elaborate environmental and social reports without corresponding improvements in actual performance (Huang & Watson, 2015; Du & Wu, 2019). By engaging independent assurors, firms aim to signal a commitment to transparency in their non-financial reporting.

Prior research underscores the competitive nature of the sustainability assurance market, with both accounting (ACCT) and non-accounting (NONACCT) providers offering assurance services (Cohen & Simnett, 2015). ACCT providers, typically audit firms, bring extensive assurance expertise and are motivated by reputational concerns tied to financial audits. In contrast, NONACCT providers often bring specialized knowledge in sustainability issues, offering a different value proposition.

External assurance is widely viewed as a mechanism to increase the reliability and comparability of CSR disclosures. Studies suggest that such assurance refines reporting definitions and methodologies, resulting in higher-quality disclosures. For example, Ballou et al. (2018) found that firms with assured sustainability reports experienced fewer financial restatements, suggesting more dependable financial reporting overall. Similarly, Garanina (2024) observed that firms providing more extensive CSR disclosures tend to weaken the negative association between earnings management and market value, as stakeholders respond positively to a credible CSR image.

Despite these benefits, not all CSR disclosures are equally credible. Some research indicates that CSR reporting can be used opportunistically. Boiral et al. (2019) argue that greenwashing compromises the integrity of CSR reports, reducing their reliability and perceived legitimacy. Clarkson et al. (2019) and Jauernig and Valentinov (2019) further emphasize that greenwashing undermines the credibility of voluntary disclosures.

Adding to the complexity, Ruwanti et al. (2019) found a positive association between CSR activities and earnings management, suggesting that some firms may engage in CSR initiatives while simultaneously manipulating earnings. In contrast, other studies present more optimistic findings. For instance, Kim et al. (2012) report that firms with stronger CSR performance tend to provide more transparent and reliable earnings, supporting a negative relationship between CSR engagement and earnings management.

Given these mixed findings, the relationship between CSR disclosure and earnings quality remains inconclusive. Building on this literature, the present study explicitly examines whether CSR assurance is associated with reduced earnings management. We also investigate whether this relationship is influenced by the type of assurance provider and the firm’s governance quality. Accordingly, we propose the following hypothesis:

Hypothesis 1. 

There is no significant relationship between the issuance of a CSR report and earnings management compared to companies that do not issue a CSR report.

Engaging independent assurance providers reinforces a company’s commitment to transparency in nonfinancial reporting. Research suggests that external assurance improves the reliability of CSR disclosures by refining reporting definitions, methodologies, and overall disclosure quality (Maroun, 2019). Ballou et al. (2018) found that firms with assured sustainability reports experience fewer financial restatements, indicating stronger financial reporting integrity. Similarly, Garanina (2024) observed that firms with extensive CSR disclosures show a weaker negative relationship between earnings management and market value, suggesting that investors positively evaluate a strong CSR image. Kim et al. (2012) further document that firms with higher CSR performance produce more transparent and reliable earnings, supporting a negative association between CSR engagement and earnings management.

However, other studies indicate that CSR disclosures can be exploited opportunistically (Boiral et al., 2019). Greenwashing undermines the integrity and credibility of CSR reports (Clarkson et al., 2019; Jauernig & Valentinov, 2019), making it difficult to discern genuine efforts from strategic image management. Ruwanti et al. (2019) found a positive relationship between CSR activities and earnings management, suggesting that some firms enhance CSR efforts while simultaneously engaging in financial manipulation. The empirical evidence on the relationship between CSR disclosure and earnings management remains inconclusive. Therefore, we propose the following hypothesis:

Hypothesis 2. 

Companies with assured CSR reports are less likely to engage in earnings management than those without assurance.

2.2. Corporate Governance and Earnings Management

Empirical research consistently demonstrates that strong corporate governance is associated with higher earnings quality. Studies show that firms with independent directors are less likely to engage in accounting fraud or aggressive accruals (Beasley, 1996; Klein, 2002). More recent findings reinforce this relationship, indicating that active board oversight and vigilant audit committees reduce earnings manipulation. Governance factors such as audit committee expertise and auditor quality are negatively associated with earnings management (Zheng et al. (2025) identify four key governance mechanisms—board independence, audit committee size, audit quality, and board gender diversity—as contributors to improved financial report clarity. These mechanisms not only mitigate earnings manipulation but also address more subtle reporting issues, such as obfuscatory disclosure language.

However, governance effectiveness varies depending on context and specific mechanisms. Some studies suggest that independent directors or CEO-Chair separation alone do not necessarily curb earnings management, particularly in firms where controlling shareholders or family ties influence board dynamics (Nabila & Saiful, 2023). While the broader trend supports the premise that strong governance reduces earnings management, governance alone is not a definitive solution. This underscores the growing interest in how governance interacts with other mechanisms, such as CSR assurance, to enhance financial reporting integrity. Hence, we propose the following hypothesis:

Hypothesis 3. 

Companies with strong governance and assured CSR reports are less likely to engage in earnings management than those with strong governance but no assurance.

2.3. Assurance Providers: Accounting vs. Non-Accounting Firms (ACCT vs. NONACCT)

When companies seek CSR assurance, they can choose between accounting (ACCT) firms, typically large audit firms (such as the Big Four), and non-accounting (NONACCT) providers, including sustainability consultancies or engineering firms. Research examines whether the type of assurer influences the quality and credibility of CSR assurance, with implications for financial reporting outcomes.

Accounting firms bring established audit methodologies and greater verification expertise, adding rigor to the assurance process. Investors tend to place higher value on sustainability reports assured by Big Four auditors, reflecting a perceived credibility premium (Clarkson et al., 2019; Handayati et al., 2025). This aligns with expectations that top-tier accounting firms adhere to stringent standards and provide more impartial assessments.

Empirical evidence suggests that CSR assurance from accounting firms is linked to better financial reporting outcomes compared to that from non-accounting providers. Cuadrado-Ballesteros et al. (2017) found that information asymmetry between firms and investors is significantly reduced when assurance is conducted by professional accountants, particularly when reasonable assurance (as opposed to limited assurance) is provided. This underscores the idea that CSR assurance enhances transparency and reduces stakeholder uncertainty, especially when conducted by high-quality accounting firms.

While most studies suggest that accounting firms provide more reliable CSR assurance, some findings indicate that provider type does not always yield significant differences in financial outcomes. For instance, Kuo et al. (2021) found that in Taiwan’s mandatory disclosure setting, while most firms chose Big Four assurors, the choice of provider had trivial effect on firms’ cost of debt. This suggests that in certain regulatory environments, assurance credibility may be influenced by factors beyond provider type. Nonetheless, the prevailing view is that accounting firms signal stronger reliability, contributing to greater constraints on earnings management. Thus, we propose the following hypotheses:

Hypothesis 4a. 

Companies assured by an ACCT provider are less likely to engage in earnings management compared to those without assurance.

Hypothesis 4b. 

Companies assured by a NONACCT provider are less likely to engage in earnings management compared to those without assurance.

2.4. Governance Mechanisms, CSR Assurance, and Earnings Management

Recent research explores how internal governance structures and CSR assurance interact to influence earnings management. The key question is whether they complement each other—reinforcing transparency—or serve as substitutes, where one mechanism compensates for the absence of the other. Current evidence suggests a complementary relationship, with both governance and assurance working together to promote financial reporting integrity.

Meqbel et al. (2024) found that while firms engaging in earnings management often seek sustainability assurance to enhance legitimacy, the presence of a dedicated CSR committee mitigates the symbolic use of assurance. This implies that without strong governance, some firms may use CSR assurance merely as a superficial gesture. However, when paired with robust governance mechanisms, assurance becomes substantive, reinforcing its role in reducing earnings management.

Other studies echo the view that governance and accountability mechanisms jointly constrain opportunistic reporting. Research in the corporate governance literature suggests that CSR-focused governance mechanisms—such as sustainability committees and independent boards—yield incremental benefits in reducing CSR “decoupling”, where reported CSR efforts do not align with actual practices (Abweny et al., 2025; Meqbel et al., 2024). Additionally, Uyar et al. (2023) found that while audit committees and CSR committees may serve overlapping monitoring roles, firms with both demonstrate the greatest improvements in reporting transparency.

Zheng et al. (2025) further highlight that governance mechanisms moderate the relationship between earnings management and poor disclosure clarity. This suggests that governance enhances the detectability of earnings manipulation, reinforcing the value of assurance and other transparency measures. However, some evidence indicates that exceptionally strong governance may render voluntary CSR assurance redundant, as such firms already enjoy reputations for reporting integrity (Zhou et al., 2016). Conversely, weak governance may undermine the credibility of assurance, limiting its impact.

The dominant perspective in recent literature is that governance and assurance are mutually reinforcing. Strong governance ensures rigorous assurance processes, while external assurance provides an additional validation layer, complementing internal controls. Hence, we propose the following hypotheses:

Hypothesis 5a. 

Companies with strong governance and assurance from an ACCT provider are less likely to engage in earnings management compared to those without assurance.

Hypothesis 5b. 

Companies with strong governance and assurance from a NONACCT provider are less likely to engage in earnings management compared to those without assurance.

3. Research Design

3.1. Data and Sample Selection

This section outlines our data sources, sample selection process, dependent variable measurement, and empirical models.

We obtain a comprehensive list of companies that publish CSR reports from the Global Reporting Initiative (GRI), covering the period from 1999 to 2015. GRI is widely recognized as the de facto standard for CSR reporting (Fonseca et al., 2014; Skouloudis et al., 2010). The GRI database provides a global listing of CSR-reporting companies, and from 2009 onward, it includes key details such as whether the report is assured, the assurance provider, the scope and level of assurance, and the assurance standard.

To identify public companies within the GRI dataset, we use the Standard & Poor’s Capital IQ (CIQ) Identifier/Converter plugin in Excel. We then cross-check sustainability assurance data for public firms against the GRI Disclosure Database (http://database.globalreporting.org/, accessed on 1 August 2016) to ensure accuracy. In cases where discrepancies exist between the two sources, we update the GRI list based on the Disclosure Database information.

Financial statement auditor names, financial data, and firm age are sourced from the CIQ database and the Compustat Global database, while ESG scores are retrieved from Bloomberg. After matching data from GRI, CIQ, and Compustat Global, we obtain a final sample of 36,563 firm-year observations spanning fiscal years 2009 to 2015. This dataset provides the necessary information to compute discretionary accruals, proxies for real activities manipulation, and relevant control variables.

Table 1. Sample Construction.

	No. of Obs
Observations common to GRI, Capital IQ, and Compustat with ESG data and ADMIRED data for years 1999–2015	53,026
No. of obs after merging with residuals calculated using the modified Jones (1991) model	51,685
No. of obs after deleting Kuwait, Malta, Morocco, Oman, Qatar, Saudi Arabia, and United Arab Emirates because they are not classifiable as having shareholder or stakeholder orientation	51,457
No. of obs after deleting South Africa and France—GSSB (2018)	49,467
No. of obs left after deleting companies with no identifying codes	49,462
No. of obs left with non-missing ABS_DA data	48,974
No. of obs from 2009–2015	42,921
No. of obs remaining with variables in regression models	36,563

presents the details of our sample construction.

3.2. Measurement of Dependent Variables

3.2.1. Discretionary Accruals

Consistent with prior research, we use discretionary accruals as a proxy for earnings management (Gordon et al., 2017; Zang, 2012; Kim et al., 2012; Kothari et al., 2005; Roychowdhury, 2006; DeFond & Subramanyam, 1998; Subramanyam, 1996). Following Zang (2012), we apply the modified Jones (1991) model to measure discretionary accruals. As in Kim et al. (2012), we use the absolute value of discretionary accruals (ABS_DA) in our primary analyses to capture both income-increasing and income-decreasing accruals.

We calculate ABS_DA as the residual from the relevant model, estimated by country, two-digit SIC code, and year (see Supplementary Materials for details). A negative association between the assurance variable (assured’) and ABS_DA indicates that firms obtaining assurance for their CSR reports engage in lower earnings management compared to those without assurance.

3.2.2. Real Activities Manipulation

Real activities manipulation (RAM) refers to managerial actions aimed at meeting or exceeding earnings thresholds (Kim et al., 2012; Roychowdhury, 2006). We adopt the framework of Kim et al. (2012) to develop RAM proxies, employing four measures: (1) abnormal operating cash flows (AB_CFO), (2) abnormal production costs (AB_PROD), (3) abnormal discretionary expenses (AB_EXP), and (4) a composite measure of RAM (COMBINED_RAM). The abnormal values of these three individual proxies are computed as the residuals from the relevant models, estimated by country, two-digit SIC code, and year (see Supplementary Materials for details).

As in Kim et al. (2012), we analyze both individual and combined proxies, computing COMBINED_RAM as AB_CFO − AB_PROD + AB_EXP. A positive association between assurance and AB_CFO, AB_EXP, and COMBINED_RAM suggests that firms with assured CSR reports engage in lower real activities manipulation. Conversely, a negative association between assurance and AB_PROD similarly implies reduced earnings management among firms with CSR assurance (Kim et al., 2012).

3.3. Empirical Models

We estimate the following models of accrual-based earnings management and real activities manipulation:

Accruals-based earnings management model:

$$\begin{array}{cc}\hfill AB{S\text{\_}DA}_t={\propto }_0& +{\propto }_1{CSR\text{\_}REPORT}_t+{\propto }_2{ASSURED}_t+{\propto }_3{GOV\text{\_}DISC\text{\_}SCORE}_t\hfill \\ & +{\propto }_4{{CSRREPORT}_t\ast GOV\text{\_}DISC\text{\_}SCORE}_t+{\propto }_5{ASSURED}_t\ast {GOV\text{\_}DISC\text{\_}SCORE}_t\hfill \\ & +{\propto }_{6}HIGH{\text{\_}GOV\text{\_}SCORE}_t+{\propto }_{7}LOW{\text{\_}CONCERN}_t+{\propto }_{8}SALES\text{\_}GROWTH\hfill \\ & +{\propto }_{9}SALES\text{\_}VOLATILITY+{\propto }_{10}CFO\text{\_}VOLATILITY+{\propto }_{11}OPERATING\text{\_}CYCLE\hfill \\ & +{\propto }_{12}COMBINED{\text{\_}RAM}_t+{\propto }_{13}{MB}_{t-1}+{\propto }_{14}{LEV}_{t-1}+{\propto }_{15}{EO}_t+{\propto }_{16}R{D\text{\_}INT}_t\hfill \\ & +{\propto }_{17}A{D\text{\_}IND\text{\_}INT}_t+{\propto }_{18}{ADMIRED}_t+{\propto }_{19}{MISS\text{\_}SALES\text{\_}GROWTH}_t\hfill \\ & +{\propto }_{20}{MISS\text{\_}SALES\text{\_}VOLATILITY}_t+{\propto }_{21}{MISS\text{\_}CFO\text{\_}VOLATILITY}_t\hfill \\ & +{\propto }_{22}{MISS\text{\_}OPERATING\text{\_}CYCLE}_t+{\propto }_{23}{MISS\text{\_}GOV\text{\_}DISC\text{\_}SCORE}_t+{\epsilon }_t\hfill \end{array}$$

(1)

$$\begin{array}{cc}\hfill AB{S\text{\_}DA}_t={\propto }_0& +{\propto }_1{CSR\text{\_}REPORT}_t+{\propto }_2{ASSURED\text{\_}NONBIG4}_t+{\propto }_3{ASSURED\text{\_}BIG4}_t+{\propto }_4{GOV\text{\_}DISC\text{\_}SCORE}_t\hfill \\ & +{\propto }_5{{CSRREPORT}_t\ast GOV\text{\_}DISC\text{\_}SCORE}_t+{\propto }_{6}ASSURED{\text{\_}NONBIG4}_t\ast {GOV\text{\_}DISC\text{\_}SCORE}_t\hfill \\ & +{\propto }_{7}A{SSURED\text{\_}BIG4}_t\ast {GOV\text{\_}DISC\text{\_}SCORE}_t+{\propto }_{8}HIGH{\text{\_}GOV\text{\_}SCORE}_t+{\propto }_{9}LOW{\text{\_}CONCERN}_t\hfill \\ & +{\propto }_{10}SALES\text{\_}GROWTH+{\propto }_{11}SALES\text{\_}VOLATILITY+{\propto }_{12}CFO\text{\_}VOLATILITY\hfill \\ & +{\propto }_{13}OPERATING\text{\_}CYCLE+{\propto }_{14}COMBINED{\text{\_}RAM}_t+{\propto }_{15}{MB}_{t-1}+{\propto }_{16}{LEV}_{t-1}+{\propto }_{17}{EO}_t\hfill \\ & +{\propto }_{18}R{D\text{\_}INT}_t+{\propto }_{19}A{D\text{\_}IND\text{\_}INT}_t+{\propto }_{20}{ADMIRED}_t+{\propto }_{21}{MISS\text{\_}SALES\text{\_}GROWTH}_t\hfill \\ & +{\propto }_{22}{MISS\text{\_}SALES\text{\_}VOLATILITY}_t+{\propto }_{23}{MISS\text{\_}CFO\text{\_}VOLATILITY}_t\hfill \\ & +{\propto }_{24}{MISS\text{\_}OPERATING\text{\_}CYCLE}_t++{\propto }_{25}{MISS\text{\_}GOV\text{\_}DISC\text{\_}SCORE}_t+{\epsilon }_t\hfill \end{array}$$

(2)

Real activities manipulation model:

$$\begin{array}{cc}\hfill {RAM\text{\_}Proxy}_t=& {\propto }_0+{\propto }_1{CSR\text{\_}REPORT}_t+{\propto }_2{ASSURED}_t+{\propto }_3{GOV\text{\_}DISC\text{\_}SCORE}_t\hfill \\ & +{\propto }_4{{CSRREPORT}_t\ast GOV\text{\_}DISC\text{\_}SCORE}_t+{\propto }_5{ASSURED}_t\ast {GOV\text{\_}DISC\text{\_}SCORE}_t\hfill \\ & +{\propto }_{6}HIGH{\text{\_}GOV\text{\_}SCORE}_t+{\propto }_{7}LOW{\text{\_}CONCERN}_t+{\propto }_{8}SALES\text{\_}GROWTH+{\propto }_{9}SALES\text{\_}VOLATILITY\hfill \\ & +{\propto }_{10}CFO\text{\_}VOLATILITY+{\propto }_{11}OPERATING\text{\_}CYCLE+{\propto }_{12}{MB}_{t-1}+{\propto }_{13}{LEV}_{t-1}+{\propto }_{14}{EO}_t\hfill \\ & +{\propto }_{15}R{D\text{\_}INT}_t+{\propto }_{16}A{D\text{\_}IND\text{\_}INT}_t+{\propto }_{17}{ADMIRED}_t+{\propto }_{18}{MISS\text{\_}SALES\text{\_}GROWTH}_t\hfill \\ & +{\propto }_{19}{MISS\text{\_}SALES\text{\_}VOLATILITY}_t+{\propto }_{20}{MISS\text{\_}CFO\text{\_}VOLATILITY}_t\hfill \\ & +{\propto }_{21}{MISS\text{\_}OPERATING\text{\_}CYCLE}_t+{\propto }_{22}{MISS\text{\_}GOV\text{\_}DISC\text{\_}SCORE}_t+{\epsilon }_t\hfill \end{array}$$

(3)

$$\begin{array}{cc}\hfill {RAM\text{\_}Proxy}_t=& {\propto }_0+{\propto }_1{CSR\text{\_}REPORT}_t+{\propto }_2{ASSURED\text{\_}NONBIG4}_t+{\propto }_3{ASSURED\text{\_}BIG4}_t\hfill \\ & +{\propto }_4{GOV\text{\_}DISC\text{\_}SCORE}_t+{\propto }_5{{CSRREPORT}_t\ast GOV\text{\_}DISC\text{\_}SCORE}_t+{\propto }_{6}ASSURED{\text{\_}NONBIG4}_t\hfill \\ & \ast {GOV\text{\_}DISC\text{\_}SCORE}_t+{\propto }_{7}A{SSURED\text{\_}BIG4}_t\ast {GOV\text{\_}DISC\text{\_}SCORE}_t+{\propto }_{8}HIGH{\text{\_}GOV\text{\_}SCORE}_t\hfill \\ & +{\propto }_{9}LOW{\text{\_}CONCERN}_t+{\propto }_{10}SALES\text{\_}GROWTH+{\propto }_{11}SALES\text{\_}VOLATILITY\hfill \\ & +{\propto }_{12}CFO\text{\_}VOLATILITY+{\propto }_{13}OPERATING\text{\_}CYCLE+{\propto }_{14}{MB}_{t-1}+{\propto }_{15}{LEV}_{t-1}+{\propto }_{16}{EO}_t\hfill \\ & +{\propto }_{17}R{D\text{\_}INT}_t+{\propto }_{18}A{D\text{\_}IND\text{\_}INT}_t+{\propto }_{19}{ADMIRED}_t+{\propto }_{20}{MISS\text{\_}SALES\text{\_}GROWTH}_t\hfill \\ & +{\propto }_{21}{MISS\text{\_}SALES\text{\_}VOLATILITY}_t+{\propto }_{22}{MISS\text{\_}CFO\text{\_}VOLATILITY}_t\hfill \\ & +{\propto }_{23}{MISS\text{\_}OPERATING\text{\_}CYCLE}_t+{\propto }_{24}{MISS\text{\_}GOV\text{\_}DISC\text{\_}SCORE}_t+{\epsilon }_t\hfill \end{array}$$

(4)

Variables are defined in Supplementary Materials S1.

The Global Sustainability Standards Board (GSSB) is an independent entity affiliated with the Global Reporting Initiative (GRI), responsible for establishing globally recognized sustainability reporting standards. The GSSB provides a list of national policies requiring assurance on non-financial information (GSSB, 2018), revealing that such assurance is mandatory only in France and South Africa. Consequently, we exclude firms from these two countries from our empirical models.

In our regression analyses, we incorporate firm-specific control variables associated with earnings quality, drawing from Kim et al. (2012) and other earnings management literature (Ashbaugh et al., 2003; Lai & Gul, 2008; Ettredge et al., 2014). These controls include sales growth, sales volatility, cash flow volatility, operating cycle, market-to-book ratio, leverage, equity offerings, R&D intensity, industry advertising intensity, and an indicator for “ADMIRED” firms. Variable definitions are provided in Supplementary Materials S1.

Since obtaining assurance on CSR reports is voluntary in the remaining countries in our sample, we apply the Heckman (1979) two-step procedure to correct for potential self-selection bias, which may otherwise distort parameter estimates. The earnings management Equations (1)–(4) represent the second step of this procedure. In the first step, we estimate a selection model based on Simnett et al. (2009), using all public companies from the GRI list that have ever sought CSR assurance.

We define EVERASSURED as an indicator variable equal to 1 from the first year a company obtains CSR assurance and for every subsequent year, and 0 otherwise. Following Simnett et al. (2009), we classify countries as stakeholder-oriented or shareholder-oriented (STAKE_SHARE, see Supplementary Materials S1 for definitions). Simnett et al. (2009, p. 944) find that firms in stakeholder-oriented countries are more likely to seek assurance due to diverse stakeholder groups exerting greater pressure for transparency.

To control for industry-specific variations in assurance demand, we incorporate industry-fixed effects. Additionally, we include variables likely to influence assurance demand:

• Return on Assets (ADJ_ROA)—More profitable firms may have greater resources to seek assurance.
• Financial Statement Auditor Size (FS_BIG)—Firms audited by Big Four auditors may also seek high-quality non-financial reporting.
• Firm Size (SIZE)—Larger firms face greater public scrutiny and are more likely to have the financial capacity to obtain assurance.

We also include CONCERN, an indicator variable for firms with an ESG score above the median, representing low concern/high ESG firms. Variable definitions are detailed in Supplementary Materials S1.

4. Results

The Relationship Between CSR Assurance and Accrual-Based Earnings Management.

Table 2. Accruals-Based Earnings Management Models. Regressions of ABS_DA Accruals-Based Earnings Management.

Dependent Variable: ABS_DA Expected Sign	Model 1 −	Model 2 −	Model 3 −	Model 4 −
CSRREPORT	−0.01347		−0.01063	−0.01069
	(−1.32025)		(−1.03357)	(−1.03896)
ASSURED		−0.00512 ***	−0.00489 ***
		(−2.89625)	(−2.75166)
ASSURED_NONACCT				−0.00520 ***
				(−2.67603)
ASSURED_ACCT				−0.00442 **
				(−2.03360)
GOV_DISC_SCORE	−0.00031	−0.00010	−0.00032	−0.00032
	(−1.34455)	(−0.57753)	(−1.39702)	(−1.39733)
CSRREPORT × GOV_DISC_SCORE	0.00035 *		0.00030	0.00030
	(1.78024)		(1.45916)	(1.46295)
ASSURED × GOV_DISC_SCORE		0.00012 ***	0.00010 ***
		(3.32885)	(2.63903)
ASSURED_NONACCT × GOV_DISC_SCORE				0.00011 ***
				(2.60141)
ASSURED_ACCT × GOV_DISC_SCORE				0.00008 *
				(1.82982)
HI_GOV_DISC_SCORE	−0.00176	−0.00198	−0.00165	−0.00164
	(−0.83107)	(−0.93541)	(−0.77726)	(−0.77416)
CONCERN	−0.00501 **	−0.00499 **	−0.00501 **	−0.00502 **
	(−2.13403)	(−2.11538)	(−2.12307)	(−2.12939)
SALES_GROWTH	0.00920 ***	0.00973 ***	0.00926 ***	0.00925 ***
	(2.94395)	(3.10446)	(2.96499)	(2.96384)
SALES_VOLATILITY	−0.00000 ***	−0.00000 ***	−0.00000 ***	−0.00000 ***
	(−3.22694)	(−3.16893)	(−3.24485)	(−3.24930)
CFO_VOLATILITY	0.00000 *	0.00000 *	0.00000 *	0.00000 *
	(1.79295)	(1.68883)	(1.84576)	(1.84991)
OPERATING_CYCLE	0.00007 ***	0.00007 ***	0.00007 ***	0.00007 ***
	(8.88863)	(9.01591)	(8.92467)	(8.91516)
COMBINED_RAM	−0.02480 ***	−0.02526 ***	−0.02503 ***	−0.02505 ***
	(−10.53663)	(−10.72662)	(−10.64723)	(−10.65869)
MB	0.00010 ***	0.00009 ***	0.00010 ***	0.00010 ***
	(4.18135)	(4.10108)	(4.19398)	(4.19557)
LEV	−0.01255 ***	−0.01249 ***	−0.01250 ***	−0.01250 ***
	(−4.52075)	(−4.49738)	(−4.49866)	(−4.49819)
EO	0.00066	0.00066	0.00075	0.00077
	(0.18766)	(0.18839)	(0.21440)	(0.22140)
RD_INT	0.05992 **	0.06445 ***	0.06042 **	0.06043 **
	(2.41049)	(2.59389)	(2.42387)	(2.42441)
AD_IND_INT	0.19767 ***	0.20534 ***	0.19964 ***	0.19914 ***
	(2.69211)	(2.80170)	(2.72470)	(2.71916)
ADMIRED	−0.00143	−0.00149	−0.00130	−0.00131
	(−1.08148)	(−1.12956)	(−0.98337)	(−0.99083)
MISSING_SALES_GROWTH	0.01705 *	0.01853 *	0.01686	0.01675
	(1.66742)	(1.81723)	(1.63958)	(1.62822)
MISSING_SALES_VOLATILITY	−0.03373 ***	−0.03374 ***	−0.03392 ***	−0.03393 ***
	(−9.22485)	(−9.17447)	(−9.53992)	(−9.57455)
MISSING_CFO_VOLATILITY	0.03483 ***	0.03485 ***	0.03503 ***	0.03503 ***
	(9.32913)	(9.28275)	(9.64327)	(9.66547)
MISSING_OPERATING_CYCLE	0.00887 ***	0.00905 ***	0.00889 ***	0.00889 ***
	(5.36867)	(5.52936)	(5.37978)	(5.37685)
MISS_GOV_DISC_SCORE	−0.00038	−0.00363	−0.00101	−0.00098
	(−0.05175)	(−0.51159)	(−0.13439)	(−0.13070)
CONSTANT	0.05409 ***	0.04856 ***	0.05479 ***	0.05480 ***
	(4.47809)	(5.47649)	(4.51666)	(4.51623)
OBSERVATIONS	35,563	35,563	35,563	35,563
CONVERGENCE ACHIEVED	Yes	Yes	Yes	Yes

Note: * p < 0.05; ** p < 0.01; *** p < 0.001. Variables are defined in Supplementary Materials S1. Below the estimated coefficients, t-statistics are presented in parentheses. All test statistics and significance levels are calculated based on standard errors adjusted by clustering at the country, industry, and fiscal year levels.

presents regression results for the absolute value of discretionary accruals (ABS_DA) using the overall sample.

4.1. Overall Sample Results (Table 2)

Results indicate that there is no significant difference between firms issuing a CSR report and those that do not, which supports H1. However, firms that obtain CSR assurance are more likely to engage in ABS_DA than those firms that do not obtain assurance (H2: coeff = −0.00512, p < 0.001), which is contrary to expectations. Similarly, firms with strong governance and an assured CSR report are less likely to manipulate ABS_DA than those with strong governance and no assurance (H3: coeff 0.00010, p < 0.001), which supports H3.

When distinguishing between firms assured by an accounting (ACCT) assurer (H4a) and those assured by a non-accounting (NonACCT) assurer (H4b), results indicate firms with either kind of assurance provider are more likely to manipulate ABS_DA than firms without assurance, thus disproving H4a and H4b. However, governance plays a role: strong governance firms with either ACCT assurance (H5a: coeff 0.00008, p < 0.1), or NonACCT assurance (H5b: coeff 0.00011, p < 0.001) exhibit lower discretionary accruals management, supporting H5a and H5b.

4.2. The Relationship Between CSR Assurance and Real Activities Manipulation

4.2.1. Overall Sample Results (Table 3)

Table 3, panel A presents second-stage regression results for COMBINED_RAM, using the Heckman procedure to correct for self-selection bias in CSR assurance decisions. Findings indicate that there is no significant difference between firms issuing a CSR report and those that do not, which supports H1. However, firms that assure their CSR reports are more likely to manipulate COMBINED_RA than those that do not, which does not support H2 (−0.03295, p < 0.001).

Table 3. Real Activities Manipulation Models.

Panel A: Regressions of Combined Real Activities Manipulation
Dependent Variable: Combined RAM Expected Sign	Model 1 +	Model 2 +	Model 3 +	Model 4 +
CSRREPORT	−0.04917		−0.02930	−0.02917
	(−0.75838)		(−0.45061)	(−0.44835)
ASSURED		−0.03295 ***	−0.03281 ***
		(−3.01698)	(−2.99398)
ASSURED_NONACCT				−0.04772 ***
				(−3.88934)
ASSURED_ACCT				−0.01061
				(−0.79676)
GOV_DISC_SCORE	−0.00069	−0.00096	−0.00083	−0.00081
	(−0.49801)	(−0.95595)	(−0.59734)	(−0.57956)
CSRREPORT × GOV_DISC_SCORE	0.00019		−0.00026	−0.00026
	(0.15344)		(−0.20671)	(−0.20898)
ASSURED × GOV_DISC_SCORE		0.00055 **	0.00072 ***
		(2.41573)	(3.07898)
ASSURED_NONACCT × GOV_DISC_SCORE				0.00102 ***
				(3.96620)
ASSURED_BIG × GOV_DISC_SCORE				0.00026
				(0.89696)
HI_GOV_DISC_SCORE	−0.01820	−0.01710	−0.01708	−0.01711
	(−1.29631)	(−1.21423)	(−1.21750)	(−1.22002)
CONCERN	−0.03813 ***	−0.04035 ***	−0.03750 **	−0.03768 **
	(−2.61054)	(−2.78115)	(−2.54937)	(−2.56285)
SALES_GROWTH	0.05491 ***	0.05139 **	0.05538 ***	0.05580 ***
	(2.64034)	(2.45901)	(2.66533)	(2.68292)
SALES_VOLATILITY	−0.00001 ***	−0.00001 ***	−0.00001 ***	−0.00001 ***
	(−8.38175)	(−8.40894)	(−8.37422)	(−8.41266)
CFO_VOLATILITY	0.00002 ***	0.00002 ***	0.00002 ***	0.00002 ***
	(3.16563)	(3.51076)	(3.23173)	(3.22966)
OPERATING_CYCLE	0.00037 ***	0.00036 ***	0.00037 ***	0.00037 ***
	(6.70527)	(6.66170)	(6.73191)	(6.71401)
MB	0.00130 ***	0.00132 ***	0.00130 ***	0.00130 ***
	(7.73250)	(7.80968)	(7.73426)	(7.74808)
LEV	−0.15158 ***	−0.15175 ***	−0.15109 ***	−0.15124 ***
	(−8.69437)	(−8.65887)	(−8.67177)	(−8.69138)
EO	−0.03121	−0.02985	−0.03058	−0.02940
	(−1.25643)	(−1.20333)	(−1.23564)	(−1.18746)
RD_INT	0.70507 ***	0.67188 ***	0.70731 ***	0.70665 ***
	(3.52302)	(3.37960)	(3.54629)	(3.54654)
AD_IND_INT	2.10479 ***	2.08020 ***	2.11839 ***	2.10076 ***
	(4.31160)	(4.27272)	(4.34031)	(4.30930)
ADMIRED	0.02366 ***	0.02254 ***	0.02324 ***	0.02338 ***
	(4.61785)	(4.36574)	(4.52111)	(4.55320)
MISSING_SALES_GROWTH	−0.20235 ***	−0.21974 ***	−0.20422 ***	−0.20505 ***
	(−3.49146)	(−3.77122)	(−3.54404)	(−3.56353)
MISSING_SALES_VOLATILITY	0.06782	0.06283	0.06651	0.06487
	(0.87372)	(0.82160)	(0.87264)	(0.87231)
MISSING_CFO_VOLATILITY	−0.08545	−0.08000	−0.08404	−0.08225
	(−1.09802)	(−1.04316)	(−1.09964)	(−1.10266)
MISSING_OPERATING_CYCLE	0.04921 ***	0.04759 ***	0.04930 ***	0.04886 ***
	(4.93980)	(4.81224)	(4.95760)	(4.91614)
MISS_GOV_DISC_SCORE	−0.07113	−0.08454 **	−0.07756 *	−0.07694 *
	(−1.56660)	(−1.98540)	(−1.69901)	(−1.68555)
CONSTANT	0.28438 ***	0.25359 ***	0.29009 ***	0.28629 ***
	(3.99934)	(4.90885)	(4.07418)	(4.01488)
OBSERVATIONS	36,184	36,184	36,184	36,184
CONVERGENCE ACHIEVED	Yes	Yes	Yes	Yes
Panel B: Regressions of ABS_CFO Earnings Management
Dependent Variable: ABS_CFO Expected Sign	Model 1 +	Model 2 +	Model 3 +	Model 4 +
CSRREPORT	−0.03295		−0.02752	−0.02851
	(−1.50539)		(−1.24812)	(−1.29103)
ASSURED		−0.00883 **	−0.00837 **
		(−2.30835)	(−2.17750)
ASSURED_NONACCT				−0.00859 **
				(−2.01151)
ASSURED_ACCT				−0.00810 *
				(−1.77190)
GOV_DISC_SCORE	−0.00106 **	−0.00070 **	−0.00112 **	−0.00113 **
	(−2.21921)	(−2.01395)	(−2.36332)	(−2.37632)
CSRREPORT × GOV_DISC_SCORE	0.00067		0.00053	0.00054
	(1.57467)		(1.22633)	(1.25796)
ASSURED × GOV_DISC_SCORE		0.00023 ***	0.00022 ***
		(2.74097)	(2.58122)
ASSURED_NONACCT × GOV_DISC_SCORE				0.00026 ***
				(2.76450)
ASSURED_ACCT × GOV_DISC_SCORE				0.00015
				(1.45878)
HI_GOV_DISC_SCORE	0.00570	0.00562	0.00623	0.00639
	(1.18933)	(1.16842)	(1.29945)	(1.33416)
CONCERN	−0.02225 ***	−0.02243 ***	−0.02187 ***	−0.02207 ***
	(−4.39034)	(−4.38811)	(−4.29047)	(−4.32531)
CONTROLS INCLUDED	Yes	Yes	Yes	Yes
OBSERVATIONS	36,227	36,227	36,227	36,227
CONVERGENCE ACHIEVED	Yes	Yes	Yes	Yes
Panel C: Regressions of ABS_PROD Earnings Management
Dependent Variable: ABS_PROD Expected Sign	Model 1 −	Model 2 −	Model 3 −	Model 4 −
CSRREPORT	0.03120		0.02283	0.02302
	(1.04994)		(0.76364)	(0.76935)
ASSURED		0.01335 **	0.01318 **
		(2.57565)	(2.52366)
ASSURED_NONACCT				0.02125 ***
				(3.65052)
ASSURED_ACCT				0.00117
				(0.18768)
GOV_DISC_SCORE	−0.00040	−0.00037	−0.00031	−0.00032
	(−0.61773)	(−0.78634)	(−0.48259)	(−0.49961)
CSRREPORT × GOV_DISC_SCORE	−0.00022		−0.00002	−0.00002
	(−0.39441)		(−0.03158)	(−0.03637)
ASSURED × GOV_DISC_SCORE		−0.00024 **	−0.00032 ***
		(−2.18857)	(−2.89376)
ASSURED_NONACCT × GOV_DISC_SCORE				−0.00049 ***
				(−4.06058)
ASSURED_ACCT × GOV_DISC_SCORE				−0.00005
				(−0.39451)
HI_GOV_DISC_SCORE	0.01671 **	0.01623 **	0.01602 **	0.01599 **
	(2.55077)	(2.47466)	(2.44675)	(2.44314)
CONCERN	0.00898	0.01009	0.00851	0.00866
	(1.27797)	(1.44883)	(1.20383)	(1.22650)
CONTROLS INCLUDED	Yes	Yes	Yes	Yes
OBSERVATIONS	36,492	36,492	36,492	36,492
CONVERGENCE ACHIEVED	Yes	Yes	Yes	Yes
Panel D: Regressions of ABS_EXP Earnings Management
Dependent Variable: ABS_EXP Expected Sign	Model 1 +	Model 2 +	Model 3 +	Model 4 +
CSRREPORT	0.01532		0.02169	0.02324
	(0.42478)		(0.60113)	(0.64393)
ASSURED		−0.01180 *	−0.01255 **
		(−1.95063)	(−2.07025)
ASSURED_NONACCT				−0.01910 ***
				(−2.83078)
ASSURED_ACCT				−0.00271
				(−0.35548)
GOV_DISC_SCORE	−0.00018	−0.00066	−0.00015	−0.00013
	(−0.23295)	(−1.17769)	(−0.19400)	(−0.16534)
CSRREPORT × GOV_DISC_SCORE	−0.00060		−0.00070	−0.00072
	(−0.82016)		(−0.95572)	(−0.98518)
ASSURED × GOV_DISC_SCORE		0.00014	0.00019
		(1.00947)	(1.40966)
ASSURED_NONACCT × GOV_DISC_SCORE				0.00027 *
				(1.80233)
ASSURED_ACCT × GOV_DISC_SCORE				0.00008
				(0.47087)
HI_GOV_DISC_SCORE	−0.00524	−0.00459	−0.00547	−0.00573
	(−0.68513)	(−0.59944)	(−0.71761)	(−0.75235)
CONCERN	−0.01088	−0.01075	−0.01129	−0.01106
	(−1.31298)	(−1.31106)	(−1.35575)	(−1.33171)
CONTROLS INCLUDED	Yes	Yes	Yes	Yes
OBSERVATIONS	36,563	36,563	36,563	36,563
CONVERGENCE ACHIEVED	Yes	Yes	Yes	Yes

Note: * p < 0.05; ** p < 0.01; *** p < 0.001. Variables are defined in Supplementary Materials S1. Additional controls include sales growth, sales volatility, CFO volatility, operating cycle, combined RAM, MB, LEV, EO, RD_INT, AD_IND_INT, ADMIRED, missing sales growth, missing sales volatility, missing operating cycle, and missing governance disclosure score. Below the estimated coefficients, t-statistics are presented in parentheses. All test statistics and significance levels are calculated based on standard errors adjusted by clustering at the country, industry, and fiscal year levels.

Table 3. Real Activities Manipulation Models.

Panel A: Regressions of Combined Real Activities Manipulation
Dependent Variable: Combined RAM Expected Sign	Model 1 +	Model 2 +	Model 3 +	Model 4 +
CSRREPORT	−0.04917		−0.02930	−0.02917
	(−0.75838)		(−0.45061)	(−0.44835)
ASSURED		−0.03295 ***	−0.03281 ***
		(−3.01698)	(−2.99398)
ASSURED_NONACCT				−0.04772 ***
				(−3.88934)
ASSURED_ACCT				−0.01061
				(−0.79676)
GOV_DISC_SCORE	−0.00069	−0.00096	−0.00083	−0.00081
	(−0.49801)	(−0.95595)	(−0.59734)	(−0.57956)
CSRREPORT × GOV_DISC_SCORE	0.00019		−0.00026	−0.00026
	(0.15344)		(−0.20671)	(−0.20898)
ASSURED × GOV_DISC_SCORE		0.00055 **	0.00072 ***
		(2.41573)	(3.07898)
ASSURED_NONACCT × GOV_DISC_SCORE				0.00102 ***
				(3.96620)
ASSURED_BIG × GOV_DISC_SCORE				0.00026
				(0.89696)
HI_GOV_DISC_SCORE	−0.01820	−0.01710	−0.01708	−0.01711
	(−1.29631)	(−1.21423)	(−1.21750)	(−1.22002)
CONCERN	−0.03813 ***	−0.04035 ***	−0.03750 **	−0.03768 **
	(−2.61054)	(−2.78115)	(−2.54937)	(−2.56285)
SALES_GROWTH	0.05491 ***	0.05139 **	0.05538 ***	0.05580 ***
	(2.64034)	(2.45901)	(2.66533)	(2.68292)
SALES_VOLATILITY	−0.00001 ***	−0.00001 ***	−0.00001 ***	−0.00001 ***
	(−8.38175)	(−8.40894)	(−8.37422)	(−8.41266)
CFO_VOLATILITY	0.00002 ***	0.00002 ***	0.00002 ***	0.00002 ***
	(3.16563)	(3.51076)	(3.23173)	(3.22966)
OPERATING_CYCLE	0.00037 ***	0.00036 ***	0.00037 ***	0.00037 ***
	(6.70527)	(6.66170)	(6.73191)	(6.71401)
MB	0.00130 ***	0.00132 ***	0.00130 ***	0.00130 ***
	(7.73250)	(7.80968)	(7.73426)	(7.74808)
LEV	−0.15158 ***	−0.15175 ***	−0.15109 ***	−0.15124 ***
	(−8.69437)	(−8.65887)	(−8.67177)	(−8.69138)
EO	−0.03121	−0.02985	−0.03058	−0.02940
	(−1.25643)	(−1.20333)	(−1.23564)	(−1.18746)
RD_INT	0.70507 ***	0.67188 ***	0.70731 ***	0.70665 ***
	(3.52302)	(3.37960)	(3.54629)	(3.54654)
AD_IND_INT	2.10479 ***	2.08020 ***	2.11839 ***	2.10076 ***
	(4.31160)	(4.27272)	(4.34031)	(4.30930)
ADMIRED	0.02366 ***	0.02254 ***	0.02324 ***	0.02338 ***
	(4.61785)	(4.36574)	(4.52111)	(4.55320)
MISSING_SALES_GROWTH	−0.20235 ***	−0.21974 ***	−0.20422 ***	−0.20505 ***
	(−3.49146)	(−3.77122)	(−3.54404)	(−3.56353)
MISSING_SALES_VOLATILITY	0.06782	0.06283	0.06651	0.06487
	(0.87372)	(0.82160)	(0.87264)	(0.87231)
MISSING_CFO_VOLATILITY	−0.08545	−0.08000	−0.08404	−0.08225
	(−1.09802)	(−1.04316)	(−1.09964)	(−1.10266)
MISSING_OPERATING_CYCLE	0.04921 ***	0.04759 ***	0.04930 ***	0.04886 ***
	(4.93980)	(4.81224)	(4.95760)	(4.91614)
MISS_GOV_DISC_SCORE	−0.07113	−0.08454 **	−0.07756 *	−0.07694 *
	(−1.56660)	(−1.98540)	(−1.69901)	(−1.68555)
CONSTANT	0.28438 ***	0.25359 ***	0.29009 ***	0.28629 ***
	(3.99934)	(4.90885)	(4.07418)	(4.01488)
OBSERVATIONS	36,184	36,184	36,184	36,184
CONVERGENCE ACHIEVED	Yes	Yes	Yes	Yes
Panel B: Regressions of ABS_CFO Earnings Management
Dependent Variable: ABS_CFO Expected Sign	Model 1 +	Model 2 +	Model 3 +	Model 4 +
CSRREPORT	−0.03295		−0.02752	−0.02851
	(−1.50539)		(−1.24812)	(−1.29103)
ASSURED		−0.00883 **	−0.00837 **
		(−2.30835)	(−2.17750)
ASSURED_NONACCT				−0.00859 **
				(−2.01151)
ASSURED_ACCT				−0.00810 *
				(−1.77190)
GOV_DISC_SCORE	−0.00106 **	−0.00070 **	−0.00112 **	−0.00113 **
	(−2.21921)	(−2.01395)	(−2.36332)	(−2.37632)
CSRREPORT × GOV_DISC_SCORE	0.00067		0.00053	0.00054
	(1.57467)		(1.22633)	(1.25796)
ASSURED × GOV_DISC_SCORE		0.00023 ***	0.00022 ***
		(2.74097)	(2.58122)
ASSURED_NONACCT × GOV_DISC_SCORE				0.00026 ***
				(2.76450)
ASSURED_ACCT × GOV_DISC_SCORE				0.00015
				(1.45878)
HI_GOV_DISC_SCORE	0.00570	0.00562	0.00623	0.00639
	(1.18933)	(1.16842)	(1.29945)	(1.33416)
CONCERN	−0.02225 ***	−0.02243 ***	−0.02187 ***	−0.02207 ***
	(−4.39034)	(−4.38811)	(−4.29047)	(−4.32531)
CONTROLS INCLUDED	Yes	Yes	Yes	Yes
OBSERVATIONS	36,227	36,227	36,227	36,227
CONVERGENCE ACHIEVED	Yes	Yes	Yes	Yes
Panel C: Regressions of ABS_PROD Earnings Management
Dependent Variable: ABS_PROD Expected Sign	Model 1 −	Model 2 −	Model 3 −	Model 4 −
CSRREPORT	0.03120		0.02283	0.02302
	(1.04994)		(0.76364)	(0.76935)
ASSURED		0.01335 **	0.01318 **
		(2.57565)	(2.52366)
ASSURED_NONACCT				0.02125 ***
				(3.65052)
ASSURED_ACCT				0.00117
				(0.18768)
GOV_DISC_SCORE	−0.00040	−0.00037	−0.00031	−0.00032
	(−0.61773)	(−0.78634)	(−0.48259)	(−0.49961)
CSRREPORT × GOV_DISC_SCORE	−0.00022		−0.00002	−0.00002
	(−0.39441)		(−0.03158)	(−0.03637)
ASSURED × GOV_DISC_SCORE		−0.00024 **	−0.00032 ***
		(−2.18857)	(−2.89376)
ASSURED_NONACCT × GOV_DISC_SCORE				−0.00049 ***
				(−4.06058)
ASSURED_ACCT × GOV_DISC_SCORE				−0.00005
				(−0.39451)
HI_GOV_DISC_SCORE	0.01671 **	0.01623 **	0.01602 **	0.01599 **
	(2.55077)	(2.47466)	(2.44675)	(2.44314)
CONCERN	0.00898	0.01009	0.00851	0.00866
	(1.27797)	(1.44883)	(1.20383)	(1.22650)
CONTROLS INCLUDED	Yes	Yes	Yes	Yes
OBSERVATIONS	36,492	36,492	36,492	36,492
CONVERGENCE ACHIEVED	Yes	Yes	Yes	Yes
Panel D: Regressions of ABS_EXP Earnings Management
Dependent Variable: ABS_EXP Expected Sign	Model 1 +	Model 2 +	Model 3 +	Model 4 +
CSRREPORT	0.01532		0.02169	0.02324
	(0.42478)		(0.60113)	(0.64393)
ASSURED		−0.01180 *	−0.01255 **
		(−1.95063)	(−2.07025)
ASSURED_NONACCT				−0.01910 ***
				(−2.83078)
ASSURED_ACCT				−0.00271
				(−0.35548)
GOV_DISC_SCORE	−0.00018	−0.00066	−0.00015	−0.00013
	(−0.23295)	(−1.17769)	(−0.19400)	(−0.16534)
CSRREPORT × GOV_DISC_SCORE	−0.00060		−0.00070	−0.00072
	(−0.82016)		(−0.95572)	(−0.98518)
ASSURED × GOV_DISC_SCORE		0.00014	0.00019
		(1.00947)	(1.40966)
ASSURED_NONACCT × GOV_DISC_SCORE				0.00027 *
				(1.80233)
ASSURED_ACCT × GOV_DISC_SCORE				0.00008
				(0.47087)
HI_GOV_DISC_SCORE	−0.00524	−0.00459	−0.00547	−0.00573
	(−0.68513)	(−0.59944)	(−0.71761)	(−0.75235)
CONCERN	−0.01088	−0.01075	−0.01129	−0.01106
	(−1.31298)	(−1.31106)	(−1.35575)	(−1.33171)
CONTROLS INCLUDED	Yes	Yes	Yes	Yes
OBSERVATIONS	36,563	36,563	36,563	36,563
CONVERGENCE ACHIEVED	Yes	Yes	Yes	Yes

Note: * p < 0.05; ** p < 0.01; *** p < 0.001. Variables are defined in Supplementary Materials S1. Additional controls include sales growth, sales volatility, CFO volatility, operating cycle, combined RAM, MB, LEV, EO, RD_INT, AD_IND_INT, ADMIRED, missing sales growth, missing sales volatility, missing operating cycle, and missing governance disclosure score. Below the estimated coefficients, t-statistics are presented in parentheses. All test statistics and significance levels are calculated based on standard errors adjusted by clustering at the country, industry, and fiscal year levels.

Firms with strong governance and CSR assurance are less likely to engage in COMBINED_RAM compared to firms with strong governance and no assurance (H3: coeff = 0.00072, p < 0.001), which supports H3. ACCT-assured firms (H4a) do not differ significantly from firms that have no assurance, which does not support H4a. But, NonACCT-assured (H4b) firms are more likely to engage in COMBINED_RAM than firms that do not seek assurance, which does not support H4b. However, strong governance firms with ACCT assurance (H5a: coeff = 0.00008, p < 0.1) and strong governance firms with NonACCT assurance (H5b: coeff = 0.00011, p < 0.001) are less likely to engage in real activities manipulation than non-assured firms, which supports H5a and H5b.

4.2.2. Abnormal Cash Flows (AB_CFO) and CSR Assurance (Table 3, Panel B)

For the overall sample (Table 3, Panel B), there is no significant difference in AB_CFO between firms that issue CSR reports and those that do not (H1), which supports H1. The coefficient of ASSURED (coeff −0.00883, p < 0.01) shows that firms that have assurance on their reports are more likely to manage earnings, which does not support H2. However, firms with strong governance and CSR assurance are less likely to manipulate abnormal cash flows (H3: coeff = 0.00022, p < 0.001) than firms with strong governance and no assurance, which supports H3.

Both firms with ACCT assurance (H4a: coeff = −0.0081, p < 0.1) and NonACCT assurance are more likely to manipulate AB_CFO (H4b: coeff = −0.00859, p < 0.01), which does not support H4a or H4b.

There is no significant difference between firms with ACCT assurance and strong governance and companies without assurance, which disproves H5a. Firms with NonACCT assurance and strong governance are less likely to manipulate AB_CFO (H5b: coeff = 0.00026, p < 0.001), which supports H5b.

4.2.3. Abnormal Production Costs (AB_PROD) and CSR Assurance (Table 3, Panel C)

In the overall sample (Table 3, Panel C), there is no significant difference between firms issuing CSR reports and those that do not, which supports H1. Firms with CSR assurance are more likely to manipulate ABS_PROD than those firm that do not have assurance, which does not support H2. Firms with strong governance and CSR assurance are less likely to manipulate AB_PROD (H3: coeff = −0.00024, p < 0.01), supporting H3.

No significant differences exist between ACCT-assured firms and non-assured firms (H4a). Firms with NonACCT assurance are more likely to manipulate ABS_PROD than firms without assurance (H4b: coeff 0.02125, p < 0.001). Thus, our results do not support H4a and H4b. There are no significant differences between firms with strong governance and ACCT assurance and firms without assurance, which does not support H5a. However, strong governance firms with NonACCT assurance are less likely to manage AB_PROD (H5b: coeff = −0.00049, p < 0.001).

4.2.4. Abnormal Discretionary Expenses (AB_EXP) and CSR Assurance (Table 3, Panel D)

In the overall sample (Table 3, Panel D), there is no significant difference between firms issuing CSR reports and those that do not, which supports H1. However, firms with CSR assurance (H2: coeff −0.01180, p < 0.1) are more likely to manipulate AB_EXP than those with no assurance, which is contrary to expectations. Firms with strong governance with assurance (H3) are not significantly different from firms that have strong governance but no assurance, which does not support H3.

Firms with ACCT assurance are not significantly different from firms without assurance, and firms with NonACCT assurance manage AB_EXP more than non-assured firms (H4b: coeff = −0.01910, p < 0.1), which does not support H4a and H4b. Additionally, firms with strong governance and NonACCT assurance manipulate AB_EXP less than non-assured firms (H5b: coeff = 0.00027, p < 0.1).

4.3. Additional Analyses

Simultaneous Equation Model of Discretionary Accruals Management and Real Activities Manipulation.

Since decisions regarding discretionary accruals management and real activities manipulation are likely made simultaneously, we employ a simultaneous equations model to capture these interrelated choices.

Table 4. Simultaneous Estimation of Accruals Earnings Management and Real Activities Manipulation.

Dependent Variable	Model 3 ABS_DA	Model 3 COMB_RAM	Model 4 ABS_DA	Model 4 COMB_RAM
CSRREPORT	−0.01056	−0.02905	−0.01062	−0.02891
	(−1.02626)	(−0.44668)	(−1.03116)	(−0.44438)
ASSURED	−0.00487 ***	−0.03283 ***
	(−2.74248)	(−2.99500)
ASSURED_NONACCT			−0.00519 ***	−0.04772 ***
			(−2.66772)	(−3.88963)
ASSURED_ACCT			−0.00440 **	−0.01064
			(−2.02541)	(−0.79847)
GOV_DISC_SCORE	−0.00033	−0.00085	−0.00033	−0.00082
	(−1.45288)	(−0.60805)	(−1.45328)	(−0.59031)
CSRREPORT × GOV_DISC_SCORE	0.00029	−0.00024	0.00029	−0.00025
	(1.45220)	(−0.19491)	(1.45566)	(−0.19713)
ASSURED × GOV_DISC_SCORE	0.00010 ***	0.00072 ***
	(2.64938)	(3.07877)
ASSURED_NONACCT × GOV_DISC_SCORE			0.00011 ***	0.00102 ***
			(2.60771)	(3.96514)
ASSURED_ACCT × GOV_DISC_SCORE			0.00009 *	0.00026
			(1.84290)	(0.89802)
HI_GOV_DISC_SCORE	−0.00144	−0.01698
	(−0.67921)	(−1.21040)
CONCERN	−0.00525 **	−0.03743 **	−0.00527 **	−0.03761 **
	(−2.23557)	(−2.54470)	(−2.24203)	(−2.55813)
CONTROLS INCLUDED	Yes	Yes	Yes	Yes
OBSERVATIONS	36,563	36,563	36,563	36,563
CONVERGENCE ACHIEVED	Yes	Yes	Yes	Yes

Note: * p < 0.05; ** p < 0.01; *** p < 0.001. Variables are defined in Supplementary Materials S1. Additional controls include sales growth, sales volatility, CFO volatility, operating cycle, combined RAM, MB, LEV, EO, RD_INT, AD_IND_INT, ADMIRED, missing sales growth, missing sales volatility, missing operating cycle, and missing governance disclosure score. Below the estimated coefficients, t-statistics are presented in parentheses. All test statistics and significance levels are calculated based on standard errors adjusted by clustering at the country, industry, and fiscal year levels.

presents the results for absolute discretionary accruals (ABS_DA) and combined real activities manipulation (COMBINED_RAM).

We find that CSRREPORT is neither significant for ABS_DA nor for COMBINED_RAM. This suggests that firms are indifferent to the form of earnings management.

The coefficient on ASSURED is significantly negative for both ABS_DA and COMBINED_RAM (H2), indicating that firms that have assurance are more likely to engage in ABS_DA and COMBINED_RAM, while ASSURED*GOV_DISC_SCORE is positive and significant for ABS_DA (coeff 0.00010, p < 0.001) and COMBINED_RAM (coeff 0.00072, p < 0.001), suggesting that firms with strong governance and assurance are less likely to engage in real activities manipulation and accrual-based management.

Coefficients for NONACCT assurance (H4b) are both negative and significant, indicating that firms that have assurance from NONAACT assurance providers are more likely to engage in ABS_DA and COMBINED_RAM. The coefficient on ACCT assurance (H4a) is negative and significant for ABS_DA (coeff −0.00440, p < 0.01), but not significant for COMBINED_RAM, which suggests that firms that are assured by ACCT assurance providers prefer to engage in ABS_DA. The coefficient on NonACCT assurance*GOV_DISC_SCORE (H5b) is positive and significant for ABS_DA (coeff 0.00011, p < 0.001) and for COMBINED_RAM (coeff = 0.00102, p < 0.001), suggesting that firms with strong governance and assurance from NONACCT assurance providers are less likely to engage in real and accrual-based earnings management. The coefficient on ACCT assurance*GOV_DISC_SCORE (H5a) is positive and significant (coeff 0.0009, p < 0.1), suggesting that firms with strong governance and assurance from ACCT assurance providers are less likely to engage in accrual-based earnings management. Our results indicate that assurance from NONACCT assurance providers reinforces strong governance, and such firms are less likely to engage in either ABS_DA or COMBINED_RAM.

5. Discussion

Discretionary accruals (ABS_DA) provide flexibility in financial reporting, but this flexibility can be exploited to distort earnings and financial stability. A common misuse is earnings smoothing, where firms shift income between periods to mask volatility—either improving weak quarters or deferring excess earnings from strong ones (Lobo & Zhou, 2010). Research has shown that firms often use discretionary accruals (ABS_DA) to meet or slightly exceed analyst forecasts, a pattern suggestive of earnings management.

A well-known example is WorldCom, which capitalized routine operating expenses—specifically line costs—to spread them over multiple periods rather than expensing them immediately. This manipulation inflated profits by an estimated USD 11 billion and contributed to the company’s eventual collapse. Similarly, AIG managed earnings by adjusting “loss reserves”, exploiting the discretion inherent in estimating future insurance claims. This led to a restatement of five years of financial statements and a USD 1.64 billion settlement (Harrington, 2009).

Discretionary accruals (ABS_DA) also tend to spike when executive compensation is tied to earning thresholds, raising concerns about personal incentive-driven manipulation. At UnitedHealth Group, discretionary accruals (ABS_DA) accompanied the backdating of stock options, helping the company achieve earnings levels that triggered executive bonuses. The scandal resulted in CEO William McGuire surrendering USD 620 million in stock gains and payments.

Unlike discretionary accruals (ABS_DA), RAM involves altering actual business operations to achieve desired financial outcomes. Because these actions have real economic consequences, they are often more damaging in the long term and raise serious concerns for investors and regulators.

Examples include the following:

Sales Manipulation: Firms may offer steep, short-term discounts to inflate sales figures before the end of reporting periods. At Sunbeam, CEO Al Dunlap employed this tactic to pull future sales into the current quarter, creating unsustainable performance spikes followed by sharp declines (Sweeney, 2000).

Timing of Asset Sales: Companies may time the sale of assets to recognize gains in strategically beneficial reporting periods. Enron infamously structured such transactions with related entities to meet quarterly targets.

Abnormal Operating Cash Flows (AB_CFO): Firms like Enron and WorldCom reported strong earnings while exhibiting persistently negative abnormal cash flows—an early red flag noted by sophisticated investors. In contrast, companies like Tesla have shown similar cash flow patterns due to aggressive expansion strategies, which markets may tolerate when aligned with communicated growth plans.

Abnormal Discretionary Expenses (AB_CFO (AB_EXP): Firms preparing for IPOs, such as Groupon and Uber, have been observed to reduce discretionary spending—especially in marketing and R&D—prior to going public. This strategy temporarily enhances margins and can support higher valuations during the offering process.

Abnormal Production Costs (AB_PROD): Manufacturing firms may deviate from expected production cost patterns to manipulate reported earnings. These deviations can signal deeper operational issues. For example, Toyota experienced unusually high production costs during the 2009–2010 acceleration crisis. Rising warranty claims and rework costs, which preceded public disclosure, were early indicators of systemic quality failures.

Together, these examples illustrate how both accrual-based and real activity-based earnings management strategies can distort financial performance, mislead stakeholders, and in some cases, foreshadow broader organizational crises.

6. Conclusions

This study examines the relationship between financial reporting transparency and CSR reports, focusing on CSR assurance, the type of assurance provider, and the interaction between assurance and governance. CSR report assurance offers a unique setting to explore the voluntary assurance of non-financial information and its potential spillover effects on financial reporting outcomes. Using a dataset of over 36,000 firm-year observations from 34 countries (2009–2015), we investigate whether firms that seek CSR assurance engage in lower accrual-based earnings management and real activities manipulation.

Our findings indicate that issuing a CSR report alone is not associated with lower accrual-based earnings management. CSR assurance, whether provided by ACCT or NonACCT assurers, does not significantly enhance financial transparency. We find no evidence that assurance reduces accrual-based earnings management and only weak evidence that it curbs real activities manipulation. However, governance reinforces CSR assurance from NONACCT assurance providers in reducing earnings management.

Interestingly, NonACCT assurance along with strong governance is associated with reduced accruals manipulation (ABS_DA) and real activities manipulation (COMBINED_RAM, ABS_CFO, AB_PROD, AB_EXP), implying firms that have both are more financially transparent and are likely working towards being better corporate citizens.

Governance plays a crucial role in mitigating combined real activities manipulation and abnormal production expenses, particularly when assurance is provided by NonACCT assurers. This suggests that while CSR assurance alone does not ensure higher financial reporting transparency, strong governance can enhance its effectiveness.

Overall, our results show that issuing a CSR report, by itself, does not significantly impact earnings management. Assurance of CSR reports, particularly by ACCT providers, does not reduce—and may even increase—earnings manipulation. Strong corporate governance plays a crucial role in limiting earnings management. The most consistent improvements in earnings quality occur when firms combine strong governance with CSR assurance from a non-accounting provider. The combination of strong governance and NonACCT assurance appears to be mutually reinforcing, suggesting a symbolic legitimacy strategy that is also substantively effective. NonACCT assurance serves as a complementary external validation mechanism, while governance strengthens internal reporting controls. This interplay signals to stakeholders a commitment to financial transparency and ESG accountability.

Our study provides initial evidence on the relationship between CSR report assurance and earnings quality. Future research could explore additional characteristics of CSR assurance, such as scope, level, and assurance standards, and investigate broader benefits of assurance beyond financial reporting quality. While our findings suggest that CSR reports—when assured by NONACCT assurance providers—are linked to financial reporting outcomes, they also indicate that CSR assurance and governance reinforce each other in influencing financial reporting quality.