Is Say On Pay All About Pay The Impact Of Firm Perfomance

IS SAY ON PAY ALL ABOUT PAY? THE IMPACT OF FIRM PERFORMANCE Abstract The Dodd-Frank Act of 2010 mandated a number of regulatory reforms including a requirement that large U.S. public companies provide their shareholders with the opportunity to cast a non-binding vote on executive compensation. The “say on pay” vote was designed to rein in excessive levels of executive compensation and to encourage boards to adopt compensation structures that tie executive pay more closely to performance. Although the literature is mixed, many studies question whether the statute has had the desired effect. Shareholders at most companies overwhelmingly approve the compensation packages, and pay levels continue to be high. Although a lack of shareholder support for executive compensation is relatively rare, say on pay votes at a number of companies have reflected low levels of shareholder support. A critical question is what factors drive a low say on pay vote. In other words, is say on pay only about pay? In this article, we examine that question by looking at the effect of three factors on voting outcomes -- pay level, sensitivity of pay to performance, and economic performance. Our key finding is the importance of economic performance to say on pay outcomes. Although pay-related variables affect the shareholder vote, even after we control for those variables, an issuer’s economic performance has a substantial effect and, perhaps most significantly, shareholders do not appear to care about executive compensation unless an issuer is performing badly. In other words, the say on pay vote is, to a large extent, say on performance. This finding has important implications. First, it raises questions about the federally-mandated shareholder voting right as a tool for concerns about executive compensation. Say on pay has limited effectiveness if it is only being used to discipline managers who are underperforming or alternatively is not a vote on outsize or inordinate pay as it was intended to be. Second, and more important, to the extent that the shareholder vote influences board behavior, granting shareholders another forum for signaling their dissatisfaction with a firm’s economic performance may be counterproductive. If investors are signaling concerns over near-term stock performance through their say on pay votes, they may be increasing director incentives to focus on short-term stock performance rather than firm value. I. Introduction One of the components of the Dodd-Frank Act of 2010 was to require publicly-traded U.S. issuers to provide their shareholders with a non-binding vote on executive compensation. The rationale for “say on pay” was that shareholder oversight would both reduce overall pay levels1 and encourage boards to tie executive pay more closely to firm performance. In other words, say on pay would increase director accountability. Issuers have now experienced five years’ of say on pay votes, and the effect of the provision remains heavily debated.2 Although shareholders at a few issuers have rejected compensation plans, shareholders at the overwhelming majority of issuers vote to approve executive compensation, and the average percentage of votes in favor exceeds 90%.3 The link between say on pay and CEO compensation is 1 See, e.g., John Carney, Why 'Say on Pay' Failed and Why That's a Good Thing, CNBC.COM (July 3, 2013), http://www.cnbc.com/id/100860959 (“From its very beginning, the "say on pay" movement was an attempt to reduce executive pay.”). 2 See, e.g., Jesse Eisinger, In Shareholder Say-on-Pay Votes, More Whispers Than Shouts, DEALBOOK, JUNE 26, 2013, https://dealbook.nytimes.com/2013/06/26/inshareholder-say-on-pay-votes-more-whispers-than-shouts/ (concluding that ““[t]he say on pay” experiment is a bust”). 3 A recent Semler Brossy Report indicates that, in 2016, the average vote result was 91% in favor. Semler Brossy, Say on Pay Reports, End of Year Report, at 2 (Feb. 1, 2017), 2 unclear – CEO pay continued to rise for the first several years after DoddFrank, declined in 2015 and, most recently, in 2016, rose to record levels.4 Despite the continued strong support for executive pay packages, some issuers have experienced low levels of support. Academic studies have reached inconsistent results about the effect of low say on pay votes but have generally failed to find conclusive evidence that issuers reduce executive pay packages in response to lower approval rates.5 Studies suggest, however, that issuers modified the structure of executive pay packages in response to the say on pay mandate. In particular current packages concentrate a greater component of pay in restricted stock and stock options. It is not clear, however, that this higher concentration of equity-based pay truly makes it performance-based6 or that the modifications are increasing shareholder value.7 The say on pay experiment may well be short-lived. President Donald Trump has vowed to repeal Dodd-Frank, which would eliminate the mandatory say on pay requirement, although issuers could voluntarily provide their shareholders with the right to vote on exeuctiive pay.8 The http://www.semlerbrossy.com/say-on-pay/2016-year-end-report-failure-rate-at-lowestlevel-since-2011/. 4 See Theo Francis & Joanna S. Lublin, It’s Good to Be a CEO, Again: Stocks Rise, and So Does Pay, WALL ST. J. (Mar. 20, 2017, 10:05 AM), https://www.wsj.com/articles/itsgood-to-be-a-ceo-again-stocks-rise-and-so-does-pay1489937848?mod=djemalertNEWS (describing median CEO pay for 2016 as “on track to set a postrecession record” and terming “the 2015 slowdown in chief executive pay [as] temporary”). 5 See, e.g., Christopher S. Armstrong, Ian D. Gow & David F. Larcker, The Efficacy of Shareholder Voting: Evidence from Equity Compensation Plans, 51 J. ACCTING RES. 909 (2013) (reporting finding “virtually no evidence that lower shareholder support for, or even the outright rejection of, proposed equity compensation plans leads to decreases in future CEO incentive compensation or firm-wide stock option grants”). 6 Id. (quoting Dieter Waizenegger, executive director of CtW Investment Group as saying that “Companies could still do more to link leaders’ pay with long-term corporate performance”). 7 See, e.g., David F. Larcker, Allan L. McCall, & Gaizka Ormazabal, Outsourcing Shareholder Voting to Proxy Advisory Firms, 58 J. L. & ECON. 173 (2015) (finding negative stock market reaction to changes to executive compensation programs adopted in response to proxy advisor guidelines and concluding that “the outsourcing of voting to proxy advisory firms appears to have the unintended economic consequence that boards of directors are induced to make choices that decrease shareholder value”). 8 Anders Melan & Caleb Melby, Investors Could Lose Influence Over Executive Pay, BLOOMBERG (Nov. 15, 2016, 2:00 AM), 3 Financial Choice Act, introduced by House Republicans in early 2017 would retain say on pay but limit its application to situations in which an issuer materially changes its executive compensation plan.9 As policymakers evaluate the decision whether to retain say on pay, it is worth examining more carefully the information that shareholders convey through their vote on executive compensation. This article examines that question. We analyze say on pay votes of S&P 1500 companies between 2010 and 2015. We find that, as might be predicted, both excess compensation and the pay-performance sensitivity affect the level of shareholder support for executive compensation packages. More surprisingly, however, we find that, even after controlling for these variables, a critical additional driver of low shareholder support for executive compensation packages is the issuer’s economic performance. Say on pay votes reflect, to a large degree, dissatisfaction with firm performance, and are not based solely on pay. We further test the role of Institutional Shareholder Services (ISS) voting recommendations. We identify two important results. First, as with voting outcomes, ISS recommendations are driven by an issuer’s economic performance, independent of pay-related variables. Second, we show that ISS’s evaluation of the CEO’s pay-performance sensitivity uses an ex post measure of sensitivity and, as such, appears to differ from shareholder preferences. Our findings have important implications. First they suggest that shareholder voting may be a poor tool to address public concerns about the size and structure of executive compensation. Because of the key role of economic performance in explaining say on pay voting outcomes, the say on pay vote operates as a signal of investor dissatisfaction with executive pay primarily in poorly-performing companies. To the extent that executive pay is too high or insufficiently tied to performance, so long as the issuer is performing well, these concerns will not lead investors to vote against the pay package.If say on pay then is about curbing excessive or inordinate compensation, it seems to be a rough and inadequate tool. https://www.bloomberg.com/news/articles/2016-11-15/say-on-pay-rules-seen-at-riskas-trump-takes-aim-at-dodd-frank. 9 Joseph A. Hall, Predictions on Dodd-Frank’s Executive Compensation Provisions, HAR. L. SCH. FORUM ON CORP. GOV. & FIN. REG. (Dec. 14, 2016), https://corpgov.law.harvard.edu/2016/12/14/predictions-on-dodd-franks-executivecompensation-provisions/. 4 Second, and perhaps more importantly, shareholder support for executive pay seems to be highly correlated with an issuer’s short term stock performance. Shareholders appear to care a lot about performance and, to an extent, they are using say on pay to punish executives for poor performance rather than excessive pay. As a result, the say on pay vote may be counterproductive to the extent that it heightens executives’ incentives to focus on short term stock price at the potential cost of working to enhance firm value. The article proceeds as follows. Part II provides a brief background on say on pay. Part III describes our empirical analysis. Part IV discusses the implications of our results for the debate over say on pay. II. Background and Effect of Say on Pay Public criticism of the size and structure of executive compensation packages at public companies increased dramatically in the 1990s.10 Institutional investors began focusing on executive compensation and seeking, through shareholder proposals and other means, to address pay practices that were viewed as problematic.11 Academics, most notably Lucian Bebchuk and Jesse Fried, argued that executive pay packages were the result of insider self-dealing rather than the product of a functioning market for executive services.12 10 See, e.g., John E. Core, Wayne Guay & David F. Larcker, The Power of the Pen and Executive Compensation (May 23, 2007) (unpublished manuscript at 2, 14, 16) (on file with author), http://ssrn.com/abstract=838347 (finding that the increase in compensation related articles from 1994 to 2002 was approximately 900% and that the percentage of those articles with a “negative tone” was 36% and 47% among major newspapers and magazines, respectively). 11 Fabrizio Ferri & David A. Maber, Say on Pay Votes and CEO Compensation: Evidence from the U.K., 17 REV. OF FIN. 527, 528 (2013). 12 LUCIAN BEBCHUK & JESSE FRIED, PAY WITHOUT PERFORMANCE: THE UNFULFILLED PROMISE OF EXECUTIVE COMPENSATION ix (2006) (“There is now recognition that many boards have employed compensation arrangements that do not serve shareholders’ interests.”). Some academics challenged this view. See., e.g., Steven N. Kaplan, CEO Pay and Corporate Governance in the U.S.: Perceptions, Facts, and Challenges, 25 J. App. Corp. Fin. 8 (2013) (citing evidence that CEO compensation is not excessive and that it is highly correlated with corporate performance). One of the authors of this article has argued that the increase in compensation was ironically linked to the requirement to disclose pay information by public issuers. See Steven Davidoff & Claire Hill, Limits of Disclosure, 36 SEATTLE U. L. REV. 599, 624 n.101 (2013). 5 The financial crisis of 2008 heightened concerns as the public learned that highly-paid executives of financial institutions13 had engaged in risky business strategies that, according to many commentators, contributed to financial instability.14 In response, Congress adopted, as part of the Dodd-Frank Act of 2010, a requirement that the Securities & Exchange Commission require that publicly-traded issuers provide their shareholders with the opportunity to cast a non-binding vote on executive compensation.15 The rules adopted by the SEC implementing say on pay, which were effective for shareholder meetings on or after January 21, 2011, provide shareholders with three seprate votes.16 Shareholders have the right to vote on executive compensation at least once every three years (say on pay). Shareholders have the right to vote on whether to have a say on pay vote yearly, biannually or triannually (say on frequency). And shareholders have the right to vote on executive severance packages (say on golden parachutes). The rules apply to issuers with more than $75 million in public equity float and to the compensation package of the issuer’s five most higly-compensated executive officers as identified in the proxy statement. The rules also provide for increased compensation disclosure to shareholders. Say on pay in the United States was modeled on the then-existing U.K say on pay procedures which commentators described as having desirable results.17 Since 2003, UK issuers have been required to provide 13 See, e.g., Louise Story & Eric Dash, Bankers Reaped Lavish Bonuses During Bailouts, N.Y. TIMES, July 30, 2009, at A1 (reporting that “Thousands of top traders and bankers on Wall Street were awarded huge bonuses and pay packages last year, even as their employers were battered by the financial crisis.”). 14 See, e.g., Board of Governors of the Federal Reserve System, Press Release, Oct. 22, 2009, https://www.federalreserve.gov/newsevents/press/bcreg/20091022a.htm (quoting Federal Reservce Chairman Ben S. Bernanke as stating that “Compensation practices at some banking organizations have led to misaligned incentives and excessive risk-taking, contributing to bank losses and financial instability”). See also Lucian A. Bebchuk & Jesse M. Fried, Paying for Long-Term Performance, 158 U. PA. L. REV. 1915, 1 (2010) (“The crisis of 2008–2009 has led to widespread recognition that pay arrange-ments that reward executives for short-term results can produce incentives to take excessive risks.”) 15 12 U.S.C. § 951 (2010). 16 Shareholder Approval of Executive Compensation and Golden Parachute Compensation, 76 Fed. Reg. 6012 (effective Feb. 2, 2011). 17 See, e.g., Ferri & Maber, supra note 10, at 530 (reporting that “UK investors perceived say on pay to be a value enhancing monitoring mechanism, and were successful in using 6 shareholders with a remuneration report, which is then approved by shareholder vote.18 Initially the shareholder vote was advisory, but in 2013, the UK amended its director remuneration rules binding for listed companies.19 Various forms of say on pay requirements have also been adopted in other jurisdictions.20 In its early years, the results of the say on pay votes do not appear to demonstrate widespread shareholder dissatisfaction with executive pay packages. Since 2011 when say on pay became mandatory, shareholders have approved executive pay at over 90% of companies every year.21 The percentage of issuers with a failed say on pay vote has never exceeded 3% and, in 2016, that number dropped to just 1.7%.22 Notably, despite frequent claims that institutional investors “blindly” follow the recommendations of proxy advisory firms with respect to say on pay votes, the voting results tell a somewhat different story. Although ISS, the largest proxy advisory firm, has recommended a negative vote at approximately 10-12% of Russell 3000 issuers per year, the percentage of Russell 3000 issuers receiving a negative vote is less than 3% annually. In addition, say on pay does not appear to have significantly reduced CEO compensation levels. Following the adoption of DoddFrank and the SEC’s implementation of the say on pay requirement, CEO say on pay votes to pressure firms to remove controversial pay practices and increase the sensitivity of pay to poor performance”). 18 Edward Greene & Cleary Gottleib, Binding Shareholder Say-on-Pay Vote in UK, HAR. L. SCH. FORUM ON CORP. GOV. & FIN. REG. (July 31, 2012), https://corpgov.law.harvard.edu/2012/07/31/binding-shareholder-say-on-pay-vote-inuk/. 19 SQUIRE PATTON BOGGS UK EXECUTIVE REMUNERATION RESOURCE CENTRE, http://www.globalcompensationinsights.com/uk-executive-remuneration-resourcecentre/#itemtwo (last visitied Nov. 28, 2016). 20 See Randall S. Thomas & Christoph Van der Elst, Say on Pay Around the World, 92 WASH. U. L. REV. 653 (2015) (describing adoption of say on pay votes around the world and evaluating their effects). Other studies have examined the effect of say on pay in multiple countries. See, e.g., Ricardo Correa & Ugur Lel, Say on pay laws, executive compensation, pay slice, and firm valuation around the world, 122 J. FIN ECON. 500 (2016) (finding that say on pay works in a study of 38 countries from 2001 through 2012). Because the composition and incentives of shareholders varies substantially around the world, it is difficult to extrapolate from the experiences in one country to the potential effects of similar legislation elsewhere. 21 Brossy, supra note 2, at 3. 22 Id. Of the remaining 98.3%, 75% passed with at least 90% support and 92 percent passed with at least 70 percent support. The average approval rate was 91%. 7 pay rose.23 CEO pay declined in 2015, but the latest statistics, from 2016, show CEO pay rebounding and rising to record levels.24 Studies suggest, however, that say on pay has influenced the structure of executive conpensation packages, finding that increases in the proportion of pay that is performance-based as well as a greater alignment between pay raises and total shareholder return.25 Although say on pay in the U.S. is relatively new, several academic studies have examined its effects on pay levels, compensation structures and firm value.26 In one early study, the authors looked at the first two years of experience with say on pay and found evidence that SOP votes are sensitive to firm risk, excessive CEO compensation, accounting quality and financial performance.27 They also found that boards react to SOP rejection votes by subsequently reducing the level of excessive compensation.28 23 Ira Kay, Did Say-on-Pay Reduce or “Compress” CEO Pay? (Mar. 27, 2017), https://corpgov.law.harvard.edu/2017/03/27/did-say-on-pay-reduce-or-compress-ceopay/ (reporting that “•Median S&P 500 CEO pay increased 27% for the 4 years after SOP implementation relative to the 3 years preceding SOP”). 24 Francis & Lublin, supra note 3. 25 See, e.g., Paul Hodgson, Surprise surprise: Say on Pay appears to be working, FORTUNE (July 8, 2015), http://fortune.com/2015/07/08/say-on-pay-ceos/ (reporting an incrase in the alignment between executive pay and total shareholder return as well as a reduction in perks, based on WSJ/Hay Group CEO pay survey). 26 A number of studies have analyzed the effect of voluntarily-adopted say on pay shareholder proposals. See, e.g., Vicente Cuñat, Mireia Gine, & Maria Guadalupe, Say pays! Shareholder voice and Firm Performance, 20 REV. FIN. 1799, https://academic.oup.com/rof/article/20/5/1799/1753294/Say-Pays-Shareholder-Voiceand-Firm-Performance. Because voluntarily-adopted governance reforms raise questions of selection effects that are difficult to separate from the effect of the reforms themselves, we do not view these studies as probative on the value of a legislativelyrequired say on pay vote. See, e.g., Stephen Choi, Jill E. Fisch, Marcel Kahan & Edward B. Rock, Does Majority Voting Improve Board Accountability?, 83 CHI. L. REV. 1119, 1148 (2016) (distinguishing between selection and causation in evaluating the effect of voluntarily-adopted majority voting reform). 27 Marinilka B. Kimbro & Danielle Xu, Shareholders Have a Say in Executive Compensation: Evidence from Say-on-Pay in the United States, 35 J. ACCTING & PUBLIC POLICY 19 (2016), http://www.sciencedirect.com/science/article/pii/S0278425415000654. 28 Id. at 36. 8 A second paper studied the effect of the say on pay rule on 2010 executive compensation.29 The authors found that management anticipated the effect of the new shareholder voting right by modifying compensation plans in advance of the vote, reducing compensation levels and making pay more performance-based. In addition, the authors documented clear relationships between the level of shareholder voting support and the characteristics of the pay plans, finding that sharheolders cast more votes against pay packages involving higher total compensation, large increases in compensation or a higher number of perks. Another paper looked at the role of peer groups in SOP votes between 2011 and 2013.30 The study found that firms that benchmark their executive pay against the pay of peer firms that experience a low SOP vote voluntarily reduce the compensation of their own executives. The changes are concentrated in firms that have higher levels of excess CEO compensation. The authors conclude that Say on Pay may “contribute to an alignment of pay practices among firms that compete with each other for managerial talent in the executive labor market.” Kimbro and Xu also look at the 2011 and 2012 say on pay votes. They find that “shareholders effectively identify firms with excessive and abnormal levels of CEO pay and expressed their dissatisfaction through SOP.”31 They further find that firms respond to shareholder dissatisfaction, as expressed through low say on pay votes, by reducing the subsequent growth of CEO pay. The results of a different, more recent study are less appealing. An examination of board responses to low say on pay voting outcomes reports that SOP does not appear to be improving compensation contracting.32 The study finds that “these votes are ineffective in reducing CEO excess compensation, which increases, on average, in the year following the low- 29 Steven Balsam, Jeff Boone, Harrison Liu, & Jennifer Yin, The impact of say-on-pay on executive compensation, 35 J. ACCTING & PUB. POLICY 162 (2016), http://www.sciencedirect.com/science/article/pii/S0278425415000915?via%3Dihub. 30 Diane K. Denis, Torsten Jochem & Anjana Rajamani, Compensation Benchmarking and the Peer Effects of Say on Pay (2017), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2909963. 31 Kimbro & Xu, supra note 26, at 37. 32 Kelly R. Brunarski, T. Colin Campbell & Yvette S. Harman, Evidence on the outcome of Say-On-Pay votes: How managers, directors, and shareholders respond, 30 J. CORP. FIN. 132 (2015), http://www.sciencedirect.com/science/article/pii/S092911991400159X. 9 support vote.”33 The study also finds that, although the firms make cosmetic changes to their R&D expenditures and dividned payouts, they show “no change in net cash flows, firm risk, or firm value”, suggesting that the responses are mere “window-dressing.”34 Further, the study finds that firms with overcompensated managers and strong SOP support decrease in value, suggesting that say on pay increases agency problems at these firms and “ultimately[] reduces shareholder wealth.”35 In understanding the effect of a low say on pay vote, it is critical to understand what the vote means. When shareholders demonstrate a low level of support for management’s compensation, what factors are driving that vote and, as a result, what information is being conveyed to the board? A recent article uses two laboratory experiments to examine the key drivers of shareholder say on pay voting.36 The authors found, in their experiments which involved MBA students who were instructed to behave as shareholders, that the subjects only reacted negatively to high CEO compensation when that compensation was linked to poor firm performance.37 Firm performance, rather than excess CEO compensation appeared to be the primary driver of the say on pay vote. Notably, the authors observed that approval rates for compensation plans when firm performance was above the mean “reflects the extent to which shareholders view CEO pay as justified, given firm performance.”38 The authors ground their analysis of the experimental results in both agency and prospect theory. They hypothesize that their subjects analyze CEO compensation less from the perspective of agency costs than from prospect theory. Terming this an “agency-normative assessment,” they posit that shareholders decide how to vote by determining whether they think “CEO pay [is] justified, given firm performance.”39 Their experimental results are consistent with this hypothesis. Although this article provides a valuable theoretical framework for understanding shareholder voting behavior, the experimental design limits the power of the findings. As the authors acknowledge, it is difficult to 33 Id. at 134. Id. 35 Id. 36 Ryan Krause, Kimberley A. Witler & Matthew Semadeni, Power to the Principals! An Experimental Look at Shareholder Say-on-Pay Voting, 57 ACAD. MGMT. J. 94 (2014). 37 Id. at 108. 38 Id. at 109 39 Id. at 100. 34 10 predict the extent to which MBA students in a laboratory setting will replicate the real world voting decisions of institutional investors. The authors observe that “analysis of actual vote outcomes, once data become available, will provide insight into whether the results of our study accurately reflect the circumstances surrounding say-on-pay votes.”40 This theoretical framework is starting point for our empirical analysis. We collect data to examine the question whether real world shareholders behave consistently with the results predicted by the Krause et al. experiment. Specifically, our experimental design allows us to test separately the effects of firm performance and the size and structure of executive compensation on the say on pay vote. We find, as detailed below, that although shareholder votes are sensitive to excess compensation, they are also highly sensitive to firm performance. Indeed, in the absence of poor economic performance, shareholders do not appear to care about excess executive compensation. If a company is doing badly, however, shareholders care significantly about executive pay. Our results further explain, in part, the substantial gap between ISS recommendations on say on pay and voting outcomes. We highlight the fact that ISS appears to evaluate pay/performance sensitivity very differently from investors. Specifically, ISS focuses on realized pay for performance rather than ex ante pay.41 For reasons that we discuss further below, it is unclear if this methodology is most appropriate for evaluating an executive compensation package. III. Empirical Analysis and Results A. Data Collection We collect data on executive compensation, the firm’s accounting and stock return performance, and shareholder meeting votes. We cover executive compensation plans effective during the time period from 2010 to 2015 which were the subject of say-on-pay votes at annual meetings from 2011 to 2016.42 For data on executive pay we use the ExecuComp 40 Id. at 110. See Yonca Ertimur, Fabrizio Ferri & David Oesch, Shareholder Votes and Proxy Advisors: Evidence from Say on Pay, 51 J. ACCT’G RES. 951, 958 (2013) (describing ISS approach of analyzing change in realized CEO pay from the prior year). 42 The SEC’s rules implementing say on pay were effective for shareholder meetings on or after January 21, 2011. See 17 C.F.R. § 240.14a-21 (“the registrant shall, for the first 41 11 database. For information on firm characteristics and accounting performance we use the Compustat database. For stock prices we use the CRSP database. Information on shareholder meetings and ISS recommendations are taken from the ISS database. We derive our sample by taking all domestically incorporated firms listed in CRSP during the time period 2010 to 2015. We then exclude firms for which full information is not available in the ExecuComp or ISS database. The final sample wherein we have all the data is 5,543 observations, consisting of 1,345 unique companies. B. Empirical Analysis Table I provides descriptive statistics on our sample. Table I: Descriptive Statistics Name Means Medians Standard Deviation Panel A: Company variables Excess compensation ($$MM) Pay-performance sensitivity ($$MM) Stock returns Return-on-Assets Ratio of debt to assets Total assets ($$Bn) R&D ($$MM) Missing R&D ISS $1,052.6 $1,071.8 0.198 0.060 0.213 $7.999 $173.83 0.545 0.110 $362.38 $241.67 0.166 0.054 0.189 $7.851 0 1 0 $4,492.62 $1,1137 0.362 0.075 0.194 $1.639 $784.1 0.498 0.313 Panel B: Voting outcomes Percent against (mean) Low Vote (<20% support) 0.091 0.136 0.389 0 0.129 0.343 annual or other meeting of shareholders on or after January 21, 2011 . . . include a separate resolution subject to shareholder advisory vote to approve the compensation of its named executive officers, as disclosed pursuant to Item 402 of Regulation S-K.”). The rules are backward-looking in that shareholders at a given annual meeting approve the compensation plan in place during the prior fiscal year. See 17 CFR § 229.402 (requiring disclosure with respect to the “last completed fiscal year”). 12 We focus in this research on two components of exeucutive pay – excess compensation aand pay-performance sensitivity.43 Excess compensation is a variable that is based on the work of Core, Guay and Larcker, who define excess compensation as the amount of compensation exceeding a predicted compensation level based on economic determinants such as “firm size, growth opportunities, stock return, accounting return, and industry controls.”44 In our sample the mean excess compensation is $1.052 million, but there is significant variation as illustrated by the high standard deviation of $4.492 million. The median compensation is $362.38 million meaning that excess compensation is skewed toward the higher end. Pay-performance sensitivity is a measurement of the Chief Executive Officer’s total pay-performance sensitivity. Our variable uses the methodology of Core and Guay, who create an options portfolio “using the precise characteristics of newly granted options and the average characteristics of previously granted unexercisable and exercisable options.”45 In our sample, the CEO’s pay performance sensitivity is also right-skewed with the average pay performance sensitivity being $1,071.8 million with a median of $241.67 and a standard deviation of $1.1137 million. Critically, this methodology measures CEO pay on an ex ante basis – expected pay for future performance.46 Stock returns is the stock return of the company for the relevant fiscal year. Average yearly stock return for the companies in our sample during the 2010-2015 time period was 19.8%. During this time period the average yearly market return for the S&P 500 was 18.41%. 43 The say on pay rules require issuers to disclose a variety of detailed information about the various components of executive compensation. The degree to which these components affect shareholder voting is unclear. At least one paper has concluded that “shareholders focus on the top-line remuneration figure when deciding how to vote [and that] None of the variables that capture the various aspects of the CEO’s remuneration package seems to influene the voting behavior of share holders significantly. Carsten Gerner-Beuerle & Tom Kirchmaier, Say on Pay: Do Shareholders Care?, FMG Discussion Paper DP71 (2016), https://ssrn.com/abstract=2720481. 44 The authors build upon a prior academic literature and define excess compensation as “the residual from an expected compensation model that controls for standard economic determinants.”). Core et. al, supra note 9, at 1, 4 & 11. 45 John E. Core & Wayne Guay, Estimating the Value of Employee Stock Option Portfolios and Their Sensitivities to Price and Volatility, 40 J. ACCT. RES. 613, 614 (2002). 46 Id. at 617 (using Black-Scholes to calculate option value at time of grant). 13 Return-on-Assets is the return on the assets of the company for a particular year as measured in the Compustat database. Return on Assets averaged 6% compared to the return on assets of the S&P 500 of 6.23% during the sample time period. Ratio of debt to assets for a company is the ratio of debt to assets measured as of year-end for each fiscal year in our sample. Total Assets is the natural logarithim of the total assets of the company as of year-end in each fiscal year in our sample. The size of the companies in our sample is large with an average total assets of $7.999 billion, which is not surprising given that ExecuComp provides CEO compensation data on S&P 1,600 companies. R&D is the ratio of R&D expenses to assets for the company. Missing R&D is a dummy variable which is 1 if information is missing for a given year for research and development figures, and 0 otherwise. ISS is a dummy variable which is 1 if ISS recommended against the proposal and 0 otherwise. In our sample ISS recommended against 11% of the proposals. In terms of voting outcomes, we define the percent vote against, calculated as against/(against+for). In our sample, the average percent vote against was 9.1% while the median percent vote against was 38.9%. We define low say on pay vote as a dummy variable equal to 1 if the percentage against vote is greater than or equal to 20%,47 and zero otherwise. Using this criteria, 13.6% of say on pay proposals had a low vaote received low say on pay votes. We note that a 20% against vote is lower than the legal standard for say on pay failure, which is a majority vote. We use this different methodology for three reasons. First, because the say on pay vote is advisory, the 50% threshold is purely symbolic, Second, very few say on pay votes receive a majority of votes against, which would greatly reduce the power of our empirical tests. Third, and perhaps most importantly, our paper is consistent with the literature finding that issuers view a vote of 47 A similar classification has been used by Diane DelGuercio, Laura Seery, and Tracie Woidtke, Do Boards Pay Attention When Instituional Investor Activistis “Just vote No”?, 90 J. FIN. ECON. 84, 89 (2008) (using withhold vote threshold of 20% as an indication of “substantial support” for a withhold campaign). See also Yonca Ertimur, Fabrizzio Ferri, and Volkan Muslu, Shareholder Activism and CEO Pay, 24 REV. FIN. STUD. 535 (2011); Kimbro & Xu, supra note 26, at 21 (terming firms with a say on pay rejection vote of more than 20% “high-dissent firms”). 14 20% against an issuer or issuer-sponsored proposal as significant48 and that such a level of dissent is substantially more likely to generate an issuer response.49 This is a lower number than the legal standard for a say on pay failure, which is a majority vote. We employ this different methodology since so few say on pay votes actually gather less than 50%, a mere 2.2% of votes in our sample. Instead, we follow the methodology of prior academic studies on this issue and characterize a high level of dissent as receiving less than 80% of the vote.50 As discussed in Part I, ISS recommendations are a significant driver of say on pay results. In an unreported regression we examine ISS’s role in negative say on pay votes. Receiving a negative recommendation from ISS generates a 337.1% increase to the average probability of a vote of greater than 20% against the say on pay proposal. Receiving a negative recommendation from ISS also generates a 285.96% increase in the probability of receiving a similar low vote. ISS recommendations are thus exceedingly important in the outcome of say on pay votes, although the extent to which these recommendations are reflective of investor sentiment on the issue or as a cause of investor sentiment is a question we discuss further below. In Table II we analyze the determinants of ISS recommendations with respect to say on pay votes. We run Probit regressions using as the dependent variable whether ISS has recommended for or against a particular say on pay proposal. Specifically, the dependent variable is set to one if ISS recommended against the say on pay proposal, and set to zero if it did not. Given that the dependent variable is binary we run a Probit regression rather than standard Ordinary Least Squares (OLS) regression. During our time period, ISS recommended a vote against 11% (610 out of 48 See Say on Pay Primer, GEORGESON 1 (July 28, 2017), http://www.georgeson.com/News/Say-on-Pay-Primer.pdf (“Thus, opposition votes of higher than 20-25 percent invites greater scrutiny by the advisory firms of a company’s compensation practices. The company is expected to engage in a shareholder outreach and likely make pay changes based on investor feedback.”). See also Stephen Choi, Jill Fisch & Marcel Kahan, Who Calls the Shots?, 3 HARV. BUS. L. REV. 35, 39 (2013) (defining “high withhold votes” as those in which more than 30% of votes are withheld from a director candidate in board elections). 49 See, e.g., Ferri & Maber, supra note 10, at 531 (reporting that a say on pay dissent of higher than 20% “results in boards implementing 75%-80% of shareholder requests to remove specific provisions.”). 50 See supra note __. 15 a total of 5,543) of the say on pay proposals (i.e., ISS recommended that shareholders vote no on the proposal). Table II: Determinants of ISS No Recommendation (Probit Regressions) Independent variables Regression coefficient 0.144∗∗∗ z‐statistic 0.177∗∗ 2.20 Stock returns ‐0.118∗∗∗ ‐7.46 Return‐on‐Assets ‐0.252∗∗∗ ‐3.72 0.033 1.39 Total assets ‐0.008∗∗ ‐2.01 R&D 0.176∗∗∗ 2.77 0.016 1.62 Constant ‐0.945∗∗∗ ‐5.37 Pseudo R2 0.122 Excess compensation Pay‐performance sensitivity Ratio of debt to assets Missing R&D 10.59 *** statistically significant at 1% level; ** statistically significant at 5% level; and * statistically significant at 10% level, respectively. Standard errors are clustered at the firm level. In our model, the coefficient on Excess compensation is positive and significant meaning that excess compensation – compensation levels beyond that predicted by the standard economic variables – is a significant driver of an ISS “no” recommendation. This is consistent with ISS’s claimed methodology for evaluating compensation packages in which it 16 evaluates compensation “relative both to market peers and to absolute shareholder returns.”51 The coefficient on Pay-performance sensitivity is also negative and significant. This result is somewhat surprising. ISS reports that its methodology is designed to capture the sensitivity of executive compensation to firm performance and that a “misalignment” between pay and performance is a substantial factor driving its say on pay recommendation.52 Our results suggest, however, that a higher sensitivity of pay to performance is more likely to produce a negative ISS recommendation. The reason for this result is likely a difference in methodology. ISS calculates pay-performance sensitivity on an ex post basis – that is, pay relative to realized performance.53 In contrast, the Core and Guay methodology is an ex ante measure – expected pay for future performance.54 It is unclear which method is more appropriate for measuring pay-performance sensitivity. On the one hand, the ex ante measure most accurately reflects the incentives created by the compensation package. To the extent that pay-performance sensitivity is designed to reduce agency costs by aligning managements’ interests with those of shareholders, ex ante sensitivity would seem to be the appropriate measure.55 On the other hand, realized compensation measures the actual money that the CEO gets to take home rather than merely a predicted value.56 Finally, the coefficients on Stock returns and Return-on-Assets are negative and statistically significant. This result means that firm economic 51 Carol Bowie, Steve Silberglied & Liz Williams, Evaluating Pay for Performance Alignment, INSTITUTIONAL SHAREHOLDER SERVICES 3 (Nov. 2014), https://www.issgovernance.com/file/publications/evaluatingpayforperformance.pdf. 52 See Bowie, et. al, supra note 50, at 3 (explaining that ISS methodology is “designed to identify outlier companies that have demonstrated significant misalignment between CEO pay and company performance over time”). 53 See Ertimur et. al, supra note 40, at 958 (explaining ISS methodology). 54 See supra note 44. 55 See, e.g., Martin J. Conyon, John E. Core & Wayne R. Guay, Are US CEOs Paid More than UK CEOs? Inferences From Risk-Adjusted Pay, 24 REV. FIN. STUD. 402, 415 (2011) (explaining the rationale for analyzing executive pay in terms the incentives it creates for CEO behavior). 56 See Kaplan, supra note 11, at 9 (explaining that realized pay is “A better measure for assessing pay for performance . . . because it is a better measure of the amount of money the CEO actually takes home in a given year—and, as such, it is more useful when considering whether CEOs are paid for performance”). 17 performance plays a significant role in driving ISS recommendations – issuers with poor economic performance are more likely to receive a negative recommendation with respect to the say on pay vote. Examining the control variables, we find that smaller companies and those with higher R&D expenses are more likely get a no recommendation from ISS. The latter is particularly troubling. Commentators have argued that shareholders that focus unduly on short term profitability may pressure executives to cut R&D, sacrificing long term growth in favor of short term profitability. Our findings imply that the ISS recommendation may contribute to this scenario. Finally, we find that a firm’s leverage ratio has no significant effect on the ISS recommendation. In sum, we find that both higher excess compensation and higher pay-performance sensitivity are correlated with an ISS no recommendation. We also show that, after controlling for these payrelated variables, economic performance remains a significant factor. We will assess the economic magnitude of these variables in Table IV after exploring their role on actual voting outcomes. Table III shows the correlation between the same economic variables and voting outcomes. We define the dependent variable %Against as votes for divided by the sum of votes for and votes against. Given that this variable is a fraction bounded between 0 and 1, we run a Tobit regression rather than the the standard Ordinary Least Squares (OLS) regression. During our time period we find a median value of 38.9% (2,156 out of a total of 5,543) of the say on pay proposals received a low vote. Table III: Determinants of Low Say-on-pay Vote Independent variables Regression coefficient 0.395∗∗∗ z‐statistic Pay‐performance sensitivity ‐0.310∗∗ ‐2.39 Stock returns ‐0.015∗∗∗ ‐5.29 Excess compensation 18 5.77 ‐0.118∗∗∗ ‐6.03 0. 004 0.52 0.002∗∗∗ 2.60 ‐0.417 ‐0.35 Missing R&D ‐0.006∗∗ ‐1.96 ISS Negative Recommendation 0.307∗∗∗ 43.61 Constant 0.047∗∗∗ 6.59 Pseudo R2 ‐0.831 Return‐on‐Assets Debt to assets Total assets R&D *** statistically significant at 1% level; ** statistically significant at 5% level; and * statistically significant at 10% level, respectively. Standard errors are clustered at the firm level. As with Table II, the coefficient on Excess compensation is again positive and significant meaning that higher levels of excess compensation are correlated with greater shareholder dissent (lower shareholder support). It should be noted that the coefficient in this table on Excess Compensation is 0.395 compared to 0.144 in Table II, indicating that excess compensation plays a more significant role in predicting low votes than ISS recommendations. The independent variable Pay-performance sensitivity is negative and significant at the five percent level. This means that contrary to the results in Table II, the probability of a low vote is lower when CEO has a greater pay-for-performance sensitivity. The results indicate that unlike ISS, investors are more likely to support pay packages in which the CEO’s pay is highly sensitive to performance on an ex ante basis, that is, when the pay package more closely aligns the CEO’s incentives with shareholder interests. As with Table II, stock returns and return on assets are significant at the 1% level. Again, this means that, after controlling for the size and structure of the pay package, shareholders are less likely to vote in favor 19 of executive compensation when the issuer has experienced poor economic performance. The results with respect to two additional variables are worth noting. Total assets is significant and positive meaning that shareholders are less likely to vote against pay packages at large issuers. The result is interesting both because pay levels tend to increase with issuer size and because the percentage of institutional ownership is also correlated with issuer size. To the extent that institutional investors are more likely to be critical of executive pay, as some commentators have observed, this result is in tension with that hypothesis. Consistent with our preceding analysis, the effect of a negative ISS recommendation is positive and significant meaning that ISS recommendatons have a substantial influence on say-on-pay voting outcomes. Notably, and consistent with other research, we find an independent effect associated with the ISS recommendation. At the same time, the ISS recommendation does not fully explain the voting results. In unreported results, we run the regressions with and without the ISS control. Consistent with our reported results, the coefficients on the other variables are larger when we do not control for ISS but, as shown here, most of the variables retain significance even when we control for the ISS effect. Table IV shows the results of a regression in which we treat the say on pay vote as a binary rather than a continuous variable. We define a low say on pay vote as a resolution that receives shareholder support of less than 80% by creating a summary variable that is equal to one if the percentage of votes against is greater than or equal to 20%, and zero otherwise.57 Given that the dependent variable is binary we need to run a Probit regression rather than the standard OLS regression. Table IV: Determinants of Low Vote (Probit Regressions) Independent variables Excess compensation Pay‐performance sensitivity 57 Francis & Lublin, supra note 3. 20 Regression coefficient 0.585∗∗∗ z‐statistic ‐0.552 ‐1.44 2.88 Stock returns ‐0.038∗∗∗ ‐2.99 Return‐on‐Assets ‐0.281∗∗∗ ‐4.08 Debt to Assets ‐0.004 ‐0.13 Total Assets 0.003 1.03 R&D ‐0.469 ‐0.11 Missing R&D s ‐0.023∗∗ ‐2.09 ISS Negative Recommendation 0.391∗∗∗ 18.96 Constant ‐1.667∗∗∗ ‐10.44 Pseudo R2 0.501 *** statistically significant at 1% level; ** statistically significant at 5% level; and * statistically significant at 10% level, respectively. Standard errors are clustered at the firm level. In the Probit regression, the coefficient on Excess compensation is positive and significant at the 1% level, meaning that higher excess compensation is correlated with an increased likelihood of greater than 20% of the shares being voted against the compensation plan. However, the variable Pay-performance sensitivity is not significant, suggesting that greater pay/performance sensitivity does not affect the likelihood of a low say on pay vote. The coefficients on Stock returns and Return-on-Assets remain significant and negative meaning that issuers with better economic performance are less likely to experience a low say on pay vote. The variable ISS is again positive and significant – a “no” recommendation from ISS is correlated with a low vote. Again this highlights the influence of ISS. With the exception of the results on Pay-performance sensitivity, the results in Table IV largely mirror those in Table III. They highlight that while Excess compensation is an important driver of low say on pay votes, so is the performance of the company even when we control for 21 excess conpensation. The absence of a significant result on the variable Pay-performance sensitivity is interesting. This result suggests the possibility that, from a shareholder perspective, actual excess compensation is more important than the structure of such compensation. On this point, the results again demonstrate that, to some extent, ISS and shareholders appear to disagree somewhat in their analysis of when a pay package is problematic. In addition, they further explain the gap between the number of issuers that receive a negative ISS recommendation and those that receive a low say on pay vote from shareholders. Thus far we have not focused on the relative importance of CEO pay and economic performance with respect to voting outcomes. Table V addresses that question. We divide our sample into four quartiles based on the amount of CEO excess compensation.58 We then sort those quartiles based on the firm’s economic performance. The resulting 4x4 matrix shows the percentage of votes against for each combination of excess pay and economic performance, measured in terms of stock price. Table V provides the results Table V CEO pay=Excess pay % against Low High Q1 Q2 Q3 Q4 Returns High Q4 4.44% 4.78% 7.41% 11.40% Q3 4.57% 4.72% 7.55% 13.90% Q2 5.29% 6.66% 8.22% 15.86% Low Q1 7.47% 9.74% 12.32% 22.42% As shown in Table V, although shareholders are somewhat sensitive to excess CEO pay when stock price performance is strong, their reaction is limited. Even firms in the highest quartile of excess CEO pay receive only 11.4% of votes against their compensation package if they are in the top quartile in terms of stock price performance. By contrast, for firms with the same level of excess pay but that are in the lowest performance quartile, the level of negative votes almost doubles. 58 Based on the results described in Table IV, we focus here on excess compensation rather than pay/performance sensitivity. 22 Relatedly, poor stock price performance appears to result in greater shareholder dissatisfaction with executive pay packages even in the absence of excess compensation. In particular, for the firms in the lowest quartile with respect to excess compensation, overall levels of say on pay dissent are quite low. Nonetheless, the percentage of votes cast against the pay package increases by 41% as we move from the highest performing firms to the lowest performers. This increase appears to be driven, not by pay, but by stock price performance. The most compelling situation is the fact that, in our sample, we have 149 cases in which even though the CEO received no excess compensation, the percentage of shares voted against the compensation package exceeded 20%. Finally, our prior tables documented the importance of both economic performance and CEO pay with respect to ISS recommendations and say on pay voting outcomes. In Table VI, we quantify how much these factors matter. 23 Table VI: Economic Effect of Key Variables ISS Negative % Against Low Vote Recommendation (using regression (using regression (using regression estimates from estimates from estimates from Table II) Table III) Table IV) Excess Compensation 58.28% 19.47% 19.23% Pay-Performance Sensitivity 17.76%% -3.79% 0.00% Total Effect of Pay (Absolute Terms) 76.04% 23.26% 19.23% Stock Returns Return on Assets -38.62% -17.09 -5.80% -9.76% -10.08% -15.43% Total Effect of Performance (Absolute Terms) 55.72% 15.55% 25.51% Table VI shows the effect of an increase of one standard deviation in each independent variable – the two compensation variables and the two performance variables from Tables II and III. As we showed in Table I, the standard deviation of excess compensation is $4.492 million and the standard deviation of pay-performane sensitivity is $1.1137 million. Table VI then shows, in each column, the effect of increasing these variables by one standard deviation, respectively, the probability of a “no” recommendation from ISS, the percentage of votes cast against say on pay, and the probability of a low say on pay vote meaning less than 80% of the shares voted in favor of the compensation package. Column one shows the effect of CEO pay and firm performance on the probability of receiving a “no” recommendation on say on pay from ISS. A one standard deviation increase in Excess compensation, results in the average probability of an ISS no recommendation to increase by 58.28%.59 This result is consistent with what we might expect – ISS is substantially more likely to recommend against a pay package when excess compensation increases. Similarly, a one standard deviation increase in Pay-performance sensitivity, results in average probability of 59 This can be calculated as (.0000144 *4492.54/.111). All other probabilities are similarly calculated. 24 ISS no recommendation increasing by 17.76%. As noted earlier, this result is somewhat counterintuitive in that we would expect greater payperformance sensitivity to be a more desirable characteristic of a pay package and therefore increased sensitivity to be negatively correlated with the probability of a “no” recommendation. The sum of these two pay-related effects is 76.04%. This means that a one standard deviation in both Excess compensation and Payperformance sensitivity will increase the probability of an ISS no recommendation to 76.04%. This finding demonstrates the importance of pay-related factors in explaining the ISS recommendation. We next examine, in rows 3 and 4 of column one, the importance of the performance variables -- Stock returns and Return-on-Assets. Again we assess the effect of increasing firm performance by one standard deviation on the probability of receiving a negative say on pay vote recommendation from ISS. Importantly, we assess these effects holding the compensation variables reflected in rows 1 and 2 constant. A one standard deviation increase of .3623 in Stock returns is associated with a 38.72% decrease in the probability of a “no” recommendation from ISS. Similarly, a one standard deviation increase in Return-on-Assets, is associated with a 17.09% decrease in the probability of a “no” recommendation from ISS. Summing these two effects, we find that our performance variables yield a combined decrease in the probabily tof a “no” recommendation of 55.72%. In the second column we examine the effect of a one standard deviation change in our independent variables on the percentage of votes cast against the company’s executive compensation plan. Using the same metrics, we find that a one standard increase in Excess compensation is associated with, on average, a 19.47% increase in the percentage of shares voted against the compensation plan. A one standard deviation in Payperformance sensitivity is associated with a 3.79% decrease in the average percentage of votes against. The joint effect of 23.26% is significantly smaller than the effect of the pay variables on the ISS recommendation, consistent with the result, observed elsewhere, that shareholder do not necessarily vote against a compensation package on the basis of a negative recommendation from ISS. We also examine the effect of the performance variables. We find that a one standard deviation increase in Stock returns is associated with an average decrease in the percentage of votes against of 5.8%. A one standard deviation increase in Return-on-Assets results in a decrease in 25 the average percentage of votes against of 9.76%. This totals to a joint effect of 15.55%. Importantly, this effect is net of the pay variables, that is, for a given level of excess compensation and pay-performance sensitivity, a one-standard deviation increase in the performance variables alone will reduce the percentage of votes against the compensation plan by an average of 15.55% . In the third column we examine the extent to which pay and performance factors contribute to the probability of a low say on pay vote, defined, as indicated above, as fewer than 80% of shares voted in favor of the compensation package. We find that a one standard increase in Excess compensation results in the average probability of a low vote increasing by 19.23%. A one standard deviation change in Pay-performance sensitivity results in the average probability of a low vote changing by zero. We mark this change as 0 since the coefficient on Pay Performance Sensitivity in Table IV is not significant. Looking at the performance variables, we find that a one standard increase Stock Returns results in the probability of a low vote decreasing by an average of 10.08%. A one standard increase in Return on Assets results in the probability of a low vote decreasing by an average of 15.43%. The joint effect of the two performance variables is 25.51%. Again, this effect is net of or in addition to the effect of the pay variables. In summary, our results highlight the fact that the size and structure of executive compensation contributes to both the ISS recomenndation and the level of shareholder support for the executive compensation plan as reflected in the outcome of the say on pay result. Our results for the pay variables, apart from the effect of pay-performance sensitivity on the ISS recommendation, are not surprising. ISS should, we would expect, be more likely to issue a “no” recommendation if the CEO is receiving a high level of excess compensation. Similarly, shareholders should be less likely to vote to approve compensation plans that provide high levels of excess compensation or in which compensation is insufficiently sensitive to firm performance. On the other hand, the results of our analysis of the performance variables is dramatic and potentially troubling. We find that a substantial driver of both ISS recommendation and shareholder votes with respect to say on pay is the issuer’s economic performance and that, even if we control for pay size and structure, the effect of performance persists. In short, the say on pay vote, which purports to provide shareholders with a vehicle to express their views on the issuer’s compensation plan is, in fact, 26 at least partially, a referendum on firm performance. We explore the implications of this finding in the next section. IV. Implications Our findings have two important implications. First, the substantial role of performance in driving both ISS recommendations and shareholder say on pay votes suggests an important limitation in the utility of say on pay as a tool for addressing the size and structure of executive pay packages. Second, to the extent that say on pay voting results matter to corporate boards, the say on pay vote may problematically encourage excess focus on short term stock price and firm value metrics. A. The Role of Performance Limits the Effectiveness of Say on Pay We show that firm economic performance is a significant factor in both ISS recommendations and voting outcomes. In particular, firms with strong stock price performance do not experience significant levels of shareholder dissent, when their CEOs receive substantial excess compensation. At the same time, shareholders react negatively to unproblematic compensation packages at issuers that underperform. To the extent that shareholder voting is largely driven by economic performance, shareholders appear to be limiting their criticism of executive compenstaion primarily to companies that are suffering from poor economic performance. At issuers with strong stock price performance, the say on pay vote is not operating as a useful tool for identifying potential problems with executive compensation,60 including structural problems that may create risks for the sustainability of that performance. Although the rationale for a separate say on pay vote is to allow shareholders who are otherwise not critical of a firm’s performance or the board’s judgment to express their views on the size and structure of executive compensation in a nuanced way, it does not appear to be serving that function. 60 See also Audrey Bout, Brian Johnson and Steve DeMaria, Does Say on Pay Failure Affect Future Share Price Performance, PAY GOVERNANCE, http://paygovernance.com/does-a-say-on-pay-failure-affect-future-share-priceperformance/ (reporting that issuers with failed SOP votes typically underperform the market prior to the vote). 27 One possibility, of course, is that shareholders do not care about the size and structure of executive pay as long as the company is performing well. Executive pay at most issuers represents a small fraction of firm revenue, and shareholders may not view pay levels as economically important. Alternatively, shareholders may view even excessive levels of pay as well deserved if they are correlated with strong economic performance. These possibilities provide reasons why say on pay may be a poor tool for reform if the objective is to reduce high levels of executive compensation or to more closely align compensation with performance.61 This is a particularly important point since the legislative purpose of say on pay was to implement these objectives.62 At the same time, shareholders may be unduly critical of pay packages at issuers that have experienced poor economic performance, even when such pay packages do not appear problematic. For example, we find 149 instances of low shareholder votes (fewer than 80% support) at issuers in which excess compensation was less than or equal to zero. At these issuers, it can plausibly be argued that investors’ negative votes on pay were driven primarily by firm performance. Similarly, we find that the level of shareholder dissent for the firms with the worst quartile of economic performance and the lowest quartile of CEO excess compensation was comparable to that of better performing firms with higher levels of excess compensation. For at least some shareholders, the say on pay vote for the worst performers seems to have been based exclusively on performance rather than pay. Our data offers real world support for the hypothesis explored by Krause et al. in the laboratory – that shareholder voting has limited power 61 See, e.g., Jill E. Fisch, The Mess at Morgan, 83 U. CINN. L. REV. 651, 678 (2015) (noting high levels of shareholder support for executive compensation packages at JP Morgan and arguing that this support may have been due to the company’s continued strong economic performance despite its involvement in a variety of scandals). 62 See, e.g., Lisa M Fairfax, Sue on Pay: Say on Pay’s Impact on Directors’ Fiduciary Duties, 55 ARIX. L. REV. 1, 4 (2013) (“The push to mandate say on pay stems from a belief that it could help curtail inappropriate pay packages and practices, while holding directors more accountable for their compensation decisions.”); Marisa Anne Pagnattaro & Stephanie Greene, “Say on Pay”: The Movement to Reform Executive Compensation in the United States and European Union, 31 NW. J. INT'L L. & BUS. 593, 597 (2011) (explaining that say on pay had its place in Dodd-Frank “because of the widespread perception that executive pay practices contributed to the financial crisis”). 28 as a tool for addressing potential agency issues identified by Bebchuk and Fried. To a certain extent, this problem may be exacerbated by the ISS methodology for defining pay-performance sensitivity in terms of realized performance which does not precisely capture the extent to which pay is structured to align management incentives with shareholder value. At the same time, focusing on pay structure is of limited value if, in the absence of poor performance, shareholders do not fully respond to that focus. These findings are more problematic if say on pay is designed to reduce overall compensation levels consistent with broader societal objectives of equity or wealth distribution.63 To the extent that shareholder voting is driven primarily by economic performance, shareholder interests are likely to be imperfectly aligned with the interests of non-shareholder stakeholders. Thus, if Dodd-Frank was motivated by an effort to protect societal interests from excessive risk-taking motivated by high-powered compensation incentives or alternatively excess or inordinate pay alone, shareholder voting is unlikely to result in the appropriate compensation reforms. B. Say on Pay May Exacerbate Problematic Incentives Ourfindings also demonsrate empirically the risk that say on pay voting may exacerbate rather than eliminate problems with executive pay structure. We show that shareholder support for executive pay is highly correlated with an issuer’s short term stock performance. The performance variables in our analysis focus on the issuer’s economic performance in the one year prior to the say on pay vote. As we document, this one-year performance variable has a dramatic effect on voting outcomes. As a result, the say on pay vote may have the effect of increasing executives’ incentives to focus on short term stock price at the potential cost of working to enhance long term firm value. Many commentators have expressed a concern that both issuers and investors have adopted a short-term perspective with respect to strategic decisions.64 More significantly, they view short-termism as 63 See, e.g., THOMAS PIKETTY, CAPITAL IN THE TWENTY-FIRST CENTURY 20-21 (Arthur Goldhammer trans., Harvard University Press 2d ed. 2014) (discussing problems associated with wealth inequality); see also Fisch, supra note 60, at 653 (discussing say on pay within the concept of corporate “publicness”). 64 See, e.g., Jack B. Jacobs, “Patient Capital”: Can Delaware Corporate Law Help Revive It?, 68 WASH. & LEE L. REV. 1645, 1651 (2011) (identifying short-termism of 29 having an adverse effect on business decision-making.65 The rationale for a say on pay vote is to hold boards accountable but, to the extent that say on pay holds boards accountable for short term firm performance, it heightens these pressures and increases the incentive for boards to sacrifice long term value in favor of immediate economic performance.66 Ironically, say on pay may operate in direct contradiction to efforts to reform the structure of executive pay packages in order to create longer term incentives.67 A further possibility is that say on pay could contribute not merely to short-termism, but to excessive risk-taking because of the correlation between risk and stock market performance. The stock market has traditionally rewarded issuers for taking risk because diversified investors are able to bear that risk. Incentive-based compensation structures that rely on stock price, especially short term stock price, may lead exeuctiives to take excessive risk in an effort to maximize that short term stock price.68 corporate shareholders and other forces that are “causing corporate managements to govern for the short-term”); Leo E. Strine, Jr., The Dangers of Denial: The Need for a Clear-Eyed Understanding of the Power and Accountability Structure Established by the Delaware General Corporation Law, 50 WAKE FOREST L. REV. 761, 772 (2015). (arguing that activist shareholder “pressure may logically lead to strategies that sacrifice long-term performance for short-term shareholder wealth.”); but see J. B. Heaton, The ‘Long Term’ in Corporate Law, 72 BUS. LAW. 353, 354 (2017) (disputing the claim that there is a conflict “between short-term wealth maximization and longterm wealth maximization”). 65 See, e.g., Strine, Jr., supra note 63, at 772 (arguing that activist shareholder “pressure may logically lead to strategies that sacrifice long-term performance for short-term shareholder wealth.”); see also William H. Donaldson, Chairman, Sec. & Exch. Comm'n, Speech by SEC Chairman: 2005 CFA INSTITUTE ANNUAL CONFERENCE (May 8, 2005), http://www.sec.gov/news/speech/spch050805whd.htm (explaining that the “focus on short-term results has, I believe, had a counter-productive influence on companies, on investors and on analysts themselves”). 66 See Jacobs, supra note 63, at 1651-1652 (identifying the risk of incentivizing corporate executiives to manage to the market). 67 See, e.g., Roberta Romano & Sanjai Bhagat, Reforming Executive Compensation: Focusing and Committing to the Long-term, 26 YALE J. ON REG. 359 (2009) (proposing such reforms). 68 See. e.g., Wm. Gerard Sanders & Donald C. Hambrick, Swinging for the Fences: The Effects of CEO Stock Options on Company Risk Taking and Performance, 50 ACAD. MGMT. J. 1055, 1076 (2007) (reporting that CEOs take excessive risks to maximize payoffs from stock option incentive compensation). See also EXXONMOBIL, EXECUTIVE COMPENSATION OVERVIEW 8 (2013), www.exxonmobil.com/Corporate/Files/news_pub_ir_execcomp2013.pdf (concluding “that a formula based approach that relies heavily on one- or three-year total 30 The situation may be worse if the market does not fully understand or reflect the riskiness of an issuer’s strategic decisions.69 Existing approaches for calculating pay/performance sensitivity may exacerbate this effect. In firm reporting on pay/performance sensitivity, and in ISS’s evaluation of such sensitivity, the principal performance metric is total shareholder return (TSR).70 This means that stock price dominates both the analysis of pay sensitivity and firm performance.71 Critically, however, TSR focuses largely on the alignment between stock price and pay rather than on the creation of long term economic value. A recent IRRC study reports results consistent with our findings, documenting that investor fail to distinguish, in their say on pay votes, between issuers that create economic value and issuers that destroy economic value.72 In fact, the IRRC study finds that most issuers do not even disclose meaningful metrics that would allow investors to focus on the creation of economic value. V. Say on Pay – The Path Forward Our findings mitigate caution both about say on pay votes and say on pay itself. Our findings support the conclusion, however, that say on pay is a vote on more than just pay but also firm stock price performance. The focus on firm performance may produce excessive risktaking. Further analysis is needed to detail whether the say on pay vote is having these effects, Moreover, some have suggested that say on pay’s primary benefit has been as an outlet for shareholder dissatisfaction.73 Commentatators shareholder return could encourage inappropriate risk taking and have a lasting and negative impact on ExxonMobil’s business by encouraging a focus on more immediate results at the expense of our long-term business model”). 69 See, e.g., Henry Hu, Risk, Time and Fiduciary Principles in Corporate Investment, 38 UCLA L. REV. 277, 324 (1990) (warning of the incentive that incentive-based compensation can create for excessive risk-taking “when compensation is highly sensitive to perceived performance, and true, risk-adjusted performance is difficult to measure.”). 70 Mark Van Clieaf, Stephen O'Byrne & Karel Leeflang, The Alignment Gap between Creating Value, Performance Measurement, and Long-Term Incentive Design, IRRCI RESEARCH REPORT 8 (2014). 71 Id. 72 Id. at 4. 73 Troutman, Sanders, Say-on-Pay: Wrong Solution, Wrong Problem, July 9, 2010, https://www.troutmansanders.com/say-on-pay-wrong-solution-wrong-problem-06-09- 31 in industry and institutional investors detail, anecdotally, increased dialogue between issuers and investors about pay levels and pay structures.74 But from a “channel” perspective we believe that dissatisfaction with issuers is better expressed through a means which conveys the actual dissatisfaction. In addition, we believe the dialogue aspect of say on pay to be meaningful but secular in nature and coincident with the rise of shareholder power more generally.75 If say on pay were to be eliminated as a legal requirement we believe that this dialogue would continue. Our results also suggest that it would be productive for ISS to examine the metrics of its say on pay recommendations. While ISS recommendations are not dispositive, they substantially influence institutional shareholder votes.76 ISS’s use of realized pay for performance may, by incorporating an ex post facto measurement linked to stock performance, exacerbate these issues. The ISS approach also appears to differ from the manner in which shareholders evaluate pay performance sensitivity. Conclusion Say on pay is in its early years, and issuers and investors are still developing their approach to allowing shareholders to vote on executive compensation. We provide an analysis of the first five years of shareholder voting in an effort to determine the key factors that influence the say on 2010/ (observing, prior to the adoption of mandatory say on pay, that “the real problem is that shareholders have no practical ability to voice their disapproval regarding company performance – other than voting with their feet – and say-on-pay proposals have evolved as a proxy for a more fundamental complaint on company performance”). 74 See, e.g., Seymour Burchman & Blair Jones, Righting the Say On Pay Ship After a “No” Vote, Sept. 2013, http://www.semlerbrossy.com/wpcontent/uploads/ATD-Shareholder-Engagement.pdf (“One of the positive outcomes of the Say on Pay provision in the Dodd-Frank legislation has been more regular dialogue between companies and shareholders.”). 75 See generally Ronald J. Gilson & Jeffrey N. Gordon, The Agency Costs of Agency Capitalism: Activist Investors and the Revaluation of Governance Rights, 113 Colum. L. Rev. 862, 897 (2013) (explaining ability of activist shareholders to leverage their power through the support of more passive institutional investors). 76 See, e.g., Stephen Choi, Jill Fisch & Marcel Kahan, The Power of Proxy Advisors: Myth or Reality?, 59 EMORY L.J. 869, 906 (2010).(finding that, in uncontested director elections, proxy advisor recommendations drive 6-10% of the vote). 32 pay vote and, in particular, the relative role of compensation factors and firm performance. Our results confirm that both compensation and economic performance are key drivers of both ISS recommendations and voting results. Critically, however, we find that economic performance is an important factor even after controlling for excess compensation. In other words, we find that say on pay is not just about pay. For under-performing firms, say on pay appears to be a useful tool for disciplining mananagement. When firms perform well, however, shareholders do not seem to care about excess pay. We argue that these findings limit the potential value of the say on pay vote. In addition, the close connection between voting results and stock price and firm performance raises the risk that the say on pay vote may increase short-termism. When say on pay is not about pay, even an advisory vote can cause real economic harm. 33