1 Univ ersit ennsy lv ani a L aw Sc hool y of P eposit aw: L ars hi p R hol ory Penn L egal Sc Faculty Scholarship 2018 Is Say on Pay All About Pay? The Impact of Firm Performance Jill E. Fisch y of P en nsylvania L aw Sc hool Universit Darius Palia Rutger usi ness Sc hool s B Steven Davidoff Solomon Ber keley L aw Follow thi s and additional works at: https://s cholarship.law.upenn.edu/fa culty_scholarship ory Citation Reposit ay? The I Fisch, Jill E.; Palia, Darius; a nd D avidoff S irm Performance" (2018). mpact of F olomon, S ay All About P teven, "Is Say on P Faculty Sc holarship . 1931. cholarship.law.upenn.edu/fa https://s culty_scholarship/1931 This Article i ree and ope n access by Penn L s brought to you for f aw: Legal Scholarship R eposit ory. It has be en accepted for inclusion in F aculty eposit aw: Legal Scholarship R mation, p e infor ory. For mor lease contact enn L strator of P ized admini y an author Scholarship b aw.upenn.edu . [email protected]
2 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 1 IS SAY ON PAY ALL ABOUT PAY? THE IMPACT OF FIRM PERFORMANCE F ISCH * J ILL P ALIA ** D ARIUS TEVEN D AVIDOFF S OLOMON *** S The Dodd-Frank Act of 2010 mandated a number of regulatory reforms including a requirement that large U.S. public issuers provide their sharehold- ers 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. Sharehold- ers at most issuers overwhelmingly approve the compensation packages, and pay levels continue to be high. Although a lack of shareholder support for executive compensation is rela- tively rare, say on pay votes at a number of issuers 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 relative to economic — 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. Perhaps most significantly — shareholders do not appear to care about executive compensation unless an is- suer 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 execu- tive compensation. Say on pay has limited effectiveness if it is only being used to discipline issuers that are underperforming, or if it is not being used as a vote on outsize or inordinate pay as it was intended to be. Second, to the extent that the shareholder vote influences board behavior, granting shareholders another fo- rum for signaling their dissatisfaction with a firm’s economic performance may be counterproductive. If shareholders are communicating 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 long- term firm value. I NTRODUCTION One of the components of the Dodd-Frank Act of 2010 (Dodd-Frank) was to require publicly-traded U.S. issuers to provide their shareholders with * Perry Golkin Professor of Law, University of Pennsylvania Law School, and Visiting Professor, University of California, Berkeley, School of Law. The authors wish to thank Fabrizio Ferri and participants at the Law and Economics seminar at Boston University School of Law for their helpful comments. Thanks also to Samantha Vega for research assistance. ** Thomas A. Renyi Endowed Chair in Banking, Rutgers Business School, and Senior Fellow, Center for Contract and Economic Organization, Columbia University School of Law. *** Professor, University of California, Berkeley, School of Law.
3 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 2 Harvard Business Law Review [Vol. 8 102 1 a “say on pay.” a non-binding vote on executive compensation The ratio- — nale for say on pay was that shareholder oversight would both reduce overall pay levels and encourage boards to tie executive pay more closely to firm 2 In other words, say on pay would increase director performance. accountability. Issuers have now experienced five years of say on pay votes, and the 3 Although shareholders at a effect of the provision remains heavily debated. few issuers have rejected compensation plans, shareholders at the over- whelming majority of issuers vote to approve executive compensation, and 4 The link between say the average percentage of votes in favor exceeds 90%. on pay and Chief Executive Officer (CEO) compensation is unclear CEO — pay continued to rise for the first several years after Dodd-Frank, declined in 5 2015, and rose to record levels in 2016. 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 6 Studies suggest, however, pay packages in response to lower approval rates. that issuers modified the structure of executive pay packages in response to 7 In particular, current packages concentrate a the say on pay mandate. 8 It is not greater component of pay in restricted stock and stock options. clear, however, that this higher concentration of equity-based pay truly 1 Dodd-Frank Wall Street Reform and Consumer Protection Act, Pub. L. No. 111-203, See § 951, 124 Stat. 1376 (2010) (codified at 15 U.S.C. § 78n-1). 2 , John Carney, Why ‘Say on Pay’ Failed and Why That’s a Good Thing , See, e.g. 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.”). 3 See, e.g. , Jesse Eisinger, In Shareholder Say-on-Pay Votes, More Whispers Than Shouts, D EALBOOK 26, 2013, https://dealbook.nytimes.com/2013/06/26/in-shareholder-say-on- , June pay-votes-more-whispers-than-shouts/ (concluding that “[t]he ‘say on pay’ experiment is a bust”). 4 A recent Semler Brossy report indicates that in 2016 the average vote result was 91% in favor. Semler Brossy, 2016 Say on Pay Results: End of Year Report , at 2, (Feb. 1, 2017), http:// www.semlerbrossy.com/wp-content/uploads/SBCG-2016-Year-End-Say-on-Pay-Report-02- 01-2017.pdf. 5 Theo Francis & Joanna S. Lublin, See It’s Good to Be a CEO, Again: Stocks Rise, and So Does Pay , W ALL T . J. (Mar. 20, 2017, 10:05 AM), https://www.wsj.com/articles/its-good-to- S be-a-ceo-again-stocks-rise-and-so-does-pay-1489937848?mod=djemalertNEWS (describing median CEO pay for 2016 as “on track to set a post[-]recession record” and terming “the 2015 slowdown in chief executive pay [as] temporary”). 6 , Christopher S. Armstrong, Ian D. Gow & David F. Larcker, See, e.g. The Efficacy of Shareholder Voting: Evidence from Equity Compensation Plans , 51 J. A CCT ES . 909 (2013) . R (reporting finding “virtually no evidence that lower shareholder support for, or even the out- right rejection of, proposed equity compensation plans leads to decreases in future CEO incen- tive compensation or firm-wide stock option grants”). 7 David F. Larcker, Allan L. McCall & Gaizka Ormazabal, Outsourcing Shareholder See Voting to Proxy Advisory Firms , 58 J. L. & E CON . 173 (2015) 8 See id. at 173 (finding that “a substantial number of firms change their compensation programs in the time period before the formal shareholder vote in a manner consistent with the features known to be favored by proxy advisory firms in an effort to avoid a negative” say on pay voting recommendation).
4 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 3 Is Say on Pay All About Pay? 103 2018] 9 makes it performance-based or that the modifications are increasing share- 10 holder value. 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 pro- 11 In contrast, vide their shareholders with the right to vote on executive pay. the Financial CHOICE Act, introduced by House Republicans in early 2017, would retain say on pay, but would only require it when an issuer materially 12 As policymakers consider changes its executive compensation plan. whether to retain the say on pay requirement, we examine the message that shareholders of S&P 1500 issuers are providing to issuers through say on pay voting. Through our analysis of say on pay votes cast between 2011 and 2016, we find that both excess compensation and pay-performance sensitivity af- fect the level of shareholder support for executive compensation packages. Surprisingly, however, we also find that, even after controlling for these variables, a critical additional driver of low shareholder support for execu- tive compensation packages is the issuer’s economic performance. Say on pay votes reflect, to a large degree, shareholder dissatisfaction with firm performance and are not based solely on pay. We also examined the influence of voting recommendations issued by Institutional Shareholder Services (ISS) the largest proxy advisory firm. — Through our examination, we identified two important results. First, as with voting outcomes, ISS’s 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 in some measure from shareholder preferences. Our findings have two 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 shareholder dissatisfaction with executive 9 See supra note 5 (quoting Dieter Waizenegger, Executive Director of R Francis & Lublin, CtW Investment Group, as saying that “[c]ompanies could still do more to link leaders’ pay with long-term corporate performance”). 10 , e.g. , Larcker, et al., supra note 7 (finding negative stock market reaction to changes R See to executive compensation programs adopted in response to proxy advisor guidelines and con- cluding 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”). 11 Anders Melin & Caleb Melby, Investors Could Lose Influence Over Executive Pay , B LOOMBERG (Nov. 15, 2016, 5:00 AM), https://www.bloomberg.com/news/articles/2016-11- 15/say-on-pay-rules-seen-at-risk-as-trump-takes-aim-at-dodd-frank. 12 Joseph A. Hall, Predictions on Dodd-Frank’s Executive Compensation Provisions , H ARV . L. S CH . F. ON C ORP . G OVERNANCE & F IN . R EG . (Dec. 14, 2016), https://corpgov.law.har vard.edu/2016/12/14/predictions-on-dodd-franks-executive-compensation-provisions/.
5 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 4 Harvard Business Law Review 104 [Vol. 8 pay primarily in poorly performing firms. To the extent that executive pay is too high or insufficiently tied to performance, these concerns will not lead shareholders to vote against the pay package as long as the issuer is perform- ing well. If say on pay is about curbing excessive or inordinate compensa- tion, it seems to be a rough and inadequate tool. Second, shareholder support for executive pay seems to be highly cor- related 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 for excessive pay. As a result, the say on pay vote may be counterproductive to the extent that it heightens an executive’s incentives to focus on short-term stock price at the potential cost of working to enhance firm value. This Article proceeds as follows: Part II provides a brief background on say on pay, Part III describes our empirical analysis, and Part IV discusses the implications of our results for the debate over say on pay. ACKGROUND AND FFECT OF S AY ON E AY P I. B Public criticism of the size and structure of executive compensation 13 By the early packages at public issuers increased dramatically in the 1990s. 2000s, institutional investors began focusing on executive compensation and seeking through shareholder proposals and other means to address pay — — 14 Academics, most notably Lu- practices that were viewed as problematic. cian Bebchuk and Jesse Fried, argued that executive pay packages were the result of insider self-dealing rather than the product of a functioning market 15 for executive services. The financial crisis of 2008 heightened concerns about executive com- pensation as the public learned that highly-paid executives of financial insti- 13 , John E. Core, Wayne Guay & David F. Larcker, The Power of the Pen and See, e.g. – Executive Compensation 14 15 (May 23, 2007) (unpublished manuscript) (on file with au- thor), http://ssrn.com/abstract=838347 (finding that the increase in compensation related arti- cles 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). 14 Fabrizio Ferri & David A. Maber, Say on Pay Votes and CEO Compensation: Evi- See dence from the UK , 17 R EV . F IN . 527 (2013). 15 L B EBCHUK & J ESSE F RIED , P AY W ITHOUT P ERFORMANCE : T UCIAN U NFULFILLED HE P E XECUTIVE C OMPENSATION ROMISE OF ix (2006) (“There is now recognition that many boards have employed compensation arrangements that do not serve shareholders’ interests.”). Some academics have challenged this view. See., e.g. , Steven N. Kaplan, CEO Pay and Cor- porate Governance in the U.S.: Perceptions, Facts, and Challenges , 25 J. A PPLIED C . F IN . ORP 8 (2013) (citing evidence that CEO compensation is not excessive and that it is highly corre- lated with corporate performance). One of the authors of this Article has argued that the in- crease in compensation was ironically linked to the requirement to disclose pay information by public issuers. See Steven M. Davidoff & Claire A. Hill, Limits of Disclosure , 36 S EATTLE U. L. R . 599, 624 n.101 (2013) (citing Kenneth R. Davis, Taking Stock Salary and Options EV — Too: The Looting Of Corporate America , 69 M D . L. R EV . 419, 445, 447 (2010)).
6 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 5 Is Say on Pay All About Pay? 105 2018] 16 tutions had engaged in risky business strategies that contributed to financial 17 In response, Congress instability, according to many commentators. — required the Securities and Exchange Commission through Dodd-Frank — (SEC) to compel publicly-traded issuers to provide their shareholders with 18 the opportunity to cast a non-binding vote on executive compensation. The rules adopted by the SEC implementing say on pay, which were effective for shareholder meetings on and after January 21, 2011, provide 19 Shareholders have: (1) the right to shareholders with three separate votes. vote on executive compensation at least once every three years (“say on 20 (2) the right to vote on whether to have a say on pay vote yearly, pay”); 21 and (3) the right to vote on biannually, or triennially (“say on frequency”); 22 The rules ap- executive severance packages (“say on golden parachutes”). ply to issuers with more than $75 million in public equity float and to the compensation packages of the issuer’s five most highly-compensated execu- 23 The rules also tive officers as identified in the issuer’s proxy statement. 24 provide for increased compensation disclosure to shareholders. Say on pay in the U.S. was modeled on the then-existing U.K. say on 25 pay procedures, which commentators described as having desirable results. Since 2003, U.K. issuers have been required to provide shareholders with a 26 Initially remuneration report, which is then approved by shareholder vote. 16 , Louise Story & Eric Dash, , See, e.g. Bankers Reaped Lavish Bonuses During Bailouts , July 30, 2009, at A1 (“Thousands of top traders and bankers on Wall Street were IMES N.Y. T awarded huge bonuses and pay packages last year, even as their employers were battered by the financial crisis.”). 17 See, e.g. , Press Release, Federal Reserve, Federal Reserve issues proposed guidance on incentive compensation (Oct. 22, 2009) (quoting Federal Reserve Chairman Ben S. Bernanke), https://www.federalreserve.gov/newsevents/press/bcreg/20091022a.htm (“Compensation prac- tices at some banking organizations have led to misaligned incentives and excessive risk-tak- See also ing, contributing to bank losses and financial instability.”). Lucian A. Bebchuk & Jesse M. Fried, Paying for Long-Term Performance , 158 U. P A EV . 1915, 1917 (2010) . L. R (“The crisis of 2008-2009 has led to widespread recognition that pay arrangements that reward executives for short-term results can produce incentives to take excessive risks.”). 18 See Dodd-Frank Wall Street Reform and Consumer Protection Act, supra note 1. R 19 Shareholder Approval of Executive Compensation and Golden Parachute Compensa- tion, 76 Fed. Reg. 6010 (Feb. 2, 2011) (to be codified at 17 C.F.R. pts. 229, 240 and 249). 20 See id. at 6013. 21 at 6017. See id. 22 – 25. See id. at 6024 23 SEC Small Entity Compliance Guide, Shareholder Approval of Executive Compen- See , sation and Golden Parachute Compensation U.S. S ECURITIES AND XCHANGE C OMMISSION E (Apr. 1, 2011), http://www.sec.gov/rules/final/2011/33-9178-secg.htm. 24 See supra note 19, at 6026 – 27. R 76 Fed. Reg., 25 See, e.g. , Ferri & Maber, supra note 14, at 530 (“UK investors perceived say on pay to R be a value enhancing monitoring mechanism and were successful in using say on pay votes to pressure firms to remove controversial pay practices and increase the sensitivity of pay to poor see also Say on Pay’s Bundling Problems , 99 K performance.”); Andrew Lund, ENTUCKY L. R . 119 (2010) (drawing on the U.K. experience on say on pay to argue that say on pay votes EV may be skewed due to shareholder hesitancy to punish executives at high-performing firms). 26 Edward F. Greene & Cleary Gottlieb Steen & Hamilton LLP, Binding Shareholder Say- on-Pay Vote in UK , H OVERNANCE ARV CH . F. ON C ORP . G . L. S & F IN . R EG . (July 31, 2012), https://corpgov.law.harvard.edu/2012/07/31/binding-shareholder-say-on-pay-vote-in-uk/.
7 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 6 Harvard Business Law Review [Vol. 8 106 the shareholder vote was advisory, but in 2013, the U.K. amended its direc- 27 Various tor remuneration rules to make the votes binding for listed firms. forms of say on pay requirements have also been adopted in other 28 jurisdictions. In its early years, the results of the say on pay votes do not appear to demonstrate widespread shareholder dissatisfaction with executive pay pack- ages. Since say on pay became mandatory in 2011, shareholders have ap- 29 The percentage of proved executive pay at over 90% of firms every year. issuers with a say on pay vote disapproving executive pay (a “negative say on pay vote”) has never exceeded 3%, and that number dropped to 1.7% in 30 2016. Notably, despite frequent claims that institutional investors blindly fol- low the recommendations of proxy advisory firms, the say on pay voting results tell a somewhat different story. Although ISS has recommended a negative vote for approximately 10% to 12% of Russell 3000 issuers per year, the percentage of those issuers receiving a negative vote is less than 31 3% annually. In addition, say on pay does not appear to have significantly reduced CEO compensation levels. Following the adoption of Dodd-Frank and the 32 In our SEC’s implementation of the say on pay requirement, CEO pay rose. sample, CEO pay rose steadily until 2015, then declined slightly for the year. Median CEO pay in 2010 was $4.215 million and rose to $5.519 million in 2014 before declining to $5.350 million in 2015. However, the latest statistics from 2016 show CEO pay rebounding and 33 Studies suggest, however, that say on pay has influ- rising to record levels. enced the structure of executive compensation packages, finding increases in 27 UK Executive Remuneration Resource Centre , C OMPENSATION Squire Patton Boggs, AND B G LOBAL I NSIGHTS ENEFITS LOG (Nov. 28, 2016) http://www.globalcompensationin- B sights.com/uk-executive-remuneration-resource-centre/#itemtwo. 28 See Randall S. Thomas & Christoph Van der Elst, Say on Pay Around the World , 92 W ASH EV . 653 (2015) (describing adoption of say on pay votes around the world and . U. L. R evaluating their effects). Other studies have examined the effect of say on pay in multiple countries. , Ricardo Correa & Ugur Lel, Say on pay laws, executive compensation, pay See, e.g. slice, and firm valuation around the world , 122 J. F E CON . 500 (2016) (finding that say on IN pay works internationally based on a study of 38 countries from 2001 through 2012). How- ever, because the composition and incentives of shareholders vary substantially around the world, it is difficult to extrapolate from the experiences in one country to the potential effects of similar legislation elsewhere. 29 Semler Brossy, supra R note 4, at 3. 30 Id . at 2. Of the remaining 98.3%, 75% passed with at least 90% support and 92% passed with at least 70% support. The average approval rate was 91%. 31 Id. 32 Ira Kay, Did Say-on-Pay Reduce or “Compress” CEO Pay? , H . L. S CH . F. ON ARV C ORP . G OVERNANCE & F IN . R EG . (Mar. 27, 2017), https://corpgov.law.harvard.edu/2017/03/ 27/did-say-on-pay-reduce-or-compress-ceo-pay/ (“Median S&P 500 CEO pay increased 27% for the 4 years after [say on pay] implementation relative to the 3 years preceding [say on pay].”). 33 Francis & Lublin, supra note 5, at 1. R
8 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 7 Is Say on Pay All About Pay? 2018] 107 the proportion of pay that is performance-based, as well as a greater align- 34 ment between pay raises and total shareholder return. Although say on pay in the U.S. is relatively new, several academic studies have examined its effects on pay levels, compensation structures, and 35 In one early study, Marinilka Kimbro and Danielle Xu looked firm value. at the first two years of say on pay and found evidence that say on pay votes are sensitive to firm risk, excessive CEO compensation, accounting quality, 36 They also found that boards react to negative and financial performance. 37 say on pay votes by subsequently reducing excessive compensation. A recent paper by Steven Balsam, Jeff Boone, Harrison Liu, and Jen- 38 In nifer Yin studied the say on pay effect on 2010 executive compensation. that paper, the authors found that management proactively anticipated the effect of say on pay by modifying compensation plans in advance of the vote, reducing compensation levels and making pay more performance- 39 In addition, the authors documented clear relationships between the based. 40 Specifically, level of shareholder support and pay plan characteristics. shareholders cast more votes against pay packages involving higher total compensation, large increases in compensation, or a higher number of 41 perks. In another paper, Diane Denis, Torsten Jochem, and Anjana Rajamani examined the influence of peer groups on say on pay votes between 2011 42 They found that firms that benchmark their executive pay against and 2013. the pay of peer firms that experience a low say on pay vote voluntarily re- 43 These changes to executive duce the compensation of their own executives. compensation are concentrated in firms that have higher levels of excess 34 , Paul Hodgson, Surprise surprise: Say on Pay appears to be working See, e.g. F ORTUNE , (July 8, 2015), http://fortune.com/2015/07/08/say-on-pay-ceos/ (reporting an increase 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). 35 A number of studies have analyzed the effect of voluntarily adopted say on pay share- holder proposals , Vicente Cu ̃ nat, Mireia Gin ́ e & Maria Guadalupe, Say Pays! Share- . See, e.g. holder Voice and Firm Performance , 20 R EV . F . 1799 (2015). Because voluntarily-adopted IN governance reforms raise questions about 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 ., Stephen J. Choi, Jill E. Fisch, Marcel Kahan a legislatively required say on pay vote. See, e.g , 83 U. Does Majority Voting Improve Board Accountability? & Edward B. Rock, C HI . L. R EV . 1119, 1148 (2016) (distinguishing between selection and causation in evaluating the effect of voluntarily adopted majority voting reform). 36 Shareholders have a say in executive compensa- Marinilka B. Kimbro & Danielle Xu, , 35 J. A tion: Evidence from say-on-pay in the United States . & P CCT . P OL ’ Y 19 (2016). UB 37 Id . at 36. 38 Steven Balsam, Jeff Boone, Harrison Liu & Jennifer Yin, The impact of say-on-pay on executive compensation , 35 J. A . & P CCT . P OL ’ Y 162 (2016). UB 39 Id. at 164. 40 at 188. Id. 41 Id. at 164. 42 Diane K. Denis, Torsten Jochem & Anjana Rajamani, Compensation Benchmarking and The Peer Effects of Say on Pay , 6, 8 – 9 (2017), https://ssrn.com/abstract=2909963. 43 Id.
9 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 8 Harvard Business Law Review [Vol. 8 108 44 CEO compensation. The authors concluded that say on pay may “contrib- ute to an alignment of pay practices among firms that compete with each 45 Kimbro and Xu other for managerial talent in the executive labor market.” 46 They found that “share- also looked at 2011 and 2012 say on pay votes. holders effectively identify firms with excessive and abnormal levels of 47 They CEO pay and expressed their dissatisfaction through [say on pay].” further found that firms responded to shareholder dissatisfaction, as ex- pressed through low say on pay votes, by reducing the subsequent growth of 48 CEO pay. The results of a more recent study are less promising. Kelly Brunarski, T. Colin Campbell, and Yvette Harman examined board responses to low say on pay voting outcomes and found that say on pay does not appear to be 49 They found that “these votes are in- improving compensation contracting. effective in reducing CEO excess compensation, which increases, on aver- 50 Brunarski, Campbell, and age, in the year following the low-support vote.” although the firms make cosmetic changes to their Harman also found that — research and development (R&D) expenditures and dividend payouts they — showed “no change in net cash flows, firm risk, or firm value,” suggesting 51 Further, the study found that the responses were mere “window-dressing.” that firms with overcompensated executives and strong say on pay support decreased in value, suggesting that say on pay increased agency problems at 52 these firms and “ultimately[ ] reduces shareholder wealth.” Research consisting of two recent laboratory experiments using Masters of Business Administration (MBA) students examined the key drivers of say 53 Ryan Krause, Kimberly Whitler, and Matthew Semadeni on pay voting. found that when MBA students were instructed to behave as shareholders, they only reacted negatively to high CEO compensation when that compen- 54 Firm performance, rather than sation was linked to poor firm performance. excess CEO compensation appeared to be the primary driver of the say on 55 Notably, the Krause, Whitler, and Semadeni observed that share- pay vote. holder approval rates “reflect[ ] the extent to which shareholders view CEO 44 Id. 45 Id. at 35. 46 supra Kimbro & Xu, R note 36, at 20. 47 Id. at 37. 48 Id. at 36. 49 Kelly R. Brunarski, T. Colin Campbell & Yvette S. Harman, Evidence on the outcome . IN . F ORP J. C of Say-On-Pay votes: How managers, directors, and shareholders respond , 30 132, 134, 147 (2015). 50 Id. at 134 (emphasis removed). 51 Id. 52 Id. 53 Ryan Krause, Kimberly A. Whitler & Matthew Semadeni, Power to the Principals! An CAD . M GMT . J . 94 (2014). A Experimental Look at Shareholder Say-on-Pay Voting , 57 54 Id. at 108. 55 See id . at 111.
10 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 9 Is Say on Pay All About Pay? 109 2018] 56 pay as justified, given firm performance.” The authors ground their analy- 57 They hy- sis of the experiment results in both agency and prospect theory. pothesize that their subjects analyze CEO compensation less from the 58 Terming this an perspective of agency costs than from prospect theory. “agency-normative assessment,” they hypothesize that shareholders decide how to vote by determining whether they think “CEO pay [is] justified, 59 The results of their experiment are consistent given firm performance.” 60 with this hypothesis. Although Krause, Whitler, and Semadeni’s analysis provides a valuable theoretical framework for understanding shareholder voting behavior, the experimental design limits the power of their findings. As they acknowl- edge, it is difficult to predict the extent to which MBA students in a labora- tory setting will replicate the real-world voting decisions of institutional 61 They observe that “analysis of actual vote outcomes, once data investors. become available, will provide insight into whether the results of our study 62 accurately reflect the circumstances surrounding say-on-pay votes.” Krause, Whitler, and Semadeni’s theoretical framework is the starting point for our empirical analysis. In understanding the effect of a say on pay vote in the real world, it is critical to understand what the vote represents. When shareholders demonstrate a low level of support for management compensation through their say on pay votes, the factors that drive their votes clarify the information that shareholders convey to the board. We col- lected data to examine this question and to determine whether real world shareholders behave consistently with the results predicted by the Krause, 63 Specifically, our empirical analysis al- Whitler, and Semadeni experiment. lowed us to separately test 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 com- pensation, 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 performing poorly, however, shareholders care significantly about executive pay. Our results further explain, in part, the substantial gap between voting outcomes and ISS recommendations on say on pay. We highlight the fact that ISS and shareholders appear to evaluate pay-performance sensitivity very differently. Specifically, ISS focuses on realized pay for performance 56 See id . at 108 – 09. This finding is in line also with the theories of Professor Lund. See Lund, supra note 25 (theorizing that shareholders would not want to discipline executives at R high performing firms through say on pay votes). 57 supra note 53, at 94. R Krause, et al. 58 Id. 59 Id. at 100. 60 Id. at 111. 61 See id. at 110. 62 Id. 63 See supra Part III.A.
11 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 10 Harvard Business Law Review [Vol. 8 110 64 rather than ex ante pay. For reasons that we discuss further below, it is unclear if this methodology is the most appropriate for evaluating executive compensation packages. MPIRICAL A NALYSIS AND R ESULTS II. E A. Data Collection We collected data on executive compensation, firm accounting and stock return performance, and shareholder meeting votes. We examined ex- ecutive compensation plans that were effective between 2010 and 2015 and that were the subjects of say on pay votes at annual meetings from 2011 to 65 For data on executive pay, we used the ExecuComp database. For 2016. information on firm characteristics and accounting performance, we used the Compustat database. For stock prices, we used the Center for Research in Security Prices (CRSP) database. Information on shareholder meetings and ISS recommendations was taken from the ISS database. We derived our sample by taking all domestically incorporated firms listed in CRSP between 2010 and 2015. We then excluded firms for which full information is not available in the ExecuComp or ISS database. The final sample represents all of the data, including 5,541 observations, consist- ing of 1,345 unique issuers. B. Empirical Analysis Table I below provides descriptive statistics on our sample. 64 See Shareholder Votes and Proxy Advi- Yonca Ertimur, Fabrizio Ferri & David Oesch, sors: Evidence from Say on Pay , 51 J. A CCT . R ES . 951, 958, 961 (2013) (describing ISS’s approach of analyzing change in realized CEO pay from the prior year). 65 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 (2017) (“[T]he registrant shall, for the first 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 C.F.R. § 229.402 (2017) (requiring disclosure with respect to the “last completed fiscal year”).
12 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 11 Is Say on Pay All About Pay? 111 2018] T I: D S TATISTICS ABLE ESCRIPTIVE Standard Name Means Medians Deviation Panel A: Company Variables $1,054.0 $362.75 $4,492.83 Excess Compensation ($thousands) Pay-Performance Sensitivity ($thousands) $1,071.9 $241.67 $1,1139 0.199 0.166 Stock Returns 0.362 Return on Assets 0.054 0.075 0.060 0.212 0.193 Ratio of Debt to Assets 0.189 $7.999 $7.851 $1.640 Logarithm of Total Assets ($Bn) R&D ($$MM) 0 $784.2 $173.89 Missing R&D 0.545 1 0.498 Delaware Incorporated 0.639 1 0.480 Dual Class Stock 0.047 0.211 0 0.983 3.386 Insider Ownership (%) 0 ISS Negative Recommendation 0.110 0 0.313 Panel B: Voting Outcomes 0.091 Fraction Against (Mean) 0.129 0.039 Low Vote (<80% Support) 0.136 0 0.343 Our research focuses on two components of executive pay: (1) excess 66 “Excess compensa- compensation and (2) pay-performance sensitivity. tion” 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 con- 67 In our sample, the mean excess compensation is $1.054 million, but trols.” there is significant variation as illustrated by the high standard deviation of 66 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] [n]one of the variables that capture the various aspects of the CEO’s remuneration package seems to influ- ence the voting behavior of shareholders significantly.” Carsten Gerner-Beuerle & Tom Kirch- Say on Pay: Do Shareholders Care? , FMG Discussion Paper DP751 (2016), https://ssrn maier, .com/abstract=2720481. 67 Core et. al, supra note 13, at 2 (defining excess compensation as “the residual from an R expected compensation model that controls for standard economic determinants”).
13 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 12 Harvard Business Law Review [Vol. 8 112 $4.493 million. The median excess compensation is $362.75 thousand, meaning that excess compensation is skewed toward the higher end. “Pay-performance sensitivity” is a measurement of the CEO’s total pay to 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 68 In our sample, the CEO’s pay- non-exercisable and exercisable options.” performance sensitivity is also right-skewed with the average pay-perform- ance sensitivity being $1.072 million with a median of $241.67 thousand and a standard deviation of $1.114 million. Critically, this methodology mea- 69 basis sures CEO pay on an ex ante expected pay for future performance. — “Stock returns” is the stock return of the company for the relevant fis- cal year. The average yearly stock return for the firms in our sample during the 2010 to 2015 time period was 19.9%. During this time period the aver- age yearly market return for the S&P 500 was 18.41%. “Return on assets” is the return on the assets of the company for a particular year as measured in the Compustat database. The return on assets we calculated averaged 6% compared to the 6.23% average of the S&P 500 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 logarithm of the total assets of the company as of year-end in each fiscal year in our sample. The size of the issuers in our sample is large with an average natural logarithm of total assets of $8 billion, which is not surprising given that ExecuComp provides CEO compensation data on S&P 1500 firms. “R&D” is the ratio of research and development expenses to company assets. “Missing R&D” is a dummy variable that is set to “1” if information is missing for a given year for research and development figures, and “0” otherwise. “Delaware Incorporated” is a dummy variable that is set to “1” if the firm is incorporated in Delaware and “0” otherwise. The percentage of com- panies incorporated in Delaware, is 63.9% which is consistent with other 70 “Dual Class Stock” is a dummy variable that is set to “1” if the studies. firm is coded in ISS as having a multiple class stock structure. These firms typically are controlled by insiders and thus may have different salary struc- tures. Additionally, due to their lack of control, minority shareholders at dual class companies may be more likely to want to express their opinions through a say on pay vote. “Insider Ownership %” is the percentage owner- 68 John E. Core & Wayne Guay, Estimating the Value of Employee Stock Option Portfolios and Their Sensitivities to Price and Volatility , 40 J. A CCT . R ES . 613, 614 (2002). 69 at 617 (using the Black-Scholes formula to calculate option value at time of grant). Id. 70 See Matthew D. Cain, Stephen McKeon & Steven Davidoff Solomon, Do takeover laws matter? Evidence from five decades of hostile takeovers , 124 J. F IN . E CON . 464, 482 (2017) (finding that, in a somewhat different sample, 54.8% of firms were incorporated in Delaware).
14 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 13 Is Say on Pay All About Pay? 113 2018] ship of the company held by officers and directors as recorded in the Com- pustat database. “ISS Negative Recommendation” is a dummy variable that is set to “1” if ISS recommended against the say on pay proposal, and “0” other- wise. In our sample, ISS recommended against 11% of the say on pay proposals. In terms of voting outcomes, we define the “fraction against,” calcu- 71 lated as votes against divided by the sum of all votes for and votes against. In our sample, the average percent vote against was 9.1% while the median percent vote against was 3.89%. We define “low say on pay vote” as a dummy variable equal to “1” if the percentage against vote is greater than or 72 and “0” otherwise. Using this criterion, 13.6% of say on pay equal to 20%, proposals 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 lower standard for three reasons. First, because the say on pay vote is advisory, the 50% thresh- old 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. Specifically, a mere 2.2% of votes in our sample actually received less than 50% of the vote. Third, and perhaps most importantly, our approach is consistent with that of existing literature on this topic that issuers view a — 73 vote of 20% against an issuer or issuer-sponsored proposal as significant, and that such a level of dissent is substantially more likely to generate an 74 We follow the methodology of prior academic studies by issuer response. characterizing a high level of dissent as receiving less than 80% of the 75 vote. As discussed in Part I, ISS recommendations are a significant driver of 76 Receiving a negative recommendation from ISS gener- say on pay results. 71 We excluded non-votes and abstentions from the denominator. 72 A similar approach has been used by others in this field of study. See, e.g. , Diane Del Guercio, Laura Seery & Tracie Woidtke, Do boards pay attention when institutional investor activists “just vote no”? , 90 J. F IN . E CON . 84, 89 (2008) (using withhold vote threshold of 20% as an indication of “substantial support” for a withhold campaign); Yonca Ertimur, Fabrizzio Ferri & Volkan Muslu, , 24 R Shareholder Activism and CEO Pay . F IN TUD . 535, 546 . S EV note 36, at 21 (terming firms with a say on pay rejection vote of supra (2011); Kimbro & Xu, R more than 20% “high-dissent firms”). 73 See EORGESON , S AY ON G AY P RIMER (July 28, 2017), http://www.georgeson.com/ P 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.”). Stephen Choi, Jill Fisch & Marcel Kahan, Who Calls the Shots? , 3 See also H ARV US . L. R EV . 35, 39 (2013) (defining “high withhold votes” as those in which more . B than 30% of votes are withheld from a director candidate in board elections). 74 See, e.g. supra note 14, at 531 (reporting that a say on pay dissent of R , Ferri & Maber, higher than 20% “results in boards implementing 75%-80% of shareholder requests to remove specific provisions”). 75 See e.g. , id. 76 See supra Part I.
15 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 12:57 Seq: 14 25-JUN-18 [Vol. 8 114 Harvard Business Law Review ates a 336.1% increase in 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 283.2% increase in the probability of receiving a similarly low vote. ISS recommendations are thus exceedingly important in the outcome of say on pay votes. We explore further below the extent to which these recommendations reflect shareholder sentiment on the issue and the causes of such sentiment. In Table II we analyze the determinants of ISS recommendations with respect to say on pay votes. We ran Probit regressions using as the depen- dent variable whether ISS has recommended for or against a particular say on pay proposal. Specifically, the dependent variable is set to “1” if ISS recommended against the say on pay proposal, and set to “0” if it did not. Given that the dependent variable is binary, we ran a Probit regression rather than a standard ordinary least squares (OLS) regression. During our time period, ISS recommended a vote against 11% (609 out of a total of 5,541) of the say on pay proposals (i.e., ISS recommended that shareholders vote “no” on the proposal). ABLE ETERMINANTS OF ISS “N O ” R ECOMMENDATION II: D T ROBIT EGRESSIONS ) (P R Regression Independent Variables Z-Statistic Coefficient -4 *** Excess Compensation (10 ) 0.141 10.48 -5 Pay-Performance Sensitivity (10 ) 0.135 1.60 *** -7.52 Stock Returns -0.117 *** -3.65 -0.245 Return on Assets * 1.67 Ratio of Debt to Assets 0.039 * -1.82 Total Assets -0.007 *** 2.81 0.000 R&D * 1.73 Missing R&D 0.017 ** 2.29 Delaware Incorporated 0.023 Dual Class Stock 0.016 0.79 Insider Ownership 0.001 1.03 *** -6.13 Constant -1.097 2 0.125 Pseudo R *** 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.
16 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 15 Is Say on Pay All About Pay? 115 2018] In our model, the coefficient on excess compensation is positive and statistically significant, meaning that excess compensation compensation — levels beyond that predicted by the standard economic variables is a large — driver of an ISS “no” recommendation. This is consistent with ISS’s stated methodology for evaluating compensation packages in which it evaluates compensation “relative both to market peers and to absolute shareholder 77 returns.” The coefficient on pay-performance sensitivity is positive, but not sta- tistically significant. This result is somewhat surprising. ISS reports that its methodology is designed to capture the sensitivity of executive compensa- tion to firm performance and that a “misalignment” between pay and per- 78 formance is a substantial factor driving its say on pay recommendations. Our results suggest, however, that pay-performance sensitivity is not a sig- nificant factor in negative ISS recommendations. The reason for this result is likely a difference between our methodology and that used by ISS. ISS cal- ex post basis that is, pay relative culates pay-performance sensitivity on an — 79 In contrast, the Core and Guay methodology that to realized performance. 80 It ex ante measure expected pay for future performance. we utilize is an — is unclear which method is more appropriate for measuring pay-performance ex ante measure most accurately reflects the sensitivity. On the one hand, the incentives created by the compensation package. To the extent that pay-per- formance sensitivity is designed to reduce agency costs by aligning the inter- ests of management and shareholders, ex ante sensitivity seems to be the 81 On the other hand, realized compensation measures appropriate measure. the actual money that CEOs take home (rather than a predicted value), which 82 Regardless, our may be more useful when assessing pay for performance. results do not indicate that the ex ante pay performance sensitivity measure is a factor in ISS “no” recommendations. Finally, the coefficients on stock returns and return on assets are nega- tive and statistically significant. This means that firm economic performance plays a substantial role in driving ISS recommendations issuers with poor — 77 Carol Bowie, Steve Silberglied & Liz Williams, Evaluating Pay for Performance Align- ment , I S’ HOLDER S ERVS . NSTITUTIONAL 3 (2014), https://www.issgovernance.com/file/publica- tions/evaluatingpayforperformance.pdf. 78 See id. (explaining that ISS methodology is “designed to identify outlier companies that have demonstrated significant misalignment between CEO pay and company performance over time”). 79 See supra note 64, at 958 (explaining ISS methodology). R Ertimur et. al, 80 See Core & Guay, supra note 68, at 614. R 81 , Martin J. Conyon, John E. Core & Wayne R. Guay, Are US CEOs Paid More See, e.g. than UK CEOs? Inferences From Risk-Adjusted Pay , 24 R EV . F IN . S TUD . 402, 415 (2011) (explaining the rationale for analyzing executive pay in terms of the incentives it creates for CEO behavior). 82 See Kaplan, supra note 15, at 9 (explaining that actual or realized pay is “a better R measure of the amount of money the CEO actually takes home in a given year,” and thus more useful when considering whether CEOs are paid for performance, because it includes the CEO’s salary, bonus, the value of the restricted stock granted and the value of the options exercised by the CEO that year).
17 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 16 Harvard Business Law Review [Vol. 8 116 economic performance are more likely to receive a negative recommenda- tion with respect to the say on pay vote. Examining the control variables, we find that smaller firms and firms with higher R&D expenses are more likely to get a “no” recommendation from ISS. Our finding on firms with higher R&D expenses is particularly troubling. Commentators have argued that shareholders who unduly focus 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 ISS recommendations 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 higher excess compensation is correlated with an ISS “no” recommendation. We also show that, after controlling for pay- related variables, economic performance remains a significant factor. We will assess the economic magnitude of these variables in Table VI after ex- ploring their influence on actual voting outcomes in Table III. Table III shows the correlation between the same economic variables and voting outcomes as Table II. We define the dependent variable “fraction against” as votes against executive compensation divided by the sum of votes for and votes against. Given that this variable is a fraction bounded between zero and one, we ran a Tobit regression rather than the standard OLS regression. We find that a median value of 3.89% (216 out of a total of 5,541) of the say on pay proposals received a “no” vote during the time period of our study. As with Table II, the coefficient on excess compensation is again posi- tive and statistically significant, meaning that higher levels of excess com- pensation are correlated with greater shareholder dissent (lower shareholder support). Notably, the excess compensation coefficient in this table is -5 -4 in Table II, , larger in magnitude than the coefficient of 0.141*10 0.401*10 indicating that the impact of excess compensation less on ISS recommenda- tions than on actual shareholder votes. The independent pay-performance sensitivity variable is negative and statistically significant at the 5% level. This means, that the probability of a “no” vote is lower when a CEO has a greater pay-performance sensitivity. In other words, unlike ISS which does not appear to take this issue into account, shareholders are more likely to support pay packages in which the ex ante basis, that is, CEO’s pay is highly sensitive to performance on an when the pay package more closely aligns the CEO’s incentives with share- holder interests. These results 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 in that shareholders, but not ISS, appear to be crediting issuers for ex ante pay sensitivity. As with Table II, stock returns and return on assets are negative and statistically significant at the 1% level. Again, this means that, after control- ling for the size and structure of the pay package, shareholders are less likely
18 \\jciprod01\productn\H\HLB\8-1\HLB102.txt Seq: 17 12:57 unknown 25-JUN-18 117 Is Say on Pay All About Pay? 2018] T ETERMINANTS OF F RACTION A GAINST ABLE III: D Regression Independent Variables Z-Statistic Coefficient -5 *** Excess Compensation (10 ) 0.401 6.30 -6 ** ) -0.279 Pay-Performance Sensitivity (10 -2.18 *** -4.90 -0.013 Stock Returns *** -6.08 -0.118 Return on Assets 0.002 0.30 Ratio of Debt to Assets ** 2.13 Total Assets 0.002 R&D -0.000 -0.08 Missing R&D -0.003 -1.19 ** 2.03 Delaware Incorporated 0.006 *** -5.52 -0.049 Dual Class Stock * -1.84 -0.001 Insider Ownership *** 45.27 0.308 ISS Negative Recommendation *** 6.49 Constant 0.048 2 -0.848 Pseudo R *** 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. to vote in favor of executive compensation when the issuer has experienced poor economic performance. It is worth noting the results of three additional variables, total assets, dual class stock and ISS negative recommendation. The total assets coeffi- cient is positive and statistically significant, meaning that shareholders are less likely to vote against pay packages at large issuers. This result is inter- esting 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 ex- 83 this result is in tension ecutive pay, as some commentators have observed, with that hypothesis. 83 See B ROADRIDGE & P RICEWATERHOUSE C OOPERS , 2016 P ROXY S EASON P REVIEW 2 (1st ed. 2016), http://media.broadridge.com/documents/ProxyPulse-First-Edition-2016.pdf (“Among the companies that failed to attain majority shareholder approval for executive com- pensation plans, retail investors cast 66% of their votes in favor of these plans while institu- tions cast 65% of their votes against.”); David Bogoslaw, Retail Investors Seen as Key to Firms Struggling on Say-on-Pay, Says ProxyPulse , IR M AGAZINE (Oct. 8, 2013), https://www .irmagazine.com/articles/proxy-voting-annual-meetings/19800/retail-investors-seen-key-firms-
19 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 18 Harvard Business Law Review [Vol. 8 118 The coefficient on dual class stock is negative and highly significant, meaning that shareholders are more likely to vote “no” on say on pay votes at dual class stock firms. This finding is consistent with the notion that say on pay votes may be used to express dissatisfaction with metrics other than pay itself. In this case, the lower votes may be a product of the lack of a governance voice in other areas. In accordance with our preceding analysis, the effect of a negative ISS recommendation is positive and statistically significant, meaning that ISS recommendations substantially influence say on pay voting outcomes. Nota- bly, and in line with other research, we find an independent effect associated 84 At the same time, the ISS recommendation with the ISS recommendation. does not fully explain the voting results. In unreported results, we ran the regressions with and without the ISS control variable. 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 statistical significance even when we control for the ISS effect. Table IV below 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 “1” if the percent- age of votes against is greater than or equal to 20%, and “0” otherwise. Given that the dependent variable is binary, we ran a Probit regression rather than a standard OLS regression. In the Probit regression, the coefficient on excess compensation is posi- tive and statistically significant at the 1% level, meaning that higher excess compensation is correlated with an increased likelihood of more than 20% of the shares being voted against the compensation plan. However, the pay- performance sensitivity variable is not statistically 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 negative and statistically significant, meaning that issuers with better economic performance are less likely to experience a low say on pay vote. Dual class stock again remains negative and highly statistically significant. a “no” recommen- The ISS variable is positive and statistically significant — dation from ISS is correlated with a low vote. Again, this highlights the influence of ISS. The results in Table IV largely mirror those in Table III with the excep- tion of the results on pay-performance sensitivity. They indicate that al- struggling-say-pay-says-proxypulse (stating that retail investors showed greater support for say-on-pay plans than institutional investors). 84 See, e.g. , Jie Cai, Jacqueline L. Garner & Ralph A. Walkling, Electing Directors , 64 J. F . 2389 (2009) (reporting that ISS recommendation against a director based on governance IN characteristics reduces director votes by 20.7% on average); Stephen Choi, Jill Fisch & Marcel Kahan, The Power of Proxy Advisors: Myth or Reality? , 59 E MORY L.J. 869, 906 (2010) (find- ing independent effect of 6-10% of ISS recommendations in uncontested director elections).
20 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 12:57 Seq: 19 25-JUN-18 119 Is Say on Pay All About Pay? 2018] T ETERMINANTS OF L OW V OTE (P ROBIT ABLE EGRESSIONS ) IV: D R Regression Independent Variables Z-Statistic Coefficient -5 *** Excess Compensation (10 4.54 ) 0.704 -5 Pay-Performance Sensitivity (10 ) 0.124 0.54 *** -2.71 Stock Returns -0.033 *** -4.04 -0.266 Return on Assets Ratio of Debt to Assets -0.002 -0.06 Total Assets 0.002 0.70 R&D 0.000 0.12 * -1.75 -0.018 Missing R&Ds ** 2.45 0.025 Delaware Incorporated *** -6.02 -0.247 Dual Class Stock Insider Ownership -0.002 -1.09 *** 18.53 ISS Negative Recommendation 0.386 *** -9.82 Constant -1.758 2 0.524 Pseudo R *** 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. though excess compensation is an important driver of low say on pay votes, so is the performance of the company, even when controlling for excess compensation. The absence of a significant result on the pay-performance sensitivity variable is interesting because it suggests the possibility that — actual excess compensation is more impor- from a shareholder perspective — tant than the structure of such compensation. On this point, the results demonstrate that, to some extent, ISS and shareholders appear to agree to some extent in their analysis of when a pay package is problematic, though as the results in Table III show, there is some divergence. Thus far, we have not focused on the relative importance of CEO pay and economic performance with respect to voting outcomes. Table V below addresses that issue. We divided our sample into four quartiles based on the 85 We then sorted those quartiles based amount of CEO excess compensation. on firm economic performance. The resulting matrix Table V shows the — — 85 See supra Table IV. Based on the results described in Table IV, we focus here on excess compensation rather than pay-performance sensitivity.
21 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 12:57 Seq: 20 25-JUN-18 [Vol. 8 120 Harvard Business Law Review percentage of votes against for each combination of excess pay and eco- nomic performance, measured in terms of stock price. V: Q ABLE A NALYSIS OF CEO P AY UARTILE T CEO Pay = Excess Pay Returns High Low Q4 Q3 Q2 Q1 4.45% 4.57% 5.30% 7.45% Q1 % against Low 4.78% 4.74% 6.67% 9.77% Q2 7.41% 7.54% 8.22% 12.32% Q3 11.40% 13.90% 15.86% 22.42% High Q4 As shown in Table V above, although shareholders are somewhat sensi- tive to excess CEO pay when stock price performance is strong, their reac- tion 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 that are in the lowest performance quartile, the level of negative votes almost doubles. 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 low- est performers. This increase appears to be driven by stock price perform- ance, not by pay. 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 compensa- tion 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 below, we quantify how much these fac- tors matter. Table VI shows the effect of an increase of one standard deviation in each independent variable from Tables II and III the compensation vari- — ables (excess compensation and pay-performance sensitivity) and the per- formance variables (stock returns and return on assets). As shown in Table I, the standard deviation of excess compensation is $4.492 million and the standard deviation of pay-performance sensitivity is $1.1139 million. Table VI then shows, in each column, the effect of increasing these variables by
22 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 21 Is Say on Pay All About Pay? 2018] 121 T ARIABLES E FFECT OF K VI: E V CONOMIC ABLE EY Low Vote % Against ISS Negative (using (using Recommendation regression regression (using regression estimates estimates estimates from from Table from Table Table II) III) IV) Excess Compensation 57.64% 19.70% 23.18% Pay-Performance Sensitivity 0.00% -3.40% 0.00% Total Effect of Pay 23.10% 23.18% 57.64% (Absolute Terms) -38.70% -8.72% Stock Returns -5.31% Return on Assets -16.78% -9.68% -14.64% Total Effect of Performance 14.99% 23.36% 55.48% (Absolute Terms) one standard deviation, respectively, the probability of a “no” recommenda- tion from ISS, the fraction of votes cast against say on pay, and the probability of a low say on pay vote, meaning that fewer than 80% of shares were voted in favor of the compensation package. The first column of Table VI 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 a 57.64% increase in the average probability of an ISS “no” rec- 86 This result is consistent with what we might expect, namely, ommendation. that ISS is substantially more likely to recommend against a pay package when excess compensation increases. Similarly, increasing pay-performance sensitivity by one standard deviation results in a zero change in the probability of an ISS “no” recommendation. The sum of these two pay- related effects is 57.64%. This means that a one standard deviation increase in both excess compensation and pay-performance sensitivity will increase the average probability of an ISS “no” recommendation by 57.64%. This finding demonstrates the importance of pay-related factors in explaining the ISS recommendation. The third and fourth rows of the first column of Table VI examine the stock returns and return on assets. importance of the performance variables — Again, we assess the effect of increasing firm performance by one standard deviation on the average probability of receiving a negative say on pay vote recommendation from ISS. Importantly, we calculated these effects while holding constant the compensation variables reflected in the first and second 86 This can be calculated as (.0000141*4492.54/.111). All other probabilities are similarly calculated.
23 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 22 Harvard Business Law Review 122 [Vol. 8 rows of Table VI. A one standard deviation increase of .3623 in stock returns is associated with a 38.70% decrease in the average probability of a “no” recommendation from ISS. Similarly, a one standard deviation increase in return on assets, is associated with a 16.78% decrease in the average probability of a “no” recommendation from ISS. Summing these two ef- fects, we find that our performance variables yield a combined 55.48% de- crease in the average probability of a “no” recommendation from ISS. In the second column of Table VI, we examine the effect of a one stan- dard deviation change in our independent variables on the fraction of votes cast against a company’s executive compensation plan. Using the same met- rics, we find that a one standard deviation increase in excess compensation correlates to, on average, a 19.70% increase in the fraction of shares voted against the compensation plan. A one standard deviation in pay-performance sensitivity is associated with a 3.40% decrease in the average fraction of votes against the proposal. The joint effect of 23.10% is significantly smaller than the effect of the pay variables on the ISS recommendation suggesting that shareholders do not necessarily vote against a compensation package on the basis of a negative recommendation from ISS. This result is consistent with our observations elsewhere. 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 fraction of votes against the compensation plan of 5.31%. A one standard deviation increase in return on assets results in a decrease in the average fraction of votes cast against the plan of 9.68%. The total is a joint effect of 14.99%. 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 re- duce the fraction of votes against the compensation plan by an average of 14.99%. In the third column of Table VI, we examine the extent to which pay and performance factors contribute to the probability of a low say on pay vote, which we defined above as fewer than 80% of shares voted in favor of the compensation package. We find that a one standard deviation increase in excess compensation results in the average probability of a low vote increas- ing by 23.18%. A one standard deviation change in pay-performance sensi- tivity 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 deviation increase in stock returns results in the probability of a low vote decreasing by an average of 8.72%. A one standard deviation increase in return on assets results in the probability of a low vote decreasing by an average of 14.64%. The joint effect of the two performance variables is 23.36%. Again, this effect is net of, or in addition to, the effect of the pay variables.
24 \\jciprod01\productn\H\HLB\8-1\HLB102.txt Seq: 23 12:57 unknown 25-JUN-18 123 Is Say on Pay All About Pay? 2018] In Table VII we reran the same models that were used to generate the results given in Table VI. However, in this model we did not use continuous variables for stock returns and return on assets. Instead, each is a dummy variable. Stock returns is set to “1” if the stock return is negative and “0” otherwise. Return on assets is similarly set to “1” if return on assets is nega- tive and “0” otherwise. We performed this test in order to assess how share- holders respond to economic performance when performance is simply negative rather than relative. ABLE VII: E CONOMIC E FFECT OF K EY V ARIABLES T U SING D UMMY V ARIABLES Fraction Low Vote ISS Negative Against (using (using Recommendation regression (using regression regression estimates estimates from estimates Table from Table from Table II) IV) III) 56.41% 19.60% 23.18% Excess Compensation 0.00% -3.63% 0.00% Pay-Performance Sensitivity Total Effect of Pay 56.41% 23.23% 23.18% (Absolute Terms) Stock Returns 75.28% 18.59% 14.92% 69.79% 27.07% 33.17% Return on Assets Total Effect of Performance 145.07% 41.98% 51.76% (Absolute Terms) The results highlight the same trend as Table VI. We find broadly simi- lar results in terms of the total effect of pay in absolute terms on the say on pay vote. However, in terms of total effect of performance, the use of dummy variables shows that performance becomes far more important. In this case, for a given level of excess compensation and pay-performance sensitivity, moving from positive performance to negative performance reduces the fraction of votes against the compensation plan by an average of 41.98% when a dummy variable is used versus 14.99% when continuous variables are utilized. The largest effect is in the ISS negative vote recom- mendation. Moving from positive performance to negative performance in- creases the likelihood of an ISS negative recommendation to 145.07% from 55.48%. These results demonstrate the dramatic effect of negative perform- ance irrespective of the compensation variables. In summary, our results highlight that the size and structure of execu- tive compensation contribute both to the ISS recommendation and to the level of shareholder support for the executive compensation plan as reflected
25 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 24 Harvard Business Law Review 124 [Vol. 8 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 recommen- dation, 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 ap- prove compensation plans that provide high levels of excess compensation or in which compensation is insufficiently sensitive to firm performance. To this extent, the say on pay vote appears to be responding appropriately to identify problematic compensation practices. On the other hand, the results of our analysis of the performance vari- ables are dramatic and potentially troubling. We find that the issuer’s eco- nomic performance is a substantial driver of both ISS recommendations and shareholder votes with respect to say on pay, and remains so even when controlling for pay size and structure. In particular, the fact of negative per- formance alone significantly increases the likelihood of a say on pay “no” vote and a negative recommendation. 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, at least in part, a referendum on firm perform- ance. We explore the implications of this finding in Part IV of this Article. MPLICATIONS III. I Our findings have two important implications. First, the substantial role of economic performance in driving both ISS recommendations and share- holder say on pay votes suggests a limitation on 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 issuers to focus excessively on short-term stock price and firm value metrics. A. The Role of Economic 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 per- formance, shareholders appear to be limiting their criticism of executive compensation primarily to firms that are suffering from poor economic per-
26 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 25 Is Say on Pay All About Pay? 125 2018] 87 formance. 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, including structural problems that may create risks for the sustainability of that performance. Although the rationale for a sepa- rate say on pay vote is to allow shareholders those 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, the say on pay vote does not appear to be serving that function. One possibility is that shareholders do not care about the size and struc- ture of executive pay as long as the company is performing well. Executive pay at most issuers represents a small fraction of firm revenue, and share- holders may not view pay levels as economically important. Alternatively, shareholders may view even excessive levels of pay as well-deserved if the company has strong economic performance. These possibilities provide rea- sons why say on pay may be a poor tool for reform if the say on pay objec- tive is either to reduce high levels of executive compensation or to have 88 This is a particularly compensation more closely align with performance. important point since the legislative purpose of say on pay was to implement 89 both of these objectives. 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 found 149 instances of low shareholder votes (less 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 shareholders’ negative votes on pay were driven primarily by firm performance. Similarly, we find that the level of shareholder dissent for firms in the lowest quartiles for both economic per- formance and 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. 87 See Audrey Bout, Brian Johnson & Steve DeMaria, Does Say on Pay Failure Affect Future Share Price Performance? , P AY G (Feb. 1, 2015), http://paygovernance OVERNANCE .com/does-a-say-on-pay-failure-affect-future-share-price-performance/ (reporting that issuers with failed say on pay votes typically underperform the market prior to the vote). 88 Jill E. Fisch, The Mess at Morgan , 83 U. C IN . L. R EV See – 79 (2015) (arguing . 651, 678 that high levels of shareholder support for executive compensation packages at J.P. Morgan may have been due to the company’s strong economic performance despite its involvement in a variety of scandals). 89 See, e.g. , Lisa M. Fairfax, Sue on Pay: Say on Pay’s Impact on Directors’ Fiduciary Duties , 55 A RIZ EV . 1, 4 (2013) (“The push to mandate say on pay stems from a belief . L. R 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 N L. & B W NT ’ L . J. I US . 593, 597 (2011) (explaining that say on pay had its place in Dodd-Frank “because of the widespread perception that executive pay prac- tices contributed to the financial crisis”).
27 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 26 Harvard Business Law Review [Vol. 8 126 Our data offers real world support for the hypothesis tested by Krause, Whitler, and Semadeni in their laboratory experiments that shareholder — voting has limited power as a tool for addressing potential agency issues 90 To a certain extent, this problem may be identified by Bebchuk and Fried. exacerbated by the ISS methodology for defining pay-performance sensitiv- ity in terms of realized performance, which does not precisely capture the extent to which pay is structured to align management incentives with share- holder 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 issue. These findings are more problematic if say on pay is designed to reduce overall compensation levels consistent with broader societal objectives of 91 To the extent that shareholder voting is driven equity or wealth distribution. 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 excessive or inordinate pay alone, shareholder voting is un- likely to result in the appropriate compensation reforms. B. Say on Pay May Exacerbate Problematic Incentives Our findings also empirically demonstrate the risk that say on pay vot- ing may exacerbate, rather than eliminate, problems with executive pay structure. We show that shareholder support for executive pay is highly cor- related with an issuer’s short-term stock performance. The performance vari- ables in our analysis focus on the issuer’s economic performance in the year prior to the say on pay vote. As we have shown, 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, instead of working to enhance long-term firm value. Many commentators have expressed a concern that both issuers and shareholders have adopted a short-term perspective with respect to strategic 92 More significantly, they view short-termism as having an ad- decisions. 90 Krause et al., supra note 53, at 98 (citing Lucian A. Bebchuk & Jesse M. Fried, R See Executive Compensation as an Agency Problem , 17 J. E CON . P ERSP . 71, 71 – 92). 91 See, e.g. T HOMAS P IKETTY , C APITAL IN THE T WENTY -F IRST C , 2 0 – 21 (Arthur ENTURY Goldhammer trans., Harvard University Press 2d ed. 2014) (discussing problems associated with wealth inequality); Fisch, supra note 88, at 653 (discussing say on pay within the concept R of corporate “publicness”). 92 See, e.g. , Jack B. Jacobs, “Patient Capital”: Can Delaware Corporate Law Help Revive It? , 68 W . & L EE ASH EV . 1645, 1651 (2011) (identifying short-termism of corporate share- L. R holders 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 , 50 W Law AKE F OREST L. R EV . 761, 772 (2015) (arguing that activist shareholder “pressure may logically lead to strategies that sacrifice long-term performance for short-term shareholder
28 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 27 Is Say on Pay All About Pay? 127 2018] 93 verse effect on business decision-making. 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 pres- sures and increases the incentive for boards to sacrifice long-term value in 94 Ironically, say on pay may op- favor of immediate economic performance. erate in direct contradiction to efforts to reform the structure of executive 95 pay packages in order to create longer-term incentives. A further possibility is that say on pay could contribute not merely to short-termism, but to excessive risk-taking because of the correlation be- tween risk and stock market performance. The stock market has traditionally rewarded issuers for taking risk because diversified shareholders are able to bear that risk. Incentive-based compensation structures that rely on stock price, especially short-term stock price, may lead executives to take exces- 96 The situation may sive risk in an effort to maximize short-term stock price. be worse if the market does not fully understand or reflect the riskiness of an 97 issuer’s strategic decisions. 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 98 This means that stock price dominates both total shareholder return (TSR). but see wealth.”); The ‘Long Term’ in Corporate Law , 72 B US . L AW . 353, 354 J.B. Heaton, (2017) (disputing the claim that there is a conflict “between short-term wealth maximization and long-term wealth maximization”). 93 William H. Donaldson, Chairman, Sec. & Exch. Comm’n, Speech by SEC Chair- See May 8, 2005), http://www.sec.gov/news/ NSTITUTE A NNUAL C ONFERENCE ( CFA I man: 2005 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”). 94 See Jacobs, note 92, at 1651 – 52 (identifying the risk of incentivizing corporate R supra executives to manage to the market). 95 See, e.g. , Roberta Romano & Sanjai Bhagat, Reforming Executive Compensation: Fo- J. ON R EG . 359 (2009) (proposing pay ALE , 26 Y cusing and Committing to the Long-term package reforms to create long-term incentives). 96 See. e.g. , Wm. Gerard Sanders & Donald C. Hambrick, Swinging for the Fences: The CAD GMT . J. . M A , 50 Effects of CEO Stock Options on Company Risk Taking and Performance 1055, 1076 (2007) (reporting that CEOs take excessive risks to maximize payoffs from stock OBIL OMPENSATION XXON M C , 2013 E XECUTIVE see also E option incentive compensation); O VERVIEW 8 (2013), www.exxonmobil.com/Corporate/Files/news_pub_ir_execcomp2013.pdf (“[A] formula based approach that relies heavily on one- or three-year total shareholder return could encourage inappropriate risk taking and have a lasting and negative impact on Exx- onMobil’s business by encouraging a focus on more immediate results at the expense of our long-term business model.”). 97 See, e.g. , Henry Hu, Risk, Time and Fiduciary Principles in Corporate Investment , 38 EV . 277, 325 (1990) (warning of the excessive risk-taking incentive that incentive- UCLA L. R based compensation can create “when compensation is highly sensitive to perceived perform- ance, and true, risk-adjusted performance is difficult to measure”). 98 See BJ Firmacion & Jim Kroll, ISS announces pay-for-performance methodology up- , Nov. 11, 2016, https://www.tower- ILLIS T OWERS W ATSON W dates for the 2017 proxy season , swatson.com/en/Insights/Newsletters/Global/executive-pay-matters/2016/ISS-announces-pay- for-performance-methodology-updates-2017-proxy-season (“Currently, ISS uses TSR as the sole financial metric in its quantitative pay-for-performance screens.”); see also Mark Van Clieaf, Stephen O’Byrne & Karel Leeflang, The Alignment Gap Between Creating Value, Per-
29 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 28 Harvard Business Law Review [Vol. 8 128 99 the analysis of pay sensitivity and firm performance. 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 study reports results consistent with our findings, documenting that shareholders fail to distinguish, in their say on pay votes, between issuers that create economic 100 In fact, the study finds that value and issuers that destroy economic value. most issuers do not even disclose meaningful performance metrics that 101 would allow shareholders to focus on the creation of economic value. AY ON P AY ORWARD F HE P IV. S ATH T — Our findings suggest caution both about say on pay votes and say on pay itself. In particular, our findings support the conclusion that say on pay is about more than just pay, it is also about firm stock price performance. The focus on firm performance may produce excessive risk-taking. Further analysis is needed to detail whether the say on pay vote is having these effects. Some have suggested that say on pay’s primary benefit has been as an 102 Industry commentators and institu- outlet for shareholder dissatisfaction. tional investors describe an increased dialogue between issuers and institu- 103 But, from a “channel” tional investors about pay levels and pay structures. perspective, we believe that dissatisfaction with issuers is better expressed through means that convey actual dissatisfaction. In addition, while we think the increased dialogue resulting from say on pay is meaningful, it is coinci- 104 If say on pay were dent with the rise of shareholder power more generally. formance Measurement, and Long-Term Incentive Design , IRRC R ESEARCH R EPORT 11, 13, 54 (2014) (explaining and criticizing reliance on TSR as a measure of economic performance). 99 See id. 100 – 52. See id. at 51 101 See id. at 8, 53. According to Ramy S. Ibrahim, an ISS Corporate Solutions Advisor, “ISS this year will begin including other financial performance metrics in the initial P4P quan- titative screen. The metrics have not been finalized yet . . . .” (statement provided by Ramy S. Ibrahim via email to authors on Nov. 6, 2017). 102 Say-on-Pay: Wrong Solution, Wrong Problem Troutman, Sanders, , (July 9, 2010), https://www.troutmansanders.com/say-on-pay-wrong-solution-wrong-problem-06-09-2010/ (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 perform- ance other than voting with their feet – and say-on-pay proposals have evolved as a proxy for a – more fundamental complaint on company performance”). 103 See, e.g. , Seymour Burchman & Blair Jones, Righting the Say On Pay Ship After a , S “No” Vote EMLER ROSSY (Sept. 2013), http://www.semlerbrossy.com/wp-content/uploads/ B ATD-Shareholder-Engagement.pdf (“One of the positive outcomes of the Say on Pay provi- sion in the Dodd-Frank legislation has been more regular dialogue between companies and shareholders.”). 104 See generally The Agency Costs of Agency Ronald J. Gilson & Jeffrey N. Gordon, Capitalism: Activist Investors and the Revaluation of Governance Rights , 113 C OLUM . L. R EV . 862, 897 (2013) (explaining ability of activist shareholders to leverage their power through the support of more passive institutional investors).
30 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown 25-JUN-18 12:57 Seq: 29 Is Say on Pay All About Pay? 2018] 129 to be eliminated as a legal requirement, we believe 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 105 ISS’s use of realized pay for performance may exacerbate the con- votes. cerns raised in this Article by incorporating an ex post measurement linked to stock performance. The ISS approach also appears to differ from the man- ner in which shareholders evaluate pay-performance sensitivity. ONCLUSION C Say on pay is in its early years, and issuers and shareholders are still developing their approach to the vote on executive compensation. This Arti- cle provides an analysis of the first five years of shareholder voting in an effort to determine the key factors that influence the say on 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 management. However, when firms per- form well, 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, stock price, and firm perform- ance 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. 105 See, e.g. , Choi, Fisch & Kahan, supra note 84 (finding that, in uncontested director R elections, proxy advisor recommendations drive 6-10% of the vote).
31 \\jciprod01\productn\H\HLB\8-1\HLB102.txt unknown Seq: 30 25-JUN-18 12:57
SEPTEMBER 2017 WHY COMPETITION IN THE POLITICS INDUSTRY IS FAILING AMERICA A strategy for reinvigorating our democracy Katherine M. Gehl and Michael E. PorterMore info »
Improving Skills Through America’s Workforce Development System Kevin Bauman and Cody Christensen SEPTEMBER 2018 AMERICAN ENTERPRISE INSTITUTEMore info »
Main Street Matters Ideas for Improving Small Business Financing AUGUST 2018More info »
All the People, All the Places A Landscape of Opportunity for Rural and Small-Town Civic Engagement 2018 by Ben GoldfarbMore info »
Attorney General’s Advisory Committee on American Indian/Alaska Native Children Exposed to Violence: Ending Violence so Children Can Thrive NOVEMBER 2014More info »
A Bid for Better Transit Improving service with contracted operationsMore info »
Review of the available Remote sensing tools, pRoducts, methodologies and data to impRove cRop pRoduction foRecastsMore info »
Institute for Civil Justice A RAND LAW, BUSINESS, AND REGULATION INSTITUTE CHILDREN AND FAMILIES The RAND Corporation is a nonprofit institution that helps improve policy and EDUCATION AND THE ARTS de...More info »