CEO Dynamics and Real Earnings Management: A Gender Diversity Perspective from Sub-Saharan Africa

Abstract

Sub-Saharan Africa’s (SSA) corporate environment, like many emerging markets, is marked by institutional voids, weak oversight structures, and patriarchal leadership norms, which heighten the risk of real earnings management (REM). This study examines how CEO characteristics and audit committee gender diversity influence REM among listed manufacturing firms in 12 SSA countries from 2012 to 2023. Anchored in agency theory and Upper Echelon Theory, this study draws on 1189 firm-year observations and employs Pooled OLS, Random Effects, Fixed Effects, Feasible Generalised Least Squares (FGLS), and System GMM estimators. Findings show that female CEOs are consistently associated with lower REM, underscoring the ethical conservatism linked to gender-inclusive leadership. CEO ownership shows a positive and significant association with REM in System GMM, though findings vary across models, indicating potential institutional effects. The firm size is negatively and significantly related to REM in Pooled, RE, and FGLS models, but becomes nonsignificant in FE and System GMM, suggesting the role of external scrutiny may be sensitive to model dynamics. Leverage exhibits a positive and significant relationship with REM in most models, but turns negative and nonsignificant under System GMM, pointing to endogeneity concerns. Interaction effects and country-specific regressions affirm that governance impacts differ across contexts. Policy reforms should prioritise gender-diverse leadership and tailored oversight mechanisms.

1. Introduction

Manufacturing firms in Sub-Saharan Africa (SSA) operate within a uniquely challenging business environment characterised by fragmented markets and limited regional integration. These structural inefficiencies amplify operational difficulties, resulting in increased production costs and reduced global competitiveness (Avordeh et al., 2024). Furthermore, restricted access to international capital markets exacerbates these challenges, limiting firms’ abilities to secure essential financing for expansion and sustainable operations (Didier et al., 2021; Kaur et al., 2023). Facing these pressures, firms may engage in real earnings management (REM), which involves the strategic manipulation of operational activities—such as overproduction, aggressive sales discounting, or cuts in discretionary expenditures—to artificially align financial outcomes with investor expectations (Roychowdhury, 2006). Although such practices can temporarily enhance reported earnings, they often the undermine long-term organisational value by distorting resource allocation and eroding stakeholder trust (B. Ali & Kamardin, 2018; Cohen & Zarowin, 2010; Mensah & Bein, 2023; Qawasmeh & Azzam, 2020). This tension between short-term financial optics and sustainable growth raises critical governance and ethical considerations, particularly in resource-constrained economies.

Within this context, the role of Chief Executive Officers (CEOs) is pivotal in either perpetuating or mitigating REM practices. CEO characteristics—including gender, ownership stakes, tenure, and nationality—significantly influence strategic financial decisions, particularly in environments prone to operational and governance challenges (A. Ali & Zhang, 2015). For instance, female CEOs, who typically exhibit risk-averse leadership styles, are generally less likely to engage in REM and are more inclined to prioritise long-term stability over short-term financial manipulation (Amelia & Eriandani, 2021; Kang et al., 2023). Conversely, CEOs with substantial equity ownership may face conflicting incentives: while higher ownership can align CEO interests with shareholder objectives, potentially reducing REM (Jensen & Meckling, 1976), concentrated equity may also provide the discretionary control needed to manipulate earnings for personal gain (Morck et al., 1988; Roychowdhury, 2006; Shleifer & Vishny, 1997).

The relationship between CEO tenure and REM is similarly nuanced. CEOs with longer tenures often accumulate valuable institutional knowledge and reputational capital, reducing their propensity to engage in REM, provided governance frameworks are robust. Conversely, CEOs with shorter tenures may feel pressured to meet short-term performance targets, increasing the likelihood of earnings manipulation (Qawasmeh & Azzam, 2020). Additionally, CEO nationality introduces further complexity; leaders originating from jurisdictions with stringent regulatory environments may import practices that limit REM, whereas those from regions with weaker oversight might adopt less stringent standards (Khanna et al., 2004; La Porta et al., 1998; Leuz et al., 2003). Thus, personal and positional attributes of CEOs interact intricately with institutional settings, significantly influencing financial reporting outcomes in SSA.

Complementing CEO influence, corporate governance mechanisms, notably gender-diverse audit committees, play a crucial role in curbing REM by imposing rigorous oversight. In SSA, characterised by fragmented institutional oversight and male-dominated governance structures, gender diversity within audit committees emerges as a critical mechanism to enhance financial integrity (Adams & Ferreira, 2009; Lakhal et al., 2015). Research indicates that gender-diverse committees exhibit greater diligence in monitoring managerial activities, demonstrate a propensity for risk-averse decision-making (Levi et al., 2014), and more effectively challenge aggressive accounting practices (Gavious et al., 2012). Such committees prioritise long-term corporate reputations and financial integrity, aligning with agency theory’s assertion that diversified governance structures mitigate managerial opportunism (Fama & Jensen, 1983). However, the effectiveness of these committees may vary due to cultural norms influencing women’s participation and members’ technical expertise in governance oversight (Bear et al., 2010; A. Musa et al., 2023).

Recent scholarship has explored corporate governance drivers of real earnings management (REM), with mixed findings. Nowicki’s (2024) meta-analysis synthesising global studies found that family ownership exhibits a statistically significant, though economically negligible, positive relationship with REM, while CEO experience and external audits show no meaningful association. These results challenge assumptions about the efficacy of traditional governance mechanisms in curbing REM, urging policymakers to look beyond ownership structures and leadership tenure for solutions. Similarly, Alquhaif and Alobaid (2024) examined the interplay between CEO power and audit committee financial expertise in Malaysian firms, finding that powerful CEOs weaken the committee’s effectiveness in curbing real earnings management via accretive share repurchases. In the context of Sub-Saharan Africa (SSA), Gerged et al. (2022) explored the relationship between board gender diversity and corporate environmental disclosure, noting that the presence of women on boards enhances transparency and accountability. Their findings suggest that gender-diverse boards are more diligent in monitoring managerial activities, which could extend to financial reporting practices.

The existing literature on CEO attributes and REM highlights four key research gaps. Firstly, although studies have explored CEO characteristics extensively in developed and Asian markets (A. Ali & Zhang, 2015; Qawasmeh & Azzam, 2020), SSA remains under-represented, despite its unique institutional context (World Bank Group, 2021). Secondly, while gender diversity in audit committees is recognised for its oversight potential (Adams & Ferreira, 2009; A. Musa et al., 2023), its role as a moderator in the CEO-REM relationship remains under-explored, especially in patriarchal governance contexts. Thirdly, the existing research largely focuses on financial and service sectors (Cohen & Zarowin, 2010), overlooking manufacturing industries, where complexities in production processes and supply chain volatility uniquely incentivise operational manipulations. Finally, mixed empirical findings, especially regarding CEO tenure and REM (A. Ali & Zhang, 2015; Qawasmeh & Azzam, 2020), underscore methodological inconsistencies and a lack of contextual granularity, necessitating robust frameworks tailored specifically to SSA’s institutional environment (P. Dechow et al., 2010; Owusu et al., 2020).

To bridge these gaps, this study sets five objectives: (1) to examine how CEO gender impacts REM in SSA-listed manufacturing firms, specifically assessing whether female CEOs’ risk aversion reduces earnings manipulation; (2) to analyse the effect of CEO ownership on REM, exploring whether ownership aligns managerial and shareholder interests or facilitates financial manipulation; (3) to investigate the influence of CEO tenure on REM, evaluating how accumulated knowledge and reputational concerns affect managerial behaviour; (4) to assess the role of CEO nationality in influencing REM through cross-border variations in governance norms; and (5) to evaluate the moderating effect of audit committee gender diversity on the relationship between CEO characteristics and REM.

This study advances academic understandings and practical governance in four pioneering dimensions. Theoretically, within Sub-Saharan Africa (SSA), it is the first to synthesise agency theory (Jensen & Meckling, 1976) and Upper Echelons Theory (Hambrick & Mason, 1984) into a unified framework, resolving the prior fragmentation in governance research. This integration demonstrates how CEO gender, share ownership, nationality, and tenure interact with board-level incentives to shape real earnings management (REM) decisions in SSA’s institutionally distinct setting, advancing theoretical discourse beyond simplistic principal–agent assumptions. Empirically, this research breaks new ground by interrogating REM drivers in SSA’s manufacturing sector, a critical yet understudied region where institutional voids magnify agency conflicts. It provides the first evidence that gender-diverse audit committees significantly moderate the relationship between CEO attributes and REM in SSA. For instance, female audit committee representation neutralises the REM propensity of male CEOs with a high share ownership, a finding that contrasts with studies in gender-egalitarian regions. This study also uncovers paradoxical dynamics: foreign-national CEOs, often assumed to prioritise transparency, engage in more REM in SSA’s opaque environments, while long-tenured CEOs with equity stakes exhibit less short-termism, contradicting Global North patterns linking tenure to entrenchment. These results challenge universalist assumptions about executive behaviour and underscore the centrality of the institutional context.

Policy wise, this study moves beyond generic calls for governance reform by proposing context-tailored interventions: (1) gender quotas for audit committees in SSA, informed by their unique moderating power over CEO gender and nationality effects and (2) regulatory frameworks tying CEO selection to diversity benchmarks (e.g., gender and nationality) and equity alignment criteria rather than purely financial metrics. These recommendations respond to a critical gap in the emerging economy policy discourse, which often adopts Global North templates without a contextual adaptation to factors like localised ownership structures or cross-national leadership dynamics. Practically, the findings equip governance practitioners with actionable tools to recalibrate oversight mechanisms. For instance, pairing foreign-national CEOs (who exhibit higher REM propensity in SSA’s opaque environments) with gender-diverse audit committees reduces manipulation risks. Similarly, aligning the CEO share ownership with tenure policies curbs short-termism: long-tenured CEOs with a substantial equity stake prioritise sustainable growth over earnings manipulation. By mapping how institutional voids in SSA amplify specific agency risks—such as the lax external monitoring of CEO nationality-driven REM—this study advances operational sustainability strategies tailored to resource-constrained settings.

This research redefines the scholarly conversation by proving that governance efficacy is neither universal nor static. It is contingent on the interplay of identity-driven leadership attributes (gender and nationality), ownership incentives, tenure dynamics, and institutional maturity. As SSA’s manufacturing sector navigates global supply chain shifts, these insights offer a roadmap for fostering transparency, attracting ethical investment, and achieving industrialisation that prioritises long-term stability over speculative gains.

This study finds that the CEO gender and audit committee diversity significantly influence real earnings management (REM) in Sub-Saharan African manufacturing firms. Specifically, female CEOs are consistently associated with lower levels of REM, supporting the view that gender-inclusive leadership reduces financial manipulation. CEO ownership shows a positive and significant association with REM in the System GMM estimation, suggesting that higher ownership stakes may, in some settings, incentivise earnings manipulation. However, its effects are inconsistent across other model specifications and may vary by country, reflecting the influence of institutional and governance contexts. Country-level robustness checks reveal that governance effects vary across institutional environments, with firm size and leverage emerging as consistent drivers of REM. These findings underscore the importance of leadership identity and governance structures in shaping earnings’ quality within weak institutional settings.

The rest of this paper is structured as follows. Section 2 presents the theoretical framework, grounding this study in agency theory and Upper Echelons Theory to contextualise the governance and leadership dimensions of real earnings management. This section also reviews the relevant empirical literature and develops the hypotheses that guide the investigation. Section 3 outlines the research methodology, including data sources, variable definitions, and estimation techniques, and further presents the regression results from both aggregate and country-level analyses. Section 4 concludes this study and provides practical recommendations for corporate governance reforms and policy interventions within Sub-Saharan Africa. Lastly, Section 5 discusses this study’s limitations and proposes directions for future research.

2. Theoretical Framework, Empirical Review, and Hypothesis Development

This section outlines the theoretical foundations and empirical evidence that underpin the study. It draws primarily on Agency Theory and Upper Echelons Theory to contextualise the governance and leadership determinants of real earnings management, and it further reviews relevant literature to formulate the study’s hypotheses.

2.1. Theoretical Framework

This study is grounded in two principal theoretical perspectives—agency theory and Upper Echelons Theory—which together offer a comprehensive foundation for analysing the relationship between CEO characteristics, audit committee dynamics, and real earnings management (REM). These frameworks guide the examination of how governance structures and leadership attributes jointly influence earnings management practices, particularly within the institutional context of Sub-Saharan Africa (SSA).

Agency theory posits a principal–agent relationship within corporate governance, wherein managers (agents) act on behalf of shareholders (principals). This relationship is inherently susceptible to conflicts of interest, as managers may pursue personal goals at the expense of shareholder value, thereby engaging in earnings management. Jensen and Meckling (1976) laid the groundwork for agency theory, while Fama and Jensen (1983) further highlighted the importance of governance mechanisms—such as boards and audit committees—as tools to mitigate agency conflicts. More contemporary contributions, including DeFond and Francis (2005) and the Committee on the Financial Aspects of Corporate Governance (1992), underscore the critical oversight role of audit committees. Within this study, audit committee gender diversity is conceptualised as a governance enhancement mechanism, improving ethical oversight and reducing REM. Female directors have been associated with stronger monitoring and higher-quality financial reporting (Lakhal et al., 2015), often attributed to their risk-averse dispositions and ethical orientation (Adams & Ferreira, 2009; Bear et al., 2010).

Upper Echelons Theory, introduced by Hambrick and Mason (1984), focuses on the impact of top executives’ characteristics on organisational outcomes. This theory contends that CEOs’ demographic attributes—such as gender, tenure, ownership, and nationality—inform their strategic decisions, ethical judgements, and management style. For instance, longer-tenured CEOs may be less inclined towards REM due to their accumulated institutional knowledge, while female CEOs are typically associated with more conservative financial approaches. This study builds on the theory by exploring how the interaction between CEO characteristics and audit committee gender diversity influences REM within manufacturing firms in SSA.

Integrating these theoretical perspectives enables a holistic assessment of the drivers of REM. Agency theory emphasises the importance of robust monitoring through governance mechanisms, while Upper Echelons Theory highlights the role of individual leadership traits in shaping financial behaviour. Together, they provide a multidimensional lens through which the moderating influence of audit committee gender diversity on the CEO–REM relationship can be analysed. This integrative framework not only addresses key gaps in the literature but also offers practical insights for enhancing corporate governance and financial transparency in emerging markets such as SSA.

2.2. Empirical Review and Hypotheses Development

This section synthesises global and regional empirical evidence, placing it within the institutional context of Sub-Saharan Africa (SSA). The aim is to identify how CEO profiles and governance structures interact to influence REM, thereby highlighting specific gaps in the literature relevant to SSA’s unique governance and market dynamics.

2.2.1. CEO Characteristics and REM

The relationship between CEO gender and REM is shaped by varying institutional and cultural factors across regions, resulting in mixed empirical findings. In many Asian and European contexts, studies have consistently shown that female CEOs are associated with reduced REM, which is attributed to risk-averse financial strategies and a stronger adherence to ethical governance practices. For instance, Altarawneh et al. (2022) find that female CEOs in Malaysia significantly reduce discretionary accruals. Similarly, research conducted in China (Xiong, 2016) and South Korea (Kyunga & Jooyeon, 2017) links gender diversity at the executive level with conservative financial behaviour. European studies reinforce this trend. Bouaziz et al. (2020), for example, argue that female CEOs tend to prioritise long-term stakeholder trust over short-term financial reporting incentives, thereby reducing REM. African studies by Githaiga et al. (2022) and Attia et al. (2022) report comparable outcomes in East Africa and Egypt, though their generalisability remains limited due to smaller sample sizes and varying regulatory environments.

In contrast, recent evidence from SSA reveals more complex and sometimes contradictory patterns. Studies based in Nigeria (A. Musa et al., 2023; Odubuasi et al., 2023) report a positive association between female CEOs and REM. These findings suggest that contextual pressures—including male-dominated governance systems, fragmented market structures, and intense investor expectations—may compel female executives to adopt short-term earnings strategies to maintain organisational viability. While agency theory would suggest that female CEOs, acting as agents of stakeholder interests, ought to prioritise transparency (Jensen & Meckling, 1976), Upper Echelons Theory (Hambrick & Mason, 1984) helps to explain these discrepancies. The theory argues that executives’ decisions are shaped by their personal values and experiences, which, in turn, are influenced by the broader institutional context. Within SSA, where audit committees often lack gender diversity (A. Musa et al., 2023) and patriarchal norms persist, female CEOs may face systemic barriers that pressure them into REM as a survivalist tactic. Based on these insights, we posit the following hypothesis:

H₁: 

Female CEOs in Sub-Saharan African-listed manufacturing firms are associated with lower levels of REM.

2.2.2. CEO Ownership and REM

The relationship between CEO ownership and REM is shaped by competing theoretical paradigms, particularly agency theory, Upper Echelons Theory, and Stewardship Theory, with institutional contingencies explaining divergent outcomes across regions. Agency theory, particularly the entrenchment hypothesis (Morck et al., 1988), suggests that greater CEO ownership may weaken external monitoring and encourage opportunistic behaviour, including REM. This is supported by studies in Jordan (Qawasmeh & Azzam, 2020), Bangladesh (Debnath et al., 2021), and Malaysia (AL-Duais et al., 2021), which document a significant positive association between CEO ownership and REM. O’Callaghan et al. (2018) similarly found that in the UK, a high level of CEO ownership is used to satisfy short-term performance pressures, fuelling manipulation.

However, findings are mixed in SSA. Nigerian studies by J. Y. Musa and Yahaya (2023) and Farouk and Ahmed (2023) report positive CEO ownership–REM associations, consistent with the view that ownership entrenches power in weak regulatory settings. In contrast, Odubuasi et al. (2023) find a negative relationship, suggesting that ownership may align interests and reduce REM when CEOs behave as accountable agents. This inconsistency can be partly explained by Upper Echelons Theory, which holds that CEO decisions reflect a fusion of their personal characteristics and institutional context. SSA’s institutional voids (Khanna & Palepu, 1997)—characterised by weak investor protection and limited enforcement—create the structural conditions for opportunism.

Stewardship Theory offers an alternative behavioural logic. It posits that managers act not as self-serving agents but as stewards whose goals are aligned with organisational and stakeholder interests (Davis et al., 1997). In SSA, cultural attributes such as communalism, social embeddedness, and long-term orientation (A. Musa et al., 2023) may provide fertile ground for stewardship behaviour, potentially reducing REM. Under this logic, CEOs with ownership stakes may exhibit greater responsibility, avoiding manipulation to preserve long-term firm value. However, the institutional preconditions for stewardship—such as strong board oversight, mutual trust, and transparent governance—are often underdeveloped in SSA. As such, while Stewardship Theory is conceptually relevant, its explanatory power is conditional upon the maturity of corporate governance mechanisms.

Given these competing theoretical perspectives, this study expects CEO ownership in SSA to show a positive association with REM, reflecting the predominance of entrenchment incentives in environments where stewardship mechanisms remain under-institutionalised. This expectation informs the second hypothesis:

H₂: 

Higher CEO ownership in Sub-Saharan African-listed manufacturing firms is positively associated with real earnings management (REM).

2.2.3. CEO Tenure and REM

The relationship between CEO tenure and REM remains empirically contested, reflecting divergent institutional dynamics and theoretical tensions. In Asia, studies present mixed findings. For instance, research conducted in China and Malaysia has shown that a longer CEO tenure correlates with reduced REM, attributing this to strategic prudence and accumulated institutional knowledge (Altarawneh et al., 2022; Xiong, 2016). In contrast, Qawasmeh and Azzam (2020) in Jordan and Nguyen et al. (2023) in Vietnam report positive associations, suggesting that entrenched CEOs leverage operational discretion to manipulate earnings. European studies, such as Bouaziz et al.’s (2020) investigation of French firms, support the negative relationship, suggesting that experienced CEOs tend to prioritise long-term value creation over short-term financial performance.

Within SSA, the evidence is similarly inconclusive. Nigerian studies highlight these contradictions: J. Y. Musa and Yahaya (2023) and Ashafoke et al. (2021) find a positive association, indicating that prolonged CEO tenure may increase REM due to the entrenched authority in poorly regulated environments. Conversely, Odubuasi et al. (2023) document a negative relationship, suggesting that a longer tenure can foster accountability and diminish the inclination to engage in earnings manipulation. According to the agency theory, extended tenure should enhance the alignment between CEOs and shareholders, thereby reducing REM. However, Entrenchment Theory argues the opposite: that longer-serving CEOs may become insulated from external oversight, facilitating opportunistic behaviour.

Upper Echelons Theory provides a contextual interpretation of these conflicting findings. It posits that the strategic choices of CEOs are shaped by their cognitive frameworks and filtered through the prevailing institutional environment. In SSA, institutional voids—such as underdeveloped governance systems, weak audit committee structures, and limited shareholder activism (Khanna & Palepu, 1997)—diminish accountability mechanisms. Consequently, long-tenured CEOs may exploit their operational latitude to achieve short-term performance goals, often at the expense of sustainable value creation (A. Musa et al., 2023). This highlights the limitations of agency theory’s alignment thesis in weak institutional settings. Based on these insights, we posit the following hypothesis:

H₃: 

A longer CEO tenure in Sub-Saharan African-listed manufacturing firms is positively associated with REM.

2.2.4. CEO Nationality and REM

The influence of CEO nationality on REM reflects a complex interplay of cultural norms, institutional familiarity, and governance efficacy, yielding divergent outcomes across global contexts. Studies such as Goel (2018) in India and Nguyen et al. (2023) in Vietnam report positive associations, attributing this to foreign CEOs leveraging operational discretion to navigate unfamiliar regulatory landscapes or respond to short-term investor pressures. Conversely, Rehmana et al. (2021) in Pakistan and Suherman et al. (2023) in Indonesia identify negative associations, suggesting that foreign CEOs may import stringent governance practices that deter REM—especially in jurisdictions with stronger institutional convergence. European evidence, particularly from Bouaziz et al. (2020), supports this view, arguing that local CEOs are more prone to engage in earnings manipulation due to cultural entrenchment and the prioritisation of stakeholder accommodation.

In SSA, the evidence remains inconclusive. Nigerian studies underscore this paradox. J. Y. Musa and Yahaya (2023) and Lawrence (2023) report positive associations, indicating that foreign CEOs may exploit governance weaknesses or conform to short-term investor demands in fragmented markets. Conversely, A. Musa et al. (2023) and Odubuasi et al. (2023) document negative or insignificant relationships, suggesting that foreign CEOs may be more constrained by ethical frameworks or face structural barriers to opportunistic behaviour due to cultural and operational unfamiliarity.

Agency theory suggests that foreign CEOs, as external agents, may align with global standards of ethical conduct and transparency, thereby reducing the likelihood of REM. However, Upper Echelons Theory introduces the institutional context into this equation, positing that CEO decisions are shaped not only by personal values but also by the environments in which they operate. Within SSA, where regulatory systems are often weak and cultural norms vary significantly, foreign CEOs may struggle to internalise local expectations or governance nuances. This lack of contextual embeddedness could inadvertently incentivise REM as a tool to meet performance expectations or close informational gaps. Hence, the contextual limitations of agency theory are evident, as institutional voids heighten informational asymmetries and increase the risk of misaligned incentives. Based on these theoretical and empirical insights, we propose the following hypothesis:

H₄: 

Foreign CEOs in Sub-Saharan African-listed manufacturing firms are positively associated with REM.

2.2.5. The Conceptual Basis for the Moderating Role of Audit Committee Gender Diversity

The relationship between CEO tenure and REM remains ambiguous, shaped by institutional vulnerabilities, sectoral differences, and variations in governance frameworks. Empirical studies across global contexts reveal inconsistent outcomes. In Sri Lanka, gender-diverse audit committees have been found to enhance oversight and reduce REM by reinforcing ethical standards and encouraging risk-sensitive decision-making (Balagobei & Keerthana, 2022). In contrast, findings from Jordan indicate that politically connected CEOs can circumvent governance checks, thereby increasing REM risks (Alhmood et al., 2020).

Such regional disparities persist in Asia and Europe. Research in China and Malaysia links an extended CEO tenure with reduced REM, attributing this to accumulated industry expertise and strategic foresight (Altarawneh et al., 2022; Xiong, 2016). Conversely, evidence from Jordan and Vietnam suggests that entrenched CEOs exploit their operational autonomy, contributing to increased REM (Nguyen et al., 2023; Qawasmeh & Azzam, 2020). French studies align with the former, noting that long-serving CEOs tend to favour sustainable growth strategies over short-term financial manipulation (Bouaziz et al., 2020).

In SSA, the evidence remains mixed. Nigerian studies illustrate tenure’s dual effect: Ashafoke et al. (2021) and J. Y. Musa and Yahaya (2023) associate a prolonged CEO tenure with increased REM in weak regulatory environments, while Odubuasi et al. (2023) report a negative association, suggesting tenure may improve accountability. Despite these insights, empirical research examining how audit committee gender diversity interacts with CEO characteristics to mitigate REM remains sparse within SSA.

Given the governance challenges prevalent in SSA—including male-dominated boards, institutional voids, and limited shareholder activism—gender-diverse audit committees are conceptually positioned to serve as an effective moderating mechanism. Drawing on agency theory and gender socialisation theory, it is reasonable to propose that audit committees with gender diversity contribute to improved scrutiny, reduced managerial opportunism, and enhanced ethical governance. These factors may diminish the capacity of entrenched or dominant CEOs to manipulate earnings. Based on this conceptual rationale, we posit the following hypothesis:

H₅: 

Audit committee gender diversity moderates the relationship between CEO characteristics and real earnings management in Sub-Saharan African-listed manufacturing firms.

3. Methodology

This section outlines the methodological framework adopted to examine the relationship between CEO characteristics, audit committee gender diversity, and REM in listed manufacturing firms across SSA. It presents the research design, data sources, variable measurement, model specification, and estimation techniques applied to ensure a rigorous and contextually grounded analysis.

3.1. Research Design

This study adopts an ex post facto research design to explore correlational relationships between CEO attributes, governance structures, and REM using historical, non-experimental data. Consistent with prior studies in corporate governance, the design relies on observable firm-level variables—such as the CEO gender and audit committee composition—to infer their influence on REM outcomes. This approach enhances ecological validity and supports the empirical testing of theory-driven hypotheses.

Through purposive sampling, 142 firms were selected based on the availability and completeness of their financial and governance disclosures. A fully balanced panel across 11 years would yield a theoretical maximum of 1562 firm-year observations (142 firms × 11 years). However, due to missing data in some firm-years and the inclusion of firms that were newly listed during the period, a total of 373 observations were excluded using a listwise deletion. That is, firm-year observations with missing values for any of the core study variables—especially those concerning CEO characteristics, audit structures, or REM metrics—were removed from the sample. This conservative approach avoids the potential bias associated with imputation or interpolation, which can distort findings when applied to qualitative governance variables. The final dataset consists of 1189 firm-year observations, resulting in an unbalanced panel structure that realistically reflects the data constraints commonly encountered in emerging markets. This enhances both the reliability and generalisability of the results across diverse institutional contexts in Sub-Saharan Africa.

The final sample is geographically representative, with major contributions from Nigeria (28.17%), South Africa (25.15%), Mauritius (12.20%), Zimbabwe (11.10%), and Kenya (8.58%), capturing the region’s varying regulatory and governance environments.

In addition to static panel estimators, this study incorporates System Generalised Method of Moments (System GMM) as a robustness check to address the potential endogeneity between CEO characteristics and REM, ensuring that dynamic relationships and feedback effects are adequately captured.

Data were collected from company annual reports, corporate governance disclosures, and audited financial statements available through firm websites, national stock exchanges, and financial databases such as Bloomberg. CEO-related variables (e.g., gender, tenure, ownership, and nationality) were manually extracted from board profile sections, while REM was operationalised using Roychowdhury’s (2006) model, ensuring methodological consistency with the established earnings management literature.

3.2. Real Earnings Management (REM) Adoption

This study employs the Roychowdhury (2006) model to measure REM, a widely validated framework for detecting operational manipulation. The model quantifies deviations from normal business activities through three proxies:

• Abnormal Cash Flows from Operations (CFO):

  $${\displaystyle \frac{{CFO}_t}{A_{t-1}}}={\alpha }_0+{\alpha }_1\left[{\displaystyle \frac{1}{A_{t-1}}}\right]+\beta 1\left[{\displaystyle \frac{S_t}{A_{t-1}}}\right]+\beta 2\left[{\displaystyle \frac{∆S_t}{A_{t-1}}}\right]+{\epsilon }_t$$
• Abnormal Production Costs (PROD):

  $${\displaystyle \frac{{PROD}_t}{A_{t-1}}}={\alpha }_0+{\alpha }_1\left[{\displaystyle \frac{1}{A_{t-1}}}\right]+{\beta }_1\left[{\displaystyle \frac{S_t}{A_{t-1}}}\right]+{\beta }_2\left[{\displaystyle \frac{∆S_t}{A_{t-1}}}\right]+{\beta }_3\left[{\displaystyle \frac{∆S_t}{A_{t-1}}}\right]+{\epsilon }_t$$
• Abnormal Discretionary Expenses (DISX):

  $${\displaystyle \frac{{DISEXP}_t}{A_{t-1}}}={\alpha }_0+{\alpha }_1\left[{\displaystyle \frac{1}{A_{t-1}}}\right]+\beta \left[{\displaystyle \frac{S_t}{A_{t-1}}}\right]+{\epsilon }_t$$

  where
  A_t: Total assets at time t;
  S_t: Total sales at time t;
  ΔS_t: Change in sales between t − 1 and t;
  ε_t: Error term.

A composite REM index is constructed by summing the absolute values of residuals from these regressions (Cohen & Zarowin, 2010; Owusu et al., 2020; Zang, 2012). This approach provides a holistic measure of earnings manipulation, capturing deviations across operational activities.

3.3. Variable Measurement

Table 1. Variable measurement.

Description	Proxy(ies)	Measurement	Reference
Dependent Variable (DV)	Real Earnings Management (REM)	Abnormal cash flows, production costs, and discretionary expenses	Roychowdhury (2006); Owusu et al. (2020)
Independent Variables (IV)	CEO Gender (CEOGen)	An indicator variable set equal to 1 if female CEO and 0 if male CEO	Odubuasi et al. (2023)
	CEO Ownership (CEOOwn)	% of stock owned by the CEO at the end of the year	Qawasmeh and Azzam (2020)
	CEO Tenure (CEOTen)	Number of years served as CEO	Alhmood et al. (2020)
	CEO Nationality (CEONat)	An indicator variable = 1 if the CEO is foreign, 0 otherwise	Amelia and Eriandani (2021)
Moderating Variable	Audit Committee Gender Diversity (A_DIV)	Ratio of female audit committee members to total audit committee size	Zalata et al. (2018)
Control Variables	Firm Size	Natural log of total assets	Bouaziz et al. (2020)
	Leverage	Debt to total assets × 100	Qawasmeh and Azzam (2020)
	Return on Assets (ROA)	Net income divided by total assets	Hünermund et al. (2024); Resfitasari et al. (2023)
	Accrual Earnings Management (AEM)	Discretionary accruals estimated using the Modified Jones Model	P. M. Dechow et al. (1995); Soeprajitno et al. (2024); Odubuasi et al. (2023)

below presents a summary of the variables, their proxies, measurement approaches, and supporting references used in this study.

3.4. Model Specification

The empirical estimation of the impact of CEO characteristics on real earnings management (REM), as moderated by audit committee gender diversity, is based on three complementary model specifications:

(i) Pooled OLS Model with Country Dummies:

$$RE{M}_{it}={\beta }_0+{\displaystyle \sum _{j=1}^4}{\beta }_{j}CEOCha{r}_{it}+{\beta }_5{\mathrm{A}\_\mathrm{D}\mathrm{I}\mathrm{V}}_{\mathrm{i}\mathrm{t}}+{\displaystyle \sum _{j=6}^9}{\beta }_j{\left(A_{DIV}\times CEOChar\right)}_{it}+{\displaystyle \sum _{k=10}^{13}}{\beta }_{k}Control{s}_{it}+{\displaystyle \sum _{c\ne Nigeria}}{\gamma }_{c}CountryDumm{y}_c+{\mathsf{ϵ}}_{\mathrm{i}\mathrm{t}}$$

where

REM_it = real earnings management for firm i at time t;

CEOChar_it denotes the set of CEO characteristics: gender, ownership, tenure, and nationality;

A_DIV_it = audit committee gender diversity;

Interactions = multiplicative terms between A_DIV and CEO variables;

Controls = firm Size (Fsize), leverage (LEV), ROA (Profitability), and accrual earnings management (AEM);

CountryDummies = set of binary indicators for country c, with Nigeria as the reference category.

This model assumes homogeneity across firms and time, ignoring individual firm effects, but controls for country-specific heterogeneity through dummies.

(ii) Random Effects (RE) Model with Country Dummies:

$$RE{M}_{it}={\beta }_0+{\displaystyle \sum _{j=1}^4}{\beta }_{j}CEOCha{r}_{it}+{\beta }_5{\mathrm{A}\_\mathrm{D}\mathrm{I}\mathrm{V}}_{\mathrm{i}\mathrm{t}}+{\displaystyle \sum _{j=6}^9}{\beta }_j{\left(A_{DIV}\times CEOChar\right)}_{it}+{\displaystyle \sum _{k=10}^{13}}{\beta }_{k}Control{s}_{it}+{\displaystyle \sum _{c\ne Nigeria}}{\gamma }_{c}CountryDumm{y}_c+u_i+{\mathsf{ϵ}}_{\mathrm{i}\mathrm{t}}$$

where

u_i = unobserved firm-specific effects;

ϵ_it = idiosyncratic error term.

This specification allows for an unobserved heterogeneity that is uncorrelated with the regressors while still estimating the effects of time-invariant variables like CEO nationality and country.

(iii) Fixed Effects (FE) Model without Country Dummies:

$$RE{M}_{it}={\beta }_0+{\displaystyle \sum _{j=1}^4}{\beta }_{j}CEOCha{r}_{it}+{\beta }_5{\mathrm{A}\_\mathrm{D}\mathrm{I}\mathrm{V}}_{\mathrm{i}\mathrm{t}}+{\displaystyle \sum _{j=6}^9}{\beta }_j{\left(A_{DIV}\times CEOChar\right)}_{it}+{\displaystyle \sum _{k=10}^{13}}{\beta }_{k}Control{s}_{it}+{\mathsf{\alpha }}_{\mathrm{i}}+{\mathsf{ϵ}}_{\mathrm{i}\mathrm{t}}$$

where:

α_i = firm-specific Fixed Effects (absorbing all time-invariant effects, including country).

Country dummies are omitted due to their perfect collinearity with Fixed Effects.

3.5. Estimation Technique

To estimate the relationship between CEO characteristics, audit committee gender diversity, and REM, five econometric approaches were employed, addressing panel data complexities:

• Pooled OLS with Country Dummies: Estimates cross-sectional effects, controlling for country heterogeneity via dummies (Nigeria as reference).
• Random Effects (RE) with Country Dummies: Accounts for time-invariant firm-level heterogeneity, assuming independence between firm effects and regressors.
• Fixed Effects (FE) without Country Dummies: Controls for unobserved firm-specific traits using within-firm variation, omitting time-invariant variables (e.g., country dummies).
• Feasible Generalised Least Squares (FGLS): Robustness check addressing heteroskedasticity and serial correlation, validated via Breusch–Pagan/Cook–Weisberg tests (χ² = 2.60, p = 0.1068).
• System GMM (Arellano & Bover, 1995; Blundell & Bond, 1998; Berhe, 2024): Dynamic panel estimator used as a robustness check to address endogeneity, simultaneity bias, and reverse causality. CEO ownership and tenure were specified as potentially endogenous. Lagged instruments were used for differenced and level equations.

System GMM employed L(2/3). (CEOOwn CEOTen AEM) as GMM-type instruments for the first-difference equation and DL. (CEOOwn CEOTen AEM) for the levels equation. All other variables, including CEO gender, audit diversity, interaction terms, firm size, leverage, and profitability, were treated as exogenous and entered as standard instruments.

Post-estimation diagnostics supported the validity of the model. The Arellano–Bond AR(1) test was significant (z = −5.90, p < 0.001), while the AR(2) test was not (z = −1.40, p = 0.161), indicating no second-order autocorrelation. The Hansen J-test of overidentifying restrictions (χ² = 87.19, p = 0.035) was within acceptable bounds. The Difference-in-Hansen tests further validated the instrument subsets: the GMM instruments for levels had a difference test result of χ²(18) = 13.07, p = 0.787, and the standard IVs passed with χ²(8) = 4.04, p = 0.853. These confirm the robustness of the instrument strategy and mitigate concerns over instrument proliferation.

The Hausman test (χ² = 23.16, p = 0.0102) favoured FE, confirming firm-specific effects correlate with regressors. FGLS complemented FE/RE results, ensuring robustness (e.g., female CEOs: β = −0.376, p < 0.10).

Diagnostics: I. Multicollinearity: Mean VIF = 2.20 (all <10), ruling out severe correlation. II. Heteroskedasticity: Residuals exhibited homoscedasticity, validated for FGLS application.

The final analysis prioritised FE estimates, supported by RE, OLS, FGLS, and System GMM, ensuring methodological rigour across 1189 firm-year observations (142 firms, 2012–2022).

3.6. Justification of Method

The FE model is prioritised to address unobserved firm-level heterogeneity and time-invariant variables (e.g., CEO nationality and institutional norms), ensuring unbiased estimates where firm-specific traits correlate with regressors. Interaction terms (e.g., audit diversity × CEO gender) enable a nuanced analysis of moderation effects, which is critical for testing governance dynamics.

To strengthen robustness, FGLS supplements FE by correcting for potential heteroskedasticity and serial correlation, validated via Breusch–Pagan/Cook–Weisberg tests (χ² = 2.60, p = 0.1068). FGLS results align with FE/RE estimates (e.g., female CEOs: β = −0.376, p < 0.10), reinforcing methodological credibility.

System GMM was employed to further enhance robustness by addressing endogeneity concerns, particularly those arising from simultaneity and omitted variable bias. It provides consistent estimates for endogenous CEO variables while validating the direction and magnitude of effects observed in static models. This approach aligns with Attia et al. (2022), who applied System GMM to examine board characteristics and earnings quality in African firms, and Berhe (2024), who adopted the same technique to investigate the determinants of bank profitability in Ethiopian banks. Additionally, A. Musa et al. (2023) employed Feasible Generalised Least Squares (FGLS) to explore similar governance–performance dynamics. Together, these studies demonstrate the value of combining static and dynamic estimators to ensure robust inference within the diverse institutional environments of Sub-Saharan Africa.

3.7. Descriptive Statistics

Table 2. Descriptive statistics.

Variable	Obs	Mean	Std. Dev.	Min	Max
REM	1189	1.246	1.214	−1.885	2.375
CEOGen	1189	0.065	0.246	0	1
CEOOwn	1189	0.029	0.099	0	0.636
CEOTen	1189	6.143	5.012	1	28
CEONat	1189	0.17	0.376	0	1
A DIV	1189	0.175	0.21	0	1
Fsize	1189	11.648	1.91	5.171	16.403
LEV	1189	47.261	15.263	11.455	74.944
ROA	1189	6.256	10.048	−98.731	57.683
AEM	1189	−0.2957649	0.4245309	−6.888581	2.10854

Source: Process data by authors via Stata 19.5.

below provides summary statistics for the key variables used in this study, including central tendencies and measures of dispersion:

The mean value of real earnings management (REM) is 1.246 with a standard deviation of 1.214, indicating moderate dispersion around the mean. The minimum and maximum values range from −1.885 to 2.376, respectively, suggesting that some firms engage in significantly high or low levels of real earnings manipulation.

For CEO characteristics, the proportion of female CEOs (CEOGen) is approximately 6.5%, highlighting a male-dominated executive landscape across SSA. CEO ownership (CEOOwn) shows a mean of 2.86%, with a maximum value of 63.6%, indicating considerable heterogeneity in managerial shareholding. CEO tenure (CEOTen) averages about 6.14 years, reflecting a stable leadership structure. CEO nationality (CEONat) reveals that 17% of CEOs are foreigners.

The mean value for audit committee gender diversity (A_DIV) is 0.175, suggesting low female representation on audit committees. Firm size (Fsize), measured as the natural log of total assets, has a mean of 11.65, while leverage (LEV) averages 47.26%. Profitability (ROA) shows an average return of 6.26%, though with a wide dispersion (SD = 10.05), ranging from −98.73% to 57.68%, indicating that some firms are experiencing extreme losses or high profits.

Accrual earnings management (AEM), estimated using the Modified Jones Model, has a mean value of −0.296 and a standard deviation of 0.425. The negative average suggests that firms generally adopt income-decreasing accrual adjustments, which may reflect conservative accounting behaviour or efforts to manage regulatory perceptions. The wide range of values, from −6.89 to 2.11, indicates substantial variability in the use of accrual-based earnings management across firms in the sample. This underscores the importance of controlling for AEM when analysing real earnings management in Sub-Saharan Africa.

Country-Level Descriptive Statistics

Table 3. Selected country-level descriptive statistics.

Kenya	Obs	Mean	Std. Dev.	Min	Max
REM	102	1.568	1.148	−0.198	2.375
CEOGen	102	0.108	0.312	0	1
CEOOwn	102	0.002	0.006	0	0.029
CEOTen	102	7.789	7.895	1.5	28
CEONat	102	0.098	0.299	0	1
A_DIV	102	0.197	0.189	0	0.75
Fsize	102	10.821	1.486	7.801	13.003
LEV	102	42.405	18.049	11.455	73.306
ROA	102	6.331	9.813	−19.463	34.579
AEM	102	−0.188	0.336	−1.102	1.129
Nigeria	Obs	Mean	Std. Dev.	Min	Max
REM	335	0.692	1.146	−1.885	2.375
CEOGen	335	0.087	0.282	0	1
CEOOwn	335	0.033	0.118	0	0.636
CEOTen	335	6.05	5.27	1.5	28
CEONat	335	0.412	0.493	0	1
A_DIV	335	0.157	0.167	0	1
Fsize	335	11.085	1.965	5.225	16.403
LEV	335	53.252	13.108	19.362	74.844
ROA	335	6.137	9.452	−34.057	53.959
AEM	335	−0.327	0.361	−1.536	1.131
South Africa	Obs	Mean	Std. Dev.	Min	Max
REM	299	1.344	1.197	−0.289	2.375
CEOGen	299	0.01	0.1	0	1
CEOOwn	299	0.052	0.115	0	0.636
CEOTen	299	6.916	3.996	1	16.58
CEONat	299	0.033	0.18	0	1
A_DIV	299	0.322	0.258	0	1
Fsize	299	13.064	1.9	7.838	16.403
LEV	299	47.37	14.383	11.455	74.17
ROA	299	4.431	8.442	−37.326	30.122
AEM	299	−0.18	0.199	−0.803	0.544
Zimbabwe	Obs	Mean	Std. Dev.	Min	Max
REM	132	1.191	1.247	−0.319	2.375
CEOGen	132	0.121	0.328	0	1
CEOOwn	132	0.009	0.025	0	0.106
CEOTen	132	5.232	5.947	1.5	24
CEONat	132	0.068	0.253	0	1
A_DIV	132	0.083	0.15	0	0.667
Fsize	132	11.227	1.345	8.605	14.067
LEV	132	42.602	12.81	11.455	72.644
ROA	132	9.554	12.444	−21.941	57.683
AEM	132	−0.448	0.883	−6.889	2.109

Source: Process data by authors via Stata 19.5.

presents subsample descriptive statistics for four key countries—Kenya, Nigeria, South Africa, and Zimbabwe—drawn from the overall panel. These country-level statistics reveal significant intra-regional heterogeneity in firm-level characteristics, governance attributes, and financial indicators. By isolating these subsamples, the analysis underscores the contextual differences across national environments in sub-Saharan Africa, which justify the application of country-specific regression models and interaction terms in later sections of this study.

Real earnings management (REM) varies across countries, with Kenya (1.568) and South Africa (1.344) showing the highest averages, suggesting aggressive REM strategies due to market pressures or lax oversight, while Nigeria reports the lowest (0.692). Female CEO representation remains minimal (1% in South Africa to 12.1% in Zimbabwe), reflecting male-dominated leadership. CEO ownership is highest in South Africa (5.2%) and Nigeria (3.3%), aligning management–shareholder interests, whereas Kenya’s longer CEO tenure (7.79 years) signals stability compared to Zimbabwe’s 5.23 years.

Foreign CEO presence is highest in Nigeria (41.2%), contrasting with Kenya (9.8%), South Africa, and Zimbabwe’s local dominance. Audit committee gender diversity (A_DIV) peaks in South Africa (32.2%), highlighting progressive governance, while Zimbabwe lags (8.3%). South Africa’s larger firm size (log assets: 13.06) reflects market maturity, trailed by Zimbabwe (11.23), Nigeria (11.09), and Kenya (10.82). Leverage is highest in Nigeria (53.3%) and South Africa (47.4%), potentially driving earnings manipulation to meet covenants, versus Kenya and Zimbabwe (~42.4%).

Zimbabwe leads profitability (ROA: 9.55%), followed by Kenya (6.33%) and Nigeria (6.14%), with South Africa being the lowest (4.43%). Accrual earnings management (AEM) is negative across all countries, indicating conservative reporting, which is most pronounced in Zimbabwe (−0.448) and mildest in South Africa (−0.180). These contrasts suggest Zimbabwe prioritises short-term gains via conservative accruals, while South Africa employs subtler adjustments. The stark country-level disparities in governance, performance, and earnings management underscore the need for institutional contextualisation, validating subsample regressions and interaction terms to dissect CEO and board impacts in Sub-Saharan Africa.

3.8. Correlation Matrix

Table 4. Correlation matrix.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
(1) REM	1.000
(2) CEOGen	−0.117	1.000
(3) CEOOwn	0.114	−0.039	1.0001
(4) CEOTen	0.019	−0.044	0.141	1.000
(5) CEONat	−0.18	0.008	−0.054	−0.192	1.000
(6) A_DIV	−0.098	0.107	0.088	−0.044	−0.059	1.000
(7) Fsize	−0.138	−0.083	−0.09	0.056	0.037	0.22	1.000
(8) LEV	−0.133	0.085	−0.008	0.115	0.064	0.106	0.189	1.000
(9) ROA	−0.094	0.094	−0.122	−0.059	0.175	−0.037	0.13	−0.214	1.000
(10) AEM	0.149	−0.139	0.088	0.026	−0.217	0.043	−0.167	−0.006	−0.365	1.000

Source: Process data by authors via Stata 19.5.

presents the pairwise correlation coefficients among all variables.

REM is negatively correlated with CEOGen (−0.117), CEONat (−0.180), A_DIV (−0.098), Fsize (−0.139), LEV (−0.133), and ROA (−0.094), indicating that firms led by female or foreign CEOs, firms with larger asset bases, and those with higher profitability are less likely to engage in real earnings management. Conversely, REM is positively correlated with CEOOwn (0.114) and AEM (0.149), suggesting that ownership-aligned CEOs and firms already engaging in accrual-based earnings management may also rely on real earnings manipulation to manage reported performances.

No pair of variables shows a correlation coefficient exceeding 0.8, indicating the absence of serious multicollinearity issues at this preliminary stage. However, the modest correlation between A_DIV and Fsize (0.220) highlights the need for further multicollinearity diagnostics during the regression estimation.

Country-Level Correlation Insights

Table 5. Correlation matrix across countries.

Variable Pair	Kenya	Nigeria	South Africa	Zimbabwe
REM ↔ CEOGen	0.104	−0.122	0.087	−0.057
REM ↔ CEOOwn	0.141	0.128	0.219	−0.122
REM ↔ CEOTen	−0.262	−0.017	0.078	−0.110
REM ↔ CEONat	−0.229	−0.097	0.161	−0.108
REM ↔ A_DIV	−0.001	−0.128	−0.209	0.138
REM ↔ Fsize	−0.432	−0.042	−0.394	−0.219
REM ↔ LEV	−0.267	0.098	−0.177	0.252
REM ↔ ROA	−0.032	−0.057	−0.019	−0.141
REM ↔ AEM	0.066	0.175	0.179	0.085
AEM ↔ CEOGen	−0.220	−0.074	0.004	−0.142
AEM ↔ ROA	−0.470	−0.409	−0.336	−0.273
AEM ↔ Leverage	0.217	−0.107	−0.027	−0.128
AEM ↔ Firm Size	−0.373	−0.472	−0.373	−0.212

Source: Process data by authors via Stata 19.5. Note. The symbol (↔) denotes the correlation coefficient between the paired variables.

presents country-level correlation matrices for Kenya, Nigeria, South Africa, and Zimbabwe, highlighting differences in the relationships between real earnings management (REM), CEO traits, and governance structures. In Kenya, REM correlates positively with CEO gender (0.104) and ownership (0.141) but negatively with CEO tenure (−0.262) and firm size (−0.432), suggesting smaller firms with newer CEOs are more prone to manipulation. Nigeria shows a negative link between foreign CEOs and both REM (−0.097) and accrual earnings management (AEM: −0.343), indicating stronger oversight. South Africa’s REM is negatively associated with audit diversity (−0.209) and firm size (−0.394), reflecting institutional strength. In Zimbabwe, CEO ownership and foreign status correlate negatively with REM, while profitability shows strong positive ties to foreign CEOs (0.510). Weak REM–AEM correlations across countries challenge the substitution view. These patterns highlight the need for context-specific reforms in governance and executive accountability across Sub-Saharan Africa.

3.9. Regression Results

This section presents the empirical findings from three primary regression models: Ordinary Least Squares (OLS, Model 1), Random Effects (RE, Model 2), and Fixed Effects (FE, Model 3), with complete results reported in

Table 6. Regression results: Pooled OLS, Random Effects, Fixed Effects, and FGLS models.

	Pooled	RE	FE	FGLS	SGMM
CEOGen	−0.924 ***	−1.001 ***	−1.063 ***	−0.372	−0.469 *
	(0.212)	(0.211)	(0.227)	(0.204)	(0.225)
CEOOwn	0.518	0.375	−0.272	0.0979	0.976 *
	(0.457)	(0.470)	(0.557)	(0.187)	(0.458)
CEOTen	−0.00748	−0.00963	−0.0396	0.000913	−0.00521
	(0.00844)	(0.0135)	(0.379)	(0.00357)	(0.0479)
CEONat	−0.0255	−0.0970		0.0173
	(0.127)	(0.200)		(0.0343)
A_DIV	−0.517	−0.352	−0.221	−0.0346	−0.436
	(0.269)	(0.315)	(0.368)	(0.0956)	(0.690)
A_DIV_CEOGen	2.641 ***	2.082 **	1.847 **	0.697	1.215
	(0.649)	(0.648)	(0.688)	(0.500)	(0.797)
A_DIV_CEOOwn	3.225 *	2.357	2.179	1.158	−0.505
	(1.303)	(1.285)	(1.519)	(0.668)	(1.560)
A_DIV_CEOTen	−0.0455	−0.0463	−0.0456	−0.00329	0.0428
	(0.0324)	(0.0397)	(0.0473)	(0.0139)	(0.101)
A_DIV_CEONat	−0.719	−1.072 *	−1.206 *	−0.836 *
	(0.507)	(0.536)	(0.578)	(0.408)
Fsize	−0.119 ***	−0.0738 *	0.0858	−0.0160 *	−0.0302
	(0.0199)	(0.0289)	(0.0477)	(0.00736)	(0.0314)
LEV	0.00551 *	0.00692 *	0.00924 **	0.000568	−0.00529
	(0.00232)	(0.00287)	(0.00349)	(0.000491)	(0.00300)
ROA	0.00292	0.00694	0.00767 *	0.000322	−0.00527
	(0.00355)	(0.00355)	(0.00378)	(0.000839)	(0.00384)
AEM	0.0681	0.0159	−0.0161	0.0210	0.0641
	(0.0828)	(0.0871)	(0.0943)	(0.0279)	(0.174)
Eswatini	1.920 ***	1.715 *		1.503 ***
	(0.360)	(0.670)		(0.0921)
Ghana	0.290	0.247		1.109 ***
	(0.212)	(0.375)		(0.278)
Kenya	0.906 ***	0.878 ***		1.339 ***
	(0.127)	(0.228)		(0.132)
Malawi	−0.131	−0.325		−0.0895
	(0.249)	(0.481)		(0.448)
Mauritius	1.654 ***	1.580 ***		1.465 ***
	(0.115)	(0.211)		(0.0893)
Namibia	1.432 ***	1.259		1.036 *
	(0.329)	(0.659)		(0.496)
Rwanda	−0.794	−0.880		−0.850 ***
	(0.535)	(0.754)		(0.163)
South Africa	0.939 ***	0.768 ***		1.121 ***
	(0.109)	(0.185)		(0.113)
Tanzania	0.601 ***	0.466		1.005 ***
	(0.181)	(0.329)		(0.198)
Uganda	−0.436	−0.499		−0.207
	(0.259)	(0.482)		(0.447)
Zambia	0.161	0.0942		0.0572
	(0.188)	(0.332)		(0.327)
Zimbabwe	0.502 ***	0.430*		0.936 ***
	(0.118)	(0.211)		(0.167)
L.REM					0.495 ***
					(0.0487)
_cons	1.938 ***	1.431 ***	0.146	1.061 ***	1.257 ***
	(0.227)	(0.349)	(2.340)	(0.112)	(0.368)
chi2		149.5		745.5	1102.6
N	1189	1189	1189	1183	955
aic	3496.9		2949.7
r2	0.280	0.035	0.0479
F-statistics	18.08	149.48	4.34	745.54
Prob > F/Prob > chi2	0.000	0.000	0.000	0.000

Note. Standard errors are reported in parentheses. *** p < 0.01, ** p < 0.05, and * p < 0.10. Pooled = Pooled Ordinary Least Squares; RE = Random Effects; FE = Fixed Effects; FGLS = Feasible Generalised Least Squares; and sGMM = System Generalised Method of Moments.

. To ensure the robustness of our conclusions, we supplement these baseline estimates with three distinct validation approaches in Section 3.10: (1) Feasible Generalised Least Squares (FGLS) to correct for heteroskedasticity and serial correlation, (2) System Generalised Method of Moments (GMM) to address potential endogeneity, and (3) country-level analyses to assess cross-national consistency. Diagnostic tests confirm the appropriateness of our model specifications.

3.9.1. CEO Gender Effects on REM

Consistent with this study’s first hypothesis (H1), the results demonstrate a robust and statistically significant negative relationship between CEO gender (female) and real earnings management (REM) across all model specifications. The coefficient for CEO gender is −0.924 (p < 0.01) in the Pooled OLS model, −1.001 (p < 0.01) in the Random Effects model, and −1.063 (p < 0.01) in the Fixed Effects model. This consistency reinforces the validity of H1, suggesting that female CEOs are systematically less likely to engage in REM practices.

This finding aligns with Upper Echelon Theory (Hambrick & Mason, 1984), which posits that top executives’ demographic traits shape strategic choices. Female CEOs, often associated with higher ethical sensitivity and risk aversion, may impose stricter internal controls and discourage aggressive earnings manipulation. Additionally, from an agency theory perspective, female executives might signal transparency and compliance to mitigate agency conflicts and enhance investor trust.

In the SSA context, where governance structures often suffer from institutional inefficiencies, the influence of CEO gender as a behavioural constraint becomes especially salient. The policy implication is clear: increasing gender diversity at the top executive level could serve as an endogenous mechanism for improving earnings quality and curbing opportunistic financial behaviour.

3.9.2. CEO Ownership Effects on REM

Contrary to Hypothesis 2 (H₂), the analysis reveals no consistent or statistically significant relationship between CEO ownership and REM. Across the four models, the coefficients vary in size and remain insignificant (e.g., 0.518 in Pooled OLS; −0.272 in Fixed Effects), undermining the expected entrenchment effect posited by agency theory.

One explanation lies in the institutional voids characteristic of many SSA countries—such as weak legal enforcement, limited shareholder activism, and underdeveloped capital markets. These voids may neutralise both the alignment and entrenchment mechanisms that CEO equity stakes are theorised to activate. Consequently, ownership by CEOs in SSA firms may serve more as a nominal feature than as a meaningful governance lever, lacking the disciplinary potency observed in developed markets.

This result suggests that ownership alone is insufficient to predict REM behaviour in environments where external monitoring and governance enforcement mechanisms are fragmented.

3.9.3. CEO Tenure Effects on REM

The empirical evidence does not support Hypothesis 3 (H₃), which expected a positive association between CEO tenure and REM based on the entrenchment logic within Upper Echelon Theory and agency theory. The coefficients for CEO tenure are small and statistically insignificant in all models (e.g., −0.040 in FE, p = 0.915), indicating no systematic relationship.

This could suggest that in SSA manufacturing firms, tenure-based entrenchment does not necessarily translate into higher earnings manipulation. Instead, longer-tenured CEOs may be constrained by reputational concerns, board oversight, or industry-level norms that discourage aggressive financial practices. Additionally, in volatile regulatory environments, even experienced CEOs may lack the institutional backing to engage in discretionary reporting without consequence.

These findings highlight that tenure’s impact on REM may be contingent on the broader governance ecosystem and thus should be interpreted with caution in low-governance contexts.

3.9.4. CEO Nationality Effects on REM

Hypothesis 4 (H₄), which anticipated that foreign CEOs would be positively associated with REM due to higher informational asymmetry and reduced social embeddedness, is not supported by the results. The coefficients for CEO nationality are statistically insignificant across models (e.g., −0.097 in RE, p = 0.589).

This outcome suggests that institutional isomorphism—the process through which foreign executives adapt to host-country norms—may counterbalance the presumed agency costs of foreignness. In SSA’s regulatory and cultural context, foreign CEOs might face strong pressures to conform to domestic expectations, especially in publicly listed firms where national legitimacy and stakeholder trust are vital.

The results challenge the assumption that foreign executives inherently elevate REM risk, implying that the effects of CEO nationality on financial behaviour are shaped more by local institutional logics than by nationality alone.

3.9.5. Interaction Effects

Hypothesis 5 (H₅) receives partial support. The direct effect of audit committee gender diversity (A_DIV) on REM is marginally significant only in the Pooled OLS model (−0.517, p < 0.10) and becomes insignificant in the RE and FE models. However, interaction terms involving A_DIV and CEO characteristics provide stronger insights.

The interaction between A_DIV and CEOGen is significantly positive (e.g., 2.641 in Pooled OLS; 1.847 in FE, both p < 0.01), indicating that female-led firms with gender-diverse audit committees exhibit lower levels of REM. This supports Upper Echelon Theory, as demographic congruence between board and executive leadership enhances ethical alignment. Similarly, the significant negative interaction between A_DIV and CEONat (−1.072 in RE, −1.206 in FE; both p < 0.05) affirms the monitoring hypothesis within agency theory, wherein audit diversity mitigates the potential informational gap introduced by foreign CEOs.

In contrast, interactions with CEO ownership and tenure are not significant, suggesting that A_DIV’s moderating power depends on specific leadership traits and may require supportive governance frameworks to be effective.

3.9.6. Control Variables and Country Dummies

Several control variables yield theoretically consistent results.

Firm size negatively correlates with REM in the Pooled OLS (−0.119, p < 0.01) and Random Effects (−0.074, p < 0.05) models, confirming agency theory’s assertion that larger firms face higher external scrutiny, reducing discretionary practices. However, the positive coefficient in the FE model (0.086, p < 0.10) suggests that firm-specific unobserved heterogeneity—such as internal governance weaknesses—may override external pressures.

Leverage shows a positive and significant relationship with REM across models, particularly in FE (0.009, p < 0.01), aligning with the debt covenant hypothesis under agency theory. Firms under financial pressure may manipulate real activities to meet creditor expectations, reinforcing the risk of earnings distortion in highly leveraged environments.

Accrual earnings management (AEM) does not significantly relate to REM, challenging Zang’s (2012) substitution hypothesis. The lack of substitution may reflect weak auditing standards and poor enforcement in SSA, where REM remains the dominant discretionary tool.

Country-level dummy variables reveal higher REM in Mauritius, Kenya, Zimbabwe, and South Africa—countries typically considered more advanced within SSA—possibly due to higher expectations or complex stakeholder demands that create incentives for real-based manipulation. Conversely, lower REM in Uganda and Rwanda may reflect tighter external aid-driven governance reforms or less aggressive financial practices.

3.10. Robustness Checks

To ensure the reliability of our findings, we conduct a series of robustness checks using alternative estimation techniques, including Feasible Generalized Least Squares (FGLS), System Generalized Method of Moments (GMM), and country-level analyses to address potential biases and validate the consistency of our results.

3.10.1. Robustness Checks Using FGLS

To ensure the reliability of the findings, this study conducted a robustness check using FGLS estimation. This method accounts for heteroskedasticity and serial correlation, which are common concerns in panel data analysis. Unlike the FE and RE models, which assume homoscedastic residuals and no cross-sectional correlation, the FGLS approach adjusts for panel-specific heteroskedasticity, making it a suitable alternative for robustness validation.

The FGLS results, reported in Table 6, align with the primary regression estimates from the Pooled OLS, RE, and FE models. CEO gender remains negatively associated with REM, although the magnitude of the effect is slightly attenuated compared to the Fixed Effects model. Similarly, the interaction between audit committee gender diversity and CEO gender remains positive, reinforcing the notion that gender-diverse audit committees strengthen the relationship between female leadership and reduced REM. CEO ownership, while positive and significant in the Pooled models, exhibits weaker statistical significance in the FGLS estimation, suggesting that the ownership–REM dynamic may be sensitive to model specification.

The findings on CEO tenure and CEO nationality interactions also remain consistent across models, with audit committee gender diversity moderating the effects of CEO nationality in a negative direction. Control variables such as firm size and leverage show comparable trends, further supporting the robustness of the main results. The FGLS results validate the primary findings, confirming that the observed relationships are not driven by model-specific assumptions but are instead robust across different econometric techniques.

3.10.2. Robustness Checks Using System GMM

To further strengthen the credibility of the findings and address potential endogeneity concerns, a dynamic panel data approach was employed using the two-step System Generalised Method of Moments (System GMM) estimator proposed by Blundell and Bond (1998). This approach is particularly appropriate for panel datasets where the number of cross-sectional units (N) exceeds the number of time periods (T), and where the dependent variable—real earnings management (REM)—exhibits potential persistence over time. The inclusion of the lagged dependent variable (L.REM) allows for the correction of autocorrelation and unobserved heterogeneity, which may otherwise bias static estimators like Pooled OLS, RE, FE, or even FGLS.

The model employed lagged levels of potentially endogenous regressors—CEO ownership, CEO tenure, and accrual-based earnings management—as GMM-type instruments, while treating time-invariant variables, such as CEO gender, audit committee gender diversity, and their interactions, as standard instruments. The Arellano–Bond AR(1) and AR(2) tests confirmed the presence of a first-order serial correlation (p < 0.01) but no significant second-order autocorrelation (p = 0.161), validating the instrument structure. Furthermore, the Hansen J-test for over-identifying restrictions (p = 0.035) falls within acceptable bounds, albeit with a caution regarding instrument proliferation, which was mitigated through limited lags.

Results from the System GMM (reported in Table 6) largely confirm the robustness of the prior estimations. CEO gender remains negatively associated with REM (β = −0.469, p < 0.05), reaffirming the mitigating influence of female leadership on earnings manipulation. CEO ownership shows a significant positive effect (β = 0.976, p < 0.05), suggesting that highly invested CEOs may engage in REM, perhaps to influence market perceptions. However, CEO tenure retains an insignificant relationship, aligning with prior Fixed Effects findings. Most interaction terms, particularly involving audit committee gender diversity, remain statistically insignificant, which could be attributed to the limited variation in these governance structures over time.

The statistically significant lagged dependent variable (β = 0.495, p < 0.01) indicates strong REM persistence, justifying the use of dynamic modelling. Control variables behave consistently with prior models, although leverage and profitability exhibit marginal significance levels. The System GMM estimator provides further validation for the core results and serves as an effective robustness check, accounting for endogeneity, autocorrelation, and dynamic adjustment mechanisms in the corporate reporting behaviour of SSA-listed manufacturing firms

3.10.3. Further Robustness Checks at Country Level

To account for cross-national heterogeneity and assess the reliability of the main findings, country-specific regressions were conducted using the same estimators used for the main models, as shown in

Table 7. Statistically significant variables by country.

Country	Significant Variables	Coefficient (Std Error)
Kenya	Fsize	−0.351 *** (0.132)
	CEOTen	−0.039 ** (0.019)
Nigeria	CEOGen	−0.782 ** (0.398)
	LEV	0.016 *** (0.005)
	AEM	0.488 ** (0.211)
South Africa	A_DIV × CEOOwn	5.459 ** (2.150)
	Fsize	−0.194 *** (0.041)
Zimbabwe	A_DIV	3.379 *** (1.236)
	LEV	0.034 *** (0.009)
	CEONat (FGLS)	−4.251 ** (2.126)

Note. This table presents country-specific regression estimates where coefficients are statistically significant across subsamples. Standard errors are reported in parentheses. *** p < 0.01, and ** p < 0.05.

. These robustness checks serve to confirm that the observed relationships are not artefacts of sample pooling or methodological bias, thereby strengthening the internal validity of this study. The results provide valuable insights into the institutional and governance dynamics specific to each sub-Saharan African (SSA) country in the sample.

In the case of Kenya, two variables consistently showed statistical significance. First, firm size (Fsize) exhibited a strong and negative association with real earnings management (REM) across all models, with a coefficient of −0.351 (p < 0.01). This aligns with agency theory, which posits that larger firms are subject to heightened scrutiny by external stakeholders, thereby limiting the scope for opportunistic financial reporting. Second, CEO tenure (CEOTen) was negatively and significantly associated with REM in the OLS and FGLS models (β = −0.039, p < 0.05). This suggests that more experienced CEOs may be less inclined to engage in earnings manipulation, which is consistent with the conservatism lens of Upper Echelon Theory. However, the lack of significance in the FE model hints at potential unobserved heterogeneity at the firm level.

In Nigeria, three factors emerged as significant determinants of REM. Female CEO leadership (CEOGen) was associated with a reduction in REM, with a robust coefficient of −0.782 (p < 0.05), supporting Hypothesis 1. This finding reinforces the ethical leadership perspective under Upper Echelon Theory, suggesting that gender diversity at the top management level may act as a deterrent to financial opportunism. Furthermore, leverage (LEV) showed a positive and significant association with REM (β = 0.016, p < 0.01), validating the debt covenant hypothesis of agency theory, whereby financially constrained firms engage in earnings manipulation to meet creditor expectations. Lastly, accrual-based earnings management (AEM) also displayed a positive relationship with REM (β = 0.488, p < 0.05), thereby contradicting the substitution hypothesis (Zang, 2012). This indicates that firms in Nigeria may employ both accrual and real manipulation concurrently, likely due to weak auditing and enforcement mechanisms.

For South Africa, governance-related interaction effects were most pronounced. Specifically, the interaction term between audit committee gender diversity and CEO ownership (A_DIV × CEOOwn) was positive and significant (β = 5.459, p < 0.05). This supports Hypothesis 5 and aligns with agency theory, suggesting that diverse audit committees can strengthen oversight in firms where CEOs hold substantial ownership stakes. Firm size also maintained a consistently negative association with REM (β = −0.194, p < 0.01), echoing findings from Kenya and reinforcing the cross-country relevance of the external scrutiny effect.

In Zimbabwe, a distinctive set of dynamics was observed. Audit committee gender diversity (A_DIV) had a strong positive association with REM (β = 3.379, p < 0.01), a finding that deviates from other countries. This may reflect institutional voids in Zimbabwe’s corporate governance environment, where diversity lacks structural empowerment to be effective. Moreover, CEO nationality (CEONat), measured using the FGLS estimator, showed a significant negative relationship with REM (β = −4.251, p < 0.05). This finding supports Upper Echelon Theory by suggesting that foreign CEOs may adopt more conservative financial reporting practices in volatile environments to preserve credibility.

Methodologically, the FGLS estimates largely corroborated those obtained through Pooled OLS and RE models, providing further assurance against concerns related to heteroskedasticity or autocorrelation. However, Fixed Effects models were generally less robust, often yielding nonsignificant coefficients—likely a result of limited within-firm variation and the omission of time-invariant predictors such as CEO nationality. This underscores the analytical limitations of FE models when applied to smaller country samples or when explanatory variables are relatively stable over time.

Theoretically, the results reinforce key propositions from both agency and upper echelon theories. Variables associated with scrutiny (e.g., firm size) and financial pressure (e.g., leverage) consistently emerged as significant across countries, validating core agency-based assumptions. In contrast, variables such as CEO gender and tenure reflected the influence of managerial characteristics on financial behaviour, a central tenet of Upper Echelon Theory. These findings support the need for targeted policy interventions. For example, Zimbabwe may benefit from strengthening audit committee structures, while Nigeria could leverage gender-based leadership reforms to curb earnings manipulation.

The country-level robustness checks affirm this study’s main findings while illuminating important national variations. The results underscore the need for context-specific governance reforms to reduce real earnings management and enhance financial transparency across Sub-Saharan Africa.

4. Conclusions and Recommendations

This study examined the impact of CEO characteristics—namely gender, ownership, tenure, and nationality—on real earnings management (REM) among listed manufacturing firms in selected Sub-Saharan African countries. It also assessed the moderating role of audit committee gender diversity. Using a panel dataset from Kenya, Nigeria, South Africa, and Zimbabwe over the period from 2012 to 2023, this analysis applied a combination of Pooled OLS, Random Effects, Fixed Effects, Feasible Generalised Least Squares (FGLS), and System Generalised Method of Moments (System GMM) models to ensure robustness, control for endogeneity, and address country-level heterogeneity.

The results demonstrate that CEO gender is a significant predictor of REM. Firms led by female CEOs exhibit lower levels of earnings manipulation, supporting the proposition of Upper Echelon Theory that managerial demographics influence strategic decisions. This finding also aligns with agency theory, where ethical leadership serves as a mechanism to reduce conflicts between managers and shareholders. The interaction between audit committee gender diversity and CEO gender further confirms that the presence of women in key oversight roles can reinforce ethical governance practices.

CEO ownership shows a positive and significant association with REM in the System GMM estimation, indicating that higher ownership stakes may, in some contexts, incentivise earnings manipulation. However, the effect is not consistent across other model specifications and may vary by country, reflecting the influence of institutional quality and governance mechanisms. Similarly, CEO tenure does not exhibit a systematic influence on REM, suggesting that experience alone is not a sufficient determinant of financial reporting behaviour in the SSA context.

The analysis also reveals the context-dependent nature of CEO nationality. While no significant effect is observed in most countries, foreign CEOs in Zimbabwe are associated with lower REM. This implies that in highly uncertain institutional environments, foreign executives may prefer conservative financial practices to manage reputational and operational risks.

Among the control variables, firm size shows a consistently negative relationship with REM in most estimators, although the effect becomes nonsignificant under Fixed Effects and System GMM. This suggests that larger firms, typically under more intense regulatory and investor scrutiny, may face reduced incentives or ability to engage in earnings manipulation. Leverage is positively and significantly associated with REM across several models, though the effect turns negative and nonsignificant in System GMM, pointing to potential endogeneity in earlier estimates. The positive association between REM and accrual earnings management (AEM) in some countries further implies that firms may deploy both techniques concurrently, particularly where enforcement mechanisms are weak.

Country-level robustness checks reinforce these findings. In Kenya, firm size and CEO tenure are significant, indicating that scrutiny and executive experience reduce REM. In Nigeria, female CEO leadership and financial leverage are key drivers. In South Africa, the interaction between audit committee diversity and CEO ownership plays a significant role. In Zimbabwe, audit committee diversity is paradoxically linked with higher REM, suggesting institutional gaps in governance effectiveness.

Based on these findings, several policy recommendations are proposed. Regulatory institutions across Sub-Saharan Africa should strengthen governance frameworks and enhance transparency mechanisms to reduce the incidence of REM. This includes improving audit committee effectiveness, enforcing compliance with corporate governance codes, and promoting gender diversity in leadership and oversight roles. The consistent association between female CEOs and lower REM underscores the need for gender-responsive reforms that encourage women’s participation in corporate leadership. Furthermore, shareholder protection and monitoring mechanisms should be improved, particularly to ensure that CEO ownership aligns managerial incentives with shareholder interests rather than entrenching control. Country-specific strategies are also critical: Zimbabwe may benefit from technical capacity-building initiatives for audit committees, whereas Nigeria should focus on ethical leadership promotion and corporate deleveraging. Tailored interventions such as these will help address institutional weaknesses, reduce manipulation incentives, and improve the financial reporting quality in the region.

5. Limitations of the Study and Directions for Future Studies

While this study provides valuable insights into the relationship between CEO characteristics, audit committee gender diversity, and real earnings management (REM) across Sub-Saharan Africa, it is not without limitations. First, the analysis is restricted to listed manufacturing firms from selected countries. This sectoral and geographical focus limits the generalisability of the findings to other industries or to unlisted firms that may operate under different governance structures or reporting incentives.

Second, although panel data techniques—specifically Fixed Effects, Feasible Generalised Least Squares (FGLS), and System GMM—were employed to control for unobserved heterogeneity and mitigate endogeneity concerns, certain limitations persist. Time-invariant variables, such as CEO nationality, were excluded from the Fixed Effects model due to collinearity, limiting the within-firm exploration of cross-cultural diversity. However, the use of System GMM as an additional robustness check helps to address the potential simultaneity bias and omitted variable bias, thereby enhancing the credibility of the empirical strategy.

Third, the absence of disaggregated demographic data on CEOs—such as age, educational background, and prior experience—constrains the scope of the individual-level analysis. This limits the ability to evaluate how nuanced personal traits influence financial reporting practices. Future studies could benefit from enriched datasets that capture these executive-level attributes.

Fourth, relatively small sample sizes for certain countries (e.g., Zimbabwe and Kenya) may have constrained the statistical power of country-level models. As a result, theoretically relevant variables, such as CEO tenure or ownership, may not have consistently emerged as statistically significant, despite their conceptual relevance. Furthermore, the use of proxies for REM and audit committee diversity in emerging market contexts—where disclosure standards are still evolving—may introduce estimation errors.

In terms of future directions, several avenues remain promising. Qualitative or mixed-method studies can enrich the understanding of behavioural mechanisms behind CEO discretion and audit committee oversight. Expanding the scope to include other regions, industries, or state-owned enterprises would allow for broader generalisation and institutional comparisons. Additional board characteristics, including ethnic diversity, tenure dispersion, and generational differences, could also be integrated to assess more complex interactions with executive traits.

Moreover, incorporating ESG indicators may reveal how broader sustainability commitments influence financial reporting practices. Finally, replicating this study in the context of the post-COVID-19 recovery or during major policy transitions—such as IFRS adoption or national governance code reforms—could offer more current insights for regulators, investors, and scholars interested in corporate accountability in Africa.