Is Busy Really Busy? Board Governance Revisited

Journal of Business Finance & Accounting

Journal of Business Finance & Accounting, 40(9) & (10), 1221–1246, November/December 2013, 0306-686Xdoi: 10.1111/jbfa.12051

Is Busy Really Busy? Board GovernanceRevisited

CHRISTIAN ANDRES, INGA VAN DEN BONGARD AND MIRCO LEHMANN∗

Abstract: We investigate the relationship between firm governance and the board’s positionin the social network of directors. Using a sample of 133 German firms over the four-year periodfrom 2003 to 2006, we find that firms with intensely connected supervisory boards are (1)associated with lower firm performance, and (2) pay their executives significantly more. Weinterpret these results as evidence of poor monitoring in firms with directors who are moreembedded in the social network. In both cases, simple measures for busy directors that wereused by other studies in the past fail to show any significant pattern. The findings suggest thatthe quality and structural position of additional board seats may play a bigger role than simplythe number of board appointments.

Keywords: director network, busy board, agency theory, executive compensation

1. INTRODUCTION

A growing literature on board characteristics investigates the question of whetherdirectors with multiple board appointments are too busy to perform their moni-toring duties (Ferris et al., 2003; and Fich and Shivdasani, 2006). The “busynesshypothesis” associates firms that have busy directors with weak corporate governance,the assumption being that directors who serve on a large number of boards maybecome overcommitted and unable to adequately comply with the requirements oftheir position as monitors of management. Fich and Shivdasani (2006) confirm thisassertion, showing firms with busier boards to exhibit significantly lower market-to-book ratios, whereas Ferris et al. (2003) find no evidence of a relationship betweennumber of board appointments and firm performance.

∗Christian Andres is at WHU – Otto Beisheim School of Management, Burgplatz 2, 56179 Vallendar,Germany. Inga van den Bongard is at the University of Mannheim, Schloss, 68131 Mannheim, Germany.Mirco Lehmann is at the University of Bonn, Adenauerallee 24–42, 53113 Bonn, Germany. The authorswould like to thank an anonymous referee, the Editor (Martin Walker), Ingolf Dittmann, Kevin McLaughlin,Ernst Maug, Stefan Petry, Harley E. Ryan Jr., Erik Theissen, seminar participants at the University ofMannheim, the University of Innsbruck, the University of Hamburg, and the University of Rotterdam,participants of the 2009 European Financial Management Association Meeting in Milan, the 2009 AnnualMeeting of the German Finance Association in Frankfurt, and the 2010 Annual Meeting of the FinancialManagement Association in New York for their helpful comments. (Paper received January, 2013, revisedversion accepted October, 2013).

Address for correspondence: Christian Andres, WHU – Otto Beisheim School of Management, Burgplatz 2,56179 Vallendar, Germany.e-mail: [email protected]

C© 2013 John Wiley & Sons Ltd 1221

1222 ANDRES, VAN DEN BONGARD AND LEHMANN

In this paper, we revisit the association between board busyness and firm gover-nance. We maintain that prior approaches, having been largely based on number ofboard seats, ignore central aspects of “busyness”, which concept has more than onedimension. We suggest that a crucial determinant of busyness is how “embedded”(Granovetter, 1985, 1992) directors are in their social networks, that is, the “extentto which [a director’s] economic action is embedded in structures of social relations”(Granovetter, 1985, p. 481). We argue that counting numbers of board seats orconnections to other directors is inadequate, that a more effective measure of busynessincorporates a detailed characterization of a director’s social network including anal-yses of individual ties and overall network architecture. More specifically, we proposethat an appropriate analysis of busyness needs to reflect not only directors’ immediateconnections to other directors within their social networks, but also connectionsamong the directors to whom they are connected. We argue that maintaining andadhering to the demands of a large social network can consume substantial temporaland cognitive resources, rendering intensely connected directors exceedingly busy andcompromising their ability to serve as efficient monitors.

To explore the foregoing hypothesis, we build on notions and methods derivedfrom social capital theory and social network analysis that facilitate the characteri-zation of directors’ social networks with respect to busyness. The variables we useattempt to capture the demands of such networks for time and cognitive resources,and may be interpreted as a proxy for time- and attention-consuming social obligationsinherent in a director’s network. For example, directors’ formal ties with one anotheroften reflect informal ties (Hwang and Kim, 2009), such as joint memberships in golfclubs or charitable organizations. Being a well-connected director probably also entailsconnections to important agents outside the boardroom, as in politics, media, sportsand so forth. Arguably, directors who spend considerable time catering to such socialdemands have less time to devote to monitoring duties.

A caveat is in order at this point. Previous literature has documented benefits ofsocial networks to agents (e.g., Cohen et al., 2010; and Horton et al., 2012), such asprivileged and timely access to information or augmentation of social capital (Adlerand Kwon, 2000). If such advantages enable directors to better fulfil their boardobligations, intense social networks may benefit corporate governance. There is noevidence, however, that such advantages accrue to directors through the network ofboard directorships, particularly in the context of a two-tier system as in Germany.Adams and Ferreira (2007, p. 235) argue that a “dual board structure allows for thecleanest separation of the board’s two roles”, that is, advising and monitoring. Thisbeing the case, it is unclear whether information that passes through the network ofdirectors1 is equally useful to directors whose primary role is to monitor management(as stipulated in article 111 of the German Stock Corporation Act). We suggest thatwhen participation in a social network simultaneously elevates busyness and bestowsinformational advantages, the busyness effect should dominate in this environment.

To empirically investigate the impact of busyness on firm governance, we collectdata on the social network of all members of the supervisory boards of all publiclytraded German firms (less financials) included in the DAX, MDAX and SDAX indices

1 For ease of notation, the terms “director” and “member of the supervisory board” are used inter-changeably throughout the paper. Members of the executive board are referred to as “management” or“executives”.

C© 2013 John Wiley & Sons Ltd

IS BUSY REALLY BUSY? BOARD GOVERNANCE REVISITED 1223

over four consecutive years (from 2003 to 2006). We map for each year the connectionsamong all of the sample firms’ directors. Our dataset includes roughly 1,600 directors,and reflects approximately 35,000 connections, per year. We aggregate at the board-level different measures used to calculate individual director’s busyness, and explorein a multivariate panel framework whether firms with busy boards exhibit weaker firmgovernance.

Using fixed-effects panel regressions, we find firms with strongly connected su-pervisory boards to tend to have significantly lower values of Tobin’s q. We furtherdocument the average remuneration of executives in firms with strongly connecteddirectors to be significantly higher. Simple measures of director “busyness” used inprior research fail to show any significant pattern in either case. We interpret ourfindings as evidence of weaker governance and poorer monitoring in firms that havedirectors who play an important role in their social network.

Our results make a number of important contributions. First, our study informsthe debate on the effects of multiple board appointments, supporting Shivdasani andYermack (1999), Core et al. (1999) and Fich and Shivdasani (2006), who find busydirectors to be associated with less effective monitoring. Our results are counter tothose of previous studies that ascribe positive value to multiple board appointments.Mace (1986), for example, argues that the number of board seats can signal directorquality, Brown and Maloney (1999) use number of board appointments as a proxyfor reputational capital, and Perry and Peyer (2005) argue that “sender” firms canbenefit from additional directorships that expose their directors to managementstyles or strategies used in other firms. Our findings also have implications for thenotion of imposing arbitrary limits on the number of directorships that can be held,indicating that such limitations are insufficient to prevent directors from becomingovercommitted.

Second, previous research in this area has been conducted almost exclusively onUS data, where (inside) firm governance is based on the one-tier system. Becausecorporate boards in this environment consist of inside directors who run the firm andoutside directors who are not part of the management team, the literature focuses onoutside directors with the premise that inside directors do not monitor management(Fich and Shivdasani, 2006). Because in the one-tier setting, management mightprefer to appoint directors to the board from the same social circle (Mace, 1986;and Barnea and Guedj, 2009), network measures capture how well connected notonly an outside (i.e., monitoring) director, but also management, is. This complicatesthe interpretation of a relationship between network variables and firm governance,as slack monitoring may reflect not so much exceeding “busyness” as members ofthe board simply being more “lenient” towards management by way of reciprocatingor positioning themselves for other board seats (Wade et al., 1990). We use datafrom Germany, in which governance structures are based on the two-tier systemwith an independent supervisory board. Most importantly, German law explicitlyforbids executives from nominating candidates for the supervisory board.2 This legalenvironment enables us to examine the effect on firm governance of position in thesocial network independent of interaction with executives.

2 According to paragraph 124(3) of the German Stock Corporation Act (AktG), executives are not allowedto nominate members of the supervisory board. Monitoring the executive board is the main responsibilityof the supervisory board.



Third, we add to the growing literature that attempts to incorporate mathematicalconcepts of sociology (see Scott, 2000, for an overview) into assessments of the impactof network structures on finance. Horton et al. (2012), using methods of socialnetwork analysis to investigate whether directors’ social networks provide valuableresources to their firms, find them to be rewarded for their connectedness, and theirconnectedness to be positively related to future firm performance. Barnea and Guedj(2009), using measures that account for directors’ importance in their social networks,find CEO remuneration to be higher and turnover less sensitive to firm performancein firms with more connected directors. We draw on social capital theory and socialnetwork analysis for methods and insights with which to extend previous work onbusy boards (Ferris et al., 2003; and Fich and Shivdasani, 2006). Our findings suggestthat detailed characterizations of directors’ social networks capture busyness moreadequately and comprehensively than previously applied definitions based on simplytallying numbers of board seats.

The remainder of the paper is organized as follows. In the next section, we developour central argument, building on notions drawn from social capital theory and socialnetwork analysis. The dataset and network measures are described, and our descriptivestatistics presented, in section 3. Section 4 reports results of the multivariate panel dataanalysis. Section 5 concludes.

2. SOCIAL NETWORKS AND BUSY BOARDS

Directors are linked both directly, for example, through boards on which theyjointly sit, and indirectly, through, for example, common acquaintances. We arguethat maintaining connections and satisfying associated obligations to other directorsdeplete limited temporal and cognitive resources directors need to monitor manage-ment, and that analyzing these connections affords a way to characterize directors’busyness.

Multiple board assignments are associated with a large number of direct obligations,most notably attending a variety of board meetings, reading corporate reports, etc.,thus rendering directors busy. However, a multitude of board seats likewise mirrorbeing intensely connected to the network of interlocking directorships. These intenseconnections reflect the direct obligations from sitting on a number of boards, butthey similarly reflect the indirect connections and the resulting obligations, forexample, joint memberships in golf clubs or a charitable organization (Hwang andKim, 2009). In addition, they are often associated with connections to importantagents outside the boardroom, e.g., in politics, media or sports. We maintain thatbecause such social demands consume considerable time, too many may driveselective allocation (Ferriani et al., 2009) which may result in neglect of monitoringduties.

In the presence of limits to human cognition (Kahneman, 1973), exposure toan abundance of information through intense network ties may impair informationprocessing in a variety of ways (for an overview, see Eppler and Mengis, 2004)including inability to identify relevant data (Schneider, 1987), inordinate attentionto irrelevant issues, overlooking large amounts of information, and “greater toleranceof error” (Sparrow, 1999, p. 144). Most recently, Oldroyd and Morris (2012) theorizedthat cognitive limitations restrict the value of network ties because the voluminous



Figure 1Network of All Directors

Notes:This graphic shows connections between all directors of 133 German firms for the year 2003. The totalnumber of directors is 1,654, the number of connections 35,106.

information generated by intense network connections, unmanaged, can devolveinto information overload, with adverse consequences for agents’ performance. Wehypothesize an analogous debilitating effect of excessive information gathered fromnetwork connections on directors’ monitoring abilities. We further posit that thenumber and quality of network connections determine the intensity of contacts andinformation and resulting busyness. We argue, for example, that a director connectedto a large number of well-connected directors is busier than a director whose networkconnections have few further connections.

The network on which our analysis is based is characterized by a strong core-periphery structure with dense, cohesive sub-groups (see Figure 1). In such environ-ments, the quality of higher-order (triadic) relations among directors, that is, relationsamong the directors to whom a director is related, are of particular importance.The literature on social capital (e.g., Coleman, 1988) has emphasized the beneficialeffects of network “closure”, by which is meant the existence of connections amonga director’s connections, because it facilitates enforcement of social norms. Consider,for example, three actors, A, B and C, the latter two having in common a connectionto A, B and C, if they are not connected to each other, have no means of imposing



sanctions on A for deviating from any social norm. But if a link is established betweenB and C, resulting in a so-called “closed” triad in which all actors are connected to eachother, B and C can combine forces to collectively sanction A. Under the assumptionthat monitoring management constitutes a social norm, one might expect directorsto devote more time to the activity in the context of a relatively “closed” network. Asystem with strongly embedded interactions offers a further incentive in that failure toproperly monitor is exposed to other directors (Raub and Weesie, 1990). Still, thereare some arguments that closure may be a detriment to efficient monitoring. First,closure may restrict directors in their ability to freely allocate resources among theirmonitoring duties. This is a particularly likely scenario among directors with largenumbers of connections in a network in which additional restrictions should, ceterisparibus, have more intense effects.

Second, we suggest that a more important avenue by which closed networks canaffect busyness is by exacerbating the social obligations that arise within the network.If, for example, the aforementioned directors B and C each host a social event, directorA will feel more compelled to attend both if B and C know each other than if they donot. Thus, by imposing more social obligations network closure may translate intoincreased busyness.3 The greater the number of a director’s connections, the moredetrimental this effect is likely to be. The propensity for members of a group to “feelobligated to help others [ . . . ] [as] a societal norm prescribes socially responsiblebehaviour” (Cialdini and Trost, 1998, p. 157), applied to the supervisory board, mighttake the form of members feeling obliged to assume tasks and respond to requests foradvice, which behaviour might be driven by the notion of reciprocity among actorswho feel indebted to return favours received in the past (Gouldner, 1960). Beinggenerous, moreover, enhances one’s social status, whereas seeking more often thanproviding help can engender a negative public image (Lee, 1997; and Flynn et al.,2006). But because time spent helping others is time not devoted to meeting one’sown responsibilities (Hansen et al., 2001), network-related activities may turn into atime-consuming social liability.

A high degree of closure can affect directors’ monitoring ability in a variety of ways.The social indebtedness implicit in close-knit social environments, for example, canhave adverse effects beyond increasing busyness (Flynn et al., 2006). In the foregoingexample, effective monitoring of directors B and C might be impeded by potentialconsequences for director A’s social life. The lines between social and professionalenvironments, because of the high transitivity of the connections, are likely to becomeblurred, dissuading directors from taking actions that are counter to the interests ofother board members. In this sense, network closure might also foster unwarrantedleniency among directors.

Our findings thus suggest that embeddedness in a social network of directorseffectively limits monitoring ability. Maintaining intense network connections mayconsume time and cognitive resources, and closed network structures restrict and limitdirectors’ ability to freely allocate resources. The metrics we employ attempt to capturethese different dimensions of being part of a network of directors.

3 Note that much of what might be subsumed under “obligations” could equally be called perquisites.From a director’s point of view, for example, attending a social event may be beneficial. This does not alter,however, the fact that such activities reduce the time available for a director’s monitoring duties.



3. DATASET AND NETWORK MEASURES

(i) Sample Selection

The sample for this investigation includes all 150 companies that were part of the threelargest German stock indices as of 31 December 2002 (30 DAX, 70 MDAX and 50SDAX).4 Excluding all companies that abandoned their listing status during the four-year period (2002–06) during which we followed them left us with a balanced panel of133 (non-financial) firms, or 532 firm–year observations.5 Accounting data are fromDatastream and Hoppenstedt Aktienfuhrer.6 Information on the composition of thesupervisory board was manually collected from the annual reports of the firms in thedataset. Panel A of Table 1 presents descriptive statistics for several firm characteristics.

(ii) Measurement of Multiple Directorships

According to section 125 of the German Stock Corporation Act (AktG), publicly tradedfirms must disclose all directorships held by members of their supervisory boards inother (for-profit) firms. We therefore count all directorships in public and non-publicfirms.7 This information enables us to detect connections between directors outsidethe 133 firms in our sample. For example, although “FC Bayern Munchen AG” is nota sample firm, four of the directors in our sample are connected by virtue of sitting onits board.

In the two-tiered German corporate board system, members of the supervisoryboard oversee the actions of the executive board, which is to say, management.Among other duties, members of the supervisory board appoint, dismiss, and setremuneration for executives, and approve management decisions.8 Employees areallocated control rights through the supervisory board by codetermination laws thatapply to public and private companies with 500 or more employees. Firms with morethan 500 but fewer than 2,000 employees must allocate one-third, firms with more than2,000 employees one-half, of supervisory board seats to employee representatives.9

Although employee representatives are often union members (in the case of largefirms, high ranking union officials), in fact, any employee can be elected to thesupervisory board. About 55% (25%) of the firms in our sample are subject to one-half (one-third) representation, and 20% have no employee representatives on theirsupervisory boards. The data in our main analysis are based on all members of the

4 The DAX (largest firms), MDAX (mid caps) and SDAX (small caps) are Deutsche Borse’s three majorindexes of firms in the classic sectors. These companies, being part of the Prime Standard segment, mustfulfil the highest transparency requirements in the EU including publishing on a quarterly basis financialreports in both German and English, applying international accounting standards (IFRS/IAS or US-GAAP),releasing a financial calendar, conducting at least one analyst conference per year, and publishing their adhoc disclosures in English as well as in German.5 Most firms terminate their listings following a squeeze-out by the dominant shareholder (10 firms). Onlyfour firms dropped out of the sample due to insolvency.6 The Hoppenstedt Aktienfuhrer is an annual publication that provides detailed information (e.g.,ownership structure, balance sheet information) on German listed firms.7 If voluntarily reported, we exclude board appointments in non-profit organizations, trusts and charitableorganizations for reasons of data consistency.8 See the German Stock Corporation Act (AktG) for a comprehensive description of the rights and dutiesof the supervisory board.9 See Gorton and Schmid (2004) for a detailed description of the system, and legal background, ofcodetermination.



Table 1Data Description

Variable Mean Median SD

Panel A

Market value of equity (Mio. €) 5,382 759 11,490Book value of assets (Mio. €) 35,602 1,317 25,167Leverage 0.66 0.68 0.18Tobin’s q 1.48 1.16 1.03Age (years since incorporation) 79.54 78.50 56.05Listing age (in years) 29.74 13.00 36.76Return on assets 3.82% 3.07% 7.04%Executive remuneration (1,000 €) 1,100.4 820.8 970.0Without options (1,000 €) 1,009.4 765.5 817.0

Panel B

Board size 13.58 13 6.06Ordinary board members 7.78 7 2.51Employee representatives 5.80 6 4.71Number of directorships 3.49 3.17 1.70

Including employee representatives 2.72 2.40 1.29Percentage of busy directors 52.44

Including employee representatives 33.07Percentage of busy boards 46.78

Including employee representatives 19.36

Notes:This table presents descriptive statistics for 532 annual observations of 133 German firms for the period2003–06. Leverage is defined as book value of total debt divided by book value of total assets. Tobin’s q iscalculated as market value of equity at the end of the year plus book value of debt divided by book value ofassets at the end of the year. Executive remuneration is the average per-capita executive remuneration peryear. Panel B provides data on 1,654 directors.

supervisory board. As a robustness test, we construct all network variables based onlyon shareholder representatives. The results, being qualitatively and quantitativelysimilar, are not tabulated.

Following Ferris et al. (2003) and the alternative methodology used by Fich andShivdasani (2006), we construct three variables that classify supervisory boards asbusy/not busy based on the number of board appointments held by members. Inline with these papers, and consistent with prior work by Core et al. (1999), weconsider as busy directors who hold three or more board appointments. Based onthis definition, 52.44% (excluding employee representatives) and 33.07% (includingemployee representatives) of the directors are classified as busy. Panel B of Table 1reports for our sample data the frequency of multiple board appointments.

The median board in our sample consists of 13 directors, seven shareholder, andsix employee, representatives. On average, directors hold 3.49 (excluding employeerepresentatives) and 2.72 (including employee representatives) directorships. In linewith Fich and Shivdasani (2006), we construct a dummy (0,1) variable set to 1 if 50%or more of a board’s shareholder representatives are busy. By this measure, about 47%of firms are governed by “busy boards”. Unlike Ferris et al. (2003), we do not find apositive correlation between multiple directorships and firm size.



(iii) Social Network Measures

A social network consists of multiple agents connected through such social relationsas direct contacts, group attachments or meetings (Scott, 2000). Network membersand their connections can be visualized as a structure of nodes and ties. In our analysis,nodes are the individual directors within a network, ties the relationships between them.Directors who serve on the same supervisory board are connected.

We collect for each firm and year in our sample, data on all directors who serve ona firm’s board.10 This yields a database of about 1,600 nodes and roughly 35,000 ties. Webuild an annual social matrix in which each director is represented by a column anda row. If two directors, i and j, serve on the same board, the value of the intersectionpoint, that is, cell (i,j), is 1, otherwise 0. Because relationships are always bilateral(director i knows director j, which implies that j must know i), this procedure resultsin a symmetric matrix, the diagonal of which (i.e., relation between i and i) is 0 bydefinition. Figure 1 visualizes the entire network of directors for the year 2003.11

We use four well-established measures to capture how embedded a director is inthe network. Our construction encompasses the overall architecture of a director’ssocial network, that is, we consider pairwise (dyadic) as well as higher-order (triadic)relations.

The first measure, Degree centrality (CD), is of the number of direct links (orcontacts) incident to a node (i.e., director), and is defined as:

CD(ni ) = di (ni ) =g∑

j=1

x ji =g∑

j=1

x ji (1)

where g is the total number of directors or nodes ni. The variable is further normalizedby dividing (1) by the number of possible connections of director ni with the otherg–1 nodes in the network. Theoretically, because it captures only the number ofconnections, this measure should be closest to the variable “busy board”. Degreecentrality describes how active agents are in the network. This measure is intuitiveand comparatively easy to implement, but its explanatory power is limited to thefirst network level, which means that two directors with the same number of directconnections will be assigned the same value CD(ni), irrespective of how well connectedtheir contacts are. The second measure, Eigenvalue centrality, addresses this issue.

Connectiveness, or Eigenvector centrality (CC), developed by Bonacich (1972, 1987),uses weighted scores that measure the “quality” of connections when assessing thecentrality of an agent, which is to say, the extent to which a director is connectedto other well-connected directors. Being connected to other well-connected agentsboosts not only a director’s centrality, but the centrality of other directors connectedto that director. Formally, the individual centrality of each director is computed as:

CC (ni ) =

g∑j=1

Wi j CC (nj )

λmax(2)

10 If a director leaves a board during a fiscal year, whether due to end of tenure, dismissal or death, weinclude both old and new director in the count for that year’s board.11 The Organizational Risk Analyzer (ORA) software v. 1.9.0, which has been developed by CASOS (2008)at Carnegie Mellon is used to visualize the social network.



where Wij stands for the intersection of row i and column j in the social matrixdiscussed above. Bonacich (1972) shows that there exists a positive Eigenvalue λ forevery Matrix W that results in a corresponding Eigenvector CC that consists only ofpositive values or 0. This condition is met for the largest positive Eigenvalue. LikeDegree centrality, Eigenvector centrality is normalized.

We calculate the local Clustering Coefficient (Watts and Strogatz, 1988) as a proxy forthe degree of triadic closure within a director’s network. Clustering Coefficient, definedas the probability that two randomly selected contacts of director A are also connectedto each other, measures the connectedness, or density, of an actor’s neighbourhood.In other words, it is the ratio of the number of pairwise relations between directorA’s contacts to the maximum possible number of relations among them. For example,director A’s neighbourhood, if director A is connected to directors B, C and D, hasthree potential pairs (BC, CD, BD). If only one pair is connected, for example, (BC),the clustering coefficient of director A is 1/3. The Clustering Coefficient can range fromzero (neither of A’s contacts is connected) to one (all of A’s contacts are connected).A larger Clustering Coefficient thus implies a higher degree of triadic closure within adirector’s neighbourhood.

We apply, as a final measure, the K-Core decomposition, which enables us to identifydense, cohesive sub-groups in the network. This procedure, proposed by Seidman(1983), is based on the minimum nodal degree within a sub-graph. A K-Core, then,is a sub-graph in which each node is connected to at least k other nodes. It can beidentified by recursively removing, or “peeling off”, all nodes with a degree smallerthan k until all remaining nodes in the sub-network have at least a degree of k. Thisapproach generates sub-graphs of gradually increasing cohesion. Hence, the higherthe K-Core score, the better connected an actor’s neighbours and the more restrictedthe actor’s actions. We normalize this measure by dividing by its largest value.

Our choice of the UCINET software package v6.171 to compute our networkmeasures is based on Huisman and van Duijn (2005), who compare different softwarepackages used for social network analysis. Table 2 lists the ten most central directorsin terms of Degree and Eigenvector centrality for each year of our sample period.12 Tofacilitate comparison of the data over different years, the figures in the table arecomputed relative to the size of the network.

The Degree measure, as can be seen in Table 2, is comparatively stable over time.The number of directorships does not seem to change dramatically among the best-connected directors. It becomes apparent, upon closer examination of Degree, thatthe best-connected directors are mostly chairmen of large DAX companies. Oursecond measure, Connectiveness, emphasizes the quality of connections. The exampleof Henning Schulte-Noelle (chairman of Allianz AG) is illustrative. In 2003, Mr.Schulte-Noelle was among the best-connected directors with respect to Degree andConnectiveness. When in 2004 he gave up several directorships and lost more than 50direct connections, Mr. Schulte-Noelle dropped off the list of directors with the highestDegree. Although Connectiveness also decreased, because he retained his positions at well-connected firms like E.ON AG and ThyssenKrupp AG, Mr. Schulte-Noelle remainedamong the best-connected directors throughout the 2004–06 period.

12 We do not tabulate Clustering Coefficient and K-Core because, owing to the definition of the variables, alarge number of directors have the largest possible values in each variable and year.



Table 2Most Central Directors

No. Name CD Name CC

Panel A: Top 10 Degree and Connectiveness for 2003 in Percent

1 Schneider, Manfred 9.56 Cromme, Gerhard 26.322 Kohlhaussen, M 9.01 Schulte-Noelle, H. 23.893 Cromme, Gerhard 8.83 Baumann, Karl-H. 23.344 Baumann, Karl-H. 8.59 Schneider, Manfred 21.785 Hartmann, Ulich 8.29 Voss, Bernd W. 21.526 Schulte-Noelle, H. 7.99 Hartmann, Ulrich 21.057 Voss, Bernd W. 7.74 Breuer, Rolf-E. 20.198 Walter, Bernhard 7.74 Kohlhaussen, Martin 20.169 Breuer, Rolf E. 7.68 Diekmann, Michael 17.0910 Schinzler, Hans-J. 7.56 Strube, Jurgen 16.82

Panel B: Top 10 Degree and Connectiveness for 2004 in Percent

1 Cromme, Gerhard 8.91 Cromme, Gerhard 28.112 Schneider, Manfred 8.85 Schneider, Manfred 23.423 Hartmann, Ulrich 8.52 Hartmann, Ulrich 22.794 Walter, Bernhard 8.00 Kuhnt, Dietmar 21.475 Kuhnt, Dietmar 7.60 Baumann, Karl-H. 20.936 Baumann, Karl-H. 7.21 Kohlhaussen, Martin 20.107 Kohlhaussen, Martin 7.21 Schulte-Noelle, H. 19.378 Strube, Jurgen 6.49 Pierer, Heinrich von 19.189 Reich, Hans W. 6.42 Neuber, Friedel 19.0410 Kley, Max Dietrich 6.36 Walter, Bernhard 18.23

Panel C: Top 10 Degree and Connectiveness for 2005 in Percent

1 Schneider, Manfred 8.98 Cromme, Gerhard 26.562 Cromme, Gerhard 8.40 Pierer, Heinrich von 25.693 Pierer, Heinrich von 8.27 Schneider, Manfred 23.304 Hartmann, Ulrich 8.08 Hartmann, Ulrich 21.975 Walter, Bernhard 7.95 Kohlhaussen, Martin 20.786 Kohlhaussen, M. 7.18 Schulz, Ekkehard D. 19.787 Schulz, Ekkehard D. 6.93 Schulte-Noelle, H. 18.488 Reich, Hans W. 6.35 Baumann, Karl-H. 17.639 Kuhnt, Dietmar 6.16 Walter, Bernhard 17.4310 Kley, Max Dietrich 6.09 Weber, Jurgen 17.19

Panel D: Top 10 Degree and Connectiveness for 2006 in Percent

1 Schneider, Manfred 9.31 Cromme, Gerhard 26.272 Cromme, Gerhard 8.67 Schneider, Manfred 25.143 Hartmann, Ulrich 8.41 Schulz, Ekkehard D. 23.364 Schulz, Ekkehard D. 8.29 Hartmann, Ulrich 22.865 Pierer, Heinrich von 7.00 Kuhnt, Dietmar 20.036 Kuhnt, Dietmar 6.74 Weber, Jurgen 19.307 Kohlhaussen, Martin 6.62 Pierer, Heinrich von 18.678 Reich, Hans W. 6.42 Kohlhaussen, Martin 18.589 Grunberg, Hubertus 6.23 Schulte-Noelle, H. 16.5210 Strube, Jurgen 5.97 Grunberg, Hubertus 15.97

Notes:This table lists the ten most central directors in the network of German firms subdivided into the Degree (C’D)and Connectiveness (C’C) centrality measures for the period 2003–06.



Table 3Normalized Firm Centrality in Percent

Variable Mean Median Lowest Decile Highest Decile SD

Average Degree 1.259 1.034 0.321 2.647 0.876Average Connectiveness 1.947 0.473 3.74 × 10−28 6.851 2.911Average Clustering Coefficient 84.533 86.059 67.556 100 11.340Average K-Core 44.353 42.183 16.129 74.866 22.96Avg. Degree (w/o empl.) 1.630 1.237 0.388 3.572 1.222Avg. Connectiveness (w/o empl.) 3.235 0.927 0 11.178 4.570Average Clustering Coefficient (w/o empl.) 73.872 76.508 46.883 100 19.261Average K-Core (w/o empl.) 37.280 33.766 14.773 62.338 18.157

Notes:This table presents descriptive statistics for the network centrality measures Degree, Connectivenes, ClusteringCoefficient and K-Core for a balanced sample of 133 German market listed firms for the period 2003–06.Thetotal number of firm-year observations is 532. All figures are normalized and aggregated at the firm level.

Our last step, aggregating at the firm level the data on individual directors, enablesus to draw conclusions about the extent to which members of the supervisory board,and thus the board as a whole, might be busy or overcommitted. Table 3 presentsdescriptive statistics (in percent) for our centrality measures.

4. MULTIVARIATE ANALYSIS

Our main hypothesis rests on the assumption that overcommitted individuals mightshirk their responsibilities as directors. Previous literature derived “busyness” simplyfrom the number of directorships. Believing this measure to be too simplistic because itdoes not capture the “quality” of other obligations, we propose an alternative approachbased on measures adopted and adapted from social network research.

In this section, we examine, in a multivariate panel regression framework, the effectof overcommitted boards on firm performance (measured by Tobin’s q) and executiveremuneration.

(i) Multiple Directorships and Firm Performance

Our first step is to apply to our dataset the methodology used by Ferris et al. (2003)and Fich and Shivdasani (2006). Specifically, we estimate firm fixed-effects modelsusing market-to-book value as the dependent variable.13 These models assume agencycosts (incurred due to poor monitoring by overcommitted directors) to be reflected ina lower market-to-book ratio. Used as a proxy for Tobin’s q, the market-to-book ratiois calculated as market value of equity at the end of the year plus book value of debtdivided by the book value of assets at the end of the year. We include as explanatoryvariables the three alternative measures for “busy boards” used by Fich and Shivdasani(2006) together with variables that control for corporate governance and financialcharacteristics. We control for firm size using the natural logarithm of total assets.Yermack’s (1996) and Fich and Shivdasani’s (2006) results suggesting that board size

13 Ferris et al. (2003) use a cross-section of 1995 data; Fich and Sivdasani (2006) analyze data from 1989–95using fixed-effects regressions.



has a negative and significant effect on firm performance, we control for board sizeusing the log of the number of directors. In a study of German codetermination,Gorton and Schmid (2004) find market valuation to decrease as the number ofemployee representatives on the supervisory board increases. Hence, we also controlfor the fraction of employee representation on the board. Several studies have shownownership characteristics to also play an important role in the German system ofcorporate governance (Franks and Mayer, 2001; and Andres, 2008). In the absenceof an active takeover market, large blockholders often act as a firm’s main monitors.Because a large blockholder, given an incentive to expropriate minority shareholders,might abuse its position, we include as control variables both the voting rights of thelargest blockholder (>5%) and its squared term. The regressions further control forthe (log) number of board meetings, age of the firm (natural logarithm of years sinceincorporation), risk (return volatility, measured as the standard deviation of shareprice returns for the previous 36 months), and capital structure (leverage, definedas book value of total debt divided by book value of total assets). Lastly, all modelsinclude year dummies. Tests for correlation between the explanatory variables andvariance inflation factors (not reported) indicate that our regression models do notsuffer from multicollinearity bias.14

The results of the multivariate panel regressions are reported in Table 4. In models(1)–(3), all variables that measure the extent to which boards are busy are based ondirectors who represent shareholder interests (i.e., exclude employee representatives).Regressions (4)–(6) are based on all directors including employee representatives.We do not find signs of a significant relationship between busy directors and firmperformance in any model specification. The coefficients for Busy board (a dummyset to 1 if 50% or more of the directors hold three or more directorships) andPercentage of busy directors (both used by Fich and Shivdasani, 2006) are negative butinsignificant. The coefficients of the variable Average number of directorships (used byFerris et al., 2003) are positive but insignificant. These findings indicate that thenumber of board appointments a director holds does not seem to influence firmperformance in Germany. Compared to US studies, the results are in line with thoseof Ferris et al. (2003), but somewhat at odds with those of Fich and Shivdasani (2006).

(ii) Director Networks and Firm Governance

The preceding results do not support the hypothesis that busy boards are associatedwith poor firm performance. A number of other research papers have tested thehypothesis that directors overcommitted by virtue of having too many board assign-ments might lack the time and attention required of vigilant monitors of management.In these studies, as noted above, boards are classified as busy/not busy based onthe number and percentage of directorships per director. We maintain, however,that these measures will be noisy if they fail to adequately capture the extent towhich directors are truly busy and overcommitted. Consider, for example, two of thedirectors in our sample, Mr. Schulte-Noelle, who held only four board appointmentsin 2006, and Michael Busch, director at Washtec AG (a producer of vehicle washing

14 The only exception is a comparatively high correlation between the variables size and board size, whichmight be too high (0.67). Estimating all regressions without board size, as a robustness test, we find theresults to not be influenced by the correlation between the two variables.



Table 4Multiple Directorships and Firm Performance

Excluding Employee Representatives All Directors

Variable (1) (2) (3) (4) (5) (6)

Busy board −0.045 −0.072(−0.60) (−0.75)

Percentage of busy −0.005 −0.012directors (−0.02) (−0.04)

Average number of 0.043 0.054directorships (1.05) (0.89)

Firm size (log of −0.288 −0.232 −0.225 −0.230 −0.232 −0.226total assets) (−1.12) (−1.24) (−1.12) (−1.14) (−1.14) (−1.12)

Board size (log of # 0.037 0.053 0.080 0.057 0.053 0.073of directors) (0.25) (0.37) (0.57) (0.42) (0.38) (0.52)

Codetermination 0.852 0.822 0.798 0.813 0.822 0.855(1.74)* (1.59) (1.65) (1.67)* (1.64) (1.63)

Leverage −0.008 −0.009 −0.029 −0.036 −0.009 −0.026(−0.01) (−0.02) (−0.05) (−0.06) (−0.02) (−0.05)

Largest blockholder 0.005 0.005 0.004 0.005 0.005 0.004(0.82) (0.82) (0.69) (0.78) (0.81) (0.69)

Largest blockholder2 −0.000 −0.000 −0.000 −0.000 −0.000 −0.000(−0.59) (−0.58) (−0.48) (−0.54) (−0.57) (−0.49)

Board meetings 0.025 0.025 0.030 0.026 0.025 0.029(1.33) (1.33) (1.44) (1.36) (1.32) (1.41)

Return volatility 0.000 0.000 0.000 0.000 0.000 0.000(1.51) (1.48) (1.52) (1.48) (1.48) (1.50)

Firm age 0.601 0.612 0.614 0.609 0.613 0.611(1.43) (1.46) (1.46) (1.46) (1.64) (1.46)

R-squared 0.142 0.142 0.147 0.142 0.141 0.146F-statistic 4.35 3.92 3.65 3.36 3.79 3.41(p-value) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)

Notes:This table presents the results of fixed-effects regressions of market-to-book value on several firm character-istics for a sample of 133 German firms for the period 2003–06. Busy board is an indicator variable set to 1if 50% (or more) of a firm’s directors hold three (or more) directorships. Percentage of busy directors is thepercentage of directors (per firm) who hold three or more directorships. In specifications (1)–(3), thesevariables are computed based on shareholder representatives only; the board variables in models (4)–(6)are based on all directors (including employee representatives). Codetermination is defined as the number ofemployee representatives divided by the number of supervisory board members. Leverage is defined as bookvalue of total debt divided by book value of total assets. Large shareholders are shareholders who hold morethan 5% of the voting equity. Return volatility is measured as the standard deviation of share price returns forthe previous 36 months. Firm age is the natural logarithm of years since incorporation. All regressions includeyear dummies. T-statistics (in parentheses) are based on White-Heteroskedasticity consistent standard errorsclustered at the firm level. Asterisks denote statistical significance at the 0.01(***), 0.05(**), and 0.10(*)-level.

equipment), who served on seven boards in the same year. Based on the number ofboard appointments, Mr. Busch is busier. Mr. Schulte-Noelle, however, is chairman ofAllianz AG and a director at Siemens AG, ThyssenKrupp AG, and E.ON AG, someof the most important German companies with connections to politics, media andsports. Being among the best-connected of directors, and sitting at the heart of an“inner circle” of organizational elites (Useem, 1984), he is more likely than Mr. Buschto be overcommitted, not so much because of the (direct) duties associated with his



directorships, but owing to the indirect obligations that stem from his central positionin the network. Because simple measures, such as number of board seats and otherclosely related metrics, fail to detect this dimension of additional directorships, we usealternative measures employed in social network analysis, namely, the variables Degree,Connectiveness, Clustering Coefficient and K-Core (aggregated at the firm level), to measurethe extent to which a firm’s directors are overcommitted.

We first examine the effect on firm performance of directors’ embeddedness inthe social network. Tables 5a and 5b report the results of fixed-effects regressions withTobin’s q as the dependent variable. In Table 5a, we use the network measures assubstitutes for the Busy board variables included in section 4(i). Model specificationsin Table 5b contain the busy dummy as an additional control variable, and use asnetwork measures average Degree, Connectiveness, Clustering Coefficient and K-Core forevery firm and year.15 All other control variables are the same as in the regressionsabove.

The influence of the number of directorships (captured by the variable Busy board)on Tobin’s q is reflected in the results in Table 5b, which confirm the precedingfindings and show insignificant coefficients in all model specifications. Negative andsignificant coefficients (at the 0.05-level) on the variables Degree and Connectivenessindicate an association between directors who are comparatively central in the networkand poor firm performance. Intense network connections, although theoreticallypotentially beneficial to firms (e.g., by affording access to financial resources), mayovercommit directors in terms of time and cognitive resources consumed. We furtherexplore the impact of a director’s higher order relations using first the local ClusteringCoefficient, which measures the degree of closure within a director’s neighbourhood.Although there is no clear prediction regarding the relation between network closureand firm performance at low levels of director connectedness, we expect a negativerelation at high levels of connectedness. Because it does not indicate to how manyother directors a director is connected, we interact Clustering Coefficient with directors’nodal degree, with both interaction variables centred at their respective means. Ourfindings suggest that the degree of triadic closure does not significantly affect theperformance of firms with an average nodal degree. The interaction term, however, isnegative and highly significant (at the 0.01-level), that is, the marginal effect of closuresignificantly decreases as a director’s number of connections increases. As stated above,we expect monitoring ability, especially of a director already highly connected, to beadversely affected by a relatively closed network structure. We illustrate this effect inFigure 2, which plots the average marginal effect of Clustering Coefficient for differentlevels of Degree (range: 10%–90% percentile of Degree).

At low levels of Degree, the marginal effect of closure is positive, but not significantlydifferent from zero. At higher levels of Degree, the effect shrinks and, if the normalizedDegree amounts to roughly 1.7 (1.5), becomes negative and significant (at the 0.05(0.10)-level). Network closure thus seems to negatively affect firm performance if aboards’ directors hold a large number of relations. These findings provide support forthe idea that closed network structures, by imposing too many time-consuming socialobligations, may impede efficient monitoring. We also find a negative and significantcoefficient (at the 0.05-level) on the variable K-Core, which implies an association

15 Because using the median of the normalized values as a robustness test does not materially change theresults, these specifications are not reported.



Table 5aDirector Networks and Firm Performance

Variable (1) (2) (3) (4)

Degree −0.449 −0.704(−2.21)** (−2.57)**

Connectiveness −0.073(−2.07)**

Clustering coefficient −0.771(−1.28)

Clustering coefficient −1.535*Degree (−2.74)***

K-core −1.076(−2.05)**

Firm size −0.232 −0.222 −0.261 −0.250(log of total assets) (−1.17) (−1.10) (−1.33) (−1.25)

Board size 0.367 0.125 0.602 0.430(log of # of directors) (1.69)* (0.84) (2.06)** (1.61)

Codetermination 0.721 0.768 0.529 0.640(1.83)* (1.93)* (1.26) (1.27)

Leverage −0.124 −0.101 −0.116 −0.033(−0.22) (−0.18) (−0.21) (−0.06)

Largest blockholder 0.006 0.005 0.007 0.006(0.96) (0.84) (1.06) (0.97)

Largest blockholder2 −0.000 −0.000 −0.000 −0.000(−0.78) (−0.66) (−0.88) (−0.72)

Board meetings 0.027 0.025 0.030 0.027(1.42) (1.32) (1.55) (1.40)

Return volatility 0.000 0.000 0.000 0.000(1.53) (1.36) (1.67)* (1.54)

Firm age 0.743 0.646 0.763 0.677(1.77)* (1.54) (1.86)* (1.66)*

R-squared 0.160 0.151 0.178 0.148F-statistic 4.44 4.28 3.89 3.47(p-value) (0.00) (0.00) (0.00) (0.00)

Notes:This table presents the results of fixed-effects regressions of market-to-book value on several firm character-istics for a sample of 133 German firms for the period 2003–06. Degree, Connectiveness, Clustering Coefficientand K-Core are measures of the embeddedness of a firm’s directors in the social network. Busy board isan indicator variable set to 1 if 50% (or more) of a firm’s directors hold three (or more) directorships.Codetermination is defined as the number of employee representatives divided by the number of supervisoryboard members. Leverage is defined as book value of total debt divided by book value of total assets. Largeshareholders are shareholders who hold more than 5% of the voting equity. Return volatility is measured as thestandard deviation of share price returns for the previous 36 months. Firm age is the natural logarithm ofyears since incorporation. All regressions include year dummies. T-statistics (in parentheses) are based onWhite-Heteroskedasticity consistent standard errors clustered at the firm level. Asterisks denote statisticalsignificance at the 0.01(***), 0.05(**), and 0.10(*)-level.

between directors with relations clustered in a cohesive sub-group and lower firmperformance. The results are robust to including the busy dummy as an additionalcontrol.

We further investigate the effect on firm governance by examining the relationbetween network embeddedness and level of executive compensation. The rationaleadvanced by prior academic research that finds weak monitoring and poor governance



Table 5bDirector Networks and Firm Performance

Variable (1) (2) (3) (4)

Degree −0.444 −0.686(−2.27)** (−2.66)***

Connectiveness −0.073(−2.10)**


Clustering coefficient −1.542*Degree (−2.77***)K-core −1.078

(−2.05)**

Busy board 0.062 0.070 0.072 0.073(0.73) (0.76) (0.90) (0.78)

Firm size −0.230 −0.220 −0.258 −0.247(log of total assets) (−1.17) (−1.10) (−1.32) (−1.25)

Board size 0.367 0.128 0.588 0.434(log of # of directors) (1.70)* (0.86) (2.08)** (1.62)

Codetermination 0.717 0.763 0.529 0.633(1.82)* (1.92)* (1.27) (1.26)

Leverage −0.145 −0.125 −0.137 −0.058(−0.26) (−0.22) (−0.25) (−0.11)



Board meetings 0.028 0.026 0.031 0.028(1.45) (1.35) (1.59) (1.43)

Return volatility 0.000 0.000 0.000 0.000(1.52) (1.35) (1.66)* (1.53)

Firm age 0.739 0.642 0.756 0.674(1.76)* (1.53) (1.84)* (1.65)


Notes:This table presents the results of fixed-effects regressions of market-to-book value on several firm character-istics for a sample of 133 German firms for the period 2003–06. Degree, Connectiveness, Clustering Coefficientand K-Core are measures of the embeddedness of a firm’s directors in the social network. Busy board isan indicator variable set to 1 if 50% (or more) of a firm’s directors hold three (or more) directorships.Codetermination is defined as the number of employee representatives divided by the number of supervisoryboard members. Leverage is defined as book value of total debt divided by book value of total assets. Largeshareholders are shareholders who hold more than 5% of the voting equity. Return volatility is measured as thestandard deviation of share price returns for the previous 36 months. Firm age is the natural logarithm ofyears since incorporation. All regressions include year dummies. T-statistics (in parentheses) are based onWhite-Heteroskedasticity consistent standard errors clustered at the firm level. Asterisks denote statisticalsignificance at the 0.01(***), 0.05(**), and 0.10(*)-level.

to be positively related to executive compensation is that executives in firms with weakcorporate governance structures and insufficient monitoring can more effectivelyinfluence their compensation committees. Core et al. (1999), for example, find CEOcompensation to be higher when boards are large and CEOs hold the position ofchairman of the board. Yermack (1997) presents related evidence of a correlation



Figure 2Average Marginal Effect (AME) of “Clustering Coefficient” on Tobin’s q with 95%

Confidence Intervals (CIs)

-6-4

-20

2

.5 1 1.5 2 2.5Degree

between CEO influence on the board and being assigned stock options at favourabletimes. In line with Core et al. (1999), Cyert et al. (2002) show, both theoreticallyand empirically, that top management compensation is higher when the CEO is alsochairman of the board, and Sapp (2008) reports a positive relationshop betweenmeasures for weaker boards and level of CEO compensation.

According to Shleifer and Vishny (1986), the presence of a large shareholder oc-casions closer monitoring of management’s performance. Bertrand and Mullainathan(2001) find CEO compensation to be less dependent on managerial effort in firms thatlack a large external blockholder, and CEOs to be less likely to be rewarded for luck inbetter governed firms. Hartzell and Starks (2003), examining the relationship betweenownership concentration and executive compensation, find higher pay-performancesensitivities and lower executive compensation with more concentrated institutionalownership, and Shivdasani (1993) finds evidence of a positive relationship betweenCEO compensation and number of directorships.

These papers, by and large, find executive compensation to be an importantcomponent of corporate governance, and poor governance to be associated withcomparatively high levels of executive pay. We further investigate the hypothesisthat overcommitted directors spend less time and effort monitoring management byexamining the relationship between directors’ embeddedness in the network and levelof executive compensation. As the dependent variable, we compute for each firm and



year the average per-capita executive compensation (total compensation includingoptions, etc.).16 As before, director busyness is measured by the variables Degree,Connectiveness, Clustering Coefficient and K-Core (aggregated at the firm level).17 Oneof the stylized facts of compensation research is the positive relationship between firmsize and executive pay. It is well documented in the empirical literature that large firmspay executives more (see, for example, Murphy, 1985 and Ryan and Wiggins, 2001).We therefore include firm size (logarithm of total assets) as a control variable. Weinclude as controls the board characteristics size (measured as the log of the numberof executives) and fraction of employee representatives. Among the explanatoryvariables, we include the fraction of voting rights of the largest blockholder and itssquared term. Although most firms’ executive compensation packages include someform of equity-based compensation, pay-performance sensitivities are generally foundto be low (e.g., Jensen and Murphy, 1990). We nevertheless control for past operatingperformance (return on assets) and stock price performance (the CDAX performanceindex is used to adjust stock returns over the past 12 months). Leverage (book valueof total debt divided by book value of total assets, John and John, 1993) and age ofthe firm (natural logarithm of years since incorporation) are also included as controlvariables. We include dummy variables to control for year fixed effects.18

The results of fixed-effects panel regressions with executive compensation as thedependent variable are presented in Table 6a and Table 6b. In Table 6a, we substituteour network measures for the Busy board variable, whereas the model specifications inTable 6b contain the busy dummy as an additional control.19 We find the hypothesizedpositive and significant relationship between the network measures and level ofexecutive compensation. Although we find no evidence of a relationship betweenClustering Coefficient and K-Core, the coefficients on Degree and Connectiveness are allpositive and significant (at the 0.10-level). This finding implies that German firmsin which members of the supervisory board are better connected in the network ofdirectors, pay their executives more. Because the regressions also control for sizeand performance, these results can be interpreted as supporting evidence for thehypothesis that well-connected directors are associated with poor monitoring. Thecoefficients on Busy board are positive but insignificant in all model specifications, andwe find, in line with the literature, a significantly positive relationship between firmsize and executive compensation. Remuneration seems to be positively influenced bystock price performance over the past 12 months.

In the aggregate, the results of the regression analysis provide evidence ofhigh levels of director embeddedness being associated with low firm performance

16 Until recently, German companies were not required to disclose executive remuneration on anindividual basis. Disclosure of aggregate executive remuneration, however, has been mandated by the GermanCommercial Code. We combine the information on aggregate compensation with information on thenumber of executives to compute average per-capita remuneration for all sample firms. From 2007 onwards,publicly listed firms have been required by law to disclose executive compensation on an individual basis (ifthe shareholders’ meeting does not grant an exemption from the disclosure requirement).17 As in the preceding regressions, using the median of the normalized network measures as a robustnesstest yields results quantitatively and qualitatively similar to those reported in the text.18 As in the previous regressions, we compute variance inflation factors and cross-correlations. The teststatistics (not reported) confirm that the coefficients obtained in our compensation regressions are notbiased by multicollinearity.19 We also run all regressions with executive compensation as the dependent variable using the threemeasures for busy boards derived from the number of directorships per director (not reported). Thecoefficients on these measures are negative but insignificant in all regressions.



Table 6aDirector Networks and Executive Compensation

Variable (1) (2) (3) (4)

Degree 230.1 151.6(1.79)* (1.11)

Connectiveness 59.31(1.71)*


Clustering coefficient −200.3*Degree (−0.37)K-core 66.58

(0.16)Firm size 376.0 361.6 363.7 373.9

(log of total assets) (2.32)** (2.21)** (2.35)** (2.37)**

Executive board size −216.0 −218.0 −211.0 −225.8(log of # of executives) (−2.18)** (−2.22)** (−2.09)** (−2.22)**

Codetermination 516.7 590.3 526.1 594.7(1.09) (1.03) (0.96) (1.16)

Leverage 240.5 256.5 229.0 181.2(0.70) (0.77) (0.67) (0.53)



Stock price 68.52 68.19 66.46 66.75performance (2.08)** (2.07)** (1.99)** (2.00)**

Return on assets 954.9 928.8 919.0 939.2(2.07)** (2.01)** (1.99)** (1.99)**

Firm age −298.2 −241.9 −283.8 −218.4(−1.35) (−1.44) (−1.30) (−1.01)


Notes:This table presents the results of fixed-effects regressions of the logarithm of average per-capita executiveremuneration on several firm characteristics for a sample of 133 German firms for the period 2003–06.Degree, Connectiveness, Clustering Coefficient and K-Core are measures of the embeddedness of a firm’s directorsin the social network. Busy board is an indicator variable set to 1 if 50% (or more) of a firm’s directors holdthree (or more) directorships. Codetermination is defined as the number of employee representatives dividedby the number of supervisory board members. Leverage is defined as book value of total debt divided bybook value of total assets. Large shareholders are shareholders who hold more than 5% of the voting equity.Stock price performance is defined as the raw stock price return minus the return of the CDAX index. Firm ageis the natural logarithm of years since incorporation. All regressions include year dummies. T-statistics (inparentheses) are based on White-Heteroskedasticity consistent standard errors clustered at the firm level.Asterisks denote statistical significance at the 0.01(***), 0.05(**), and 0.10(*)-level.

and high levels of executive compensation. The results further suggest that theexplanatory power of social network measures exceeds that of simple measuresof director busyness. We interpret our findings as evidence of weaker governanceand poorer monitoring in firms with directors intensely connected in their socialnetworks.



Table 6bDirector Networks and Executive Compensation

Variable (1) (2) (3) (4)

Degree 236.8 161.5(1.84)* (1.16)

Connectiveness 59.93(1.76)*


Clustering coefficient −214.4*Degree (−0.39)K-core 75.93

(0.19)Busy board 83.44 77.87 68.98 75.59

(1.38) (1.33) (1.13) (1.24)Firm size 379.0 364.2 366.7 376.8

(log of total assets) (2.29)** (2.19)** (2.33)** (2.35)**

Executive board size −222.2 −224.0 −215.7 −231.8(log of # of executives) (−2.27)** (−2.30)** (−2.15)** (−2.30)**

Codetermination 517.1 592.8 524.4 596.3(1.10) (1.04) (0.97) (1.18)

Leverage 219.9 236.4 213.1 161.3(0.64) (0.71) (0.62) (0.47)



Stock price 67.23 66.97 65.79 65.49performance (2.05)** (2.04)** (1.98)** (1.97)*

Return on assets 951.0 924.5 916.8 936.2(2.05)** (1.99)** (1.97)* (1.98)**

Firm age −305.3 −246.5 −290.3 −223.6(−1.38) (−1.16) (−1.33) (−1.04)


Notes:This table presents the results of fixed-effects regressions of the logarithm of average per-capita executiveremuneration on several firm characteristics for a sample of 133 German firms for the period 2003–06.Degree, Connectiveness, Clustering Coefficient and K-Core are measures of the embeddedness of a firm’s directorsin the social network. Busy board is an indicator variable set to 1 if 50% (or more) of a firm’s directors holdthree (or more) directorships. Codetermination is defined as the number of employee representatives dividedby the number of supervisory board members. Leverage is defined as book value of total debt divided bybook value of total assets. Large shareholders are shareholders who hold more than 5% of the voting equity.Stock price performance is defined as the raw stock price return minus the return of the CDAX index. Firm ageis the natural logarithm of years since incorporation. All regressions include year dummies. T-statistics (inparentheses) are based on White-Heteroskedasticity consistent standard errors clustered at the firm level.Asterisks denote statistical significance at the 0.01(***), 0.05(**), and 0.10(*)-level.

(iii) Robustness Tests

The findings presented above potentially suffer from the problem of endogeneity.Poorly performing firms might seek to populate their boards with well-connecteddirectors with reputations for, and experience in, improving corporate performance.



Table 7Effect of Past Performance on Network Centrality

Change in Change in Change in Change inDegree Connectiveness Clustering coefficient K-core

Dependent Variable (1) (2) (3) (4)

Abnormal stock return 0.023 0.241 0.012 0.013(current year) (0.51) (1.17) (0.74) (0.06)

Abnormal stock return 0.004 −0.002 −0.003 −0.194(previous year) (0.18) (−0.04) (−0.34) (−1.49)

Change in Tobin’s q −0.038 −0.298 0.002 0.058(−0.96) (−1.58) (0.16) (0.34)

Change in size −0.00 −0.00 0.00 −0.00(−1.66)** (−0.15) (0.99) (−0.76)

Industry dummies yes yes yes yesYear dummies yes yes yes yesR-squared 0.021 0.032 0.031 0.137

Notes:This table presents the results of ordinary least square regressions of changes in network centrality on severalfirm characteristics for a sample of 133 German firms for the period 2003–06. Degree, Connectiveness, ClusteringCoefficient and K-Core are measures of the embeddedness of a firm’s directors in the social network. TheAbnormal stock return is computed as the raw stock price return during the current year minus the return ofthe CDAX-index. The Change in size is measured by the log of total assets. All regressions include year andindustry dummies. T-statistics (in parentheses) are based on White-Heteroskedasticity consistent standarderrors clustered at the firm level. The total number of observations is 399. Asterisks denote statisticalsignificance at the 0.01(***), 0.05(**), and 0.10(*)-level.

This reverse causality would imply that performance is not worse due to betterconnected boards, but rather that only poorly performing firms seek the advice of,and thereby boost their boards’ network embeddedness through the appointment of,well-connected directors. We test the causality of our results by estimating regressionmodels of the relationship between past performance and changes in networkembeddedness. Roughly following Yermack (1996), we take the total annual change inthe network measure (from t = −1 to t = 0) as the dependent variable and include asexplanatory variables the abnormal stock price performance (raw stock price returnminus the return of the CDAX index) during the same year (t = −1 to t = 0) andin the previous year (t = −2 to t = −1). If poorly performing companies appointedwell-connected directors, performance either in the year of the board change or in theprevious year (and, hence, the coefficients of these two variables) should show somelevel of significance. We include as additional control variables, changes in Tobin’s qand firm size (log of total assets).

Table 7 reports the results of these OLS regressions. The models provide no evi-dence that firms alter the centrality of their boards in response to poor performance.In other words, we find no evidence of reverse causality, well-connected directors notseeming to be appointed in response to bad performance. Adding lagged changes inTobin’s q (i.e., from t = −2 to t = −1) as additional control variables yields qualitativelysimilar results. There is weak evidence that Degree tends to decrease in response to(positive) changes in firm size.

As an additional robustness test, we identify the top and bottom deciles for 2003abnormal stock price returns and compute the average (and median) annual changein network measures for these groups. As argued above, if the findings presented



Table 8Top and Bottom Performance Decile 2003 and Network Centrality

Bottom TopPerformance Performance Difference in Means

Decile Decile (T-Values)

2003Average abnormal stock returns −40.65% +152.32% 9.34***Degree 1.035 1.070 0.09Connectiveness 0.967 1.237 0.28Clustering Coefficient 0.027 0.004 0.60K-Core 0.274 0.319 0.53

2003 –06 change inDegree −0.247 − 0.142 0.73Connectiveness − 0.070 − 0.067 1.12Clustering Coefficient 0.005 0.002 0.33K-Core −0.042 −0.030 0.47

Notes:This table reports changes in the network measures over the period 2003–06 for the best and worstperforming companies in 2003. Performance is defined as abnormal stock price performance (raw stockprice return minus the return of the CDAX index) during the first year of the sample period (2003).

in section 4(ii) are the result of reverse causality, companies should appoint well-connected directors in the wake of poor performance. Table 8 reports the averagechange in network centrality for the deciles with the best and worst abnormal stockperformance during 2003. This table displays the longest possible interval (2003–06).This is the most conservative measure, as board changes usually take time, and a four-year time window theoretically allows for replacement of the entire board (maximumtenure is five years). That all network measures except Clustering Coefficient decreasefor these groups over the 2003–06 period is not surprising, given the break up ofso-called “Deutschland AG”, with its strong inter-firm connections (Dittmann et al.,2010). However, changes in the network measures are almost indistinguishable (andstatistically insignificant) between the best and worst performers. Performing similarcalculations for other periods that relate the top and bottom performance deciles interms of abnormal stock price performance and changes in Tobin’s q to subsequentchanges in board embeddedness yields similar results.20 This can be regarded assupporting evidence of the main finding that the negative relationship between firmperformance and board centrality is not the result of reverse causality.

5. CONCLUSION

In the wake of a number of corporate scandals, governance structures have recentlybecome a subject of public debate in many countries. Shareholder activists andorganizations that defend the interests of minority shareholders often criticize firmsfor appointing as directors individuals who hold directorships in other firms, arguingthat serving on too many boards may impair directors’ effectiveness as monitors of

20 Tables for these computations, which are not reported as the results are qualitatively similar to thosereported in Table 8, are available upon request.



management. Corporate governance policy advocates have therefore called for thenumber of directorships a director can hold to be limited.

The debate has also spurred empirical research on whether busy directors areassociated with poor governance. Empirical evidence thus far is scarce and limited tothe United States. More important, previous papers use as a measure of busyness onlythe number of directorships, which metric fails to capture the quality and structuralposition of additional board seats. We propose an alternative approach based onmeasures from the social network literature.

We examine a sample of 133 German firms over four years, which yields a network ofapproximately 1,600 directors and 35,000 connections. We use these data, aggregatedat the firm level, to investigate the relationship between firm governance and directors’embeddedness in the network. Our results show boards populated by well-connected(in terms of Degree and Connectedness) directors to be associated with lower perfor-mance (as measured by Tobin’s q) and higher executive compensation. We interpretthese findings as evidence of poor monitoring in firms with well-connected directors.

We by no means claim to have found the “perfect” measure of director busyness.Supervisors with weak or no connections to the corporate elite may be busier thanother directors owing to such activities as hobbies and volunteerism. Nevertheless, ourresults are evidence of an important contribution of network measures to the complexand difficult to assess issue of busyness by helping to explain dimensions beyond simplythe number of board appointments.

These findings have direct policy implications. They suggest, for example, that callsto impose limits on multiple directorships that do not take into account the qualityof the attendant obligations are ill advised. Our results also imply that it might notbe optimal for shareholders to select directors solely on the basis of reputation, asthe most reputable directors will likely already be serving on multiple boards andtherefore have to balance the interests of many parties. Having a preponderance ofintensely connected directors, if it translates into an overcommitted board, may becounterproductive.

REFERENCES

Adams, R. and D. Ferreira (2007), ‘A Theory of Friendly Boards’, Journal of Finance, Vol. 62, No.1 (February), pp. 217–50.

Adler, P. S. and S. Kwon (2000), ‘Social Capital: The Good, the Bad, and the Ugly’, in Lesser,E. (ed.), Knowledge and Social Capital: Foundations and Applications (Boston: Butterworth-Heineman), pp. 89–115.

Andres, C. (2008), ‘Large Shareholders and Firm Performance – An Empirical Examination ofFounding-Family Ownership’, Journal of Corporate Finance, Vol. 14, No. 4 (September), pp.431–45.

Barnea, A. and I. Guedj (2009), ‘Director Networks’, SSRN Working Pa-per. Available at SSRN: http://papers.ssrn.com/sol3/papers.cfm?abstractid=966555.

Bertrand, M. and S. Mullainathan (2001), ‘Are CEOs Rewarded for Luck? The Ones WithoutPrincipals Are’, Quarterly Journal of Economics, Vol. 116, No. 3 (August), pp. 901–32.

Bonacich, P. (1972), ‘Factoring and Weighting Approaches to Status Scores and CliqueIdentification’, Journal of Mathematical Sociology, Vol. 2, No. 1, pp. 113–20.

——— (1987), ‘Power and Centrality: A Family of Measures’, American Journal of Sociology, Vol.92, No. 5 (March), pp. 1,170–82.



Brown, W. O. and M. T. Maloney (1999), ‘Exit, Voice, and the Role of Corporate Direc-tors: Evidence from Acquisition Performance’, SSRN Working Paper. Available at SSRN:http://papers.ssrn.com/sol3/papers.cfm?abstract˙id=160308.

CASOS (2008), ‘Organizational Risk Analyzer v1.9.0’, Carnegie Mellon University, Schoolof Computer Science, and online resource, available at: http://www.casos.cs.cmu.edu/projects/ora/.

Cialdini, R. B. and M. R. Trost (1998), ‘Social Influence: Social Norms, Conformity andCompliance’, in D. T. Gilbert, S. T. Fiske and G. Lindzey (eds.), The Handbook of SocialPsychology, Vol. 2 (Boston, MA: McGraw-Hill), pp. 151–92.

Cohen, L. H., C. J. Malloy and A. Frazzini (2010), ‘Sell-Side School Ties’, Journal of Finance, Vol.65, No. 4 (August), pp. 1,409–37.

Coleman, J. S. (1988), ‘Social Capital in the Creation of Human Capital’, The American Journal ofSociology, Vol. 94, No. 1 (January), Supplement: Organizations and Institutions: Sociologicaland Economic Approaches to the Analysis of Social Structure, pp. 95–120.

Core J., R. Holthausen and D. Larcker (1999), ‘Corporate Governance, Chief Executive OfficerCompensation, and Firm Performance’, Journal of Financial Economics, Vol. 51, No. 3(March), pp. 371–406.

Cyert, R. M., S. H. Kang and P. Kumar (2002), ‘Corporate Governance, Takeovers, and Top-Management Compensation: Theory and Evidence’, Management Science, Vol. 48, No. 4(April), pp. 453–69.

Dittmann, I., E. Maug and C. Schneider (2010), ‘Bankers on the Boards of German Firms: WhatThey Do, What They Are Worth, and Why They Are (Still) There’, Review of Finance, Vol. 14,No. 1 (May), pp. 35–71.

Eppler, M. J. and J. Mengis (2004), ‘The Concept of Information Overload: A Review of Litera-ture from Organization Science, Accounting, Marketing, MIS, and Related Disciplines’, TheInformation Society: An International Journal, Vol. 20, No. 5 (November), pp. 325–44.

Ferriani, S., G. Cattani and C. Baden-Fuller (2009), ‘The Relational Antecedents of Project-Entrepreneurship: Network Centrality, Team Composition and Project Performance’,Research Policy, Vol. 38, No. 10 (December), pp. 1,545–58.

Ferris, S. P., M. Jagannathan and A. C. Pritchard (2003), ‘Too Busy to Mind the Business?Monitoring by Directors with Multiple Board Appointments’, Journal of Finance, Vol. 58,No. 3 (June), pp. 1,087–111.

Fich, E. M. and A. Shivdasani (2006), ‘Are Busy Boards Effective Monitors?’ Journal of Finance,Vol. 61, No. 2 (April), pp. 689–724.

Flynn, F. J., R. E. Reagans, E. T. Amanattulah and D. R. Ames (2006), ‘Helping One’s Wayto the Top: Self-Monitors Achieve Status by Helping Others and Knowing Who HelpsWhom’, Journal of Personality and Social Psychology, Vol. 91, No. 6 (December), pp. 1,123–37.

Franks, J. and C. Mayer (2001), ‘Ownership and Control of German Corporations’, Review ofFinancial Studies, Vol. 14, No. 4 (October), pp. 943–77.

Gorton, G., and F. A. Schmid (2004), ‘Capital, Labor, and the Firm: A Study of GermanCodetermination’, Journal of the European Economic Association, Vol. 2, No. 5 (September),pp. 863–904.

Gouldner, A. W. (1960), ‘The Norm of Reciprocity: A Preliminary Statement’, AmericanSociological Review, Vol. 25, No. 2 (April), pp. 161–78.

Granovetter, M. (1985), ‘Economic Action and Social Structure: The Problem ofEmbedded-ness’, American Journal of Sociology, Vol. 91, No. 3 (November), pp. 481–510.

——— (1992), ‘Problems of Explanation in Economic Sociology’, in N. Nohria and R. G. Eccles(eds.), Networks and Organizations: Structure, Form, and Action (Boston, MA: Harvard BusinessSchool Press), pp. 25–56.

Hansen, M. T., J. M. Podolny and J. Pfeffer (2001), ‘So Many Ties, So Little Time: A TaskContingency Perspective on Corporate Social Capital in Organizations’, in S. M. Gabbayand R. T. A. J. Leenders (eds.), Social Capital of Organizations: Research in the Sociology ofOrganizations, Vol. 18 (Amsterdam: Emerald Group Publishing Limited), pp. 21–57.

Hartzell, J. C. and L. T. Starks (2003), ‘Institutional Investors and Executive Compensation’,Journal of Finance, Vol. 58, No. 6 (December), pp. 2,351–74.



Horton, J., Y. Millo and G. Serafeim (2012), ‘Resources or Power? Implications of Social Net-works on Compensation and Firm Performance’, Journal of Business Finance and Accounting,Vol. 39, Nos. 3&4 (April/May), pp. 399–426.

Huisman, M. and M. A. J. van Duijn (2005), ‘Software for Social Network Analysis’, in P. J.Carrington, J. Scott and S. Wasserman (eds.), Models and Methods in Social Network Analysis(New York, NY: Cambridge University Press), pp. 270–316.

Hwang, B.-H. and S. Kim (2009), ‘It Pays to Have Friends’, Journal of Financial Economics, Vol. 93,No. 1 (July), pp. 138–58.

Jensen, M. C. and K. J. Murphy (1990), ‘Performance Pay and Top-Management Incentives’,Journal of Political Economy, Vol. 98, No. 2 (April), pp. 225–64.

John, T. A. and K. John (1993), ‘Top-Management Compensation and Capital Structure’, Journalof Finance, Vol. 48, No. 3 (July), pp. 949–74.

Kahneman, D. (1973), Attention and Effort (Englewood Cliffs, NJ: Prentice-Hall).Lee, F. (1997), ‘When the Going Gets Tough, Do the Tough Ask for Help? Help Seeking and

Power Motivation in Organizations’, Organizational Behavior and Human Decision Processes,Vol. 72, No. 3 (December), pp. 336–63.

Mace, M. L. (1986), Directors: Myth and Reality, 2nd edition (Boston, MA: Harvard Business SchoolPress).

Murphy, K. J. (1985), ‘Corporate Performance and Managerial Remuneration – An EmpiricalAnalysis’, Journal of Accounting and Economics, Vol. 7, Nos. 1–3 (April), pp. 11–42.

Oldroyd, J. and S. Morris (2012), ‘Catching Falling Stars: A Human Resource Response toSocial Capital’s Detrimental Effect of Information Overload for Star Employees’, Academy ofManagement Review, Vol. 37, No. 3 (July), pp. 396–418.

Perry, T. and U. Peyer (2005), ‘Board Seat Accumulation by Executives: A Shareholder’sPerspective’, Journal of Finance, Vol. 60, No. 4 (August), pp. 2,083–123.

Raub, W. and J. Weesie (1990), ‘Reputation and Efficiency in Social Interactions: An Exampleof Network Effects’, American Journal of Sociology, Vol. 96, No. 3 (November), pp. 626–54.

Ryan, H. E. and R. A. Wiggins (2001), ‘The Influence of Firm- and Manager-Specific Character-istics on the Structure of Executive Compensation’, Journal of Corporate Finance, Vol. 7, No.2 (June), pp. 101–23.

Sapp, S. G. (2008), ‘The Impact of Corporate Governance on Executive Compensation’,European Financial Management, Vol. 14, No. 4 (September), pp. 710–46.

Schneider, S. C. (1987), ‘Information Overload: Causes and Consequences’, Human SystemsManagement, Vol. 7, No. 2 (January), pp. 143–53.

Scott, J. P. (2000), Social Network Analysis – A Handbook, 2nd edition (London: Sage Publications).Seidman, S. B. (1983), ‘Network Structure and Minimum Degree’, Social Networks, Vol. 5, No. 3

(September), pp. 269–87.Shleifer, A. and R. Vishny (1986), ‘Large Shareholders and Corporate Control’, Journal of

Political Economy, Vol. 94, No. 3 (June), pp. 461–88.Shivdasani, A. (1993), ‘Board Composition, Ownership Structure and Hostile Takeovers’, Journal

of Accounting and Economics, Vol. 16, Nos. 1–3 (January-July), pp. 167–98.——— and D. Yermack (1999), ‘CEO Involvement in the Selection of New Board Members: An

Empirical Analysis’, Journal of Finance, Vol. 54, No. 5 (October), pp. 1,829–53.Sparrow, P. R. (1999), ‘Strategy and Cognition: Understanding the Role of Management

Knowledge Structures, Organizational Memory and Information Overload’, Creativity andInnovation Management, Vol. 8, No. 2 (June), pp. 140–49.

Useem, M. (1984), ‘The Inner Circle: Large Corporations and the Rise of Business PoliticalActivity’ (New York, Oxford: Oxford University Press).

Wade, J. B., C. A. O’Reilly and I. Chandratat (1990), ‘Golden Parachutes: CEOs and the Exerciseof Social Influence’, Administrative Science Quarterly, Vol. 35, No. 4 (December), pp. 587–603.

Watts, D.J. and S. H. Strogatz (1988), ‘Collective Dynamics of “Small-World” Networks’, Nature,Vol. 393, No. 6,684 (June), pp. 440–42.

Yermack, D. (1996), ‘Higher Market Valuations of Companies with a Small Board of Directors’,Journal of Financial Economics, Vol. 40, No. 2 (February), pp. 185–221.

——— (1997), ‘Good Timing: CEO Stock Option Awards and Company News Announcements’,Journal of Finance, Vol. 52, No. 2 (June), pp. 449–76.


Documents

Is Busy Really Busy? Board Governance Revisited