This study examines whether board interlocks are related to earnings management among KOSPI-listed firms in Korea from 2019 to 2024. Board interlocks may bring external information, experience, and reputational capital into the boardroom. The same ties, however, can weaken oversight when directors hold multiple appointments or when reporting practices travel through director networks. This study tests which interpretation better explains firms' financial reporting behavior. Earnings management is measured primarily by discretionary accruals from the Modified Jones model. Additional tests use the Kothari model, signed discretionary accruals and real earnings management. Board interlocks are measured at the total-board, inside-director, and outside-director levels, with further tests using alternative interlock definitions. The results show a positive relation between board interlocks and accrual-based earnings management. This relation remains evident across alternative interlock definitions and the Kothari-based measure. Signed accrual tests show that interlocks are related to larger discretionary accruals in both income-increasing and income-decreasing directions, rather than to one reporting direction alone. By contrast, the evidence for real earnings management is weaker, suggesting that the main results are concentrated in accrual-based reporting discretion. Director-type analyses indicate that inside- and outside-director interlocks capture different aspects of board networks, although neither type dominates across all specifications. Taken together, the evidence suggests that board interlocks in Korea are not merely advisory channels. They are also tied to weaker monitoring and the diffusion of reporting practices.
1. Introduction
Earnings management matters because it changes how outside investors read reported performance. When managers use discretionary accounting choices to alter earnings, financial statements become less useful for assessing firm value and governance quality. Prior studies have examined board independence, audit committee characteristics, and ownership structure as governance mechanisms related to earnings management. More recent work has shifted attention to board networks. This study examines whether board interlocks, one of the most visible forms of board networks, are related to earnings management.
A board interlock arises when one or more directors serve on the boards of two or more firms, creating ties across firms. These ties can shape information exchange, resource access, strategic decisions, and the diffusion of governance practices (Mizruchi, 1996; Gulati and Westphal, 1999). Resource dependence theory treats interlocks as channels for advice, legitimacy, and external knowledge. Directors with multiple board appointments may bring expertise and reputational capital that strengthen monitoring and advisory functions (Haunschild, 1993; Haunschild and Beckman, 1998; Gulati and Westphal, 1999; Burt, 2000; Carpenter and Westphal, 2001; Dass et al., 2014; Field et al., 2013). Ferris et al. (2003) similarly argue that multiple directorships do not necessarily impair monitoring because reputational concerns can discipline director behavior.
The busyness hypothesis gives the opposite prediction. Directors who serve on multiple boards have limited time and attention, and their monitoring effectiveness may decline as outside commitments increase (Core et al., 1999; Fich and Shivdasani, 2006; Jiraporn et al., 2008; Cashman et al., 2012; Ferris and Liao, 2019). A network perspective adds another concern. Interlocked directors can transmit not only useful knowledge but also reporting practices and managerial norms across firms (Chiu et al., 2013; Shi et al., 2013; Ma et al., 2024). Chiu et al. (2013) show that earnings management can spread through shared directors. Reporting discretion, therefore, may be learned and normalized through board networks, not only driven by firm-specific incentives.
Korea offers a setting in which these competing mechanisms can be observed together. Korean listed firms prepare financial statements under K-IFRS and are subject to external audit and public disclosure requirements. These rules support transparency, but they do not remove managerial discretion in accrual estimates and reporting choices. Ownership is also concentrated in many Korean firms, and controlling shareholders often influence corporate decisions. At the same time, directors frequently hold appointments across firms, creating board networks within and across business groups. In this environment, interlocks can serve as information channels, but they can also increase director busyness and transmit reporting practices.
The existing literature leaves two issues unresolved. First, evidence on the direct link between board interlocks and earnings management remains limited. Prior studies do not clearly show whether monitoring benefits outweigh busyness and diffusion costs. Second, many studies rely on aggregate interlock measures and do not distinguish director roles. This distinction matters. Inside directors hold firm-specific information and participate in managerial decisions, so their network ties may transmit strategic reporting practices. Outside directors are expected to monitor management, but their external appointments may reduce monitoring capacity or introduce practices from other firms. An aggregate interlock measure can hide these channels.
This study addresses these issues using Korean listed firms. Earnings management is measured primarily by discretionary accruals from the Modified Jones model. Additional analyses use the performance-matched Kothari model, signed discretionary accruals, and real earnings management. Board interlocks are measured at three levels: total board interlocks, inside-director interlocks, and outside-director interlocks. The analysis also uses alternative interlock definitions that incorporate non-registered directors and interlocks with KOSDAQ-listed firms.
The study contributes to the literature in three ways. First, it links board interlocks to earnings management, a direct financial reporting outcome. Second, it separates inside-director and outside-director interlocks, allowing the analysis to distinguish managerial information channels from external monitoring channels. Third, it provides evidence from Korea, where concentrated ownership and dense director networks make resource dependence, director busyness, and diffusion arguments all relevant.
The results show that board interlocks are positively related to accrual-based earnings management. The relation appears across alternative interlock definitions and remains generally robust when discretionary accruals are measured using the Kothari model. Signed discretionary accrual analyses show larger discretionary accruals in both income-increasing and income-decreasing directions, rather than a shift toward one direction. Additional tests using real earnings management and propensity score matching qualify the interpretation but do not overturn the main evidence. Overall, the findings indicate that board interlocks are linked to firms' reporting discretion and fit better with weakened monitoring and network diffusion explanations than with a pure monitoring-enhancement view.
2. Literature review and hypothesis
2.1 Theoretical background: resource dependence, busyness, and network diffusion
Board interlocks create interorganizational ties when directors serve on multiple boards (Mizruchi, 1996). Prior research usually explains their effects through three channels: resource access, director busyness, and network-based diffusion. These channels lead to different predictions about financial reporting quality.
Resource dependence theory views interlocks as channels for information, advice, legitimacy, and strategic resources. Directors with multiple appointments may bring industry expertise, network capital, and reputational incentives to the board. Carpenter and Westphal (2001) show that strategically relevant external experience increases director involvement and monitoring, while Dass et al. (2014) find that interlocks with related industries are associated with higher firm value and profitability. Other studies also suggest that multiple directorships can reflect director quality rather than overcommitment (Ferris et al., 2003; Lei and Deng, 2014). These benefits are not automatic. They depend on firm-specific needs, including life-cycle stage and demand for advisory functions (Field et al., 2013).
The busyness hypothesis makes the opposite claim. Directors with many outside appointments have less time to review firm-specific information and monitor management. Empirical evidence links busy boards to lower firm value, weaker CEO turnover-performance sensitivity, and lower profitability (Fich and Shivdasani, 2006; Jiraporn et al., 2008). Cashman et al. (2012) further show that the relation between multiple directorships and firm value often becomes negative after addressing selection and specification issues. Multiple appointments, therefore, may signal monitoring overload rather than director quality.
Network-based diffusion provides a third explanation. Interlocks connect firms not only to resources but also to practices, norms, and behavioral templates. Interlocked directors can transmit both beneficial and harmful practices across firms (Shropshire, 2010). Ma et al. (2024) show that interlocks may improve governance and reduce uncertainty, but may also promote reduced disclosure, earnings management, and tax avoidance. In the financial reporting domain, interlocks influence audit and reporting environments, suggesting that accounting-related practices can converge through director networks (Park, 2025).
These perspectives imply that board interlocks have no single governance effect. They can improve advice and monitoring, but they can also weaken oversight or transmit opportunistic reporting practices. This dual view is consistent with Korean evidence that relationship-based social networking can facilitate information flow and trust, while also increasing opacity and agency problems (Lee, 2021). Whether interlocks constrain or facilitate earnings management therefore depends on which mechanism dominates in a given institutional setting.
2.2 Board interlock and earnings management
Earnings management is not driven only by managerial incentives. Board-level governance also matters. Recent studies show that board remuneration, remuneration committees, board diversity, and nomination committee quality are related to accrual earnings management (Putra et al., 2024; Putra and Setiawan, 2025). These findings suggest that reporting discretion depends on how boards are structured, incentivized, and monitored. This study extends that line of work by examining board interlocks as a network-based board characteristic.
Evidence on board interlocks and earnings management remains mixed. From a resource dependence perspective, interlocked directors can reduce earnings management by bringing external knowledge, monitoring experience, and reputational incentives to the board. Tham et al. (2019), for example, find that Australian firms with more extensive board interlocks exhibit lower earnings management.
Other studies point in the opposite direction. Ferris and Liao (2019) report that firms with a higher proportion of busy directors engage in more earnings management and have lower reporting quality. Sarkar et al. (2008) also find that boards with more interlocked directors are associated with higher discretionary accruals in emerging markets, while more diligent boards mitigate earnings management. These findings support the concern that multiple board appointments can reduce directors' ability to challenge managerial reporting choices.
A separate stream emphasizes diffusion through director networks. Chiu et al. (2013) show that firms sharing directors with earnings-managing firms are more likely to engage in similar practices, whereas ties to non-manipulating firms reduce that likelihood. Shi et al. (2013) further show that board and audit committee interlocks increase similarity in discretionary accruals and accounting quality across firms, especially after interlocks are formed. Their evidence suggests that unfavorable accounting practices may travel more easily through director networks than favorable ones.
Prior studies therefore do not point to a single prediction. The monitoring-enhancement view predicts a negative relation between board interlocks and earnings management. The busyness and diffusion views predict a positive relation. The direction of the effect should depend on institutional context and on the role played by the interlocked director.
2.3 Hypothesis
Board interlocks affect financial reporting through competing channels. Resource dependence theory predicts that interlocked directors improve monitoring and advice by bringing external knowledge, experience, and reputational capital to the board. The busyness hypothesis predicts the opposite: directors with multiple appointments have less time and attention for firm-specific reporting issues. Network diffusion arguments add another channel. Interlocked directors carry accounting practices, reporting norms, and discretionary reporting strategies across firms. Which effect dominates depends on the institutional setting.
This study focuses on KOSPI-listed firms. Many Korean listed firms have concentrated ownership, and controlling shareholders often influence corporate decisions. Although outside directors are formally institutionalized, their substantive independence remains a concern. In this setting, board interlocks are less likely to operate only as independent monitoring devices. They can also become channels through which reporting practices and managerial norms circulate.
Diffusion is likely to occur through shared decision-makers. Directors who sit on multiple boards observe accounting choices used elsewhere and bring those experiences into later board discussions. When aggressive but acceptable reporting choices are observed repeatedly, they become more familiar and easier to justify. Repeated participation in financial reporting and governance decisions can also normalize similar discretionary practices within director networks. Over time, connected firms may converge in reporting behavior.
Weakened monitoring reinforces this process. Directors with multiple appointments face time and attention constraints. These constraints limit their ability to review complex accounting estimates, question managerial assumptions, and challenge discretionary reporting choices. In a governance environment marked by ownership concentration and insider influence, interlocks may reinforce elite networks rather than strengthen board independence. The monitoring benefits predicted by resource dependence theory are therefore likely to be limited, while busyness and diffusion mechanisms become more salient.
Accordingly, this study predicts a positive relation between board interlocks and earnings management.
Board interlocks are positively associated with earnings management.
The effect may also differ by director type. Inside directors participate directly in managerial decisions and possess firm-specific information. Their interlocks may serve as direct channels for transferring discretionary reporting practices. Outside directors are expected to monitor management, but multiple appointments may reduce their monitoring capacity. Their network ties may also introduce reporting practices from other firms. Thus, both inside-director and outside-director interlocks are expected to be positively related to earnings management.
Inside-director interlocks are positively associated with earnings management.
Outside-director interlocks are positively associated with earnings management.
3. Study design and sample selection
3.1 Study model
To examine the effect of board interlocks on earnings management, we estimate the following regression model:
Earnings Management: EM represents earnings management, measured as the absolute value of discretionary accruals estimated using the Modified Jones model and the performance-matched Kothari model.
AvgID_Total: The average number of interlocking directorates per director in firm i in year t (total number of interlocking directorates divided by total number of directors)
AvgID_Inside: The average number of interlocking directorates per inside director.
AvgID_Outside: The average number of interlocking directorates per outside director.
SIZE: The natural logarithm of the total assets
LEV: Total liabilities divided by total assets
CFOt-1: Operating cash flow scaled by total assets at t-1
MB: Market-to-book ratio
ROAt-1: Return on assets in year t-1
PPEt-1: (Property, plant, and equipment – land – construction in progress) scaled by total assets at t-1
INTANt-1: Intangible assets scaled by total assets at t-1
OWN: Ownership percentage of the largest shareholder
FOR: Ownership percentage of foreign shareholders
AGE: Natural logarithm of listing age
LOSS: Indicator variable equal to 1 if net income is negative, and 0 otherwise
Big4: Indicator variable equal to 1 if the firm is audited by a Big 4 accounting firm, and 0 otherwise
Year: Year fixed effects
Industry: Industry fixed effects
Earnings management is measured primarily by the absolute value of discretionary accruals estimated from the Modified Jones model of Dechow et al. (1995). The performance-matched Kothari model of Kothari et al. (2005) is used as an alternative accrual-based measure. Prior Korean evidence also suggests that accrual-based measures may be sensitive to model specification and earnings-recognition properties (Paek, 2020). The two accrual models are specified in Equations (2) and (3):
The residuals from Equations (2) and (3) are used as discretionary accruals. The main analyses use their absolute values because the theoretical focus is on the magnitude of discretionary reporting rather than on a specific income-increasing or income-decreasing direction. Because this aggregation may hide directional differences, additional tests examine signed, positive, and negative discretionary accruals. Real earnings management is also examined to assess whether the results extend beyond accrual-based reporting discretion. This distinction is relevant because accrual-based and real earnings management reflect different managerial choices and may have different implications for firm value (Sohn et al., 2011).
This study uses a continuous measure of interlock intensity rather than a binary indicator of interlock presence. Scaling interlocks at the director level reduces the mechanical influence of board size and improves comparability across firms. It also fits the diffusion and busyness mechanisms examined in this study because AvgID captures directors' average exposure to other firms' board practices and external board commitments.
For the director-type decomposition, inside and outside directors are classified based on their roles in each focal firm-year. The two groups are mutually exclusive within the same firm-year, so a director does not contribute to both (AvgID_Inside) and (AvgID_Outside) for the same observation. If the same individual serves in different roles across firms, the classification follows the director's role in each focal firm-year.
The study also distinguishes registered and non-registered directors for the expanded interlock definition. Registered directors are formally registered board members disclosed in corporate filings. Non-registered directors are disclosed executives who participate in managerial decision-making but do not hold a formally registered board seat. The baseline definition uses registered directors, while the expanded definition includes both groups to capture broader managerial and governance-related network ties.
Control variables are selected to account for firm characteristics associated with accrual behavior and reporting incentives. Firm size (SIZE) captures organizational complexity and political cost considerations. Leverage (LEV) reflects incentives arising from debt contracting. Operating cash flow (CFOt-1) and lagged return on assets (ROAt-1) control for underlying performance, allowing us to separate discretionary reporting from performance-driven accruals. The market-to-book ratio (MB) proxies for growth opportunities and market expectations. Property, plant, and equipment (PPEt-1) and intangible assets (INTANt-1) control for structural differences in accrual-generating processes. Governance-related controls include the largest shareholder's ownership (OWN) and foreign ownership (FOR), both of which proxy for monitoring incentives and governance-related ownership effects (Choi and Seo, 2008; Bang et al., 2021; Lee et al., 2022; Noh and Park, 2024). Similar Korean listed-firm studies also control for foreign ownership when examining firm value and governance-related outcomes (Kwon and Yoon, 2026). Listing age (AGE) captures organizational maturity and reporting stability, while a loss indicator (LOSS) controls for reporting pressures specific to loss firms. Audit quality is controlled using an indicator for Big4 auditors. Year and industry fixed effects are included to account for macroeconomic shocks and industry-specific accounting environments. All continuous variables are winsorized at the top and bottom 1% to mitigate the influence of outliers.
3.2 Sample selection
This study examines KOSPI-listed firms from 2019 to 2024. Korea is a relevant setting for this analysis because many listed firms have concentrated ownership, business group affiliations, and active director networks. Yet evidence on how board interlocks relate to financial reporting behavior in Korea remains limited.
The sample period covers a phase in which reporting and audit regulation became more stringent. Since 2019, reforms related to internal control over financial reporting, external audit quality, and governance disclosure have increased scrutiny of firms' reporting processes. This setting allows us to examine whether board interlocks are linked to discretionary reporting behavior under a stronger monitoring environment.
KOSPI firms also provide detailed financial and director-level disclosures, which improves the reliability of discretionary accrual estimation and interlock measurement. The sample includes firms that meet the following criteria: (1) listed on the KOSPI between 2019 and 2024; (2) non-financial firms with December fiscal year-ends; and (3) firms with available financial, ownership, audit, and director-level data from TS-2000 and related disclosure sources. The final sample consists of 3,195 firm-year observations. Appendix Table A1 reports the industry distribution.
4. Results of the empirical analysis
4.1 Descriptive statistics and correlation analysis
Table 1 presents descriptive statistics for the main variables. The mean and median values of earnings management (EM) are 0.04 and 0.03, respectively, with a standard deviation of 0.05. Most firms report modest discretionary accruals, but the 99th percentile reaches 0.29. This upper-tail value indicates substantial reporting discretion among a small subset of firms.
Descriptive statistics
| Stats | Mean | sd | p1 | p25 | p50 | p75 | p99 |
|---|---|---|---|---|---|---|---|
| EM | 0.04 | 0.05 | 0.00 | 0.01 | 0.03 | 0.06 | 0.29 |
| AvgID_Total | 0.21 | 0.24 | 0.00 | 0.00 | 0.14 | 0.33 | 1.13 |
| AvgID_Inside | 0.26 | 0.38 | 0.00 | 0.00 | 0.00 | 0.50 | 1.75 |
| AvgID_Outside | 0.14 | 0.23 | 0.00 | 0.00 | 0.00 | 0.25 | 1.00 |
| SIZE | 20.86 | 1.57 | 17.82 | 19.77 | 20.65 | 21.83 | 25.27 |
| LEV | 0.47 | 0.20 | 0.06 | 0.31 | 0.48 | 0.62 | 0.90 |
| CFOt-1 | 0.05 | 0.07 | −0.14 | 0.02 | 0.05 | 0.09 | 0.24 |
| MB | 1.31 | 1.81 | 0.05 | 0.40 | 0.68 | 1.47 | 11.99 |
| ROAt-1 | 0.02 | 0.07 | −0.26 | 0.00 | 0.03 | 0.06 | 0.24 |
| PPEt-1 | 0.34 | 0.19 | 0.01 | 0.19 | 0.33 | 0.46 | 0.88 |
| INTANt-1 | 0.03 | 0.03 | 0.00 | 0.00 | 0.01 | 0.03 | 0.19 |
| OWN | 0.44 | 0.16 | 0.08 | 0.33 | 0.45 | 0.55 | 0.82 |
| FOR | 0.10 | 0.12 | 0.00 | 0.02 | 0.06 | 0.14 | 0.55 |
| AGE | 3.08 | 0.78 | 0.69 | 2.64 | 3.30 | 3.64 | 4.04 |
| LOSS | 0.25 | 0.43 | 0.00 | 0.00 | 0.00 | 0.00 | 1.00 |
| BIG4 | 0.67 | 0.47 | 0.00 | 0.00 | 1.00 | 1.00 | 1.00 |
The mean value of AvgID_Total is 0.21, with a median of 0.14 and a 99th percentile of 1.13. Board interlocks are therefore present among KOSPI-listed firms, but they are unevenly distributed. The director-type measures show a similar pattern. Inside-director interlocks are more prevalent than outside-director interlocks on average, while the medians of both AvgID_Inside and AvgID_Outside are zero. Many firms have no interlocks, and interlock intensity is concentrated among a smaller group.
The prevalence of interlocks increases when broader definitions are applied. Of the 3,195 firm-year observations, 1,903 have registered director–registered director interlocks. This number rises to 2,022 when registered–non-registered director connections are included and to 2,312 when KOSPI–KOSDAQ interlocks are considered. These figures show that many firms are embedded in director networks, even though the intensity of those networks varies widely.
The control variables are broadly consistent with the characteristics of KOSPI-listed firms. The sample consists of large firms with moderate leverage and heterogeneous growth opportunities. Ownership is concentrated, with the largest shareholder holding 44% of shares on average. Foreign ownership averages 10%, 25% of observations report losses, and 67% of firms are audited by Big 4 auditors. Overall, Table 1 shows meaningful variation in both earnings management and board interlock intensity, providing a basis for the multivariate tests.
Table 2 reports Pearson correlations among the main variables. Pairwise correlations are moderate. Most coefficients are below 0.6, and the correlations between AvgID_Total and the control variables are mostly below 0.3. EM is positively correlated with AvgID_Total (r = 0.02). The magnitude is small, but the direction is consistent with H1, which calls for multivariate tests with firm characteristics, governance variables, and fixed effects.
Correlation analysis
| EM | AvgID_Total | SIZE | LEV | CFOt-1 | MB | ROAt-1 | PPEt-1 | INTANt-1 | OWN | FOR | AGE | LOSS | BIG4 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| EM | 1 | |||||||||||||
| AvgID_Total | 0.02 | 1 | ||||||||||||
| SIZE | −0.069*** | 0.263*** | 1 | |||||||||||
| LEV | 0.115*** | −0.027 | 0.178*** | 1 | ||||||||||
| CFOt-1 | −0.068*** | 0.071*** | 0.188*** | −0.150*** | 1 | |||||||||
| MB | 0.156*** | −0.103*** | −0.120*** | 0.087*** | 0.012 | 1 | ||||||||
| ROAt-1 | −0.167*** | 0.095*** | 0.223*** | −0.315*** | 0.405*** | −0.107*** | 1 | |||||||
| PPEt-1 | −0.152*** | 0.039** | 0.063*** | 0.201*** | 0.130*** | −0.037** | 0.018 | 1 | ||||||
| INTANt-1 | 0.030* | 0.071*** | 0.264*** | 0.022 | 0.120*** | 0.077*** | 0.033* | −0.125*** | 1 | |||||
| OWN | −0.121*** | 0.184*** | 0.008 | −0.061*** | 0.066*** | −0.107*** | 0.128*** | 0.070*** | −0.048*** | 1 | ||||
| FOR | −0.053*** | 0.090*** | 0.525*** | −0.113*** | 0.236*** | 0.012 | 0.208*** | −0.038** | 0.168*** | −0.247*** | 1 | |||
| AGE | −0.095*** | 0.016 | −0.021 | −0.009 | −0.102*** | −0.190*** | −0.073*** | 0.026 | −0.137*** | −0.121*** | −0.026 | 1 | ||
| LOSS | 0.140*** | −0.128*** | −0.223*** | 0.236*** | −0.393*** | 0.095*** | −0.449*** | 0.009 | −0.047*** | −0.115*** | −0.204*** | 0.014 | 1 | |
| BIG4 | 0.031* | 0.132*** | 0.469*** | 0.052*** | 0.117*** | 0.032* | 0.122*** | −0.069*** | 0.218*** | 0.008 | 0.272*** | −0.138*** | −0.123*** | 1 |
Note(s): *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively
The correlations between EM and the control variables are also in line with prior earnings management research. EM is negatively correlated with SIZE, CFOt-1, and ROAt-1, and positively correlated with LEV, MB, and LOSS. OWN and FOR are negatively correlated with EM, suggesting that ownership concentration and foreign ownership may constrain discretionary reporting.
The correlation between SIZE and FOR is relatively high at 0.525. We therefore calculate VIF statistics to assess multicollinearity. As reported in the note to Table 3, the mean VIF is 1.92, and the VIFs for AvgID_Total, SIZE, and FOR are 1.13, 2.15, and 1.71, respectively. These values are well below conventional concern thresholds and indicate that multicollinearity is unlikely to affect the estimates.
OLS and fixed-effects results – baseline interlock definition
| Variables | OLS | Panel – Fixed effect | ||||||
|---|---|---|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | |
| AvgID_Total | 0.018*** | 0.016** | ||||||
| (5.07) | (1.98) | |||||||
| AvgID_Inside | 0.011*** | 0.011*** | 0.006 | 0.006 | ||||
| (4.75) | (4.86) | (1.07) | (1.00) | |||||
| AvgID_Outside | 0.011*** | 0.011*** | 0.011** | 0.011** | ||||
| (2.85) | (2.92) | (1.97) | (1.99) | |||||
| SIZE | −0.002** | −0.002** | −0.002** | −0.002** | 0.011 | 0.011 | 0.011 | 0.011 |
| (−2.39) | (−2.11) | (−2.07) | (−2.40) | (1.41) | (1.44) | (1.42) | (1.44) | |
| LEV | 0.024*** | 0.024*** | 0.023*** | 0.024*** | 0.063*** | 0.063*** | 0.062*** | 0.063*** |
| (4.98) | (4.94) | (4.69) | (4.86) | (3.11) | (3.11) | (3.09) | (2.95) | |
| CFOt-1 | 0.032** | 0.033** | 0.031** | 0.032** | 0.018 | 0.018 | 0.017 | 0.018 |
| (2.19) | (2.23) | (2.14) | (2.17) | (0.60) | (0.60) | (0.57) | (0.60) | |
| MB | 0.003*** | 0.003*** | 0.003*** | 0.004*** | 0.001 | 0.001 | 0.001 | 0.003 |
| (5.79) | (5.82) | (5.57) | (4.20) | (1.35) | (1.35) | (1.32) | (1.49) | |
| ROAt-1 | −0.059*** | −0.058*** | −0.060*** | −0.064*** | 0.015 | 0.016 | 0.013 | 0.013 |
| (−4.21) | (−4.18) | (−4.33) | (−4.59) | (0.64) | (0.68) | (0.56) | (0.54) | |
| PPEt-1 | −0.042*** | −0.042*** | −0.041*** | −0.043*** | −0.006 | −0.006 | −0.006 | −0.007 |
| (−9.18) | (−9.22) | (−8.95) | (−9.47) | (−0.42) | (−0.40) | (−0.41) | (−0.43) | |
| INTANt-1 | −0.009 | −0.008 | −0.007 | −0.006 | 0.216*** | 0.215*** | 0.209*** | 0.216*** |
| (−0.38) | (−0.33) | (−0.30) | (−0.24) | (2.72) | (2.70) | (2.63) | (2.73) | |
| OWN | −0.033*** | −0.033*** | −0.029*** | −0.034*** | −0.038 | −0.038 | −0.036 | −0.035 |
| (−5.93) | (−5.91) | (−5.20) | (−5.92) | (−1.22) | (−1.23) | (−1.13) | (−1.11) | |
| FOR | −0.016* | −0.016* | −0.016* | −0.018** | −0.006 | −0.007 | −0.006 | −0.008 |
| (−1.76) | (−1.82) | (−1.76) | (−1.98) | (−0.21) | (−0.28) | (−0.23) | (−0.31) | |
| AGE | −0.005*** | −0.005*** | −0.005*** | −0.005*** | −0.001 | −0.001 | −0.001 | −0.000 |
| (−4.80) | (−4.78) | (−4.62) | (−4.72) | (−0.16) | (−0.09) | (−0.12) | (−0.04) | |
| LOSS | 0.008*** | 0.008*** | 0.008*** | 0.008*** | 0.002 | 0.002 | 0.002 | 0.002 |
| (3.63) | (3.59) | (3.46) | (3.73) | (0.85) | (0.83) | (0.86) | (0.94) | |
| BIG4 | 0.004* | 0.004** | 0.004** | 0.004* | – | – | – | – |
| (1.96) | (1.99) | (2.04) | (1.96) | – | – | – | – | |
| N | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 |
| Fixed Effects | ||||||||
| Year | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Industry | Yes | Yes | Yes | Yes | No | No | No | No |
| Firm | No | No | No | No | Yes | Yes | Yes | Yes |
| Adj R2 | 0.11 | 0.11 | 0.10 | 0.10 | – | – | – | – |
| within R2 | – | – | – | – | 0.03 | 0.03 | 0.03 | 0.03 |
| F-Statistics | 16.68*** | 16.53*** | 15.83*** | 15.48*** | 2.69*** | 2.33*** | 2.52*** | 2.76*** |
Note(s): Standard errors are clustered at the firm level
Additional VIF diagnostics indicate that multicollinearity is unlikely to affect the results. The mean VIF is 1.92, and the VIFs for AvgID_Total, SIZE, and FOR are 1.13, 2.15, and 1.71, respectively. The highest VIF among substantive explanatory variables is 2.15 for SIZE
*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively
4.2 Main analysis
Table 3 reports OLS and firm fixed-effects (FE) estimates of the relation between board interlocks and earnings management (EM). OLS estimates capture cross-sectional differences across firms, while FE estimates examine whether within-firm changes in interlock intensity are related to changes in EM. Hausman tests reject the null that the random-effects estimator is consistent. For the baseline specification, χ2 = 117.95 with p < 0.001, and untabulated tests for the other main specifications yield similar conclusions. This result supports reporting FE estimates to control for time-invariant unobserved firm characteristics.
In the OLS specifications, AvgID_Total is positive and significant. AvgID_Inside and AvgID_Outside also show positive and significant coefficients when director types are examined separately. Firms whose directors hold more external board appointments therefore report higher accrual-based earnings management in cross-sectional comparisons. This pattern is consistent with the busyness and network diffusion arguments: multiple appointments can weaken monitoring capacity, while interlocked directors can transmit discretionary reporting practices across firms.
The FE estimates absorb time-invariant firm characteristics. AvgID_Total remains positive and significant, although its magnitude declines. This suggests that the aggregate interlock effect is not driven only by persistent differences across firms. In the director-type specifications, AvgID_Inside becomes insignificant, whereas AvgID_Outside remains positive and significant. The within-firm evidence is therefore stronger for outside-director interlocks. However, Wald tests do not reject equality between the AvgID_Inside and AvgID_Outside coefficients in either the OLS or FE specifications. The results therefore do not establish that one director type has a statistically stronger effect than the other.
The coefficient magnitude is modest but meaningful. In the baseline OLS specification, the coefficient on AvgID_Total is 0.018. Given the standard deviation of AvgID_Total of 0.24, a one-standard-deviation increase in interlock intensity corresponds to an increase in EM of about 0.004. This equals approximately 10.8% of the sample mean of EM and 8.6% of its standard deviation. Because the R2 values range from 0.03 to 0.11, board interlocks should not be viewed as the dominant determinant of earnings management. They are better interpreted as one governance-related channel associated with a limited but nontrivial portion of reporting discretion.
Overall, Table 3 supports H1. Board interlocks are positively related to accrual-based earnings management, and the evidence fits better with monitoring-weakening and diffusion-based explanations than with a pure monitoring-enhancement view. The director-type results remain informative but should be interpreted cautiously because the coefficient differences are not statistically significant.
Table 4 reports OLS and firm fixed-effects (FE) estimates using the expanded interlock definition, which includes ties involving both registered and non-registered directors. This test addresses whether the baseline results depend on the narrower registered-director definition.
OLS and fixed-effects results using expanded interlock definition
| Variables | OLS | Panel – Fixed effect | ||||||
|---|---|---|---|---|---|---|---|---|
| (9) | (10) | (11) | (12) | (13) | (14) | (15) | (16) | |
| AvgID_Total | 0.017*** | 0.017** | ||||||
| (5.45) | (2.23) | |||||||
| AvgID_Inside | 0.010*** | 0.011*** | 0.008 | 0.008 | ||||
| (5.50) | (5.56) | (1.47) | (1.45) | |||||
| AvgID_Outside | 0.010*** | 0.010*** | 0.011** | 0.011** | ||||
| (2.74) | (2.85) | (2.27) | (2.26) | |||||
| SIZE | −0.002** | −0.002** | −0.002** | −0.002*** | 0.011 | 0.011 | 0.011 | 0.011 |
| (−2.41) | (−2.30) | (−2.03) | (−2.87) | (1.41) | (1.46) | (1.40) | (1.40) | |
| LEV | 0.024*** | 0.024*** | 0.023*** | 0.025*** | 0.063*** | 0.063*** | 0.062*** | 0.063*** |
| (5.05) | (5.06) | (4.73) | (5.13) | (3.10) | (3.11) | (3.09) | (3.12) | |
| CFOt-1 | 0.032** | 0.033** | 0.031** | 0.032** | 0.017 | 0.017 | 0.017 | 0.017 |
| (2.19) | (2.24) | (2.14) | (2.23) | (0.58) | (0.59) | (0.57) | (0.58) | |
| MB | 0.003*** | 0.003*** | 0.003*** | 0.003*** | 0.001 | 0.001 | 0.001 | 0.001 |
| (5.82) | (5.87) | (5.55) | (5.87) | (1.35) | (1.37) | (1.30) | (1.31) | |
| ROAt-1 | −0.058*** | −0.057*** | −0.060*** | −0.057*** | 0.016 | 0.017 | 0.013 | 0.015 |
| (−4.17) | (−4.08) | (−4.32) | (−4.08) | (0.69) | (0.72) | (0.56) | (0.64) | |
| PPEt-1 | −0.042*** | −0.042*** | −0.041*** | −0.042*** | −0.006 | −0.006 | −0.006 | −0.007 |
| (−9.20) | (−9.26) | (−8.92) | (−9.19) | (−0.41) | (−0.41) | (−0.38) | (−0.42) | |
| INTANt-1 | −0.011 | −0.011 | −0.007 | −0.011 | 0.217*** | 0.214*** | 0.209*** | 0.214*** |
| (−0.44) | (−0.43) | (−0.27) | (−0.43) | (2.70) | (2.68) | (2.63) | (2.67) | |
| OWN | −0.034*** | −0.034*** | −0.029*** | −0.034*** | −0.039 | −0.040 | −0.034 | −0.037 |
| (−6.01) | (−6.06) | (−5.19) | (−6.09) | (−1.24) | (−1.28) | (−1.09) | (−1.20) | |
| FOR | −0.015* | −0.015* | −0.016* | −0.014 | −0.006 | −0.008 | −0.006 | −0.006 |
| (−1.69) | (−1.73) | (−1.73) | (−1.61) | (−0.23) | (−0.31) | (−0.23) | (−0.22) | |
| AGE | −0.005*** | −0.005*** | −0.005*** | −0.005*** | −0.001 | −0.001 | −0.001 | −0.002 |
| (−4.74) | (−4.69) | (−4.63) | (−4.72) | (−0.17) | (−0.14) | (−0.09) | (−0.19) | |
| LOSS | 0.008*** | 0.008*** | 0.008*** | 0.008*** | 0.002 | 0.002 | 0.002 | 0.002 |
| (3.64) | (3.59) | (3.45) | (3.63) | (0.84) | (0.83) | (0.86) | (0.84) | |
| BIG4 | 0.004* | 0.004* | 0.004** | 0.004* | – | – | – | |
| (1.84) | (1.89) | (1.99) | (1.82) | – | – | – | ||
| N | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 |
| Fixed Effects | ||||||||
| Year | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Industry | Yes | Yes | Yes | Yes | No | No | No | No |
| Firm | No | No | No | No | Yes | Yes | Yes | Yes |
| Adj R2 | 0.10 | 0.10 | 0.10 | 0.10 | – | – | – | – |
| within R2 | – | – | – | – | 0.03 | 0.03 | 0.03 | 0.03 |
| F-Statistics | 16.88*** | 16.90*** | 15.80*** | 16.57*** | 2.69*** | 2.33*** | 2.52*** | 2.52*** |
Note(s): Standard errors are clustered at the firm level
*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively
The OLS results are close to those in Table 3. AvgID_Total remains positive and significant, with a coefficient magnitude similar to the baseline estimate. AvgID_Inside and AvgID_Outside are also positive and significant, both separately and jointly. Thus, including non-registered directors does not change the positive cross-sectional relation between interlock intensity and EM.
The FE results tell a similar story. AvgID_Total remains positive and significant, although the coefficient is smaller than in the OLS specification. AvgID_Inside loses significance, whereas AvgID_Outside remains positive and significant. This result again points to stronger within-firm evidence for outside-director interlocks.
Overall, Table 4 shows that the main findings are not driven by the registered-director restriction. The director-type pattern remains informative, but it should not be read as conclusive evidence that outside-director interlocks are statistically stronger than inside-director interlocks, because the coefficient equality tests do not show a significant difference.
Table 5 expands the interlock network to include both KOSPI- and KOSDAQ-listed firms. This test examines whether the baseline results are sensitive to the KOSPI-only network boundary.
OLS and fixed-effects results using interlocks extended to KOSDAQ firms
| Variables | OLS | Panel – Fixed effect | ||||||
|---|---|---|---|---|---|---|---|---|
| (17) | (18) | (19) | (20) | (21) | (22) | (23) | (24) | |
| AvgID_Total | 0.016*** | 0.018*** | ||||||
| (5.99) | (2.99) | |||||||
| AvgID_Inside | 0.009*** | 0.009*** | 0.008* | 0.008* | ||||
| (5.38) | (5.53) | (1.84) | (1.71) | |||||
| AvgID_Outside | 0.010*** | 0.010*** | 0.012** | 0.011** | ||||
| (3.23) | (3.47) | (2.51) | (2.42) | |||||
| SIZE | −0.002** | −0.002** | −0.001** | −0.002*** | 0.010 | 0.010 | 0.011 | 0.010 |
| (−2.28) | (−2.04) | (−1.97) | (−2.60) | (1.31) | (1.37) | (1.42) | (1.34) | |
| LEV | 0.024*** | 0.024*** | 0.023*** | 0.024*** | 0.063*** | 0.063*** | 0.063*** | 0.064*** |
| (5.04) | (4.96) | (4.68) | (4.99) | (3.16) | (3.14) | (3.14) | (3.19) | |
| CFOt-1 | 0.031** | 0.033** | 0.031** | 0.032** | 0.016 | 0.017 | 0.016 | 0.016 |
| (2.16) | (2.23) | (2.11) | (2.18) | (0.55) | (0.58) | (0.54) | (0.54) | |
| MB | 0.003*** | 0.003*** | 0.003*** | 0.003*** | 0.001 | 0.001 | 0.001 | 0.001 |
| (5.86) | (5.85) | (5.64) | (5.94) | (1.33) | (1.35) | (1.32) | (1.32) | |
| ROAt-1 | −0.059*** | −0.058*** | −0.061*** | −0.059*** | 0.016 | 0.016 | 0.014 | 0.015 |
| (−4.22) | (−4.20) | (−4.35) | (−4.23) | (0.68) | (0.70) | (0.58) | (0.65) | |
| PPEt-1 | −0.041*** | −0.042*** | −0.041*** | −0.041*** | −0.006 | −0.006 | −0.006 | −0.007 |
| (−9.12) | (−9.21) | (−8.93) | (−9.14) | (−0.42) | (−0.41) | (−0.42) | (−0.45) | |
| INTANt-1 | −0.011 | −0.008 | −0.010 | −0.011 | 0.219*** | 0.217*** | 0.209*** | 0.216*** |
| (−0.44) | (−0.33) | (−0.40) | (−0.46) | (2.72) | (2.72) | (2.63) | (2.70) | |
| OWN | −0.031*** | −0.031*** | −0.028*** | −0.031*** | −0.036 | −0.038 | −0.034 | −0.035 |
| (−5.53) | (−5.61) | (−5.06) | (−5.51) | (−1.14) | (−1.20) | (−1.07) | (−1.11) | |
| FOR | −0.013 | −0.014 | −0.015* | −0.013 | −0.004 | −0.007 | −0.005 | −0.004 |
| (−1.47) | (−1.61) | (−1.69) | (−1.43) | (−0.14) | (−0.26) | (−0.20) | (−0.16) | |
| AGE | −0.005*** | −0.005*** | −0.005*** | −0.005*** | −0.002 | −0.001 | −0.002 | −0.003 |
| (−4.64) | (−4.67) | (−4.55) | (−4.62) | (−0.31) | (−0.16) | (−0.24) | (−0.35) | |
| LOSS | 0.008*** | 0.008*** | 0.008*** | 0.008*** | 0.002 | 0.002 | 0.002 | 0.002 |
| (3.54) | (3.51) | (3.48) | (3.59) | (0.79) | (0.78) | (0.90) | (0.84) | |
| BIG4 | 0.004** | 0.004** | 0.004** | 0.004** | – | – | – | |
| (2.10) | (2.17) | (1.99) | (2.10) | – | – | – | ||
| N | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 |
| Fixed Effects | ||||||||
| Year | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Industry | Yes | Yes | Yes | Yes | No | No | No | No |
| Firm | No | No | No | No | Yes | Yes | Yes | Yes |
| Adj R2 | 0.10 | 0.10 | 0.10 | 0.10 | – | – | – | – |
| within R2 | – | – | – | – | 0.03 | 0.03 | 0.03 | 0.03 |
| F-Statistics | 17.18*** | 16.84*** | 15.95*** | 16.70*** | 2.88*** | 2.40*** | 2.64*** | 2.69*** |
Note(s): Standard errors are clustered at the firm level
*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively
The OLS results remain consistent with the main analysis. AvgID_Total is positive and significant, and both AvgID_Inside and AvgID_Outside retain positive and significant coefficients. The positive cross-sectional relation between interlock intensity and EM therefore persists when the network boundary is broadened.
The FE results also support the baseline inference. AvgID_Total remains positive and significant. In the director-type specifications, both AvgID_Inside and AvgID_Outside show positive associations, but the evidence is again more stable for outside-director interlocks. In the joint specification, AvgID_Inside weakens, while AvgID_Outside remains positive and significant.
Overall, Table 5 shows that the main results are not limited to KOSPI-only interlocks. Director networks extending across Korean listed markets are related to accrual-based earnings management. As in Tables 3 and 4, the within-firm evidence is stronger for outside-director interlocks, but the results do not establish that one director type dominates the other.
Table 6 reports OLS and firm fixed-effects results using discretionary accruals estimated from the performance-matched Kothari model. The table presents the main interlock coefficients under the baseline, expanded, and KOSDAQ-extended interlock definitions. Full regression results, including all control variables, are reported in Appendix Tables A2–A4.
OLS and fixed-effects results – Kothari-based earnings management
| Variables | Baseline | Expanded | KOSDAQ | |||
|---|---|---|---|---|---|---|
| OLS | FE | OLS | FE | OLS | FE | |
| (25) | (26) | (27) | (28) | (29) | (30) | |
| AvgID_Total | 0.015*** | 0.017** | 0.014*** | 0.018** | 0.014*** | 0.019*** |
| (4.21) | (2.28) | (4.56) | (2.38) | (5.15) | (3.06) | |
| AvgID_Inside | 0.009*** | 0.006 | 0.009*** | 0.007 | 0.008*** | 0.008* |
| (3.93) | (1.14) | (4.68) | (1.37) | (4.63) | (1.69) | |
| AvgID_Outside | 0.010** | 0.012** | 0.008** | 0.012** | 0.008*** | 0.012** |
| (2.55) | (2.07) | (2.28) | (2.30) | (2.82) | (2.57) | |
| Controls | Include | Include | Include | Include | Include | Include |
| N | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 |
| Adj. R2 | 0.09 | – | 0.09 | – | 0.09 | – |
| within R2 | – | 0.03 | – | 0.03 | – | 0.03 |
| Firm FE | No | Yes | No | Yes | No | Yes |
| Industry | Yes | No | Yes | No | Yes | No |
| Year | Yes | Yes | Yes | Yes | Yes | Yes |
Note(s): Standard errors are clustered at the firm level
For brevity, this table reports only the coefficients on the main interlock variables. Full regression results including all control variables are reported in Appendix Tables A2–A4
*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively
The OLS results are consistent with the main analyses. AvgID_Total remains positive and significant across all interlock definitions. AvgID_Inside and AvgID_Outside also show positive and significant coefficients. The cross-sectional relation between board interlocks and earnings management therefore does not depend on the Modified Jones measure alone.
The firm fixed-effects results show a similar pattern for the aggregate measure. AvgID_Total remains positive and significant across all definitions. The director-type results are less uniform. AvgID_Inside is insignificant under the baseline and expanded definitions and only weakly significant when KOSDAQ-listed firms are included. AvgID_Outside remains positive and significant across specifications, which is consistent with the within-firm patterns reported in Tables 3–5.
Appendix Tables A2–A4 show that the main control variables have signs broadly consistent with the baseline analyses. Overall, Table 6 suggests that the positive relation between board interlocks and accrual-based earnings management is not driven by the choice of accrual model.
Table 7 examines whether board interlocks are related to the direction of earnings management. In an untabulated signed discretionary accrual model, the interlock variables are not statistically significant. This suggests that income-increasing and income-decreasing accrual choices offset each other when combined in a single signed measure. We therefore estimate separate models for positive and negative discretionary accruals.
OLS results using signed earnings management
| Variables | Negative DA | Positive DA | ||||||
|---|---|---|---|---|---|---|---|---|
| (31) | (32) | (33) | (34) | (35) | (36) | (37) | (38) | |
| AvgID_Total | −0.020*** | 0.014*** | ||||||
| (−3.27) | (2.83) | |||||||
| AvgID_Inside | −0.012*** | −0.012*** | 0.009*** | 0.009*** | ||||
| (−3.16) | (−3.26) | (3.11) | (3.09) | |||||
| AvgID_Outside | −0.010* | −0.011* | 0.007 | 0.007 | ||||
| (−1.72) | (−1.89) | (0.91) | (0.88) | |||||
| SIZE | 0.003** | 0.003** | 0.003** | 0.003*** | 0.000 | 0.001 | 0.001 | 0.000 |
| (2.53) | (2.34) | (2.20) | (2.74) | (0.23) | (0.32) | (0.33) | (0.13) | |
| LEV | −0.033*** | −0.033*** | −0.031*** | −0.033*** | −0.008 | −0.009 | −0.009 | −0.008 |
| (−4.20) | (−4.18) | (−4.00) | (−4.20) | (−1.12) | (−1.15) | (−1.22) | (−1.13) | |
| CFOt-1 | −0.531*** | −0.532*** | −0.532*** | −0.531*** | −0.377*** | −0.377*** | −0.377*** | −0.377*** |
| (−19.77) | (−19.78) | (−19.73) | (−19.78) | (−2.93) | (−2.93) | (−2.94) | (−2.93) | |
| MB | −0.003*** | −0.004*** | −0.003*** | −0.004*** | 0.002*** | 0.003*** | 0.002*** | 0.003*** |
| (−4.84) | (−4.85) | (−4.71) | (−4.89) | (2.94) | (2.99) | (2.83) | (2.98) | |
| ROAt-1 | 0.268*** | 0.267*** | 0.274*** | 0.266*** | 0.101*** | 0.101*** | 0.102*** | 0.101*** |
| (11.69) | (11.62) | (11.95) | (11.62) | (3.64) | (3.65) | (3.66) | (3.64) | |
| PPEt-1 | 0.069*** | 0.069*** | 0.068*** | 0.069*** | −0.005 | −0.006 | −0.005 | −0.005 |
| (9.12) | (9.15) | (8.95) | (9.14) | (−0.40) | (−0.42) | (−0.35) | (−0.40) | |
| INTANt-1 | −0.061* | −0.061* | −0.062* | −0.061* | −0.047 | −0.044 | −0.043 | −0.045 |
| (−1.70) | (−1.70) | (−1.72) | (−1.70) | (−1.10) | (−1.05) | (−1.00) | (−1.06) | |
| OWN | 0.049*** | 0.049*** | 0.045*** | 0.050*** | −0.013 | −0.013 | −0.009 | −0.013 |
| (5.53) | (5.54) | (5.12) | (5.56) | (−0.61) | (−0.63) | (−0.42) | (−0.63) | |
| FOR | 0.036** | 0.036*** | 0.035** | 0.035** | −0.009 | −0.010 | −0.010 | −0.009 |
| (2.57) | (2.60) | (2.50) | (2.54) | (−0.33) | (−0.35) | (−0.36) | (−0.32) | |
| AGE | 0.000 | 0.000 | 0.000 | 0.000 | −0.004 | −0.004 | −0.004 | −0.004 |
| (0.08) | (0.09) | (0.03) | (0.09) | (−1.53) | (−1.54) | (−1.45) | (−1.54) | |
| LOSS | −0.042*** | −0.042*** | −0.041*** | −0.042*** | −0.027*** | −0.027*** | −0.027*** | −0.027*** |
| (−11.85) | (−11.82) | (−11.70) | (−11.88) | (−3.22) | (−3.22) | (−3.24) | (−3.23) | |
| BIG4 | 0.000 | 0.000 | 0.001 | 0.000 | 0.002 | 0.002 | 0.003 | 0.002 |
| (0.12) | (0.13) | (0.19) | (0.07) | (0.67) | (0.72) | (0.79) | (0.67) | |
| N | 1,576 | 1,576 | 1,576 | 1,576 | 1,619 | 1,619 | 1,619 | 1,619 |
| Fixed Effects | ||||||||
| Year | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Industry | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Firm | No | No | No | No | No | No | No | No |
| Adj R2 | 0.34 | 0.34 | 0.34 | 0.34 | 0.09 | 0.09 | 0.08 | 0.07 |
| F-Statistics | 36.87*** | 36.83*** | 36.36*** | 35.50*** | 10.71*** | 10.77*** | 10.66*** | 10.53*** |
Note(s): Standard errors are clustered at the firm level
*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively
The separate results show that interlocks are related to discretionary accruals in both directions. For negative discretionary accruals, AvgID_Total, AvgID_Inside, and AvgID_Outside have negative coefficients, indicating larger income-decreasing discretionary accruals. For positive discretionary accruals, AvgID_Total and AvgID_Inside are positive and significant, while AvgID_Outside is positive but insignificant. These findings suggest that board interlocks are linked to greater reporting discretion rather than to one directional reporting incentive.
Wald tests do not reject equality between the positive and negative discretionary accrual coefficients for AvgID_Total (χ2 = 0.63, p = 0.427), AvgID_Inside (χ2 = 0.57, p = 0.450), or AvgID_Outside (χ2 = 0.16, p = 0.690). Thus, the evidence does not show that interlocks favor either income-increasing or income-decreasing earnings management. Instead, the results support the interpretation that interlocks are associated with accrual-based reporting discretion in both directions.
Table 8 reports additional analyses using real earnings management (REM) as the dependent variable. This test addresses the possibility that managers shift from accrual-based earnings management to real activities manipulation. Unlike discretionary accruals, REM captures deviations in actual business activities, such as abnormal cash flows, production costs, and discretionary expenditures.
OLS and fixed-effects results using real earnings management
| Variables | OLS | Panel – Fixed effect | ||||||
|---|---|---|---|---|---|---|---|---|
| (39) | (40) | (41) | (42) | (43) | (44) | (45) | (46) | |
| AvgID_Total | −0.101*** | 0.007 | ||||||
| (−2.67) | (0.56) | |||||||
| AvgID_Inside | −0.069*** | −0.069*** | 0.000 | 0.000 | ||||
| (−2.65) | (−2.67) | (0.06) | (0.05) | |||||
| AvgID_Outside | −0.059* | −0.061* | 0.002 | 0.002 | ||||
| (−1.88) | (−1.93) | (0.32) | (0.32) | |||||
| SIZE | 0.023*** | 0.022*** | 0.022*** | 0.025*** | 0.013 | 0.013 | 0.013 | 0.013 |
| (2.80) | (2.75) | (2.84) | (3.08) | (0.89) | (0.90) | (0.89) | (0.89) | |
| LEV | 0.070 | 0.070 | 0.079 | 0.068 | −0.004 | −0.004 | −0.004 | −0.004 |
| (1.41) | (1.41) | (1.62) | (1.38) | (−0.11) | (−0.12) | (−0.12) | (−0.12) | |
| CFOt-1 | −1.540*** | −1.546*** | −1.542*** | −1.541*** | −1.369*** | −1.369*** | −1.369*** | −1.369*** |
| (−14.79) | (−14.82) | (−14.70) | (−14.76) | (−29.88) | (−29.87) | (−29.87) | (−29.87) | |
| MB | −0.017*** | −0.017*** | −0.016** | −0.017*** | −0.001 | −0.001 | −0.001 | −0.001 |
| (−2.59) | (−2.61) | (−2.50) | (−2.62) | (−0.54) | (−0.54) | (−0.54) | (−0.54) | |
| ROAt-1 | 0.128 | 0.122 | 0.136 | 0.124 | −0.032 | −0.033 | −0.033 | −0.033 |
| (1.35) | (1.30) | (1.42) | (1.32) | (−0.74) | (−0.74) | (−0.75) | (−0.74) | |
| PPEt-1 | 0.035 | 0.037 | 0.026 | 0.035 | −0.014 | −0.013 | −0.013 | −0.013 |
| (0.71) | (0.74) | (0.54) | (0.71) | (−0.53) | (−0.51) | (−0.51) | (−0.51) | |
| INTANt-1 | −1.574*** | −1.583*** | −1.593*** | −1.573*** | −0.541*** | −0.543*** | −0.544*** | −0.544*** |
| (−4.56) | (−4.60) | (−4.57) | (−4.60) | (−2.65) | (−2.65) | (−2.66) | (−2.66) | |
| OWN | 0.054 | 0.058 | 0.029 | 0.059 | 0.050 | 0.051 | 0.051 | 0.051 |
| (0.85) | (0.89) | (0.43) | (0.91) | (1.05) | (1.06) | (1.07) | (1.06) | |
| FOR | −0.295*** | −0.293*** | −0.298*** | −0.298*** | −0.073 | −0.074 | −0.074 | −0.074 |
| (−2.96) | (−2.96) | (−2.92) | (−3.02) | (−1.41) | (−1.42) | (−1.42) | (−1.42) | |
| AGE | 0.003 | 0.003 | 0.001 | 0.003 | −0.006 | −0.006 | −0.006 | −0.006 |
| (0.21) | (0.21) | (0.10) | (0.23) | (−0.41) | (−0.38) | (−0.39) | (−0.39) | |
| LOSS | −0.020 | −0.020 | −0.017 | −0.020 | 0.016*** | 0.016*** | 0.016*** | 0.016*** |
| (−1.37) | (−1.36) | (−1.19) | (−1.39) | (3.79) | (3.79) | (3.80) | (3.79) | |
| BIG4 | −0.027 | −0.028 | −0.030 | −0.027 | ||||
| (−1.40) | (−1.44) | (−1.54) | (−1.42) | |||||
| N | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 | 3,195 |
| Fixed Effects | ||||||||
| Year | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Industry | Yes | Yes | Yes | Yes | No | No | No | No |
| Firm | No | No | No | No | Yes | Yes | Yes | Yes |
| Adj R2 | 0.24 | 0.24 | 0.23 | 0.24 | – | – | – | – |
| within R2 | – | – | – | – | 0.56 | 0.56 | 0.56 | 0.56 |
| F-Statistics | 16.68*** | 16.72*** | 16.99*** | 16.08*** | 64.07*** | 63.88*** | 63.94*** | 60.40*** |
Note(s): Standard errors are clustered at the firm level
*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively
The OLS results show that AvgID_Total is negative and significant. When interlocks are decomposed by director type, AvgID_Inside and AvgID_Outside also have negative coefficients, with stronger significance for inside-director interlocks. These results suggest that firms with more intensive board interlocks report lower REM in cross-sectional comparisons.
The FE results, however, do not show a significant relation between board interlocks and REM. The coefficients on AvgID_Total, AvgID_Inside, and AvgID_Outside become small and insignificant after firm fixed effects are included. The OLS association therefore appears to reflect persistent differences across firms rather than within-firm changes in interlock intensity.
Overall, Table 8 indicates that board interlocks are more consistently related to accrual-based reporting discretion than to real operating manipulation. This difference is plausible because AEM depends on accounting estimates and reporting choices, whereas REM involves operating decisions tied to firm-specific production, sales, and investment conditions. The REM results therefore qualify, rather than overturn, the main finding: the positive interlock–earnings management relation is concentrated in accrual-based discretion.
Table 9 reports the propensity score matching (PSM) estimates. This analysis complements the firm fixed-effects regressions by addressing observable differences between firms with and without board interlocks. Firm fixed effects control for time-invariant unobserved firm characteristics, whereas PSM improves covariate balance on observable firm characteristics. These methods do not remove all endogeneity concerns, but they help assess whether the interlock–earnings management relation is driven by observable selection.
Treatment effects of board interlocks – propensity score matching estimates
| Variable | Treated | Control | ATET | Std. Err. | z | p | 95% CI |
|---|---|---|---|---|---|---|---|
| AvgID_Total | 570 | 2,624 | 0.007 | 0.003 | 2.38 | 0.017 | [0.001, 0.013] |
| AvgID_Inside | 584 | 2,611 | 0.007 | 0.003 | 2.76 | 0.006 | [0.002, 0.012] |
| AvgID_Outside | 433 | 2,762 | 0.002 | 0.003 | 0.62 | 0.533 | [−0.004, 0.009] |
The ATET estimates show that firms with AvgID_Total interlocks exhibit higher earnings management than matched control firms. The effect is also positive and significant for AvgID_Inside. In contrast, the ATET for AvgID_Outside is positive but statistically insignificant. These results suggest that the average treatment difference is more evident for total and inside-director interlocks.
Table 10 reports standardized mean differences before and after matching. Covariate balance improves substantially after matching. The mean bias declines from 18.1% to 3.4% for AvgID_Total, from 15.8% to 4.1% for AvgID_Inside, and from 20.0% to 3.6% for AvgID_Outside. Most individual covariate imbalances also fall below conventional thresholds after matching. Appendix Table A5 reports additional balance diagnostics, including pseudo-R2, Rubin's B, Rubin's R, and LR χ2 tests, which show the same improvement.
Standardized mean differences before and after matching
| Variable | AvgID_Total | AvgID_Inside | AvgID_Outside | |||
|---|---|---|---|---|---|---|
| Bias before (%) | Bias after (%) | Bias before (%) | Bias after (%) | Bias before (%) | Bias after (%) | |
| SIZE | 42.7 | 3.9 | 34.7 | 4.6 | 66.4 | −0.2 |
| LEV | −2.2 | 2.3 | 5.0 | −9.3 | 10.1 | −9.1 |
| CFO | 8.1 | −2.0 | −2.1 | 0.8 | 15.8 | −8.5 |
| MB | −27.0 | 1.9 | −30.0 | −4.7 | −17.7 | −5.6 |
| ROAt-1 | 10.0 | −0.4 | 6.9 | 3.3 | 18.7 | 4.6 |
| PPE t-1 | 12.3 | 3.1 | 17.6 | −1.5 | 3.2 | −0.9 |
| INTANt-1 | 12.8 | 3.1 | 3.6 | 1.2 | 20.4 | −0.8 |
| OWN | 39.3 | −3.8 | 47.2 | −8.7 | 1.4 | −5.8 |
| FOR | 7.6 | 1.3 | 0.1 | −0.3 | 26.4 | −0.3 |
| AGE | 8.1 | −10.0 | 8.5 | 6.5 | 6.1 | 1.2 |
| LOSS | −23.2 | −2.1 | −15.5 | −3.7 | −19.6 | 5.0 |
| BIG4 | 23.9 | 6.6 | 18.4 | 5.2 | 34.3 | 1.0 |
Note(s): The table reports standardized mean differences (SMDs) between treated and control firms before and after propensity score matching. Values closer to zero indicate better covariate balance. Following common practice, absolute values below 10% are considered indicative of satisfactory balance
Table 11 reports OLS and firm fixed-effects regressions using the matched samples. These regressions differ from the ATET estimates because they examine whether interlock intensity is associated with earnings management within the matched samples, rather than estimating the average difference between treated and matched control firms. AvgID_Total and AvgID_Inside are weaker and less stable than in the ATET analysis. By contrast, AvgID_Outside remains positively associated with earnings management in the matched-sample regressions.
OLS and fixed-effects results – after propensity score matching
| Variables | OLS | Panel – Fixed effect | ||||
|---|---|---|---|---|---|---|
| (31) | (32) | (33) | (34) | (35) | (36) | |
| AvgID_Total | 0.004 | 0.004 | ||||
| (1.40) | (0.85) | |||||
| AvgID_Inside | 0.002 | −0.007 | ||||
| (0.59) | (−1.25) | |||||
| AvgID_Outside | 0.006* | 0.010** | ||||
| (1.88) | (2.00) | |||||
| SIZE | −0.003* | −0.003** | −0.000 | 0.008 | 0.015 | 0.006 |
| (−1.77) | (−2.02) | (−0.33) | (0.47) | (1.09) | (0.29) | |
| LEV | 0.006 | 0.028*** | 0.006 | 0.015 | 0.081* | 0.039 |
| (0.69) | (2.91) | (0.66) | (0.32) | (1.91) | (0.86) | |
| CFOt-1 | −0.010 | −0.037 | −0.016 | −0.069 | 0.019 | 0.014 |
| (−0.36) | (−1.35) | (−0.60) | (−1.24) | (0.30) | (0.27) | |
| MB | 0.004*** | 0.002* | 0.002** | −0.001 | −0.000 | 0.001 |
| (3.43) | (1.74) | (2.36) | (−0.47) | (−0.05) | (0.75) | |
| ROAt-1 | −0.013 | 0.029 | 0.028 | −0.008 | −0.056 | 0.075 |
| (−0.52) | (1.07) | (0.98) | (−0.13) | (−1.10) | (1.19) | |
| PPEt-1 | −0.028*** | −0.043*** | −0.028*** | −0.010 | −0.004 | −0.029 |
| (−3.41) | (−4.94) | (−3.36) | (−0.24) | (−0.10) | (−0.95) | |
| INTANt-1 | 0.012 | −0.068 | −0.053 | 0.232 | −0.044 | 0.065 |
| (0.29) | (−1.40) | (−1.31) | (1.61) | (−0.25) | (0.50) | |
| OWN | −0.027** | −0.032*** | −0.043*** | 0.021 | 0.075 | −0.052 |
| (−2.36) | (−2.96) | (−3.99) | (0.13) | (0.54) | (−1.06) | |
| FOR | −0.009 | 0.004 | −0.007 | 0.070 | 0.075 | 0.077 |
| (−0.41) | (0.22) | (−0.40) | (1.14) | (1.25) | (1.55) | |
| AGE | −0.001 | −0.005** | −0.006*** | 0.032 | −0.025 | −0.002 |
| (−0.57) | (−2.37) | (−3.12) | (1.61) | (−1.19) | (−0.21) | |
| LOSS | 0.000 | 0.002 | 0.011** | −0.011* | −0.001 | 0.010** |
| (0.10) | (0.38) | (2.57) | (−1.76) | (−0.16) | (1.98) | |
| BIG4 | 0.003 | 0.005 | −0.000 | – | – | – |
| (0.89) | (1.34) | (−0.11) | – | – | – | |
| N | 902 | 935 | 1,022 | 902 | 935 | 1,022 |
| Fixed Effects | ||||||
| Year | Yes | Yes | Yes | Yes | Yes | Yes |
| Industry | Yes | Yes | Yes | No | No | No |
| Firm | No | No | No | Yes | Yes | Yes |
| Adj R2 | 0.04 | 0.07 | 0.05 | – | – | – |
| within R2 | – | – | – | 0.05 | 0.07 | 0.06 |
| F-Statistics | 2.73*** | 4.04*** | 3.48*** | 2.09*** | 2.44*** | 2.27*** |
Note(s): Standard errors are clustered at the firm level
*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively
This divergence is informative. The significant ATET for AvgID_Inside suggests that inside-director interlocks distinguish firms with higher average discretionary reporting from comparable non-interlocked firms. This pattern is consistent with the view that inside directors, who participate directly in managerial decision-making, may transmit internal reporting practices across firms. The stronger matched-sample regression result for AvgID_Outside points to a different margin: among otherwise comparable firms, variation in outside-director interlock intensity is more closely related to earnings management. This result fits a monitoring-capacity interpretation, because outside directors with more external board commitments may have less time and attention for firm-specific reporting oversight.
Taken together, Tables 9–11 suggest that the main findings are not solely attributable to observable differences between interlocked and non-interlocked firms. The mixed PSM evidence also cautions against strong causal claims. The results are best interpreted as showing that board interlocks remain related to earnings management after observable covariate balance is improved, rather than as definitive causal evidence.
5. Conclusions
This study examines whether board interlocks are related to earnings management among KOSPI-listed firms. Prior literature offers competing predictions. Resource dependence theory views interlocked directors as sources of external knowledge, advice, and reputational capital. The busyness hypothesis and network diffusion arguments, however, suggest that interlocks can weaken monitoring or transmit discretionary reporting practices across firms. In Korea, where ownership is concentrated and director networks are common, weakened monitoring and diffusion mechanisms are likely to play an important role.
The empirical results show a positive relation between board interlocks and accrual-based earnings management. The baseline OLS and firm fixed-effects results indicate that firms with higher interlock intensity report greater discretionary accruals. This pattern remains evident under alternative interlock definitions and when discretionary accruals are estimated using the performance-matched Kothari model. The effect is not large, but it is economically meaningful. A one-standard-deviation increase in AvgID_Total corresponds to an increase in discretionary accruals of about 0.004, or 10.8% of the sample mean of earnings management.
Additional analyses refine this interpretation. Signed discretionary accrual tests show that interlocks are related to reporting discretion in both income-increasing and income-decreasing directions, not to a systematic shift in one direction. The REM results are weaker. Interlocks are negatively related to REM in OLS models, but this relation disappears after firm fixed effects are included. The main evidence is therefore concentrated in accrual-based reporting discretion. The PSM analyses also show mixed but informative results. ATET estimates are stronger for total and inside-director interlocks, whereas matched-sample regressions show more stable evidence for outside-director interlocks. This difference suggests that interlocks operate through more than one margin: average differences between connected and unconnected firms and variation in monitoring capacity among comparable firms.
This study contributes to the literature in three ways. First, it links board interlocks to earnings management, a direct financial reporting outcome. Second, it separates inside-director and outside-director interlocks, showing that aggregate interlock measures can hide director-type differences. Third, it provides Korean evidence from a setting where concentrated ownership, governance reforms, and dense director networks allow monitoring, busyness, and diffusion mechanisms to coexist.
The findings also offer practical implications. Firms should not evaluate board networks only as sources of advice or reputation. Director workload and repeated appointments can affect financial reporting oversight. For regulators, the results indicate that formal board independence may not be enough when directors are embedded in dense interfirm networks. Better disclosure of director commitments and closer attention to repeated board appointments may strengthen reporting oversight. For investors and auditors, board interlocks provide useful information about reporting risk and should be considered alongside conventional governance indicators.
Several limitations remain. Firm fixed effects, PSM, matched-sample regressions, and alternative earnings management measures reduce some concerns, but they do not establish definitive causality. Time-varying omitted variables, simultaneity, and dynamic selection may remain. The available data also do not allow a direct test of whether firms share directors with prior earnings-managing firms, as in diffusion-based designs. Future research could use exogenous director departures, regulatory shocks, or richer director-level network data to identify causal mechanisms more directly. It could also examine network centrality, director-level concentration of interlocks, and other reporting quality outcomes to clarify how board networks shape corporate financial reporting.
The supplementary material for this article can be found online.
