Monitoring the Monitor: Distracted Institutional Investors and Board Governance * Claire Yang Liu UNSW Business School UNSW Australia yang.liu11@unsw.edu.au Angie Low Nanyang Business School Nanyang Technological University aaclow@ntu.edu.sg Ron Masulis UNSW Business School UNSW Australia ron.masulis@unsw.edu.au Le Zhang UNSW Business School UNSW Australia le.zhang@unsw.edu.au This Version: May 18, 2017 * We thank Chang-Mo Kang, E. Han Kim, Bo Li, Mark Humphrey-Jenner, Ji Yeol Jimmy Oh, Paul Karehnke, Peter Pham, David Reeb, Rik Sen, Elvira Sojli, Buvaneshwaran Venugopal, Wing Wah Tham, and Robert Tumarkin, as well as conference participants from the FIRN Annual Conference, Conference on Asia-Pacific Financial Markets, Australasian Finance and Banking Conference, and seminar participants at Georgia Institute of Technology and UNSW Australia for helpful comments.
Monitoring the Monitor: Distracted Institutional Investors and Board Governance Abstract While boards are crucial to shareholder wealth, the literature often overlooks how continuous shareholder oversight can improve director incentives. Using exogenous industry shocks in investors portfolio firms, we find that institutional investor distraction weakens board oversight. Distracted institutions are less likely to use their votes to discipline ineffective independent directors, and directors with poor voting outcomes are less likely to turnover. As a result, directors have less monitoring incentives with independent directors missing more meetings and boards holding fewer meetings. Board structures are also compromised with the appointment of more ineffective independent directors and lead. to worse governance outcomes such as greater earnings management, higher excess CEO pay, and lower firm valuation. Our findings suggest that continuous institutional investor attention represents an important determinant of board monitoring incentives. Keywords: Board of directors, Shareholder activism, Institutional investors, Board monitoring, Shareholder voting, Corporate governance. JEL classification: G23, G34.
1. Introduction Large shareholders and the board of directors are two key corporate governance mechanisms. Due to the legal and regulatory limits on the power of large shareholders in the United States, 1 the board of directors is generally considered to have a critical role in corporate governance. The board serves as gatekeeper of all shareholder proposals to amend the charter and also as the approver of almost all major corporate decisions. The board is also charged with monitoring management, hiring and firing of CEOs, and setting executive compensation. While the board is a powerful governance mechanism for monitoring managers, director monitoring incentives do not appear to be particularly strong. This raises important questions on how reliable are boards of directors in representing shareholder interests. What motivates directors to monitor? Who monitors the board monitors? To address these questions, we examine whether institutional investor monitoring on a regular basis improves board incentives to exert greater effort in monitoring and motivating senior management. Several recent studies report evidence that institutional investors in general improve corporate decision making and thus, firm value. 2 The existing literature tends to emphasize shareholder actions during extreme events, e.g. proxy contests and lawsuits. However, those actions are likely to serve as a weapon of last resort. Given the significant powers boards have over major corporate decisions, it is more effective and less costly for shareholders to take precautionary actions to monitor the board on a regular basis. Our study helps to further our current understanding of director monitoring incentives, and how institutional investors improve firm performance by continually monitoring the board. 1 In the U.S. shareholders cannot directly initiate business transactions or charter amendments, nor modify those proposed by directors. In contrast, shareholders in the U.K., Japan, and France can force these decisions if the proposals receive a majority vote (Hansmann and Kraakman, 2001). 2 For instance, Doidge, Dyck, Mahmudi and Virani (2015); Appel, Gormley and Keim (2016); Kempf, Manconi and Spalt (2016); and Li, Liu, and Wu (2016). 1
Prior studies that go behind-the-scenes generally report evidence that institutional investors actively intervene by engaging management and directors in active discussions. 3 Less is understood about whether such interventions result in improved director incentives and board monitoring effectiveness. A board of directors is the only representative that shareholders have in the firm who can make or approve major corporate decisions. Thus, the incentives of boards to act in shareholder interests are very important in minimizing shareholder-manager agency problems. Yet, it is unclear, whether the labour market for directors sufficiently punishes poorly-performing directors (Fahlenbrach, Low, and Stulz, 2016). 4 Furthermore, independent directors generally have weak financial incentives to actively monitor a firm s management on a consistent basis (Yermack, 2004). Moreover, the existing governance literature documents that certain types of directors are ineffective monitors. For example, independent directors who are more socially dependent on the CEO, busy directors, non-domestic directors, co-opted directors and possibly old directors tend to be less effective monitors. 5 In the absence of effective board monitoring, institutional investors can be exposed to severe agency problems and experience significant losses. Thus, monitoring boards to ensure they perform their fiduciary duties should be a critical channel through which outside investors seek to maximize their returns on investments. Several studies have argued that institutional investors may not actively intervene to improve firm governance due to the classical free-rider problem as discussed in Grossman and Hart (1980) and Shleifer and Vishny (1986). Furthermore, the threat of exit makes active intervention less important as the potential stock price declines that result from exit may be a 3 See e.g. Becht et al. (2010); Carleton, Nelson, and Weisbach (1998); McCahery, Sautner, and Starks (2016). 4 Past literature on the directorial labor market impact of poor monitoring by directors mostly focuses on extreme events such as earnings restatements (Srinivasan, 2005), financial fraud lawsuits (Fich and Shivdasani, 2007), bankruptcies (Gilson, 1990), and option backdating (Ertimur, Ferri, and Maber, 2012). 5 See for example, Chidambaran, Kedia and Prabhala (2011), Coles, Daniel, and Naveen (2014), Falato, Kadyrzhanova, and Lel (2014), Fich and Shivdasani (2006), Fracassi and Tate (2012), Hwang and Kim (2009) (2012), Masulis and Mobbs (2014), and Nguyen (2012). 2
sufficient threat to motivate managers to work hard (Kahn and Winton, 1998; Maug, 1998). Even if institutional investors actively intervene to change firm policies, it is unclear whether shareholder monitoring and board monitoring are close substitutes for mitigating managershareholder agency conflicts. Institutional shareholders can directly monitor management and thus, obtain direct influence over major corporate decisions, which negates the need to otherwise intervene in board governance. Also, activist institutional investors can join the boards to directly effect changes (Gow, Shin, and Srinivasan, 2014) without the need to influence monitoring by other board members. Thus, it is unclear whether institutions actively intervene to affect board governance. We construct an exogenous measure of institutional shareholder monitoring to test whether institutional monitoring affects board monitoring incentives. Following Kempf, Manconi, and Spalt (2016), we address such endogeneity concerns by utilizing exogenous variations in institutional shareholder attention, caused by unrelated industry shocks to other portfolio firms held by the institutional investors. We use these shocks to capture reductions in the level of institutional shareholder monitoring. The following example illustrates how such an exogenous shock to outside monitoring of the board can occur. Suppose a mutual fund investor has two large stockholdings belonging to two unrelated industries, for example a bank and a pharmaceutical firm. When the pharmaceutical industry is experiencing a large return shock, the mutual fund manager needs to allocate more time and effort to monitor the pharmaceutical firm. Assuming the attention and effort of a fund manager is in limited supply, we expect the bank to receive less attention. Hence, this shock reduces a mutual fund manager s monitoring intensity of the bank. 6 To the extent that these industry-level shocks to a fund s 6 In a similar manner, Kacperczyk, Van Nieuwerburgh, and Veldkamp (2016) model how mutual fund managers, due to their finite mental capacity, optimally choose to allocate their limited attention to different information depending on the business cycle. Consistent with the idea of limited resources on the part of institutional investors, in their survey of institutional investors, McCahery, Sautner, and Starks (2016) find that limited resources (personnel) and too many firms in our portfolio rank as important impediments to shareholder activism. 3
portfolio firms are unrelated to the focal firm s fundamentals, this measure captures the exogenous variation in institutional investor monitoring intensity that is orthogonal to the focal firm s fundamentals. Generalizing on the above example, we aggregate industry shocks using the weights of the shocked industries in an institutional shareholder s portfolio, to construct an investor-level measure of exogenous distractions experienced by each institutional shareholder towards a given firm in a given quarter. Next, we construct a focal firm-level investor distraction measure by summing the distraction levels of all of its institutional investors. Kempf, Maconi, and Spalt (2016) convincingly show that this distraction measure is negatively related to the amount of attention institutional investors spend on monitoring a firm s activities such as participating in conference calls and initiating governance-related proposals. To investigate whether investor distraction affects director incentives to monitor, we first examine the voting behavior of institutional investors in director elections. Shareholder voting in director elections represents one primary mechanism for institutional shareholders to exert influence over a firm s board and ultimately over its corporate decisions. While directors rarely fail to be re-elected, except in proxy fights, experiencing disciplinary votes can be embarrassing to a director and adversely affect her reputation and likelihood of being renominated in the future (Aggarwal, Dahiya, and Prabhala, 2015). Therefore, shareholder voting may motivate independent directors to act in shareholder interests due to these adverse reputational impacts. Using an investor-level measure of mutual fund distraction, we find that mutual funds are less likely to discipline directors with negative votes when they are distracted. Economically, a one standard deviation rise in a mutual fund investor s distraction level is associated with a fall in the likelihood that the investor withholds or votes against an independent director candidate in the annual shareholder meeting by 3.3%. This effect is 4
stronger for problematic director candidates, defined as those who are busy directors, directors socially linked to the CEO, and foreign directors. 7 Such directors are shown in prior studies to be ineffective monitors. Next, we examine how the distraction by multiple institutional investors of a firm affects director voting outcomes. We find that independent directors in general receive significantly fewer disciplinary votes from institutional shareholders when these investors are distracted. Economically, one standard deviation increase in investor distraction reduces the likelihood of an independent director receiving poor vote results by almost 5%. Consistent with our fund vote results, this effect is stronger for problematic director candidates. In addition, the sensitivity of director departure to poor election vote results is also significantly lower when institutional shareholders are distracted, consistent with weaker disciplinary effects of shareholder votes. Taken together, the voting results indicate that independent directors, especially problematic ones, are significantly less likely to be disciplined by voting outcomes when institutional shareholders are distracted. Next, we examine whether weakened board oversight by institutional shareholders affects director monitoring incentives. We find that independent directors miss more board meetings and boards hold fewer meetings when institutional shareholders are distracted. For example, we find that one standard deviation rise in institutional investor distractions is associated with an almost 16% rise in the likelihood of poor director meeting attendance. Furthermore, institutional investor distraction leads to more problematic independent directors on the board, due to the increased likelihood of a new 7 Endogeneity issues such as unobserved heterogeneity across mutual funds, directors, and firms are unlikely to be driving our results. As our dataset comprises of the voting outcomes of each mutual fund for individual director elections for each firm and each year, we exploit within-investor differences in monitoring intensity across their different portfolio firms for each year by controlling for fund-year fixed effects. We also exploit within-election proposal differences in monitoring intensity across the mutual funds voting on each election proposal by including individual director election proposal fixed effects, which essentially accounts for time-varying director and firm characteristics. 5
problematic director being appointed to the board and retention of existing problematic directors. These findings suggest that the institutional investor monitoring has important implications for director monitoring incentives and efforts. Finally, we examine how the reduction in board monitoring efforts and incentives due to investor distraction affects several governance outcomes. Firms with distracted investors exhibit significantly greater earnings management, grant their CEOs higher compensation, but with less performance-based incentives, and generally have lower equity valuation. Importantly, the negative impact of investor distraction on these governance outcomes are amplified in firms where problematic directors sit on key committees or where board monitoring efforts are compromised. Taken together, our results suggest that institutional shareholder distraction leads to significantly poorer board monitoring quality and effectiveness. Our study contributes to the literature on corporate governance in several ways. First, we extend our understanding of what motivates independent directors to do their job well and monitor management carefully. While it is well known that boards make important corporate decisions which have economically large impacts on shareholder value, director incentives to monitor managers are not well-understood. It is unclear why directors with limited financial incentives are willing to exert sufficient effort to closely monitor managers (Yermack, 2004). Fama and Jensen (1983) argue that director reputational concerns provide a strong motivation, and recent studies show that reputational concerns affect director incentives to perform their roles as effective monitors. 8 We advance our understanding of director incentives by showing that institutional investor oversight of boards significantly improves director incentives to exert efforts in monitoring the senior management. 8 For instance, Jiang, Wan, and Zhao (2016); Masulis and Mobbs (2014); Levitt and Malenko (2016). 6
Second, our study furthers our understanding of how institutional investors intervene to improve board governance and firm value. Existing studies that examine shareholder interventions in corporate governance emphasize the actions of shareholder activists during extreme events, including the use of proxy contests and law suits. 9 However, evidence on shareholder actions to improve board functioning on a regular basis is surprisingly scarce. 10 We help fill this gap in the existing literature, and show that institutional investors use their disciplinary votes to monitor and discipline directors on a regular basis. Our study also differs substantially from Kempf, Manconi, and Spalt (2016) who show that firms are more likely to undertake diversifying acquisitions and grant their CEOs opportunistically-timed equity grants when institutional investors are distracted. Given that acquisitions and CEO pay are within the board s decision domain, we show that the underlying mechanism through which distracted shareholders can impact firm policies is through board monitoring and consultations. In particular, our study highlights the important role that institutional investors play in monitoring the board. We show that reduced institutional investor monitoring leads to poorer board effectiveness, in part due to independent directors reducing their own monitoring efforts in response to reduced voting pressure, and in part due to poorly chosen board appointments. Our study is also related to the literature that explores how corporate governance mechanisms interact with each other. Existing evidence is mixed as to how governance mechanisms interact (Agrawal and Knoeber, 1996; Cremers and Nair, 2005; Gillan, Hartzell, and Starks, 2011), and it remains unclear whether the monitoring roles of the board and 9 See e.g., Brav et al. (2008) on hedge fund activism, Del Guercio and Hawkins (1999) on the impact of shareholder proposals put forward by public pension funds, Doidge et al. (2016) on the activities of Canadian Coalition for Good Governance, a formal collective action organization of institutional investors, Del Guercio, Seery, and Woidtke (2008) on vote-no campaigns, Gillian and Starks (2000) on detailed analysis of shareholder proposal voting outcomes. See also Gillian and Starks (2007), Denes, Karpoff, and McWilliams (2016) for comprehensive reviews on shareholder activism. 10 Outside shareholders can submit Rule 14a-8 shareholder proposals relating to board independence and other board issues but they are often ineffective in eliciting change (Gillian and Starks, 2007; Denes, Karpoff, and McWilliams, 2016). Activist shareholders can also organize just vote no campaigns to withhold votes from directors but Del Guercio, Seery, and Woidtke (2008) show that such campaigns often target at large, poorly performing firms and are typically sponsored by public pension funds. 7
institutional shareholders are complements or substitutes. We contribute to this literature by showing that the monitoring function of corporate boards, a key internal governance mechanism, depends crucially on the effectiveness of institutional shareholder monitoring of directors, which acts as a complementary governance mechanism. Lastly, we contribute to the literature that examines the impact of institutional investor monitoring on firm policies and governance outcomes. Monitoring by institutional investors is found to have a positive impact on a firm s governance indices, CEO compensation, mergers and acquisitions profitability, firm risk-taking, and earnings management 11 However, endogeneity makes it difficult to assess the causal dimension of these effects. For example, Chung and Zhang (2011) conclude that the positive association between institutional ownership and good governance structure is driven by institutional investors gravitating towards firms with good governance so as to minimize their own monitoring costs. Recent studies have used the annual reconstitution of the Russell 1000 and 2000 indexes as an exogenous shock to examine how changes in passive institutional ownership affect firm governance and policies. 12 Our study differs from theirs in that we focus on how institutional shareholder monitoring intensity affects board governance as a crucial internal governance mechanism. Additionally, we examine a wide range of institutional investor classes, not just passive index investors. 11 Examples include: Hartzell and Starks (2003), Aggarwal et al. (2011), Chen, Harford, and Li (2007), Cornett, Marcus, Tehranian (2008), Aghion, Van Reenen, and Zingales (2013), and Kim and Lu (2011). 12 Appel, Gormley, and Keim (2016) find that an increase in passive institutional ownership increases board independence while Schmidt and Fahlenbrach (2016) find no change in board independence and appointments of independent directors are met with worse announcement returns when passive institutional ownership increases. 8
2. Variable Construction, Data, and Descriptive Statistics 2.1. Construction of institutional investor distraction We follow Kempf, Manconi, and Spalt (2016) to construct a measure for institutional investor s distraction in three steps. For each institutional investor i in a given firm f, we first identify extreme returns for industry sectors that are unrelated to firm f in investor i s portfolio. These unrelated industry shocks are likely to cause investor i to shift his attention away from firm f. To obtain investor i s level of distraction away from firm f, we then aggregate these industry shocks across all the unrelated industry sectors in investor i s portfolio. Finally, to get the firm-level distraction measure of firm f, we aggregate the investor-level distraction measure across all institutional investors of the firm. Specifically, we define the level of distraction experienced by the institutional investors of firm f in a given quarter q as: Total Distraction f,q = w i,f,q 1 i f IND INDf w IND i,q 1 IS q IND where i denotes firm f s institutional investor at the end of quarter q-1, IND denotes Fama- French 12 (FF12) industry sector, and INDf denotes the industry sector in which firm f belongs. FF12 industry sector represents a broad industry classification. It follows that industry-level events should generally be unrelated to the fundamentals of individual firms in other FF12 industries. 13 IS q IND is an indicator variable equals to one if industry IND experiences a shock in quarter q, and zero otherwise. An industry is deemed to have experienced a shock if the industry s return for the quarter is either the highest or lowest across all FF12 industry sectors. We weight these shocks to capture their economic importance. The variable w IND i,q 1 denotes the weight of industry sector IND in investor i s portfolio in the prior quarter q-1, which represents the importance of industry sector IND in institutional investor i s portfolio. 13 The potential supply-chain links between firms across different Fama-French 12 industries will reduce the shifts of investor attention to the shocked industries, thus biasing against us finding results. 9
IND IND The sum of the products of w i,q 1 and IS q across all industries unrelated to firm f, capture investor i s level of distraction away from firm f, when events occur in unrelated industries of investor i s portfolio. To aggregate across institutional investors of a firm, we weight the level of distraction of each investor i in firm f by w i,f,q 1 which measures the importance of investor i in firm f in the prior quarter, q-1. Intuitively, investor i have more weight if 1) firm f has more weight in i s portfolio and 2) investor i owns a larger fraction of firm f s shares. Following Kempf et al. (2016), we compute w i,f,q 1 as: w i,f,q 1 = QPFweight i,f,q 1 + QPercOwn i,f,q 1 (QPFweight i,f,q 1 + QPercOwn i,f,q 1 ) i F,q 1 where PFweight i,f,q 1 is the weight of the market value of firm f in investor i s portfolio, and PercOwn i,f,q 1 is investor i s percentage ownership in firm f. The former measures how much time the investor is likely to spend in analysing firm f, and the latter measures how much influence investor i can have in the firm. We sort all stocks held by investor i into quintiles by PFweight i,f,q 1 and all investors of firm f into quintiles by PercOwn i,f,q 1. QPFweight i,f,q 1 and QPercOwn i,f,q 1 takes a value from 1 to 5 depending on which quintile PFweight i,f,q 1 and PercOwn i,f,q 1, respectively, falls in. The weights w i,f,q 1 sum to 1 for each firm f after scaling by the denominator. It follows that higher values of Total Distraction indicates that the institutional shareholders in firm f are more distracted by attention-grabbing returns from unrelated industry sectors, and therefore their overall monitoring intensity of firm f s board is reduced. As we have stockholding data and voting data by each individual mutual fund, in addition to our main firm-level distraction measure, we can also calculate an investor-level distraction measure for each fund using a similar approach as above so as to examine how fundlevel distractions affect their voting behaviour. 10
2.2. Validity of the distraction measure There are two important advantages of measuring institutional investor distractions in this way. First, to the extent that the return shocks are in unrelated industries, this measure captures exogenous variation in institutional shareholder monitoring, since the return shocks from other FF12 industries are unlikely to be driven by the focal firm s fundamentals. This helps to alleviate reverse causality issues and issues relating to omitted variables, which might affect both institutional investor monitoring levels and firm behaviour. Second, the investorlevel distraction measure differs across the firms held by each investor due to the way it is constructed. Thus, we are able to compare the within-investor difference in levels of distractions towards each firm held in the investor s portfolio, essentially taking into account the selection issues whereby certain institutions select into firms with specific features due to their preferences. We further evaluate whether return shocks in unrelated industries can cause permanent changes in the focal firm s board governance and performance. We first check the length of the extreme industry-level returns, and find that each industry return shock on average only last for 1.25 quarters and the maximum length of continuous return shocks is only 2 quarters. These short-lived industry-level return shocks are likely to be random events, and unlikely to cause institutional investors to significantly rebalance their portfolios. Consistent with our expectation, Kempf, Manconi, and Spalt (2016) find that the focal firm is unlikely to experience a large change in its portfolio weight in the institution s portfolio when the institutional investor is distracted by events elsewhere. Therefore, unrelated industry shocks are unlikely to drive changes in the focal firm s board governance through changes in the stockholdings of the distracted institutional investor. Nevertheless, non-shocked industries can continuously experience lowered institutional investor monitoring intensities for relatively longer periods, as investors 11
continuously shift their attention to different shocked industries over time. Suppose a mutual fund investor has 4 stocks in unrelated industries, for example a bank, a pharmaceutical firm, a food retailer, and a manufacturing firm. Suppose the pharmaceutical industry experiences a large return shock in quarter 1, the food retailer experiences a large return shock in quarter 2, and the manufacturing industry experiences a large return shock in quarter 3. As the mutual fund manager continuously shifts her attention to different industries other than the financial services industry, the bank is being overlooked during this period. In untablulated analysis, we find that this distraction period on average lasts for 7 quarters. Thus, looser investor monitoring due to distractions can lead to significant changes in the focal firm s board governance and its long-run operating performance. 2.3. Data and sample formation Our main sample comes from linking several well-known databases. We start with firmyears in the RiskMetrics director database, which contains information on board structure and director characteristics. We obtain information on institutional investor shareholdings from the Thomson-Reuters Institutional Holdings (13F) database to construct the investor distraction measure. We then merge the director and institutional holding data with firm accounting and stock returns data from Compustat and CRSP, respectively. We drop firm-years with missing information for our distraction measure, board structure, or key financial and stock return variables. We also exclude the heavily regulated finance and utility industries. Further, we exclude firms with dual-class share structures and closely-held firms where insiders or directors as a group hold more than 50% of shares, because institutional shareholders are unlikely to have much influence on the corporate governance of these firms. 14 We focus on independent 14 These exclusions reduce the initial sample size by about 10% and leads to a cleaner sample for the purpose of our analysis. 12
directors in this study because they are the primary board monitors. 15 Our final sample consists of 121,205 independent director-firm-years from 15,575 firm-year observations for the period 1996 to 2013. However, the sample size may vary across tests due to availability of control variables and dependent variables. To examine board characteristics and composition, we require CEO-director social ties information from BoardEx. Although BoardEx has data from 2000, it becomes better populated only from 2003 and onwards. Therefore, for most of our tests involving board structure, we start the sample period from 2003 and end in 2013 resulting in 7,663 firm-year observations. To examine voting outcomes, we also obtain shareholder voting data from ISS Voting Analytics for the period 2003 to 2012. We merge ISS Voting Analytics with RiskMetrics director data using company and director names, and further merged this data with CEOdirector social ties information from BoardEx. After requiring non-missing voting data and social ties data, we end up with 30,310 individual director elections. We call this sample the director votes sample. In these elections, we observe 21,127 distinct mutual fund-years and 2,081,611 votes. 2.4. Descriptive statistics Table 1 provides summary statistics for our key variables. Detailed descriptions of all of the variables can be found in Appendix A. We winsorize all the continuous dependent and control variables at 1% and 99%. Panel A summarizes the means, medians, 25 th and 75 th percentiles, and the standard deviation of the institutional investor distraction measures. For our fund-level distraction measure, the mean and median mutual fund distraction levels in our 15 We follow the director classification in Riskmetrics. Independent directors are outside directors that have no family or economic ties to management or the firm they monitor other than that through their directorship. RiskMetrics primarily relies on the NYSE and Nasdaq listing rules to classify independent directors, and identifies independent directors based on proxy statements and disclosures of related transactions. Examples of nonindependence using the RiskMetrics definition include but are not limited to: former employees of the firm or subsidiaries, major shareholders, customers, suppliers, and family members of executives. 13
sample are both 0.14, with a standard deviation of 0.06. The mean and median firm-level distraction measure, Total Distraction, is 0.17 and 0.16 respectively, while the 25th percentile and 75th percentile values are 0.13 and 0.21, respectively. By construction, Total Distraction has a minimum value of 0.05. The distribution of our distraction measure across firms is in line with Kempf, Manconi, and Spalt (2016). Panel B reports summary statistics for director election outcomes and director characteristics for our director voting sample with data on voting outcomes and social ties. Among the 30,310 independent director election votes, on average 5% were No votes, i.e., defined as against or withheld votes. Across all director elections, the median director election sees only 2% of the votes being cast as No votes and only 5% cast as negative votes at the 75 th percentile. Therefore, negative votes are rare. At the maximum, 74% of the votes were negative votes. We also find that only 11% of directors receive a poor voting outcome, which we define as elections where a director receives more than 10% of No votes. Furthermore, 6% of the ISS director election recommendations are negative. These proposallevel statistics are similar to those in Cai, Garner, and Walkling (2009). Almost 25% of the independent directors in the director voting sample are considered problematic which we define as an independent director who is busy, socially connected to the CEO, and/or foreign based. A busy director is one who holds 3 or more directorships in total during the year (Fich and Shivdasani, 2006). Directors who attend the same educational institutions and/or the same non-business organization as the CEO are deemed to be socially connected to the CEO. 16 A foreign director is one who primarily works for a non-u.s. employer 16 Although BoardEx uses a unique identifier for each of the educational institutions in the database, the same educational institution may appear under different identifiers (e.g. Harvard University and Harvard Business School). To remedy this problem, we manually match the names of the educational institutions and create new identifiers that uniquely identify each educational institution. For shared social ties at non-business organizations, we include connections at charities, social clubs, and armed forces, and exclude those compulsory professional and industrial organizations where social interaction is unlikely due to the compulsory nature of membership in the organizations (e.g. American Bar Association). 14
or last worked for a foreign company if already retired (Masulis, Wang, and Xie, 2012). Among the problematic directors, 18% are busy directors, 13% are socially dependent directors, and 3% are foreign directors. About 16% of directors leave the board and are not renominated by the board in the subsequent director elections. We use data from Voting Analytics to determine whether a particular director is being renominated by the board or not. In Panel C, we report the director characteristics for our full sample of director-firmyear observations from 1996 to 2013. We find that only 1% of the independent directors in our sample have attendance problems defined as missing 25% or more of board meetings. Independent directors have an average of 1.6 total directorships, while the median number of directorships is 1. In addition, the average director is 63 years old with board tenure averaging 11 years. Moreover, independent directors generally have weak firm financial incentives with a director s mean (median) equity stake in the firm representing only 0.07% (0.02%) of outstanding shares. We report board characteristics for the subsample of firm-years with information from BoardEx. On average 21% of the independent directors on the board are considered problematic. About 10% of new director appointments and 66% renominations are problematic. We also look at the distribution of problematic directors on the major committees. On average about 21% to 22% of the audit, compensation, and nomination committee members are problematic. In Panel E, we report descriptive statistics for firm characteristics at the firm-year level. Boards on average hold 8 meetings per year. The average board has 9 members, out of whom 71% are classified as independent. CEOs own about 2.5% of shares outstanding. Finally, about 60% of the firm-years have a staggered board in our sample. 15
3. Shareholder Distraction and Voting 3.1. Investor-level mutual fund distraction and fund votes In this section, we examine voting behavior of institutional investors in director elections as one critical channel of investor monitoring of director quality and performance, and a disciplinary mechanism for poorly-performing directors. Although the vast majority of director elections are uncontested and the rejection of a standing director is extremely rare, Aggarwal, Dahiya, and Prabhala (2015) show poor vote results have disciplinary effects in themselves. Directors with weak voting support tend to depart from the board and even if they remain on the board, they get demoted to less important board positions. These poorlyperforming directors also suffer from significant reputational losses in the directorial labor market in that they gain fewer additional board appointments and lose more of their other board seats than other independent directors. We first look at investor-level distraction and their individual voting behavior in Table 2 Panel A. Since 2003, mutual funds are required to publicly disclose their voting behavior through N-PX filings with the Securities and Exchange Commission (SEC). Therefore, our analysis of mutual fund voting starts in 2003. 17 We focus on actively-managed equity funds as such funds are most likely to gain from active intervention. We provide more details about the formation of our voting sample in Appendix B. For each director election proposal, depending on whether it is under plurality or majority voting, mutual funds can vote For, Against, or withhold their vote. 18 According to Cai, Garner, and Walkling (2009), shareholders can express their dissatisfaction by withholding votes, thus Withhold is functionally equivalent to Against. Therefore, we classify Withhold as well as Against votes as No or negative votes. Our dependent variable in Panel A is Oppose Director, which is an indicator variable 17 In the earlier years, Voting Analytics only covers the largest fund families. 18 We include director elections involving plurality voting and majority voting. Under plurality voting, shareholders can vote For or withhold their votes while in majority voting, shareholders can vote For or vote Against a director (Ertimur, Ferri, and Oesch, 2015). 16
that equals one if the mutual fund cast a negative vote for a particular director in a given annual election and is zero otherwise. We find that Oppose Director is equal to one in 6% of cases. Due to their relative rarity, negative votes by mutual funds should have strong disciplinary effects on independent directors. The key independent variable in Table 2 Panel A is Mutual Fund Distraction, which is the level of distraction experienced by the voting mutual fund in the past four quarters immediately preceding the voting date. We report linear probability model (LPM) estimates with two sets of fixed effects and with clustering of standard errors at the fund level. The first set of fixed effects is interacted fund and year fixed effects, which allows us to compare the vote outcomes at firms where the voting mutual fund is more distracted against vote outcomes at firms where the same fund is less distracted during the same year. The second set of fixed effects are director election proposal fixed effects, which allow us to control for the characteristics of each election proposal, including time-varying director characteristics and time-varying firm characteristics including performance. For instance, the ISS recommendation for the director candidate and shareholder activist events like just vote no campaigns are taken into account using director election proposal fixed effects. In Model (1) of Panel A, we find that the coefficient on Mutual Fund Distraction is -0.033 and statistically significant at the 5% level, suggesting that when mutual funds are distracted, they are less likely to oppose management recommendations in independent director elections. Economically, a one standard deviation increase (decrease) in institutional investor distractions reduces the chance that the investor votes against (for) an independent director candidate in the election by 3.3% (0.033*0.06/6%), adjusted for the unconditional probability of voting against an election proposal (6%). We then examine whether problematic directors in particular are less likely to be disciplined when institutional investors are distracted. We identify three types of independent 17
directors whose monitoring incentives are likely to be compromised. The first type of problematic director is an independent director who is socially connected to the CEO. Socially connected directors tend to be friendlier to management and thus, are ineffective monitors (Hwang and Kim 2009, 2012; Chidambaran, Kedia, and Prabhala, 2011; Fracassi and Tate, 2012; Nguyen, 2012). For example, studies show that CEO-director ties lead to a decline in CEO turnover-performance sensitivity, and CEOs in such firms also extract more private benefits, which significantly reduce firm values. The second type of problematic director is an independent director who is excessively busy due to sitting on multiple boards. Busy directors, defined as those with 3 or more total directorships during the year, are likely to be overcommitted and tend to be less effective monitors (Fich and Shivdasani, 2006; Falato, Kadyrzhanova, and Lel, 2014; Masulis and Mobbs, 2014). The third type of problematic director is a foreign independent director whose primary employer is a non-us company. Masulis, Wang and Xie (2012) find that foreign directors are less effective board monitors, and they tend to miss more board meetings, accept greater earnings management, higher CEO pay, leading to poor firm performance. In Models (2) and (3), we split the independent director sample into problematic and non-problematic candidates. We find that most of the significant effects we found earlier come from the subsample of problematic director candidates in Model (2), and we observe no significant effects for non-problematic director candidates in Model (3). Given that most investors tend to support the director election proposals in general, non-problematic directors are likely to receive support regardless of whether the investor is distracted. Therefore, we do not observe a significant impact of mutual fund distraction on voting outcomes for the nonproblematic directors. However, problematic directors often receive less support in general as vigilant investors vote against these underperforming directors. Yet, when investors are distracted they tend to ignore the underperformance of these problematic directors and support 18
them in the elections. These findings suggest that directors, especially problematic directors, are less likely to be disciplined during general elections when mutual fund investors are distracted. 19 Mutual funds within the same fund families may have similar voting patterns. To account for voting patterns by fund families, we aggregate the mutual fund distraction and voting behavior at fund family level and report the results in Panel B. The dependent variable in Panel B is the average of Oppose Director across all funds in the same family, and the key independent variable is Mean Mutual Fund Distraction, the average distraction of all the funds in the same family. We use interacted fund family, year fixed effects and election proposal fixed effects, and standard errors are clustered by fund family. After accounting for the similar voting patterns within the same mutual fund families, we find similar and consistent evidence that mutual fund investors are less likely to vote against directors, especially problematic ones, when they are distracted. In untabulated tests, we exclude family firms where institutional investors are likely to have more limited influence and we find that the link between mutual fund distraction and their voting pattern is stronger. We also use a larger sample where we include both actively-managed and passively-managed funds. We find that the results are weaker, but remain qualitatively similar. Additionally, using Bushee s classification of institution types, we find a stronger distraction effect for mutual funds that are affiliated to institutions who are more likely to monitor, such as investment companies, independent investment advisors, and public pension funds. 19 Another possible measure of ineffective directors is to use ISS recommendations. However, ISS recommendations have important limitations. For example, ISS have been accused of making blanket recommendations that are uniform recommendations for or against certain types of directors or firm characteristics. Furthermore, using ISS recommendations to measure ineffective directors would require the additional assumption that investors are active voters and do not rely on ISS recommendations when they are not distracted. But this assumption is unlikely to be valid as Iliev and Lowry (2014) find that mutual fund investors vary greatly on their reliance on ISS recommendations with a substantial portion of funds relying solely on ISS recommendations. 19
3.2. Firm-level investor distraction and director election results In Table 3, we examine the impact of distractions by a firm s institutional investors on director election outcomes at the firm-level. Our dependent variable is the Fraction of "No" Votes for a Director and the key independent variable is Total Distraction, defined as the average distraction of the firm s institutional investors over the past 4 quarters immediately before the meeting date. In Models (1) and (2), we examine the impact of investor distraction on director election outcomes for all independent directors. In Models (3) to (6), we compare the impact of investor distraction for the election outcomes of problematic versus nonproblematic independent directors. We report LPM estimates and include firm and year fixed effects in all models as well as director fixed effects in the even-numbered models. We further control for director and firm characteristics that are likely to affect the director voting results, and the average fraction of No votes for all the independent directors in the firm standing for election during the year. In all these regressions, we cluster standard errors at the director level. In Models (1) and (2), we find that the coefficient on Total Distraction is negative and statistically significant, suggesting that there are fewer No votes for director elections when a firm s institutional investors are distracted. 20 Economically, a one standard deviation increase in institutional investor distractions increases the negative votes received by a director by 1.7% to 2.8% (0.014*0.06/5% or 0.023*0.06/5%). For control variables, we find that director election results are influenced by Average Director No Votes, indicating that shareholders overall satisfaction with the board members significantly affects individual director election outcomes. Also, we find that the coefficients on Negative ISS and Poor Meeting Attendance are positive and significant, which is consistent with the Cai, Garner, and Walking (2009) who 20 Model 2 has fewer observations than Model 1 because directors who only appear once in elections are dropped in Model 2 with director fixed effects. 20
find that ISS recommendations and director attendance are important determinants of director voting results. In Models (3) to (6), we find that the coefficient on investor distraction is statistically significant for the subsample of problematic independent directors, while it is insignificant for the subsample of non-problematic directors. The economic magnitude are also much bigger for the problematic directors. These findings suggest that the effect of investor distraction on director election outcomes is more pronounced for problematic independent directors, who are more likely to face negative shareholder assessments. We then repeat the above analysis using an alternative measure of director disciplinary vote, Poor Vote Result, as the dependent variable in Panel B. Poor Vote Result is an indicator variable equals to one if the percent of No votes received by a director exceeds 10% of total votes cast, and zero otherwise. 21 We obtain similar results as in Panel A. Economically, a one standard deviation increase in total distraction leads to a 4.3% or a 4.9% (0.089*0.06/11% or 0.079*0.06/11%) decrease in the likelihood that the independent director receives a poor election result, after we adjust for the unconditional probability of poor vote result (11%). Again, the results are concentrated among the sample of problematic directors. Overall, our findings support the view that directors generally face less pressure to perform in director elections when institutional investors are distracted, and this effect is more pronounced for more problematic directors who would have otherwise face more dissent when investors are more vigilant. 21 The choice of cutoffs is arbitrary. We find similar results using the 30% cutoff as in Aggarwal, Dahiya, and Prabhala (2015). 21
3.3. Sensitivity of director departures to vote results Aggarwal, Dahiya, and Prabhala (2015) find that directors are more likely to depart after they face a more negative shareholder reception in director elections and therefore, annual elections serve as a potentially important disciplining mechanism for directors. We conjecture that this disciplinary effect will be weakened when outside institutional investors are distracted. In Table 4, we examine the impact of shareholder distraction on the sensitivity of director departures to election results. Our dependent variable is Director Departure, an indicator variable equals to one if the director is not renominated by the board in the subsequent elections and zero otherwise. 22 We construct this variable by following each director from one election to the next in Voting Analytics and checking whether the management proposals include putting up the director for election. 23 The focal variables in our analysis are the interaction terms between Total Distraction and weak vote outcomes, which we measure using Fraction No Votes for a Director and Poor Vote Result Indicator. We only include directors who are aged 70 and below to exclude departures due to mandatory retirements (Fahlenbrach, Low, and Stulz, 2016). Since we require vote resultdata for this test, our sample is restricted to directors who stand up for re-elections at the end of their terms. Thus, our sample is unlikely to include director departures due to health problems and deaths. We estimate LPMs with either firm and year fixed effects or director, firm, and year fixed effects. We cluster standard errors at the director level. Table 4 Panel A presents the results. Consistent with Aggarwal, Dahiya, and Prabhala (2015), we find in all models that directors who receive less shareholder support during elections are less likely to be subsequently renominated. Consistent with our conjecture, we 22 Our results are robust to using a departure indicator using RiskMetrics Director data instead, i.e., we follow each director from one year to the next in the RiskMetrics Directors database and define a departure as having occurred if the director is no longer on the board in the subsequent years. 23 We lost some observations because we need to check 2 or 3 subsequent years to see whether the director is renominated in firms with staggered boards. 22