Do Small and Large Shareholders Have a Say on Pay?

Transcription

1 Do Small and Large Shareholders Have a Say on Pay? Miriam Schwartz-Ziv and Russ Wermers 1 October 31, 2014 Abstract This paper investigates the voting patterns of shareholders on the recently enacted Say-On-Pay (SOP) for publicly traded corporations, and the efficacy of vote outcomes on rationalizing executive compensation. We find that small shareholders are more likely than large shareholders to use the non-binding SOP vote to govern their companies: small shareholders are more likely to vote for a more frequent annual SOP vote, and more likely to vote against SOP (i.e., to disapprove executive compensation). Further, we find that low support for management in the SOP vote is more likely to be followed by a decrease in excess compensation, and by a more reasonable selection of peer companies for determining compensation, when ownership is more concentrated. Hence, the nonbinding SOP vote offers a convenient mechanism for small shareholders to voice their opinions, yet, larger shareholders must be present to compel the Board to take action. Thus, diffuse shareholders are able to coordinate on the SOP vote to employ the threat that large shareholders represent to management. 1 Miriam Schwartz-Ziv is from the Eli Broad College of Business at Michigan State University, Russ Wermers is from the Smith School of Business at the University of Maryland. We thank Tim Adam, Jackie Cook, Charlie Hadlock, James McRitchie, David Stolin, and Yuehua Tang, and participants in the 2014 Mutual and Hedge Fund Conference, and Michigan State University seminar for helpful comments, and Corrine Carr, Sam Floyd and particularly Jinming Xue for research assistance. Miriam Schwartz-Ziv thanks the Women and Public Policy Program at Harvard Kennedy School and the Edmond J. Safra Center for Ethics at Harvard University for hosting her as a fellow during the initial stages of the preparation of this paper. 1

2 1. Introduction Beyond selling their stake in a company, shareholders have very few opportunities to take a stand, and to exhibit their disapproval of the management of their company. Recently, shareholders were given such an opportunity: Following extensive criticism concerning executive pay, beginning January 2011, companies listed in the United States with public free float exceeding $75 million are required to hold a non-binding shareholder Say-On-Pay (SOP) vote. 2 The formal goal of this vote is to allow shareholders to express whether they approve or disapprove of the compensation awarded to the named executive officers during the previous year. 3 The SOP vote is the only non-binding vote that the majority of U.S. companies hold every year, since it does not depend on a shareholder submitting a proposal to hold such a vote. 4 That is, the Securities and Exchange Commission, by adopting the SOP rule, provides a coordination device for shareholders who might, otherwise, not bother (due to the free-rider problem) to voice their opinion about executive pay. Accordingly, the SOP vote is the best opportunity shareholders have to provide some kind of indication, on a regular annual basis, on whether they are satisfied with management s performance. 5 We start by documenting that even after controlling for ISS s recommendation, shareholders are particularly likely to vote against SOP when compensation is large and excessive, and when 2 Smaller companies are also required to hold a SOP vote, but have until their annual meeting on or after January 21, 2013 to comply. 3 The named executives include the CEO, the CFO, and the three additional executive officers that receive the highest compensation. 4 Specifically, as Panel A of Table 1 documents, 57.9% of the companies with a market capitalization that exceeds 75 million USD now hold an annual SOP vote, while the remaining 42.1% of the companies hold a bi- or tri-annual SOP vote. 5 The only other issues that are raised routinely (i.e., annually), at shareholder meetings are whether to approve the slate of directors recommended by management, and whether to ratify the auditors. Since, in more than 99.7% of the cases, shareholders are presented with only one slate of directors and a single option of auditors, shareholders do not have obvious alternative options for which to vote with respect to these items. (Interestingly, the SEC has recently adopted Rule 14a-11, which gives shareholders the right to place nominees on the ballot for directors. However, the U.S. Court of Appeals for the DC Circuit vacated the rule.) However, for the SOP vote, shareholders have clearer options they can vote either for or against the compensation awarded. Accordingly, although the SOP vote is formally about the compensation awarded to management, it is the best opportunity shareholders have to express their broader views on management performance. 2

3 recent firm performance is poor. These two findings indicate that shareholders view the SOP vote as one that reflects both their views on the level of compensation of corporate executives as well as their general satisfaction about the company s performance. We continue by exploring how share ownership structure affects voting. We show that, on the aggregate level, companies that have a larger percentage of their shares held by blockholders that each hold at least 5% of the outstanding shares (henceforth, 5% blockholders), have shareholders that are more likely to vote to in support of SOP. Similarly, we find that the larger the fraction of shares held by 5% blockholders, the more likely that shareholders cast votes supporting a bi- or triannual SOP vote frequency, as opposed to an annual SOP frequency (henceforth, the frequency SOP vote ). Our interpretation of these results is that large shareholders prefer to confront management directly and in private, even when compensation is large and excessive. This approach may be driven by a desire to avoid a negative vote outcome, followed by a share price decrease since, presumably, large blockholders have established a toehold for a longer-term purpose. To understand how the structure of ownership affects the votes cast by individual shareholders, we examine the votes cast by mutual funds. We focus on voting by mutual funds for two reasons: First, we estimate, in our paper, that mutual fund voting amounts to 36% of all SOP votes cast, and that they represent, to a great extent, the votes cast by all institutions, 6 which cast 87.8% of all the votes cast. 7 And, second, we use mutual funds voting as a proxy for the aggregate votes of all institutions, since they are, currently, the only shareholders that are required to publicly report the votes they cast. 6 Management companies tend to vote consistently, see column 5 of Table 2. 7 The vast majority of votes are cast electronically, and these must be cast via Broadridge. Broadridge reports that 90% of institutional shares are voted, while 29% of retail shares are voted, and that institutions own, on average, 70% of the outstanding shares of the S&P 1500 companies. Hence, 87.8% (=(0.7*0.9)/(0.7* *0.29)) of all votes cast are cast by institutions (see Broadhttp://proxypulse.broadridge.com/#ownership). 3

4 In contrast to our results for all shareholders, we find that mutual funds are more likely to vote against the SOP vote when a large percentage of the shares is held by 5% blockholders. Put differently, mutual funds are particularly likely to express their dissatisfaction when ownership is concentrated. The fact that, on the aggregate level, the results are in the opposite direction, i.e., shareholders of companies with a concentrated ownership structure are more likely to vote in favor of management on the SOP, further reinforces the conclusion that large blockholders, who can have a dramatic effect on the aggregate level results because of their large holdings, vote in support of management on the SOP. Indeed, we find further evidence for this interpretation by examining how the magnitude of the shareholder s investment, in our case, the mutual fund s investment, affects its vote. We find that the larger the portfolio weight held in a given company, the more likely the fund is to vote in favor of management on the SOP. Put differently, the larger the magnitude of the holding from the shareholder s perspective, the less likely the shareholder is to confront management publicly via the SOP vote. We also examine the impact of the shares held by executives on the votes cast. We find that, on the aggregate level of shareholder votes, the larger the fraction of shares held by executives, the more likely shareholders are to vote in support of SOP, and to support a bi/tri annual frequency SOP vote, as opposed to an annual SOP. However, we find that mutual funds do not follow this pattern. In fact, we find some evidence that mutual funds are more likely to vote against management when a greater fraction of shares is held by management. This indicates that, at least from an institutional investor's perspective, there is a difference between tying compensation to performance, versus having already a large fraction of shares held by executives. Mutual fund s votes convey the perception that the latter does not improve the compensation packages that are currently awarded to the executives. 4

5 Finally, we examine the effect of ownership structure on the efficacy of the SOP vote in disciplining management. We find that particularly when ownership is concentrated in the year following a SOP vote that garnered low management support: (1) companies are likely to award less excessive compensation, and (2) companies pick peer companies that more closely resemble their own company, thereby decreasing the phenomena of artificially inflating executive pay. These findings indicate that, while blockholders are less likely to vote against management on SOP, the board appears to be especially anxious to quell any perceived shareholder dissatisfaction in the presence of blockholders, possibly to avoid a confrontation. In addition, we find evidence that suggests that CEO turnover is more likely to occur in companies that receive low support rates for the SOP vote when ownership is concentrated. In sum, our paper documents that small shareholders are especially likely to use the SOP vote as a means of governing management. We find evidence that companies respond to low support rates for the SOP vote, especially in the presence of large blockholders, by decreasing excessive compensation and selecting peers that more closely resemble their own company, thereby reducing compensation inflation. Hence, the non-binding SOP vote seems to offer an easy mechanism for small shareholders to voice their opinion, however, the SOP vote is likely to be associated with a change, such as a decrease in excess compensation, particularly if ownership is concentrated. 2. Literature Review Several studies have examined which factors impact upon the ultimate outcome of SOP proposals. Larcker, McCall, and Ormazabal (2012), Ertimur, Ferri, and Oesch (2012), and Thomas, Palmiter, and Cotter (2012), all document that proxy advisers, specifically, Institutional Shareholders Services (ISS) and, to a lesser extent, Glass, Lewis & Co., have a significant impact on the final outcomes. 5

6 These papers examine how the recommendations made by proxy advising firms impact aggregate shareholder voting patterns. Iliev and Lowry (2014) examine votes cast on the mutual fund level in the pre-sop era. They document that, in a significant number of cases, mutual funds cast votes which are inconsistent with ISS recommendations, which they define as active voting. They show that active voting is particularly likely to occur when mutual funds have a relatively large stake in the company. Dimmock et al. (2014) use a different approach for defining active voting voting against management. They show that mutual funds which have a high fraction of investors that are locked in the fund for the long term due to tax considerations, are more likely to be active voters, i.e., vote against management s recommendations. Davis and Kim (2007) and Ashraf, Jayaraman, and Ryan (2012) point to additional factors that may affect the voting patterns of mutual funds. They document that mutual funds are more manager friendly if they have pension fund business ties with a firm once the latter exists, mutual funds are more likely to vote with management. In similar spirit, Chou, Ng, and Wang (2011) show that funds with weaker governance are more likely to vote with management. Aggarwal, Erel, and Starks (2014) find that shareholder votes, including those cast by mutual funds, are significantly affected by media coverage. Matvos and Ostrovsky (2010) also examine votes cast on the mutual fund level, specifically, votes on the election of directors. They find that mutual funds tend to vote against the proposed directors when they expect other funds to do the same. Finally, Yafeh and Hamdani (2013) show that institutional investors take a special interest in compensation: they examine votes cast by Israeli institutional investors holding Israeli companies that mostly have a dominant shareholder. They find that institutional investors are particularly likely to vote against 6

7 compensation-related proposals, even when the minority shareholders do not have the mass required to fail the SOP proposal. One may wonder if the SOP vote is followed by a change that would indicate that shareholders are able to govern their companies. Cuñat, Gine and Guadalupe (2013) examine the impact of voluntary adoption of a SOP vote in the period. They find that voluntary adoption increases the market value and profitability of a company, but not the level and structure of pay. Similarly, Iliev and Vitanova (2013) find that, in the United States, markets reacted positively to the requirement to comply with SOP, yet this regulation did not affect the level or composition of average CEO pay. Ferri and Maber (2013) document that in the UK, following the submission of say on pay regulations, firms with excess CEO pay, and particularly those that exhibited poor performance, experienced positive abnormal returns. In addition, Ertimur, Ferri, and Muslu (2011) examine the period when the SOP vote was voluntary. In some companies during this period, shareholders initiated a proxy fight and demanded to hold a SOP vote. They find that shareholders are more likely to initiate a SOP vote if the total CEO compensation is high. However, they find that the voters are likely to oppose the compensation awarded only if CEO compensation is excessively high, but not when the predicted CEO compensation is high. The authors interpret these findings as demonstrating that shareholders are able to distinguish between cases in which CEOs deserve to be awarded a large compensation, versus the cases they do not. In sum, these studies indicate that holding the SOP vote does provide shareholders an opportunity to at least attempt to discipline corporate management at least to some extent. 3. Data 7

8 Companies report the aggregate vote outcomes for each proposal that came up for vote at a shareholders meeting in their 8-K filings. 8 Rarely, the vote results are published in a 10Q filing that is filed shortly after the meeting. We obtain these aggregate voting outcomes from the ISS Voting Analytics database. In addition, since 2003, mutual funds are required by the Securities and Exchange Commission (SEC) to report the votes they cast at shareholders meetings. Mutual funds report their votes in the N-PX form they submit annually to the SEC. 9 The N-PX form includes management s recommendation on how shareholders should vote for each item/issue (henceforth proposal or resolution ) on the proxy ballot. ISS has extracted the votes cast by each of the mutual funds and the corresponding management recommendations, and has compiled this data into the ISS voting analytics database that we use in this study. With respect to the SOP votes examined in this paper, the ISS voting analytics database includes the votes cast by 8,307 mutual funds that are offered by the 357 largest U.S. mutual fund management companies. Unfortunately, the ISS voting analytics dataset on mutual fund votes does not include conventional identifiers for mutual funds. However, ISS does provide links to the N-PX form (on SEC s Edgar archiving system) in which the votes are reported. Virtually all funds report in their N- PX filing a fund family CIK code and a mutual fund Seriesid identifier. 10 However, reporting a fund symbol/ticker in the N-PX filing is voluntary, and most mutual funds do not do so. To increase the number of funds for which we are able to obtain a ticker, using a fund s Seriesid, we search for a 8 For an example, see this link 9 See for an example. 10 The Seriesid identifier is assigned by the SEC, and uniquely identifies a mutual fund. To the best of our knowledge the Seriesid identifier is not included in any of the mutual fund databases commonly available to academics. 8

9 ticker in the company filings section of the Edgar database 11. This is the same procedure used by Matvos and Ostrovsky (2010), and Iliev and Lowry (2014). To further increase the number of mutual funds for which we are able to locate an identifier, we manually search several sources for a ticker that is associated with a fund that matches both the fund name and the fund family name as our source fund from the N-PX filing. We search for these fund tickers in the CRSP Survivorship Bias Free Mutual Fund Database, Thomson Reuters Database on mutual fund holdings (S12, available from WRDS), Factset, and the internet using the name of the fund. In some cases, we found a ticker using these approaches. After we locate as many fund tickers as possible, we match the ISS voting analytics dataset to external databases by first matching the tickers to WFICNs using the MFLINK table available from Wharton Research Data Services. Data on companies is obtained from CRSP and Compustat. Data on mutual funds is obtained from CRSP mutual funds and Thompson s-12. Data on executive compensation is obtained from Execucomp. Data on institutional shareholders is obtained from the Thompson Reuters 13-f database. Data on the percentage of shares held by 5% blockholders is obtained from GMI ratings. Data on the peer-companies each company selected to determine the compensation to be awarded to the named executives is obtained from ISS. This data is extracted from the DEF 14-A filing Descriptive Statistics 11 In Edgar, one may type a seriesid in the Fast Search box, which leads to the hyperlink List all Funds and Classes/Contracts which details the tickers of all funds branching from the fund family. 12 See p. 43 of the DEF 14-A of AAR Corp. for an example of peer-companies disclosure The compensation and peer groups disclosed in the proxy statement of a given year pertain to the compensation awarded to the executives in the prior year. 9

10 Starting January 21, 2011, companies listed in the United States with public free float exceeding $75 million are required to include the following two issues on their ballots: (1) The frequency SOP vote: the frequency shareholders wish to hold a SOP vote, i.e., every one, two, or three years, 13 and (2) a SOP vote: an advisory vote on whether shareholders approve or disapprove the compensation awarded to the named executives in the prior year. 14 Panel A of Table 1 documents, for the period, the aggregate vote outcomes regarding the frequency SOP vote, given management s recommendation. We note that ISS has a policy of always recommending that SOP votes be held annually. As mentioned, the ISS recommendation is of importance, since numerous studies have documented that the ISS recommendation has a significant impact upon the votes cast by institutional investors, including mutual funds. Perhaps not surprisingly, Panel A of Table 1 shows managers prefer that their compensation be discussed less frequently than ISS recommends, i.e., only 79.4% of the managers recommend an annual SOP vote. This panel also shows clear evidence that the management recommendation influences the votes cast by shareholders: In almost all cases in which management recommends that the SOP vote be held every three years (19.9% of all cases), the frequency SOP vote resulted in a three year frequency outcome (19.7% of all cases), and when management recommended an annual SOP vote (79.4% of all cases), in most cases (57.7% of all cases) shareholders voted to hold an annual SOP vote. Figure 1 documents how mutual funds govern their companies, compared to all shareholders on the aggregate level (the latter include the mutual funds). As this figure documents, mutual funds enforce more stringent governance, compared to other shareholders. In each of years 13 The frequency SOP vote must be held every 1-6 years. 14 Companies are also required to hold a golden parachute SOP vote if executives receive a golden parachute in connection to a merger or an acquisition. However, this is an irregular vote, which is not the main focus of our current paper. However, we will address this vote later in the paper. 10

11 , mutual funds were more likely to vote that the SOP vote be held every year, as opposed to every two or three years. 15 Panel B of Table 1 documents for the period, the ultimate results of the SOP vote, broken down according to ISS s recommendation (i.e., vote against or for the compensation awarded). Unsurprisingly, in all cases examined, management recommended that shareholders vote for the proposed compensation. Panel B of Table 1 documents that, when ISS recommended to vote against the proposed compensation, the SOP vote failed (i.e., received less than 50% support from shareholders) only 15.5% of the time. In contrast, when ISS recommended to vote for the SOP resolution, the SOP vote failed only 0.6% of the time. In other words, ISS has a large, but not completely deterministic impact on the SOP vote outcome. Figure 2 documents that mutual funds usually enforce slightly stricter governance practices, compared to all shareholders: In each of the years , compared to all shareholders, mutual funds were more likely to vote against the compensation awarded, however this was not the case in The figure also highlights that the actual objection rates of all shareholders and also of mutual funds are somewhere between the objection rates recommended by management and those recommended by ISS. Panel C of Table 1 highlights the potential impact that mutual funds may have on the outcome of votes. This panel documents that, on average (median), mutual funds own 27.5% (27.7%) of the shares of the companies that held a SOP vote during the period. These figures are calculated first on the company-year level by dividing the aggregate number of shares held by mutual funds by the total number of shares outstanding (both figures are obtained as documented in the Thompson s12 database), and then calculating the average/median across all 15 As mentioned, companies are required to hold the frequency SOP vote every 1-6 years, and the bottom panel of Figure 1 documents that most companies chose to hold the frequency SOP vote only in 2011, and did not repeat this vote in 2012 and

12 stock-years. Moreover, an average (median) of 36% (35.4%) of the votes cast on the share level were cast by mutual funds, consistent with the view that mutual funds are more active voters compared to other shareholders. Table 2 focuses on the votes cast at the fund management company level. Column 2 of Table 2 provides insight into the extent to which management companies do not follow ISS recommendations, and accordingly, most likely conduct their own monitoring. Column 3 of Table 2 reports, for the 20 management companies which have participated in largest number of SOP votes, the frequency they either: voted in support of a SOP proposal when ISS recommended to vote against a proposal, voted against a SOP proposal when ISS recommended to vote for the proposal, or voted that the SOP vote be held every two or three years, while ISS recommended that the vote be held every year. As Table 2 documents, some fund advisors, such as Dimensional Fund Advisors, never vote against ISS recommendations. In contrast, other advisors do so much more frequently (for example, Black Rock votes in the opposite direction of ISS recommendation 34% of the time). On average, in 11% of the cases, mutual funds voted in the opposite direction of ISS recommendation. Similarly, in Column 4 of Table 2, we examine the percentage of cases in which mutual funds voted against management recommendation. We find that some mutual funds always follow management s recommendation (e.g., Rydex Investments), while others oppose management quite frequently (e.g., Jackson National Asset Management voted in the opposite direction of management recommendation in 37% of the cases). On average, mutual funds were likely to vote in 15% of the cases against management s recommendation. Finally, Column 5 of Table 2 documents that in the vast majority of cases, all funds within a management company vote in the same direction. This suggests that the management company generally determines how the votes of all the funds within 12

13 the management company should be cast, consistent with economies-of-scale in analyzing the details of executive pay. Taken together, this section documents that management and ISS recommendations affect the SOP votes cast both by mutual funds, and by all shareholders. 5. Results 5.1. The SOP Vote and SOP Frequency Vote Outcome Table 3 starts by examining the factors that impact upon the fraction of votes cast against SOP, that is, the fraction opposing the prior-year compensation awarded to the CEO and the top five named executives, during the period. Not surprisingly, Regressions 1-2 show that, the larger the compensation awarded to the CEO (Regression 1), and to the named executives (Regression 2), the larger the fraction of votes cast against the compensation awarded. We continue by examining how awarding excessive compensation impacts upon the SOP vote. We follow Core, Guay, and Larcker (2008) to define excess compensation as: Residual (Compensation it ) = Compensation it Expected Compensation it. 16 The logic of this regression is that CEOs that are overpaid (have a positive residual) may be more likely to have shareholders who vote against approving the compensation awarded. We model compensation as a function of: fraction shares held by executives, market capitalization of equity, fraction shares held by blockholders, fraction shares held by institutions, number of institutions holding shares, CEO age, CEO tenure, ISS recommendation, and management recommendation at year t-1; and ROA during year t-1. Regressions 1-3 of Table 3 show the results of this model, using different metrics of executive compensation. First, shareholders are more likely to vote against management on SOP when total compensation is high, or when residual compensation is high. However, companies 16 Expected compensation is estimated by regressing the log of the total compensation awarded to the CEO/named executives on the: lagged ROA, market capitalization, age of CEO and tenure of CEO. 13

14 having strong operating profitability, i.e., those exhibiting highly positive ROA of company t-1, experience lower opposition to SOP, which suggests that shareholders view the SOP vote as an opportunity to express their general satisfaction with a company s performance, rather than the SOP vote being confined to the compensation awarded. Regression 3 confirms this, since residual compensation captures the effect of poor corporate performance (i.e., since predicted compensation is modeled on ROA, residual compensation captures the effect of compensation that exceeds that commensurate to lagged performance). Indeed, Regression 3 of Table 3 documents that shareholders are particularly likely to vote against SOP, not only when CEO compensation is large in absolute terms ( predicted CEO compensation t-1 ), but also when it is excessive relative to our model ( residual of CEO compensation t-1 ). Similar to the effect of lagged ROA, lagged fraction executive ownership also negatively predicts opposition to SOP, but its significance disappears once residual compensation is included in regression 3. Thus, residual compensation acts as a proxy for (poor) governance. We continue by exploring how shareholder structure impacts upon the votes cast. To do this, we include in Table 3 the variable Fraction of shares held by shareholders holding at least 5% of equity. This variable captures the aggregate percentage of shares that are held by all shareholders that each hold at least 5% of the company s outstanding shares. We find that, when ownership is more concentrated (i.e., large values for the variable Fraction shares held by shareholders holding at least 5% of equity ), the SOP vote is likely to garner larger support rates. In particular, the results document that if an additional 1% of the outstanding shares is held by 5% blockholders, the likelihood that the votes are in support of management increase by 0.03%. This means that if a company that previously had 21% of its outstanding shares held by 5% blockholders (21% is the average percentage held by 5% blockholders), now has none of its shares held by 5% blockholders (which is the case for 7.9% of the companies), we expect to observe an 14

15 increase of 0.64% (21*0.035%) in the frequency shareholder vote against the SOP proposal. Since during , on average only 9% of the votes cast were against SOP (see Figure 2), this change is equivalent to an increase in the likelihood that shareholders oppose SOP by approximately 7.8% (0.63%/9%). The flip side of this finding is that, when ownership is dispersed, shareholders are particularly likely to vote against SOP. This indicates that, when ownership is dispersed, the SOP vote is particularly beneficial in coordinating a large number of small shareholders, by offering them a relatively simple and cheap way to communicate their (dis-)satisfaction with the executive s compensation awarded and/or the company s performance. In contrast, when ownership is concentrated, the SOP vote may not be as crucial, since large shareholders probably have direct ongoing access to management, and are not confined to relying on the annual, publicly observable, SOP vote to govern their company. In fact, one interpretation is that blockholders prefer to negotiate privately with management, rather than to publicly air their grievances. Regressions 4-5 document a similar pattern with respect to the frequency SOP vote, that is, the vote on how frequent the SOP vote must be held. In companies that have a large percentage of their shares held by 5% blockholders, shareholders are significantly more likely to vote against an annual SOP vote (in other words, they are more likely to support a bi- or tri- annual SOP vote). The flip side of this finding is that, in companies with dispersed ownership, shareholder are likely to vote in support of an annual SOP vote, indicating that they wish to have an opportunity to frequently express their views concerning compensation and, perhaps, performance. Overall, our results of Table 3 indicate that, when ownership is dispersed, shareholders use the SOP votes to coordinate and provide feedback, and their votes are more likely to be opposed to management s wishes. Finally, we point out that Regression 3 documents that the larger the fraction of shares held by executives, the more likely the SOP vote is to pass, and the more likely shareholders are to 15

16 support a bi- or tri- annual SOP vote. This pattern may be explained by (at least) one of the following explanations: (1) Managers who have received stocks as part of their compensation package, or may even have bought them, vote in support of SOP. (2) Companies that have awarded in the past compensation packages that included stocks and options that have materialized to stocks, are companies that currently grant compensation packages which shareholders support more strongly. In the following section we will further examine these two interpretations Votes Cast by Mutual Funds In this section we examine the votes cast by mutual funds. This level of analysis allows us to examine which factors affect the votes cast, from the perspective of the shareholder-voter. Moreover, as mentioned above, a substantial percentage of the shares are held and cast by mutual funds, and mutual funds are a proxy for how all institutional investors vote (the latter are not required to disclose their votes). For this reason we have special interest in the votes cast by mutual funds. In Table 4, we repeat the analysis conducted in Table 3, but now the dependent variable equals 1 if an individual mutual fund voted against management on SOP, and zero otherwise. The results in Table 4 show that, similar to the aggregate level results (Table 3), mutual funds are likely to vote against SOP when compensation is large and excessive, and when performance is weak. In contrast to Table 3, in all regressions of Table 4 the sign of the coefficient for Fraction shares held by shareholders holding at least 5% of stocks is positive (and significant). This result is important, as it indicates that, from the perspective of the typical mutual fund shareholder which holds only a tiny fraction of a company s outstanding shares, (mutual funds hold on average 0.14% of a given company), having a large percentage of the stocks held by 5% blockholders increases the likelihood that shareholders publicly confront managers in the SOP vote 16

17 (Regressions 1-3), and also vote to have an annual SOP vote (Regressions 4-5). Thus, institutional shareholders are more likely to vote against management when large blockholders are present, perhaps because these large blockholders represent a credible threat to management. That is, large blockholders, who prefer to negotiate privately, are notified of the dissatisfaction of diffuse institutional shareholders and perhaps pressured to take action. The corresponding results on the aggregate level (Table 3) the larger the percentage of the 5% blockholders, the more likely that shareholders vote for the SOP, and for a bi- or triannual SOP frequency vote seem to be particularly driven by the votes cast by the large shareholders. It seems that the large shareholders do not wish to govern their companies via a public vote. We will further explore this interpretation in Table 5. As mentioned, Table 3 documents that on the aggregate level, the larger the fraction of shares held by executives, shareholders are significantly more likely to vote in support of SOP, and a bi/tri annual frequency SOP vote. Table 4 documents that mutual funds do not follow this pattern: The sign of the coefficient for fraction of shares held by executives shows that the larger the fraction of shares held by executives, the more likely mutual funds are to oppose the SOP vote and to support an annual SOP vote. These results are significant only in Regression 1 of Table 4. Nevertheless, these results highlight that from a shareholder s perspective, having already a large fraction of shares held by executives does not improve the compensation package awarded. This may imply that on the aggregate level of the votes (analyzed in Table 3), it is the executives that vote in support of SOP and a bi/tri- annual SOP frequency, and not the institutional shareholders. Table 4 also documents that fund characteristics affect the SOP vote: funds that have low returns ( fund 12 month return ) are particularly likely to vote in support of management on SOP. This finding indicates that, when a mutual fund is a larger shareholder, similar to a blockholder, the fund prefers to avoid confronting management publicly. 17

18 Also, the fact that management always recommends to vote in support of the SOP vote (and that is not always the case for other resolutions), allows a relatively clean identification of the impact of ISS recommendations on the voting patterns of mutual funds versus other investors. Comparing the votes on the aggregate level to those on the mutual fund level leads to the conclusion that ISS recommendations particularly impact the votes cast by mutual funds: On the aggregate level, a negative SOP recommendation issued by ISS increases the likelihood that shareholders vote against SOP by approximately 30% (see Regressions 1-3 of Table 3), whereas on the mutual fund level, this figure is found to be approximately 44% (see Regressions 1-3 of Table 4). 17 These figures indicate that non-mutual-fund shareholders are likely to follow the ISS recommendation on SOP in only about 23% of the cases. 18 Thus, mutual funds are particularly likely to follow ISS, consistent with Gilson and Gordon (2013), who argue that institutional investors have little incentive to monitor their portfolios. With respect to the frequency of the SOP vote, ISS recommends that all companies have an annual SOP vote (and ISS does not typically issue blanket recommendations on the issues it considers). This allows a relatively clean identification of the impact of management recommendations on the voting patterns of mutual funds versus the non-mutual funds. We find that management recommendations impact more strongly the non-mutual-fund shareholders. On the aggregate level of shareholders, a recommendation issued by management on the frequency SOP vote to hold an annual SOP vote increases the likelihood that shareholders support an annual SOP vote by approximately 28% (see Regressions 4-5 of Table 3). On the mutual fund level, such a management recommendation increases the likelihood that mutual funds vote in support of an annual SOP vote by approximately 12% (see Regressions 4-5 of Table 4). Hence, non-mutual fund 17 The latter magnitude is similar also when we exclude the mutual funds variables from the regression. 18 Since according to Panel A of Table 1, 36% of the votes are cast by mutual funds, if we define X as the percentage of non-mutual-fund votes following ISS recommendation, we may find a rough estimate for the latter: 36% 44% + 64% X = 30% X = 22.7%. 18

19 voters follow management recommendation to a substantially greater extent on the frequency of vote issue than mutual funds do 37% versus 12%, respectively. 19 We also find evidence that shareholders actively monitor their companies independently of ISS: shareholders were more likely to vote against the compensation awarded when companies held a golden parachute SOP vote in the same year. Such a vote would be held if a company awarded executives a golden parachute in connection to a merger. In Table A in the Appendix, we document that ISS SOP recommendations are not impacted significantly by a golden parachute SOP vote (this has also been confirmed to us by ISS). However, when shareholders vote for the SOP vote they seem to be concerned about golden parachute awards granted in connection to a merger, and they are significantly more likely to vote against the SOP if a company also held a golden parachute SOP vote in the same year. Next, we further investigate how the magnitude of the investment held by a shareholder, in our case, a mutual fund, affects voting behavior. Iliev and Lowry (2014) document that, the larger the fund s stake in a company, the more likely the mutual fund is to disagree with ISS, i.e., to vote with management similar to our results for large blockholders. In our analysis, in Table 5 we include three holdings variables that capture the magnitude of the fund s stake in a company: the percentage of the outstanding shares of a given company held by a given fund (equals on average 0.14%), the dollar value of equity in a given company held by a given fund (equals on average 9.4 million dollars), and the percentage the value of a company s stock represents of a fund s assets under management (equals on average 2.8%). Because these three holding variables are correlated, in Table 5 we repeat Regression 1 of Table 4, but include, in each specification, only one of the three holdings variables. Table 5 includes, but does not report, all variables included in Table 4. Regressions 1-3 of Table 5 examine how the 19 If we repeat the estimation we conducted above, and define X as the percentage of non-mutual-funds votes following management recommendation we may find that: 36% 12% + 64% X = 28% X = 37%. 19

20 percentage of outstanding equity held by a given fund affects the SOP vote it casts. Regression 1, which includes all observations (after winsorizing 1% of the top and bottom tails), does not document a significant relation between percent of company held and the SOP vote cast. We are especially interested in the voting pattern of mutual funds with large holdings in a given company, because their votes have a particularly large effect on the aggregate results. And, it is possible that the relation between ownership and voting is non-linear. For these reasons, we focus, in Regressions 2 and 3, on the votes cast by mutual funds that have relatively large holdings. Specifically, Regression 2, includes only the top 10% holding observations, as measured by the percent of company held, i.e. the observations that document a mutual fund holding in a given company that exceeds 0.4% of the outstanding stocks, and Regression 3 includes only the top 1% of the holdings, i.e., the observations that document a mutual fund holding in a given company that exceeds 1.4% of the outstanding stock. As in Regression 1, Regressions 2-3 of Table 5 do not document a significant relation between the holding variable and the SOP vote cast by the fund. Regressions 4-6 measure the magnitude of the fund s holding by the dollar value of the stocks. Regression 4 includes all observations, Regression 5 includes the top 10% holding observations, and Regression 6 includes only the top 1% holding observations, all measured by the value of the stocks (Table 5 specifies the average value of stocks included in each regression). Regressions 4-5 both document that the larger the dollar value of the stock held by the mutual fund, the less likely the fund is to vote against the compensation awarded. This finding is consistent with the findings presented above large shareholders do not wish to use the publicly observable SOP vote as an opportunity to govern management. Finally, in regressions 7-9 we follow Fich, Harford, and Tran (2014), who argue that funds are particularly likely to monitor companies in which they have invested a large percentage of the assets they manage (measured by the value of the stocks held in a given company/value of total 20

21 assets under management). Regression 7 includes all observations, Regression 8 the top 10% of the holdings, and Regression 9 the top 1% of the holdings, all measured by the percentage of the value of a given stock consists of all assets under management. Regressions 7-9 all document that the larger the percentage of the value of equity held in a given company of all assets under management, the less likely the fund is to vote against the SOP. Because the holding variable is significant in each of Regressions 7-9 of Table 5, this seems to imply that from a shareholder s perspective, the magnitude of his investment is perceived as the percentage of the value of a given stock of all assets under management Are Changes Observed Following the SOP Vote? In this section, we examine whether the SOP vote has a different effect, given the ownership structure. As will be detailed in this section, to examine the effect of SOP we use two variables: The first is the CEO residual compensation, as described in Section 5.1, while the second variable is the extent the company inflates the executive compensation by cherry picking peer companies that allow inflating compensation. In Regressions 1-3 of Table 6, we examine how the results of the SOP vote in the prior year affect the CEO s residual compensation. The logic of these regressions is that the larger the percentage of votes cast against SOP, the smaller the residual of the CEO compensation (i.e., excess compensation). In these regressions, we split the sample to three quintiles broken down by to the total percentage of outstanding equity held by 5% blockholders: Regression 1 includes only observations in which 5% blockholders hold less than 15% of the outstanding shares, in other words, the relatively dispersed companies. Each quintile has a similar number of observations. Regression 1 shows that when ownership is dispersed, CEO s residual compensation may even increase following low SOP support rates (the results are positive but not significant). In fact, when 21

22 we repeat Regression 1, and restrict the sample to observations in which 5% blockholders hold less than 7%, we find that low support for SOP significantly increase the likelihood that the CEO be awarded residual pay. This further points out to the conclusion that the SOP vote does not govern companies when ownership is extremely dispersed. In Regression 2 of Table 6 we includes only the observations in which blockholders hold 15%-30% of the outstanding shares, while in Regression 3 we include only observations in which the 5% blockholders hold more than 30% of the outstanding shares. Both of these regressions document that low support rates for the SOP vote significantly decrease CEO residual compensation. Put differently, both these regressions, which examine companies with relatively concentrated shareholder structure document that the residual compensation decreases following low support rates for the SOP vote. This indicates that the SOP vote can govern CEO pay, when shareholder structure is relatively concentrated. Our second approach for examining the impact of SOP is to examine how the peer companies selected by each company for benchmarking and determining the compensation of the named executives changes following a SOP vote that garners low support rates. Since 2006, the SEC requires companies to disclose which peer-companies they use to benchmark and determine the compensation of their named executives. Both Faulkender and Yang (2010) and Bizjak, Lemmon, and Ngujen (2011) find that companies generally choose peer-companies that have more highly paid CEOs. Moreover, Faulkender and Yang (2010) find that companies with more powerful CEOs are more likely to pick highly paid peer-groups. Perhaps the SOP vote offers an opportunity to govern such cherry-picking of peer-companies. A company (henceforth origin-company ) can inflate the compensation awarded to its CEO and/or the named executives, by cherry-picking a certain type of peer-companies (henceforth peer-companies ). For example, an origin-company can inflate executives compensation by picking 22

23 companies that have better performance, are larger, are in different (i.e., higher executive compensation) industries, etc. To capture the extent of this manipulation, we define two pay gap variables that capture the difference between the predicted peer-compensation versus the predicted origin-compensation: these are the (1) Difference between the predicted CEO compensation of peer and predicted CEO compensation of origin company, and the (2) Difference between the predicted compensation of named executives of peer and predicted compensation of named executives of origin companies. The larger each of these pay gap variables, the more the origin-company is cheating when selecting the peer-companies, since each pay gap variable represents abnormal compensation for the origin company, based on its profitability, stock-price performance, and industry, compared to its chosen peers. Table B of the Appendix documents that the larger each of the two pay gap variables (estimated in a set of first-stage regressions not reported), indicating increased cheating by origincompanies that pick peer-companies that have a large predicted compensation, the more likely shareholders are to vote against SOP. This provides evidence that our pay gap variables capture a phenomenon that is recognized and disliked by shareholders, and accordingly, increases the likelihood that they vote against SOP. In Regressions 4-6 of Table 6 we examine whether the extent companies game the selection of peer companies changes following low support rates for the SOP vote. Our dependent variable in these regressions is the average of the difference between predicted CEO compensation of peer and predicted CEO compensation of origin company for a given company in a given year, specifically, in the year following the SOP vote. While we focus on the first pay gap variable described above, the results are similar when we use the second pay gap variable. Regression 4, which includes the companies with the most diffuse ownership structure (i.e., less than 15% of the 23

24 stocks are held by 5% blockholders), documents that these companies are significantly more likely to inflate CEO compensation in the year following the vote by selecting unreasonable peers. In other words, when ownership is dispersed, low support rates for the SOP vote increase the likelihood that companies cherry-pick peer companies that allow inflating executive compensation. Hence, once again, in companies with dispersed ownership we do not find evidence that the SOP vote governs companies. In contrast, in regressions 5 and 6, which examine companies with a relatively concentrated ownership structure (15%-30% held by 5% blockholders in Regression 5, and above 30% held by 5% blockholders in Regression 6), we do not find that that low support rates for the SOP vote significantly increase the likelihood that CEO compensation be inflated by selecting unreasonable peer companies. Put differently, following a SOP vote that garners low support rates, companies that have dispersed ownership continue to inflate compensation, while those that have concentrated ownership do not continue do so. This provides further evidence that the SOP vote can be effective in governing companies that have a concentrated ownership structure, but not in those which have a dispersed ownership structure. Last, we address the question whether the SOP vote may be associated with a more dramatic change as CEO turnover. In Table C of the Appendix we examine whether companies that garnered low support rates for the SOP vote were more likely to replace their CEOs in the following year. Unfortunately, we currently only have observations for one year (the effect of the votes cast in 2001 on CEO turnover in 2012), and this may impede upon our ability to obtain significant results. The dependent variable in these regressions equals one if CEO turnover occurred in the year following the SOP vote. Regression 1 of Table A, which includes all observations, does not document a significant relation between the fraction of votes cast against SOP and CEO turnover in the following year. 24