Director Network Centrality and Corporate Tax Avoidance

Director Network Centrality and Corporate Tax Avoidance | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Director Network Centrality and Corporate Tax Avoidance Ahmed Bouteska, Shikuan Zhao This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9196981/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract This study investigates the relationship between director network centrality and corporate tax avoidance, using data from publicly listed firms in Taiwan, an emerging market characterized by high ownership concentration, strong informal networks, and institutional voids. Drawing on the integrated agency-resource dependence framework, we propose and test a non-linear association between board connectivity and tax behavior. Using multiple measures of network centrality (degree, betweenness, closeness, and eigenvector) and tax avoidance (ETR, CETR, CAETR), we confirm the robustness of this non-linear effect using lagged specifications, alternative centrality metrics, and formal nonlinearity tests (RESET, AIC/BIC, Lind-Mehlum slope changes). Graphical analysis of predictive margins further supports the curvilinear pattern. We find a robust inverted-U relationship: moderate board connectivity appears to strengthen reputational discipline and reduce tax avoidance, whereas excessive connectivity corresponds with greater capacity for complex tax planning. We temper our claims: these results suggest that a strictly linear view of governance may be incomplete in institutional contexts characterized by concentrated ownership, informal director networks, and weaker enforcement such as Taiwan. We therefore present our findings as context-bounded evidence that highlights boundary conditions for the governance literature, rather than as a universal refutation of linear governance models. Tax avoidance director networks board centrality social capital emerging markets corporate governance Figures Figure 1 1. Introduction Taxes represent one of the most significant and persistent costs for corporations and are often perceived as a constraint on profitability (Chen et al., 2010 ). In response, firms increasingly engage in tax minimization strategies to enhance after-tax income, among them, tax avoidance, which involves exploiting legal gray areas, manipulating tax planning tools, or leveraging complex financial structures to reduce tax liabilities. According to Dyreng et al. ( 2008 ), corporate tax avoidance refers to any strategic action that lowers a firm's tax obligations relative to its pre-tax earnings. Although tax avoidance can improve cash flows and shareholder value in the short term, it also introduces reputational and regulatory risks (Armstrong et al., 2015 ), especially when it borders on aggressive or opaque behavior. Prior studies reveal that tax avoidance is shaped by various firm-specific and governance-related characteristics, such as size, leverage, profitability (Manzon and Plesko, 2002 ), capital intensity (Rego, 2003 ), board composition (Minnick and Noga, 2010 ; Lanis and Richardson, 2011 ), audit quality (Kanagaretnam et al., 2016 ; Gaaya et al., 2017 ), institutional ownership (Khurana and Moser, 2013 ), and external financing pressures (Aparicio and Kim, 2023 ). More recently, researchers have begun to explore how the social and structural positions of directors—particularly their embeddedness in interlocking board networks—may shape corporate risk-taking and ethical behavior (Ferris et al., 2017 ). However, the specific influence of director network centrality on tax avoidance remains largely unexplored, despite its theoretical relevance to both corporate governance and financial strategy. Drawing on social capital theory and the integrated agency-resource dependence perspective, our study investigates whether director connectivity reduces or encourages corporate tax avoidance. Prior literature offers two contrasting predictions. On the one hand, directors who are well-networked may be more constrained by reputational concerns and social scrutiny, thereby deterring them from engaging in controversial tax strategies (Bertrand and Mullainathan, 2003 ; Masulis and Mobbs, 2014 ). On the other hand, central directors may possess privileged access to information, technical expertise, and peer-driven strategies, potentially enabling more sophisticated and aggressive forms of tax planning (Desai and Dharmapala, 2009 ; Fracassi and Tate, 2012 ). This study contributes to a growing literature that links board networks to corporate financial decision-making, including innovation (Han et al., 2015 ), capital access (Li et al., 2019 ; Goncalves et al., 2019 ), and post-merger performance (El-Khatib et al., 2015 ). In contrast to prior studies that often assume a linear relationship between governance mechanisms and outcomes, we propose and empirically demonstrate that the link between director network centrality and tax avoidance is nonlinear and exhibits an inverted-U shape. Specifically, we show that firms with moderately connected directors experience lower levels of tax avoidance, likely due to enhanced board oversight and reputational constraints. However, beyond a certain threshold, higher centrality correlates with increased tax avoidance, suggesting that social capital may transform into a vehicle for strategic opportunism. To test this hypothesis, we analyze a panel of 3,422 firm-year observations from 313 firms listed on the Taiwan Stock Exchange between 2012 and 2023. Taiwan represents a particularly rich empirical context: it is an emerging market with high tax rates (OECD, 2020 ), dense informal networks, concentrated ownership, and institutional gaps in regulatory enforcement (Rajagopalan and Zhang, 2008 ; Kale, 2019 ). These features make it an ideal case for understanding the dual role of director networks, as both monitors and enablers, in shaping corporate tax behavior. While our empirical setting is Taiwan, we argue that the findings offer insights applicable to a broader set of emerging markets with comparable institutional profiles. For instance, India, Brazil, Indonesia, Mexico, South Africa, and Turkey exhibit three features that mirror Taiwan’s context: (i) ownership concentration and prevalence of family-controlled firms (La Porta et al., 1999; Khanna & Yafeh, 2007), (ii) dense informal networks and relational governance that substitute for underdeveloped formal institutions (Khanna & Palepu, 2010), and (iii) regulatory and enforcement gaps that create opportunities for opacity in corporate behavior (Fan, Wong, & Zhang, 2007 ; Aguilera & Crespi-Cladera, 2016). In these markets, directors often operate within elite networks that simultaneously constrain and enable corporate strategies. Hence, Taiwan serves not only as a theoretically rich case but also as a proxy for other emerging economies where relational governance plays a dominant role. Our contribution is threefold. First, we extend the literature on tax avoidance by introducing network centrality as a dynamic, nonlinear determinant of tax behavior, going beyond standard governance variables. Second, we provide methodological rigor through multiple centrality measures, lagged regressions, and formal nonlinearity tests (RESET, AIC/BIC, Lind-Mehlum slope tests), ensuring robustness and theoretical validity. Third, we highlight the global relevance of director social capital by anchoring our findings within the context of emerging-market governance and regulatory design. Finally, we emphasize that our empirical claims are context-sensitive. Taiwan’s institutional architecture, high ownership concentration, pronounced informal director networks, and uneven enforcement dynamics, provides a particularly informative testbed for network effects on tax policy. Accordingly, while our inverted-U finding identifies an important boundary condition for governance theory, the interpretation should be conditional: the mechanism we identify is most likely to operate in emerging markets or other settings where informal social capital substitutes for formal controls. We explicitly discuss external validity and provide robustness diagnostics and interaction tests to illustrate the heterogeneity of effects across institutional settings. The remainder of the paper is structured as follows. Section 2 reviews the relevant literature and provides the institutional background. Section 3 outlines the data, variables, and empirical methodology. Section 4 presents the baseline regression results and the inverted-U analysis. Section 5 includes robustness checks and graphical validations. Section 6 concludes the paper and discusses theoretical, policy, and cross-national implications. 2. Literature Review and Institutional Background The structure, behavior, and influence of board networks have drawn increasing scholarly interest in recent years, particularly as they relate to corporate ethics and financial discretion. Directors embedded in powerful networks often bring informational advantages, resource access, and reputational constraints that shape firm behavior in complex ways. From the standpoint of labor market dynamics, well-connected directors are actively sought after by companies due to their social capital and access to privileged knowledge (Andersen et al., 2022 ). In turn, these directors are incentivized to protect their reputational capital by avoiding associations with unethical or high-risk firms. As Masulis and Mobbs ( 2014 ) note, directors who become linked to scandal-tainted companies face reduced chances of future board appointments and diminished compensation. From this perspective, board centrality serves a disciplinary function: directors with high centrality may exercise greater oversight, deter opportunistic management behavior, and safeguard firm reputation. This interpretation is supported by the "quiet life" hypothesis (Bertrand and Mullainathan, 2003 ), which suggests that powerful directors with secure career trajectories prefer to avoid risky or aggressive practices, such as corporate tax avoidance. Consequently, one would expect that firms governed by highly central directors would demonstrate lower levels of tax aggressiveness. However, emerging insights from the agency-resource dependence framework complicate this view. As Tao et al. ( 2019 ) argue, networked directors may simultaneously become channels for transmitting opaque strategies and information asymmetries. In such settings, board interlocks may no longer serve as monitors but rather as vehicles for normalizing opportunistic practices. Peer influence among elite networks can facilitate the diffusion of tax planning tactics across firms (Fracassi and Tate, 2012 ), including strategies that border on artificial tax avoidance. Cheng et al. ( 2019 ) document that such networks have been instrumental in the dissemination of unethical practices such as earnings manipulation and options backdating. These findings suggest a dual role of board centrality, initially constraining but eventually enabling opportunistic behavior as influence and insulation grow. This duality becomes particularly salient in the context of corporate tax avoidance. As Desai and Dharmapala ( 2006 ) explain, tax avoidance often requires both “elaborateness” and “opacity”—features that thrive in environments of weak monitoring and strong insider access. When networks become too dense or powerful, the same directors who once imposed constraints may instead facilitate access to elite tax professionals, regulatory arbitrage opportunities, and high-level tax shelters. Hence, we posit that the relationship between director network centrality and tax avoidance is nonlinear and inverted-U shaped: social capital constrains behavior up to a threshold, after which it begins to enable strategic exploitation. Despite the expanding literature on tax avoidance, much of the evidence stems from developed economies with formalized regulatory systems (Hanlon & Heitzman, 2010 ; Richardson et al., 2013 ). In contrast, emerging markets remain underexplored, even though they present distinctive features such as informal enforcement systems, concentrated ownership structures, and dependence on relational governance. Taiwan, in this regard, provides a theoretically rich and policy-relevant context. With one of the highest statutory corporate tax rates globally (OECD, 2020 ), a complex tax code (Kale, 2019 ), and institutional voids in transparency and enforcement (Rajagopalan and Zhang, 2008 ), Taiwan mirrors many challenges faced by emerging markets. Board structures in Taiwanese firms often serve as rubber stamps for controlling shareholders, and directors frequently hold positions across multiple firms, creating dense informal networks. These networks substitute for underdeveloped formal institutions and are often the primary source of strategic coordination (Fan, Wong, & Zhang, 2007 ). Therefore, Taiwan is not only a relevant empirical setting—it is also a proxy case for other network-dependent governance environments across Asia, Latin America, and Africa. This study addresses the critical gap in understanding how director network centrality influences tax behavior in such environments. Using panel data from 313 firms over 12 years (2012–2023), we empirically test the hypothesis that board centrality first reduces, then increases corporate tax avoidance. By deploying multiple centrality metrics, degree, closeness, betweenness, and eigenvector, and three widely accepted tax avoidance proxies (ETR, CETR, CAETR), we uncover robust evidence of an inverted-U shaped relationship. Our findings are validated through multiple robustness tests, including nonlinear specification checks, lagged regressions, and graphical modeling. The results suggest that director networks serve as a double-edged sword: they initially operate as mechanisms of discipline, but beyond a threshold, they evolve into enablers of sophisticated tax planning. Social-capital saturation: micro-mechanisms for an inverted-U We reconcile the seemingly contradictory forces of reputational discipline and privileged access by introducing a social-capital saturation argument. Three linked micro-mechanisms generate an inverted-U relationship between board centrality and tax avoidance. First, visibility and reputational monitoring (low→moderate centrality). Directors with a small to moderate number of interlocks serve as bridges across firms, increasing public visibility, peer scrutiny, and the likelihood that questionable choices will be observed and sanctioned. Because reputational penalties are salient, modest increases in centrality strengthen informal monitoring and reduce incentives for aggressive tax strategies. Second, redundancy, brokerage power, and elite capture (high centrality). As centrality grows beyond a threshold, directors accumulate redundant ties, increasing brokerage power and private informational access (technical know-how, advisor networks, insider flows). Redundant ties reduce exposure to independent observers, and dense, reciprocal relationships normalize risk-taking. Brokerage and elite membership also facilitate coordinated strategies that are difficult for external monitors to detect, effectively enabling sophisticated tax planning. Third, conditional moderators and dynamic feedback. The switch from constraint to facilitation is more likely where formal enforcement is weak, ownership is concentrated, and audit or market monitoring is limited. Conversely, strong third-party oversight raises the cost of opportunism and shifts the tipping point. Repeated appointments and network reinforcement can lock in clique norms, making the enabling effect persistent. These mechanisms yield the theoretical prediction that the relationship between director network centrality and tax avoidance is nonlinear and inverted-U shaped, with moderate connectivity constraining tax aggressiveness and high connectivity enabling sophisticated tax planning. 3. Sample and Methodology Data and Sample Selection This study is based on an unbalanced panel dataset of 313 non-financial, non-utility companies listed on the Taiwan Stock Exchange (TWSE) between 2012 and 2023, yielding 3,422 firm-year observations. Financial and utility firms are excluded due to their distinct tax regulations, disclosure requirements, and capital structures (similar to Hanlon and Heitzman, 2010 ). Observations with negative pretax income are also excluded, as effective tax rate (ETR) measures are undefined or distorted in such cases. We construct our director network measures using board interlock data derived from Bloomberg, which maps the affiliations of each director across firms and years. Governance and firm-level financial data, including board structure, tax figures, and accounting metrics, are extracted from the Taiwan Economic Journal (TEJ) database, consistent with prior studies on Taiwanese corporate governance (Tao et al., 2019 ; Kale, 2019 ). The final sample enables a comprehensive examination of the relationship between board network centrality and tax behavior in a developing market context characterized by complex tax structures and network-dependent governance. Sample Considerations and Institutional Context While our empirical setting is Taiwan, this context provides a theoretically informative case for understanding board networks in emerging markets. Taiwanese firms typically exhibit concentrated ownership, dense interlocking director networks, and partially informal enforcement systems. These characteristics make Taiwan an ideal laboratory for examining how director social capital influences tax avoidance. We note, however, that the findings may not translate directly to developed markets with formalized governance structures. Nonetheless, the mechanisms uncovered, moderate connectivity fostering reputational monitoring and high connectivity enabling strategic coordination, are likely applicable in other emerging markets with similar relational governance dynamics, such as parts of Latin America, Africa, and Asia. Empirical Models and Variable Definitions To examine the influence of director networks on corporate tax avoidance, we estimate the following baseline panel regression model using Generalized Least Squares (GLS) with robust standard errors clustered at the firm level to correct for intra-firm autocorrelation and heteroskedasticity: $$\:{TA}_{it}\:=\:{\alpha\:}_{it}+{{\beta\:}_{1}CB}_{it}+\:{{\beta\:}_{2}CB}_{it}^{2}\:+\gamma\:{X}_{it}+{\mu\:}_{i}+{\lambda\:}_{t}+{\epsilon\:}_{it}$$ 1 Where, the Tax avoidance (TA) is the tax avoidance proxy, measured alternately using ETR, CETR, and CAETR. CB represents centrality metrics (DCB, CCB, BCB, ECB). CB² captures nonlinearity (quadratic specification) to test for the inverted-U relationship. X is a vector of firm- and board-level control variables. µ and λ are firm and year fixed effects, respectively. ε is the error term. This specification is estimated separately for each centrality type to mitigate multicollinearity concerns between network metrics. Following recent research (Arora and Gill, 2022 ; Aparicio and Kim, 2023 ), we also estimate lagged specifications in robustness tests to address endogeneity and reverse causality concerns. Our baseline specification is estimated using Generalized Least Squares (GLS) with firm-clustered robust standard errors. We adopt GLS because tax avoidance measures exhibit both cross-sectional heteroskedasticity and within-firm serial correlation, which GLS efficiently accommodates. Moreover, we retain firm-fixed and year-fixed effects within GLS, ensuring control for time-invariant firm traits and common shocks. This estimator has precedent in the governance literature examining director networks and firm policies in emerging markets (Fracassi & Tate, 2012 ; Cheng et al., 2019 ; Arora & Gill, 2022 ). To address concerns that results may depend on GLS, we also re-estimate our models using fixed-effects OLS, dynamic system-GMM, IV-2SLS, propensity-score matched regressions, and inverse-probability weighting. The inverted-U relationship persists across all methods (see Tables 14 – 16 and A1 ). Thus, while GLS provides efficiency gains under our data structure, our substantive conclusions are not sensitive to the estimation strategy. In line with prior studies (Frank et al., 2009 ; Dyreng et al., 2017 ; Minnick and Noga, 2010 ), we employ three distinct measures of effective tax rates: ETR: Total tax expense divided by pretax income; CETR: Current tax expense divided by pretax income; CAETR: Cash taxes paid divided by pretax income. Lower values of these measures indicate more aggressive tax avoidance 1. These metrics are widely used in both developed and emerging market tax avoidance studies and offer a robust triangulation of firms' tax planning intensity. We construct four centrality measures, normalized annually at the director level and aggregated to the board level: Degree Centrality (DCB): Number of direct ties a director maintains; Closeness Centrality (CCB): Inverse of the average distance to all other directors, capturing information access speed; Betweenness Centrality (BCB): Frequency with which a director lies on the shortest paths between others, indicating brokerage power; Eigenvector Centrality (ECB): Influence of a director based on both direct and indirect ties to highly connected individuals. Each metric is squared to capture nonlinear (inverted-U) dynamics. Aggregation is performed across all board members, following Godigbe et al. ( 2018 ) and Tsai et al. ( 2019 ). We include the following controls, drawn from established literature (Armstrong et al., 2015 ; Richardson et al., 2013 ; Jarboui et al., 2020 ): SB: Number of board members; IDB: Ratio of independent directors; FDB: Ratio of female directors; PH: Promoter shareholding (% of total); SIZE: Log of total assets; ROA: Return on assets; LEV: Total debt to total assets; R&D: R&D expenditure to total assets; PPE: Net property, plant, and equipment to total assets; AGE: Years since firm incorporation; Year and firm fixed effects are included to control for macroeconomic shocks and time-invariant firm traits. The definitions of the variables are provided in Table 1 in appendix. Director-background controls: To control for potential omitted heterogeneity at the director level, we also construct four board-aggregated measures from director CVs and board membership records (aggregated to the firm–year): (i) PCT_FINANCE, share of directors with prior financial/accounting experience (CFO, audit partner, investment banker, Big-4 auditor); (ii) PCT_POLITICAL, share of directors with political ties (former public office, political appointments, or close official affiliations); (iii) AVG_TENURE, average board tenure (years) of directors; and (iv) PCT_ADVDEG, share of directors with advanced business/technical degrees (MBA, MSc, PhD). These variables are added to the control vector ( \(\:{X}_{i,t}\) ) in all baseline, quadratic, and robustness specifications. (Please Insert Table 1 Here) 4. Empirical Results Preliminary Results We begin by presenting the descriptive statistics for the full sample in Table 2 . The mean values for the tax avoidance proxies, ETR (0.452), CETR (0.399), and CAETR (0.288), suggest substantial variability in effective tax burdens across firms. While the median tax rates (ETR: 44.9%, CETR: 38.9%, CAETR: 28.6 28.6%) are broadly consistent with Taiwan’s statutory corporate tax rate (ranging from 17% to 45% during the period), the extremely high standard deviations and wide min-max ranges indicate that a sizable subset of firms engage in aggressive tax strategies, consistent with prior findings by Chen et al. ( 2010 ) and Arora and Gill ( 2022 ). The dispersion across the centrality measures, degree centrality (DCB), closeness centrality (CCB), eigenvector centrality (ECB), and betweenness centrality (BCB), also highlights significant heterogeneity in network structures across boards. For example, the standard deviation of BCB is large (4,793.84), reflecting considerable differences in strategic positioning within board networks. These descriptive statistics suggest a ripe setting for testing nonlinear patterns of influence across firms and over time. Table 3 presents the Pearson correlation coefficients for all primary variables. As expected, the tax avoidance proxies (ETR, CETR, CAETR) are positively correlated with one another (p < 0.01) but not perfectly collinear (ETR–CETR ≈ 0.01; ETR–CAETR ≈ − 0.01, affirming that they capture overlapping dimensions of tax behavior. Importantly, the correlations between the centrality variables and tax avoidance metrics are low to moderate in magnitude, suggesting that network centrality explains variation in tax behavior beyond firm fundamentals. The observed relationships also preliminarily support a nonlinear effect, with weak or mixed signs across linear terms, consistent with our later hypothesis of an inverted-U pattern. The centrality measures also are modestly correlated with firm characteristics, a pattern consistent with the multivariate regression results. We also compute variance inflation factors (VIFs) for all predictors to test for multicollinearity. All VIF values are well below the commonly accepted threshold of 5 (maximum VIF = 1.64), confirming the absence of multicollinearity and strengthening the reliability of our regression estimates. These diagnostic checks validate the soundness of our empirical model and variable construction. Taken together, the preliminary analysis provides early evidence that (1) there is significant variation in both tax avoidance and network centrality across firms, (2) the relationships are likely more complex than linearity would suggest, and (3) our data structure is robust enough for advanced multivariate modeling, including the non-linear GLS regressions presented in the next section. (Please Insert Table 2 Here) (Please Insert Table 3 Here) Baseline Results Table 4 presents the results of our baseline GLS regressions, estimating the relationship between director network centrality and corporate tax avoidance using three measures: ETR, CETR, and CAETR 2. All models include firm-fixed effects and year-fixed effects, with robust standard errors clustered at the firm level. We find that degree centrality (DCB) is positively and significantly associated with each tax rate measure (p < 0.01), suggesting that as network centrality increases, firms report higher effective tax rates—implying less tax avoidance. This is consistent with the argument that well-connected directors have stronger reputational incentives to avoid risky or unethical behavior (Masulis & Mobbs, 2014 ), and may exert greater oversight in boardrooms. These results align with the "quiet life" hypothesis (Bertrand & Mullainathan, 2003 ), which posits that influential directors are less inclined to engage in complex, controversial strategies such as aggressive tax avoidance. Control variables perform as expected. Larger firms (SIZE) are more likely to avoid taxes, consistent with prior literature that links organizational complexity with aggressive planning (Richardson et al., 2013 ). Higher profitability (ROA) is associated with greater tax avoidance, which may reflect firms in emerging markets allocating surplus resources toward financial optimization. Leverage (LEV) also shows a significant positive correlation with ETR and CETR, indicating reduced tax avoidance, consistent with Badertscher et al. ( 2013 ). R&D intensity (R&D) is negatively associated with tax rates, supporting the idea that firms with intangible-heavy business models may exploit more tax-saving opportunities. Older firms (AGE) are modestly associated with higher ETRs, echoing the finding that mature firms may avoid reputational risk and regulatory scrutiny (Higgins et al., 2015 ). When we include the board-aggregated director-background controls described above (see Table 18 ), the sign, significance and economic magnitude of the DCB and \(\:{DCB}^{2}\) terms remain essentially unchanged, demonstrating that our core inverted-U result is not driven by observable director heterogeneity. (Please Insert Table 4 Here) To test the core hypothesis that the effect of director networks on tax avoidance is nonlinear, we include the squared term of DCB in the regressions. Table 5 presents the results. The linear term remains significantly positive, while the quadratic term (DCB²) is significantly negative for ETR and CETR, confirming a robust inverted-U relationship. This finding reveals a crucial dynamic: initially, as board connectivity rises, tax avoidance decreases—likely due to improved governance and reputational constraints. However, beyond a certain point, further increases in centrality reverse this effect, with firms exhibiting lower effective tax rates, suggesting more aggressive tax behavior. In line with Fracassi and Tate ( 2012 ) and Cheng et al. ( 2019 ), this implies that excessive centrality transforms board networks into enablers of opportunistic behavior, such as the diffusion of complex tax strategies. This nonlinear finding is also supported by our predictive margins analysis and formal tests (presented in Section 5), which show that the turning point lies well within the data range, and the model fit improves significantly compared to linear specifications. Directors with elevated centrality may access sophisticated tax advisors, private knowledge flows, or insider-based tax schemes (Cai & Sevilir, 2012 ), which would explain the shift from constraint to facilitation. These results confirm the theoretical premise of this study: director social capital is a double-edged sword, it may enhance ethical governance up to a point, but beyond that, it fosters strategic rent-seeking and regulatory arbitrage. (Please Insert Table 5 Here) Additional Analyses and Tests To verify the robustness and generalizability of our main findings, we conduct a set of additional analyses using alternative measures of director network centrality: closeness centrality (CCB), betweenness centrality (BCB), and eigenvector centrality (ECB). Each metric captures distinct dimensions of a director’s network position. CCB reflects the speed of information flow, BCB measures brokerage and influence over network flows, while ECB indicates indirect influence through connections to other well-connected directors. Table 6 reports the results of GLS regressions using these centrality proxies. The coefficients for CCB and BCB are consistently positive and statistically significant at the 1% level across all three tax avoidance proxies (ETR, CETR, and CAETR). This supports our core result that increased director connectivity is associated with higher reported tax rates, and hence lower tax avoidance, in the linear portion of the relationship. These findings reinforce the idea that moderate levels of board connectivity enhance governance by facilitating transparency and reputational accountability (Ferris et al., 2017 ). Interestingly, the ECB coefficient is not statistically significant for ETR and CETR and only weakly so for CAETR. Since ECB reflects indirect, rather than direct, influence, this may suggest that peripheral or second-order connections are less effective in shaping observable tax policy decisions. The differing influence of centrality types highlights the strategic relevance of direct vs. indirect network positions in shaping corporate financial behavior. To further test our hypothesis of nonlinearity, Table 7 presents regression results incorporating the squared terms for CCB, BCB, and ECB. As with our earlier findings for DCB, the results show a statistically significant inverted-U relationship for CCB and BCB. Specifically, the coefficients on CCB² and BCB² are negative and significant at the 1% level, suggesting that the disciplining effect of board networks reverses at higher levels of centrality. Directors with excessive proximity to others or excessive brokerage power may accumulate strategic information and develop coalitions that enable sophisticated tax planning techniques—consistent with findings in Fracassi and Tate ( 2012 ) and Cheng et al. ( 2019 ). In contrast, ECB² does not display a consistent pattern, further supporting the notion that indirect or less transparent forms of connectivity may exert weaker or more ambiguous influence on tax decisions. Overall, these additional tests reinforce the robustness and multi-dimensional validity of our core result: the relationship between director network centrality and corporate tax avoidance is nonlinear, inverted-U shaped, and manifests consistently across different definitions of centrality. These findings align with broader insights from social network theory, where influence and constraint are contingent on the nature and strength of relational ties (Adler & Kwon, 2002 ; Cai & Sevilir, 2012 ). (Please Insert Table 6 Here) (Please Insert Table 7 Here) 5. Robustness Tests Formal Nonlinearity Checks To statistically validate the inverted-U relationship between director network centrality and corporate tax avoidance, we conduct formal nonlinearity tests to compare the performance of linear versus quadratic specifications of our models. These tests allow us to move beyond visual or coefficient-based arguments and provide concrete evidence that the quadratic form is both better-fitting and theoretically appropriate. We begin by applying the Ramsey Regression Equation Specification Error Test (RESET) to the linear model specifications using each of the three tax avoidance proxies (ETR, CETR, CAETR) as dependent variables. The RESET test detects omitted nonlinear terms by including powers of the fitted values in the regression and testing whether they significantly improve model fit. In Table 8 , the RESET test returns statistically significant F-statistics (p < 0.01) in all three models, indicating model misspecification under the linear assumption. This supports our hypothesis that a purely linear specification fails to capture the actual data-generating process and that a non-linear (quadratic) form is statistically justified. To complement the RESET test, we compare the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) for the linear and quadratic models. Both AIC and BIC penalize model complexity while rewarding improved fit; lower values indicate better model performance. In Table 9 , the quadratic models consistently outperform the linear specifications in both AIC and BIC across all dependent variables. This provides strong additional evidence that the inclusion of the squared centrality term improves the explanatory power of the models without overfitting. Together, the results from the RESET and AIC/BIC tests confirm that the relationship between director network centrality and tax avoidance is nonlinear, and that models omitting the quadratic term suffer from functional misspecification. These findings substantiate the theoretical framing of an inverted-U, where social capital initially constrains but later enables more aggressive financial strategies (Desai & Dharmapala, 2009 ; Tao et al., 2019 ). Our robustness tests reinforce that the inverted-U shape is not only statistically detectable but also economically meaningful and robust to various model specifications. (Please Insert Table 8 Here) (Please Insert Table 9 Here) Controlling for Endogeneity We re-specify the regression model as follows: $$\:{TA}_{it}\:=\:{\alpha\:}_{it}+{{\beta\:}_{1}CB}_{it-1}+\:{{\beta\:}_{2}CB}_{it-1}^{2}\:+\gamma\:{X}_{it}+{\mu\:}_{i}+{\lambda\:}_{t}+{\epsilon\:}_{it}$$ 1 Where, \(\:\:{CB}_{it-1}\) and \(\:{CB}_{it-1}^{2}\) are the lagged centrality terms (degree centrality and its square), \(\:{TA}_{it}\) remains the current year's effective tax rate (ETR, CETR, CAETR). All control variables \(\:{X}_{it}\) , firm fixed effects \(\:{\mu\:}_{i}\) , and year fixed effects \(\:{\lambda\:}_{t}\) are included as in prior models. The regression results are presented in Table 10 . The coefficient on the lagged DCB term remains positive and significant, while the coefficient on lagged DCB² is negative and statistically significant (p < 0.01) for both ETR and CETR models. These results are nearly identical in sign and magnitude to our original contemporaneous models, confirming that the inverted-U relationship persists when controlling for endogeneity. The results for CAETR remain statistically weaker, which is expected given the noisier nature of cash-based tax proxies and the dominance of accrual tax planning in many Taiwanese firms (Kale, 2019 ). Nevertheless, the strength and stability of the inverted-U relationship using lagged variables significantly strengthens our identification strategy. The use of lagged independent variables helps reduce concerns of simultaneity and potential reverse causality, lending greater confidence that director network centrality precedes and influences tax avoidance, rather than the other way around. Although this does not eliminate all forms of endogeneity (e.g., omitted variable bias), it represents a meaningful improvement in causal inference and aligns with best practices in recent governance and accounting research (Arora & Gill, 2022 ; Tao et al., 2019 ). (Please Insert Table 10 Here) Winsorized Sample Outliers and extreme values in financial and governance data can often bias coefficient estimates or inflate standard errors, especially in emerging market settings where disclosure quality and firm heterogeneity are high (Richardson et al., 2013 ). To assess whether our findings are robust to such distortions, we conduct winsorization of all continuous independent variables at the 1st and 99th percentiles, following best practice in empirical accounting and finance literature (Hanlon & Heitzman, 2010 ; Badertscher et al., 2013 ). We re-estimate the quadratic specification of our core model after winsorizing the following variables: DCB and DCB²; SIZE, ROA, LEV, R&D, PPE, AGE, PH; ETR, CETR, and CAETR (dependent variables). Winsorization replaces extreme values with the nearest observations within the top and bottom 1% range, preserving the sample size and variance structure while removing extreme leverage points. Table 11 presents the regression results after winsorization. The main coefficients of interest—DCB and DCB²—retain their expected signs and statistical significance, with only marginal changes in magnitude. The inverted-U relationship between board centrality and tax avoidance remains intact, demonstrating that our original findings are not driven by a few extreme firms. The winsorized re-estimation confirms that our central findings are not sensitive to outliers, the inverted-U effect remains statistically strong, and the data-generating process is stable across the distribution of firms. These results further support the reliability and credibility of our conclusions regarding the dual effects of director networks on corporate tax avoidance. (Please Insert Table 11 Here) Subsample Analysis To further evaluate the robustness and generalizability of our core findings, we conduct subsample analyses by systematically removing subsets of firms that may disproportionately influence the observed relationships. These checks are particularly relevant in Taiwan’s market context, where: Technology firms dominate in scale and R&D intensity, and Firm size correlates with board structure, political ties, and disclosure sophistication. Accordingly, we perform the following two exclusion tests: First, we exclude the largest 10% of firms (based on average total assets over 2012–2023) to assess whether our findings are driven by outlier firms with excessive board complexity or global tax structures. Results from Table 12 show the inverted-U shape remains strong and statistically significant. Both the linear (DCB) and quadratic (DCB²) terms remain significant at the 1% level. Second, we remove firms classified under technology, electronics, or information services sectors (TEJ codes 2, 3, and 4), which comprise a large portion of Taiwan’s export economy and may use industry-specific tax strategies. Results from Table 13 indicate once again, the inverted-U relationship remains statistically and economically significant for ETR and CETR. CAETR is still not significant — consistent with earlier tests. These two exclusion-based robustness tests confirm that the inverted-U shaped relationship between director network centrality and corporate tax avoidance is not driven by firm size or sector concentration. This finding reinforces that our results reflect a generalizable pattern of director network effects, rather than being artifacts of large-cap tech firms that dominate Taiwan’s public markets. It also enhances the international relevance of our findings, especially for emerging markets where similar governance and sector dynamics exist (Fan et al., 2007 ; Arora & Gill, 2022 ). (Please Insert Table 12 Here) (Please Insert Table 13 Here) Predictive Margins Plot To further substantiate the nonlinear relationship between director network centrality and corporate tax avoidance, we present Fig. 1 , which visualizes the predictive margins of the effective tax rate (ETR) as a function of board degree centrality (DCB), with 95% confidence intervals. As shown, the relationship between board centrality and ETR takes the form of an inverted-U curve, confirming our empirical results from earlier GLS estimations. Specifically, at low to moderate levels of board centrality, increases in director connectivity are associated with higher effective tax rates, indicative of lower tax avoidance. However, beyond a certain threshold of centrality, the marginal effect becomes negative: higher connectivity is now linked to lower ETR values, reflecting increased corporate tax avoidance. This turning point, marked by the red dashed line in the figure, supports the theoretical proposition that board networks serve as a double-edged sword. Initially, connectivity enhances governance and reputational discipline. But as network centrality intensifies, it becomes a conduit for complex, and potentially aggressive, financial behaviors—including tax avoidance. These findings align with Fracassi and Tate ( 2012 ) and Cheng et al. ( 2019 ), who demonstrate that dense director networks can foster the diffusion of both efficient and opportunistic practices. Furthermore, the confidence band surrounding the predicted ETR values remains narrow and stable across most of the centrality distribution, confirming the statistical precision and reliability of the model’s estimates. Economic significance and prevalence of the tipping point. To help interpret the inverted-U, we compute the estimated turning point of the quadratic DCB specification and place it in the DCB distribution. Across our main specifications (Tables 5 and 18 ) the turning point is approximately 0.12 (DCB units). This value lies near the 98th percentile of the sample DCB distribution: only ~ 1.8% of firm-year observations (≈ 61/3,422) exceed this level, while 55 unique firms (≈ 17.6% of the 313 firms) have at least one year above the threshold. In economic magnitude, the model predicts that increasing DCB from the sample median (0.041) to the turning point (≈ 0.12) changes the predicted ETR by ≈ + 0.40 (ETR units), indicating a substantial effect. These statistics show two linked facts: the nonlinearity is statistically precise and economically important where it occurs, but the “high-centrality” enabling regime is relatively rare in our sample. We report bootstrap confidence intervals for the turning point and the percentile counts in Appendix A2. This Table A2 summarizes the economic significance of the turning point in the inverted-U relationship between board centrality and corporate tax avoidance. The estimated turning point of director centrality (DCB) is approximately 0.12, corresponding to the 98th percentile of the sample distribution. Only about 1.8% of firm-year observations (61 out of 3,422) lie above this threshold, representing 55 unique firms (≈ 17.6% of the sample). Moving from the sample median to the turning point is associated with an increase in effective tax rate (ETR) of roughly 0.40 units (≈ 40 percentage points), after which the marginal effect becomes negative. Bootstrap percentile confidence intervals for the turning point are reported in Table A2 . (Please Insert Fig. 1 Here) (Please Insert Table A2 Here) External validity and generalizability There is a key concern about whether our Taiwan-based results generalize to most firms. We accept that institutional context conditions inference and therefore add two clarifying elements. First, subsample investigations (Tables 12 – 13 ) show that excluding the largest 10% of firms or removing technology/electronics sectors yields qualitatively similar inverted-U patterns for ETR and CETR. These checks reduce the concern that a small set of large, export-oriented firms drives the result. Second, we add further diagnostics. Dynamic GMM (Table 14 ) addresses persistence and dynamic endogeneity; IV-2SLS (Table 15 ) corrects for potential simultaneity bias; and PSM regressions (Table 16 ) compare matched firms on observables. All three approaches confirm the inverted-U pattern, with turning points interior to the observed distribution. To assess representativeness, we implement inverse-probability weighting (Appendix A1) so that our listed-firm sample better approximates the broader firm population. The quadratic shape remains intact under IPW. We also interact centrality with institutional proxies (ownership concentration, family ownership, political ties), finding the inverted-U is strongest where enforcement is weaker. Finally, we report bootstrapped confidence intervals for the turning point, which consistently lie within the observed centrality range. Together, these checks support the view that the inverted-U effect is not an artifact of a few firms and clarify the institutional conditions under which it is strongest. Nevertheless, we reiterate that generalization beyond markets with similar institutional architectures should be cautious. The GMM results in Table 14 confirm the robustness of our baseline findings. Degree centrality (DCB) exhibits a positive and significant association with both ETR and CETR, while its squared term (DCB²) is negative and significant, reinforcing the inverted U-shaped relationship between board centrality and tax avoidance. The lagged dependent variables are significant, highlighting persistence in tax behavior. Control variables behave as expected: profitability (ROA) and R&D intensity reduce tax avoidance, while leverage increases it. The Hansen and AR(2) tests validate the instruments and specification, confirming that the results are not driven by dynamic endogeneity. The IV-2SLS estimates from Table 15 corroborate our main findings while addressing remaining endogeneity concerns. The second-stage coefficients show a positive linear DCB effect and a negative quadratic (DCB²) effect for ETR and CETR, yielding an inverted-U pattern. Instrument diagnostics (first-stage F = 28.4; Kleibergen–Paap p = 0.001) indicate instrument strength, and Hansen J tests (p > 0.20) do not reject instrument exogeneity. The IV-derived turning point for ETR (≈ 0.133) lies well inside the observed DCB distribution, implying the nonlinearity is economically meaningful and robust to IV correction. In Table 16 , PSM-based matched-sample results further support the inverted-U finding: DCB is positive and significant while DCB² is negative and significant for ETR and CETR in the matched sample. The estimated turning point for ETR is approximately 0.132 (bootstrap 95% CI: [0.115, 0.150]), which again is interior to the empirical DCB distribution. The matched-sample results show that differences in observed covariates between high- and low-centrality firms are unlikely to generate the observed nonlinearity; the inverted-U pattern survives after balancing observable characteristics. The IPW results in Table A1 confirm the inverted-U relationship. The coefficients on DCB are positive and significant, while those on DCB² are negative and significant. The estimated turning points (~ 4) are within the observed range of the centrality distribution. This indicates that the nonlinear pattern is not an artifact of sample composition. Even when re-weighted to better approximate the population of Taiwanese firms, the inverted-U holds. (Please Insert Tables from 14 to 16 Here) (Please Insert Table A1 Here) Winsorization Robustness To ensure our results are not driven by extreme observations, we winsorize all continuous variables at the 1st and 99th percentiles and re-estimate the baseline specification with the full set of controls. The results, reported in Table 17 , remain qualitatively unchanged. The coefficients on degree centrality and its squared term continue to support an inverted-U relationship between director connectivity and tax avoidance, and the estimated turning point remains well within the range of observed centrality values. These findings confirm that our main conclusions are not artifacts of outliers but reflect robust underlying patterns. More specifically, the results from remain consistent after winsorization: the inverted-U relationship between board centrality and tax avoidance is robust. All main coefficients retain significance and similar magnitudes, confirming that our findings are not driven by extreme outliers. (Please Insert Table 17 Here) Director-background controls We reproduce the baseline specification (ETR, CETR, CAETR) with the board-aggregated director background controls included. As an additional robustness test, Table 18 adds the board-aggregated director background controls. Coefficients on DCB and DCB² remain positive and negative (p < 0.01), respectively, and the turning point stays interior to the DCB distribution. More specifically, This Table 18 reports the estimation results of the impact of board centrality on corporate tax avoidance after including board-aggregated director background controls: PCT_FINANCE (share of directors with financial/accounting experience), PCT_POLITICAL (share with political ties), AVG_TENURE (average years on the board), and PCT_ADVDEG (share with advanced business/technical degrees). Results confirm the inverted-U relationship: the coefficient on DCB is positive and significant, while DCB² is negative and significant for ETR and CETR. CAETR results remain directionally consistent. Among director controls, boards with higher financial expertise are associated with less tax avoidance, while political ties modestly increase it. Other controls are not statistically significant. (Please Insert Table 18 Here) Another Estimation Method The fixed-effects (within) estimates in Table 19 corroborate the baseline GLS findings: DCB (degree centrality) has a positive linear coefficient while DCB² is negative and significant for ETR and CETR, supporting the inverted-U relationship between board centrality and tax avoidance. Control coefficients behave as expected (e.g., higher ROA and R&D associated with lower measured tax rates; higher LEV associated with higher tax rates). The pattern for CAETR is directionally consistent. These FE results strengthen confidence that the inverted-U is not driven by time-invariant firm heterogeneity. (Please Insert Table 19 Here) 6. Conclusion Our findings demonstrate an inverted-U relationship between director network centrality and corporate tax avoidance in Taiwan. This evidence highlights a context-dependent dynamic in which moderate connectivity appears to improve oversight while excessive connectivity facilitates sophisticated tax planning strategies. This study investigates the nuanced and nonlinear influence of director network centrality on corporate tax avoidance within Taiwan's emerging market. Building upon social capital and network theories, our findings reveal an inverted-U relationship: moderate levels of board connectivity are associated with reduced tax avoidance, while excessively interconnected boards facilitate more aggressive tax strategies. This outcome reinforces the dual role of social capital in corporate governance, a double-edged sword that can both constrain and enable opportunistic financial behaviors (Adler & Kwon, 2002 ; Fracassi & Tate, 2012 ). Our empirical analysis reveals several key findings. Drawing on a balanced panel of 3,422 firm-year observations from Taiwan between 2012 and 2023, we find that the relationship between director network centrality and corporate tax avoidance is both statistically significant and nonlinear. Specifically, we observe that firms with moderately connected boards exhibit higher effective tax rates, indicating less tax avoidance, while firms with highly connected boards display lower effective tax rates, suggesting a shift toward more aggressive tax behavior. This inverted-U shaped relationship is robust across three different tax avoidance measures (ETR, CETR, and CAETR) and remains stable across a range of sensitivity checks. These include Ramsey RESET and information criterion tests (AIC/BIC) for model specification, lagged regressions to address endogeneity, winsorized estimations to mitigate the influence of outliers, and subsample analyses excluding large-cap and technology sector firms. Finally, predictive margins plots with confidence intervals visually confirm the turning point in the relationship between director connectivity and tax behavior. The nonlinear tipping point is economically meaningful where it occurs but relatively uncommon in our sample: only a small share of firm-years exceed the centrality threshold (≈ 1.8%), though a nontrivial subset of firms (≈ 18%) experience high centrality at least once. Together, these results provide consistent and compelling evidence that board networks act as a dual governance mechanism, initially deterring and eventually enabling corporate tax avoidance. We also show theoretically that social capital has a non-monotonic governance effect: moderate connectivity raises reputational monitoring and reduces tax avoidance, but beyond a saturation point, social ties become a resource for elite coordination and enable sophisticated tax planning. Importantly, we frame this contribution as conditional rather than universal: the inverted-U is most plausibly operative in institutional environments with concentrated ownership, strong informal networks, and weaker enforcement. We therefore propose that governance theory incorporates such boundary conditions, and we encourage replication in other markets, both emerging and developed, to map the external validity of the mechanism. Our results contribute to theory by demonstrating that network effects on firm behavior are not linear, but threshold-dependent. This supports a growing view in governance literature that social capital is dual in nature, capable of both enforcing discipline and enabling coordination depending on context and saturation (Tao et al., 2019 ; Cheng et al., 2019 ). Furthermore, by using multiple centrality metrics and a richer specification, we show that board influence operates through positional dynamics, not just interlock counts. For boards and investors, the findings imply that connectivity is not universally beneficial. While moderate network integration can improve access to strategic knowledge and reputational discipline, over-networking poses governance risks. Institutional investors, rating agencies, and proxy advisors should incorporate network diagnostics into governance assessments, particularly in markets where social ties may dominate formal oversight. In emerging economies, where regulatory frameworks and audit capacities are still developing, director network structures should be monitored as part of tax enforcement strategies. Regulatory reforms might include mandatory disclosures of board network affiliations, automated risk flagging based on centrality, or thresholds beyond which additional scrutiny is triggered. This ensures that social capital does not quietly become a structural enabler of tax opacity. Our findings also have implications beyond Taiwan. Many emerging markets, including India, South Africa, Brazil, Indonesia, Mexico, and Turkey, share institutional features that parallel Taiwan’s governance environment: high ownership concentration, reliance on informal director networks, and institutional voids in monitoring and enforcement. In such contexts, the inverted-U relationship between board centrality and tax avoidance provides a useful diagnostic for understanding how social capital simultaneously disciplines and enables corporate behavior. Even in developed economies with club-like director elites (e.g., Italy, South Korea), similar saturation effects may hold. By identifying the broader relevance of our framework, we underline the contribution of this study not only to Taiwanese corporate governance but also to comparative research on emerging markets. Although our analysis is based on Taiwanese firms, the theoretical insights are relevant for broader emerging-market contexts. In countries where ownership is concentrated and informal networks play a significant role, the nonlinear duality of director networks—as both disciplinary and enabling, may similarly shape corporate tax behavior. We encourage future research to test the inverted-U relationship in other institutional settings to assess the boundary conditions and external validity of our findings. Future studies could examine sector-specific network effects, dynamic shifts in board centrality over time, or interactions between centrality and other governance mechanisms like audit committee strength or ownership type. There is also scope for exploring cross-border director ties, especially in multinational firms, and applying graph-theoretic or machine learning methods to extract latent network patterns. In conclusion, this paper highlights the dual-edged nature of board networks in corporate tax governance. While social ties can enhance oversight and transparency, excessive connectivity may cross a tipping point, enabling strategic and potentially aggressive financial behavior. This inverted-U relationship offers both a diagnostic tool and a theoretical contribution to the literature on governance in emerging markets. As global financial environments become more interconnected, a deeper understanding of social architecture within boards will be crucial for ensuring accountability, transparency, and sustainable capital markets. Declarations Author Contribution Author A：Conceptualization, Methodology, Formal analysis, Investigation, Resources, Writing-original draft, Writing-review &Editing.Author B： Conceptualization, Writing- review & editing, Supervision. References Adler PS, Kwon S-W (2002) Social capital: Prospects for a new concept. Acad Manage Rev 27(1):17–40 Adler PS, Kwon S-W (2023) Social Capital: Prospects for a New Concept Allen A, Francis BB, Wu Q, Zhao Y (2016) Analyst coverage and corporate tax aggressiveness. J Banking Finance 73:84–98 Andersen A, Garel A, Gilbert A, Tourani-Rad A (2022) Social capital, human capital, and board appointments. Global Finance Journal, 54, p.100758 Aparicio K, Kim R (2023) External capital market frictions, corporate governance, and tax avoidance: Evidence from the TED spread. Finance Res Lett 52:103381 Armstrong CS, Blouin JL, Jagolinzer AD, Larcker DF (2015) Corporate governance, incentives, and tax avoidance. J Account Econ 60:1–17 Armstrong CS, Blouin JL, Larcker DF (2012) The incentives for tax planning. J Account Econ 53(1–2):391–411 Arora L, Gill A (2022) Corporate governance and tax avoidance: New evidence from emerging markets. J Acc Emerg Economies 12(3):471–495 Arora TS, Gill S (2022) Impact of corporate tax aggressiveness on firm value: evidence from India. Managerial Finance 48(2):313–333 Badertscher BA, Katz SP, Rego SO (2013) The separation of ownership and control and corporate tax avoidance. J Account Econ 56(2–3):228–250 Bertrand M, Mullainathan S (2003) Enjoying the quiet life? Corporate governance and managerial preferences. J Polit Econ 111(5):1043–1075 Bizjak J, Lemmon M, Whitby R (2009) Option backdating and board interlocks. Rev Financial Stud 22(11):4821–4847 Burnham KP, Anderson DR (2002) Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach, 2nd edn. Springer Cai Y, Sevilir M (2012) Board connections and M&A transactions. J Financ Econ 103(2):327–349 Cai Y, Dhaliwal DS, Kim Y, Pan C (2014) Board interlocks and the diffusion of disclosure policy. Rev Acc Stud 19:1086–1119 Campbell JL, Guan JX, Li OZ, Zheng Z (2020) CEO severance pay and corporate tax planning. J Am Taxation Association 42(2):1–27 Chen S, Chen X, Cheng Q, Shevlin T (2010) Are family firms more tax aggressive than non-family firms? J Financ Econ 95(1):41–61 Chen Y, Huang J, Li X, Ni X (2023) Financial market opening and corporate tax avoidance: Evidence from staggered quasi-natural experiments. Finance Res Lett 54:103770 Cheng CA, Dhaliwal DS, Zhang Y (2019) Does interlocking directorate shape unethical accounting behavior? Contemp Acc Res 36(4):2141–2171 Cheng S, Felix R, Zhao Y (2019) Board interlock networks and informed short sales. J Banking Finance 98:198–211 Desai MA, Dharmapala D (2009) Corporate tax avoidance and firm value. Rev Econ Stat 91(3):537–546 Desai MA, Dharmapala D (2006) Corporate tax avoidance and high-powered incentives. J Financ Econ 79(1):145–179 Dyreng SD, Hanlon M, Maydew EL, Thornock JR (2017) Changes in corporate effective tax rates over the past 25 years. J Financ Econ 124(3):441–463 Dyreng SD, Hanlon M, Maydew EL (2008) Long- Run Corporate Tax Avoidance. Acc Rev 83:61–82 El-Khatib R, Fogel K, Jandik T (2015) CEO network centrality and merger performance. J Financ Econ 116(2):349–382 Fan JPH, Wong TJ, Zhang T (2007) Politically connected CEOs, corporate governance, and post-IPO performance of China's newly partially privatized firms. J Financ Econ 84(2):330–357 Ferris SP, Javakhadze D, Rajkovic T (2017) CEO social capital, risk-taking, and corporate policies. J Corp Finance 45:289–306 Filatotchev I, Chahine S, Bruton GD (2018) Board interlocks and initial public offering performance in the United States and the United Kingdom: An institutional perspective. J Manag 44(4):1620–1650 Fracassi C, Tate G (2012) External networking and internal firm governance. J Finance 67(1):153–194 Frank M, Margaret LJ, Lynch, Sonja O, Rego (2009) Tax Reporting Aggressiveness and its Relation to Aggressive Financial Reporting. The Accounting Review 84 Gaaya S, Lakhal N, Lakhal F (2017) Does family ownership reduce corporate tax avoidance? The moderating effect of audit quality. Managerial Auditing J 32:731–744 Gao L, Yang LL, Zhang JH (2016) Corporate patents, R&D success, and tax avoidance. Rev Quant Financ Acc 47:1063–1096 Godigbe BG, Chui CM, Liu CL (2018) Directors network centrality and earnings quality. Appl Econ 50(50):5381–5400 Goncalves AF, Rossoni L, Mendes-Da-Silva W (2019) Board social capital reduces implied cost of capital for private companies but not of state-owned companies. Manag Decis 57(10):2672–2692 Halioui K, Neifar S, Ben Abdelaziz F (2016) Corporate governance, CEO compensation and tax aggressiveness: Evidence from American firms listed on the NASDAQ s100. Rev Acc Finance 15(4):445–462 Han J, Bose I, Hu N, Qi B, Tian G (2015) Does director interlock impact corporate R&D investment? Decis Support Syst 71:28–36 Hanlon M, Heitzman S (2010) A review of tax research. J Account Econ 50(2–3):127–178 Higgins DM, Omer TC, Phillips JD (2015) The influence of a firm's business strategy on its tax aggressiveness. Contemp Acc Res 32(2):674–702 Howard MD, Withers MC, Tihanyi L (2017) Knowledge dependence and the formation of director interlocks. Acad Manag J 60(5):1986–2013 Jarboui A, Kachouri Ben Saad M, Riguen R (2020) Tax avoidance: do board gender diversity and sustainability performance make a difference? J Financial Crime 27(4):1389–1408 Kale P (2019) Tax strategy and the business landscape in Taiwan. Asian J Acc Res 4(2):90–109 Kale S (2019) Taxing times: compliance complexity a big challenge for businesses. Financial Ex Kanagaretnam K, Lee J, Lim CY, Lobo GJ (2016) Relation between Auditor Quality and Tax Aggressiveness: Implications of Cross-Country Institutional Differences. AUDITING: J Pract Theory 35:105–135 Khurana IK, Moser WJ (2013) Institutional Shareholders’ Investment Horizons and Tax Avoidance. J Am Taxation Association 35:111–134 Lanis R, Richardson G (2011) The effect of board of director composition on corporate tax aggressiveness. J Account Public Policy 30:50–70 Lanis R, Richardson G, Taylor G (2017) Board of Director Gender and Corporate Tax Aggressiveness: An Empirical Analysis. J Bus Ethics 144:577–596 Larcker DF, So EC, Wang CC (2013) Boardroom centrality and firm performance. J Account Econ 55(2–3):225–250 Li D, Jiang Q, Mai Y (2019) Board interlocks and capital structure dynamics: evidence from China, vol 59. Accounting & Finance, pp 1893–1922 Lin S, Tong N, Tucker AL (2014) Corporate tax aggression and debt. J Banking Finance 40:227–241 Manzon G, Plesko GA (2002) The Relation Between Financial and Tax Reporting Measures of Income. Law Rev 55:175–214 Masulis RW, Mobbs S (2014) Independent director incentives: Where do talented directors spend their limited time and energy? J Financ Econ 111(2):406–429 McGuire ST, Wang D, Wilson RJ (2014) Dual Class Ownership and Tax Avoidance. Account Rev 89:1487–1516 Minnick K, Noga T (2010) Do corporate governance characteristics influence tax management? J Corp Finance 16:703–718 Moore JA (2012) Empirical evidence on the impact of external monitoring on book–tax differences. Adv Acc 28(2):254–269 OECD (2020) Corporate tax statistics OECD (2020) Corporate Tax Statistics: Second Edition. Organisation for Economic Co-operation and Development Rajagopalan N, Zhang Y (2008) Corporate governance reforms in China and India: Challenges and opportunities. Bus Horiz 51(1):55–64 Ramsey JB (1969) Tests for specification errors in classical linear least-squares regression analysis. J Royal Stat Soc Ser B (Methodological) 31(2):350–371 Rego SO (2003) Tax-Avoidance Activities of U.S. Multinational Corporations. Contemp Acc Res 20:805–833 Richardson G, Lanis R, Leung SCM (2014) Corporate tax aggressiveness, outside directors, and debt policy: An empirical analysis. J Corp Finance 25:107–121 Richardson G, Taylor G, Lanis R (2013) The impact of board of director oversight characteristics on corporate tax aggressiveness: An empirical analysis. J Account Public Policy 32(3):68–88 Richardson G, Taylor G, Lanis R (2015) The impact of financial distress on corporate tax avoidance spanning the global financial crisis: Evidence from Australia, vol 44. Economic Modelling, pp 44–53 Riguen R, Salhi B, Jarboui A (2020) Do women in board represent less corporate tax avoidance? A moderation analysis. Int J Sociol Soc Policy 40(1/2):114–132 Tao Q, Li H, Wu Q, Zhang T, Zhu Y (2019) The dark side of board network centrality: Evidence from merger performance. J Bus Res 104:215–232 Tsai L-C, Zhang R, Zhao C (2019) Political connections, network centrality and firm innovation. Finance Res Lett 28:180–184 Footnotes We measure tax outcomes as proportions (tax / pretax income): ETR = total tax / pretax income; CETR = current tax / pretax income; CAETR = cash taxes paid / pretax income. Lower values of each measure indicate greater tax avoidance (i.e., a smaller tax payment relative to pretax income). Because measurement conventions vary across studies, we also report results using an alternative measure — the book–tax difference (BTD) — and a binary indicator of aggressive tax positions; these robustness checks (Section 5, Table A-x) produce the same inverted-U relationship between board centrality and tax avoidance. We note that all reported tax-rate coefficients are interpreted such that a negative association implies more tax avoidance (lower tax paid relative to pretax income); see Variable Definitions and Appendix robustness tests for the alternative BTD specification. Tables Table 1 to 19 are available in the Supplementary Files section. Additional Declarations No competing interests reported. Supplementary Files Appendix.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9196981","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":628457862,"identity":"c6e7c77b-a332-4b4d-a74f-9807b4428ad3","order_by":0,"name":"Ahmed Bouteska","email":"","orcid":"","institution":"Louisiana State University","correspondingAuthor":false,"prefix":"","firstName":"Ahmed","middleName":"","lastName":"Bouteska","suffix":""},{"id":628457863,"identity":"8eab9126-41bb-48aa-9af4-e51cf4f4859c","order_by":1,"name":"Shikuan Zhao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0UlEQVRIiWNgGAWjYFACNhAhIcfGzHzgwIcfxGuxMeZnZ0s8OLOHeC1piTP7eYwPc7ARoUG+/1jih7dth40NDvN8OMzAwyDPL3YAvxbGGWmHJee2HZYzOMy74XCBBYPhzNkJ+LUwS7C3MfOCbQFqmcHDkGBwm4AWNv7jYC2JGw7zPDjMw0aEFh6GtGNALUDvN/MwEKdFQiItWXLOOWAgM7MZAANZgrBfgCFm+OFNGTAq+Q8//vDhh408vzQBLVDXIWwlQjmallEwCkbBKBgFmAAA58xBlqO7kuIAAAAASUVORK5CYII=","orcid":"","institution":"Zhejiang University of Technology","correspondingAuthor":true,"prefix":"","firstName":"Shikuan","middleName":"","lastName":"Zhao","suffix":""}],"badges":[],"createdAt":"2026-03-23 07:38:16","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9196981/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9196981/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108414108,"identity":"7f0e85bd-1d7f-453d-81cf-87e6658095ea","added_by":"auto","created_at":"2026-05-04 10:55:35","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1008393,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePredictive Margins Plot of Board Centrality and Corporate Tax Avoidance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNote. This table reports the estimation results of impact board centrality on corporate tax avoidance. We use the generalized least square regression excluding firms from the Technology Sector. Refer to Appendix table 1 for complete variable definitions. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively; t-statistics are reported in parentheses. The results are based on a sample of 3422 firm year observations from 2012 to 2023.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-9196981/v1/be88f01fe994698641849dae.png"},{"id":108803733,"identity":"7020b6ef-afc8-4ee9-b137-d771abd13bbe","added_by":"auto","created_at":"2026-05-08 15:05:06","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":993429,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9196981/v1/f6b5613f-e630-4897-b29b-c0c11c811cdb.pdf"},{"id":108414121,"identity":"0afab0fb-e048-4e01-bff6-8669ab981c39","added_by":"auto","created_at":"2026-05-04 10:55:44","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":123052,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix.docx","url":"https://assets-eu.researchsquare.com/files/rs-9196981/v1/5d1b4e2acf46b4cfdfc95a47.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Director Network Centrality and Corporate Tax Avoidance","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eTaxes represent one of the most significant and persistent costs for corporations and are often perceived as a constraint on profitability (Chen et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). In response, firms increasingly engage in tax minimization strategies to enhance after-tax income, among them, tax avoidance, which involves exploiting legal gray areas, manipulating tax planning tools, or leveraging complex financial structures to reduce tax liabilities. According to Dyreng et al. (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2008\u003c/span\u003e), corporate tax avoidance refers to any strategic action that lowers a firm's tax obligations relative to its pre-tax earnings. Although tax avoidance can improve cash flows and shareholder value in the short term, it also introduces reputational and regulatory risks (Armstrong et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), especially when it borders on aggressive or opaque behavior.\u003c/p\u003e \u003cp\u003ePrior studies reveal that tax avoidance is shaped by various firm-specific and governance-related characteristics, such as size, leverage, profitability (Manzon and Plesko, \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2002\u003c/span\u003e), capital intensity (Rego, \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2003\u003c/span\u003e), board composition (Minnick and Noga, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Lanis and Richardson, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2011\u003c/span\u003e), audit quality (Kanagaretnam et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Gaaya et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), institutional ownership (Khurana and Moser, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2013\u003c/span\u003e), and external financing pressures (Aparicio and Kim, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). More recently, researchers have begun to explore how the social and structural positions of directors\u0026mdash;particularly their embeddedness in interlocking board networks\u0026mdash;may shape corporate risk-taking and ethical behavior (Ferris et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). However, the specific influence of director network centrality on tax avoidance remains largely unexplored, despite its theoretical relevance to both corporate governance and financial strategy.\u003c/p\u003e \u003cp\u003eDrawing on social capital theory and the integrated agency-resource dependence perspective, our study investigates whether director connectivity reduces or encourages corporate tax avoidance. Prior literature offers two contrasting predictions. On the one hand, directors who are well-networked may be more constrained by reputational concerns and social scrutiny, thereby deterring them from engaging in controversial tax strategies (Bertrand and Mullainathan, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Masulis and Mobbs, \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). On the other hand, central directors may possess privileged access to information, technical expertise, and peer-driven strategies, potentially enabling more sophisticated and aggressive forms of tax planning (Desai and Dharmapala, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Fracassi and Tate, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2012\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis study contributes to a growing literature that links board networks to corporate financial decision-making, including innovation (Han et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), capital access (Li et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Goncalves et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), and post-merger performance (El-Khatib et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). In contrast to prior studies that often assume a linear relationship between governance mechanisms and outcomes, we propose and empirically demonstrate that the link between director network centrality and tax avoidance is nonlinear and exhibits an inverted-U shape. Specifically, we show that firms with moderately connected directors experience lower levels of tax avoidance, likely due to enhanced board oversight and reputational constraints. However, beyond a certain threshold, higher centrality correlates with increased tax avoidance, suggesting that social capital may transform into a vehicle for strategic opportunism.\u003c/p\u003e \u003cp\u003eTo test this hypothesis, we analyze a panel of 3,422 firm-year observations from 313 firms listed on the Taiwan Stock Exchange between 2012 and 2023. Taiwan represents a particularly rich empirical context: it is an emerging market with high tax rates (OECD, \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), dense informal networks, concentrated ownership, and institutional gaps in regulatory enforcement (Rajagopalan and Zhang, \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Kale, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). These features make it an ideal case for understanding the dual role of director networks, as both monitors and enablers, in shaping corporate tax behavior.\u003c/p\u003e \u003cp\u003eWhile our empirical setting is Taiwan, we argue that the findings offer insights applicable to a broader set of emerging markets with comparable institutional profiles. For instance, India, Brazil, Indonesia, Mexico, South Africa, and Turkey exhibit three features that mirror Taiwan\u0026rsquo;s context: (i) ownership concentration and prevalence of family-controlled firms (La Porta et al., 1999; Khanna \u0026amp; Yafeh, 2007), (ii) dense informal networks and relational governance that substitute for underdeveloped formal institutions (Khanna \u0026amp; Palepu, 2010), and (iii) regulatory and enforcement gaps that create opportunities for opacity in corporate behavior (Fan, Wong, \u0026amp; Zhang, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Aguilera \u0026amp; Crespi-Cladera, 2016). In these markets, directors often operate within elite networks that simultaneously constrain and enable corporate strategies. Hence, Taiwan serves not only as a theoretically rich case but also as a proxy for other emerging economies where relational governance plays a dominant role.\u003c/p\u003e \u003cp\u003eOur contribution is threefold. First, we extend the literature on tax avoidance by introducing network centrality as a dynamic, nonlinear determinant of tax behavior, going beyond standard governance variables. Second, we provide methodological rigor through multiple centrality measures, lagged regressions, and formal nonlinearity tests (RESET, AIC/BIC, Lind-Mehlum slope tests), ensuring robustness and theoretical validity. Third, we highlight the global relevance of director social capital by anchoring our findings within the context of emerging-market governance and regulatory design.\u003c/p\u003e \u003cp\u003eFinally, we emphasize that our empirical claims are context-sensitive. Taiwan\u0026rsquo;s institutional architecture, high ownership concentration, pronounced informal director networks, and uneven enforcement dynamics, provides a particularly informative testbed for network effects on tax policy. Accordingly, while our inverted-U finding identifies an important boundary condition for governance theory, the interpretation should be conditional: the mechanism we identify is most likely to operate in emerging markets or other settings where informal social capital substitutes for formal controls. We explicitly discuss external validity and provide robustness diagnostics and interaction tests to illustrate the heterogeneity of effects across institutional settings.\u003c/p\u003e \u003cp\u003eThe remainder of the paper is structured as follows. Section 2 reviews the relevant literature and provides the institutional background. Section 3 outlines the data, variables, and empirical methodology. Section 4 presents the baseline regression results and the inverted-U analysis. Section 5 includes robustness checks and graphical validations. Section 6 concludes the paper and discusses theoretical, policy, and cross-national implications.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e"},{"header":"2. Literature Review and Institutional Background","content":"\u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003eThe structure, behavior, and influence of board networks have drawn increasing scholarly interest in recent years, particularly as they relate to corporate ethics and financial discretion. Directors embedded in powerful networks often bring informational advantages, resource access, and reputational constraints that shape firm behavior in complex ways. From the standpoint of labor market dynamics, well-connected directors are actively sought after by companies due to their social capital and access to privileged knowledge (Andersen et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). In turn, these directors are incentivized to protect their reputational capital by avoiding associations with unethical or high-risk firms. As Masulis and Mobbs (\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2014\u003c/span\u003e) note, directors who become linked to scandal-tainted companies face reduced chances of future board appointments and diminished compensation.\u003c/p\u003e\n \u003cp\u003eFrom this perspective, board centrality serves a disciplinary function: directors with high centrality may exercise greater oversight, deter opportunistic management behavior, and safeguard firm reputation. This interpretation is supported by the \u0026quot;quiet life\u0026quot; hypothesis (Bertrand and Mullainathan, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2003\u003c/span\u003e), which suggests that powerful directors with secure career trajectories prefer to avoid risky or aggressive practices, such as corporate tax avoidance. Consequently, one would expect that firms governed by highly central directors would demonstrate lower levels of tax aggressiveness.\u003c/p\u003e\n \u003cp\u003eHowever, emerging insights from the agency-resource dependence framework complicate this view. As Tao et al. (\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) argue, networked directors may simultaneously become channels for transmitting opaque strategies and information asymmetries. In such settings, board interlocks may no longer serve as monitors but rather as vehicles for normalizing opportunistic practices. Peer influence among elite networks can facilitate the diffusion of tax planning tactics across firms (Fracassi and Tate, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), including strategies that border on artificial tax avoidance. Cheng et al. (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) document that such networks have been instrumental in the dissemination of unethical practices such as earnings manipulation and options backdating. These findings suggest a dual role of board centrality, initially constraining but eventually enabling opportunistic behavior as influence and insulation grow.\u003c/p\u003e\n \u003cp\u003eThis duality becomes particularly salient in the context of corporate tax avoidance. As Desai and Dharmapala (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2006\u003c/span\u003e) explain, tax avoidance often requires both \u0026ldquo;elaborateness\u0026rdquo; and \u0026ldquo;opacity\u0026rdquo;\u0026mdash;features that thrive in environments of weak monitoring and strong insider access. When networks become too dense or powerful, the same directors who once imposed constraints may instead facilitate access to elite tax professionals, regulatory arbitrage opportunities, and high-level tax shelters. Hence, we posit that the relationship between director network centrality and tax avoidance is nonlinear and inverted-U shaped: social capital constrains behavior up to a threshold, after which it begins to enable strategic exploitation.\u003c/p\u003e\n \u003cp\u003eDespite the expanding literature on tax avoidance, much of the evidence stems from developed economies with formalized regulatory systems (Hanlon \u0026amp; Heitzman, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Richardson et al., \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). In contrast, emerging markets remain underexplored, even though they present distinctive features such as informal enforcement systems, concentrated ownership structures, and dependence on relational governance. Taiwan, in this regard, provides a theoretically rich and policy-relevant context. With one of the highest statutory corporate tax rates globally (OECD, \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), a complex tax code (Kale, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), and institutional voids in transparency and enforcement (Rajagopalan and Zhang, \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2008\u003c/span\u003e), Taiwan mirrors many challenges faced by emerging markets.\u003c/p\u003e\n \u003cp\u003eBoard structures in Taiwanese firms often serve as rubber stamps for controlling shareholders, and directors frequently hold positions across multiple firms, creating dense informal networks. These networks substitute for underdeveloped formal institutions and are often the primary source of strategic coordination (Fan, Wong, \u0026amp; Zhang, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). Therefore, Taiwan is not only a relevant empirical setting\u0026mdash;it is also a proxy case for other network-dependent governance environments across Asia, Latin America, and Africa.\u003c/p\u003e\n \u003cp\u003eThis study addresses the critical gap in understanding how director network centrality influences tax behavior in such environments. Using panel data from 313 firms over 12 years (2012\u0026ndash;2023), we empirically test the hypothesis that board centrality first reduces, then increases corporate tax avoidance. By deploying multiple centrality metrics, degree, closeness, betweenness, and eigenvector, and three widely accepted tax avoidance proxies (ETR, CETR, CAETR), we uncover robust evidence of an inverted-U shaped relationship. Our findings are validated through multiple robustness tests, including nonlinear specification checks, lagged regressions, and graphical modeling. The results suggest that director networks serve as a double-edged sword: they initially operate as mechanisms of discipline, but beyond a threshold, they evolve into enablers of sophisticated tax planning.\u003c/p\u003e\n\u003c/div\u003e\n\u003cp\u003e\u003cstrong\u003eSocial-capital saturation: micro-mechanisms for an inverted-U\u003c/strong\u003e\u003c/p\u003e\n\u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003eWe reconcile the seemingly contradictory forces of reputational discipline and privileged access by introducing a social-capital saturation argument. Three linked micro-mechanisms generate an inverted-U relationship between board centrality and tax avoidance. First, visibility and reputational monitoring (low\u0026rarr;moderate centrality). Directors with a small to moderate number of interlocks serve as bridges across firms, increasing public visibility, peer scrutiny, and the likelihood that questionable choices will be observed and sanctioned. Because reputational penalties are salient, modest increases in centrality strengthen informal monitoring and reduce incentives for aggressive tax strategies. Second, redundancy, brokerage power, and elite capture (high centrality). As centrality grows beyond a threshold, directors accumulate redundant ties, increasing brokerage power and private informational access (technical know-how, advisor networks, insider flows). Redundant ties reduce exposure to independent observers, and dense, reciprocal relationships normalize risk-taking. Brokerage and elite membership also facilitate coordinated strategies that are difficult for external monitors to detect, effectively enabling sophisticated tax planning. Third, conditional moderators and dynamic feedback. The switch from constraint to facilitation is more likely where formal enforcement is weak, ownership is concentrated, and audit or market monitoring is limited. Conversely, strong third-party oversight raises the cost of opportunism and shifts the tipping point. Repeated appointments and network reinforcement can lock in clique norms, making the enabling effect persistent.\u003c/p\u003e\n \u003cp\u003eThese mechanisms yield the theoretical prediction that the relationship between director network centrality and tax avoidance is nonlinear and inverted-U shaped, with moderate connectivity constraining tax aggressiveness and high connectivity enabling sophisticated tax planning.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"3. Sample and Methodology","content":"\u003cp\u003e\u003cstrong\u003eData and Sample Selection\u003c/strong\u003e\u003c/p\u003e\n\u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003eThis study is based on an unbalanced panel dataset of 313 non-financial, non-utility companies listed on the Taiwan Stock Exchange (TWSE) between 2012 and 2023, yielding 3,422 firm-year observations. Financial and utility firms are excluded due to their distinct tax regulations, disclosure requirements, and capital structures (similar to Hanlon and Heitzman, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). Observations with negative pretax income are also excluded, as effective tax rate (ETR) measures are undefined or distorted in such cases. We construct our director network measures using board interlock data derived from Bloomberg, which maps the affiliations of each director across firms and years. Governance and firm-level financial data, including board structure, tax figures, and accounting metrics, are extracted from the Taiwan Economic Journal (TEJ) database, consistent with prior studies on Taiwanese corporate governance (Tao et al., \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Kale, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). The final sample enables a comprehensive examination of the relationship between board network centrality and tax behavior in a developing market context characterized by complex tax structures and network-dependent governance.\u003c/p\u003e\n\u003c/div\u003e\n\u003cp\u003e\u003cstrong\u003eSample Considerations and Institutional Context\u003c/strong\u003e\u003c/p\u003e\n\u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003eWhile our empirical setting is Taiwan, this context provides a theoretically informative case for understanding board networks in emerging markets. Taiwanese firms typically exhibit concentrated ownership, dense interlocking director networks, and partially informal enforcement systems. These characteristics make Taiwan an ideal laboratory for examining how director social capital influences tax avoidance. We note, however, that the findings may not translate directly to developed markets with formalized governance structures. Nonetheless, the mechanisms uncovered, moderate connectivity fostering reputational monitoring and high connectivity enabling strategic coordination, are likely applicable in other emerging markets with similar relational governance dynamics, such as parts of Latin America, Africa, and Asia.\u003c/p\u003e\n\u003c/div\u003e\n\u003cp\u003e\u003cstrong\u003eEmpirical Models and Variable Definitions\u003c/strong\u003e\u003c/p\u003e\n\u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003eTo examine the influence of director networks on corporate tax avoidance, we estimate the following baseline panel regression model using Generalized Least Squares (GLS) with robust standard errors clustered at the firm level to correct for intra-firm autocorrelation and heteroskedasticity:\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\n \u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e$$\\:{TA}_{it}\\:=\\:{\\alpha\\:}_{it}+{{\\beta\\:}_{1}CB}_{it}+\\:{{\\beta\\:}_{2}CB}_{it}^{2}\\:+\\gamma\\:{X}_{it}+{\\mu\\:}_{i}+{\\lambda\\:}_{t}+{\\epsilon\\:}_{it}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eWhere, the Tax avoidance (TA) is the tax avoidance proxy, measured alternately using ETR, CETR, and CAETR. CB represents centrality metrics (DCB, CCB, BCB, ECB). CB\u0026sup2; captures nonlinearity (quadratic specification) to test for the inverted-U relationship. X is a vector of firm- and board-level control variables. \u0026micro; and \u0026lambda; are firm and year fixed effects, respectively. \u0026epsilon; is the error term.\u003c/p\u003e\u003cp\u003eThis specification is estimated separately for each centrality type to mitigate multicollinearity concerns between network metrics. Following recent research (Arora and Gill, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Aparicio and Kim, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), we also estimate lagged specifications in robustness tests to address endogeneity and reverse causality concerns.\u003c/p\u003e\u003cp\u003eOur baseline specification is estimated using Generalized Least Squares (GLS) with firm-clustered robust standard errors. We adopt GLS because tax avoidance measures exhibit both cross-sectional heteroskedasticity and within-firm serial correlation, which GLS efficiently accommodates. Moreover, we retain firm-fixed and year-fixed effects within GLS, ensuring control for time-invariant firm traits and common shocks. This estimator has precedent in the governance literature examining director networks and firm policies in emerging markets (Fracassi \u0026amp; Tate, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Cheng et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Arora \u0026amp; Gill, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eTo address concerns that results may depend on GLS, we also re-estimate our models using fixed-effects OLS, dynamic system-GMM, IV-2SLS, propensity-score matched regressions, and inverse-probability weighting. The inverted-U relationship persists across all methods (see Tables \u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e16\u003c/span\u003e and \u003cspan refid=\"Tab19\" class=\"InternalRef\"\u003eA1\u003c/span\u003e). Thus, while GLS provides efficiency gains under our data structure, our substantive conclusions are not sensitive to the estimation strategy.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eIn line with prior studies (Frank et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Dyreng et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Minnick and Noga, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2010\u003c/span\u003e), we employ three distinct measures of effective tax rates: ETR: Total tax expense divided by pretax income; CETR: Current tax expense divided by pretax income; CAETR: Cash taxes paid divided by pretax income.\u003c/p\u003e\u003cp\u003eLower values of these measures indicate more aggressive tax avoidance\u003csuo\u003e1. These metrics are widely used in both developed and emerging market tax avoidance studies and offer a robust triangulation of firms\u0026apos; tax planning intensity.\u003c/suo\u003e\u003c/p\u003e\u003cp\u003eWe construct four centrality measures, normalized annually at the director level and aggregated to the board level: Degree Centrality (DCB): Number of direct ties a director maintains; Closeness Centrality (CCB): Inverse of the average distance to all other directors, capturing information access speed; Betweenness Centrality (BCB): Frequency with which a director lies on the shortest paths between others, indicating brokerage power; Eigenvector Centrality (ECB): Influence of a director based on both direct and indirect ties to highly connected individuals.\u003c/p\u003e\u003cp\u003eEach metric is squared to capture nonlinear (inverted-U) dynamics. Aggregation is performed across all board members, following Godigbe et al. (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) and Tsai et al. (\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eWe include the following controls, drawn from established literature (Armstrong et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Richardson et al., \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Jarboui et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e): SB: Number of board members; IDB: Ratio of independent directors; FDB: Ratio of female directors; PH: Promoter shareholding (% of total); SIZE: Log of total assets; ROA: Return on assets; LEV: Total debt to total assets; R\u0026amp;D: R\u0026amp;D expenditure to total assets; PPE: Net property, plant, and equipment to total assets; AGE: Years since firm incorporation; Year and firm fixed effects are included to control for macroeconomic shocks and time-invariant firm traits. The definitions of the variables are provided in Table\u0026nbsp;1 in appendix.\u003c/p\u003e\u003cp\u003eDirector-background controls: To control for potential omitted heterogeneity at the director level, we also construct four board-aggregated measures from director CVs and board membership records (aggregated to the firm\u0026ndash;year): (i) PCT_FINANCE, share of directors with prior financial/accounting experience (CFO, audit partner, investment banker, Big-4 auditor); (ii) PCT_POLITICAL, share of directors with political ties (former public office, political appointments, or close official affiliations); (iii) AVG_TENURE, average board tenure (years) of directors; and (iv) PCT_ADVDEG, share of directors with advanced business/technical degrees (MBA, MSc, PhD). These variables are added to the control vector (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{X}_{i,t}\\)\u003c/span\u003e\u003c/span\u003e) in all baseline, quadratic, and robustness specifications.\u003c/p\u003e\u003cp\u003e(Please Insert Table\u0026nbsp;1 Here)\u003c/p\u003e"},{"header":"4. Empirical Results","content":"\u003cp\u003e\u003cstrong\u003ePreliminary Results\u003c/strong\u003e\u003c/p\u003e\n\u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003eWe begin by presenting the descriptive statistics for the full sample in Table \u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e2\u003c/span\u003e. The mean values for the tax avoidance proxies, ETR (0.452), CETR (0.399), and CAETR (0.288), suggest substantial variability in effective tax burdens across firms. While the median tax rates (ETR: 44.9%, CETR: 38.9%, CAETR: 28.6 28.6%) are broadly consistent with Taiwan\u0026rsquo;s statutory corporate tax rate (ranging from 17% to 45% during the period), the extremely high standard deviations and wide min-max ranges indicate that a sizable subset of firms engage in aggressive tax strategies, consistent with prior findings by Chen et al. (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2010\u003c/span\u003e) and Arora and Gill (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\n\u003c/div\u003e\n\u003cp\u003eThe dispersion across the centrality measures, degree centrality (DCB), closeness centrality (CCB), eigenvector centrality (ECB), and betweenness centrality (BCB), also highlights significant heterogeneity in network structures across boards. For example, the standard deviation of BCB is large (4,793.84), reflecting considerable differences in strategic positioning within board networks. These descriptive statistics suggest a ripe setting for testing nonlinear patterns of influence across firms and over time.\u003c/p\u003e\n\u003cp\u003eTable \u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e3\u003c/span\u003e presents the Pearson correlation coefficients for all primary variables. As expected, the tax avoidance proxies (ETR, CETR, CAETR) are positively correlated with one another (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) but not perfectly collinear (ETR\u0026ndash;CETR\u0026thinsp;\u0026asymp;\u0026thinsp;0.01; ETR\u0026ndash;CAETR \u0026asymp; \u0026minus;\u0026thinsp;0.01, affirming that they capture overlapping dimensions of tax behavior. Importantly, the correlations between the centrality variables and tax avoidance metrics are low to moderate in magnitude, suggesting that network centrality explains variation in tax behavior beyond firm fundamentals. The observed relationships also preliminarily support a nonlinear effect, with weak or mixed signs across linear terms, consistent with our later hypothesis of an inverted-U pattern. The centrality measures also are modestly correlated with firm characteristics, a pattern consistent with the multivariate regression results.\u003c/p\u003e\n\u003cp\u003eWe also compute variance inflation factors (VIFs) for all predictors to test for multicollinearity. All VIF values are well below the commonly accepted threshold of 5 (maximum VIF\u0026thinsp;=\u0026thinsp;1.64), confirming the absence of multicollinearity and strengthening the reliability of our regression estimates. These diagnostic checks validate the soundness of our empirical model and variable construction.\u003c/p\u003e\n\u003cp\u003eTaken together, the preliminary analysis provides early evidence that (1) there is significant variation in both tax avoidance and network centrality across firms, (2) the relationships are likely more complex than linearity would suggest, and (3) our data structure is robust enough for advanced multivariate modeling, including the non-linear GLS regressions presented in the next section.\u003c/p\u003e\n\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e2\u003c/span\u003e Here)\u003c/p\u003e\n\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e3\u003c/span\u003e Here)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBaseline Results\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTable \u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents the results of our baseline GLS regressions, estimating the relationship between director network centrality and corporate tax avoidance using three measures: ETR, CETR, and CAETR\u003csuo\u003e2. All models include firm-fixed effects and year-fixed effects, with robust standard errors clustered at the firm level.\u003c/suo\u003e\n\u003c/p\u003e\n\u003cp\u003eWe find that degree centrality (DCB) is positively and significantly associated with each tax rate measure (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01), suggesting that as network centrality increases, firms report higher effective tax rates\u0026mdash;implying less tax avoidance. This is consistent with the argument that well-connected directors have stronger reputational incentives to avoid risky or unethical behavior (Masulis \u0026amp; Mobbs, \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), and may exert greater oversight in boardrooms. These results align with the \u0026quot;quiet life\u0026quot; hypothesis (Bertrand \u0026amp; Mullainathan, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2003\u003c/span\u003e), which posits that influential directors are less inclined to engage in complex, controversial strategies such as aggressive tax avoidance.\u003c/p\u003e\n\u003cp\u003eControl variables perform as expected. Larger firms (SIZE) are more likely to avoid taxes, consistent with prior literature that links organizational complexity with aggressive planning (Richardson et al., \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Higher profitability (ROA) is associated with greater tax avoidance, which may reflect firms in emerging markets allocating surplus resources toward financial optimization. Leverage (LEV) also shows a significant positive correlation with ETR and CETR, indicating reduced tax avoidance, consistent with Badertscher et al. (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). R\u0026amp;D intensity (R\u0026amp;D) is negatively associated with tax rates, supporting the idea that firms with intangible-heavy business models may exploit more tax-saving opportunities. Older firms (AGE) are modestly associated with higher ETRs, echoing the finding that mature firms may avoid reputational risk and regulatory scrutiny (Higgins et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2015\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eWhen we include the board-aggregated director-background controls described above (see Table \u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e18\u003c/span\u003e), the sign, significance and economic magnitude of the DCB and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{DCB}^{2}\\)\u003c/span\u003e\u003c/span\u003e terms remain essentially unchanged, demonstrating that our core inverted-U result is not driven by observable director heterogeneity.\u003c/p\u003e\n\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e4\u003c/span\u003e Here)\u003c/p\u003e\n\u003cp\u003eTo test the core hypothesis that the effect of director networks on tax avoidance is nonlinear, we include the squared term of DCB in the regressions. Table \u003cspan refid=\"Tab7\" class=\"InternalRef\"\u003e5\u003c/span\u003e presents the results. The linear term remains significantly positive, while the quadratic term (DCB\u0026sup2;) is significantly negative for ETR and CETR, confirming a robust inverted-U relationship.\u003c/p\u003e\n\u003cp\u003eThis finding reveals a crucial dynamic: initially, as board connectivity rises, tax avoidance decreases\u0026mdash;likely due to improved governance and reputational constraints. However, beyond a certain point, further increases in centrality reverse this effect, with firms exhibiting lower effective tax rates, suggesting more aggressive tax behavior. In line with Fracassi and Tate (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2012\u003c/span\u003e) and Cheng et al. (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), this implies that excessive centrality transforms board networks into enablers of opportunistic behavior, such as the diffusion of complex tax strategies.\u003c/p\u003e\n\u003cp\u003eThis nonlinear finding is also supported by our predictive margins analysis and formal tests (presented in Section 5), which show that the turning point lies well within the data range, and the model fit improves significantly compared to linear specifications. Directors with elevated centrality may access sophisticated tax advisors, private knowledge flows, or insider-based tax schemes (Cai \u0026amp; Sevilir, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), which would explain the shift from constraint to facilitation.\u003c/p\u003e\n\u003cp\u003eThese results confirm the theoretical premise of this study: director social capital is a double-edged sword, it may enhance ethical governance up to a point, but beyond that, it fosters strategic rent-seeking and regulatory arbitrage.\u003c/p\u003e\n\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab7\" class=\"InternalRef\"\u003e5\u003c/span\u003e Here)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAdditional Analyses and Tests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo verify the robustness and generalizability of our main findings, we conduct a set of additional analyses using alternative measures of director network centrality: closeness centrality (CCB), betweenness centrality (BCB), and eigenvector centrality (ECB). Each metric captures distinct dimensions of a director\u0026rsquo;s network position. CCB reflects the speed of information flow, BCB measures brokerage and influence over network flows, while ECB indicates indirect influence through connections to other well-connected directors.\u003c/p\u003e\n\u003cp\u003eTable \u003cspan refid=\"Tab8\" class=\"InternalRef\"\u003e6\u003c/span\u003e reports the results of GLS regressions using these centrality proxies. The coefficients for CCB and BCB are consistently positive and statistically significant at the 1% level across all three tax avoidance proxies (ETR, CETR, and CAETR). This supports our core result that increased director connectivity is associated with higher reported tax rates, and hence lower tax avoidance, in the linear portion of the relationship. These findings reinforce the idea that moderate levels of board connectivity enhance governance by facilitating transparency and reputational accountability (Ferris et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eInterestingly, the ECB coefficient is not statistically significant for ETR and CETR and only weakly so for CAETR. Since ECB reflects indirect, rather than direct, influence, this may suggest that peripheral or second-order connections are less effective in shaping observable tax policy decisions. The differing influence of centrality types highlights the strategic relevance of direct vs. indirect network positions in shaping corporate financial behavior.\u003c/p\u003e\n\u003cp\u003eTo further test our hypothesis of nonlinearity, Table \u003cspan refid=\"Tab9\" class=\"InternalRef\"\u003e7\u003c/span\u003e presents regression results incorporating the squared terms for CCB, BCB, and ECB. As with our earlier findings for DCB, the results show a statistically significant inverted-U relationship for CCB and BCB. Specifically, the coefficients on CCB\u0026sup2; and BCB\u0026sup2; are negative and significant at the 1% level, suggesting that the disciplining effect of board networks reverses at higher levels of centrality. Directors with excessive proximity to others or excessive brokerage power may accumulate strategic information and develop coalitions that enable sophisticated tax planning techniques\u0026mdash;consistent with findings in Fracassi and Tate (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2012\u003c/span\u003e) and Cheng et al. (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eIn contrast, ECB\u0026sup2; does not display a consistent pattern, further supporting the notion that indirect or less transparent forms of connectivity may exert weaker or more ambiguous influence on tax decisions.\u003c/p\u003e\n\u003cp\u003eOverall, these additional tests reinforce the robustness and multi-dimensional validity of our core result: the relationship between director network centrality and corporate tax avoidance is nonlinear, inverted-U shaped, and manifests consistently across different definitions of centrality. These findings align with broader insights from social network theory, where influence and constraint are contingent on the nature and strength of relational ties (Adler \u0026amp; Kwon, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2002\u003c/span\u003e; Cai \u0026amp; Sevilir, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2012\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab8\" class=\"InternalRef\"\u003e6\u003c/span\u003e Here)\u003c/p\u003e\n\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab9\" class=\"InternalRef\"\u003e7\u003c/span\u003e Here)\u003c/p\u003e"},{"header":"5. Robustness Tests","content":"\u003cp\u003e\u003cstrong\u003eFormal Nonlinearity Checks\u003c/strong\u003e\u003c/p\u003e\n\u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003eTo statistically validate the inverted-U relationship between director network centrality and corporate tax avoidance, we conduct formal nonlinearity tests to compare the performance of linear versus quadratic specifications of our models. These tests allow us to move beyond visual or coefficient-based arguments and provide concrete evidence that the quadratic form is both better-fitting and theoretically appropriate.\u003c/p\u003e\n \u003cp\u003eWe begin by applying the Ramsey Regression Equation Specification Error Test (RESET) to the linear model specifications using each of the three tax avoidance proxies (ETR, CETR, CAETR) as dependent variables. The RESET test detects omitted nonlinear terms by including powers of the fitted values in the regression and testing whether they significantly improve model fit. In Table \u003cspan refid=\"Tab10\" class=\"InternalRef\"\u003e8\u003c/span\u003e, the RESET test returns statistically significant F-statistics (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) in all three models, indicating model misspecification under the linear assumption. This supports our hypothesis that a purely linear specification fails to capture the actual data-generating process and that a non-linear (quadratic) form is statistically justified.\u003c/p\u003e\n\u003c/div\u003e\n\u003cp\u003eTo complement the RESET test, we compare the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) for the linear and quadratic models. Both AIC and BIC penalize model complexity while rewarding improved fit; lower values indicate better model performance. In Table \u003cspan refid=\"Tab11\" class=\"InternalRef\"\u003e9\u003c/span\u003e, the quadratic models consistently outperform the linear specifications in both AIC and BIC across all dependent variables. This provides strong additional evidence that the inclusion of the squared centrality term improves the explanatory power of the models without overfitting.\u003c/p\u003e\n\u003cp\u003eTogether, the results from the RESET and AIC/BIC tests confirm that the relationship between director network centrality and tax avoidance is nonlinear, and that models omitting the quadratic term suffer from functional misspecification. These findings substantiate the theoretical framing of an inverted-U, where social capital initially constrains but later enables more aggressive financial strategies (Desai \u0026amp; Dharmapala, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Tao et al., \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eOur robustness tests reinforce that the inverted-U shape is not only statistically detectable but also economically meaningful and robust to various model specifications.\u003c/p\u003e\n\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab10\" class=\"InternalRef\"\u003e8\u003c/span\u003e Here)\u003c/p\u003e\n\u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab11\" class=\"InternalRef\"\u003e9\u003c/span\u003e Here)\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eControlling for Endogeneity\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eWe re-specify the regression model as follows:\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\n \u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e$$\\:{TA}_{it}\\:=\\:{\\alpha\\:}_{it}+{{\\beta\\:}_{1}CB}_{it-1}+\\:{{\\beta\\:}_{2}CB}_{it-1}^{2}\\:+\\gamma\\:{X}_{it}+{\\mu\\:}_{i}+{\\lambda\\:}_{t}+{\\epsilon\\:}_{it}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003cp\u003eWhere,\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:{CB}_{it-1}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{CB}_{it-1}^{2}\\)\u003c/span\u003e\u003c/span\u003e are the lagged centrality terms (degree centrality and its square), \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{TA}_{it}\\)\u003c/span\u003e\u003c/span\u003e remains the current year\u0026apos;s effective tax rate (ETR, CETR, CAETR). All control variables \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{X}_{it}\\)\u003c/span\u003e\u003c/span\u003e, firm fixed effects \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\mu\\:}_{i}\\)\u003c/span\u003e\u003c/span\u003e, and year fixed effects \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\lambda\\:}_{t}\\)\u003c/span\u003e\u003c/span\u003e are included as in prior models.\u003c/p\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eThe regression results are presented in Table \u003cspan refid=\"Tab12\" class=\"InternalRef\"\u003e10\u003c/span\u003e. The coefficient on the lagged DCB term remains positive and significant, while the coefficient on lagged DCB\u0026sup2; is negative and statistically significant (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) for both ETR and CETR models. These results are nearly identical in sign and magnitude to our original contemporaneous models, confirming that the inverted-U relationship persists when controlling for endogeneity. The results for CAETR remain statistically weaker, which is expected given the noisier nature of cash-based tax proxies and the dominance of accrual tax planning in many Taiwanese firms (Kale, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Nevertheless, the strength and stability of the inverted-U relationship using lagged variables significantly strengthens our identification strategy. The use of lagged independent variables helps reduce concerns of simultaneity and potential reverse causality, lending greater confidence that director network centrality precedes and influences tax avoidance, rather than the other way around. Although this does not eliminate all forms of endogeneity (e.g., omitted variable bias), it represents a meaningful improvement in causal inference and aligns with best practices in recent governance and accounting research (Arora \u0026amp; Gill, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Tao et al., \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab12\" class=\"InternalRef\"\u003e10\u003c/span\u003e Here)\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eWinsorized Sample\u003c/strong\u003e\u003c/p\u003e\u003cp\u003eOutliers and extreme values in financial and governance data can often bias coefficient estimates or inflate standard errors, especially in emerging market settings where disclosure quality and firm heterogeneity are high (Richardson et al., \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). To assess whether our findings are robust to such distortions, we conduct winsorization of all continuous independent variables at the 1st and 99th percentiles, following best practice in empirical accounting and finance literature (Hanlon \u0026amp; Heitzman, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Badertscher et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). We re-estimate the quadratic specification of our core model after winsorizing the following variables: DCB and DCB\u0026sup2;; SIZE, ROA, LEV, R\u0026amp;D, PPE, AGE, PH; ETR, CETR, and CAETR (dependent variables). Winsorization replaces extreme values with the nearest observations within the top and bottom 1% range, preserving the sample size and variance structure while removing extreme leverage points.\u003c/p\u003e\u003cp\u003eTable \u003cspan refid=\"Tab13\" class=\"InternalRef\"\u003e11\u003c/span\u003e presents the regression results after winsorization. The main coefficients of interest\u0026mdash;DCB and DCB\u0026sup2;\u0026mdash;retain their expected signs and statistical significance, with only marginal changes in magnitude. The inverted-U relationship between board centrality and tax avoidance remains intact, demonstrating that our original findings are not driven by a few extreme firms. The winsorized re-estimation confirms that our central findings are not sensitive to outliers, the inverted-U effect remains statistically strong, and the data-generating process is stable across the distribution of firms. These results further support the reliability and credibility of our conclusions regarding the dual effects of director networks on corporate tax avoidance.\u003c/p\u003e\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab13\" class=\"InternalRef\"\u003e11\u003c/span\u003e Here)\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eSubsample Analysis\u003c/strong\u003e\u003c/p\u003e\u003cp\u003eTo further evaluate the robustness and generalizability of our core findings, we conduct subsample analyses by systematically removing subsets of firms that may disproportionately influence the observed relationships. These checks are particularly relevant in Taiwan\u0026rsquo;s market context, where: Technology firms dominate in scale and R\u0026amp;D intensity, and Firm size correlates with board structure, political ties, and disclosure sophistication. Accordingly, we perform the following two exclusion tests:\u003c/p\u003e\u003cp\u003eFirst, we exclude the largest 10% of firms (based on average total assets over 2012\u0026ndash;2023) to assess whether our findings are driven by outlier firms with excessive board complexity or global tax structures. Results from Table \u003cspan refid=\"Tab14\" class=\"InternalRef\"\u003e12\u003c/span\u003e show the inverted-U shape remains strong and statistically significant. Both the linear (DCB) and quadratic (DCB\u0026sup2;) terms remain significant at the 1% level.\u003c/p\u003e\u003cp\u003eSecond, we remove firms classified under technology, electronics, or information services sectors (TEJ codes 2, 3, and 4), which comprise a large portion of Taiwan\u0026rsquo;s export economy and may use industry-specific tax strategies. Results from Table \u003cspan refid=\"Tab15\" class=\"InternalRef\"\u003e13\u003c/span\u003e indicate once again, the inverted-U relationship remains statistically and economically significant for ETR and CETR. CAETR is still not significant \u0026mdash; consistent with earlier tests.\u003c/p\u003e\u003cp\u003eThese two exclusion-based robustness tests confirm that the inverted-U shaped relationship between director network centrality and corporate tax avoidance is not driven by firm size or sector concentration. This finding reinforces that our results reflect a generalizable pattern of director network effects, rather than being artifacts of large-cap tech firms that dominate Taiwan\u0026rsquo;s public markets. It also enhances the international relevance of our findings, especially for emerging markets where similar governance and sector dynamics exist (Fan et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Arora \u0026amp; Gill, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab14\" class=\"InternalRef\"\u003e12\u003c/span\u003e Here)\u003c/p\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab15\" class=\"InternalRef\"\u003e13\u003c/span\u003e Here)\u003c/p\u003e\u003cp\u003e\u003cstrong\u003ePredictive Margins Plot\u003c/strong\u003e\u003c/p\u003e\u003cp\u003eTo further substantiate the nonlinear relationship between director network centrality and corporate tax avoidance, we present Fig. \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, which visualizes the predictive margins of the effective tax rate (ETR) as a function of board degree centrality (DCB), with 95% confidence intervals. As shown, the relationship between board centrality and ETR takes the form of an inverted-U curve, confirming our empirical results from earlier GLS estimations. Specifically, at low to moderate levels of board centrality, increases in director connectivity are associated with higher effective tax rates, indicative of lower tax avoidance. However, beyond a certain threshold of centrality, the marginal effect becomes negative: higher connectivity is now linked to lower ETR values, reflecting increased corporate tax avoidance. This turning point, marked by the red dashed line in the figure, supports the theoretical proposition that board networks serve as a double-edged sword. Initially, connectivity enhances governance and reputational discipline. But as network centrality intensifies, it becomes a conduit for complex, and potentially aggressive, financial behaviors\u0026mdash;including tax avoidance. These findings align with Fracassi and Tate (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2012\u003c/span\u003e) and Cheng et al. (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), who demonstrate that dense director networks can foster the diffusion of both efficient and opportunistic practices. Furthermore, the confidence band surrounding the predicted ETR values remains narrow and stable across most of the centrality distribution, confirming the statistical precision and reliability of the model\u0026rsquo;s estimates.\u003c/p\u003e\u003c/div\u003e\u003cp\u003eEconomic significance and prevalence of the tipping point. To help interpret the inverted-U, we compute the estimated turning point of the quadratic DCB specification and place it in the DCB distribution. Across our main specifications (Tables \u003cspan refid=\"Tab7\" class=\"InternalRef\"\u003e5\u003c/span\u003e and \u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e18\u003c/span\u003e) the turning point is approximately 0.12 (DCB units). This value lies near the 98th percentile of the sample DCB distribution: only\u0026thinsp;~\u0026thinsp;1.8% of firm-year observations (\u0026asymp;\u0026thinsp;61/3,422) exceed this level, while 55 unique firms (\u0026asymp;\u0026thinsp;17.6% of the 313 firms) have at least one year above the threshold. In economic magnitude, the model predicts that increasing DCB from the sample median (0.041) to the turning point (\u0026asymp;\u0026thinsp;0.12) changes the predicted ETR by \u0026asymp;\u0026thinsp;+\u0026thinsp;0.40 (ETR units), indicating a substantial effect. These statistics show two linked facts: the nonlinearity is statistically precise and economically important where it occurs, but the \u0026ldquo;high-centrality\u0026rdquo; enabling regime is relatively rare in our sample. We report bootstrap confidence intervals for the turning point and the percentile counts in Appendix A2.\u003c/p\u003e\u003cp\u003eThis Table \u003cspan refid=\"Tab20\" class=\"InternalRef\"\u003eA2\u003c/span\u003e summarizes the economic significance of the turning point in the inverted-U relationship between board centrality and corporate tax avoidance. The estimated turning point of director centrality (DCB) is approximately 0.12, corresponding to the 98th percentile of the sample distribution. Only about 1.8% of firm-year observations (61 out of 3,422) lie above this threshold, representing 55 unique firms (\u0026asymp;\u0026thinsp;17.6% of the sample). Moving from the sample median to the turning point is associated with an increase in effective tax rate (ETR) of roughly 0.40 units (\u0026asymp;\u0026thinsp;40 percentage points), after which the marginal effect becomes negative. Bootstrap percentile confidence intervals for the turning point are reported in Table \u003cspan refid=\"Tab20\" class=\"InternalRef\"\u003eA2\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e(Please Insert Fig. \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e Here)\u003c/p\u003e\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab20\" class=\"InternalRef\"\u003eA2\u003c/span\u003e Here)\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eExternal validity and generalizability\u003c/strong\u003e\u003c/p\u003e\u003cp\u003eThere is a key concern about whether our Taiwan-based results generalize to most firms. We accept that institutional context conditions inference and therefore add two clarifying elements. First, subsample investigations (Tables \u003cspan refid=\"Tab14\" class=\"InternalRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan refid=\"Tab15\" class=\"InternalRef\"\u003e13\u003c/span\u003e) show that excluding the largest 10% of firms or removing technology/electronics sectors yields qualitatively similar inverted-U patterns for ETR and CETR. These checks reduce the concern that a small set of large, export-oriented firms drives the result. Second, we add further diagnostics. Dynamic GMM (Table \u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e14\u003c/span\u003e) addresses persistence and dynamic endogeneity; IV-2SLS (Table \u003cspan refid=\"Tab16\" class=\"InternalRef\"\u003e15\u003c/span\u003e) corrects for potential simultaneity bias; and PSM regressions (Table \u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e16\u003c/span\u003e) compare matched firms on observables. All three approaches confirm the inverted-U pattern, with turning points interior to the observed distribution. To assess representativeness, we implement inverse-probability weighting (Appendix A1) so that our listed-firm sample better approximates the broader firm population. The quadratic shape remains intact under IPW. We also interact centrality with institutional proxies (ownership concentration, family ownership, political ties), finding the inverted-U is strongest where enforcement is weaker. Finally, we report bootstrapped confidence intervals for the turning point, which consistently lie within the observed centrality range. Together, these checks support the view that the inverted-U effect is not an artifact of a few firms and clarify the institutional conditions under which it is strongest. Nevertheless, we reiterate that generalization beyond markets with similar institutional architectures should be cautious.\u003c/p\u003e\u003cp\u003eThe GMM results in Table \u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e14\u003c/span\u003e confirm the robustness of our baseline findings. Degree centrality (DCB) exhibits a positive and significant association with both ETR and CETR, while its squared term (DCB\u0026sup2;) is negative and significant, reinforcing the inverted U-shaped relationship between board centrality and tax avoidance. The lagged dependent variables are significant, highlighting persistence in tax behavior. Control variables behave as expected: profitability (ROA) and R\u0026amp;D intensity reduce tax avoidance, while leverage increases it. The Hansen and AR(2) tests validate the instruments and specification, confirming that the results are not driven by dynamic endogeneity. The IV-2SLS estimates from Table \u003cspan refid=\"Tab16\" class=\"InternalRef\"\u003e15\u003c/span\u003e corroborate our main findings while addressing remaining endogeneity concerns. The second-stage coefficients show a positive linear DCB effect and a negative quadratic (DCB\u0026sup2;) effect for ETR and CETR, yielding an inverted-U pattern. Instrument diagnostics (first-stage F\u0026thinsp;=\u0026thinsp;28.4; Kleibergen\u0026ndash;Paap p\u0026thinsp;=\u0026thinsp;0.001) indicate instrument strength, and Hansen J tests (p\u0026thinsp;\u0026gt;\u0026thinsp;0.20) do not reject instrument exogeneity. The IV-derived turning point for ETR (\u0026asymp;\u0026thinsp;0.133) lies well inside the observed DCB distribution, implying the nonlinearity is economically meaningful and robust to IV correction. In Table \u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e16\u003c/span\u003e, PSM-based matched-sample results further support the inverted-U finding: DCB is positive and significant while DCB\u0026sup2; is negative and significant for ETR and CETR in the matched sample. The estimated turning point for ETR is approximately 0.132 (bootstrap 95% CI: [0.115, 0.150]), which again is interior to the empirical DCB distribution. The matched-sample results show that differences in observed covariates between high- and low-centrality firms are unlikely to generate the observed nonlinearity; the inverted-U pattern survives after balancing observable characteristics. The IPW results in Table \u003cspan refid=\"Tab19\" class=\"InternalRef\"\u003eA1\u003c/span\u003e confirm the inverted-U relationship. The coefficients on DCB are positive and significant, while those on DCB\u0026sup2; are negative and significant. The estimated turning points (~\u0026thinsp;4) are within the observed range of the centrality distribution. This indicates that the nonlinear pattern is not an artifact of sample composition. Even when re-weighted to better approximate the population of Taiwanese firms, the inverted-U holds.\u003c/p\u003e\u003cp\u003e(Please Insert Tables from 14 to 16 Here)\u003c/p\u003e\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab19\" class=\"InternalRef\"\u003eA1\u003c/span\u003e Here)\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eWinsorization Robustness\u003c/strong\u003e\u003c/p\u003e\u003cp\u003eTo ensure our results are not driven by extreme observations, we winsorize all continuous variables at the 1st and 99th percentiles and re-estimate the baseline specification with the full set of controls. The results, reported in Table \u003cspan refid=\"Tab17\" class=\"InternalRef\"\u003e17\u003c/span\u003e, remain qualitatively unchanged. The coefficients on degree centrality and its squared term continue to support an inverted-U relationship between director connectivity and tax avoidance, and the estimated turning point remains well within the range of observed centrality values. These findings confirm that our main conclusions are not artifacts of outliers but reflect robust underlying patterns. More specifically, the results from remain consistent after winsorization: the inverted-U relationship between board centrality and tax avoidance is robust. All main coefficients retain significance and similar magnitudes, confirming that our findings are not driven by extreme outliers.\u003c/p\u003e\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab17\" class=\"InternalRef\"\u003e17\u003c/span\u003e Here)\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eDirector-background controls\u003c/strong\u003e\u003c/p\u003e\u003cp\u003eWe reproduce the baseline specification (ETR, CETR, CAETR) with the board-aggregated director background controls included. As an additional robustness test, Table \u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e18\u003c/span\u003e adds the board-aggregated director background controls. Coefficients on DCB and DCB\u0026sup2; remain positive and negative (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01), respectively, and the turning point stays interior to the DCB distribution. More specifically, This Table \u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e18\u003c/span\u003e reports the estimation results of the impact of board centrality on corporate tax avoidance after including board-aggregated director background controls: PCT_FINANCE (share of directors with financial/accounting experience), PCT_POLITICAL (share with political ties), AVG_TENURE (average years on the board), and PCT_ADVDEG (share with advanced business/technical degrees). Results confirm the inverted-U relationship: the coefficient on DCB is positive and significant, while DCB\u0026sup2; is negative and significant for ETR and CETR. CAETR results remain directionally consistent. Among director controls, boards with higher financial expertise are associated with less tax avoidance, while political ties modestly increase it. Other controls are not statistically significant.\u003c/p\u003e\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e18\u003c/span\u003e Here)\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eAnother Estimation Method\u003c/strong\u003e\u003c/p\u003e\u003cp\u003eThe fixed-effects (within) estimates in Table \u003cspan refid=\"Tab18\" class=\"InternalRef\"\u003e19\u003c/span\u003e corroborate the baseline GLS findings: DCB (degree centrality) has a positive linear coefficient while DCB\u0026sup2; is negative and significant for ETR and CETR, supporting the inverted-U relationship between board centrality and tax avoidance. Control coefficients behave as expected (e.g., higher ROA and R\u0026amp;D associated with lower measured tax rates; higher LEV associated with higher tax rates). The pattern for CAETR is directionally consistent. These FE results strengthen confidence that the inverted-U is not driven by time-invariant firm heterogeneity.\u003c/p\u003e\u003cp\u003e(Please Insert Table \u003cspan refid=\"Tab18\" class=\"InternalRef\"\u003e19\u003c/span\u003e Here)\u003c/p\u003e"},{"header":"6. Conclusion","content":"\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eOur findings demonstrate an inverted-U relationship between director network centrality and corporate tax avoidance in Taiwan. This evidence highlights a context-dependent dynamic in which moderate connectivity appears to improve oversight while excessive connectivity facilitates sophisticated tax planning strategies. This study investigates the nuanced and nonlinear influence of director network centrality on corporate tax avoidance within Taiwan's emerging market. Building upon social capital and network theories, our findings reveal an inverted-U relationship: moderate levels of board connectivity are associated with reduced tax avoidance, while excessively interconnected boards facilitate more aggressive tax strategies. This outcome reinforces the dual role of social capital in corporate governance, a double-edged sword that can both constrain and enable opportunistic financial behaviors (Adler \u0026amp; Kwon, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2002\u003c/span\u003e; Fracassi \u0026amp; Tate, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2012\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eOur empirical analysis reveals several key findings. Drawing on a balanced panel of 3,422 firm-year observations from Taiwan between 2012 and 2023, we find that the relationship between director network centrality and corporate tax avoidance is both statistically significant and nonlinear. Specifically, we observe that firms with moderately connected boards exhibit higher effective tax rates, indicating less tax avoidance, while firms with highly connected boards display lower effective tax rates, suggesting a shift toward more aggressive tax behavior. This inverted-U shaped relationship is robust across three different tax avoidance measures (ETR, CETR, and CAETR) and remains stable across a range of sensitivity checks. These include Ramsey RESET and information criterion tests (AIC/BIC) for model specification, lagged regressions to address endogeneity, winsorized estimations to mitigate the influence of outliers, and subsample analyses excluding large-cap and technology sector firms. Finally, predictive margins plots with confidence intervals visually confirm the turning point in the relationship between director connectivity and tax behavior. The nonlinear tipping point is economically meaningful where it occurs but relatively uncommon in our sample: only a small share of firm-years exceed the centrality threshold (\u0026asymp;\u0026thinsp;1.8%), though a nontrivial subset of firms (\u0026asymp;\u0026thinsp;18%) experience high centrality at least once. Together, these results provide consistent and compelling evidence that board networks act as a dual governance mechanism, initially deterring and eventually enabling corporate tax avoidance. We also show theoretically that social capital has a non-monotonic governance effect: moderate connectivity raises reputational monitoring and reduces tax avoidance, but beyond a saturation point, social ties become a resource for elite coordination and enable sophisticated tax planning.\u003c/p\u003e \u003cp\u003eImportantly, we frame this contribution as conditional rather than universal: the inverted-U is most plausibly operative in institutional environments with concentrated ownership, strong informal networks, and weaker enforcement. We therefore propose that governance theory incorporates such boundary conditions, and we encourage replication in other markets, both emerging and developed, to map the external validity of the mechanism.\u003c/p\u003e \u003cp\u003eOur results contribute to theory by demonstrating that network effects on firm behavior are not linear, but threshold-dependent. This supports a growing view in governance literature that social capital is dual in nature, capable of both enforcing discipline and enabling coordination depending on context and saturation (Tao et al., \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Cheng et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Furthermore, by using multiple centrality metrics and a richer specification, we show that board influence operates through positional dynamics, not just interlock counts.\u003c/p\u003e \u003cp\u003eFor boards and investors, the findings imply that connectivity is not universally beneficial. While moderate network integration can improve access to strategic knowledge and reputational discipline, over-networking poses governance risks. Institutional investors, rating agencies, and proxy advisors should incorporate network diagnostics into governance assessments, particularly in markets where social ties may dominate formal oversight.\u003c/p\u003e \u003cp\u003eIn emerging economies, where regulatory frameworks and audit capacities are still developing, director network structures should be monitored as part of tax enforcement strategies. Regulatory reforms might include mandatory disclosures of board network affiliations, automated risk flagging based on centrality, or thresholds beyond which additional scrutiny is triggered. This ensures that social capital does not quietly become a structural enabler of tax opacity.\u003c/p\u003e \u003cp\u003eOur findings also have implications beyond Taiwan. Many emerging markets, including India, South Africa, Brazil, Indonesia, Mexico, and Turkey, share institutional features that parallel Taiwan\u0026rsquo;s governance environment: high ownership concentration, reliance on informal director networks, and institutional voids in monitoring and enforcement. In such contexts, the inverted-U relationship between board centrality and tax avoidance provides a useful diagnostic for understanding how social capital simultaneously disciplines and enables corporate behavior. Even in developed economies with club-like director elites (e.g., Italy, South Korea), similar saturation effects may hold. By identifying the broader relevance of our framework, we underline the contribution of this study not only to Taiwanese corporate governance but also to comparative research on emerging markets.\u003c/p\u003e \u003cp\u003eAlthough our analysis is based on Taiwanese firms, the theoretical insights are relevant for broader emerging-market contexts. In countries where ownership is concentrated and informal networks play a significant role, the nonlinear duality of director networks\u0026mdash;as both disciplinary and enabling, may similarly shape corporate tax behavior. We encourage future research to test the inverted-U relationship in other institutional settings to assess the boundary conditions and external validity of our findings.\u003c/p\u003e \u003cp\u003eFuture studies could examine sector-specific network effects, dynamic shifts in board centrality over time, or interactions between centrality and other governance mechanisms like audit committee strength or ownership type. There is also scope for exploring cross-border director ties, especially in multinational firms, and applying graph-theoretic or machine learning methods to extract latent network patterns.\u003c/p\u003e \u003cp\u003eIn conclusion, this paper highlights the dual-edged nature of board networks in corporate tax governance. While social ties can enhance oversight and transparency, excessive connectivity may cross a tipping point, enabling strategic and potentially aggressive financial behavior. This inverted-U relationship offers both a diagnostic tool and a theoretical contribution to the literature on governance in emerging markets. As global financial environments become more interconnected, a deeper understanding of social architecture within boards will be crucial for ensuring accountability, transparency, and sustainable capital markets.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eAuthor A：Conceptualization, Methodology, Formal analysis, Investigation, Resources, Writing-original draft, Writing-review \u0026amp;Editing.Author B： Conceptualization, Writing- review \u0026amp; editing, Supervision.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAdler PS, Kwon S-W (2002) Social capital: Prospects for a new concept. Acad Manage Rev 27(1):17\u0026ndash;40\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAdler PS, Kwon S-W (2023) Social Capital: Prospects for a New Concept\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAllen A, Francis BB, Wu Q, Zhao Y (2016) Analyst coverage and corporate tax aggressiveness. J Banking Finance 73:84\u0026ndash;98\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAndersen A, Garel A, Gilbert A, Tourani-Rad A (2022) Social capital, human capital, and board appointments. Global Finance Journal, 54, p.100758\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAparicio K, Kim R (2023) External capital market frictions, corporate governance, and tax avoidance: Evidence from the TED spread. Finance Res Lett 52:103381\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArmstrong CS, Blouin JL, Jagolinzer AD, Larcker DF (2015) Corporate governance, incentives, and tax avoidance. J Account Econ 60:1\u0026ndash;17\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArmstrong CS, Blouin JL, Larcker DF (2012) The incentives for tax planning. J Account Econ 53(1\u0026ndash;2):391\u0026ndash;411\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArora L, Gill A (2022) Corporate governance and tax avoidance: New evidence from emerging markets. J Acc Emerg Economies 12(3):471\u0026ndash;495\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArora TS, Gill S (2022) Impact of corporate tax aggressiveness on firm value: evidence from India. Managerial Finance 48(2):313\u0026ndash;333\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBadertscher BA, Katz SP, Rego SO (2013) The separation of ownership and control and corporate tax avoidance. J Account Econ 56(2\u0026ndash;3):228\u0026ndash;250\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBertrand M, Mullainathan S (2003) Enjoying the quiet life? Corporate governance and managerial preferences. J Polit Econ 111(5):1043\u0026ndash;1075\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBizjak J, Lemmon M, Whitby R (2009) Option backdating and board interlocks. Rev Financial Stud 22(11):4821\u0026ndash;4847\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBurnham KP, Anderson DR (2002) Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach, 2nd edn. Springer\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCai Y, Sevilir M (2012) Board connections and M\u0026amp;A transactions. J Financ Econ 103(2):327\u0026ndash;349\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCai Y, Dhaliwal DS, Kim Y, Pan C (2014) Board interlocks and the diffusion of disclosure policy. Rev Acc Stud 19:1086\u0026ndash;1119\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCampbell JL, Guan JX, Li OZ, Zheng Z (2020) CEO severance pay and corporate tax planning. J Am Taxation Association 42(2):1\u0026ndash;27\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen S, Chen X, Cheng Q, Shevlin T (2010) Are family firms more tax aggressive than non-family firms? J Financ Econ 95(1):41\u0026ndash;61\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen Y, Huang J, Li X, Ni X (2023) Financial market opening and corporate tax avoidance: Evidence from staggered quasi-natural experiments. Finance Res Lett 54:103770\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheng CA, Dhaliwal DS, Zhang Y (2019) Does interlocking directorate shape unethical accounting behavior? Contemp Acc Res 36(4):2141\u0026ndash;2171\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheng S, Felix R, Zhao Y (2019) Board interlock networks and informed short sales. J Banking Finance 98:198\u0026ndash;211\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDesai MA, Dharmapala D (2009) Corporate tax avoidance and firm value. Rev Econ Stat 91(3):537\u0026ndash;546\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDesai MA, Dharmapala D (2006) Corporate tax avoidance and high-powered incentives. J Financ Econ 79(1):145\u0026ndash;179\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDyreng SD, Hanlon M, Maydew EL, Thornock JR (2017) Changes in corporate effective tax rates over the past 25 years. J Financ Econ 124(3):441\u0026ndash;463\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDyreng SD, Hanlon M, Maydew EL (2008) Long- Run Corporate Tax Avoidance. Acc Rev 83:61\u0026ndash;82\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEl-Khatib R, Fogel K, Jandik T (2015) CEO network centrality and merger performance. J Financ Econ 116(2):349\u0026ndash;382\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFan JPH, Wong TJ, Zhang T (2007) Politically connected CEOs, corporate governance, and post-IPO performance of China's newly partially privatized firms. J Financ Econ 84(2):330\u0026ndash;357\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFerris SP, Javakhadze D, Rajkovic T (2017) CEO social capital, risk-taking, and corporate policies. J Corp Finance 45:289\u0026ndash;306\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFilatotchev I, Chahine S, Bruton GD (2018) Board interlocks and initial public offering performance in the United States and the United Kingdom: An institutional perspective. J Manag 44(4):1620\u0026ndash;1650\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFracassi C, Tate G (2012) External networking and internal firm governance. J Finance 67(1):153\u0026ndash;194\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFrank M, Margaret LJ, Lynch, Sonja O, Rego (2009) Tax Reporting Aggressiveness and its Relation to Aggressive Financial Reporting. The Accounting Review 84\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGaaya S, Lakhal N, Lakhal F (2017) Does family ownership reduce corporate tax avoidance? The moderating effect of audit quality. Managerial Auditing J 32:731\u0026ndash;744\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGao L, Yang LL, Zhang JH (2016) Corporate patents, R\u0026amp;D success, and tax avoidance. Rev Quant Financ Acc 47:1063\u0026ndash;1096\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGodigbe BG, Chui CM, Liu CL (2018) Directors network centrality and earnings quality. Appl Econ 50(50):5381\u0026ndash;5400\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoncalves AF, Rossoni L, Mendes-Da-Silva W (2019) Board social capital reduces implied cost of capital for private companies but not of state-owned companies. Manag Decis 57(10):2672\u0026ndash;2692\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHalioui K, Neifar S, Ben Abdelaziz F (2016) Corporate governance, CEO compensation and tax aggressiveness: Evidence from American firms listed on the NASDAQ s100. Rev Acc Finance 15(4):445\u0026ndash;462\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHan J, Bose I, Hu N, Qi B, Tian G (2015) Does director interlock impact corporate R\u0026amp;D investment? Decis Support Syst 71:28\u0026ndash;36\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHanlon M, Heitzman S (2010) A review of tax research. J Account Econ 50(2\u0026ndash;3):127\u0026ndash;178\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHiggins DM, Omer TC, Phillips JD (2015) The influence of a firm's business strategy on its tax aggressiveness. Contemp Acc Res 32(2):674\u0026ndash;702\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoward MD, Withers MC, Tihanyi L (2017) Knowledge dependence and the formation of director interlocks. Acad Manag J 60(5):1986\u0026ndash;2013\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJarboui A, Kachouri Ben Saad M, Riguen R (2020) Tax avoidance: do board gender diversity and sustainability performance make a difference? J Financial Crime 27(4):1389\u0026ndash;1408\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKale P (2019) Tax strategy and the business landscape in Taiwan. Asian J Acc Res 4(2):90\u0026ndash;109\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKale S (2019) Taxing times: compliance complexity a big challenge for businesses. Financial Ex\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKanagaretnam K, Lee J, Lim CY, Lobo GJ (2016) Relation between Auditor Quality and Tax Aggressiveness: Implications of Cross-Country Institutional Differences. AUDITING: J Pract Theory 35:105\u0026ndash;135\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKhurana IK, Moser WJ (2013) Institutional Shareholders\u0026rsquo; Investment Horizons and Tax Avoidance. J Am Taxation Association 35:111\u0026ndash;134\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLanis R, Richardson G (2011) The effect of board of director composition on corporate tax aggressiveness. J Account Public Policy 30:50\u0026ndash;70\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLanis R, Richardson G, Taylor G (2017) Board of Director Gender and Corporate Tax Aggressiveness: An Empirical Analysis. J Bus Ethics 144:577\u0026ndash;596\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLarcker DF, So EC, Wang CC (2013) Boardroom centrality and firm performance. J Account Econ 55(2\u0026ndash;3):225\u0026ndash;250\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi D, Jiang Q, Mai Y (2019) Board interlocks and capital structure dynamics: evidence from China, vol 59. Accounting \u0026amp; Finance, pp 1893\u0026ndash;1922\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLin S, Tong N, Tucker AL (2014) Corporate tax aggression and debt. J Banking Finance 40:227\u0026ndash;241\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eManzon G, Plesko GA (2002) The Relation Between Financial and Tax Reporting Measures of Income. Law Rev 55:175\u0026ndash;214\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMasulis RW, Mobbs S (2014) Independent director incentives: Where do talented directors spend their limited time and energy? J Financ Econ 111(2):406\u0026ndash;429\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcGuire ST, Wang D, Wilson RJ (2014) Dual Class Ownership and Tax Avoidance. Account Rev 89:1487\u0026ndash;1516\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMinnick K, Noga T (2010) Do corporate governance characteristics influence tax management? J Corp Finance 16:703\u0026ndash;718\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoore JA (2012) Empirical evidence on the impact of external monitoring on book\u0026ndash;tax differences. Adv Acc 28(2):254\u0026ndash;269\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOECD (2020) Corporate tax statistics\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOECD (2020) Corporate Tax Statistics: Second Edition. Organisation for Economic Co-operation and Development\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRajagopalan N, Zhang Y (2008) Corporate governance reforms in China and India: Challenges and opportunities. Bus Horiz 51(1):55\u0026ndash;64\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRamsey JB (1969) Tests for specification errors in classical linear least-squares regression analysis. J Royal Stat Soc Ser B (Methodological) 31(2):350\u0026ndash;371\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRego SO (2003) Tax-Avoidance Activities of U.S. Multinational Corporations. Contemp Acc Res 20:805\u0026ndash;833\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRichardson G, Lanis R, Leung SCM (2014) Corporate tax aggressiveness, outside directors, and debt policy: An empirical analysis. J Corp Finance 25:107\u0026ndash;121\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRichardson G, Taylor G, Lanis R (2013) The impact of board of director oversight characteristics on corporate tax aggressiveness: An empirical analysis. J Account Public Policy 32(3):68\u0026ndash;88\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRichardson G, Taylor G, Lanis R (2015) The impact of financial distress on corporate tax avoidance spanning the global financial crisis: Evidence from Australia, vol 44. Economic Modelling, pp 44\u0026ndash;53\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRiguen R, Salhi B, Jarboui A (2020) Do women in board represent less corporate tax avoidance? A moderation analysis. Int J Sociol Soc Policy 40(1/2):114\u0026ndash;132\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTao Q, Li H, Wu Q, Zhang T, Zhu Y (2019) The dark side of board network centrality: Evidence from merger performance. J Bus Res 104:215\u0026ndash;232\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTsai L-C, Zhang R, Zhao C (2019) Political connections, network centrality and firm innovation. Finance Res Lett 28:180\u0026ndash;184\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Footnotes","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003e We measure tax outcomes as proportions (tax / pretax income): ETR\u0026thinsp;=\u0026thinsp;total tax / pretax income; CETR\u0026thinsp;=\u0026thinsp;current tax / pretax income; CAETR\u0026thinsp;=\u0026thinsp;cash taxes paid / pretax income. Lower values of each measure indicate greater tax avoidance (i.e., a smaller tax payment relative to pretax income). Because measurement conventions vary across studies, we also report results using an alternative measure \u0026mdash; the book\u0026ndash;tax difference (BTD) \u0026mdash; and a binary indicator of aggressive tax positions; these robustness checks (Section 5, Table A-x) produce the same inverted-U relationship between board centrality and tax avoidance.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e We note that all reported tax-rate coefficients are interpreted such that a negative association implies more tax avoidance (lower tax paid relative to pretax income); see Variable Definitions and Appendix robustness tests for the alternative BTD specification.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTable 1 to 19 are available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Tax avoidance, director networks, board centrality, social capital, emerging markets, corporate governance","lastPublishedDoi":"10.21203/rs.3.rs-9196981/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9196981/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study investigates the relationship between director network centrality and corporate tax avoidance, using data from publicly listed firms in Taiwan, an emerging market characterized by high ownership concentration, strong informal networks, and institutional voids. Drawing on the integrated agency-resource dependence framework, we propose and test a non-linear association between board connectivity and tax behavior. Using multiple measures of network centrality (degree, betweenness, closeness, and eigenvector) and tax avoidance (ETR, CETR, CAETR), we confirm the robustness of this non-linear effect using lagged specifications, alternative centrality metrics, and formal nonlinearity tests (RESET, AIC/BIC, Lind-Mehlum slope changes). Graphical analysis of predictive margins further supports the curvilinear pattern. We find a robust inverted-U relationship: moderate board connectivity appears to strengthen reputational discipline and reduce tax avoidance, whereas excessive connectivity corresponds with greater capacity for complex tax planning. We temper our claims: these results suggest that a strictly linear view of governance may be incomplete in institutional contexts characterized by concentrated ownership, informal director networks, and weaker enforcement such as Taiwan. We therefore present our findings as context-bounded evidence that highlights boundary conditions for the governance literature, rather than as a universal refutation of linear governance models.\u003c/p\u003e","manuscriptTitle":"Director Network Centrality and Corporate Tax Avoidance","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-04 10:55:12","doi":"10.21203/rs.3.rs-9196981/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"5c4db7da-0189-491d-bb58-c4881962077c","owner":[],"postedDate":"May 4th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-05-04T10:55:13+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-04 10:55:12","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9196981","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9196981","identity":"rs-9196981","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}