Emory Law Journal

The Power of Proxy Advisors: Myth or Reality?
Stephen Choi,
Jill Fisch,
Marcel Kahan Murray and Kathleen Bring Professor of Law, New York University School of Law.Perry Golkin Professor of Law, University of Pennsylvania Law School.George T. Lowy Professor of Law, New York University School of Law.The authors thank Glass Lewis, Proxy Governance, and Egan Jones for providing access to their voting recommendations and policies. Earlier drafts of this paper were presented at Duke University School of Law, University of Virginia School of Law, UCLA School of Law, and Yale Law School, and we received many useful comments at each session. We also thank Un Kyung Park for her helpful comments. Marcel Kahan thanks the Milton and Miriam Handler Foundation for financial support. The Filomen D'Agostino and Max E. Greenberg Research Fund provided summer research support for this Article.

Abstract

Recent regulatory changes increasing shareholder voting authority have focused attention on the role of proxy advisors. In particular, greater shareholder empowerment raises the question of how much proxy advisors influence voting outcomes. This Article analyzes the significance of voting recommendations issued by four proxy advisory firms in connection with uncontested director elections. We find, consistent with press reports, that Institutional Shareholder Services (ISS) is the most powerful proxy advisor and that, of the others, only Glass, Lewis & Co. seems to have a meaningful impact on shareholder voting.

This Article also attempts to measure the impact of voting recommendations on voting outcomes. Unlike prior literature, it distinguishes correlation from causality by examining both the recommendation itself and the underlying factors that may influence a shareholder’s vote. Using several different tests, we conclude that popular accounts substantially overstate the influence of ISS. Our findings reveal that the impact of an ISS recommendation is reduced greatly once company- and firm-specific factors important to investors are taken into consideration. Overall, we estimate that an ISS recommendation shifts 6%–10% of shareholder votes. We also determine that a major component of ISS’s influence stems from its role as an information agent, aggregating factors that its subscribers consider important.

Introduction

Proxy advisors—private firms that analyze corporate elections and advise investor clients on how to vote their shares—are recent and potentially powerful new players in the corporate governance world. 1 See Albert Verdam, An Exploration of the Role of Proxy Advisors in Proxy Voting (Feb. 2007) (unpublished manuscript, available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=978835) (describing the emergence of proxy advisors); see also U.S. Gov’t Accountability Office, Corporate Shareholder Meetings: Issues Relating to Firms That Advise Institutional Investors on Proxy Voting 6–12 (2007) [hereinafter GAO, Corporate Shareholder], available at http://www.gao.gov/new.items/d07765.pdf (exploring competition and potential conflicts of interest in the proxy advisor market); Colin Diamond & Irina Yevmenenko, Who Is Overseeing the Proxy Advisors?, 3 Bloomberg Corp. L.J. 606, 608 (2008) (highlighting the proxy advisor market). Institutional investors, which hold an increasing percentage of the shares of U.S. companies, 2Marcel Kahan & Edward Rock, Embattled CEOs, 88 Tex. L. Rev. 987 (2010); see also Paul Rose, The Corporate Governance Industry, 32 J. Corp. L. 887, 897 (2007) (“In 1965, institutional investors held 16% of U.S. equities; by 2001, institutional investors held 61%.”). wield substantial voting power but often lack the appropriate incentives to cast informed ballots with respect to their portfolio companies. 3According to conventional wisdom, these institutional investors generally do not care enough about their votes to cast an informed ballot. They hold shares in too many companies, so any particular stake represents a small fraction of their portfolio, and how they vote is unlikely to affect the outcome and even if it did, the effect on the value of their portfolio would be minimal. Researching the issues on a company’s annual meeting agenda is costly, and institutions may also lack the necessary expertise to evaluate these issues adequately. See, e.g., Rose, supra note 2, at 897 (“Unless an institutional investor believes that it can conduct research for less, or that more expensive but discerning research will enable it to obtain better returns (after subtracting its own research costs), the investor may be better off outsourcing its corporate governance research.”); Omari Scott Simmons, Taking the Blue Pill: The Imponderable Impact of Executive Compensation Reform, 62 SMU L. Rev. 299, 354 (2009) (“Institutional investors, despite having greater capacity to monitor and gather information, may have too small a stake in a company or too limited industry expertise to monitor it actively.”). Instead, many institutional investors employ the services of proxy advisors to assist them in exercising their voting rights. 4 See, e.g., GAO, Corporate Shareholder, supra note 1, at 13 (describing ISS’s client base as consisting of an estimated 1,700 institutional investors). The services of proxy advisors include providing research, helping investors develop voting guidelines, handling the mechanics of the voting process, and offering recommendations on each issue on a company’s agenda. 5 See, e.g., Glass, Lewis & Co., Proxy Paper: Proxy Research and Voting Recommendations on Global Proxies, http://www.glasslewis.com/solutions/proxypaper.php (last visited Aug. 11, 2009) (describing Glass Lewis’s proxy research, voting recommendations, and voting platform for voting subscribers’ shares); RiskMetrics Group, Custom Proxy Advisory, http://www.riskmetrics.com/custom_proxy_advisory (last visited Aug. 11, 2009) (describing how ISS works with clients to develop customized voting guidelines). In some cases, institutional investors may even subcontract their voting decisions to proxy advisors. 6RiskMetrics Group, Proxy Advisory Services, http://www.riskmetrics.com/proxy_advisory/options (last visited Aug. 11, 2009) (detailing the choice of ISS guidelines that subscribers can use and incorporate into “RiskMetrics’ turnkey voting agency services”).

As a result of their capacity to influence voting, proxy advisors are regarded as very powerful. 7 See generally GAO, Corporate Shareholder, supra note 1. The popular, business, and academic media describe ISS (Institutional Shareholder Services, a division of RiskMetrics), the proxy advisor with the largest client base, 8 See id. at 4. and Glass, Lewis & Co., which has the second largest client base, 9Stephen Davis, White Knight Swoops in for Glass Lewis, Directorship, Dec. 2007/Jan. 2008, at 7 (“Glass Lewis is the world’s second biggest proxy adviser next to RiskMetrics . . . .”). as “influential,” 10 See, e.g., Pallavi Gogoi, Support for Bank of America CEO Wanes; Shareholders Meet Today, and Many Want Him Out, USA Today, Apr. 29, 2009, at B1 (describing RiskMetrics as “[i]nfluential”); Robert D. Hershey Jr., A Little Industry with a Lot of Sway on Proxy Votes, N.Y. Times, June 18, 2006, § 3, at 6 (quoting David W. Smith, president of the Society of Corporate Secretaries and Governance Professionals, stating that “‘[t]he influence [proxy] advisers wield is extraordinary’”). “powerful,” 11 See, e.g., Matt O’Sullivan, When Only a Corporate Jet Will Do, Sydney Morning Herald, May 28, 2009, at 25 (describing RiskMetrics as “America’s most powerful shareholder voting adviser”). and having great “clout.” 12 See, e.g., Kim Clark, Reading Proxies for Fun and Profit, U.S. News & World Rep., May 22, 2006, at EE10 (describing Glass Lewis’s “growing clout”); Joann S. Lublin, RiskMetrics’s Head Faces His Day of Shareholder Judgment, Wall St. J., June 2, 2008, at C1 (“ISS Governance Services . . . exerts tremendous clout in advising institutional investors on proxy fights . . . .”). Commentators have claimed that ISS alone is able to influence shareholder votes by 19%, 13Jie Cai, Jacqueline L. Garner & Ralph A. Walkling, Electing Directors, 64 J. Fin. 2389, 2404 (2009). 13.6 to 20.6%, 14Jennifer E. Bethel & Stuart L. Gillan, The Impact of the Institutional and Regulatory Environment on Shareholder Voting, Fin. Mgmt, Winter 2002, at 29, 30. 30%, 15Posting of William J. Holstein to BNET: The Corner Office, http://blogs.bnet.com/ceo/?p=1100&tag=content;col1 (Feb. 7, 2008, 08:03). and even “a third or more.” 16 See Rose, supra note 2, at 889 (attributing this view to executives). The collective power of proxy advisors arguably is even greater. As a result of this influence, management and shareholder activists alike frequently lobby ISS to endorse their respective positions. As related by Delaware’s Vice-Chancellor Leo Strine:

[P]owerful CEOs come on bended knee to Rockville, Maryland, where ISS resides, to persuade the managers of ISS of the merits of their views about issues like proposed mergers, executive compensation, and poison pills. They do so because the CEOs recognize that some institutional investors will simply follow ISS’s advice rather than do any thinking of their own. ISS has been so successful that it now has a California rival, Glass Lewis. 17Leo E. Strine, Jr., The Delaware Way: How We Do Corporate Law and Some of the New Challenges We (and Europe) Face, 30 Del. J. Corp. L. 673, 688 (2005).

Similarly, commentators have observed that “boards may do what they believe ISS wants them to in order to keep their seats, whether or not their belief is justified.” 18Diamond & Yevmenenko, supra note 1, at 617.

This influence is troubling in light of the limited accountability of proxy advisors. Proxy advisors do not have a financial stake in the companies about which they provide voting advice; they owe no fiduciary duties to the shareholders of these companies; 19Leo E. Strine, Jr., Toward a True Corporate Republic: A Traditionalist Response to Bebchuk’s Solution for Improving Corporate America, 119 Harv. L. Rev. 1759, 1765 (2006) (“Unlike corporate managers, neither institutional investors as stockholders nor ISS as a voting advisor owe fiduciary duties to the corporations whose policies they seek to influence.”). and they are not subject to any meaningful regulation. 20 See GAO, Corporate Shareholder, supra note 1, at 8–9 (observing that, as pension consultants, ISS and Proxy Governance, Inc. (PGI) are registered with the SEC as investment advisors while Glass Lewis and Egan-Jones are not registered as investment advisors). Moreover, it is not clear that the proxy advisory industry is sufficiently competitive and transparent to subject advisory firms—ISS in particular—to substantial market discipline. 21 See id. at 14 (acknowledging that “newer proxy advisory firms may face challenges attracting clients and establishing themselves in the industry”). Institutional investors, for the reasons outlined above, may lack sufficient interest in voting to scrutinize advisors’ recommendations carefully. In addition, ISS has, until recently, enjoyed a near-monopoly position and still remains the dominant firm providing voting advice. 22 Id. at 7; Rose, supra note 2, at 899 (“ISS is the dominant firm in the corporate governance industry . . . .”).

The ability of proxy advisors to influence investor voting becomes particularly significant as the importance of shareholder voting increases. With respect to director elections, most U.S. companies have shifted in recent years from plurality to majority voting. 23In 2005, more than 90% of S&P 500 companies employed plurality voting. See, e.g., Brooke A. Masters, Shareholders Flex Muscles; Proxy Measures Pushing Corporate Accountability Gain Support, Wash. Post, June 17, 2006, at D1 (stating that, as of the start of 2005, fewer than thirty S&P 500 companies had majority voting or director resignation policies in place). By 2008, over 80% had moved away from plurality voting. Kahan & Rock, supra note 2, at 23. Under plurality voting, the nominees who win the most votes are elected, regardless of the number of votes that are “withheld.” 24Under plurality voting, a shareholder in an uncontested election may cast a vote in favor of a director candidate or withhold voting authority but may not cast a vote “against” the nominees. See, e.g., Jill E. Fisch, The Transamerica Case, in The Iconic Cases in Corporate Law 46, 68 (Jonathan R. Macey ed., 2008) (explaining the concept of “withheld” votes). Thus, in an uncontested election, a single vote in favor is enough to assure a nominee’s election. By contrast, a majority standard requires a nominee to receive a majority of the votes cast. 25 See id. at 69 (explaining majority voting). Under this standard, shareholders can prevent the election of a nominee even without nominating a competing candidate; the voters simply must cast a sufficient number of “withhold” votes. As a consequence, the shift to a majority standard substantially increases the importance of shareholder voting in uncontested elections.

Over the same time period, a large number of companies dismantled their staggered boards. 26Commentators have described effective staggered boards as the most powerful anti-takeover device and thus the most effective mechanism by which boards can insulate themselves from shareholder voice. See, e.g., Lucian Arye Bebchuk, John C. Coates IV & Guhan Subramanian, The Powerful Antitakeover Force of Staggered Boards: Theory, Evidence, and Policy, 54 Stan. L. Rev. 887, 889–91 (2002). The percentage of S&P 500 companies with staggered boards declined from 55% in 2005 to 40% in 2007. 27Stephen Taub, Revival of Classified Boards? Well, Maybe Not, Compliance Wk., Sept. 11, 2007, http://www.complianceweek.com/article/3647/revival-of-classified-boards-well-maybe-not. In companies with staggered boards, typically only one-third of the board is up for election in any given year. 28 See, e.g., Del. Code Ann. tit. 8, § 141(d) (2009); see also Gregory T. Carrott, The Case for and Against Staggered Boards, Directorship, Sept. 22, 2009, http://www.directorship.com/against-staggered-boards/ (explaining that, most often, staggered boards provide directors with three year terms). With a non-staggered board, the whole board is up for election. Dismantling the staggered board increases the number of directors up for election each year, thereby increasing the opportunity for shareholders to exercise their franchise. Indeed, the move from the typical three-year staggered board to non-staggered, annual elections triples the potential impact of the shareholder vote.

Finally, the New York Stock Exchange (NYSE) has adopted a rule that eliminates discretionary broker voting in uncontested director elections. 29In October 2006, the New York Stock Exchange (NYSE) proposed to amend Rule 452 governing broker votes to redefine all director elections as “non-routine,” which would eliminate the ability of brokers to cast discretionary votes. Proxy Working Group, Report and Recommendations of the Proxy Working Group to the New York Stock Exchange 3 (2006), available at http://www.nyse.com/pdfs/PWG_REPORT.pdf. On July 1, 2009, the SEC finally approved the amendments, effectively ending broker discretionary voting in director elections. Order Approving Proposed Rule Change, as Modified by Amendment No. 4, to Amend NYSE Rule 452 and Corresponding Listed Company Manual Section 402.08 to Eliminate Broker Discretionary Voting for the Election of Directors, Except for Companies Registered Under the Investment Company Act of 1940, and to Codify Two Previously Published Interpretations that Do Not Permit Broker Discretionary Voting for Material Amendments to Investment Advisory Contracts with an Investment Company, Exchange Act Release No. 34-60215, 74 Fed. Reg. 33,293 (July 1, 2009). Historically, brokers who did not receive voting instructions from the beneficial owners of shares in their brokerage accounts were permitted to vote these shares in their discretion. 30NYSE, Inc., Rule 452 (Mar. 6, 2003). Brokers generally exercised their discretion to vote the shares in favor of the slate nominated by the company—the so-called management slate. 31Kahan & Rock, supra note 2. These discretionary broker votes are estimated to amount to about 19% of the votes cast at annual meetings. 32 See Posting of Ted Allen to RiskMetrics Group, http://blog.riskmetrics.com/gov/2007/05/sec-hears-testimony-on-broker-votessubmitted-by-ted-allen-director-of-publications.html (May 25, 2007, 10:58) (attributing this figure to Broadridge Financial). Under the revised NYSE rules, companies will lose a sizeable block of automatic votes in favor of their nominees, shifting power to those shareholders who do vote. 33 See David A. Katz & Laura A. McIntosh, A Seismic Shift in Mechanics of Electing Directors, N.Y. L.J., July 27, 2006, at 5 (“If, in the aftermath of NYSE rule changes as proposed, issuers indeed are unable to contact or obtain voting instructions from large numbers of individual shareholders, the effect will be a massive shift of voting power from brokers to institutions, and, therefore, to proxy advisory services such as ISS, Glass, Lewis & Co., and Proxy Governance.”). The effect of broker voting is illustrated dramatically by the Citigroup 2009 annual meeting in which broker votes comprised 46% of votes cast. 34Citigroup, Inc., First Quarter of 2009 (Form 10-Q), at 156–57 (May 11, 2009), available at http://www.citigroup.com/citi/fin/data/q0901c.pdf?ieNocache=643. (indicating broker votes of 1.732 billion shares). Had the NYSE rule been in effect, two of the Citigroup nominees would not have won re-election.

As the Citigroup annual meeting demonstrates, the number of directors who receive a large percentage of withhold votes has increased. According to Georgeson, Inc., one of the leading proxy solicitation firms, 35For information on Georgeson, see http://www.georgeson.com/. a record 612 directors at S&P 1500 companies received withhold votes in excess of 15% in the 2008 proxy season. 36Georgeson, Annual Corporate Governance Review 7 (2008), available at http://www.georgeson.com/usa/download/acgr/acgr2008.pdf. Thirty directors failed to receive a majority of the votes cast (up from fifteen in 2007). 37 Id. at 7–8. In 2004, twelve directors failed to receive a majority of votes cast. Fisch, supra note 24, at 68. Additionally, the number of contested elections, though still relatively small, continues to increase. For 2008, Georgeson reported an all-time high of fifty-six contested solicitations, following a previous all-time high of forty-six contested solicitations in 2007. 38Georgeson, supra note 36, at 8. In comparison, between 1995 and 1999, the number of contested solicitations averaged twenty-five per year. 39 Id. at 46.

In addition to voting in director elections, shareholders vote on shareholder proposals introduced pursuant to Rule 14a-8 of the Securities Exchange Act. 40Rules and Regulations Under the Securities Exchange Act of 1934: Solicitations of Proxies, 17 C.F.R. § 240.14a-8 (2008). As institutional activism increases, the character of these shareholder proposals has shifted from social policy issues 41 See, e.g., Jill E. Fisch, From Legitimacy to Logic: Reconstructing Proxy Regulation, 46 Vand. L. Rev. 1129, 1152–55 (1993) (describing the rise in shareholder use of social policy proposals in the 1950s and 1960s). to proposals dealing with core economic and governance questions, 42 See, e.g., A. A. Sommer, Jr., Corporate Governance in the Nineties: Managers vs. Institutions, 59 U. Cin. L. Rev. 357, 371 (1990) (describing the shift from proposals “having a social dimension” to those dealing with corporate governance). such as executive compensation, 43 See, e.g., BNA, Annual Meeting Voting Compels More Accountability, 11 Corp. Governance Rep. 30 (2008) (listing “say on pay” executive compensation proposals as one of the top three issues on corporate ballots for 2008). shareholder nomination rights, 44 See, e.g., Am. Fed’n of State, County & Mun. Employees v. Am. Int’l Group, 462 F.3d 121, 123 (2d Cir. 2006) (addressing a shareholder proposal on proxy access). and other corporate governance matters. 45 See Georgeson, supra note 36, at 14 fig.3 (detailing corporate governance proposals from 2004 to 2008). These proposals are receiving increasing attention and support from shareholders. The number of proposals receiving majority shareholder support at S&P 1500 companies has increased from twenty-five in 2001 to eighty-six in 2008. 46Kahan & Rock, supra note 2, at 27 tbl.4 (citing Georgeson, Corporate Governance: Annual Meeting Season Wrap Up (2001), available at http://www.georgeson.com/usa/download/acgr/acgr2001.pdf; Georgeson, supra note 36). Prior to 2001, Georgeson prepared a similar report, but it analyzed only corporate governance proposals made by institutional investors. Georgeson, Corporate Governance: Annual Meeting Season Wrap Up (2000), available at http://www.georgeson.com/usa/download/acgr/acgr2000.pdf. More importantly, boards have become more responsive to proposals receiving majority support. The number of implemented proposals rose from three in 2001 to forty-three in 2008. 47Kahan & Rock, supra note 2. As a result of these increases, shareholder power to introduce proposals is beginning to have a noticeable effect on the governance of U.S. corporations.

Two regulatory initiatives have the potential to increase the significance of shareholder votes even more. Under the first initiative—so-called “proxy access”—shareholders are likely to gain some ability to introduce candidates for the board of directors in a company’s proxy statement. Although shareholders have traditionally been able to mount an election contest by nominating competing candidates, a company is not required to include the challenger’s nominees on the company proxy statement, and the challenge requires an independent (and costly) proxy solicitation. For many years shareholders have sought the power to compel the inclusion of their nominees on the company’s proxy statement. 48Fisch, supra note 37, at 63–67 (reviewing the history of proxy access proposals); Kahan & Rock, supra note 2 (reviewing the history of proxy access proposals). The SEC first considered a proposed rule permitting proxy access in 1942. Fisch, supra note 37, at 63. In 2003, the SEC solicited comments on a complex proposal for proxy access; the proposal stalled due to opposition from corporations and lack of support from Republican commissioners. Id. at 65–66. After several unsuccessful attempts to persuade the SEC to adopt a rule providing for proxy access, institutional investors began to seek proxy access by introducing amendments to individual companies’ bylaws. 49Fisch, supra note 24, at 65–66. Although these efforts were upheld in court, 50Am. Fed’n of State, County & Mun. Employees v. Am. Int’l Group, 462 F.3d 121, 123 (2d Cir. 2006) (holding that shareholders can introduce proxy access proposals under Rule 14a-8). in 2007, the Republican-controlled SEC amended the proxy rules to prohibit shareholders from using SEC Rule 14a-8 to introduce such bylaw amendments. 51Rules and Regulations Under the Securities Exchange Act of 1934: Solicitations of Proxies, 17 C.F.R. § 240.14a-8 (2008).

In 2009, proxy access received a dramatic boost when the Delaware legislature amended its corporation law to authorize proxy access bylaws explicitly. 52 Del. Code Ann. tit. 8, § 112 (2009). The legislature also adopted a provision authorizing bylaws that provide for reimbursement of a shareholder’s proxy solicitation expenses. Del. Code Ann. tit. 8, § 113(a) (2009). Subsequently, the new Democratically-controlled SEC introduced a revised proxy access proposal which, if adopted, would require proxy access under specified conditions. 53Facilitating Shareholder Director Nominations, 17 C.F.R. §§ 200, 232, 240, 249, 274 (2009). If the SEC adopts a proxy access rule, it would mean that for companies with majority voting, shareholders would not only have the power to reject a company’s nominees to the board, but also the power to select nominees of their own choosing.

The second regulatory initiative—“say-on-pay”—enables shareholders to vote on executive compensation. Say-on-pay, which is modeled on a procedure adopted in England in 2002, provides for an annual advisory shareholder vote on the compensation packages paid to top corporate executives. 54Fisch, supra note 37, at 71 (describing say-on-pay). Institutional investors have introduced shareholder proposals seeking say-on-pay at a substantial number of companies. 55 See, e.g., Robert Kropp, Shareowner Resolutions on Say on Pay Gain Widespread Support, May 6, 2009, http://www.socialfunds.com/news/article.cgi?sfArticleId=2690 (reporting that seventy-nine say-on-pay resolutions were introduced in 2008, and more than one hundred have been filed in 2009). Some of these proposals have received majority support, 56According to a preliminary count, as of May 2009, ten of the twenty-nine proposals that came up for a vote received majority support. Press Release, AFSCME, Say on Pay Shareholder Proposals Garner Record Support During Tumultuous Shareholder Season (May 4, 2009). and several companies have already implemented say-on-pay. 57 See, e.g., Editorial, Stockholders Should Demand a Say on Executive Pay, Seattle Times, May 14, 2009, available at http://seattletimes.nwsource.com/html/editorialsopinion/2009221158_editb15sayonpay.html (noting that resolutions had been approved by fifteen companies this year and that Hewlett-Packard and Occidental Petroleum had agreed to adopt say-on-pay without a shareholder vote); SmartPros.com, Say-On-Pay Is on the Way, http://accounting.smartpros.com/x65641/xml (last visited Feb. 7, 2010) (listing Occidental Petroleum, Intel, Hewlett-Packard, MBIA, Motorola, and Ingersoll-Rand as companies that have adopted say-on-pay). Furthermore, Congress may implement some form of say-on-pay requirement, either directly through legislation or indirectly via an SEC rule. 58 See Lawrence Bard et al, Morrison Foerster, Administration Proposals on Compensation Committees and Say on Pay Would Affect All Public Companies, July 30, 2009, http://www.mofo.com/news/updates/files/15793.html (describing the Treasury Department’s draft legislation that would require the SEC to adopt rules mandating say-on-pay for all publicly traded companies). The House approved a say-on-pay bill in 2007, 59Shareholder Vote on Executive Compensation Act, H.R. 1257, 110th Cong. §2 (2007); Shareholder Vote on Executive Compensation Act, S. 1181, 110th Cong. §2 (2007). and President Obama has indicated his support for such legislation. 60Stephen Taub, Obama Pushes Say on Pay Legislation, CFO.com, Apr. 11, 2008, http://www.cfo.com/article.cfm/11037327/c_11036422 (reporting then-Senator Obama’s support for say-on-pay). Similarly, the Emergency Economic Stabilization Act of 2008 61Emergency Economic Stabilization Act of 2008, Pub. L. No. 110343, 122 Stat. 3765, 110th Cong. (2008). required companies receiving financial assistance under the Troubled Asset Relief Program to permit a shareholder advisory vote on executive compensation. 62Shareholder Approval of Executive Compensation of TARP Recipients, Exchange Act Release No. 34-61335, 75 Fed. Reg. 2789 (Jan. 12, 2010) (describing requirement of Section 111(e) of the Emergency Economic Stabilization Act of 2008 and amending federal proxy rules to implement the requirement).

Viewed in the context of the increasing importance of the shareholder franchise, claims about proxy advisor power paint a frightening picture. A few entities with limited accountability and broad discretion control a huge portion of the shareholder vote. And the shareholder vote they control influences an ever-increasing range of issues.

Yet, despite the assertions that proxy advisors are powerful, little systematic study of their actual influence on shareholder votes has been conducted. Only a handful of academic papers analyze ISS recommendations empirically. In one article, Jennifer Bethel and Stuart Gillian 63Bethel & Gillan, supra note 14, at 29. examine votes on shareholder proposals during the 1998 proxy season. Bethel and Gillan conclude that a negative ISS recommendation was associated with 13.6% to 20.6% fewer shares voted in favor of management proposals. 64 Id. at 46. Another recent paper analyzes the role of ISS recommendations in proxy contests. 65Cindy R. Alexander et al., The Role of Advisory Services in Proxy Voting (Nat’l Bureau of Econ. Research, Working Paper No. 15143, 2008), available at http://www.nber.org/paper/w15143. The authors find that ISS recommendations have significant explanatory value for contest outcomes. 66 Id. at 34–35. Finally, Jie Cai, Jacqueline Garner, and Ralph Walking examine the factors that determine the percentage of “for” votes cast in uncontested director elections. 67 See Cai et al., supra note 13. After controlling for several other factors, they find that a negative ISS recommendation reduces the vote in favor of directors by 19%. 68 Id. at 19.

These studies, as well as the other less systematic claims about the effect of proxy advisors, suffer in varying degrees from two problems. First, they focus only on ISS and do not consider the effect of other proxy advisors on shareholder voting. Second, and more importantly, they fail to deal with the issue of what is meant by the “power” or “influence” of proxy advisors. In particular, the studies do not distinguish between correlation and causation. Thus, although they demonstrate that proxy advisor recommendations are correlated with voting outcomes, they do not fully address the underlying factors—firm performance, director attendance, and the like—that are likely to influence both the recommendations and the ultimate vote. 69The article by Cai and others partially examines other factors that may influence the vote. See Cai, et al., supra note 13.

In this Article, we try to correct for these problems in two ways. First, we examine the relationship between shareholder votes and the recommendations of proxy advisors, including not merely ISS, but also Glass Lewis, Proxy Governance, and Egan Jones. (Glass Lewis is reputedly the second most influential proxy advisor; Proxy Governance and Egan Jones also provide proxy advice.) 70 See GAO, Corporate Shareholder, supra note 1, at 7 (describing ISS, Glass Lewis, Proxy Governance, and Egan Jones as among the “five major firms” comprising the proxy advisory industry). The fifth firm included in the GAO report is Marco Consulting Group (MCG), which provides investment consulting services to Taft-Hartley funds and a number of public benefit plans but does not publicly issue voting recommendations. See Marco Consulting, Company History, http://www.marcoconsulting.com/1.2.html (last visited May 5, 2010) (“MCG only provides investment consulting and proxy voting services.”). Second, we try to disentangle the difference between correlation and causation both conceptually and empirically. 71We explore the relationship between these factors and proxy advisor recommendations in a prior article. Stephen J. Choi, Jill E. Fisch & Marcel Kahan, Director Elections and the Role of Proxy Advisors, 82 S. Cal. L. Rev. 649, 650–51 (2009).

Part I discusses the distinction between correlation and causation and posits four possible relationships between proxy advisor recommendations and the subsequent shareholder vote. Part II describes our dataset and provides summary statistics on advisor recommendations and voting outcomes. Part III incorporates factors that, we hypothesize, are likely to influence voting outcomes and, using multivariate regression analysis, analyzes the role these factors and advisor recommendations play in influencing voting outcomes. Part IV focuses on ISS in particular and introduces an alternative methodology for measuring ISS’s power by distinguishing institutional voting behavior from that of individual retail investors.

I. Correlation and Causation: Four Types of “Power”

Proxy advisor recommendations may correlate with the shareholder vote for four conceptually distinct reasons. First, the same director nominee and company characteristics may independently influence both the proxy advisors’ recommendation and the shareholder vote. Second, proxy advisors may gather information that investors use to make their voting decisions. Third, investors may select a proxy advisor based on their ex ante agreement with the bases upon which the advisor formulates its recommendations. Finally, investors may view the advisor’s recommendation alone as a basis for deciding how to vote, independent of the underlying factors upon which that recommendation is based. It is only this last reason that can truly be characterized as causality.

There is reason to believe a substantial overlap exists between the factors that proxy advisors consider important and those that matter to voters. To start, there is extensive corporate governance literature examining board composition and effectiveness. 72 See, e.g., Laura Lin, The Effectiveness of Outside Directors as a Corporate Governance Mechanism:Theories and Evidence, 90 Nw. U. L. Rev. 898, 921–39 (1996) (collecting empirical studies of board composition and effectiveness); cf. Sanjai Bhagat & Bernard Black, The Non-Correlation Between Board Independence and Long-Term Firm Performance, 27 J. Corp. L. 231, 263 (2002) (finding no correlation between director independence and long-term firm performance). Although precise specifications of the characteristics that increase director effectiveness are difficult to identify, many commentators agree on baseline attributes. 73These include director independence both from the company and the CEO, limited service on other corporate boards, and regular attendance at board meetings. See, e.g., In the Matter of the Walt Disney Co., Exchange Act Release No. 50882 (Dec. 20, 2004) (“The independence of directors is a linchpin of sound corporate governance, and is crucial to the objective oversight of management.”); Paul W. MacAvoy & Ira M. Millstein, The Recurrent Crisis in Corporate Governance 22–23 (2003) (stating that directors should act “independently of management”); Stephen P. Ferris et al., Too Busy to Mind the Business? Monitoring by Directors with Multiple Board Appointments, 58 J. Fin. 1087 (2003) (finding no evidence, contrary to popular wisdom, that multiple directors shirk their responsibilities); Renée B. Adams & Daniel Ferreira, Regulatory Pressure and Bank Directors’ Incentives to Attend Board Meetings 304 (European Corporate Governance Inst. Working Paper Series, Working Paper No. 203/2008, 2008), available at http://ssrn.com/abstract=936261 (discussing various directives that directors attend board meetings regularly). The federal proxy rules require issuers to disclose whether any director has attended fewer than 75% of the board meetings held during the prior fiscal year. Schedule 14A. Information Required in Proxy Statement, 17 C.F.R. § 240.14a-101, Item 7(f) (2009). The rules also require disclosure of outside directorships. Schedule 14A. Information Required in Proxy Statement, 17 C.F.R. § 240.14a-101, Item 22(b) (2009). In addition, while shareholders may be dissatisfied with a board of directors for many reasons, common reasons for concern include poor financial performance; corporate misconduct, such as securities fraud; excessive executive compensation; and a lack of responsiveness to shareholders. 74 See, e.g., Mark Anderson, Eli Lilly Heads CalPERS’ ‘Underperforming’ List, Sacramento Bus. J., Mar. 19, 2009, http://sacramento.bizjournals.com/sacramento/stories/2009/03/16/daily56.html (describing CalPERS’s (the California Public Employees’ Retirement System) watch list as targeting companies with corporate governance defects that also “show weakness with profitability, transparency and/or management”).

In an earlier article, we examined the relationship between these factors and proxy advisor recommendations in uncontested director elections. 75 See Choi, Fisch & Kahan, supra note 71, at 650–51. In particular, we examined the effect on recommendations of twenty-three factors, including director-specific factors such as age and attendance, and firm-specific factors such as financial performance, the existence of antitakeover defenses, and the board’s failure to implement a previously approved shareholder proposal. We found that the majority of our factors affected the likelihood that at least one proxy advisor would issue a withhold recommendation—although firm antitakeover defenses did not seem to play a significant role. Moreover, while all of the proxy advisors considered a few specific factors important—such as poor director attendance—on most issues there was substantial variation. For example, ISS was significantly more likely to issue a withhold recommendation when the company board had refused to implement a shareholder resolution that had received majority shareholder support. Glass Lewis was significantly more likely to issue a withhold recommendation if the nominee was an inside director (other than the CEO). Egan Jones was significantly more likely to issue a withhold recommendation if the nominee was a board member at three or more other major companies. Proxy Governance was significantly more likely to issue a withhold recommendation if the company CEO received abnormally high compensation. 76 Id. at 664–70.

We found a substantial correlation between proxy advisor recommendations and the factors that academics, policy makers, and the media have identified as important. This correlation challenges the view that ISS and the other proxy advisors are causally significant in determining the shareholder vote because shareholders may themselves directly consider these factors important. To the extent that the same factors independently affect both shareholders’ voting behavior and the proxy advisor’s recommendation, shareholder votes and recommendations will be correlated. However, the recommendation will not be the cause of the shareholder vote. Any power or influence inferred from such a correlation would be illusory. 77Paul W. Holland, Statistics and Causal Inference, 81 J. Am. Stat. Ass’n 945, 945 (1986).

Of course, proxy advisors may be the source of the information underlying shareholder voting decisions. When proxy advisors issue recommendations, they provide more than a bottom line—more than a mere vote “for” or “withhold.” Proxy advisors also provide additional information about the basis for their recommendation. 78 See, e.g., ISS Governance Services, Proxy Alert, Citigroup Inc. 18 (Apr. 10, 2009) (stating that ISS recommends shareholders withhold their votes for board nominee Anne Mulcahy because she may be overextended as she sits on more than three boards and serves as CEO of Xerox Corporation). For example, a proxy advisor may explain that it issued a withhold recommendation because the director is a member of a board that failed to implement a shareholder resolution adopted with majority shareholder support. Thus, a shareholder who cares about responsiveness to such resolutions, but has neither the time nor the interest to research whether the resolution won majority support and, if so, whether it has been implemented, may obtain that information from the proxy advisor’s report. The relevant underlying information is generally available to the public, but as long as the shareholder is not willing to conduct the requisite research, the proxy advisor’s report is likely to become the exclusive source of information relevant to shareholder voting decisions. Under this circumstance, had the shareholder not subscribed to the services of the advisor, the shareholder would not have learned of the information.

In such a case, the proxy advisor may well be the “but for” cause of the shareholder vote. Nonetheless, it still may be inappropriate to attribute the shareholder’s voting decision to the “power” of the proxy advisor. The advisor is acting as a mere information agent. The underlying information provided by the proxy advisor—not the bottom-line conclusion—is what affects the shareholder vote.

The proxy advisor nonetheless exercises power as an information agent by selecting, in its discretion, which information to report. For example, a proxy advisor could, as a general matter, choose not to provide any information on whether a board failed to implement a shareholder proposal, or it could provide this information selectively. 79For example, the proxy advisor could provide information on the board’s failure to implement a shareholder proposal only when the advisor was recommending a withhold vote and not when the advisor was recommending a vote in favor of the nominees. In theory, proxy advisors could also misreport information. The ability of advisors to exercise power consistently by misreporting is quite limited, however. In addition to the market competition provided by other advisors, the company itself has a strong incentive to correct inaccuracies, and the media is likely to report any substantial errors. Thus proxy advisors have incentives to avoid recommendations that can be described as erroneous. Indeed, ISS received substantial adverse media attention for its recommendation that shareholders withhold their votes from Warren Buffett, a nominee to the Coca-Cola board, because of business relationships between Coca-Cola and some Berkshire Hathaway subsidiaries. See, e.g., Donald E. Graham, The Gray Lady’s Virtue, Wall St. J., Apr. 23, 2007, at A17, available at http://online.wsj.com/article/SB117728391033378436.html (describing ISS’s recommendation as “perhaps the single silliest recommendation ever made to shareholders”). In either case, assuming that shareholders do not otherwise obtain the underlying information, the proxy advisor is exercising some power over the shareholder vote. In sum, to the extent that the information provided by a proxy advisor affects the shareholder vote, the proxy advisor has some limited influence, but inferring from this correlation that the advisor has power over the shareholder vote is an overstatement.

Alternatively, some institutional investors may just look at the bottom line of the proxy advisor and vote accordingly. That is, shareholders may rely on the proxy advisor’s assessment of the underlying information, rather than evaluating that information themselves. 80The proxy voting guidelines of the Nathan Cummings Foundation, for example, reflect this role for the proxy advisor, indicating that the Foundation will vote for a director nominee if the company does not have a staggered board, if the company is not recommending against an issue proposal supported by the Foundation, and if RiskMetrics supports the nominee. Proxy Voting Practices, The Nathan Cummings Foundation, available at http://www.nathancummings.org/shareholders/pvgandvr/VotingGuidelines.pdf. Even in such cases, however, the extent of the proxy advisor’s power may be overstated. At least some investors will have substantial information about proxy advisors’ recommendations and the bases on which they are issued, and they may choose to follow the recommendations of an advisor because they have concluded that they usually agree with the proxy advisor’s decisions. Proxy advisors prepare and distribute annual explanations of their voting policies to their clients, identifying the factors that they consider important. 81 See, e.g., ISS Governance Services, 2008 U.S. Proxy Voting Guidelines Summary (2007), available at http://www.riskmetrics.com/sites/default/files/2008PolicyUSSummaryGuidelines.pdf. Recognizing that different institutions potentially have different objectives (primarily with respect to shareholder resolutions), ISS in fact now offers different guidelines tailored to the needs of union pension funds, public pension funds, and socially responsible institutional investors. 82RiskMetrics, Proxy Advisory Services, http://www.riskmetrics.com/proxy_advisory/options (last visited Aug. 11, 2009) (describing different voting guideline options). For most of these institutional investors, many of which hold securities in hundreds or even thousands of issuers, the most efficient way of deciding how to vote is to determine which proxy advisor has a voting policy they most agree with and then to follow its recommendations.

Anecdotal evidence also suggests that institutions sometimes choose to follow an advisor that has adopted certain voting policies to further their business interests. For example, according to the SEC, INTECH, an investment advisor, switched to ISS’s union fund voting guidelines in an effort to generate more advisory business from union funds. 83 See Thompson Hine, SEC Enters Order Against Adviser Related to Proxy Voting, May 22, 2009, http://www.thompsonhine.com/publications/publication1818.html (describing SEC action); Press Release, U.S. Sec. & Exch. Comm’n, SEC Charges Investment Adviser for Proxy Voting Rule Violations (May 8, 2009), available at http://www.sec.gov/news/press/2009/2009-105.htm. While this may reflect a conflict of interest between INTECH and its clients, 84The SEC noted that the INTECH decision created a potential conflict of interest in that “not all clients would agree with votes made pursuant to the ISS-PVS Guidelines and that voting proxies pursuant to the ISS-PVS Guidelines would benefit INTECH in obtaining and retaining union-affiliated clients.” Thompson Hine, supra note 83. it also indicates that the choice of advisor is correlated with the advisor’s voting policies.

To the extent that an institutional investor chooses a proxy advisor based on its voting policies, the proxy advisor exercises a degree of power, but this power is contingent in two respects. First, the power derives from an ex ante assessment by the advisor’s client that it is in general agreement with the way the proxy advisor makes the recommendations. Second, to the extent that the client ceases to be in agreement—because the client’s view (or its business objective) has changed, because the advisor’s methodology has changed, or because the client believes that there is a different advisor whose recommendations coincide with the client’s views more closely—the client may switch to another proxy advisor.

The degree of contingent power held by a proxy advisor depends on the nature of competition in the market for proxy advisory services. An increase in the number of proxy advisory firms, the extent to which the recommendations of different advisors vary, and the transparency of the bases of these recommendations will each increase the ability of an institution to achieve a closer match between its voting preferences and the recommendations of an advisor. 85 See GAO, Corporate Shareholder, supra note 1, at 13–14 (describing the market for proxy advisory services). To the extent the market for proxy advisory services is sufficiently competitive, market forces will discipline proxy advisors to make recommendations that conform to the preferences of current and potential clients. Indeed, this analysis suggests that those proxy advisors who appear to exercise the most power—i.e., those whose recommendations are followed most often by shareholders—may have this apparent power not because they exercise discretion in making voting recommendations, but rather because they base their recommendations on criteria important to their clients. To the extent this conclusion is correct, the criticism of proxy advisors as being both powerful and unaccountable to shareholders would be substantially muted.

Lastly, some shareholders may not care about how they vote their shares. They may lack the resources, time, or expertise to evaluate voting decisions, or they may engage in an investment strategy in which the outcome of shareholder voting is irrelevant. Although some such investors simply refrain from voting, 86William Baue, Report Urges Foundations to Vote Their Proxies, Social Funds, Mar. 4, 2004, http://www.socialfunds.com/news/article.cgi/1358.html (describing the low level of proxy voting by foundations and suggesting purchasing proxy voting service from ISS as a superior and reasonably priced alternative to refraining from voting). others are legally required to make an informed vote. 87 See Rose, supra note 16, at 897–98 (noting Department of Labor and SEC regulations). Subscribing to a proxy advisor and, in some cases, delegating complete voting authority to that advisor, 88 See Stephen J. Choi & Jill E. Fisch, On Beyond CalPERS: Survey Evidence on the Developing Role of Public Pension Funds in Corporate Governance, 61 Vand. L. Rev. 315, 324 (2008) (reporting that 20% of public pension funds surveyed reported delegating complete voting authority to ISS or a similar organization). may be the most cost effective way of complying with that requirement. 89The SEC has specifically stated that investment advisors can comply with their fiduciary obligations by using a ‘“predetermined voting policy,’ such as a third-party proxy voting service’s platform, to vote proxies provided that the predetermined policy is ‘designed to further the interests of clients rather than the adviser.’” INTECH Inv. Mgmt. LLC, Investment Advisers Act Rel. No. 2872, at 5 (May 7, 2009) (quoting Final Rule: Proxy Voting by Investment Managers, Investment Advisers Act Rel. No. 2106 (Jan. 31, 2003)), available at http://www.sec.gov/litigation/admin/2009/ia-2872.pdf. To the extent that the choice of the proxy advisor is unrelated to the voting recommendations it issues, a proxy advisor may have absolute power. The advisor may base its recommendations on factors that it (or its staff) considers important and would face no short- or long-term pressure to modify these factors because they do not mesh with the interests of its clients. The causal power of proxy advisors to affect a shareholder vote is strongest in this last form of proxy advisor influence. Note, however, that even this absolute power is limited as long as a proxy advisor has other clients who will periodically review its recommendations to determine whether they coincide with their interests, and the advisor issues the same recommendation to both sets of clients.

II. Advisor Recommendations and the Shareholder Vote

We now proceed to examine the power of proxy advisors empirically. Our dataset examines uncontested director elections in 2005 and 2006. We focus only on director elections for companies listed in the S&P 1500 as of June 30 for the year prior to the director election (June 30, 2004 and June 30, 2005, respectively). For each director in our sample of S&P 1500 companies, we collected information about whether the director received a “for” or withhold recommendation (or no recommendation) from ISS, Glass Lewis, Egan Jones, and Proxy Governance. 90Institutional Shareholder Services recommendations were obtained through LEXIS. Glass Lewis, Egan Jones, and Proxy Governance provided us with their recommendations for the period in question. All of the companies in our sample that conducted a director election in 2005 had a proxy advisor recommendation for at least one of their directors.

Table 1, Panel A presents some summary statistics on the coverage rates and recommendations of the four proxy advisors. ISS, Glass Lewis, and Egan Jones provided extensive coverage, issuing recommendations on 88% to 99% of the director nominees in the sample. Proxy Governance, by contrast, provided much more limited coverage—issuing recommendations on only 34% of the director nominees in the sample.

The advisors also differed significantly in their proclivity to issue a withhold recommendation. Institutional Shareholder Services issued such recommendations for only 6.8% of the directors it covered, and Proxy Governance issued even fewer withholds at 3.7% of its recommendations. By contrast, Glass Lewis’s withholds accounted for 18.8% of its recommendations, and for Egan Jones, withholds accounted for 11%. Panel B presents a correlation matrix of the recommendations made by the proxy advisors. The correlation is uniformly positive, but low, indicating that advisors make different decisions about whether to issue a withhold recommendation.

These findings—together with the findings in our prior article that proxy advisors base their recommendations on different factors 91 See Choi, Fisch & Kahan, supra note 71, at 649. —highlight that institutional investors have a real choice in selecting proxy advisors. They can pick among advisors that differ both in how critical they are of board nominees (as demonstrated by the overall rate of their withhold recommendations) and in the criteria they use to assess those nominees. As a result, even institutions that do not want to examine the bases for recommendations on a case-by-case basis can nonetheless choose an advisor, or combination of advisors, to match their preferences.

Table 1, Panels C and D, explore the general correlation between withhold recommendations and the subsequent shareholder vote by providing data on the relationship between the recommendations and the vote outcome. Panel C shows the average percentage of “for” votes 92Percentage “for” vote is defined as the “for” votes as a percentage of the sum of “for” and withhold votes. when a proxy advisor has issued a “for” and a “withhold” recommendation. The last column of that table displays the difference in these percentages as the marginal impact of a withhold recommendation. As Panel C shows, an ISS withhold recommendation is associated with a 20.3% drop in the “for” vote. This drop reflects a far higher percentage than for any of the other advisors. For Glass Lewis, the drop is 6.2%, and for Egan Jones and Proxy Governance, it is 4.7% and 3.5% respectively. The data in this table are consistent with the press characterizations of ISS as the most powerful and Glass Lewis as the second most powerful proxy advisor, 93 See supra text accompanying notes 9–12. and the marginal impact is within the range of votes—albeit at the lower end—that media and prior academic reports have claimed ISS controls. 94 See supra text accompanying notes 12–16.

Note, however, that Panel C measures correlation, not causation. This correlation reflects the combined effect of all of the relationships between the shareholder vote and the ISS recommendation detailed in Part I above. Thus, the 20% effect of an ISS recommendation may be due to a combination of the following: (1) some shareholders conducting an independent analysis and voting the way that ISS recommends without considering (or even knowing about) the ISS recommendation; (2) some shareholders learning information from ISS that affected their own assessment as well as the ISS recommendation; (3) some shareholders following ISS based on their general assessment of ISS’s voting policies; 95These shareholders may switch if they find that ISS’s voting policies in fact do not match their preferences. (who may switch if they find that ISS’s voting policies in fact do not match their preferences); and (4) some shareholders following ISS recommendations without regard to (or without having) their own views on the issues.

In Panel D, we consider the combined effect of recommendations by multiple proxy advisors. For ISS, the marginal impact of a recommendation is pretty stable, regardless of what the other proxy advisors do—ranging from 17.6% to 21.4% depending upon whether another advisor (and which advisor) has issued a “for” or a withhold recommendation. By contrast, the impact of the other advisors seems to decline when the ISS recommendation is taken into account. Thus, holding the ISS recommendation constant, withhold recommendations by Egan Jones and Proxy Governance affect less than 2% of the vote. A withhold recommendation by Glass Lewis retains its earlier effect (6.2% in Panel C compared to 6.5% in Panel D) when ISS also issued a withhold recommendation. But the marginal impact of a Glass Lewis withhold recommendation is only 3.6% of the vote when ISS issues a “for” recommendation. In sum, when we combine the recommendations, the ISS effect clearly dominates those of the other advisors. Although not conclusive, these data suggest either that ISS’s recommendations are more closely aligned with shareholders’ preferences, that other proxy advisors are far less influential than ISS, or both.

Table 1, Panel E presents data on the distribution of shareholder votes. In 2005 and 2006, most directors were elected with a very high vote margin—an unsurprising outcome given that we examined uncontested elections during an economic bubble in a period when broker discretionary voting was permitted. 96 See infra text accompanying note 135. Since 2006, the number of directors with high withhold votes has increased. See Georgeson, supra note 36, at 7 (describing the increase in the number of high withhold votes). For 72% of the nominees, the margin is 95% or more of the vote, and for 89% of the nominees, the margin exceeds 90%. Only 4% of the nominees received a “for” vote of less than 80%. It is important to remember that, in uncontested elections, shareholders make a significant statement simply by withholding a higher percentage of votes than normal. Thus, given that the average “for” vote is 95%, a “for” vote of 80% could be considered a rebuff or an embarrassment to a director. 97Georgeson, supra note 36, at 7 (charting the number of directors who received a withhold vote of 20% or more). Indeed, issuers have become increasingly responsive to substantial (but less than majority) withhold votes, even though such votes have no direct impact on the composition of the board. 98 See Cai et al., supra note 13, at 2390; Diane Del Guercio et al., Do Boards Pay Attention When Institutional Investor Activists “Just Vote No”?, 90 J. Fin. Econ. 84 (2008) (finding operating performance improvement and increased CEO turnover in response to successful “vote no” campaigns).

III. Independent Factors Affecting the Shareholder Vote

We now probe further into the effect of underlying firm and director factors and advisor recommendations on vote outcomes. As in our prior research, we collected information about various publicly available factors that, based on corporate governance literature, we posit may influence the shareholder vote. We obtained data regarding the characteristics of both individual director nominees and the company for which the director was being nominated. With respect to individual directors, we obtained data 99These data were obtained from the RiskMetrics-Investor Responsibility Research Center (IRRC) director database, available to subscribers of Wharton Research Data Services. The IRRC dataset consists of data on individual board directors from 1996 to 2006. The data include “a range of variables related to individual board directors (e.g., name, age, tenure, gender, committee memberships, independence classification, primary employer and title, number of other public company boards serving on, shares owned, etc.).” See RiskMetrics-Directors Legacy Data Request, Wharton Research Data Services (on file with authors). on: (1) whether the director was the CEO (CEO), a non-executive chairman (Chairman Only), an employee of the company other than the CEO (Empl_Dir), an outside director with certain links to the company (OutDirLink), or a new Director (New Director); (2) whether the director was a member of the audit committee (AuditMbr), the compensation committee (CompMbr), or the nominating committee (NomMbr); and (3) whether the director was a member of at least three other “major” company boards during the year prior to the annual meeting date (ManyBds), 100We use the IRRC data on other “major” company boards held by directors for the year prior to the annual meeting. whether the director attended less than 75% of the director meetings (Attendance), whether the director held at least 20% of the company’s stock (BlockDir), whether the director was an interlocking director (Interlock), and whether the director was 75 years or older (Age75).

For each company in our sample, we obtained data 101These data were obtained from SEC filings, press releases, the IRRC Governance database, the Georgeson Annual Corporate Governance Reviews, and the Center for Research in Security Prices (CRSP). All of the data are publicly available or based on publicly available sources. on (1) whether the first public report of a restatement to the company’s financial statement occurred within two years prior to the annual meeting (Prior Restat), whether the first public statement of an SEC investigation or enforcement action occurred within two years prior to the annual meeting (Prior SEC), and whether the company rejected an issue proposal that had received majority shareholder support in the last year (IP No); (2) whether the company had a classified board (ClassBd), a poison pill (PPill), cumulative voting (CumVote), or golden parachutes (GP); (3) whether the company was in the top or bottom 5% of the companies ranked based on the abnormal holding period return for the three-year period prior to the meeting date for the year of the recommendation (Top5AbRet, Bot5Abret); 102The abnormal return is defined as the difference between the raw three-year holding period return for the company in question and the three-year holding period return for the CRSP value weighted market index. and (4) whether the CEO for the company was in the top 5% for total excess compensation (Top5AbComp). 103Top5AbComp is an indicator variable defined as equal to 1 if the total excess compensation for the CEO for the company in question is in the top 5% of the sample and 0 otherwise. We define total excess CEO compensation as the difference between the total CEO compensation for the year prior to the meeting date (as provided by the Compustat Executive Compensation database) minus the expected total CEO compensation. We calculate the expected total CEO compensation by (1) estimating an Ordinary Least Squares (OLS) model for: ln(Total CEO compensation) = α + β1ln(Market_Capitalization) +β2One_Year_Abnormal_Holding_Period_Return + β3Year_2006 + ε and (2) using the predicted Total CEO compensation based on this model as the expected Total CEO compensation.

We hypothesize that all factors other than new director, CEO, non-executive chairman, and top 5% abnormal return are associated with a decline in “for” votes for a particular director. As most shareholders typically vote for a company’s nominees in an uncontested election, 104 See, e.g., Georgeson, supra note 36, at 8 (reporting that, in 2008, only thirty directors at S&P 1500 companies failed to receive a majority of “for” votes, compared to fifteen directors in 2007). it is likely that withhold votes are triggered by specific problems with a particular director or the company as a whole. Directors who may not perform their duties as effectively as other directors (due to low attendance, posts on multiple boards, or old age) may receive a greater proportion of withhold votes. Shareholders may look negatively on directors who lack independence or have conflicts of interest (including employee directors other than the CEO, outside directors with linked affiliations with the company, directors with substantial block shareholdings, and directors that have interlocking board relationships with the company). Company problems such as poor performance, a restatement, or an SEC investigation may trigger a withhold vote, as may a lack of board responsiveness to investors, indicated by the failure to adopt a shareholder-approved issue proposal. Shareholders, particularly institutional investors, may also view the presence of antitakeover mechanisms as a lack of board responsiveness.

We also hypothesize that shareholders tailor their voting to hold directors who sit on key committees more responsible for certain problems. Thus shareholders may be more likely to hold members of audit committees responsible for audit-related problems, or they may be more likely to withhold votes from members of the compensation committee if a company overpays its CEO.

We view shareholder voting for a CEO-director as categorically different. A significant withhold vote on the CEO may both send a strong signal of dissatisfaction (because the CEO, in many ways, personifies the current management policy of the company), but it may also entail greater costs (leading to the CEO’s resignation, possibly without a successor in place). We hypothesize a decreased likelihood of a withhold recommendation for new directors because shareholders are not likely to hold them responsible for prior problems. We also hypothesize a decreased likelihood of a withhold recommendation for non-executive chairmen because they are likely both to reflect company responsiveness to shareholder demands and to be selected for factors valued by shareholders such as expertise and independence. For obvious reasons, we similarly hypothesize a decreased likelihood of withhold recommendations for directors of companies that rank in the top 5% of abnormal return.

In Table 2, we present some summary statistics about the distribution of these variables as well as a univariate analysis of the relationship between these variables and the voting outcome. For the variables for which we had a prediction, all but five yield a statistically significant difference in the vote outcome in the predicted direction. The remaining five variables (audit committee membership, chairman only, cumulative vote, golden parachute, and top abnormal returns) do not yield a statistically significant difference.

The results of the univariate analysis, however, should be viewed with care. This is especially true for the variables associated with board or employment status (CEO; membership on the audit, compensation, or nominating committee; chairman only; employee director; and outside-linked director) because these variables are negatively correlated with each other. For example, a CEO cannot also be a chairman-only, an employee director, or an outside-linked director. As the key committees tend to consist only of independent directors, a CEO or an employee director will generally not be on the audit, compensation, or nominating committee. 105 See, e.g., NYSE, Inc., Listed Company Manual §§ 303A.03, 303A.07(b) (2009); NASDAQ, Inc., Stock Market Equity Rules § 5605 (Mar. 12, 2009) (describing the composition of the audit committee (5605(c)(2)), independent director executive compensation (5605(d)), and independent director oversight of director nominations (5605(d))), available at http://nasdaq.cchwallstreet.com/NASDAQTools/PlatformViewer.asp?selectednode=chp_1_1_4_2&manual=%2Fnasdaq%2Fmain%2Fnasdaq-equityrules%2F. Additionally, given some notion of fair distribution of work among outside directors, a director generally does not serve on more than one of these committees at the same time.

We also hypothesize interactions between these variables. Specifically, we hypothesize: (1) that the presence of audit and disclosure-related problems (prior audits or restatements) may have a particularly strong adverse impact on members of the audit committee; 106Studies have demonstrated relationships between audit committee composition and audit-related problems. See, e.g., Bradley Pomeroy & Daniel B. Thornton, Meta-Analysis and the Accounting Literature: The Case of Audit Committee Independence and Financial Reporting Quality, 17 Eur. Acct. Rev. 305, 310–11 (2008) (summarizing twenty-seven empirical studies examining the relationship between audit committee independence and financial reporting quality); Joseph V. Carcello et al., Audit Committee Financial Expertise, Competing Corporate Governance Mechanisms, and Earnings Management (Working paper, 2006), available at http://ssrn.com/abstract=887512 (finding that “independent audit committee members with financial expertise are most effective in mitigating earnings management”). (2) that the presence of compensation-related problems (abnormally high compensation) may have a particularly strong adverse impact on members of the compensation committee; 107 See, e.g., Ronald C. Anderson & John M. Bizjak, An Empirical Examination of the Role of the CEO and the Compensation Committee in Structuring Executive Pay, 27 J. Banking & Fin. 1323, 1332–36 (2003) (discussing compensation committee independence and CEO presence on the compensation committee as mechanisms for dealing with potential agency issues in setting CEO pay). (3) that an abnormal positive or negative return may have a particularly strong impact on the CEO; 108 See, e.g., Michael S. Weisbach, Outside Directors and CEO Turnover, 20 J. Fin. Econ. 431, 453–54 (1988) (finding that firms with independent boards are more likely to remove the CEO on the basis of poor stock performance). and (4) that membership on many boards may have a different impact on the CEO than on other board members. (This impact could be more positive because it serves as a signal of quality or more negative because of concerns that the CEO is spending excess time on non-company business.) Table 2.1 reports that the “for” vote outcome correlates significantly with three of these interaction terms (Prior Restat x AuditMbr, Prior SEC x AuditMbr, and Top5AbComp x CompMbr).

We will refer to variables and interacted variables other than the vote recommendations as “underlying factors.” Our prior research demonstrates that most of these variables (other than those related to takeovers) 109We regard ClassBd, PPill, CumVote, and GP as takeover-related factors. are significantly related to a withhold recommendation by at least one proxy advisor. 110 See Choi, Fisch & Kahan, supra note 71, at 649. As at least some proxy advisors base their recommendations on these variables, it is plausible that shareholders may give independent weight to these factors in determining their votes—either because they have independent information about these underlying factors or because they obtain this information through the proxy advisor’s analysis. Finally, even though takeover-related factors do not appear to affect the recommendations of proxy advisors, we nevertheless include them in our analysis because these factors are often identified as important indicators of governance quality, 111 See Paul Gompers et al., Corporate Governance and Equity Prices, 118 Q.J. Econ. 107 (2003). may affect firm value, 112 See id. (finding a relationship between equity prices and various corporate governance variables). and are within the control of the board. 113Boards generally can adopt poison pills and golden parachutes without shareholder approval. See, e.g., Moran v. Household Int’l, Inc., 500 A.2d 1346, 1351 (Del. 1985) (holding that the board of directors has the power to adopt a poison pill under Delaware law); see also Unocal Corp. v. Mesa Petroleum Co., 493 A.2d 946, 953–54 (Del. 1985) (outlining a board’s broad powers to act unilaterally). In contrast, addition or removal of a classified board (otherwise known as a staggered board) or a cumulative voting structure typically requires both board and shareholder approval. See, e.g., Lucian Arye Bebchuk et al., The Powerful Antitakeover Force of Staggered Boards: Theory, Evidence, and Policy, 54 Stan. L. Rev. 887, 894 (2002) (“[D]ismantling [a staggered board] that is in the charter requires both a shareholder vote and a board vote . . . .”); Jeffrey N. Gordon, Institutions as Relational Investors: A New Look at Cumulative Voting, 94 Colum. L. Rev. 124, 161 (1994) (“[T]he elimination of cumulative voting in a specific firm ordinarily requires shareholder approval . . . .”).

We next examine (in Table 3) the relationship between the “for” vote outcome and our identified, publicly available underlying factors in a multivariate model. We first estimate a regression with a log odds transformation of the “for” vote outcome as a dependent variable. 114While the vote outcome for any director election is continuous, the vote outcome is bounded by zero and one. Estimating an ordinary least squares model on a bounded dependent variable results in biased coefficients. We employ a log odds transformation of the vote outcome to generate an unbounded, continuous variable allowing for ordinary least squares estimation of the relationship between the vote outcome and our independent variables of interest. We compute the log odds of the vote outcome as follows: For the dependent variable for a particular proxy advisor, we compute a term X = 0.5/n, where n is the number of data points where “for” vote data exists for directors with a recommendation from the particular proxy advisor in question. We then use the log((VOTE + X)/(1-VOTE + X)) as the dependent variable (to avoid division by zero problems when the “for” vote fraction is equal to 1).To control for the possibility that errors for directors in the same company may be correlated we use standard errors clustered by company in the models of Table 3. Unreported, we re-estimate the base model (Model 1 of Table 3) using non-clustered, robust standard errors and obtain similar qualitative results, except that both PPill and CumVote are now significant (at the <1% and 10% levels respectively), and Bot5AbRet x CEO is insignificant. For independent variables we use publicly available factors with the following additions: We add interaction variables for Prior Restat x AuditMbr, Prior SEC x AuditMbr, Top5AbComp x CompMbr, Bot5AbRet x CEO, Top5AbRet x CEO, and ManyBds x CEO. As further controls, we add variables for the percentage of shares held by institutional investors (InstHold); the percentage of the vote held by all board members (Tot_Dir_Shs); firm size (lmktcap, the log of the firm’s market capitalization); risk (sdret, the standard deviation of the company’s returns for the one-year period prior to the annual meeting date). We also add a dummy variable for whether the election took place in 2005 or 2006 (Year06).

For the base model (reported as Model 1 in Table 3), we do not include any proxy advisor recommendations. In the base model, virtually every underlying factor significantly affects the shareholder vote, either on its own or as part of an interaction variable. As predicted, the following are associated with a reduced “for” vote: membership on audit, compensation, or nominating committees; status as outside-linked or employee director; poor attendance; age 75 years or older; a prior SEC investigation; a prior restatement (for audit committee members only); payment of abnormally high compensation (for compensation committee members only); membership on many boards (for non-CEOs); ignoring a shareholder proposal; and abnormally low stock returns. Status as a new director and abnormally high stock returns are each associated with an increased “for” vote. In addition, we find that CEOs get a lower percentage of “for” votes than other directors. With regard to the takeover-related factors, only the presence of a classified board is associated with a significant decrease in the “for” vote. No underlying factor is significant in the opposite of the predicted direction.

As with the analysis in Part II, however, the associations in the base regressions between the underlying factor variables and the shareholder vote reflect correlation, not causation. Significantly, the impact of these variables may be mediated because they affect proxy advisor recommendations rather than directly influencing shareholder votes. We address this issue in Model 2 by adding an indicator variable for the ISS recommendation (VoteISS): assigning a value of 1 if ISS issued a withhold recommendation and 0 if ISS issued a “for” recommendation. In Models 3, 4, and 5, we do the same for recommendations by Glass Lewis, Egan Jones, and Proxy Governance.

Having both the ISS recommendation and the underlying factors in the same regression permits us, to some extent, to separate the effects of the two types of variables on the election results. To the extent that the underlying factors affect the vote outcome independently of the ISS recommendation—either because voters pay direct attention to these factors or because voters pay attention to other proxy advisors who pay attention to these factors—the effect should persist even after controlling for the ISS recommendation. Indeed, when we add the variable for the ISS recommendation, there is almost no qualitative distinction between Models 1 and 2 in the significance levels of the underlying factors. All variables retain their statistical significance with the exception that one variable (Interlock) that was not significant in the base model is now significant at the 10% level (in the predicted direction) in Model 2. The levels of significance change from Model 1 to Model 2 for only Prior Restat x AuditMbr (which decreases from a 5% to a 10% level) and Bot5AbRet x CEO (which increases from a 10% to a 5% level). To test the continuing importance of the underlying factors even with the ISS recommendation, we perform an F-test of the joint hypothesis that no independent variable except the ISS recommendation variable is significantly different from zero. The p-value of the F-test is 0.0000, which indicates that the other independent variables add significance to the explanatory power of the “for” vote ordinary least squares model.

The variable for the ISS recommendation in Model 2 is also highly significant. This provides compelling evidence that the ISS recommendation has independent significance—that vote outcome is not driven exclusively by the underlying factors included in our regression. In addition, the magnitude of the ISS recommendation variable is higher than any other single variable, and inclusion of the ISS recommendation greatly increases the predictive power of the regression (the adjusted R-squared increases from .109 to .185). In short, ISS’s recommendation matters.

A. The Impact of an ISS Recommendation

Finding that the ISS recommendation matters leads to the next question: How much does it matter? To get a better sense of the quantitative impact of the ISS withhold recommendation on the “for” vote percentage, we calculate the predicted change in the “for” vote outcome—depending on whether ISS makes a “for” or withhold recommendation. We find that this effect varies depending on the overall level of the vote in favor of the director candidate. Thus Table 4 reports the effect at various points along the log-odds “for” vote distribution. For example, calculations show that the ISS withhold recommendation reduces the predicted “for” vote by 13.1% (from 98.1% to 85.4%) at the fiftieth percentile of the log-odds “for” vote distribution. 115These percentiles were calculated using the actual distribution of all independent variables except the variable for the ISS vote recommendation, which was set to zero (the baseline “for” recommendation). At the twenty-fifth percentile, an ISS withhold recommendation has a stronger impact, reducing the predicted “for” vote by 17.0%. On the other hand, at the seventy-fifth percentile, an ISS withhold recommendation has a weaker impact, reducing the predicted “for” vote by 10.1%.

The quantitative impact of the ISS variable reflected in Table 4 likely overstates the actual impact of the ISS recommendation. One of the challenges of the multivariate regression models in Table 3 (used to compute the marginal impacts reported in Table 4) is that they are incomplete. Although we have endeavored to identify many of the publicly available factors that may influence the shareholder vote, it is likely that we have failed to identify and control for all such factors. 116Research indicates, for example, that shareholders affiliated with the AFL-CIO may consider the interests of union workers when voting in director elections. See, e.g., Ashwini K. Agrawal, Corporate Governance Objectives of Labor Union Shareholders: Evidence from Proxy Voting 30 (N.Y.U. Stern Working Paper Series, No. FIN-08-006, 2009), available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1285084 (finding evidence that labor relations affect the voting behavior of some union shareholders). Considerations of corporate social responsibility may influence other shareholders. See Thomas W. Joo, Corporate Hierarchy and Racial Justice, 79 St. John’s L. Rev. 955, 956–57 (2005) (describing the potential role of shareholder power in increasing racial justice and social responsibility). Shareholders may care about the board’s position on current as well as previously submitted shareholder proposals. See Nathan Cummings Found., supra note 80, at 1 (indicating that the Foundation will vote for company nominees if, inter alia, “[t]he board does not recommend a vote AGAINST a shareholder proposal that the Foundation supports”). This reflects the standard omitted variable problem in regression analyses. As long as we do not control for these factors, the ISS variable will include both the direct effect of the variable and the effect of these omitted factors—thus potentially overstating the importance of the ISS recommendation in explaining the “for” vote outcome. The coefficient estimates for the VoteISS dummy variable represent the upper bound of any direct effect of the ISS recommendation, but the true effect of the ISS recommendation may be lower, even much lower.

The extent to which our model overstates the significance of the ISS recommendation depends on how many underlying factors we have omitted from the regression, the importance of these factors, and their correlation with the ISS recommendation. In this regard, it is important to keep in mind that some of the variables we have identified and used in the regressions are imprecise proxies for an actual problem with a director or company. This is true specifically for the variables for CEO status and for membership on the audit, compensation, or nominating committee. For example, shareholders are not automatically going to vote against a director because the director is a member of the compensation committee (nor is ISS more likely to recommend a withhold vote against such a director because of his or her committee membership). Rather, membership on the audit, compensation, or nominating committee may result in a withhold vote or withhold recommendation because voters or ISS hold the committee responsible for problems under its purview. In our regression, we control for only a few potential problems: high CEO compensation for compensation committee members; restatements and SEC investigations for audit committee members; and performance and membership on other boards for CEOs.

In addition to being underinclusive, our proxies are overinclusive—not every restatement reflects adversely on the current audit committee. More generally, given the nature of our empirical analysis and the size of our data set, we can include only the factors that are easily available, quantifiable, and generalizable across a large number of firms and directors. Neither proxy advisors (which have a sizeable full-time staff) nor shareholders are confined in this manner. We thus expect that our regressions fail to include a large number of important underlying factors that presumably also affect the ISS recommendation.

B. Contingent Versus Absolute Power

We did not find any variable that was both (i) associated with an increased likelihood of an ISS withhold recommendation as reported in our earlier article 117Choi, Fisch & Kahan, supra note 71, at 665. and (ii) associated with a reduced “for” vote in the regressions reported in Table 3. Similarly, we did not find any variable that was both (i) associated with a reduced likelihood of an ISS withhold recommendation and (ii) associated with an increased “for” vote in the regressions reported in Table 3. This is true whether or not we control for the ISS recommendation. Thus, we have not identified any factor that ISS views as negative but shareholders view as positive (or vice versa). This, incidentally, is not true for Glass Lewis. In our earlier article, we found that Glass Lewis is less likely to issue withhold recommendations for CEOs, 118 Id. at 695. but here we find that CEOs receive a significantly higher withhold vote from shareholders than do non-CEOs for all the models reported in Table 3.

Moreover, most of the factors that we identified in our earlier article as having a statistically significant impact on the ISS recommendation 119 Id. at 665. remain significant in explaining the voting outcome even after controlling for the ISS recommendation in Model 2 of Table 3. Specifically, the following factors are associated with a lower likelihood of a “for” recommendation by ISS and, after controlling for the ISS recommendation, with a lower “for” shareholder vote percentage: CEO status, membership on the compensation committee, abnormal compensation (for compensation committee members), lack of attendance, membership on multiple boards (for non-CEOs), membership on the nominating committee, status as an employee or outside-linked director, ignoring a shareholder proposal, and having a classified board. New director status is associated with a higher likelihood of a “for” recommendation by ISS (as identified in our earlier article) and, after controlling for the ISS recommendation, a higher “for” vote percentage in Model 2 of Table 3. 120Only two factors that were significant for the ISS recommendation—status as non-executive chairman and golden parachutes, both of which reduced the likelihood of a withhold recommendation—are not significant in the vote regressions. Compare id. at 665, with id. at 671–72. Furthermore, the four most important factors affecting an ISS recommendation—ignoring a shareholder proposal, poor attendance at board meetings, status as outside-linked director, and status as employee director 121 Id. at 671–72. —were also among the six most important factors (together with the ISS recommendation itself and status as CEO) affecting the vote outcome. Institutional Shareholder Services is the only advisor for which CEO status is associated with an increased likelihood of a withhold recommendation as reported in our earlier article, and it is associated with a reduction in the “for” vote percentage in Model 2 of Table 3. The relationship between the factors that independently affect the ISS recommendations and the vote outcome suggests that ISS is in sync with the sentiments of shareholders. In essence, ISS generally gives the same directional weight to company and director attributes in making its voting recommendation as do shareholders in making their voting decisions. 122This alignment is unlikely to be coincidental. Institutional Shareholder Services explicitly seeks shareholder input in formulating its voting policies, surveying institutional investors on a yearly basis. See RiskMetrics Group, 2009–2010 RiskMetrics Policy Survey 4 (2009) (describing how feedback from both institutional investors and issuers is part of RiskMetrics’s annual policy-formulation process).

The results for the separate regressions involving the other proxy advisors (reported in Table 3 as Models 3, 4, and 5) follow the same basic pattern as the results for ISS. Specifically, for each advisor, the dummy variable for a withhold recommendation is negative and significant, and most attribute variables that were significant in the base regression remain so. The marginal effect of a withhold recommendation by the advisors, calculated in the same manner as discussed above for ISS, is reported in Table 4. In each case, the upper-bound estimate of the direct effect is significantly smaller than the respective estimate for ISS. Glass Lewis has a larger upper-bound effect, and the estimates for Proxy Governance and Egan Jones are similar to each other. 123As a robustness test, we re-estimated the “for” vote outcomes for Proxy Governance (PGI), Glass Lewis (GL), and Egan Jones (EJ) recommendations using a Tobit model. Unreported, the coefficients on VotePGI (-0.025), VoteGL (-0.052), and VoteEJ (-0.037) are all significantly different from zero. Note that the upper bounds of influence for PGI (2.5 percentage points), GL (5.2 percentage points), and EJ (3.7 percentage points) are again smaller than for ISS. We also re-estimated the “for” vote outcomes for PGI, GL, and EJ recommendations using an OLS model with the untransformed “for” vote outcome as the dependent variable. Unreported, the coefficients on VotePGI (-0.023), VoteGL (-0.050), and VoteEJ (-0.035) are all significantly different from zero. Note that the upper bounds of influence for PGI (2.3 percentage points), GL (5.0 percentage points), and EJ (3.5 percentage points) are again smaller than for ISS.

The results further suggest that these advisors are less in sync with shareholders than ISS. For example, the four most important factors affecting the recommendations of Egan Jones and Proxy Governance do not correspond closely to the factors affecting the shareholder vote. 124The most important factors affecting Egan Jones’s recommendation are attendance, membership on multiple boards, outside-linked status, and membership on the nominating committee. The most important factors for Proxy Governance’s recommendation are attendance, high compensation, membership on the compensation committee, and age. Choi, Fisch & Kahan, supra note 71, at 671–72. Although ignoring a shareholder proposal is an important factor in explaining the shareholder vote in all the regressions of Table 3, it is not a significant factor in explaining recommendations by Egan Jones and Proxy Governance. 125 Id. at 672. As to Glass Lewis and ISS, the regressions of Table 3 show a significant overlap in the most important factors affecting the recommendation. 126The four most important factors affecting the ISS recommendation were also among the five most important factors accounting for a withhold recommendation for Glass Lewis. Id. at 671–72. The fifth Glass Lewis factor significantly associated with a withhold recommendation, board interlock, was present only in 25% of the sample. But Glass Lewis gives strong positive weight to CEO status (i.e., CEOs are less likely to receive a withhold recommendation), 127 Id. at 671. whereas CEO status is associated with a lower “for” vote.

These findings have two implications: First, they suggest that the effect of an ISS recommendation, as reflected in our measurements, may include a fair degree of contingent power. Our results indicate that shareholders are basing their votes on considerations similar to those that ISS uses in making its recommendations, whether shareholders are following ISS’s recommendation or not. Most of the factors we identified as affecting the ISS recommendation also independently affect the shareholder vote, and both ISS and shareholders consider the same factors as most important. This allows us to infer that many institutions that follow ISS’s recommendations do so because they generally agree with the basis for ISS’s voting recommendations. Second, the findings suggest that ISS’s market position, and to a lesser extent Glass Lewis’s market position, may be due, at least in part, to the fact that their recommendations reflect client views better than those of the other proxy advisors. While catering to clients’ views may explain ISS’s market dominance, it also suggests the limits of such dominance—if ISS were to shift its recommendations away from the views of its clients, it would likely lose those clients to competing advisory firms.

C. ISS and Glass Lewis

The data in Table 1, Panels A and B, suggest that the marginal impact of Glass Lewis’s recommendations may be affected by the recommendation made by ISS. In Table 5, Panel A, we start with the base model (reported as Model 1 in Table 3) and include separate indicator variables for the recommendations of ISS and Glass Lewis as well as an interacted indicator variable taking the value of 1 if both ISS and Glass Lewis issued a withhold recommendation. In this regression, the indicator variable for the ISS recommendation (VoteISS) measures the impact of an ISS withhold recommendation on the “for” vote outcome when Glass Lewis has issued a “for” recommendation. Similarly, the indicator variable for Glass Lewis (VoteGL) measures the impact of a Glass Lewis withhold recommendation on the “for” vote outcome when ISS has issued a “for” recommendation. The sum of the indicator variable for ISS (VoteISS) and the interacted indicator variable (VoteISS x VoteGL) measures the impact of an ISS withhold recommendation when Glass Lewis has also issued a withhold recommendation. The sum of the indicator variable for Glass Lewis (VoteGL) plus the interacted indicator variable (VoteISS x VoteGL) measures the impact of a Glass Lewis withhold recommendation when ISS has also issued a withhold recommendation.

In Table 5, Panel A, the variables for both the ISS and Glass Lewis recommendations are negative and significant, indicating that a withhold recommendation by either advisor reduces the “for” vote percentage. At the median of the log-odds “for” vote distribution, assuming that Glass Lewis has issued a “for” recommendation, the predicted change in the “for” vote outcome is -14.5 percentage points when ISS issues a withhold recommendation. In contrast, assuming ISS has issued a “for” recommendation, the predicted change in the “for” vote outcome is -3.1 percentage points when Glass Lewis issues a withhold recommendation. If ISS issues a withhold recommendation, the predicted marginal effect of Glass Lewis also issuing a withhold recommendation (the sum of VoteGL and VoteISS x VoteGL) is insignificant. However, if Glass Lewis issues a withhold recommendation, the predicted marginal effect of ISS also issuing a withhold recommendation (the sum of VoteISS and VoteISS x VoteGL) is negative and significant; the predicted change in the “for” vote outcome (measured at the mean level of the other control variables) is -13.2 percentage points. 128 See infra tbl.5, Panel B.

These results are consistent with those in the univariate analysis (Table 2) and suggest that a Glass Lewis withhold recommendation has a greater impact on the vote if ISS has issued a “for” recommendation than if ISS has issued a withhold recommendation. This suggests the possibility that some institutional investors automatically will vote in favor of the board’s nominees if both ISS and Glass Lewis issue “for” recommendations, but not if one of them issues a withhold recommendation. Alternatively, it may indicate that there are some underlying factors that both Glass Lewis and shareholders (but not ISS) consider relevant when voting. The recommendations by the other two proxy advisors have only a small, if any, effect on the vote outcome.

IV. Institutional Versus Individual Test

Proxy advisors provide recommendations and supporting research to their subscribers, which include mutual funds, pension funds, foundations, and other institutional investors. 129 GAO, Corporate Shareholder, supra note 1, at 9–10. Individual shareholders generally do not employ the services of these advisors, and advisors typically do not provide public access to their recommendations and underlying research. 130 See Alexander, et al., supra note 65, at 8 (“The core business of ISS and other proxy advisors is to supply institutional investors with vote recommendations on a subscription basis.”). Institutional Shareholder Services’s recommendations and reports are now available on LEXIS and are also available on Westlaw through a premium subscription. In some high profile elections such as those involving a proxy contest 131 See, e.g., Press Release, Starboard Value and Opportunity Master Fund Ltd. and Ramius Capital Group LLC, Institutional Shareholder Services (ISS), A Leading Independent Proxy Advisory Firm, Supports Ramius’ Independent Nominees for Election to the A. Schulman Board of Directors (Jan. 7, 2008), available at http://www.euroinvestor.co.uk/news/story.aspx?id=9692387&bw=20080107005892 (reporting that ISS and Glass Lewis supported the appointment of dissident nominees to the A. Schulman Board of Directors). Media reports typically are a response to a press release, and press releases are most common in contested elections. or merger, 132 See, e.g., Press Release, Arris, ISS and Glass Lewis Each Recommend Merger of ARRIS and C-COR (Dec. 6, 2007), available at http://www.arrisi.com/press_events/press_releases/pressdetail.asp?id=389 (reporting ISS and Glass Lewis recommendations in favor of proposed merger). interested parties may issue a press release disclosing a proxy advisor’s recommendation. It is thus likely that recommendations directly affect only the vote of institutional investors (some of which are clients of these advisors), not the vote of individual investors (who are not clients).

We therefore construct a test designed to measure the power of ISS by capturing the differential in voting between individual investors and institutional investors. We estimate the base regression (Model 1 of Table 3) by substituting a dummy variable for the recommendation with two interaction variables. First, we multiply a dummy variable for the recommendation (VoteISS taking a value of 1 if the ISS recommendation is withhold and 0 if the ISS recommendation is “for”) by the fraction of shares held by institutional investors (Insthold). Second, we multiply a dummy variable for the recommendation by the fraction of shares held neither by institutional investors nor by board members as a proxy for holdings by individual investors (Indivhold). 133“Insthold” is defined as the fraction of outstanding shares of the company in question in the hands of institutional investors, measured using Form 13-F data obtained from Thomson Financial for the time period immediately prior to the annual meeting date. “Indivhold” is defined as 1 – InstHold – Tot_Dir_Shs. Given the assumption that individuals do not directly receive the ISS recommendation, we posit that any relationship between an ISS withhold recommendation and votes by individual investors (Indivhold x VoteISS) must be the result of individuals responding to some other observable factor that is not directly included in our regressions, 134 See discussion of potentially omitted variables supra Part III.A. but for which the ISS recommendation in our model acts as a proxy. We then use the differential between Insthold x VoteISS and Indivhold x VoteISS to estimate the effect of ISS’s influence on the proxy vote.

Table 6 reports the results of the regression with the Insthold x VoteISS and Indivhold x VoteISS interaction variables. The coefficients for the Insthold x VoteISS interaction terms are more negative than the coefficients for the Indivhold x VoteISS interaction terms. In unreported F-tests, the difference in coefficients is significant at the <1% level. These results are consistent with the hypothesis that an ISS withhold recommendation has a greater impact on voting by institutions than by individuals.

The results also enable us to estimate the effect of an ISS recommendation. We start by making the following three assumptions: First, a recommendation affects the vote of some institutional—but not any individual—investors. Second, if they did not follow ISS, institutional investors would base their votes on the same underlying factors as individual investors. Third, ISS does not provide to its clients any additional information about these underlying factors that is not known to individual investors.

Under these assumptions, the voting record of individual investors is a perfect proxy for how institutional investors would vote if ISS did not exist. This is so because the votes by individual investors are not themselves affected by ISS (first assumption) and because institutions would vote the same way as individuals but for ISS (second assumption). Moreover, any influence of ISS is entirely due to its bottom-line recommendation, not to any information and analysis accompanying its recommendation (third assumption). The power of ISS thus can be measured by the difference in the coefficients for the interaction term with institutional holdings and the interaction term with individual holdings (for example, a difference of -1.889 for ISS). This variable will measure any absolute power by ISS as well as any contingent power, but only to the extent that ISS clients would have voted differently had they not followed ISS.

Because our dependent variable is the log odds of the “for” vote, we use the following methodology to quantify ISS’s power. We start with the overall marginal impact of 13.1 percentage points from an ISS withhold recommendation on the base “for” vote model measured at the median of the log-odds “for” vote distribution (as reported in Table 4). We then apportion the overall marginal impact of an ISS withhold recommendation between the effect on institutional and individual holdings. For our entire sample, the mean fraction of institutional ownership is 0.60, and the mean fraction of individual holdings is 0.35. In the model, the coefficient estimate for Insthold x ISS is -3.137, and for Indivhold x ISS it is -1.248. Therefore, we calculate the relative contribution of the Insthold x ISS variable on the overall marginal impact of an ISS withhold recommendation as (3.137*0.60)/(3.137*0.60 + 1.248*0.35) = 81.2%. Stated differently, the marginal impact of an ISS withhold recommendation is 10.6 points of the 13.1 overall marginal impact. The relative contribution of the Indivhold x ISS variable is 18.8% (or 2.5 percentage points of the 13.1 overall marginal impact). This suggests that an ISS withhold recommendation reduces the “for” vote of institutional holders by 17.7% (10.6/60) and the “for” vote by individual holders by 7.0 % (2.5/35). If, as assumed, the relationship between the ISS recommendation and the vote is due to other factors correlated with the ISS recommendation, not the recommendation itself, and these factors have the same impact on the institutional vote, then the real effect of an ISS withhold recommendation is to reduce the institutional “for” vote by 10.7 percentage points of the institutional vote. Multiplying ISS’s relatively greater influence with institutional investors (the 10.7 percentage points) by the fraction of votes held on average by institutional investors (60% of the votes) yields 6.4 % of the overall vote.

Note that this result is critically dependent on our foundational assumptions. To the extent that the first assumption is incorrect, and some individual investors follow the ISS recommendations (or some institutional investors are misclassified in our data as individuals), our result would understate the magnitude of ISS’s power. We think this is unlikely because ISS recommendations are rarely publicized in uncontested elections and because individual investors are unlikely to automate their voting decisions. Thus, they would only learn of an ISS recommendation through independent research.

To the extent that the second assumption is incorrect and institutional investors who follow ISS vote differently from individuals, our result would overstate ISS’s power to the degree that institutional investors that follow ISS pay more attention to the factors that affect an ISS recommendation than do individual investors. It would also understate ISS’s power to the extent that institutional investors that follow ISS would pay less attention to the factors that affect an ISS recommendation than do individual investors.

There are two reasons to believe that the second assumption is at least partially incorrect and that it biases our results towards overstating ISS’s power. First, a significant portion of the votes attributed to individuals in our methodology are actually brokers’ discretionary votes. 135 See discussion of discretionary broker voting in director elections, supra text accompanying notes 29–34. According to one estimate, an average of 19% of all votes cast are broker discretionary votes. 136 See Allen, supra note 32. Traditionally, brokers exercised their discretionary voting authority in accordance with management recommendations, that is, for the board nominees. 137 See Marcel Kahan & Edward Rock, The Hanging Chads of Corporate Voting, 96 Geo. L.J. 1227, 1269 (2008) (“[B]rokers tend to vote in accordance with management recommendations . . . .”). Although a few brokers have adopted other voting measures and either abstain from voting or vote uninstructed shares in the same proportion as shares for which they have obtained voting instructions, 138 See Kahan & Rock, supra note 2, at 30 (discussing the shift in some brokers’ voting strategies). We are not aware of any information to suggest that brokers vote the shares of their clients for which they received no voting instructions in accordance with ISS recommendations. Charles Schwab policy dictates that it votes securities held in its customers’ brokerage accounts, for which it has not received voting instructions, in proportion to “all instructed shares held by Schwab.” N.Y. Stock Exch., Report and Recommendations of the Proxy Working Group to the New York Stock Exchange 16 (2006) (describing Schwab’s adoption of proportional voting in 2005). generally broker votes are more favorable to management than shares voted by their beneficial owners. 139 See N.Y. Stock Exch., supra note 138, at 13 (describing the anticipated effect on uncontested director elections of eliminating broker discretionary voting). If one divides the shareholdings of individuals into broker votes—which are automatically votes “for” each nominee—and remaining shares, the coefficient estimate for the interaction between the ISS recommendation and the individual shareholdings not part of such broker votes would be higher than the coefficient estimate in Table 6. Therefore, the estimate of the difference between that coefficient and the coefficient for Insthold x VoteISS—the measure of ISS’s power—would be lower.

Second, institutional investors who follow ISS have made an affirmative choice to do so. While some institutions may have followed the ISS recommendation as an easy way to satisfy their duty to cast an informed vote, others may have done so because they are in overall philosophical agreement with the way in which ISS makes voting recommendations. Even institutions that for practical reasons want to follow the recommendations of some advisor can choose which advisor to follow. Thus, it is likely that those institutions that choose to follow ISS differ in their voting preferences from—and are closer to the voting preference of ISS than—those shareholders who have made no such choice.

Finally, to the extent that our third assumption is incorrect and ISS provides additional information to its subscribers that individual investors do not have, our estimate of ISS’s power would also include the following two components. First, it would include the votes by clients for which ISS acts as a pure information agent. Votes by these investors, as discussed in Part II, are not based on the bottom-line ISS recommendation, but rather on the information provided by ISS. Second, our estimate would include the votes by ISS clients who base their votes on the ISS recommendation (but not on the information provided by ISS), but who would have voted the same way ISS recommended if they had known of the additional information provided by ISS.

Conclusion

In this paper we analyze the significance of voting recommendations issued by proxy advisors. Our examination includes four advisory firms: ISS, Glass Lewis, Egan Jones, and Proxy Governance. We find, consistent with press reports, that ISS is the most powerful proxy advisor. Of the others, only Glass Lewis seems to have a meaningful impact on the shareholder voting.

We conduct several tests to quantify the impact of an ISS recommendation. Although superficial analyses suggest that an ISS recommendation can have a marginal impact of as much as 20%, and press reports state that ISS has the power to shift 20% to 30% of the shareholder vote, we conclude that these numbers are substantially overstated. In particular, our findings reveal that although an ISS recommendation has independent value, this value is greatly reduced once we take into account the company- and firm-specific factors that are important to investors. Depending on the test, we find that the impact of an ISS recommendation ranges from 6% to 13% for the median company. Overall, we consider it likely that an ISS recommendation shifts 6% to 10% of shareholder votes—a material percentage but far less than commonly attributed to ISS.

Furthermore, we find evidence that ISS’s power is partially due to the fact that ISS (to a greater extent than other advisors) bases its recommendations on factors that shareholders consider important. This fact and competition among proxy advisors place upper bounds on ISS’s power. Institutional Shareholder Services cannot issue recommendations arbitrarily if it wants to retain its market position. Doing so would lead institutional investors to seek the services of other proxy advisory firms. Thus, ISS is not so much a Pied Piper followed blindly by institutional investors as it is an information agent and guide, helping investors to identify voting decisions that are consistent with their existing preferences.

Table 1

Panel A: Coverage and Withhold Rates

N

Coverage Rate

Number of

Withhold Recs.

Number of “For” Recs.

Withhold Rate

All

16038

1.00

ISS

15823

0.99

1073

14750

0.068

GL

15722

0.98

2956

12766

0.188

EJ

14147

0.88

1551

12596

0.110

PGI

5437

0.34

202

5235

0.037

Panel B: Correlation Matrix of Recommendations

VoteISS

VoteGL

VoteEJ

VotePGI

VoteISS

1

VoteGL

0.1683

1

VoteEJ

0.1803

0.1425

1

VotePGI

0.1057

0.0736

0.0548

1

VoteISS equals 1 if ISS gives a Withhold recommendation and 0 if ISS gives a “for” recommendation. VoteGL, VoteEJ, and VotePGI are defined similarly.

Panel C: Recommendation and Percentage “For” Vote—Single Advisor

Percentage of “For” Votes (mean)

Percentage of “For” Votes where Advisor Rec. =

For (mean)

Percentage of “For” Votes where Advisor Rec. = Withhold (mean)

Marginal Impact

Total

95.12%

ISS

96.44%

76.14%

20.3%

GL

96.25%

90.05%

6.2%

EJ

95.75%

91.02%

4.73%

PGI

95.39%

91.90%

3.49%

choi-fisch-kahan-fig1

Table 2

Variable

=0

=1

N

ForVote

N

ForVote

% Difference

p-value

CEO

12566

95.0

1471

95.9

0.9

0.0000

New Director

13050

94.9

2232

96.6

1.7

0.0000

AuditMbr

8389

95.0

5564

95.4

0.4

0.0009

Prior Restat

13441

95.2

1841

94.4

-0.8

0.0000

Prior SEC

14210

95.2

1072

94.4

-0.8

0.0008

CompMbr

8585

95.5

5368

94.5

-1.0

0.0000

Top5AbComp

13870

95.2

707

94.4

-0.8

0.0033

Attendance

13820

95.2

93

79.3

-15.9

0.0000

ManyBds

12107

95.2

1280

94.2

-1.0

0.0000

ManyBds x

CEO

13334

95.2

53

96.7

1.6

0.0977

Age75

13488

95.2

1794

94.9

-0.3

0.0967

NomMbr

8460

95.4

5493

94.7

-0.7

0.0000

Empl_Dir

13065

95.1

888

94.7

-0.4

0.0495

OutDirLink

12489

95.6

1464

90.7

-4.9

0.0000

Interlock

13916

95.1

37

92.7

-2.4

0.0366

Chairman_

Only

13719

95.1

318

94.9

-0.2

0.5173

IP No

15136

95.2

146

82.8

-12.4

0.0000

ClassBd

9459

95.3

5516

94.8

-0.5

0.0000

PPill

7242

95.4

7733

94.9

-0.6

0.0000

CumVote

13345

95.1

1630

95.1

0.0

0.8105

GP

3700

94.9

11275

95.2

0.3

0.0710

Top5AbRet

14505

95.1

755

95.0

-0.1

0.6463

Bot5AbRet

14536

95.2

724

93.3

-1.9

0.0000

Table 2.1

Interaction Variable

=0

=1

N

ForVote

N

ForVote

%

Difference

p-value

Prior Restat x AuditMbr

13305

95.2

648

94.3

-0.8

0.0025

Prior SEC x AuditMbr

13607

95.1

346

94.4

-0.7

0.0490

Top5AbComp x CompMbr

13083

95.2

241

92.9

-2.3

0.0000

Top5AbRet x CEO

13947

95.1

85

95.6

0.5

0.5169

Bot5AbRet x CEO

13963

95.1

69

94.5

-0.6

0.4434

The =1 group is where the variable in question is equal to 1 (For example, Prior Restat x AuditMbr=1 means the director is a member of the audit committee, and the company experienced a first public announcement of an accounting restatement within the two years prior to the annual meeting.). The =0 group is where the variable in question is equal to 0.

The p-value is from a two-sided t-test of the difference in the mean ForVote between the =0 and =1 groups for each in the interaction variables.

Table 3: “For” Vote Outcome

Variable

Model 1

No Advisor

Model 2

ISS

Model 3

GL

Model 4

EJ

Model 5

PGI

CEO

-0.714**

-0.648**

-0.777**

-0.709**

-0.574**

(-9.84)

(-9.20)

(-10.73)

(-9.54)

(-5.93)

New Director

0.335**

0.278**

0.218**

0.295**

0.321**

(6.61)

(5.61)

(4.34)

(5.91)

(5.11)

AuditMbr

-0.251**

-0.264**

-0.193**

-0.201**

-0.126*

(-5.29)

(-5.70)

(-4.26)

(-4.40)

(-1.97)

Prior Restat

-0.0961

-0.125

-0.0722

-0.137

-0.226

(-0.68)

(-0.93)

(-0.53)

(-0.94)

(-1.46)

Prior SEC

-0.282**

-0.260**

-0.234*

-0.287**

0.0536

(-2.83)

(-2.71)

(-2.57)

(-2.92)

(0.36)

Prior Restat

-0.235*

-0.192+

-0.126

-0.250*

-0.119

x AuditMbr

(-2.06)

(-1.73)

(-1.20)

(-2.25)

(-0.99)

Prior SEC

-0.0959

-0.0914

-0.103

-0.0788

-0.381*

x AuditMbr

(-1.08)

(-1.13)

(-1.25)

(-0.92)

(-2.39)

CompMbr

-0.381**

-0.334**

-0.329**

-0.354**

-0.265**

(-8.24)

(-7.38)

(-7.35)

(-7.31)

(-4.40)

Top5AbComp

-0.105

-0.0690

-0.109

-0.0733

0.00164

(-0.66)

(-0.51)

(-0.69)

(-0.46)

(0.01)

Top5AbComp

-0.462**

-0.376**

-0.310+

-0.463**

-0.493

x CompMbr

(-2.78)

(-3.10)

(-1.85)

(-2.79)

(-1.52)

Attendance

-1.907**

-1.188**

-1.272**

-1.512**

-1.801**

(-10.64)

(-9.27)

(-8.33)

(-8.26)

(-8.84)

ManyBds

-0.394**

-0.266**

-0.298**

-0.164**

-0.287**

(-7.14)

(-5.47)

(-5.56)

(-2.77)

(-3.81)

ManyBds x

0.264

0.0428

0.231

0.177

0.239

CEO

(1.56)

(0.26)

(1.36)

(1.09)

(1.03)

Age75

-0.351**

-0.341**

-0.306**

-0.355**

-0.199

(-3.05)

(-3.02)

(-2.75)

(-2.89)

(-1.01)

NomMbr

-0.168**

-0.126**

-0.108**

-0.134**

-0.113*

(-4.22)

(-3.29)

(-2.77)

(-3.24)

(-2.06)

Empl_Dir

-1.030**

-0.848**

-0.885**

-0.957**

-0.943**

(-11.04)

(-10.05)

(-9.89)

(-10.09)

(-7.30)

OutDirLink

-1.303**

-0.967**

-1.022**

-1.156**

-1.245**

(-18.24)

(-14.72)

(-14.96)

(-15.41)

(-10.82)

Tot_Dir_Shs

1.183**

1.424**

1.246**

1.172*

1.855*

(2.60)

(2.95)

(2.68)

(2.50)

(2.43)

Interlock

-0.110

-0.306+

0.334+

-0.127

0.121

(-0.61)

(-1.77)

(1.92)

(-0.78)

(0.42)

Chairman_

0.0349

-0.0803

-0.102

-0.00752

-0.0894

Only

(0.31)

(-0.80)

(-0.91)

(-0.06)

(-0.50)

IP No

-1.507**

-0.631**

-1.360**

-1.381**

-1.893**

(-5.10)

(-4.54)

(-5.14)

(-4.83)

(-9.00)

ClassBd

-0.263**

-0.208**

-0.252**

-0.277**

-0.146

(-3.44)

(-2.89)

(-3.37)

(-3.53)

(-1.32)

PPill

-0.0929

-0.0770

-0.0950

-0.121

-0.179

(-1.10)

(-0.97)

(-1.14)

(-1.37)

(-1.42)

CumVote

-0.0859

-0.0157

-0.100

-0.0711

0.113

(-0.64)

(-0.12)

(-0.76)

(-0.50)

(0.69)

GP

-0.0592

-0.126

-0.0638

-0.0536

-0.0110

(-0.61)

(-1.37)

(-0.67)

(-0.53)

(-0.07)

Top5AbRet

0.445*

0.414*

0.394*

0.385*

0.475*

(2.57)

(2.44)

(2.31)

(2.25)

(2.18)

Bot5AbRet

-0.484**

-0.438**

-0.391**

-0.563**

-0.677**

(-3.47)

(-3.22)

(-2.94)

(-3.15)

(-3.82)

Top5AbRet x

-0.134

-0.109

-0.0603

-0.0317

-0.232

CEO

(-0.66)

(-0.55)

(-0.30)

(-0.16)

(-1.26)

Bot5AbRet x

0.310+

0.318*

0.177

0.212

0.119

CEO

(1.92)

(2.05)

(1.09)

(1.59)

(0.49)

Sdret

-31.62**

-26.91**

-28.99**

-29.53**

-17.93+

(-4.48)

(-3.91)

(-4.14)

(-3.82)

(-1.81)

ln(Market

-0.126**

-0.125**

-0.132**

-0.131**

-0.119*

Capitalization)

(-3.93)

(-4.17)

(-4.16)

(-3.69)

(-2.55)

InstHold

0.465

0.460

0.474

0.363

0.486

(1.42)

(1.48)

(1.46)

(1.04)

(1.05)

Year06

0.0197

0.00358

0.0420

0.00388

0.0910

(0.30)

(0.06)

(0.65)

(0.06)

(1.05)

VoteISS

-2.216**

(-25.54)

VoteGL

-1.182**

(-21.52)

VoteEJ

-0.595**

(-9.22)

VotePG

-0.559**

(-3.22)

Constant

6.003**

5.946**

6.084**

6.078**

5.309**

(11.80)

(12.34)

(11.96)

(10.84)

(7.60)

N

12644

12605

12563

11447

4624

adj. R2

0.109

0.185

0.163

0.119

0.143

t statistics in parentheses: + p < 0.10, * p < 0.05, ** p < 0.01

Table 4: Marginal Impact of a Withhold Recommendation at Varying Points on “For” Vote Distribution (Calculated Using a “For” Recommendation)

“For” Vote Distribution

ISS Withhold

GL Withhold

EJ Withhold

PGI

Withhold

5%

0.2547

0.0859

0.0403

0.0437

10%

0.2171

0.0692

0.0312

0.0340

25%

0.1699

0.0505

0.0216

0.0221

50%

0.1310

0.0368

0.0154

0.0157

75%

0.1015

0.0273

0.0113

0.0117

90%

0.0780

0.0209

0.0085

0.0087

95%

0.0660

0.0180

0.0071

0.0074

Table 5

Panel A: “For” Vote Outcome

Variable

Model

CEO

-0.709**

(-10.07)

New Director

0.173**

(3.51)

AuditMbr

-0.210**

(-4.72)

Prior Restat

-0.104

(-0.80)

Prior SEC

-0.229**

(-2.62)

Prior Restat x AuditMbr

-0.0837

(-0.83)

Prior SEC x AuditMbr

-0.0901

(-1.20)

CompMbr

-0.290**

(-6.62)

Top5AbComp

-0.0718

(-0.52)

Top5AbComp x CompMbr

-0.251*

(-2.04)

Attendance

-0.850**

(-6.98)

ManyBds

-0.190**

(-4.00)

ManyBds x CEO

0.0192

(0.12)

Age75

-0.296**

(-2.71)

NomMbr

-0.0838*

(-2.24)

Empl_Dir

-0.724**

(-9.01)

OutDirLink

-0.763**

(-11.93)

Tot_Dir_Shs

1.499**

(3.03)

Interlock

0.116

(0.68)

Chairman_Only

-0.188+

(-1.90)

IP No

-0.557**

(-4.11)

ClassBd

-0.200**

(-2.83)

PPill

-0.0809

(-1.03)

CumVote

-0.0394

(-0.31)

GP

-0.126

(-1.40)

Top5AbRet

0.369*

(2.21)

Bot5AbRet

-0.350**

(-2.74)

Top5AbRet x CEO

-0.0444

(-0.22)

Bot5AbRet x CEO

0.211

(1.35)

Sdret

-24.60**

(-3.61)

ln(Market Capitalization)

-0.131**

(-4.39)

InstHold

0.470

(1.51)

Year06

0.0344

(0.54)

VoteISS

-2.473**

(-26.62)

VoteGL

-1.123**

(-21.94)

VoteISS x VoteGL

0.995**

(6.74)

Constant

6.024**

(12.51)

N

12524

adj. R2

0.230

t statistics in parentheses: + p < 0.10, * p < 0.05, ** p < 0.01. Panel B: Expected and Marginal “For” Votes

ISS “For” Rec.

ISS Withhold Rec.

Marginal Impact of ISS Withhold Rec.

Glass Lewis

“For” Rec.

98.4%

83.9%

-14.5%

Glass Lewis Withhold Rec.

95.3%

82.1%

-13.2%

Marginal Impact of Glass Lewis Withhold Rec.

-3.1%

-1.8%

All expected and marginal “For” votes are calculated at the median of the log odds for vote distribution.

Table 6: Institutional v. Non-Institutional Investor Model

Variable

Model

Tot_Dir_Shs

1.290**

(2.77)

InstHold

0.583+

(1.81)

InstHold x VoteISS

-3.137**

(-15.29)

IndivHold x VoteISS

-1.248**

(-4.52)

Constant

5.895**

(12.21)

N

12605

adj. R2

0.187

t statistics in parentheses: + p < 0.10, * p < 0.05, ** p < 0.01. Unreported, the models all include the same independent variables as in the base log-odds for vote model (reported above as Model 1 of Table 3).

Footnotes

Murray and Kathleen Bring Professor of Law, New York University School of Law.Perry Golkin Professor of Law, University of Pennsylvania Law School.George T. Lowy Professor of Law, New York University School of Law.The authors thank Glass Lewis, Proxy Governance, and Egan Jones for providing access to their voting recommendations and policies. Earlier drafts of this paper were presented at Duke University School of Law, University of Virginia School of Law, UCLA School of Law, and Yale Law School, and we received many useful comments at each session. We also thank Un Kyung Park for her helpful comments. Marcel Kahan thanks the Milton and Miriam Handler Foundation for financial support. The Filomen D'Agostino and Max E. Greenberg Research Fund provided summer research support for this Article.

1 See Albert Verdam, An Exploration of the Role of Proxy Advisors in Proxy Voting (Feb. 2007) (unpublished manuscript, available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=978835) (describing the emergence of proxy advisors); see also U.S. Gov’t Accountability Office, Corporate Shareholder Meetings: Issues Relating to Firms That Advise Institutional Investors on Proxy Voting 6–12 (2007) [hereinafter GAO, Corporate Shareholder], available at http://www.gao.gov/new.items/d07765.pdf (exploring competition and potential conflicts of interest in the proxy advisor market); Colin Diamond & Irina Yevmenenko, Who Is Overseeing the Proxy Advisors?, 3 Bloomberg Corp. L.J. 606, 608 (2008) (highlighting the proxy advisor market).

2Marcel Kahan & Edward Rock, Embattled CEOs, 88 Tex. L. Rev. 987 (2010); see also Paul Rose, The Corporate Governance Industry, 32 J. Corp. L. 887, 897 (2007) (“In 1965, institutional investors held 16% of U.S. equities; by 2001, institutional investors held 61%.”).

3According to conventional wisdom, these institutional investors generally do not care enough about their votes to cast an informed ballot. They hold shares in too many companies, so any particular stake represents a small fraction of their portfolio, and how they vote is unlikely to affect the outcome and even if it did, the effect on the value of their portfolio would be minimal. Researching the issues on a company’s annual meeting agenda is costly, and institutions may also lack the necessary expertise to evaluate these issues adequately. See, e.g., Rose, supra note 2, at 897 (“Unless an institutional investor believes that it can conduct research for less, or that more expensive but discerning research will enable it to obtain better returns (after subtracting its own research costs), the investor may be better off outsourcing its corporate governance research.”); Omari Scott Simmons, Taking the Blue Pill: The Imponderable Impact of Executive Compensation Reform, 62 SMU L. Rev. 299, 354 (2009) (“Institutional investors, despite having greater capacity to monitor and gather information, may have too small a stake in a company or too limited industry expertise to monitor it actively.”).

4 See, e.g., GAO, Corporate Shareholder, supra note 1, at 13 (describing ISS’s client base as consisting of an estimated 1,700 institutional investors).

5 See, e.g., Glass, Lewis & Co., Proxy Paper: Proxy Research and Voting Recommendations on Global Proxies, http://www.glasslewis.com/solutions/proxypaper.php (last visited Aug. 11, 2009) (describing Glass Lewis’s proxy research, voting recommendations, and voting platform for voting subscribers’ shares); RiskMetrics Group, Custom Proxy Advisory, http://www.riskmetrics.com/custom_proxy_advisory (last visited Aug. 11, 2009) (describing how ISS works with clients to develop customized voting guidelines).

6RiskMetrics Group, Proxy Advisory Services, http://www.riskmetrics.com/proxy_advisory/options (last visited Aug. 11, 2009) (detailing the choice of ISS guidelines that subscribers can use and incorporate into “RiskMetrics’ turnkey voting agency services”).

7 See generally GAO, Corporate Shareholder, supra note 1.

8 See id. at 4.

9Stephen Davis, White Knight Swoops in for Glass Lewis, Directorship, Dec. 2007/Jan. 2008, at 7 (“Glass Lewis is the world’s second biggest proxy adviser next to RiskMetrics . . . .”).

10 See, e.g., Pallavi Gogoi, Support for Bank of America CEO Wanes; Shareholders Meet Today, and Many Want Him Out, USA Today, Apr. 29, 2009, at B1 (describing RiskMetrics as “[i]nfluential”); Robert D. Hershey Jr., A Little Industry with a Lot of Sway on Proxy Votes, N.Y. Times, June 18, 2006, § 3, at 6 (quoting David W. Smith, president of the Society of Corporate Secretaries and Governance Professionals, stating that “‘[t]he influence [proxy] advisers wield is extraordinary’”).

11 See, e.g., Matt O’Sullivan, When Only a Corporate Jet Will Do, Sydney Morning Herald, May 28, 2009, at 25 (describing RiskMetrics as “America’s most powerful shareholder voting adviser”).

12 See, e.g., Kim Clark, Reading Proxies for Fun and Profit, U.S. News & World Rep., May 22, 2006, at EE10 (describing Glass Lewis’s “growing clout”); Joann S. Lublin, RiskMetrics’s Head Faces His Day of Shareholder Judgment, Wall St. J., June 2, 2008, at C1 (“ISS Governance Services . . . exerts tremendous clout in advising institutional investors on proxy fights . . . .”).

13Jie Cai, Jacqueline L. Garner & Ralph A. Walkling, Electing Directors, 64 J. Fin. 2389, 2404 (2009).

14Jennifer E. Bethel & Stuart L. Gillan, The Impact of the Institutional and Regulatory Environment on Shareholder Voting, Fin. Mgmt, Winter 2002, at 29, 30.

15Posting of William J. Holstein to BNET: The Corner Office, http://blogs.bnet.com/ceo/?p=1100&tag=content;col1 (Feb. 7, 2008, 08:03).

16 See Rose, supra note 2, at 889 (attributing this view to executives).

17Leo E. Strine, Jr., The Delaware Way: How We Do Corporate Law and Some of the New Challenges We (and Europe) Face, 30 Del. J. Corp. L. 673, 688 (2005).

18Diamond & Yevmenenko, supra note 1, at 617.

19Leo E. Strine, Jr., Toward a True Corporate Republic: A Traditionalist Response to Bebchuk’s Solution for Improving Corporate America, 119 Harv. L. Rev. 1759, 1765 (2006) (“Unlike corporate managers, neither institutional investors as stockholders nor ISS as a voting advisor owe fiduciary duties to the corporations whose policies they seek to influence.”).

20 See GAO, Corporate Shareholder, supra note 1, at 8–9 (observing that, as pension consultants, ISS and Proxy Governance, Inc. (PGI) are registered with the SEC as investment advisors while Glass Lewis and Egan-Jones are not registered as investment advisors).

21 See id. at 14 (acknowledging that “newer proxy advisory firms may face challenges attracting clients and establishing themselves in the industry”).

22 Id. at 7; Rose, supra note 2, at 899 (“ISS is the dominant firm in the corporate governance industry . . . .”).

23In 2005, more than 90% of S&P 500 companies employed plurality voting. See, e.g., Brooke A. Masters, Shareholders Flex Muscles; Proxy Measures Pushing Corporate Accountability Gain Support, Wash. Post, June 17, 2006, at D1 (stating that, as of the start of 2005, fewer than thirty S&P 500 companies had majority voting or director resignation policies in place). By 2008, over 80% had moved away from plurality voting. Kahan & Rock, supra note 2, at 23.

24Under plurality voting, a shareholder in an uncontested election may cast a vote in favor of a director candidate or withhold voting authority but may not cast a vote “against” the nominees. See, e.g., Jill E. Fisch, The Transamerica Case, in The Iconic Cases in Corporate Law 46, 68 (Jonathan R. Macey ed., 2008) (explaining the concept of “withheld” votes).

25 See id. at 69 (explaining majority voting).

26Commentators have described effective staggered boards as the most powerful anti-takeover device and thus the most effective mechanism by which boards can insulate themselves from shareholder voice. See, e.g., Lucian Arye Bebchuk, John C. Coates IV & Guhan Subramanian, The Powerful Antitakeover Force of Staggered Boards: Theory, Evidence, and Policy, 54 Stan. L. Rev. 887, 889–91 (2002).

27Stephen Taub, Revival of Classified Boards? Well, Maybe Not, Compliance Wk., Sept. 11, 2007, http://www.complianceweek.com/article/3647/revival-of-classified-boards-well-maybe-not.

28 See, e.g., Del. Code Ann. tit. 8, § 141(d) (2009); see also Gregory T. Carrott, The Case for and Against Staggered Boards, Directorship, Sept. 22, 2009, http://www.directorship.com/against-staggered-boards/ (explaining that, most often, staggered boards provide directors with three year terms).

29In October 2006, the New York Stock Exchange (NYSE) proposed to amend Rule 452 governing broker votes to redefine all director elections as “non-routine,” which would eliminate the ability of brokers to cast discretionary votes. Proxy Working Group, Report and Recommendations of the Proxy Working Group to the New York Stock Exchange 3 (2006), available at http://www.nyse.com/pdfs/PWG_REPORT.pdf. On July 1, 2009, the SEC finally approved the amendments, effectively ending broker discretionary voting in director elections. Order Approving Proposed Rule Change, as Modified by Amendment No. 4, to Amend NYSE Rule 452 and Corresponding Listed Company Manual Section 402.08 to Eliminate Broker Discretionary Voting for the Election of Directors, Except for Companies Registered Under the Investment Company Act of 1940, and to Codify Two Previously Published Interpretations that Do Not Permit Broker Discretionary Voting for Material Amendments to Investment Advisory Contracts with an Investment Company, Exchange Act Release No. 34-60215, 74 Fed. Reg. 33,293 (July 1, 2009).

30NYSE, Inc., Rule 452 (Mar. 6, 2003).

31Kahan & Rock, supra note 2.

32 See Posting of Ted Allen to RiskMetrics Group, http://blog.riskmetrics.com/gov/2007/05/sec-hears-testimony-on-broker-votessubmitted-by-ted-allen-director-of-publications.html (May 25, 2007, 10:58) (attributing this figure to Broadridge Financial).

33 See David A. Katz & Laura A. McIntosh, A Seismic Shift in Mechanics of Electing Directors, N.Y. L.J., July 27, 2006, at 5 (“If, in the aftermath of NYSE rule changes as proposed, issuers indeed are unable to contact or obtain voting instructions from large numbers of individual shareholders, the effect will be a massive shift of voting power from brokers to institutions, and, therefore, to proxy advisory services such as ISS, Glass, Lewis & Co., and Proxy Governance.”).

34Citigroup, Inc., First Quarter of 2009 (Form 10-Q), at 156–57 (May 11, 2009), available at http://www.citigroup.com/citi/fin/data/q0901c.pdf?ieNocache=643. (indicating broker votes of 1.732 billion shares).

35For information on Georgeson, see http://www.georgeson.com/.

36Georgeson, Annual Corporate Governance Review 7 (2008), available at http://www.georgeson.com/usa/download/acgr/acgr2008.pdf.

37 Id. at 7–8. In 2004, twelve directors failed to receive a majority of votes cast. Fisch, supra note 24, at 68.

38Georgeson, supra note 36, at 8.

39 Id. at 46.

40Rules and Regulations Under the Securities Exchange Act of 1934: Solicitations of Proxies, 17 C.F.R. § 240.14a-8 (2008).

41 See, e.g., Jill E. Fisch, From Legitimacy to Logic: Reconstructing Proxy Regulation, 46 Vand. L. Rev. 1129, 1152–55 (1993) (describing the rise in shareholder use of social policy proposals in the 1950s and 1960s).

42 See, e.g., A. A. Sommer, Jr., Corporate Governance in the Nineties: Managers vs. Institutions, 59 U. Cin. L. Rev. 357, 371 (1990) (describing the shift from proposals “having a social dimension” to those dealing with corporate governance).

43 See, e.g., BNA, Annual Meeting Voting Compels More Accountability, 11 Corp. Governance Rep. 30 (2008) (listing “say on pay” executive compensation proposals as one of the top three issues on corporate ballots for 2008).

44 See, e.g., Am. Fed’n of State, County & Mun. Employees v. Am. Int’l Group, 462 F.3d 121, 123 (2d Cir. 2006) (addressing a shareholder proposal on proxy access).

45 See Georgeson, supra note 36, at 14 fig.3 (detailing corporate governance proposals from 2004 to 2008).

46Kahan & Rock, supra note 2, at 27 tbl.4 (citing Georgeson, Corporate Governance: Annual Meeting Season Wrap Up (2001), available at http://www.georgeson.com/usa/download/acgr/acgr2001.pdf; Georgeson, supra note 36). Prior to 2001, Georgeson prepared a similar report, but it analyzed only corporate governance proposals made by institutional investors. Georgeson, Corporate Governance: Annual Meeting Season Wrap Up (2000), available at http://www.georgeson.com/usa/download/acgr/acgr2000.pdf.

47Kahan & Rock, supra note 2.

48Fisch, supra note 37, at 63–67 (reviewing the history of proxy access proposals); Kahan & Rock, supra note 2 (reviewing the history of proxy access proposals). The SEC first considered a proposed rule permitting proxy access in 1942. Fisch, supra note 37, at 63. In 2003, the SEC solicited comments on a complex proposal for proxy access; the proposal stalled due to opposition from corporations and lack of support from Republican commissioners. Id. at 65–66.

49Fisch, supra note 24, at 65–66.

50Am. Fed’n of State, County & Mun. Employees v. Am. Int’l Group, 462 F.3d 121, 123 (2d Cir. 2006) (holding that shareholders can introduce proxy access proposals under Rule 14a-8).

51Rules and Regulations Under the Securities Exchange Act of 1934: Solicitations of Proxies, 17 C.F.R. § 240.14a-8 (2008).

52 Del. Code Ann. tit. 8, § 112 (2009). The legislature also adopted a provision authorizing bylaws that provide for reimbursement of a shareholder’s proxy solicitation expenses. Del. Code Ann. tit. 8, § 113(a) (2009).

53Facilitating Shareholder Director Nominations, 17 C.F.R. §§ 200, 232, 240, 249, 274 (2009).

54Fisch, supra note 37, at 71 (describing say-on-pay).

55 See, e.g., Robert Kropp, Shareowner Resolutions on Say on Pay Gain Widespread Support, May 6, 2009, http://www.socialfunds.com/news/article.cgi?sfArticleId=2690 (reporting that seventy-nine say-on-pay resolutions were introduced in 2008, and more than one hundred have been filed in 2009).

56According to a preliminary count, as of May 2009, ten of the twenty-nine proposals that came up for a vote received majority support. Press Release, AFSCME, Say on Pay Shareholder Proposals Garner Record Support During Tumultuous Shareholder Season (May 4, 2009).

57 See, e.g., Editorial, Stockholders Should Demand a Say on Executive Pay, Seattle Times, May 14, 2009, available at http://seattletimes.nwsource.com/html/editorialsopinion/2009221158_editb15sayonpay.html (noting that resolutions had been approved by fifteen companies this year and that Hewlett-Packard and Occidental Petroleum had agreed to adopt say-on-pay without a shareholder vote); SmartPros.com, Say-On-Pay Is on the Way, http://accounting.smartpros.com/x65641/xml (last visited Feb. 7, 2010) (listing Occidental Petroleum, Intel, Hewlett-Packard, MBIA, Motorola, and Ingersoll-Rand as companies that have adopted say-on-pay).

58 See Lawrence Bard et al, Morrison Foerster, Administration Proposals on Compensation Committees and Say on Pay Would Affect All Public Companies, July 30, 2009, http://www.mofo.com/news/updates/files/15793.html (describing the Treasury Department’s draft legislation that would require the SEC to adopt rules mandating say-on-pay for all publicly traded companies).

59Shareholder Vote on Executive Compensation Act, H.R. 1257, 110th Cong. §2 (2007); Shareholder Vote on Executive Compensation Act, S. 1181, 110th Cong. §2 (2007).

60Stephen Taub, Obama Pushes Say on Pay Legislation, CFO.com, Apr. 11, 2008, http://www.cfo.com/article.cfm/11037327/c_11036422 (reporting then-Senator Obama’s support for say-on-pay).

61Emergency Economic Stabilization Act of 2008, Pub. L. No. 110343, 122 Stat. 3765, 110th Cong. (2008).

62Shareholder Approval of Executive Compensation of TARP Recipients, Exchange Act Release No. 34-61335, 75 Fed. Reg. 2789 (Jan. 12, 2010) (describing requirement of Section 111(e) of the Emergency Economic Stabilization Act of 2008 and amending federal proxy rules to implement the requirement).

63Bethel & Gillan, supra note 14, at 29.

64 Id. at 46.

65Cindy R. Alexander et al., The Role of Advisory Services in Proxy Voting (Nat’l Bureau of Econ. Research, Working Paper No. 15143, 2008), available at http://www.nber.org/paper/w15143.

66 Id. at 34–35.

67 See Cai et al., supra note 13.

68 Id. at 19.

69The article by Cai and others partially examines other factors that may influence the vote. See Cai, et al., supra note 13.

70 See GAO, Corporate Shareholder, supra note 1, at 7 (describing ISS, Glass Lewis, Proxy Governance, and Egan Jones as among the “five major firms” comprising the proxy advisory industry). The fifth firm included in the GAO report is Marco Consulting Group (MCG), which provides investment consulting services to Taft-Hartley funds and a number of public benefit plans but does not publicly issue voting recommendations. See Marco Consulting, Company History, http://www.marcoconsulting.com/1.2.html (last visited May 5, 2010) (“MCG only provides investment consulting and proxy voting services.”).

71We explore the relationship between these factors and proxy advisor recommendations in a prior article. Stephen J. Choi, Jill E. Fisch & Marcel Kahan, Director Elections and the Role of Proxy Advisors, 82 S. Cal. L. Rev. 649, 650–51 (2009).

72 See, e.g., Laura Lin, The Effectiveness of Outside Directors as a Corporate Governance Mechanism:Theories and Evidence, 90 Nw. U. L. Rev. 898, 921–39 (1996) (collecting empirical studies of board composition and effectiveness); cf. Sanjai Bhagat & Bernard Black, The Non-Correlation Between Board Independence and Long-Term Firm Performance, 27 J. Corp. L. 231, 263 (2002) (finding no correlation between director independence and long-term firm performance).

73These include director independence both from the company and the CEO, limited service on other corporate boards, and regular attendance at board meetings. See, e.g., In the Matter of the Walt Disney Co., Exchange Act Release No. 50882 (Dec. 20, 2004) (“The independence of directors is a linchpin of sound corporate governance, and is crucial to the objective oversight of management.”); Paul W. MacAvoy & Ira M. Millstein, The Recurrent Crisis in Corporate Governance 22–23 (2003) (stating that directors should act “independently of management”); Stephen P. Ferris et al., Too Busy to Mind the Business? Monitoring by Directors with Multiple Board Appointments, 58 J. Fin. 1087 (2003) (finding no evidence, contrary to popular wisdom, that multiple directors shirk their responsibilities); Renée B. Adams & Daniel Ferreira, Regulatory Pressure and Bank Directors’ Incentives to Attend Board Meetings 304 (European Corporate Governance Inst. Working Paper Series, Working Paper No. 203/2008, 2008), available at http://ssrn.com/abstract=936261 (discussing various directives that directors attend board meetings regularly). The federal proxy rules require issuers to disclose whether any director has attended fewer than 75% of the board meetings held during the prior fiscal year. Schedule 14A. Information Required in Proxy Statement, 17 C.F.R. § 240.14a-101, Item 7(f) (2009). The rules also require disclosure of outside directorships. Schedule 14A. Information Required in Proxy Statement, 17 C.F.R. § 240.14a-101, Item 22(b) (2009).

74 See, e.g., Mark Anderson, Eli Lilly Heads CalPERS’ ‘Underperforming’ List, Sacramento Bus. J., Mar. 19, 2009, http://sacramento.bizjournals.com/sacramento/stories/2009/03/16/daily56.html (describing CalPERS’s (the California Public Employees’ Retirement System) watch list as targeting companies with corporate governance defects that also “show weakness with profitability, transparency and/or management”).

75 See Choi, Fisch & Kahan, supra note 71, at 650–51.

76 Id. at 664–70.

77Paul W. Holland, Statistics and Causal Inference, 81 J. Am. Stat. Ass’n 945, 945 (1986).

78 See, e.g., ISS Governance Services, Proxy Alert, Citigroup Inc. 18 (Apr. 10, 2009) (stating that ISS recommends shareholders withhold their votes for board nominee Anne Mulcahy because she may be overextended as she sits on more than three boards and serves as CEO of Xerox Corporation).

79For example, the proxy advisor could provide information on the board’s failure to implement a shareholder proposal only when the advisor was recommending a withhold vote and not when the advisor was recommending a vote in favor of the nominees. In theory, proxy advisors could also misreport information. The ability of advisors to exercise power consistently by misreporting is quite limited, however. In addition to the market competition provided by other advisors, the company itself has a strong incentive to correct inaccuracies, and the media is likely to report any substantial errors. Thus proxy advisors have incentives to avoid recommendations that can be described as erroneous. Indeed, ISS received substantial adverse media attention for its recommendation that shareholders withhold their votes from Warren Buffett, a nominee to the Coca-Cola board, because of business relationships between Coca-Cola and some Berkshire Hathaway subsidiaries. See, e.g., Donald E. Graham, The Gray Lady’s Virtue, Wall St. J., Apr. 23, 2007, at A17, available at http://online.wsj.com/article/SB117728391033378436.html (describing ISS’s recommendation as “perhaps the single silliest recommendation ever made to shareholders”).

80The proxy voting guidelines of the Nathan Cummings Foundation, for example, reflect this role for the proxy advisor, indicating that the Foundation will vote for a director nominee if the company does not have a staggered board, if the company is not recommending against an issue proposal supported by the Foundation, and if RiskMetrics supports the nominee. Proxy Voting Practices, The Nathan Cummings Foundation, available at http://www.nathancummings.org/shareholders/pvgandvr/VotingGuidelines.pdf.

81 See, e.g., ISS Governance Services, 2008 U.S. Proxy Voting Guidelines Summary (2007), available at http://www.riskmetrics.com/sites/default/files/2008PolicyUSSummaryGuidelines.pdf.

82RiskMetrics, Proxy Advisory Services, http://www.riskmetrics.com/proxy_advisory/options (last visited Aug. 11, 2009) (describing different voting guideline options).

83 See Thompson Hine, SEC Enters Order Against Adviser Related to Proxy Voting, May 22, 2009, http://www.thompsonhine.com/publications/publication1818.html (describing SEC action); Press Release, U.S. Sec. & Exch. Comm’n, SEC Charges Investment Adviser for Proxy Voting Rule Violations (May 8, 2009), available at http://www.sec.gov/news/press/2009/2009-105.htm.

84The SEC noted that the INTECH decision created a potential conflict of interest in that “not all clients would agree with votes made pursuant to the ISS-PVS Guidelines and that voting proxies pursuant to the ISS-PVS Guidelines would benefit INTECH in obtaining and retaining union-affiliated clients.” Thompson Hine, supra note 83.

85 See GAO, Corporate Shareholder, supra note 1, at 13–14 (describing the market for proxy advisory services).

86William Baue, Report Urges Foundations to Vote Their Proxies, Social Funds, Mar. 4, 2004, http://www.socialfunds.com/news/article.cgi/1358.html (describing the low level of proxy voting by foundations and suggesting purchasing proxy voting service from ISS as a superior and reasonably priced alternative to refraining from voting).

87 See Rose, supra note 16, at 897–98 (noting Department of Labor and SEC regulations).

88 See Stephen J. Choi & Jill E. Fisch, On Beyond CalPERS: Survey Evidence on the Developing Role of Public Pension Funds in Corporate Governance, 61 Vand. L. Rev. 315, 324 (2008) (reporting that 20% of public pension funds surveyed reported delegating complete voting authority to ISS or a similar organization).

89The SEC has specifically stated that investment advisors can comply with their fiduciary obligations by using a ‘“predetermined voting policy,’ such as a third-party proxy voting service’s platform, to vote proxies provided that the predetermined policy is ‘designed to further the interests of clients rather than the adviser.’” INTECH Inv. Mgmt. LLC, Investment Advisers Act Rel. No. 2872, at 5 (May 7, 2009) (quoting Final Rule: Proxy Voting by Investment Managers, Investment Advisers Act Rel. No. 2106 (Jan. 31, 2003)), available at http://www.sec.gov/litigation/admin/2009/ia-2872.pdf.

90Institutional Shareholder Services recommendations were obtained through LEXIS. Glass Lewis, Egan Jones, and Proxy Governance provided us with their recommendations for the period in question. All of the companies in our sample that conducted a director election in 2005 had a proxy advisor recommendation for at least one of their directors.

91 See Choi, Fisch & Kahan, supra note 71, at 649.

92Percentage “for” vote is defined as the “for” votes as a percentage of the sum of “for” and withhold votes.

93 See supra text accompanying notes 9–12.

94 See supra text accompanying notes 12–16.

95These shareholders may switch if they find that ISS’s voting policies in fact do not match their preferences.

96 See infra text accompanying note 135. Since 2006, the number of directors with high withhold votes has increased. See Georgeson, supra note 36, at 7 (describing the increase in the number of high withhold votes).

97Georgeson, supra note 36, at 7 (charting the number of directors who received a withhold vote of 20% or more).

98 See Cai et al., supra note 13, at 2390; Diane Del Guercio et al., Do Boards Pay Attention When Institutional Investor Activists “Just Vote No”?, 90 J. Fin. Econ. 84 (2008) (finding operating performance improvement and increased CEO turnover in response to successful “vote no” campaigns).

99These data were obtained from the RiskMetrics-Investor Responsibility Research Center (IRRC) director database, available to subscribers of Wharton Research Data Services. The IRRC dataset consists of data on individual board directors from 1996 to 2006. The data include “a range of variables related to individual board directors (e.g., name, age, tenure, gender, committee memberships, independence classification, primary employer and title, number of other public company boards serving on, shares owned, etc.).” See RiskMetrics-Directors Legacy Data Request, Wharton Research Data Services (on file with authors).

100We use the IRRC data on other “major” company boards held by directors for the year prior to the annual meeting.

101These data were obtained from SEC filings, press releases, the IRRC Governance database, the Georgeson Annual Corporate Governance Reviews, and the Center for Research in Security Prices (CRSP). All of the data are publicly available or based on publicly available sources.

102The abnormal return is defined as the difference between the raw three-year holding period return for the company in question and the three-year holding period return for the CRSP value weighted market index.

103Top5AbComp is an indicator variable defined as equal to 1 if the total excess compensation for the CEO for the company in question is in the top 5% of the sample and 0 otherwise. We define total excess CEO compensation as the difference between the total CEO compensation for the year prior to the meeting date (as provided by the Compustat Executive Compensation database) minus the expected total CEO compensation. We calculate the expected total CEO compensation by (1) estimating an Ordinary Least Squares (OLS) model for: ln(Total CEO compensation) = α + β1ln(Market_Capitalization) +β2One_Year_Abnormal_Holding_Period_Return + β3Year_2006 + ε and (2) using the predicted Total CEO compensation based on this model as the expected Total CEO compensation.

104 See, e.g., Georgeson, supra note 36, at 8 (reporting that, in 2008, only thirty directors at S&P 1500 companies failed to receive a majority of “for” votes, compared to fifteen directors in 2007).

105 See, e.g., NYSE, Inc., Listed Company Manual §§ 303A.03, 303A.07(b) (2009); NASDAQ, Inc., Stock Market Equity Rules § 5605 (Mar. 12, 2009) (describing the composition of the audit committee (5605(c)(2)), independent director executive compensation (5605(d)), and independent director oversight of director nominations (5605(d))), available at http://nasdaq.cchwallstreet.com/NASDAQTools/PlatformViewer.asp?selectednode=chp_1_1_4_2&manual=%2Fnasdaq%2Fmain%2Fnasdaq-equityrules%2F.

106Studies have demonstrated relationships between audit committee composition and audit-related problems. See, e.g., Bradley Pomeroy & Daniel B. Thornton, Meta-Analysis and the Accounting Literature: The Case of Audit Committee Independence and Financial Reporting Quality, 17 Eur. Acct. Rev. 305, 310–11 (2008) (summarizing twenty-seven empirical studies examining the relationship between audit committee independence and financial reporting quality); Joseph V. Carcello et al., Audit Committee Financial Expertise, Competing Corporate Governance Mechanisms, and Earnings Management (Working paper, 2006), available at http://ssrn.com/abstract=887512 (finding that “independent audit committee members with financial expertise are most effective in mitigating earnings management”).

107 See, e.g., Ronald C. Anderson & John M. Bizjak, An Empirical Examination of the Role of the CEO and the Compensation Committee in Structuring Executive Pay, 27 J. Banking & Fin. 1323, 1332–36 (2003) (discussing compensation committee independence and CEO presence on the compensation committee as mechanisms for dealing with potential agency issues in setting CEO pay).

108 See, e.g., Michael S. Weisbach, Outside Directors and CEO Turnover, 20 J. Fin. Econ. 431, 453–54 (1988) (finding that firms with independent boards are more likely to remove the CEO on the basis of poor stock performance).

109We regard ClassBd, PPill, CumVote, and GP as takeover-related factors.

110 See Choi, Fisch & Kahan, supra note 71, at 649.

111 See Paul Gompers et al., Corporate Governance and Equity Prices, 118 Q.J. Econ. 107 (2003).

112 See id. (finding a relationship between equity prices and various corporate governance variables).

113Boards generally can adopt poison pills and golden parachutes without shareholder approval. See, e.g., Moran v. Household Int’l, Inc., 500 A.2d 1346, 1351 (Del. 1985) (holding that the board of directors has the power to adopt a poison pill under Delaware law); see also Unocal Corp. v. Mesa Petroleum Co., 493 A.2d 946, 953–54 (Del. 1985) (outlining a board’s broad powers to act unilaterally). In contrast, addition or removal of a classified board (otherwise known as a staggered board) or a cumulative voting structure typically requires both board and shareholder approval. See, e.g., Lucian Arye Bebchuk et al., The Powerful Antitakeover Force of Staggered Boards: Theory, Evidence, and Policy, 54 Stan. L. Rev. 887, 894 (2002) (“[D]ismantling [a staggered board] that is in the charter requires both a shareholder vote and a board vote . . . .”); Jeffrey N. Gordon, Institutions as Relational Investors: A New Look at Cumulative Voting, 94 Colum. L. Rev. 124, 161 (1994) (“[T]he elimination of cumulative voting in a specific firm ordinarily requires shareholder approval . . . .”).

114While the vote outcome for any director election is continuous, the vote outcome is bounded by zero and one. Estimating an ordinary least squares model on a bounded dependent variable results in biased coefficients. We employ a log odds transformation of the vote outcome to generate an unbounded, continuous variable allowing for ordinary least squares estimation of the relationship between the vote outcome and our independent variables of interest. We compute the log odds of the vote outcome as follows: For the dependent variable for a particular proxy advisor, we compute a term X = 0.5/n, where n is the number of data points where “for” vote data exists for directors with a recommendation from the particular proxy advisor in question. We then use the log((VOTE + X)/(1-VOTE + X)) as the dependent variable (to avoid division by zero problems when the “for” vote fraction is equal to 1).To control for the possibility that errors for directors in the same company may be correlated we use standard errors clustered by company in the models of Table 3. Unreported, we re-estimate the base model (Model 1 of Table 3) using non-clustered, robust standard errors and obtain similar qualitative results, except that both PPill and CumVote are now significant (at the <1% and 10% levels respectively), and Bot5AbRet x CEO is insignificant.

115These percentiles were calculated using the actual distribution of all independent variables except the variable for the ISS vote recommendation, which was set to zero (the baseline “for” recommendation).

116Research indicates, for example, that shareholders affiliated with the AFL-CIO may consider the interests of union workers when voting in director elections. See, e.g., Ashwini K. Agrawal, Corporate Governance Objectives of Labor Union Shareholders: Evidence from Proxy Voting 30 (N.Y.U. Stern Working Paper Series, No. FIN-08-006, 2009), available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1285084 (finding evidence that labor relations affect the voting behavior of some union shareholders). Considerations of corporate social responsibility may influence other shareholders. See Thomas W. Joo, Corporate Hierarchy and Racial Justice, 79 St. John’s L. Rev. 955, 956–57 (2005) (describing the potential role of shareholder power in increasing racial justice and social responsibility). Shareholders may care about the board’s position on current as well as previously submitted shareholder proposals. See Nathan Cummings Found., supra note 80, at 1 (indicating that the Foundation will vote for company nominees if, inter alia, “[t]he board does not recommend a vote AGAINST a shareholder proposal that the Foundation supports”).

117Choi, Fisch & Kahan, supra note 71, at 665.

118 Id. at 695.

119 Id. at 665.

120Only two factors that were significant for the ISS recommendation—status as non-executive chairman and golden parachutes, both of which reduced the likelihood of a withhold recommendation—are not significant in the vote regressions. Compare id. at 665, with id. at 671–72.

121 Id. at 671–72.

122This alignment is unlikely to be coincidental. Institutional Shareholder Services explicitly seeks shareholder input in formulating its voting policies, surveying institutional investors on a yearly basis. See RiskMetrics Group, 2009–2010 RiskMetrics Policy Survey 4 (2009) (describing how feedback from both institutional investors and issuers is part of RiskMetrics’s annual policy-formulation process).

123As a robustness test, we re-estimated the “for” vote outcomes for Proxy Governance (PGI), Glass Lewis (GL), and Egan Jones (EJ) recommendations using a Tobit model. Unreported, the coefficients on VotePGI (-0.025), VoteGL (-0.052), and VoteEJ (-0.037) are all significantly different from zero. Note that the upper bounds of influence for PGI (2.5 percentage points), GL (5.2 percentage points), and EJ (3.7 percentage points) are again smaller than for ISS. We also re-estimated the “for” vote outcomes for PGI, GL, and EJ recommendations using an OLS model with the untransformed “for” vote outcome as the dependent variable. Unreported, the coefficients on VotePGI (-0.023), VoteGL (-0.050), and VoteEJ (-0.035) are all significantly different from zero. Note that the upper bounds of influence for PGI (2.3 percentage points), GL (5.0 percentage points), and EJ (3.5 percentage points) are again smaller than for ISS.

124The most important factors affecting Egan Jones’s recommendation are attendance, membership on multiple boards, outside-linked status, and membership on the nominating committee. The most important factors for Proxy Governance’s recommendation are attendance, high compensation, membership on the compensation committee, and age. Choi, Fisch & Kahan, supra note 71, at 671–72.

125 Id. at 672.

126The four most important factors affecting the ISS recommendation were also among the five most important factors accounting for a withhold recommendation for Glass Lewis. Id. at 671–72. The fifth Glass Lewis factor significantly associated with a withhold recommendation, board interlock, was present only in 25% of the sample.

127 Id. at 671.

128 See infra tbl.5, Panel B.

129 GAO, Corporate Shareholder, supra note 1, at 9–10.

130 See Alexander, et al., supra note 65, at 8 (“The core business of ISS and other proxy advisors is to supply institutional investors with vote recommendations on a subscription basis.”). Institutional Shareholder Services’s recommendations and reports are now available on LEXIS and are also available on Westlaw through a premium subscription.

131 See, e.g., Press Release, Starboard Value and Opportunity Master Fund Ltd. and Ramius Capital Group LLC, Institutional Shareholder Services (ISS), A Leading Independent Proxy Advisory Firm, Supports Ramius’ Independent Nominees for Election to the A. Schulman Board of Directors (Jan. 7, 2008), available at http://www.euroinvestor.co.uk/news/story.aspx?id=9692387&bw=20080107005892 (reporting that ISS and Glass Lewis supported the appointment of dissident nominees to the A. Schulman Board of Directors). Media reports typically are a response to a press release, and press releases are most common in contested elections.

132 See, e.g., Press Release, Arris, ISS and Glass Lewis Each Recommend Merger of ARRIS and C-COR (Dec. 6, 2007), available at http://www.arrisi.com/press_events/press_releases/pressdetail.asp?id=389 (reporting ISS and Glass Lewis recommendations in favor of proposed merger).

133“Insthold” is defined as the fraction of outstanding shares of the company in question in the hands of institutional investors, measured using Form 13-F data obtained from Thomson Financial for the time period immediately prior to the annual meeting date. “Indivhold” is defined as 1 – InstHold – Tot_Dir_Shs.

134 See discussion of potentially omitted variables supra Part III.A.

135 See discussion of discretionary broker voting in director elections, supra text accompanying notes 29–34.

136 See Allen, supra note 32.

137 See Marcel Kahan & Edward Rock, The Hanging Chads of Corporate Voting, 96 Geo. L.J. 1227, 1269 (2008) (“[B]rokers tend to vote in accordance with management recommendations . . . .”).

138 See Kahan & Rock, supra note 2, at 30 (discussing the shift in some brokers’ voting strategies). We are not aware of any information to suggest that brokers vote the shares of their clients for which they received no voting instructions in accordance with ISS recommendations. Charles Schwab policy dictates that it votes securities held in its customers’ brokerage accounts, for which it has not received voting instructions, in proportion to “all instructed shares held by Schwab.” N.Y. Stock Exch., Report and Recommendations of the Proxy Working Group to the New York Stock Exchange 16 (2006) (describing Schwab’s adoption of proportional voting in 2005).

139 See N.Y. Stock Exch., supra note 138, at 13 (describing the anticipated effect on uncontested director elections of eliminating broker discretionary voting).