On the Right Side of the Faultline: Effects of Subgroups and CEO Inclusion on CEO Compensation

This study examines how subgroup formation resulting from group faultlines can affect corporate boards’ CEO compensation-setting process. Group faultlines are hypothetical dividing lines that can split the board into subgroups based on directors’ diversity characteristics. Using a sample of US firms for the period 2007–2019, we find that if the CEO demographically belongs to the same faultline-based subgroup as corporate directors who are members of the compensation committee, strong faultlines lead to higher CEO compensation. Furthermore, we show that joint tenure of directors and the CEO strengthens this relationship. Our results are robust to endogeneity concerns and a battery of sensitivity tests.Our findings make importantcontributions to the literature on CEO compensation, corporate governance and faultlines, as we consider not only the existence of faultlines in the board context but also how the composition of identity-based subgroups affects CEO compensation.


Introduction
Interest in determining executive compensation packages has increased substantially, as CEO compensation has outpaced stock prices or corporate profits in recent years (Mishel and Schieder, 2018).Boards of directors are held accountable for executive compensation, as determining the correct compensation packages is within the purview of their control role (Daily et al., 1998).However, board decision-making regarding CEO pay is fundamentally ambiguous, as directors' interpretations can be subject to sociopsychological processes (Lewellyn and Muller-Kahle, 2021).Our knowledge of the social influence dynamics operating between the CEO and the rest of the board remains insufficient (Fiss, 2006;Kolev et al., 2019;Lewellyn and Muller-Kahle, 2021;O'Reilly III and Main, 2010).In fact, the CEO compensation-setting process can be affected by the existence of faultlines within the corporate board.
Faultlines are hypothetical lines that can split a board into subgroups based on the alignment of multiple diversity characteristics (Lau and Murnighan, 1998).The Authorship is alphabetical; the two authors contributed equally to the research.concept of faultlines goes beyond standard diversity measures because it considers the alignment of multiple diversity attributes simultaneously.The main theoretical framework underlying the concept of faultlines is based on social identity and categorization theories (Lau and Murnighan, 1998).If multiple attributes align, a strong faultline is created, which can lead to salient identity-based subgroups, which, in turn, may become a basis for social identity and categorization processes (Hogg and Terry, 2000;Lau and Murnighan, 1998).Such subgroup formation can have detrimental effects on decision-making (Carton and Cummings, 2012).Therefore, we use the concept of faultlines to detect identity-based subgroups within corporate boards and investigate how their composition affects board decision-making regarding CEO compensation.Specifically, we examine the particular demographic subgroup to which the CEO belongs and argue that this belonging can impact how directors evaluate the CEO's performance, thus influencing CEO pay.
The literature on corporate boards is increasingly examining the effects of faultlines on corporate boards' decision-making (i.e.Arena, Garcia-Torea and Michelon, 2024;Barroso-Castro et al., 2020;Pu, Xie and Wang, 2023;Shin and You, 2022;Van Peteghem, 2 C. Antoons and A. Vandebeek Bruynseels and Gaeremynck, 2018;Vandebeek et al., 2021Vandebeek et al., , 2024;;Xue et al., 2024).However, the relationship among faultlines, the resulting subgroups and CEO compensation remains unclear.Our study will shed more light on this relationship, by investigating faultlines based on three attributes: gender, age and board-committee membership.Gender and age are two social identity attributes that generate social categorization and are widely recognized as the most predominant and influential characteristics in studies examining social categorization processes (Thatcher and Patel, 2012).If these attributes align with board committee membership, committee members may form strong and cohesive subgroups within the board.Particularly, the compensation-setting process relies heavily on the compensation committee's deliberations as a subcommittee of the board (Murphy, 1999); this subcommittee must assess the CEO's performance according to their expectations (Kolev et al., 2019).Therefore, compensation committee membership can provide a sense of power in CEO pay decision-making (Spira and Bender, 2004).Importantly, we expect that the effects of faultlines on CEO compensation will depend on whether the CEO belongs to a strong subgroup within the board that contains compensation committee members.
This study makes several contributions to the literature.We contribute to corporate governance research by considering not only the existence of faultlines in the corporate board context but also how the composition of identity-based subgroups affects boards' decisionmaking regarding CEO compensation.In this way, we specifically add to the extant literature on CEO compensation by considering the effects of faultlines and the composition of corresponding subgroups on the CEO compensation-setting process.While prior studies have focused on isolated similarity measures (i.e.Belliveau, O'Reilly III and Wade, 1996;Westphal and Zajac, 1995;Young and Buchholtz, 2002), our study presents a more nuanced view by examining multiple diversity attributes simultaneously using the concept of faultlines.Thus, we also add to board diversity research (e.g.Dah, Dah and Frye, 2023;Firoozi and Keddie, 2022;Guest, 2019;Gull et al., 2023;Nanda, Prevost and Upadhyay, 2023;Pandey et al., 2023).Moreover, we contribute to prior research by examining the CEO's belonging to a faultline-based subgroup as a factor that can influence assessments of CEO performance.Next, we contribute to faultline research by identifying a new combination of attributes (board-committee membership, age and gender) in faultlines, as well as by elaborating on the concept of faultlines in corporate boards, which is an understudied topic.Therefore, we respond to calls to integrate the concept of faultlines into different settings (Murnighan and Lau, 2017).

Board faultlines and CEO compensation
Early research on CEO compensation examined how demographic and social similarities between CEOs and directors affect compensation, generating inconclusive results.Westphal and Zajac (1995) showed that CEOs prefer demographically similar directors, while such increased similarity between the CEO and the board is likely to result in more generous CEO compensation owing to social influence.However, Belliveau, O'Reilly III and Wade (1996) found no effects of social similarity on CEO compensation.Some studies only consider the similarities between the CEO and board chair (e.g.Fiss, 2006).Young and Buchholtz (2002) examined how demographic dissimilarity between the CEO and the members of the compensation committee affected pay for performance and found mixed results.The authors concluded that greater dissimilarity in age led to a weaker relationship between CEO pay and performance, and dissimilarity in gender had no influence.One key limitation of these studies is that diversity attributes were examined in isolation, using traditional measures of similarity.Social identity comprises a complex bundle of demographic attributes, and the measures used in prior studies do not capture the interaction among attributes (Jehn and Rupert, 2008).The concept of faultlines overcomes these limitations by simultaneously examining the alignment of multiple attributes (Lau and Murnighan, 1998;Thatcher, Jehn and Zanutto, 2003).Additionally, boards are becoming increasingly diverse, potentially attenuating the effects of CEO-board similarity (Fiss, 2006;Westphal and Zajac, 1995).However, we argue that social influence processes are currently at play within corporate boards owing to an important downside of diversity, namely, the existence of faultlines that can create subgroups within boards.Similarity within subgroups, combined with strong out-group differences, may affect decisionmaking in increasingly diverse boards, which we will explain in the following paragraphs.
The theoretical foundation for the concept of faultlines is rooted in social identity and categorization theories (Lau and Murnighan, 1998).These theories explain how group members perceive each other as part of different social categories and how subgroups form when members share a common identity, values and social characteristics (Ashforth and Mael, 1989;Hogg and Terry, 2000;Tajfel and Turner, 2004).Members of one subgroup often feel scarce personal attachment to members of other subgroups (i.e. the out-group) and favour interactions with their own subgroup members (i.e. the in-group; Carton and Cummings, 2012;Kunze and Bruch, 2010).Such subgroup identification can hinder collective thinking and affect decision-making, as subgroup members are expected to promote the aims of said subgroup (Crucke and Knockaert, 2016).From a social categorization perspective, social category faultlines can 'separate' the board because they denote different social identities and induce members to see each other as belonging to a particular social category (Harrison and Klein, 2007;Tajfel and Turner, 2004), thus creating identity-based subgroups.Such separating effects are less common for other types of faultlines, such as informational faultlines, which are based on attributes such as educational specialization (Qu and Liu, 2017).Our study focuses on social category faultlines because they are the most relevant in groups aiming to accomplish a specific task (in our case, deciding on CEO compensation), whereas informational faultlines are more relevant for strategic decisions in which the integration of alternative perspectives is vital (e.g.different directors have different pieces of information that need to be integrated for strategic decision-making; Bezrukova et al., 2009;Vandebeek et al., 2021).Moreover, we aim to study the effects of social influence and the detrimental effect of separating social identity-based subgroups on the board's compensation-setting process.By solely examining social category attributes, we can theorize more precisely regarding the effects of social category faultlines, as 'measuring a single faultline based on characteristics that reflect very different aspects of individuals may hinder interpretation of its effect' (Hutzschenreuter and Horstkotte, 2013, p. 719).
Our study investigates the effects of social category faultlines based on three social category attributes.Following prior faultline research, we included age and gender as social attributes (Thatcher and Patel, 2012), as these characteristics play a key role in first impressions and in how individuals categorize themselves and other directors into identity-based subgroups (Bezrukova et al., 2009;Thatcher and Patel, 2012).Next, we included committee membership as an attribute, inspired by Crucke and Knockaert (2016), who suggested that stakeholder representation on boards is an important social identity attribute leading to social category faultline emergence.Compensation setting relies heavily on the compensation committee's deliberations as a subcommittee of the board (Murphy, 1999) that must interpret how well the CEO has performed according to their expectations (Kolev et al., 2019).Therefore, committee membership may contribute to subgroup emergence because it can provide a sense of power (Spira and Bender, 2004), while directors can derive part of their social identity from their membership (Hillman, Nicholson and Shropshire, 2008).Therefore, committee membership is an important social category attribute in the compensation-setting process.Faultlines become stronger as more attributes align (Lau and Murnighan, 1998).If compensation committee membership aligns with gender and age, we expect strong faultlines to generate strong identity-based subgroups.
Members of a strong identity-based subgroup resulting from strong faultlines exhibit stronger interdependence (Meyer et al., 2015), while subgroup members may have a greater incentive to comply with compensation arrangements that favour the CEO if the latter is part of their subgroup.Within subgroups, social and psychological mechanisms such as friendship, loyalty and collegiality may be stronger, thus making CEOs better positioned to negotiate compensation arrangements in their favour (O'Reilly III and Main, 2010).Furthermore, numerous studies have provided evidence that people tend to have more favourable views towards in-group members than towards out-group members (Carton and Cummings, 2012;Tajfel and Turner, 2004).In-group members have a significant advantage in receiving rewards and positive evaluations compared with out-group members (Bodenhausen, Kang and Peery, 2012;Zhu, Shen and Hillman, 2014).If the subgroup of compensation committee members includes the CEO, it is likely that in-group members will overly positively evaluate the CEO's performance, resulting in higher compensation.In contrast, CEOs not included in this subgroup will be less likely to be appreciated for their positive behaviour and more likely to be blamed for negative results (Abrams et al., 1990).Consequently, it is likely that out-group CEOs receive lower compensation, on average.Conversely, on boards with weak faultlines, directors identify less with their subgroup and are more likely to focus on the entire group (Lau and Murnighan, 2005), thereby reducing the likelihood of in-group favouritism.
For example, in a scenario involving a board that includes two female directors in their 40s who both serve on the compensation committee and two male directors in their 60s who do not serve on this committee, a strong faultline will emerge based on the attributes of age, gender and compensation committee membership.We expect that, in this scenario, if the CEO is a 40-yearold woman, she demographically belongs to the same identity-based subgroup as members of the compensation committee; thus, strong faultlines may have a positive effect on CEO compensation, as, in this case, members perceive strong differences with out-group members and evaluate the in-group CEO more positively.By contrast, if the CEO in this scenario is a 60-year-old man and he belongs to a different subgroup than the members of the compensation committee, we expect a lower evaluation of the CEO and lower compensation.Therefore, we propose: H1: If a CEO is part of a faultline-based subgroup of compensation committee members, stronger faultlines will be associated with higher CEO compensation.However, if the CEO is not included in 4 C. Antoons and A. Vandebeek this subgroup, stronger faultlines will be associated with lower CEO compensation.

The role of joint CEO-director tenure
According to Lau and Murnighan (1998, p. 333), 'over time, subgroup identification can grow, augmented by polarization.Subsequent interactions, then, may act to legitimize the subgroups, and conflict between them may continue to be likely'.Therefore, increased subsequent interactions due to higher joint CEO-director tenure can alter social influence patterns (Westphal, 1999).Moreover, social category faultlines may lead to subgroup entrenchment, in which group members agree on the existence and composition of strong and stable subgroups (Meister et al., 2020(Meister et al., , p. 1475).Therefore, a faultline-based subgroup containing compensation committee members can become more cohesive as directors become more integrated within the subgroup over time (Carton and Cummings, 2012).
We argue that joint CEO-director tenure will positively moderate the interaction between social category faultlines and the CEO's membership to the compensation committee members' subgroup vis-à-vis CEO compensation.That is, if the CEO is part of the faultline-based subgroup and joint tenure is high, favouritism towards the CEO as part of that in-group can be higher, as intergroup bias (i.e. the systematic tendency to behave more positively towards the in-group) strengthens over repeated contact (Dorrough et al., 2015;Tajfel and Turner, 2004).Therefore, we argue that stronger faultlines will be associated with higher CEO compensation if the CEO is included in the faultlinebased subgroup of the compensation committee and joint tenure is high because in-group members will evaluate the performance of those CEOs positively owing to increased subgroup identification.However, in the case of strong faultlines, CEOs who are not included in the subgroup of the compensation committee members will always be considered out-group members because of the salient basis for in-and out-group categorization that social category membership provides (Bezrukova et al., 2009), thus leading to lower compensation.In our previous scenario (i.e.all compensation committee members are women in their 40s), we expected that a new 40-yearold female CEO (i.e.low joint tenure) would benefit less from a more positive in-group evaluation than a 40year-old woman CEO with a longer joint tenure with directors.Conversely, the 60-year-old man in this scenario will always belong to a different identity-based subgroup than the compensation committee members, thus lowering his compensation.Therefore, we propose: H2: If a CEO is part of the subgroup of the compensation committee members and the joint CEOdirector tenure is higher, stronger faultlines will be associated with higher CEO compensation.However, if the CEO is not part of this subgroup, stronger faultlines will be associated with lower CEO compensation.

Sample and data
We obtained information on CEO compensation from Execucomp for the period 2007−2019. 1We then added the company financial data from Compustat.We obtained detailed data on director characteristics of US companies from BoardEx.To compute our faultline measures, we required each firm to have at least three board members, as this is the minimum size that allows the emergence of subgroups (Meyer and Glenz, 2013).In line with prior research (e.g.Gull et al., 2023;Nanda, Prevost and Upadhyay, 2023), we discarded observations with missing data and excluded financial firms, 2 obtaining a final sample of 14,253 firmyear observations.Table 1 outlines the sample selection process.

Measurement of the main variables
CEO pay.Our dependent variable is total CEO compensation, measured as the natural logarithm of total annual compensation, which includes salary, bonuses, the Black-Scholes value of stock-option grants, restricted stock grants and long-term incentive plans (Gupta and Wowak, 2017;Shin and You, 2017;Van Essen, Otten and Carberry, 2015).
Faultline measure.Our faultline measure follows a cluster-based approach, namely, average silhouette width (ASW) faultline clustering.The cluster analysis groups board members into subgroups according to their similarity, such that the subgroups have maximum internal homogeneity and between-cluster heterogeneity.For each board in each year, the strength of the faultline was calculated.ASW goes beyond other faultline measures, as it does not need the attributes to be categorical or assume the existence of only two subgroups (Meyer and Glenz, 2013).The procedure for the ASW algorithm to detect subgroups and calculate the faultline strength involved several steps.First, cluster analysis determined possible subgroup splits based on the distribution of board members' attributes, while all 1 We focused on this sample period because in 2006, the U.S. Securities and Exchange Commission (SEC) adopted enhanced executive compensation disclosure requirements in proxy statements.These requirements became effective for fiscal years ending on or after 15 December 2006.As of fiscal year 2007, all companies in our sample reported CEO compensation following the new format.
2 Firms with SIC codes 6000-6999.members were placed in their own subgroup of size 1.Next, clusters with very similar directors were merged into bigger subgroups.Finally, all members became part of a cluster that encompassed the entire group.ASW is the average of each director's individual silhouette width, which quantifies how well the director fits into a subgroup.Individual i's silhouette width s is defined as

Effects of Subgroups and CEO Inclusion
where a i denotes the average dissimilarity of i to all members of Cluster A, while b i denotes the average dissimilarity of i to all members of Cluster B (Meyer and Glenz, 2013).The ASW value ranges between −1 and 1; the closer the ASW value is to 1, the more homogenous the subgroups.If the ASW value is below 0, members of the same subgroup are more dissimilar than members of different subgroups.The algorithm then selected the subgroup configuration that yielded the highest ASW.All the ASW calculations were performed using the asw.cluster package for faultline calculations in R (Meyer and Glenz, 2013).For more details and illustrative examples of these calculations, refer to Meyer and Glenz (2013).Other studies that have used the ASW measure in board settings include Shin and You (2022) and Vandebeek et al. (2021).Our measure of faultline strength is based on three characteristics: gender, age and committee membership.We included gender and age, which are widely recognized as the most predominant and influential characteristics in studies examining social categorization processes (Bezrukova et al., 2009;Li and Hambrick, 2005;Shin and You, 2022;Veltrop et al., 2015), because these attributes represent social categories that lead to social identity attribution and similarity attraction (Crucke and Knockaert, 2016;Vandebeek et al., 2016).The ages of the board members in our sample ranged from 25 to 103 years (average: 62 years).Furthermore, approximately 15% of the board members were women.Next, we included committee membership because this attribute may contribute to subgroup emergence, as it can provide a sense of power (Spira and Bender, 2004), and members can derive part of their social identity from belonging to this subgroup (Hillman, Nicholson and Shropshire, 2008).
CEO in same subgroup.In addition to faultline strength, the algorithm yields information on the subgroup configuration that resulted in the highest ASW value, namely the number and sizes of subgroups and which member was placed into which subgroup.We used these data to detect whether the CEO belonged to the same subgroup as the members of the compensation committee.First, we identified the subgroup to which the CEO belonged; similarly, we identified the subgroup to which each compensation committee member belonged.We then defined the subgroup with the largest number of compensation committee members.Our measure (i.e.CEO in same subgroup) equalled 1 if the CEO belonged to this subgroup and 0 otherwise.Importantly, if all members belong to the same subgroup, the CEO belonged by default to the same subgroup as the members of the compensation committee.However, there were no outgroups in this case.Therefore, we retained only observations in which at least two subgroups were present.In 17% of observations, the CEO was included in the subgroup with the largest number of compensationcommittee members.
Joint CEO-director tenure.Finally, to test H2, we used the measure proposed by Carroll and Harrison (1998) to capture the shared tenure of the CEO and board members.This measure was calculated as the natural logarithm of the average number of years each pair had worked together on the board of directors. 3The average (median) joint tenure in our sample was 5.46 (5.01) years.We used this measure to split the sample into subsamples with low (i.e.lower than or equal to the median value) and high (i.e. higher than the median value) joint tenures. 4

Control variables
The regression models controlled for several factors that may affect CEO compensation.First, we included a set 6 C. Antoons and A. Vandebeek of firm-level control variables identified in the literature as economic determinants of total compensation (Core, Holthausen and Larcker, 1999;Dah, Dah and Frye, 2023;Van Peteghem, Bruynseels and Gaeremynck, 2018).Firm size was proxied using the natural logarithm of total sales.To control for firm performance, we included accounting (i.e. return on assets) and market (i.e. stock market return) measures.We also included the standard deviation of stock returns over the past 3 years as a measure of firm risk (i.e. standard deviation of return).Next, we included leverage, measured as the debt-to-capital ratio and a firm's investment opportunities, inversely proxied by the book-to-market ratio.Finally, we controlled for firm age and Tobin's Q.
Furthermore, we included a set of board characteristics relevant to decision-making (Hutzschenreuter and Horstkotte, 2013;Shin and You, 2017).We controlled for board size, the proportion of independent directors on the board (i.e.percent independent) and the proportion of board members appointed after the CEO takes office (i.e.percent appointed after CEO), as these directors may feel loyal to the CEO (Shin and You, 2022).Next, we added the number of compensation committee members (i.e.compensation committee size), as this may affect the distribution of our variables of interest.We also included average director tenure on the board.
The models control for several CEO characteristics identified in the literature (Gupta and Wowak, 2017;Nanda, Prevost and Upadhyay, 2023;Zajac and Westphal, 1996).Numerous studies argue that powerful CEOs can increase their compensation levels (e.g.Van Essen, Otten and Carberry, 2015).To capture CEO power, we included CEO duality and the natural logarithm of CEO tenure.We also included the natural logarithm of CEO age, and CEO gender.Finally, we included an indicator variable for outside CEOs (i.e.CEO hired from outside), who often receive higher compensation levels than internally promoted CEOs (Peters and Wagner, 2014).
We also included a set of traditional board diversity and CEO-board similarity measures to ascertain the distinct effects of faultlines and board subgroups on CEO pay.We added the proportion of female directors (i.e.percent women) and director age heterogeneity, measured as the coefficient of variation of age among all directors on a given board.Finally, we included a measure that captured average CEO-board similarity in terms of age and gender.To construct a measure of age similarity, we computed the Euclidean distance between the CEO's age and the average age of the board members.For gender similarity, we first identified whether each board member was of the same gender as the CEO.Then, we calculated the squared value of the proportion of same-gender dyads across the board.Our CEOboard similarity measure was then calculated as the average of the standardized age and gender similarity measures.
Finally, in line with prior studies (Bezrukova et al., 2009;Chung et al., 2015), we controlled for informational faultlines (Shin and You, 2022;Vandebeek et al., 2024).Our informational faultline measure considered directorship, board tenure, educational level and specialization. 5All continuous variables were winsorized at the 1st and 99th percentiles to account for possible outliers.All variables are defined in Appendix A. Table 2 provides the summary statistics for all firm-year observations in our sample.The mean (median) of annual CEO pay in our sample was approximately $5,890,602 ($4,297,279).Table 3 presents a pairwise correlation matrix.The correlation among all variables was less than the threshold of 0.7 (Gull et al., 2023).

Empirical approach
We first assessed whether there was a multicollinearity problem in our sample and computed the variance inflation factors for all variables.All variance inflation factors were below 2.74, which is lower than the recommended value of 5 (O'Brien, 2007), indicating that there was no multicollinearity in our dataset.Endogeneity may be a potential issue, as previous research has documented that governance characteristics are not randomly distributed among firms and that the appointment of directors to committees may not be random (Adams and Ferreira, 2009;Gull et al., 2023).We took several steps to address this concern.First, we lagged all the faultline-related variables by 1 year relative to the dependent variable to facilitate causal inferences. 6Second, in addition to using ordinary least squares (OLS) regression models with industry fixed effects and robust standard errors (White, 1980), we employed firm-level fixedeffect models to control for the effects of time-invariant variables. 75 Directorship type was divided into two categories: 'executive director' and 'independent director'.Tenure was measured as the number of years the director had served on the board.To measure educational level, we classified each director into one of five educational levels: high school, college, academic bachelor's degree, academic master's degree and PhD.Finally, we identified the field of study in which directors received their highest degree.We categorized directors into seven educational areas, following Hambrick, Cho and Chen (1996): engineering, sciences, business, economics, liberal arts, law and 'other' (i.e.director obtained degree in another area). 6Lagging all independent variables, including control variables, by 1 year, produces qualitatively similar results (available upon request).

Results
Table 4 reports the regression results for the moderating effects of CEO in same subgroup.Models 1 and 3 are baseline models with only control variables.
The coefficient estimates show a pattern identified in prior research (e.g.Faleye, Hoitash and Hoitash, 2011; Van Peteghem, Bruynseels and Gaeremynck, 2018), namely, that CEOs at larger firms and CEOs at firms with stronger financial performance generally receive higher compensation.Models 2 (OLS) and 4 (fixed effects) present regression results testing H1, which predicted that stronger faultlines would lead to higher CEO compensation if the CEO was in the subgroup of the compensation committee members.In both models, the coefficient for the interaction term was positive and significant (β = 0.683, p = 0.023 in Model 2; β = 0.467, p = 0.015 in Model 4), whereas the main effects of social category faultlines were negative and significant (β = −0.439,p = 0.002 and β = −0.156,p = 0.043, respectively).These results support H1.When the CEO was not in the same subgroup as the compensation committee members, we noted a negative effect of social category faultlines: stronger faultlines were associated with lower pay.Using the fixed-effect results, a one standard deviation increase in social category faultline strength implies a 2.62% ($154,430) increase in CEO compensation from its mean value for CEOs in the subgroup of the compensation committee members.To facilitate the interpretation of these results, we provide a graphical representation of this relationship in Figure 1.This illustrates that the effect of subgroups is highly contingent upon the subgroup to which the CEO belongs.
Table 5 reports the regression results for the effects across joint tenure.The results indicate that stronger faultlines were associated with higher CEO compensation when the CEO was part of the compensation committee's subgroup, but only in the case of high joint tenure (β = 1.118, p = 0.021 in Model 6; β = 0.889, p = 0.081 in Model 8). 8When joint tenure was low, strong faultlines were not associated with higher CEO pay if the CEO was part of the compensation committee subgroup (β = 0.278, p > 0.10 in Model 2; β = 0.317, p > 0.10 in Model 4).To compare coefficients across the two groups, we used the Stata command suest to test whether the coefficients in Models 2 and 6 were equal.The results indicate that the difference was significant at the 8 C. Antoons and A. Vandebeek  10% level (p = 0.080).When comparing the coefficients in Models 4 and 8, we found that the difference was significant at the 5% level (p = 0.040).These results support H2.

Additional analyses
Alternative compensation measures.Increased CEO pay may be justified by a firm's improved performance; then, the documented associations would be in line with efficient contracting (Faleye, Hoitash and Hoitash, 2011).Several tests were conducted to test this hypothesis.First, we examined whether social category faultlines were positively associated with excess CEO compensation when the CEO belonged to a compensation committee subgroup.To do so, we first estimated a total compensation model by regressing the natural logarithm of CEO compensation on the observable economic determinants of CEO pay (Faleye, Hoitash and Hoitash, 2011). 9The residuals of this regression Note: * , ** and *** represent significance at the 10%, 5% and 1% levels, respectively.The dependent variable is the natural logarithm of CEO compensation.Models 1 and 2 were estimated using pooled OLS with standard errors computed according to White (1980) to account for possible heteroscedasticity.Models 3 and 4 are estimated using firm fixed effects.The T-statistics are presented in parentheses.6 reports the estimation results.Next, we repeated our main analyses using the residuals of Model 1 as a measure of excess compensation and report the results for Models 2 and 3 in Table 6.We found that stronger faultlines were associated with higher excess compensation when the CEO was included in the subgroup of compensation committee members.In sum, our results indicate that the effects documented in the main analyses are unlikely to be owed to efficient 10 Untabulated tests showed that if the CEO was included in the subgroup of the compensation committee, stronger faultlines were associated with lower delta and vega, indicating that pay-for-performance sensitivity and risk-taking incentives were weakened.Additionally, when examining the positive residuals from Model 1 in Table 6 (i.e.CEO overpayment), we found that if the CEO was included in the subgroup of the compensation committee, stronger faultlines were associated with an increased likelihood of overpayment (untabulated).
contracting but rather highlight the effects of social identification and categorization due to the existence of faultlines and subgroups. 11 Board monitoring incentives.Next, we examined whether the positive association between social category faultlines and CEO compensation when the CEO belonged to the compensation committee subgroup varied with the strength of board monitoring incentives.Following earlier research (e.g.Ahmed and Duellman, 2013;Hope and Lu, 2020), we established five board characteristics to represent directors' monitoring incentives.We classified observations as firms with 'strong board monitoring incentives' if they met at least three of the following criteria: (1) a higher percentage of independent directors than for the median firm in the sample, (2) no CEO duality, (3) a lower number of average additional directorships held by board members than for the median firm in the sample, (4) a higher percentage of outside director ownership than for the median firm in the sample and ( 5) a lower proportion of directors appointed after the CEO took office than for the median firm in the sample.Of the 10,309 observations in the sample, 12 5014 were classified Note: * , ** and *** represent significance at the 10%, 5% and 1% levels, respectively.The dependent variable is the natural logarithm of CEO compensation.Models 1−2 and 5−6 were estimated using pooled OLS with standard errors computed according to White (1980) to account for possible heteroscedasticity.Models 3−4 and 7−8 were estimated using firm fixed effects.In Models 1−4, we examined the subsample of observations with joint board tenure lower than the median.In Models 5−8, we examine the subsample of observations with a joint board tenure higher than the median.The T-statistics are presented in parentheses.
The T-statistics are presented in parentheses.
as having strong board monitoring incentives.Table 7 reports the results of the split-sample analyses based on this measure.The results show that these effects were mainly present in the subsample of observations with weak board monitoring incentives.That is, we found that stronger faultlines were only associated with higher CEO compensation when the CEO was part of the compensation committee subgroup in the case of weak board monitoring incentives (β = 1.076, p = 0.026 in Model 1; β = 0.644, p = 0.027 in Model 3).When board monitoring incentives were strong, strong faultlines were not associated with higher CEO pay if the CEO was part of the compensation committee subgroup (β = 0.014, p = 0.976 in Model 2; β = −0.245,p = 0.487 in Model 4). 13 13 When testing whether coefficients across subsamples were equal, we found that the coefficients for the regression term are significantly different between columns 1 and 2 (p = 0.096) and columns 3 and 4 (p = 0.012).

Robustness checks
Stricter CEO in same subgroup measure.It is possible that all compensation committee members belong to different subgroups, and our measure would equal 1 if one of these subgroups includes the CEO (i.e. the CEO and one compensation committee member are part of the same subgroup).As a robustness check, we included a more conservative measure: we only considered observations in which at least half of the compensation committee members belonged to the subgroup that contained the CEO to have a value of 1.All inferences remain (untabulated).
Faultlines versus standard diversity measures.We also tested whether there were any effects of traditional diversity measures by replacing our faultlines with two distinct diversity indices based on board directors' gender, age and compensation committee membership.Specifically, to construct the first index, we normalized each component by its mean and standard deviation.Note: * , ** and *** represent significance at the 10%, 5% and 1% levels, respectively.Models 1 and 2 were estimated using pooled OLS with standard errors computed according to White (1980) to account for possible heteroscedasticity.Models 2 and 4 were estimated using firm fixed effects.Firms are classified as 'strong board monitoring incentives' observations if the firm meets three or more of the criteria: (1) a higher percentage of independent directors than for the median firm in the sample, (2) no CEO duality, (3) a lower number of average additional directorships held by board members than for the median firm in the sample, (4) a higher percentage of outside director ownership than for the median firm in the sample and (5) a lower proportion of directors appointed after the CEO took office than for the median firm in the sample.In Models 1 and 3, we examine the subsample of observations with weak board monitoring incentives (i.e.meets fewer than three of the above criteria).In Models 2 and 4, we examine the subsample of observations with strong monitoring incentives.The T-statistics are presented in parentheses.
calculated as the sum of the three normalized characteristics (Bernile, Bhagwat and Yonker, 2018).Second, we constructed a diversity index using Blau's index of heterogeneity, following the approach of Harjoto, Laksmana and Lee (2015).Specifically, for each of the three characteristics, we calculated Blau's indices as 1 -Pi 2 , where P is the proportion of directors in a specific category and i is the number of categories (Blau, 1977).
To do so, we divided age into the following categories: younger than 40 years, 40−49, 50−59, 60−69 and 70 years and older.The resulting indices had values between 0 and 1; that is, when there was only one category within a diversity dimension (e.g.only female directors on a board), the value of the index was 0, suggesting that the group was perfectly homogeneous.A value closer to 1 indicates a more heterogeneous group.Because individual indices have different ranges (dependent on the number of categories of that characteristic), we standardized each index to a value between 0 and 1 by dividing each index by its maximum.We then summed the standardized indices to construct our second composite diversity index, Blau's index.The results are presented for Models 1−4 in Table 8.None of the interaction terms was statistically significant, indicating that our faultline measure is distinct and captures effects beyond standard diversity measures.We also replaced our faultline measure with average CEO-board similarity in Models 5 and 6 of Table 8, again finding no significant results.
Endogeneity tests.Although our main tests address endogeneity by including lagged faultline measures and multiple econometric specifications, we took additional steps to ensure that endogeneity did not induce biased coefficient estimates (Wintoki, Linck and Netter, 2012).First, we utilized a dynamic generalized method of moments (GMM) estimator (Arellano and Bond, 1991), allowing us to control for the dynamic nature of the relationship between board composition and CEO compensation (dynamic endogeneity) and address unobservable heterogeneity.The model included twoperiod lagged CEO pay values.We assumed that all explanatory variables were endogenous and used lags of t -3 and t -4 as instruments.We specified a two-step GMM estimator and applied Windmeijer-corrected standard errors to control for downward bias in the estimator (Windmeijer, 2005).The results of the system GMM estimation are presented in Table 9.The positive and significant interaction terms in Models 1 and 3 further support our findings.Second, we used an entropy balancing approach to alleviate the concern that firms in which the CEO was included in the compensation committee subgroup were systematically different from firms in which the CEO was not, allowing us to control for the selection of observables (Hainmueller, 2012;Lennox, Francis and Wang, 2012). 14We used the covariate means and variances of all control variables to balance the samples.15Note: * , ** and *** represent significance at the 10%, 5% and 1% levels, respectively.Models 1, 3 and 5 were estimated using pooled OLS with standard errors computed according to White (1980) to account for possible heteroscedasticity.Models 2, 4 and 6 were estimated using firm fixed effects.The T-statistics are presented in parentheses.Panel A of Table 10 presents descriptive statistics for the treatment and control groups before and after entropy ables, except for the variables social category faultlines and CEO in same subgroup.The results were robust to using entropy balancing to eliminate differences in the first three moments of the distribution of all control variables.Our results are available upon request.
balancing.The results shown in Panel B of Table 10 confirm these main findings.

Discussion
Boards of directors are increasingly being held accountable for executive compensation, as determining compensation packages is within the purview of their con- Note: Panel A presents the differences in the covariates between the treatment and control groups before and after entropy balancing.Panel B presents the results for the entropy-balanced sample using the reweighted observations.Models 1, 3 and 5 were estimated using pooled OLS with standard errors computed according to White (1980) to account for possible heteroscedasticity.Models 2, 4 and 6 were estimated using firm fixed effects.The T-statistics are presented in parentheses.
trol role (Daily et al., 1998).It has been argued that the CEO compensation-setting process relies heavily on compensation committee deliberations and that their interpretations can be influenced by psychological factors because of the ambiguous nature of CEO pay decisions (Lewellyn and Muller-Kahle, 2021;O'Reilly and Tushman, 2008).However, previous executive compensation research has not examined how the existence of faultlines and subgroups within the board could influence the compensation-setting process.If board members' di-versity attributes align, a strong faultline may emerge, which could lead to salient identity-based subgroups, which, in turn, may become a basis for self-identity and social categorization processes (Hogg and Terry, 2000).Therefore, we used the concept of faultlines to identify identity-based subgroups within the board and investigate how the composition of these subgroups affects decision-making regarding CEO compensation.Drawing on social identity and self-categorization theories to explain how CEO evaluation differs for in-group C. Antoons and A. Vandebeek members (vis-à-vis out-group members), we predicted that CEO compensation would be higher if the CEO and compensation committee members formed a subgroup because of strong faultlines.As individuals tend to have a more favourable view of in-group members, CEOs included in the compensation committee member subgroup tend to be advantaged in receiving compensation.Subgroup members may have more incentives to go along with compensation arrangements that favour the CEO, while social and psychological mechanisms such as friendship, loyalty and collegiality may be more common within the subgroup, thus making those CEOs better positioned to negotiate compensation arrangements that are in their favour (O'Reilly III and Main, 2010).Our results strongly suggest that in the case of strong faultlines based on gender, age and compensation committee membership, CEO compensation is higher if the CEO is part of the compensation committee members' faultline-based subgroup, whereas CEO compensation is lower if the CEO is part of a different subgroup.Furthermore, we found that if a CEO is part of the faultlinebased subgroup of compensation committee members and the joint CEO-director tenure is higher, stronger faultlines are associated with higher CEO compensation.CEOs that are not included in the faultline-based subgroup will, in the case of strong social category faultlines, always be considered out-group members because of the salient basis for in-and out-group categorization that social category membership provides (Bezrukova et al., 2009), thus leading to lower compensation.
This study contributes to existing literature in several ways.First, we add to the CEO compensation literature by identifying faultlines and their corresponding subgroups as important factors that can influence CEO compensation.CEO compensation research has examined the composition of the compensation committee and its relationship with CEO compensation (Newman and Mozes, 1999).Our study takes a different path and finds that compensation committee members can socially identify with CEOs if they are part of a strong identity-based subgroup based on the basis of their committee membership, gender and age.Thus, we add to the corporate governance literature by pointing to the effect of committee membership on relationships and structures within the board and vis-à-vis the CEO (Kolev et al., 2019).Furthermore, we provide valuable insights into the influence of faultlines and, particularly, into subgroup formation on corporate boards.Our study also adds to the faultline literature by elaborating on the concept of faultlines in the context of corporate boards, which has been understudied.Moreover, we introduce board committee membership as an important identity-based attribute and investigate whether the alignment of committee membership and two other social category attributes affects board decision-making.Our study also responds to calls within faultline the-ory to integrate the concept of faultlines within different team settings (Murnighan and Lau, 2017).Finally, our study examined the effect of faultlines and subgroup inclusion on joint tenure, thereby responding to calls for more research that examines the effect of faultlines over time (Murnighan and Lau, 2017;Van Peteghem, Bruynseels and Gaeremynck, 2018).
Our study also has implications for practitioners because our results reveal the potentially negative effects of board committee composition, because board committee membership can lead to the emergence of faultlines.While installing board committees can help directors perform their board roles adequately, one must be attentive to the potential decision-making biases that may result from the existence of subgroups.Therefore, it may be crucial to strive for a committee composition that minimizes internal homogeneity in each subgroup while simultaneously promoting between-subgroup heterogeneity and considering the important position of the CEO as a subgroup member.
Finally, our study had some limitations that may guide future research.First, although we examined the CEO's inclusion within a particular subgroup, we did not examine all the subjective psychological factors involved in the social relationships between directors and the CEO.It would be interesting to examine whether additional social factors are involved.Qualitative research designs could be used to examine the internal dynamics of these relationships.For example, by observing board meetings, we can gain a better understanding of the deep-level social and psychological processes underlying faultlines and identity-based subgroups.Second, while we specifically chose to examine the CEO's role in CEO compensation, the board chair within the subgroups may also play a role, as the chair is in fact positioned as the leader of the board (Banerjee, Nordqvist and Hellerstedt, 2020;Meyer et al., 2015).Third, the results show that joint tenure has an important moderating effect.Future research could examine other conditions that influence the effects of faultlines on corporate boards.For example, activities that reinforce the repeated interactions of a subgroup over time will make members more accustomed to working in and interacting with their subgroups.Such activities could include regular board committee meetings that include only members of a specific board committee and the CEO.

Conclusion
This study used the concept of faultlines to identify identity-based subgroups within corporate boards and investigated how the composition of these subgroups affects CEO compensation decision-making.The results show that if a CEO is part of a faultline-based sub-group of compensation committee members, stronger faultlines are associated with higher CEO compensation.However, if the CEO is not included in this subgroup, stronger faultlines are associated with lower CEO compensation.Furthermore, we found that if a CEO is part of the faultline-based subgroup of compensation committee members and the joint CEO-director tenure is higher, stronger faultlines are associated with higher CEO compensation.In this way, our study enriches CEO compensation, governance, and faultline literature.

Figure 1 .
Figure 1.The moderating effect of CEO in same subgroup on the relationship between social category faultlines and CEO pay Furthermore, in Models 4−7, we decomposed CEO compensation into cash pay (salary and bonuses) and equity pay.The results indicate that the effects documented earlier were mainly owed to increased equity compensation. 10Finally, in Models 8−9, we examined the effects on pay slice, which is the percentage of CEO pay in the total compensation of the top five executives.The extant literature (e.g.Bebchuk, Cremers and Peyer, 2011) suggests that an increased pay slice in the C-suite signals inferior efficiency in CEO compensation design.The results from Model 9 indicate a positive and significant interaction term (β = 0.084, p = 0.084).

Table 2 .
7Including firm fixed effects filters out potential omitted variable bias originating from fundamental firm characteristics, while controlling for time-invariant heterogeneity.Summary statistics © 2024 The Author(s).British Journal of Management published by John Wiley & Sons Ltd on behalf of British Academy of Management.14678551, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/1467-8551.12851by Erasmus University Rotterdam Universiteitsbibliotheek, Wiley Online Library on [11/07/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Effects of Subgroups and CEO Inclusion 7 Note: N = 14,253.This table reports the mean, standard deviation and minimum, median and maximum values of each variable.

Table 4 .
The moderating effect of CEO in same subgroup on CEO pay 9 Specifically, we included firm size, return on assets, stock market return, book-to-market, standard deviation of return, leverage, log of firm age, Tobin's Q and year and firm fixed effects as determinants.The results were similar using industry instead of firm fixed effects.© 2024 The Author(s).British Journal of Management published by John Wiley & Sons Ltd on behalf of British Academy of Management.14678551, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/1467-8551.12851by Erasmus University Rotterdam Universiteitsbibliotheek, Wiley Online Library on [11/07/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Effects of Subgroups and CEO Inclusion 9

Table 6 .
Alternative compensation measures , 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/1467-8551.12851by Erasmus University Rotterdam Universiteitsbibliotheek, Wiley Online Library on [11/07/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License The resulting diversity index, Diversity index, was then © 2024 The Author(s).British Journal of Management published by John Wiley & Sons Ltd on behalf of British Academy of Management.14678551

Table 8 .
Index measures of diversity

Table 9 .
System GMM , ** and *** represent significance at the 10%, 5% and 1% levels, respectively.This table presents the results of the system GMM model.We instrumented the endogenous variables using lags at t -3 and t -4.AR(1) and AR(2) are tests of the first-and second-order serial correlations in the first-differenced residuals.The Hansen J-test is a test for over-identifying restrictions.p-values of these tests are reported.The T-statistics are presented in parentheses. *