Corporate Governance and Dividend Policy Under Concentrated Ownership: Evidence from Post-Reform Korea

Abstract

This study investigates how ownership structure conditions the transmission of corporate governance mechanisms into dividend policy within the context of South Korea’s evolving regulatory environment. Using a balanced panel of 5022 firm-year observations from 558 non-financial KOSPI-listed firms over 2011–2019, we analyze governance quality using data from the Korea Corporate Governance Service. We employ both an aggregate score and four constituent dimensions: board effectiveness, shareholder rights protection, audit committee competency, and disclosure transparency. The empirical framework combines firm fixed effects estimation, binary logistic regressions, and a two-step dynamic System GMM approach to account for unobserved heterogeneity, payout persistence, and endogeneity. The results reveal systematic heterogeneity across ownership regimes. Among non-Chaebol firms, higher governance quality across all dimensions is associated with higher dividend payouts, consistent with the governance outcome hypothesis. In contrast, among Chaebol-affiliated firms, the effectiveness of governance mechanisms is selective rather than uniform. While the aggregate governance score and shareholder rights protection retain explanatory power for dividend outcomes, internal oversight mechanisms related to board structure, audit competency, and disclosure do not exert independent influences once ownership structure is taken into account. These findings show that concentrated ownership structures condition which governance mechanisms remain effective in shaping payout policy. Regulators seeking to mitigate valuation discounts in conglomerate-dominated economies should prioritize the substantive empowerment of minority shareholder rights, as these mechanisms retain influence over payout policy even under concentrated ownership structures.

1. Introduction

How do multidimensional corporate governance mechanisms interact with concentrated ownership structures to shape the credibility of dividend policy? This question is central to understanding the persistence of the “Korea discount,” a valuation gap commonly associated with weak payout commitments and entrenched conglomerate control (Claessens et al., 2000; Faccio et al., 2001; Kang, 2023; La Porta et al., 2000). While much of the governance literature assumes that stronger governance uniformly enhances dividend outcomes (Mitton, 2004; Renneboog & Szilagyi, 2020; Shleifer & Vishny, 1997), such an assumption is difficult to reconcile with ownership environments characterized by cross-shareholdings and controlling shareholders. We argue that the effectiveness of governance mechanisms is conditional on ownership architecture, such that identical governance structures may transmit differently into payout behavior depending on the degree of control concentration. In markets dominated by business groups, formal governance improvements need not translate mechanically into higher or more credible dividend payouts (Council, 2016; Lewis, 2024).

The extant literature is divided between two competing theoretical perspectives on the governance–payout relationship. The governance outcome hypothesis posits that rigorous monitoring frameworks, proxied by governance scores reflecting effective monitoring capacity, compel managers to disgorge free cash flow to mitigate agency costs (Faccio et al., 2001; M. C. Jensen, 1986; La Porta et al., 2000). In contrast, the substitution hypothesis suggests that in concentrated ownership systems, internal monitoring by blockholders may render formal governance mechanisms such as board or audit committee oversight redundant (Villalonga & Amit, 2006). While prior studies document both positive and muted effects of governance on payouts, they offer limited consensus on the conditions under which each outcome prevails (Almeida et al., 2011; Kang, 2023; Mili et al., 2017). This study responds by investigating whether ownership heterogeneity, distinguishing between Chaebol and non-Chaebol firms, constitutes a boundary condition that governs the transmission of governance quality into dividend policy.

To address the fragmented evidence in prior studies that rely on limited governance proxies or narrower institutional windows (Hwang et al., 2013; S. M. Kim & Lee, 2008; W. Kim et al., 2021; Y. S. Kim & Wee, 2019), we assemble a comprehensive panel of firm-year observations spanning 2011–2019, a period characterized by heightened scrutiny of corporate governance in the Korean capital market (Kang, 2023). We draw on governance data supplied by the Korea Corporate Governance Service (KCGS), employing its full set of five indices, an aggregate score and four constituent pillars capturing board effectiveness, shareholder rights protection, audit committee competency, and disclosure transparency (Korea Corporate Governance Service [KCGS], 2021). This multidimensional structure allows us to move beyond aggregate governance measures and examine how distinct governance mechanisms differentially transmit into dividend policy across ownership regimes.

Our econometric framework integrates firm fixed effect estimation, binary choice models, and a dynamic System GMM approach to address the endogeneity arising from the joint determination of governance quality and dividend policy (Jeon et al., 2011; Lee, 2022; Song & Chun, 2025). Firm fixed effects absorb time-invariant unobserved heterogeneity, while binary logistic specifications assess whether governance affects the propensity to maintain dividend-paying status rather than payout levels alone. The dynamic System GMM estimator further accounts for payout persistence and potential reverse causality, allowing us to distinguish governance effects operating through intertemporal payout credibility from those reflecting static firm characteristics. This multi-estimator strategy ensures that the documented governance–dividend relationship is not an artifact of a specific functional form or identification strategy, but reflects a robust structural association (Ararat et al., 2021; Chintrakarn et al., 2022; Claessens et al., 2000; Claessens & Yurtoglu, 2013; Gonzalez et al., 2017; Miller & Modigliani, 1961; Mitton, 2004; Renneboog & Szilagyi, 2020).

Empirically, we uncover pronounced heterogeneity in how governance mechanisms translate into dividend outcomes across ownership regimes. Among non-Chaebol firms, stronger governance across all constituent pillars is associated with higher dividend yields, consistent with the view that formal governance mechanisms function as binding constraints on managerial discretion. In contrast, among Chaebol-affiliated firms, the governance–dividend linkage is sharply mechanism-specific. While the aggregate governance score and shareholder rights protection retain economic and statistical significance, internal monitoring mechanisms, such as board structure, audit oversight, and disclosure transparency, lose incremental explanatory power once ownership concentration is taken into account. This pattern indicates that conglomerate control structures fundamentally condition which governance mechanisms remain effective in shaping payout policy.

This study contributes to the corporate finance literature in three important ways. First, it demonstrates that the relationship between governance quality and dividend policy is conditional on ownership structure rather than uniform across firms. Second, by decomposing governance into its constituent dimensions, the analysis shows that governance substitution within business groups is mechanism-specific: under concentrated ownership, internal monitoring channels are largely redundant, whereas shareholder-facing mechanisms continue to discipline payout behavior. Third, the findings contribute to the broader investor protection literature by clarifying how ownership concentration conditions the informational and disciplinary content of governance signals, with direct implications for economies characterized by dominant controlling shareholders.

In sum, our results show that dividend policy reflects the interaction between governance quality and ownership structure rather than governance quality in isolation. In the Korean context, the persistence of the “Korea discount” appears linked not merely to weaknesses in formal governance metrics, but to structural features of conglomerate control that constrain the effectiveness of internal monitoring mechanisms in shaping payout credibility. These findings suggest that efforts to enhance dividend credibility and shareholder value require attention not only to governance reforms themselves, but also to the ownership environments in which such reforms are embedded.

The remainder of the article is structured as follows. Section 2 develops the theoretical foundations and reviews the literature. Section 3 details the data sources and empirical methodology. Section 4 presents the primary findings while Section 5 provides robustness analyses. Section 6 concludes with implications for policy and institutional development.

2. Theoretical Foundations and Hypothesis Develpment

2.1. Dividend Policy as an Agency and Commitment Mechanisms

Dividend policy has long occupied a central position in corporate finance theory, serving as a pivotal mechanism for mitigating agency conflicts between insiders and outside shareholders. In the presence of excess free cash flow, managers and controlling owners face incentives to pursue value-reducing investments or extract private benefits unless constrained by credible payout commitments (Fama & Jensen, 1983; M. Jensen & Meckling, 1976; M. C. Jensen, 1986). By channeling cash to shareholders rather than retaining it, dividends reduce managerial discretion, impose discipline on insiders, and transform payout decisions into an internal governance outcome rather than a mere financial choice (Faccio et al., 2001; La Porta et al., 2000; Shleifer & Vishny, 1997).

In emerging and civil law markets such as Korea, where external investor protection remains comparatively weak, dividend payouts assume heightened importance as substitute governance mechanisms to safeguard minority shareholders (Claessens & Yurtoglu, 2013; Mitton, 2004). Empirical evidence from East Asia consistently demonstrates that payout policy is not neutral but instead reflects underlying agency conditions shaped by ownership concentration, business group affiliation, and institutional enforcement (Almeida et al., 2011; Claessens et al., 2000; Hwang et al., 2013).

2.2. Governance Mechanisms as Credibility Channels

Corporate governance mechanisms critically shape dividend policy by influencing the credibility of payout commitments. Formal governance structures, board oversight, audit quality, disclosure practices, and shareholder rights, constrain opportunistic behavior and enhance transparency in cash allocation (Faccio et al., 2001; Healy & Palepu, 2001; La Porta et al., 2000; Shleifer & Vishny, 1997). Stronger governance increases the likelihood that dividends function as credible commitments to mitigate agency costs rather than as discretionary or symbolic distributions (Mitton, 2004; Renneboog & Szilagyi, 2020).

In Korea, reforms such as the 2016 Stewardship Code and subsequent governance initiatives explicitly target improvements in payout credibility, aiming to minimize persistent discounts in equity valuations (Council, 2016; Kang, 2023). Governance mechanisms thus operate as credibility channels, signaling internal control quality and positioning governance strength as a determinant of payout reliability.

Hypothesis 1.

Corporate governance quality improves the credibility of dividend payouts.

2.3. Ownership Structure as an Institutional Regime

Ownership structure fundamentally conditions how governance mechanisms translate into corporate outcomes. In dispersed ownership contexts, formal governance arrangements play a primary role in constraining managerial discretion and aligning payout decisions with shareholder interests (Edmans, 2014; Fama & Jensen, 1983).

By contrast, firms with concentrated ownership, particularly Chaebol affiliates, exhibit internalized control, where dominant shareholders already exercise monitoring and decision authority (Atanassov & Mandell, 2018; Burkart et al., 1997; Gonzalez et al., 2017; Jeon et al., 2011; Shleifer & Vishny, 1986). In such settings, governance mechanisms do not uniformly enhance payout outcomes; their effectiveness depends on whether they complement or merely duplicate existing control (Mili et al., 2017). Internal monitoring channels such as boards, audit oversight, and disclosure may be saturated under concentrated ownership, while shareholder-facing mechanisms, shareholder rights and dividend signaling, retain binding power (Kwon & Han, 2020).

Crisis period evidence supports this conditional logic. Mili et al. (2017) show that during financial crises in emerging markets, dividend decisions respond heterogeneously to governance depending on firm-specific control structures. For example, in firms with CEO duality or high ownership concentration, board independence and activity exert a weaker influence on dividends, whereas shareholder-facing commitments remain effective in disciplining managerial discretion. This emphasizes the mechanism-specific nature of governance: effectiveness arises not from uniform implementation but from filling gaps left by internal control.

Accordingly, ownership structure functions as an institutional regime that shapes the governance–dividend relationship. Dividend responses to governance arise only when governance represents the marginal channel of control. Consequently, governance effectiveness is mechanism-specific and contingent on ownership, directly linking to payout credibility and extending prior evidence within the Korean context.

Hypothesis 2.

The effectiveness of governance mechanisms in shaping dividend payouts is conditional on the firm’s ownership regime.

2.4. Review Summary

Korea’s hybrid ownership and governance system offers a valuable setting to study how governance mechanisms affect dividend policy under persistent institutional frictions. While formal governance strengthens transparency and shareholder rights, concentrated ownership shapes which mechanisms remain effective. Evidence shows that dividend payouts reflect both governance quality and ownership structure: internal monitoring channels may be saturated under concentrated ownership, whereas shareholder-facing mechanisms retain influence. This study thus examines how governance interacts with ownership to enhance the credibility of dividend payouts.

3. Research Design

3.1. Data and Sample Construction

This study employs a balanced panel of 558 non-financial firms listed on the Korea Composite Stock Price Index (KOSPI) over the period 2011–2019, yielding 5022 firm-year observations. The sample period is chosen for methodological rather than policy-driven reasons. First, it represents the longest window over which corporate governance scores from the Korea Corporate Governance Service (KCGS) are consistently available using a stable scoring methodology. Second, it allows sufficient within-firm variation in governance quality and dividend behavior to support fixed effects and dynamic panel estimation. Together, these features ensure internal consistency and econometric comparability across firms and over time.

Governance variables are obtained from KCGS, which provides annually updated, standardized governance indices widely used in empirical corporate finance research on Korea (Korea Corporate Governance Service [KCGS], 2021). Financial and accounting data are sourced from Kis-Value (ValueSearch), a comprehensive database covering publicly listed Korean firms (OECD, 2018; Value Search, 2023). The combination of these two data sources enables a precise alignment between governance indicators and firm-level financial outcomes while minimizing measurement inconsistencies across datasets.

Sample construction follows the multistage screening process detailed in Appendix A and summarized in Table A1. Financial institutions are excluded due to their distinct regulatory environment and payout constraints, which are not comparable to those of industrial firms. Firms with incomplete governance or financial disclosures are removed to ensure balanced panel estimation and to avoid bias arising from intermittent reporting. Only firms with continuous observations over the sample period are retained. All continuous variables are winsorized at the 5th and 95th percentiles to reduce the influence of extreme values without truncating economically meaningful variation.

3.2. Variable Measurement

Dividend policy is measured using two complementary dependent variables. Dividend yield (DYD), defined as cash dividends divided by market capitalization, captures the intensity of payout policy in continuous form. In addition, a binary indicator for high dividend yield (HDYD) is constructed, taking the value of one when a firm’s dividend yield exceeds the annual sample median and zero otherwise. This binary specification mitigates the impact of skewness in dividend distributions, accommodates the prevalence of zero or low dividend payers, and reflects the discrete nature of dividend decisions documented in the literature (Atanassov & Mandell, 2018).

Corporate governance is measured using indices provided by the Korea Corporate Governance Service (KCGS). All KCGS governance scores are standardized on a scale ranging from 0 to 100, with higher values indicating stronger governance quality. Five governance variables are employed: the total corporate governance score (TCGS), which captures overall governance quality; board composition (BOD), reflecting board independence and effectiveness; the protection of shareholder rights (PSR), measuring mechanisms designed to safeguard minority shareholders; audit committee quality (AUDIT), assessing audit committee independence and expertise; and corporate disclosure (DISC), evaluating the transparency and scope of firm disclosures (Hwang et al., 2013; Korea Corporate Governance Service [KCGS], 2021; Lee & Tulcanaza-Prieto, 2024; OECD, 2018). These measures allow the analysis to distinguish between shareholder-facing governance mechanisms and internal monitoring structures, which is central to the study’s hypotheses.

Ownership concentration (OWN) is measured as the percentage of shares held by the largest shareholder or controlling group. This variable captures cross-firm heterogeneity in control structures and operationalizes the ownership dimension underlying the moderating hypothesis (Kwon & Han, 2020; Njoku & Lee, 2024). Control variables include leverage (DEBT), free cash flow (FCF), asset intensity (ASSINT), and profitability (ROE), which are standard determinants of dividend policy and payout capacity in both developed and emerging markets (Jabbouri, 2016; Nazar, 2021). The definitions and construction of all variables are summarized in Table 1.

Table 1. Variables and measurements.

Variable	Symbol	Measurement/Description
Dependent Variables
Dividend Yield	DYD	Cash dividends divided by market capitalization
High Dividend Yield	HDYD	Dummy = 1 if firm’s dividend yield exceeds sample median; 0 otherwise
Independent Variables
Corporate Governance Score	TCGS	Composite KCGS index capturing overall governance quality
Board Composition	BOD	KCGS score reflecting proportion and effectiveness of outside directors
Protection of Shareholder Rights	PSR	KCGS score measuring mechanisms safeguarding minority shareholders
Audit Committee Quality	AUDIT	KCGS score capturing audit committee independence and expertise
Corporate Disclosure	DISC	KCGS index evaluating transparency and disclosure practices
Ownership Concentration	OWN	Shareholding percentage of the largest or controlling shareholder group
Control Variables
Debt Ratio	DEBT	Total liabilities divided by total assets
Free Cash Flow	FCF	(Operating cash flow minus dividends paid) divided by total assets
Return on Equity	ROE	Net income divided by shareholders’ equity
Asset Intensity	ASSINT	Total assets divided by total sales
Firm Size	SIZE	Natural logarithm of total assets

Note: This table defines the variables used in this study, including their measurement. Supporting literature is cited in the main text.

3.3. Empirical Strategy

This study adopts a layered empirical strategy designed to evaluate the governance–dividend relationship and its conditional dependence on ownership structure. Rather than relying on a single estimator, the approach combines complementary methods to address distinct econometric concerns, unobserved heterogeneity, nonlinear payout behavior, and dynamic persistence, while keeping the fixed effects framework as the primary source of identification (Fulgence et al., 2024; Kwon & Han, 2020; Lee & Tulcanaza-Prieto, 2024; Njoku & Lee, 2024).

3.3.1. Baseline Fixed Effects Framework

The core empirical analysis relies on firm and year fixed effects panel regressions. This framework exploits within-firm variation over time to identify how changes in governance quality are associated with changes in dividend payouts, net of time-invariant firm characteristics such as organizational culture, business group affiliation, or persistent managerial preferences, as well as economy-wide shocks common to all firms.

Firm fixed effects absorb unobservable, stable traits that jointly influence governance structures and payout policy, thereby mitigating omitted variable bias. Year fixed effects control for macroeconomic conditions, regulatory changes, and market-wide payout norms. This specification is well suited to the Korean setting, where governance reforms and payout practices evolve gradually within firms rather than through abrupt cross-sectional shifts.

The baseline model evaluates whether governance quality and its constituent dimensions are systematically associated with dividend yield, consistent with governance operating as a credibility channel for payout commitments. Coefficient interpretation is therefore economic rather than mechanical: estimated effects capture how improvements in governance within a firm translate into changes in payout behavior, holding constant ownership structure and financial fundamentals (Arslan et al., 2019; Chen et al., 2005; Jabbouri, 2016; Nazar, 2021).

The baseline equation (model 1) is specified as follows:

$$\begin{array}{cc}\hfill DY{D}_{i,t}& ={\beta }_0+{\beta }_{1}TCG{S}_{i,t}+{\beta }_{2}BO{D}_{i,t}+{\beta }_{3}PS{R}_{i,t}+{\beta }_{4}AUDI{T}_{i,t}+{\beta }_{5}DIS{C}_{i,t}+{\beta }_{6}OW{N}_{i,t}\hfill \\ \hfill & +\sum _{j=1}^4{\gamma }_j{\mathrm{Controls}}_{j,i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}\hfill \end{array}$$

(1)

In this specification, $DY{D}_{i,t}$ denotes the dividend yield, calculated as cash dividends divided by market capitalization for firm i in year t. The main explanatory variables include the total corporate governance score ($TCG{S}_{i,t}$), along with its subdimensions: board effectiveness ($BO{D}_{i,t}$), shareholder rights ($PS{R}_{i,t}$), audit oversight ($AUDI{T}_{i,t}$), and disclosure quality ($DIS{C}_{i,t}$). Ownership concentration ($OW{N}_{i,t}$) is also included as a key explanatory variable, reflecting the heterogeneity in ownership structures across firms.

The control vector Controls_i,t comprises debt ratio (DEBT), free cash flow (FCF), asset intensity (ASSINT), and return on equity (ROE), all standard determinants of dividend policy in the literature. Firm fixed effects ${\mu }_i$ account for unobserved, time-invariant firm characteristics, while year fixed effects ${\lambda }_t$ absorb macroeconomic and regulatory shocks common to all firms. The error term ${\epsilon }_{i,t}$ captures idiosyncratic disturbances (Fulgence et al., 2024; Kwon & Han, 2020; Lee & Tulcanaza-Prieto, 2024; Njoku & Lee, 2024). A positive and statistically significant coefficient on $TCGS$ and its subcomponents, would support Hypotheses H₁ indicating that stronger governance mechanisms are associated with greater dividend payouts. This aligns with the view that effective governance reduces agency conflicts and, through sustained improvements, engenders consistent dividend discipline over time.

Ownership-Conditioned Identification

To test whether governance effectiveness depends on ownership architecture, we stratify the sample into Chaebol and non-Chaebol firms using Korea Fair Trade Commission (KFTC) classifications. KFTC criteria defined Chaebol affiliates as business groups with ownership concentration exceeding 30% of issued shares and total assets greater than KRW 5 trillion (Council, 2016; Kang, 2023; Korea Corporate Governance Service [KCGS], 2021; Lee & Tulcanaza-Prieto, 2024; Njoku & Lee, 2024). This subsample design operationalizes ownership structure as an institutional regime rather than as a simple continuous control.

We estimated identical specifications across ownership regimes to enable a direct comparison of governance coefficients without relying on functional-form assumptions through interaction terms. By doing so, we were able to interpret differences in coefficient magnitude or significance as evidence that governance mechanisms transmit into payout behavior differently depending on whether control is externally enforced or internally concentrated.

This approach is taken to mitigate concerns that binary ownership indicators may oversimplify complex control structures, so we ground the classification in regulatory definitions that jointly reflect ownership concentration and economic significance. This analysis does not claim that ownership concentration causes payout outcomes; instead, it tests whether governance mechanisms are binding margins of control under different ownership regimes.

3.3.2. High Dividend Threshold Analysis

Dividend yield may obscure nonlinear or threshold behavior, particularly where firms cluster around stable payout policies. To address this concern, a logistic specification models the probability that a firm crosses a high dividend threshold.

This specification does not replace the linear model; it complements it by examining whether governance quality influences the likelihood of adopting or maintaining a high-payout regime rather than marginal changes in payout levels. The binary model therefore enhances interpretability by linking governance to economically meaningful payout states, not just average effects.

$$\mathrm{Prob}(HDY{D}_{i,t}=1)={\displaystyle \frac{1}{1+exp\left(-\left[{\alpha }_0+{\sum }_{k=1}^5{\alpha }_k{\mathrm{GOV}}_{k,i,t}+{\alpha }_{6}OW{N}_{i,t}+{\sum }_{j=1}^4{\gamma }_j{\mathrm{Controls}}_{j,i,t}+{\mu }_i+{\lambda }_t\right]\right)}}$$

(2)

In this specification, $HDY{D}_{i,t}$ is a binary indicator equal to 1 if firm i in year t exhibits a high dividend yield, defined by crossing a predetermined payout threshold. The model estimates the probability that a firm adopts a sustained high-payout policy as a function of governance quality and ownership structure. The governance indicators ${\mathrm{GOV}}_{k,i,t}$ include the total corporate governance score ($TCGS$), board effectiveness ($BOD$), protection of shareholder rights ($PSR$), audit oversight ($AUDIT$), and disclosure quality ($DISC$). Ownership concentration ($OW{N}_{i,t}$) is also included as a key explanatory variable, reflecting the heterogeneity in ownership structures across firms. The control variables include debt ratio, free cash flow, asset intensity, and return on equity. Firm fixed effects ${\mu }_i$ and year fixed effects ${\lambda }_t$ are incorporated to account for unobserved heterogeneity and macroeconomic shocks, respectively. A positive and statistically significant coefficient on any governance indicator or $OWN$ implies that stronger governance or concentrated ownership increases the likelihood of a firm maintaining high dividend payouts. This robustness check complements the baseline linear model by capturing nonlinear effects and threshold behavior. That way, we can understand how governance and ownership shape dividend policy in practice.

3.4. Dynamic GMM as a Robustness Layer

Dividend policy is known to exhibit persistence, and governance reforms may be endogenously related to firms’ payout preferences. To address these concerns, this study employs a dynamic System GMM estimator as a robustness exercise rather than as a primary identification strategy. The GMM framework incorporates lagged dividend yield to capture payout inertia and uses internal instruments to mitigate reverse causality between governance quality and dividend policy.

Lagged levels and differences in endogenous variables are employed as internal instruments, with instrument validity and serial correlation assessed using the Hansen J-test and the Arellano–Bond AR(2) test. Accordingly, the role of the GMM analysis is confirmatory, reinforcing the credibility of the baseline findings under intertemporal adjustment.

The dynamic specification is given by

$$DY{D}_{i,t}={\alpha }_0+{\alpha }_{1}DY{D}_{i,t-2}+\sum _{k=1}^K{\theta }_k{\mathrm{GOV}}_{k,i,t-1}+\beta {\mathrm{OWN}}_{i,t-1}+\sum _{j=1}^J{\gamma }_j{\mathrm{Controls}}_{j,i,t-1}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(3)

where $DY{D}_{i,t}$ denotes the dividend yield of firm i in year t, and $DY{D}_{i,t-2}$ captures the persistence of dividend policy. $GO{V}_{k,i,t-1}$ represents lagged corporate governance measures, $OW{N}_{i,t-1}$ denotes ownership structure, and $Control{s}_{j,i,t-1}$ include standard firm-level covariates. Firm fixed effects ${\mu }_i$ control for time-invariant heterogeneity, year fixed effects ${\lambda }_t$ absorb common shocks, and ${\epsilon }_{i,t}$ is the idiosyncratic error term.

Consistency between the dynamic GMM and fixed effects estimates is interpreted as reassurance that the governance–dividend relationship is not driven by mechanical persistence or dynamic misspecification. The GMM analysis is therefore confirmatory, reinforcing the credibility of the baseline results under intertemporal adjustment (Arellano & Bond, 1991; Arellano & Bover, 1995).

The results from this robustness exercise are discussed in Section 5 alongside complementary binary logistic regressions that examine dividend payment likelihood. Together, these analyses demonstrate that the core findings are not sensitive to alternative functional forms or estimation strategies, but reflect a stable association between governance quality, ownership structure, and dividend credibility in the Korean context.

3.5. Diagnostic Mean Comparisons

Equality-of-means tests between Chaebol and non-Chaebol firms provide descriptive validation for the ownership-conditioned analysis. Demonstrating systematic differences in governance quality, ownership concentration, and dividend behavior motivates the regime-based empirical design and supports the interpretation of ownership structure as a conditioning environment rather than a simple control variable. This test evaluates whether the sample means of the two groups differ significantly using the following standard two-sample t-statistic:

$$t={\displaystyle \frac{{\overline{X}}_1-{\overline{X}}_2}{\sqrt{\frac{s_1^2}{n_1}+\frac{s_2^2}{n_2}}}}$$

(4)

where

• ${\overline{X}}_1=\mathrm{Sample}\mathrm{means}\mathrm{of}\mathrm{the}\mathrm{Chaebol}\mathrm{groups}.$
• ${\overline{X}}_2=\mathrm{Sample}\mathrm{means}\mathrm{of}\mathrm{the}\mathrm{non}\text{-}\mathrm{Chaebol}\mathrm{groups}.$
• $s_1^2=\mathrm{Variances}\mathrm{for}\mathrm{Chaebol}\mathrm{group}.$
• $s_2^2=\mathrm{Variances}\mathrm{for}\mathrm{non}\text{-}\mathrm{Chaebol}\mathrm{group}.$
• $n_1=\mathrm{Sample}\mathrm{size}\mathrm{for}\mathrm{Chaebol}\mathrm{group}.$
• $n_2=\mathrm{Sample}\mathrm{size}\mathrm{for}\mathrm{non}\text{-}\mathrm{Chaebol}\mathrm{group}.$

4. Empirical Results

4.1. Descriptive Analysis

Table 2 summarizes the descriptive statistics for the 5022 firm-year observations used in this study, offering an overview of dividend practices, governance quality, ownership structures, and financial characteristics among Korean listed firms. The average dividend yield is 1.2% (median 1.0%), indicating stable payout patterns consistent with the conservative distribution norms in Asian markets. The relatively low dispersion (SD = 1.3%) underlines the persistence of modest payout policies.

Table 2. Descriptive statistics.

Variable	N	Mean	Median	Std. Dev.
DYD	5022	0.012	0.010	0.013
TCGS	5022	29.073	28.670	8.378
BOD	5022	13.800	12.500	7.739
PSR	5022	45.112	45.000	11.931
AUDIT	5022	44.648	42.000	15.116
DISC	5022	24.237	20.000	13.269
OWN	5022	28.095	24.890	13.736
DEBT	5022	0.406	0.408	0.203
FCF	5022	0.043	0.039	0.056
ASSINT	5022	2.814	1.312	4.360
ROE	5022	0.024	0.031	0.101
SIZE	5022	26.242	26.165	1.402

Note: N represents the number of sample observations.

Corporate governance quality, measured using the total corporate governance score (TCGS), shows an average of 29.07 (median = 28.67), reflecting moderate alignment with post-reform governance expectations. Among the subdimensions, board effectiveness (BOD) exhibits the highest variability (SD = 7.739), suggesting heterogeneity in board structures. Shareholder rights protection (PSR = 45.11) and audit committee quality (AUDIT = 44.65) indicate relatively strong institutional safeguards. However, corporate disclosure quality (DISC = 24.24) comparatively remains the weakest component, reinforcing ongoing concerns about transparency gaps in Korean corporations.

Ownership concentration averages 28.1%, revealing the structural dominance of major shareholders, an important sustainability consideration given the implications for power asymmetries and minority shareholder vulnerability. Firms show moderate leverage (DEBT = 40.6%), positive free cash flow (FCF = 4.3%), and modest profitability (ROE = 2.4%). Asset intensity (ASSINT = 2.81) varies widely (SD = 4.36), reflecting sectoral differences in capital deployment. To minimize the impact of extreme observations and ensure the validity of the estimates, we winsorize all continuous variables at the 5% and 95% levels.

4.2. Comparison of Mean Values Between Chaebol and Non-Chaebol Entities

Table 3 provides strong evidence of systematic structural differences across ownership regimes. Dividend yields are significantly higher in Chaebol firms (mean = 0.0140) than in non-Chaebol firms (mean = 0.0115; t = 6.76 ***), suggesting more generous payout policies, possibly reflecting internal capital market dynamics or strategic signaling in business groups. Corporate governance quality, measured by TCGS, is marginally higher among Chaebol firms (mean = 29.33 vs. 28.91; t = 1.74 *), with the most pronounced difference appearing in disclosure quality (DISC; t = 3.31 ***). Differences in board structure (BOD), shareholder rights (PSR), and audit oversight (AUDIT) are statistically insignificant, implying similar formal governance structures across firm types.

Table 3. Statistical tests for mean differences between Chaebol and Non-Chaebol entities.

Variable	Total Sample [N = 5022]		Chaebol Sample [N = 1918]		Non-Chaebol Sample [N = 3101]		Mean Diff.	Std. Error	t-Value
	Mean	Std. Dev	Mean	Std. Dev	Mean	Std. Dev
DYD	0.0125	0.0125	0.0140	0.0131	0.0115	0.0121	0.0025	0.00037	6.76 ***
TCGS	29.0727	8.3781	29.3326	8.3063	28.9093	8.4225	0.4233	0.2426	1.74 *
BOD	13.8000	7.7392	14.0046	7.7927	13.6771	7.7075	0.3275	0.2254	1.45
PSR	45.1121	11.9308	44.7771	11.5413	45.3102	12.1612	−0.5331	0.3423	−1.56
AUDIT	44.6482	15.1164	44.9965	14.8766	44.4405	15.2630	0.5559	0.4365	1.27
DISC	24.2368	13.2689	25.0325	13.7068	23.7428	12.9727	1.2896	0.3902	3.31 ***
OWN	28.0952	13.7364	42.8804	9.2086	18.9503	5.8384	23.9301	0.2350	101.85 ***
DEBT	0.4062	0.2034	0.4085	0.2079	0.4048	0.2006	0.0037	0.0060	0.62
FCF	0.0426	0.0559	0.0472	0.0567	0.0397	0.0552	0.0076	0.0016	4.60 ***
ASSINT	2.8140	4.3597	2.9375	4.6999	2.7335	4.1266	0.2039	0.1304	1.56
ROE	0.0245	0.1014	0.0328	0.0978	0.0193	0.1033	0.0134	0.0029	4.65 ***

Note: Significance at the 1% and 10% levels is denoted by *** and *, respectively.

Ownership concentration shows the strongest divergence: Chaebol firms average 42.88% vs. 18.95% for non-Chaebols (t = 101.85 ***), reflecting entrenched control structures typical of Korean conglomerates. Financial characteristics also vary: Chaebol firms exhibit significantly higher free cash flow (t = 4.60 ***) and profitability (t = 4.65 ***), though leverage (DEBT; t = 0.62) and asset intensity (ASSINT; t = 1.56) show no meaningful differences. These results justify the moderation analysis in Section 4.5 by demonstrating structural differences with implications for governance effectiveness and sustainability-aligned financial behavior.

4.3. Correlation and Multicollinearity Analysis

Table 4 presents correlations among the main variables. Dividend yield (DYD) exhibits positive associations with all governance metrics and key financial controls, suggesting that firms with stronger governance and healthier financial profiles tend to distribute higher dividends. DYD is negatively related to leverage (DEBT), consistent with the financial constraint logic that debt reduces payout capacity. Governance dimensions, board effectiveness (BOD), shareholder rights (PSR), and disclosure (DISC), display moderate intercorrelations, indicating related but distinct institutional mechanisms. Ownership concentration (OWN) correlates negatively with BOD and PSR, These correlations are descriptive, serving to identify data relationships and guide subsequent model formulation.

Table 4. Pearson’s correlation matrix of study variables.

Variables	TCGS	BOD	PSR	AUDIT	DISC	OWN	DEBT	FCF	ASSINT	ROE	DYD
TCGS	1
BOD	0.744 ***	1
PSR	0.762 ***	0.484 ***	1
AUDIT	0.582 ***	0.556 ***	0.238 ***	1
DISC	0.524 ***	0.524 ***	0.226 ***	0.374 ***	1
OWN	−0.015	−0.037 ***	−0.055 ***	−0.012	−0.003	1
DEBT	0.039 ***	0.137 ***	−0.075 ***	0.101 ***	0.073 ***	0.017	1
FCF	0.108 ***	0.078 ***	0.099 ***	0.093 ***	0.131 ***	0.045 ***	−0.138 ***	1
ASSINT	0.009	−0.009	0.041 ***	0.008	−0.043 ***	0.016	−0.352 ***	−0.156 ***	1
ROE	0.125 ***	0.055 ***	0.113 ***	0.072 ***	0.116 ***	0.065 ***	−0.247 ***	0.418 ***	−0.044 ***	1
DYD	0.145 ***	0.047 ***	0.196 ***	0.033 **	0.031 **	0.098 ***	−0.252 ***	0.227 ***	0.036 ***	0.322 ***	1

Note: Statistical significance at the 1% and 5% levels is denoted by *** and ** based on Pearson’s correlation coefficients.

The multicollinearity diagnostic results provided in Table A3 within Appendix A including variance inflation factors and condition indices confirm the absence of problematic multicollinearity among the primary explanatory variables. These diagnostics support the reliability of the regression estimates because all variance inflation factor values remain below the conventional benchmark of 10. The highest recorded value is 6.171 for the total corporate governance score which validates the suitability of the governance metrics for rigorous empirical analysis (Kwon & Han, 2020; Lee & Tulcanaza-Prieto, 2024; Njoku & Lee, 2024).

4.4. Baseline Regression Analysis

Table 5 reports the fixed effects regression estimates assessing the influence of governance quality on dividend yield. Across all models, governance emerges as a significant predictor of dividend policy.

Table 5. The results of OLS estimation using the full sample.

Variable	Model 1		Model 2		Model 3		Model 4		Model 5		Model 6
	Beta	t-Stat	Beta	t-Stat	Beta	t-Stat	Beta	t-Stat	Beta	t-Stat	Beta	t-Stat
Intercept	−0.0051	(−1.091)	0.0136 ***	(12.527)	0.0088 ***	(6.873)	0.0144 ***	(12.35)	0.0144 ***	(13.535)	0.0060 ***	(6.320)
TCGS	0.0040 ***	(4.242)									0.00009 **	(1.960)
BOD			0.0001 ***	(4.522)							0.00005	(1.491)
PSR					0.00012 ***	(6.693)					0.00014 ***	(5.791)
AUDIT							0.00002 **	(2.232)			0.00001	(0.566)
DISC									0.00001	(0.441)	0.00004 **	(2.276)
OWN	0.0001 ***	(4.183)	0.0001 ***	(4.234)	0.0001 ***	(3.979)	0.0001 ***	(4.196)	0.0001 ***	(4.200)	0.00008 ***	(6.928)
DEBT	−0.0109 ***	(−6.703)	−0.0110 ***	(−6.754)	−0.0104 ***	(−6.437)	−0.0109 ***	(−6.747)	−0.0109 ***	(−6.749)	−0.0108 ***	(−11.839)
FCF	0.0077 ***	(2.768)	0.0078 ***	(2.813)	0.0071 ***	(2.583)	0.0078 ***	(2.804)	0.0078 ***	(2.810)	0.0214 ***	(6.586)
ASSINT	−0.0009 ***	(−2.336)	−0.0009 ***	(−2.529)	−0.0009 ***	(−2.451)	−0.0009 ***	(−2.570)	−0.0009 ***	(−2.573)	−0.00003 *	(−1.775)
ROE	0.0125 ***	(8.078)	0.0122	(7.935)	0.0127 ***	(8.264)	0.0122 ***	(7.886)	0.0122 ***	(7.883)	0.0268 ***	(14.844)
Year Fixed Effect	Yes		Yes		Yes		Yes		Yes		Yes
Firm Fixed Effect	Yes		Yes		Yes		Yes		Yes		Yes
R2	0.644		0.6428		0.6462		0.6426		0.6426		0.6460
Adj. R2	0.5983		0.5969		0.6007		0.5967		0.5967		0.6002
F-statistics	14.0915 ***		14.0143 ***		14.2228 ***		14.0035 ***		14.0036 ***		14.1033
Prob(F-stat.)	0.0000		0.0000		0.0000		0.0000		0.0000		0.0000
Hausman Chi2	61.2238 ***		68.1368 ***		76.2252 ***		72.2505 ***		77.2058 ***		109.0763 ***

Note: The reported coefficients are represented by Beta and the associated t-statistics are provided in parentheses where the symbols *** and ** along with * mark statistical significance at the 1% and 5% and 10% thresholds.

In Model 1, the total corporate governance score (TCGS) is positively related to dividend yield ($\beta =0.0040$, $t=4.242$, $p<0.01$). Subdimension tests show similarly positive outcomes: board effectiveness (BOD) in Model 2 ($\beta =0.0001$, $t=4.522$, $p<0.01$), shareholder rights protection (PSR) in Model 3 ($\beta =0.00012$, $t=6.693$, $p<0.01$), and audit committee quality (AUDIT) in Model 4 ($\beta =0.00002$, $t=2.232$, $p<0.05$). Corporate disclosure (DISC) becomes significant only when all governance variables are jointly included (Model 6: $\beta =0.00004$, $t=2.276$, $p<0.05$), implying interdependency across governance mechanisms. Ownership concentration (OWN), included as an independent variable in all models, shows a positive and significant coefficient.

Control variables behave as expected. Free cash flow (FCF), and profitability (ROE) exhibit strong positive associations with dividend yield. Leverage (DEBT) and asset intensity (ASSINT) reduce dividend payouts, consistent with financing constraints and capital intensiveness.

Model diagnostics further validate these results: $R^2$ values approach 0.64, all F-statistics are highly significant, and the Hausman test (61.2238, $p<0.001$) strongly favors fixed effects. Durbin–Watson statistics approximate 2, and all VIF values remain within acceptable thresholds. All reported t-statistics use firm-clustered standard errors (clustered at the firm level) to address heteroskedasticity and within-firm serial correlation. Coefficients are therefore robust to common violations of classical assumptions in panel data.

4.5. Ownership Structure Analysis

Table 6 and Table 7 evaluate whether ownership heterogeneity affects the governance–dividend relationship, directly testing the second hypothesis. Distinguishing Chaebol and non-Chaebol firms reveals meaningful asymmetries.

Table 6. Ownership structure effect under non-Chaebol group.

Variable	Model 1		Model 2		Model 3		Model 4		Model 5		Model 6
	Beta	t-Value	Beta	t-Value	Beta	t-Value	Beta	t-Value	Beta	t-Value	Beta	t-Value
Intercept	−0.0124 ***	(−3.309)	0.0114 ***	(12.295)	0.0045 ***	(3.980)	0.0116 ***	(11.457)	0.0125 ***	(13.297)	0.0080 ***	(4.232)
TCGS	0.0053 ***	(6.931)									0.00003 *	(1.679)
BOD			0.00013 ***	(4.765)							0.00004 *	(1.943)
PSR					0.00018 ***	(11.125)					0.0001 ***	(5.025)
AUDIT							0.00003 ***	(2.447)			0.00003	(1.426)
DISC									0.00002	(1.357)	0.00004 *	(1.870)
OWN	0.0001 ***	(3.634)	0.0001 ***	(3.669)	0.0001 ***	(3.586)	0.0001 ***	(3.232)	0.0001 ***	(3.230)	0.00008	(1.549)
DEBT	−0.0120 ***	(−10.664)	−0.0120 ***	(−10.513)	−0.0108 ***	(−9.774)	−0.0115 ***	(−10.100)	−0.0113 ***	(−9.951)	−0.0047 **	(−2.356)
FCF	0.0202 ***	(4.963)	0.0218 ***	(5.359)	0.0203 ***	(5.088)	0.0230 ***	(5.640)	0.0236 ***	(5.796)	0.0093 ***	(2.756)
ASSINT	−0.0007 ***	(−2.482)	−0.0006 ***	(−2.405)	−0.0006 ***	(−2.115)	−0.0006 ***	(−2.121)	−0.0005 ***	(−1.980)	−0.0001 *	(−1.824)
ROE	0.0243 ***	(10.982)	0.0254 ***	(11.509)	0.0244 ***	(11.240)	0.0257 ***	(11.590)	0.0257 ***	(11.565)	0.0128 ***	(6.855)
Year Fixed Effect	Yes		Yes		Yes		Yes		Yes		Yes
Firm Fixed Effect	Yes		Yes		Yes		Yes		Yes		Yes
R2	0.659		0.6585		0.6619		0.6587		0.6587		0.6627
Adj. R2	0.6047		0.6041		0.6081		0.6043		0.6043		0.6084
F-statistics	12.1312 ***		12.1031 ***		12.2894 ***		12.1140 ***		12.1151 ***		12.2000 ***
Prob > F	0.0000		0.0000		0.0000		0.0000		0.0000		0.0000
Hausman Chi2	48.4718 ***		51.3603 ***		62.6630 ***		59.4319 ***		64.6210 ***		76.3321 ***

Note: Beta denotes the regression coefficients and the associated t-statistics are provided in parentheses where the symbols *** and ** as well as * indicate significance at the 1% and 5% and 10% levels, respectively.

Table 7. Ownership structure effect under Chaebol group.

Variable	Model 1		Model 2		Model 3		Model 4		Model 5		Model 6
	Beta	t-Value	Beta	t-Value	Beta	t-Value	Beta	t-Value	Beta	t-Value	Beta	t-Value
Intercept	−0.006	(−0.728)	0.0201 ***	(6.765)	0.0159 ***	(5.041)	0.0199 ***	(6.582)	0.0208 ***	(7.112)	0.0120 ***	(5.527)
TCGS	0.0056 ***	(3.557)									0.0003 ***	(3.877)
BOD			0.00008	(1.455)							0.00002	(0.307)
PSR					0.00012 ***	(3.887)					0.00009 ***	(2.653)
AUDIT							0.00003	(1.432)			0.00002	(0.668)
DISC									0.00003	(0.970)	0.000004	(0.119)
OWN	0.000045	(0.826)	0.000042	(0.771)	0.000042	(0.774)	0.000038	(0.709)	0.000038	(0.699)	0.00005	(0.850)
DEBT	−0.0213 ***	(−6.974)	−0.0216 ***	(−7.041)	−0.0211 ***	(−6.898)	−0.0216 ***	(−7.055)	−0.0215 ***	(−7.017)	−0.0208 ***	(−6.787)
FCF	0.0163 ***	(2.905)	0.0168 ***	(2.999)	0.0168 ***	(3.021)	0.0170 ***	(3.019)	0.0175 ***	(3.109)	0.0187 ***	(3.383)
ASSINT	−0.0017 **	(−2.052)	−0.0018 ***	(−2.285)	−0.0018 **	(−2.261)	−0.0018 **	(−2.224)	−0.0018 **	(−2.220)	−0.00028 *	(−1.724)
ROE	0.0147 ***	(5.274)	0.0142 ***	(5.093)	0.0148 ***	(5.296)	0.0141 ***	(5.042)	0.0142 ***	(5.075)	0.0156 ***	(5.649)
Year Fixed Effect	Yes		Yes		Yes		Yes		Yes		Yes
Firm Fixed Effect	Yes		Yes		Yes		Yes		Yes		Yes
$R^2$	0.6867		0.6847		0.6872		0.6847		0.6845		0.6875
Adj. $R^2$	0.6272		0.6248		0.6278		0.6248		0.6245		0.6272
F-statistics	11.5411 ***		11.4321 ***		11.5665 ***		11.4314 ***		11.4199 ***		11.4045
Prob(F-statistics)	0.0000		0.0000		0.0000		0.0000		0.0000		0.0000
Hausman ${\chi }^2$	28.3887 ***		31.9550 ***		30.6965 ***		31.8650 ***		32.7748 ***		52.7429 ***

Note: Beta denotes the regression coefficients and the associated t-statistics are provided in parentheses where the symbols *** and ** as well as * indicate significance at the 1% and 5% and 10% levels, respectively.

4.5.1. Non-Chaebol Firms

For firms outside Chaebol networks (Table 6), governance mechanisms exert strong and consistent effects on dividend policy. The composite governance score (TCGS) is highly significant ($\beta =0.0053$, $t=6.931$, $p<0.01$), as shown in Model 1. This governance influence is echoed in Model 2, where board effectiveness (BOD) maintains a significant positive association ($\beta =0.00013$, $t=4.765$, $p<0.01$). Model 3 reveals shareholder rights protection (PSR) as the most influential governance factor ($\beta =0.00018$, $t=11.125$, $p<0.01$). Model 4’s audit committee quality (AUDIT) remains significant ($\beta =0.00003$, $t=2.447$, $p<0.01$), whereas corporate disclosure (DISC) shows a weakly positive effect ($\beta =0.00002$, $t=1.357$) in Model 5. However, when all governance dimensions are jointly included in Model 6, DISC becomes statistically significant at the 10% level ($\beta =0.00004$, $t=1.870$), consistent with the baseline regression specification. Ownership concentration (OWN) retains significance across models ($\beta \approx 0.0001$, t between 3.232 and 3.669).

Financial variables consistently contribute to dividend policy predictions. Debt ratio (DEBT) displays a strong negative effect ($\beta =-0.0120$, $t=-10.664$, $p<0.01$). Free cash flow (FCF) and return on equity (ROE) serve as strong positive predictors ($\beta =0.0202$, $t=4.963$; $\beta =0.0243$, $t=10.982$, both $p<0.01$). Asset intensity (ASSINT) negatively associates with dividend yield ($\beta =-0.0007$, $t=-2.482$, $p<0.01$).

4.5.2. Chaebol Firms

In Chaebol-affiliated firms (Table 7), the influence of governance is more selective. The composite score (TCGS) remains significant ($\beta =0.0056$, $t=3.557$, $p<0.01$), as does shareholder rights protection (PSR; $\beta =0.00012$, $t=3.887$, $p<0.01$). Board effectiveness (BOD) is not statistically significant ($\beta =0.00008$, $t=1.455$). Audit committee quality (AUDIT) and corporate disclosure (DISC) do not show significant impacts ($\beta =0.00003$, $t=1.432$; $\beta =0.00003$, $t=0.970$).

Financial variables play a consistent role. Debt ratio (DEBT) is strongly negatively related to dividend yield ($\beta$ ranges from $-0.0213$ to $-0.0215$, t from $-6.898$ to $-7.055$, $p<0.01$). Free cash flow (FCF) shows positive effects ($\beta$ between $0.0163$ and $0.0175$, t from $2.905$ to $3.109$, $p<0.01$), as does return on equity (ROE; $\beta$ from $0.0141$ to $0.0148$, t between $5.042$ and $5.296$, $p<0.01$). Ownership concentration (OWN) is not significant ($\beta$ ranges from $0.000038$ to $0.000045$; $t<1.0$). Asset intensity (ASSINT) negatively associates with dividends ($\beta$ between $-0.0017$ and $-0.0018$, t from $-2.052$ to $-2.285$, $p<0.05$).

The models exhibit strong fit, with $R^2$ values around 0.685 and adjusted $R^2$ between 0.6245 and 0.6278. F-tests are highly significant, and Hausman’s tests favor fixed effects estimation.

4.6. Discussion

4.6.1. Corporate Governance Quality and Dividend Credibility

This study posits that corporate governance quality enhances the credibility of dividend payouts by constraining insider discretion and strengthening firms’ commitment to cash distributions. The full sample results in Table 5 provide strong empirical support for this hypothesis. Across specifications, the aggregate governance score (TCGS) enters positively and significantly, indicating that firms with stronger governance frameworks distribute higher dividend yields after controlling for firm and year fixed effects, leverage, profitability, and cash flow conditions.

The economic magnitude of this relationship is substantial. Using the full sample fixed effects estimate ($\beta =0.00040$), a one-standard-deviation increase in governance quality (8.378 points; Table 2) raises dividend yield by 0.335 percentage points, or 33.5 basis points. Relative to the sample mean dividend yield of 1.20%, this represents an increase of approximately 28%, confirming that governance improvements translate into economically meaningful payout commitments rather than marginal statistical effects. This finding is consistent with agency-based theories that view dividends as a mechanism for reducing free cash flow problems and enhancing payout credibility.

Disaggregated governance components reinforce this interpretation. Board effectiveness, shareholder rights protection, and audit oversight are each positively associated with dividend yield in the full sample, indicating that governance operates through multiple channels, such as monitoring, rights protection, and financial oversight, to enhance dividend credibility. Disclosure quality alone exhibits weaker standalone effects, suggesting that transparency without enforcement capacity is insufficient to generate credible payout commitments.

In sum, the full sample evidence strongly supports H1, showing that dividend policy functions as an internal governance outcome reflecting sustained discipline in cash allocation.

4.6.2. Ownership Regimes and Conditional Governance Effectiveness

While governance quality improves dividend credibility on average, H2 predicts that its effectiveness is conditional on the firm’s ownership regime. Table 6 and Table 7 provide clear evidence of this conditionality by contrasting governance effects between non-Chaebol and Chaebol firms.

In non-Chaebol firms (Table 6), governance mechanisms exert a strong and broad influence on dividend payouts. The aggregate governance score is highly significant, and its economic impact exceeds that of the full sample. A one-standard-deviation increase in TCGS ($\beta =0.00053$) raises dividend yield by 44.4 basis points, equivalent to approximately 39% of the subgroup’s mean payout. Moreover, multiple governance components, namely board effectiveness, shareholder rights, audit oversight, and disclosure, remain statistically significant. This pattern indicates that when ownership is relatively dispersed, formal governance structures serve as the primary channels through which agency conflicts are mitigated and dividend commitments are enforced.

By contrast, the Chaebol subsample exhibits a sharply different mechanism-level pattern (Table 7). Although the aggregate governance score remains economically large ($\beta =0.00056$, implying a 46.9-basis-point increase per standard deviation), most internal monitoring mechanisms, such as boards, audit oversight, and disclosure—lose statistical significance once ownership is concentrated. Shareholder rights emerge as the only consistently effective governance channel. Ownership concentration itself becomes insignificant, a pointer to the saturation of internal monitoring under dominant control.

The comparison summarized in Table 8 highlights this divergence. While governance remains relevant in Chaebol firms only through shareholder-facing mechanisms, internal monitoring channels provide little incremental discipline. This evidence confirms that governance effectiveness is conditional and mechanism-specific, rather than uniformly diminished.

Table 8. Governance mechanism effects across ownership regimes.

Panel A. Governance Variables ($\mathit{\beta }$ with t-Statistics)
Variable	Full Sample	Non-Chaebol	Chaebol
TCGS	0.0040 ***	0.0053 ***	0.0056 **
	(4.24)	(6.93)	(3.56)
BOD	0.0010 ***	0.0013 ***	0.0008
	(4.52)	(4.77)	(1.46)
PSR	0.00012 ***	0.00018 ***	0.00012 ***
	(6.69)	(11.13)	(3.89)
AUDIT	0.00002 **	0.00003 **	0.00003
	(2.23)	(2.45)	(1.43)
DISC	0.0001	0.0002	0.0003
	(0.44)	(1.36)	(0.97)
OWN	0.00010 ***	0.00010 ***	0.00005
	(4.18)	(3.63)	(0.83)

Note: This table reports governance mechanism effects across ownership regimes. t-statistics are reported in parentheses below each coefficient. *** and ** denote significance at the 1% and 5% levels, respectively.

These findings support H2, showing that the governance–dividend relationship is systematically conditioned by ownership regime.

4.6.3. Dividend Credibility Under Institutional Constraints

The evidence across Table 3, Table 4, Table 5, Table 6, Table 7 and Table 8 indicates that corporate governance shapes dividend policy primarily through credibility rather than payout levels per se. In firms with relatively dispersed ownership, governance mechanisms function as binding constraints that discipline insider discretion and convert dividends into credible, repeatable commitments. Under concentrated ownership, by contrast, dividend policy reflects the strategic allocation choices of controlling shareholders, and only governance mechanisms that directly protect minority investors retain marginal effectiveness.

This distinction explains why aggregate governance remains economically significant among Chaebol affiliates even as several individual mechanisms lose statistical power. Governance does not vanish under concentrated control; instead, its disciplining role becomes selective and mechanism-specific. Evidently, boards and audit committees are largely influenced by controlling shareholders and therefore provide limited additional discipline. By contrast, shareholder rights continue to matter because they directly limit insiders’ ability to divert or tunnel firm resources, suppress dividends, or redistribute cash within business groups at the expense of minority investors. Dividend payouts thus retain credibility in Chaebol firms only through governance mechanisms that impose external, rule-based constraints on controlling owners, rather than through internal monitoring structures.

The associated economic magnitudes are substantial. A one-standard-deviation improvement in governance is associated with an increase in dividend yield of approximately 33.5 basis points in the full sample (about 28% of the mean yield), 44.4 basis points in non-Chaebol firms (roughly 39% of the mean), and 46.9 basis points in Chaebol firms (around 34% of the mean). These effects are not cosmetic: they are large enough to influence valuation, affect payout stability, and remain economically interpretable to both asset-pricing and corporate governance audiences.

Taken together, these findings substantiate both H1 and H2. Dividend credibility emerges from the interaction between governance quality and ownership structure, not from governance in isolation. Having quantified the economic magnitudes and isolated mechanism-specific effects, this study offers a more granular account of payout behavior in concentrated ownership systems and advances the literature on corporate governance and dividend policy in the Republic of Korea.

5. Robustness Analysis of Alternative Outcomes and Dynamic Endogeneity

To ensure that the documented governance–dividend relationship is not driven by model specification, payout measurement, or endogeneity, we conduct two complementary robustness analyses using (i) a binary payout commitment framework and (ii) a dynamic panel System GMM estimator.

5.1. Binary Dividend Commitment

Table 9 presents fixed effects binary logistic regressions where the dependent variable equals one if a firm maintains a high dividend yield relative to the annual sample median. This specification captures dividend commitment rather than payout magnitude and allows a direct economic interpretation of governance effects on the probability of sustaining high dividends.

Table 9. Binary logistic regression results for high dividend likelihood.

Variable	Model 1		Model 2		Model 3		Model 4		Model 5		Model 6
	Beta	z-Stat	Beta	z-Stat	Beta	z-Stat	Beta	z-Stat	Beta	z-Stat	Beta	z-Stat
Intercept	−0.303 **	(−2.022)	0.361 ***	(3.033)	−0.900 ***	(−5.412)	0.266 **	(1.951)	0.457 ***	(3.779)	−1.181 ***	(−6.444)
TCGS	0.032 ***	(8.354)									0.003 *	(1.928)
BOD			0.017 ***	(4.144)							0.014 **	(2.162)
PSR					0.031 ***	(11.67)					0.036 ***	(7.700)
AUDIT							0.007 ***	(3.342)			0.005 *	(1.726)
DISC									0.004	(1.497)	0.001	(0.194)
OWN	0.008 ***	(3.659)	0.008 ***	(3.587)	0.010 ***	(4.183)	0.008 ***	(3.478)	0.008 ***	(3.435)	0.010 ***	(4.227)
DEBT	−2.643 ***	(−14.709)	−2.618 ***	(−14.453)	−2.485 ***	(−13.837)	−2.564 ***	(−14.271)	−2.501 ***	(−14.026)	−2.259 ***	(−12.712)
FCF	2.603 ***	(4.094)	2.730 ***	(4.308)	2.548 ***	(3.981)	2.763 ***	(4.364)	2.842 ***	(4.481)	2.831 ***	(4.423)
ASSINT	−0.121 ***	(−3.010)	−0.116 ***	(−2.915)	−0.114 ***	(−2.839)	−0.111 ***	(−2.781)	−0.100 ***	(−2.531)	−0.007	(−0.902)
ROE	5.687 ***	(13.955)	5.867 ***	(14.430)	5.799 ***	(14.156)	5.905 ***	(14.490)	5.931 ***	(14.543)	5.954 ***	(14.484)
Year Fixed Effect	Yes		Yes		Yes		Yes		Yes		Yes
Firm Fixed Effect	Yes		Yes		Yes		Yes		Yes		Yes
McFadden $R^2$	0.129		0.121		0.139		0.120		0.119		0.139
Akaike Information Criterion	1.210		1.221		1.196		1.222		1.224		1.198
LR Statistic	897.950 ***		844.251 ***		967.398 ***		838.153 ***		829.193 ***		968.502
Prob(LR Statistic)	0.0000		0.0000		0.0000		0.0000		0.0000		0.0000
Log-likelihood	−3029.931		−3056.780		−2995.207		−3059.829		−3064.309		−2994.654

Note: This table reports estimates from a binary logistic regression predicting the probability of a firm maintaining high-dividend status. The dependent variable equals one if dividend yield exceeds the sample median and zero otherwise. Coefficient estimates (Beta) are shown with z-statistics in parentheses. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

The estimated coefficient on the composite governance score (TCGS) is positive and statistically significant across specifications. In Model 1, a one-unit increase in TCGS raises the log-odds of maintaining a high dividend by $0.032$ ($z=8.354$, $p<0.01$), while the effect remains significant, though smaller, in the fully specified model ($\beta =0.003$, $z=1.928$, $p<0.10$). This implies that improvements in governance materially increase the likelihood of high-dividend status even after controlling for firm fundamentals.

Disaggregated governance mechanisms display economically meaningful and heterogeneous effects. Stronger shareholder rights (PSR) exert the largest influence: in Model 3, a one-unit increase in PSR raises the log-odds of high dividends by $0.031$ ($z=11.67$, $p<0.01$), increasing to $0.036$ ($z=7.700$, $p<0.01$) when all governance dimensions are jointly included. Board effectiveness (BOD) also significantly enhances dividend commitment ($\beta =0.017$, $z=4.144$ in Model 2; $\beta =0.014$, $z=2.162$ in Model 6), while audit quality has a smaller but positive effect ($\beta =0.007$, $z=3.342$), consistent with its monitoring role.

By contrast, disclosure quality (DISC) is statistically insignificant across models (z-statistics below $1.5$), indicating that transparency alone does not translate into stronger dividend commitment in the absence of enforcement or control rights. Control variables exhibit strong economic effects. Leverage substantially reduces the probability of high dividends ($\beta \approx -2.26$ to $-2.64$, $\left|z\right|>12$), consistent with debt-induced payout constraints. Free cash flow has a large positive effect ($\beta \approx 2.60$–$2.84$, $z\approx 4.1$–$4.5$), while asset intensity negatively affects dividend likelihood, reflecting reinvestment needs.

Overall, the magnitude and significance of the coefficients confirm that governance influences not only payout levels but also the probability of sustaining dividends, supporting the credibility and agency-based interpretations advanced earlier.

5.2. Dynamic Endogeneity and Payout Persistence

Table 10 reports results from a two-step System GMM estimation that explicitly addresses dividend persistence and potential endogeneity. The coefficient on the second lag of dividend yield is $0.409$ ($t=8.924$, $p<0.01$), indicating strong payout inertia: approximately 41% of prior dividend behavior persists over time.

Table 10. Dynamic panel GMM estimation.

Variable	Beta	t-Stat	Std. Error	Prob.
$DY{D}_{(-2)}$	0.4091 ***	(8.924)	0.0458	0.0000
$TCG{S}_{(-1)}$	0.0002 ***	(2.832)	0.0001	0.0047
$OW{N}_{(-1)}$	0.0002 *	(1.770)	0.0001	0.0768
$DEB{T}_{(-1)}$	−0.0213 *	(−1.662)	0.0128	0.0965
$FC{F}_{(-1)}$	0.0069 ***	(2.515)	0.0027	0.0119
$ASSIN{T}_{(-1)}$	−0.0052 ***	(−2.520)	0.0021	0.0118
$RO{E}_{(-1)}$	0.0088 ***	(5.887)	0.0015	0.0000
Year Fixed Effect	Yes
Firm Fixed Effect	Yes
S.E. of Regression	0.0101
J-statistic	47.938
Prob(J-statistic)	0.7801
AR(1) test	−10.3491 ***		$\rho =-0.1629$, SE = 0.0157	0.0000
AR(2) test	1.0474		$\rho =0.0074$, SE = 0.0071	0.2949

Note: This table reports results from a two-step System GMM estimation addressing the dynamic nature of dividend policy. The framework controls for unobserved firm-level effects and potential endogeneity of regressors. Dividend yield (DYD) is the dependent variable, with lagged governance and control covariates used as internal instruments. t-statistics are reported in parentheses. ***, and * denote statistical significance at the 1%, and 10% levels, respectively. Model validity is assessed using Arellano–Bond tests for serial correlation [AR(1) and AR(2)] and the Hansen J-test for over-identification restrictions.

Importantly, governance quality retains a positive and statistically significant dynamic effect. The lagged governance score enters with a coefficient of $0.0002$ ($t=2.832$, $p<0.01$), implying that improvements in governance translate into higher future dividend payouts rather than merely contemporaneous correlations. Although the magnitude is modest, its persistence and significance in a dynamic framework indicate a structurally meaningful effect.

Ownership concentration displays a weaker but positive influence ($\beta =0.0002$, $t=1.77$, $p<0.10$), while leverage continues to exert a negative effect on dividend payouts ($\beta =-0.021$, $t=-1.66$, $p<0.10$). Free cash flow and profitability remain positive and significant, whereas asset intensity reduces dividend payouts, reinforcing the investment–payout tradeoff.

Diagnostic tests confirm model validity. The Hansen J-statistic ($p=0.780$) supports instrument exogeneity, while the Arellano–Bond tests indicate first-order but no second-order serial correlation, satisfying the necessary conditions for consistent GMM estimation.

5.3. Summary of Robustness Analyses

Robustness tests confirm that our core findings on governance and dividend credibility are not artifacts of estimation choice or sample specification. Binary logistic regressions (Table 9) show that higher governance quality significantly increases the probability of sustaining high dividends, with shareholder rights and board effectiveness exerting the strongest effects. Dynamic System GMM estimates (Table 10) demonstrate that governance improvements persistently raise future dividend payouts, even after accounting for payout inertia, endogeneity, and firm fixed effects. Control variables, including leverage, free cash flow, and profitability, retain economically meaningful signs and magnitudes across specifications.

Subsample analyses by Chaebol and non-Chaebol groups were also conducted as part of these robustness checks. While the results are consistent with the main findings, confirming H1 and H2, they are not reported here for brevity but are fully acknowledged. Collectively, these results reinforce that governance enhances dividend credibility and that its effectiveness is conditional on ownership regimes, with effects that are both statistically robust and economically interpretable.

6. Conclusions

This study investigated how corporate governance quality interacts with ownership structure to shape dividend policy in Korean listed firms. It provides a more granular understanding of payout credibility under varying institutional constraints. We show that governance mechanisms, particularly shareholder rights, board effectiveness, audit oversight, and disclosure, translate into credible dividend commitments when ownership is dispersed. In Chaebol-affiliated firms, governance influence becomes mechanism-specific: shareholder rights continue to matter, while internal monitoring channels are largely subsumed by concentrated control. This conditionality clarifies how formal governance complements or substitutes for internalized oversight in complex conglomerate structures.

Importantly, this study quantifies the economic impact of governance improvements, demonstrating that improved governance translates into economically meaningful dividend yield effects, reinforcing payout stability and providing interpretable signals to investors and analysts. By combining fixed effect estimation, binary logistic regressions, and dynamic System GMM, we mitigate endogeneity concerns and confirm that these effects persist across multiple empirical approaches and ownership subsamples, ensuring the robustness of our findings. Subsample analyses further confirm that non-Chaebol firms benefit from broad governance effectiveness, whereas Chaebol affiliates exhibit selective mechanism-specific transmission.

These results carry clear implications. For investors, dividend behavior in dispersed ownership firms reliably signals governance quality, whereas in conglomerate contexts, interpreting dividends requires attention to selective mechanism effects. For policymakers and regulators, strengthening disclosure, board independence, audit credibility, and minority shareholder protections remains critical for boosting payout credibility and market integrity.

Limitations remain. This analysis focuses on listed firms and formal governance metrics, which may not fully capture informal monitoring, intra-group capital flows, or complex pyramidal control structures. Future research could incorporate richer measures of ownership depth, explore alternative payout channels such as share repurchases, and extend the framework to comparative or unlisted settings to better understand governance–dividend dynamics in diverse institutional contexts.

Overall, our study shows that dividend credibility emerges from the interaction between governance quality and ownership structure, offering a mechanism-specific, economically interpretable account of payout behavior. These findings advance the literature on corporate governance and provide practical guidance for boosting market confidence, investor protection, and sustainable payout practices in economies characterized by concentrated corporate control.