CEO Equity-based Incentives and Shareholder Say-on-pay in the U.S.

Transcription

1 CEO Equity-based Incentives and Shareholder Say-on-pay in the U.S. Denton Collins Blair B. Marquardt Xu Niu May 30, 2017 Abstract This paper examines the relationship between shareholder voting outcome (percent in favor or opposed) and executive equity-based incentives (pay-performance sensitivity and pay-risk sensitivity) as part of the say-on-pay provision of the 2010 Dodd-Frank Act. Our setting allows us to directly test whether shareholders perceive equitybased incentives as a source of or solution to agency problems. Consistent with our hypotheses, we provide evidence that shareholders tend to approve of compensation packages that are more sensitive to changes in stock price (pay-performance sensitivity) and changes in stock volatility (pay-risk sensitivity). Our findings are consistent with theoretical predictions that outside owners approve of equity incentives as a means of aligning managers interests with those of shareholders and as a way to mitigate potential agency costs. We also demonstrate that future changes to equity-based incentives are related to voting outcomes. We provide meaningful evidence of the importance of equity-based incentives from the perspective of those most concerned with firm value and of the effectiveness of say-on-pay as a governance mechanism. JEL classification: G34, G38, M48, M52 Keywords: pay-performance sensitivity, pay-risk sensitivity, executive compensation, shareholder voting We thank Bruce Billings, Shira Cohen, Jack Cooney, Jon Durrant, Hal Elkins, Josh Fairbanks, Brett Myers, Volkan Muslu, Derek Oler, Juan Manuel Sanchez, and workshop participants at the 2017 AAA Financial Accounting and Reporting Section Conference and Texas Tech University for their helpful comments. School of Accounting, Rawls College of Business Administration, Texas Tech University, Lubbock, Texas 79409, denton.collins@ttu.edu. Corresponding author at School of Accounting, Rawls College of Business Administration, Texas Tech University, Lubbock, Texas 79409, blair.marquardt@ttu.edu. Area of Finance, Rawls College of Business Administration, Texas Tech University, Lubbock, Texas 79409, xu.niu@ttu.edu.

2 1. Introduction President Barack Obama signed into law the Dodd-Frank Wall Street Reform and Consumer Protection Act (here forward the Dodd-Frank Act or the Act ) on July 21, 2010 in response to the 2008 financial crisis. The legislation requires shareholder say-onpay, a mandated advisory vote by shareholders on the compensation of top executives of public companies. We examine the outcomes of these votes to study shareholder preferences for pay-performance sensitivity (Core and Guay, 1999) and pay-risk sensitivity (Guay, 1999). Pay-performance sensitivity ties executive wealth to stock returns, while pay-risk sensitivity aligns the executive s risk appetite with that of the firm. These incentives arise from stockbased compensation, such as employee stock options and shares of stock. Together, they are intended to mitigate agency costs by linking manager and shareholder wealth. Our motivation is threefold. First, we utilize this setting to study whether shareholders express views consistent with performance- and risk-based equity incentives as net beneficial or costly. Given the abundance of research dedicated to the effects of equity-based incentives, it is worthwhile to consider the shareholders perceptions as expressed in the say-on-pay vote. Theory suggests that shareholders are concerned with a potential misalignment of interests with management and attempt to mitigate this via incentive compensation that ties executive wealth to improvements in firm value (Core and Guay, 1999; Murphy, 2013). Additionally, theory suggests that incentive compensation can encourage investment in value-enhancing, risky projects (Guay, 1999; Prendergast, 2002). While some empirical evidence has supported these assertions (e.g., Jensen and Murphy, 1990; Coles, Daniel, and Naveen, 2006), an abundance of value-destroying consequences have also been documented (e.g., Bergstresser and Phillippon, 2006; Armstrong and Vashishtha, 2012). Equity-based incentives may also be less intuitive and transparent than other aspects of compensation (e.g., bonuses). Thus, it is unclear ex ante whether shareholders approve of equity-based incentives, 1

3 on average. Assuming that shareholders are most concerned with the total returns to their investments (Financial Accounting Standards Board, 2010), we consider this a relatively clean and powerful setting to examine the value-relevance of equity-based compensation. Second, our research addresses the effectiveness of shareholder voting. Theory suggests that shareholders are concerned with management moral hazard (Murphy, 2013). Furthermore, boards of directors at times award executives compensation packages that perpetuate, rather than mitigate, agency costs (Bebchuk and Fried, 2003). Shareholder voting provides an opportunity for shareholders to express discontent with such misalignment, as well as express their opinions on these types of incentives generally. The effectiveness of shareholder voting as a governance mechanism is debated, but growing evidence demonstrates the power of voting to effect change. Our research assesses the power of shareholder voting in an important governance context, incentive compensation. Third, we assess whether shareholders vote in a manner consistent with the intent of the regulation. The Dodd-Frank Act s intention was to promote greater responsibility and accountability of executives (US Senate Committee on Banking, 2010). Regulators have expressed concerns about potential societal damage due to rent extraction and excessive risk taking by insulated executives, and appear to support stronger alignment of manager and shareholder interests (US Senate Committee on Banking, 2010; Securities and Exchange Commission, 2015). We investigate whether shareholders appear sensitive to equity-based incentive in their votes and whether shareholder voting is related to subsequent changes in equity-based incentives. Thus, our study seeks to inform regulatory efforts in this area, notably regulation regarding expanded disclosure of pay-performance sensitivity metrics (Securities and Exchange Commission, 2015). To our knowledge, our research represents a first step in examining shareholder response to the underlying incentives embedded in equity compensation, a feature critical to evaluating the appropriateness of a compensation package (Murphy, 2013). 2

4 To shed light on these issues, we examine how the underlying pay-performance sensitivity and pay-risk sensitivity of a CEO s compensation package explain the proportion of favorable votes. The large cross-section of voting data that have become available after the effective date of the regulation reduces selection bias and increases generalizability of voting outcomes regarding executive compensation. This setting provides us with a window into shareholder perceptions of performance-based and risk-based equity incentives, a perspective not previously examined. We hypothesize that the favorable vote will be increasing in the strength of the linkage between pay and performance and in the strength of the risk-taking incentive. We measure these incentives using CEO wealth delta and vega, scaled by total compensation (Core and Guay, 1999; Guay, 1999), respectively. 1 Assuming that the underlying characteristics of the compensation package will drive shareholder perceptions of the acceptability of the package, we regress the proportion of favorable votes on scaled CEO delta and vega, plus a vector of variables designed to control for other characteristics expected to influence shareholder voting. Our results support our hypotheses. We document a significant positive relationship between CEO pay-performance sensitivity (scaled delta) and the proportion of favorable votes, implying shareholders generally approve of compensation packages designed to align manager and shareholder wealth. We also document a significant positive relationship between CEO pay-risk sensitivity (scaled vega) and the proportion of favorable votes, which we interpret as evidence that shareholders generally approve of packages that align of manager and firm risk appetites. Furthermore, the voting outcome is driven by the increase in pay-performance sensitivity relative to historical levels, which suggests that shareholders 1 Delta is the change in dollar value of equity incentives of the CEO for a 1% change in stock price. It measures the sensitivity of the CEO s compensation to changes in the stock price. Vega is the change in dollar value of the CEO s incentive options for a 0.01 change in the annualized standard deviation of stock price. It measures the risk-taking incentive of the CEO s compensation. 3

5 focus on the most recent equity awards. Additional cross-sectional tests reveal that these effects are generally stronger in settings where demand for CEO incentives is expected to be higher. For example, the effects are stronger in the presence of sophisticated ownership, as represented by high institutional ownership, where the shareholder base is presumed better able to understand and evaluate these complex incentives. The effects are also stronger in firms with weak financial performance, as represented by industry-adjusted return on assets, where shareholders are expected to be especially critical of executives shielded from losses. Together, these findings support our hypothesis that shareholders value performance-based equity compensation as a tool for mitigating agency costs. Assuming shareholders votes capture their informed judgments of the impact of equity-based compensation on firm value, our results further support the interpretation that the benefits of equity-based incentives outweigh the unintended consequences, on average. We also extend prior research by conducting analyses of shareholder voting on pay packages of executives other than the CEO. Specifically, we find that the pay-performance sensitivities of the CFO and the remaining Top Five executives play incrementally important roles in shareholder voting. We also document evidence that shareholders perceive the returns on pay-performance sensitivity as diminishing. Specifically, we identify a concave relationship between the proportion favorable vote and CEO pay-performance sensitivity. This is consistent with recent research suggesting diminishing returns on investments in equity compensation (Billings, Moon, Morton, and Wallace, 2016), and further implies shareholders are sensitive to these diminishing returns in their voting. Finally, we examine the relationship between shareholder voting outcomes and future changes to CEO equity-based awards. We find that large shareholder dissent is associated with greater increases to pay-performance sensitivity in the subsequent period. This implies that boards do respond to shareholder opinion, which is a critical finding in the debate of the effectiveness of say-on-pay. 4

6 Our findings inform discussion and debate on shareholder activism as a monitoring mechanism. In the presence of powerful management or biased directors, additional external controls may be necessary to ensure protection of shareholder wealth (Bebchuk and Fried, 2003). A growing body of research has examined whether and how shareholders can use their say-on-pay vote as one such external check on management and an unresponsive board (Ertimur, Ferri, and Oesch, 2013; Al-Issa, 2015; Iliev and Vitanova, 2015; Balsam, Boone, Liu, and Yin, 2016; Correa and Lel, 2016; Kimbro and Xu, 2016). Our research expands this knowledge to the area of incentive compensation. To our knowledge, we are the first to directly examine the say-on-pay vote in relation to important equity-based incentives, pay-performance sensitivity and pay-risk sensitivity (Core and Guay, 1999; Guay, 1999). These represent distinct and important constructs relative to those studied by prior research (Balsam, Boone, Liu, and Yin, 2016; Kimbro and Xu, 2016). Our research examines shareholders perceptions of the incentive effects of equity, holding constant the dollar values of compensation. Consequently, our study emphasizes the use of equity compensation as a governance and control mechanism, and not just a reward for past performance. Our results and interpretation suggest that shareholders have a relatively sophisticated understanding of compensation packages and how those packages can potentially mitigate agency risks. Our research also provides insights into the effectiveness of regulations attempting to restrain perceived excessive executive compensation. Our results imply that say-on-pay, mandated through the Dodd-Frank Act, has provided shareholders with a mechanism to express their collective judgment on the specifics of compensation packages, and that shareholders are using the vehicle to express their concerns on the relation between compensation and incentive alignment. Furthermore, we document some evidence of board responsiveness to such concerns, implying that say-on-pay voting is able to influence subsequent changes in incentive compensation design. Our findings complement those of other researchers attempting to understand the effectiveness of international regulatory 5

7 efforts regarding compensation, e.g., Correa and Lel (2016). Additional research will be necessary, however, to evaluate the long-term effectiveness of the vote in restraining rent extraction, encouraging and optimizing incentivized pay of top executives, and aligning the interests of managers and shareholders. Thus, the results of our study contribute to the ongoing discussions of shareholder activism and regulatory compensation reform. 2. Prior Literature and Hypothesis Development Say-on-pay refers to the shareholder vote on the compensation packages of top executives. Say-on-pay was mandated by the Dodd-Frank Act of 2010, and final regulations were issued by the SEC in January 2011 (Securities and Exchange Commission, 2011). Traditionally, the role of monitoring, including the design of executive compensation, is administered by the elected board of directors. However, shareholder voting serves as an additional monitoring mechanism to shareholders outside the board of directors, as the board may be unduly influenced by top executives or have ulterior motives (Bebchuk, Grinstein, and Peyer, 2010; Coles, Daniel, and Naveen, 2014). The purpose of mandated say-onpay is to enhance this existing check-and-balance mechanism (US Senate Committee on Banking, 2010). We study the outcomes of the votes to understand shareholders preferences in compensation design. Specifically, our research addresses whether shareholders say-onpay votes are sensitive to equity-based incentives Agency Conflict and Equity-based Incentives Theory regarding executive compensation relates to the agency conflict (Jensen and Meckling, 1976). Management, as a self-interested and effort-averse party, has inherently different motives from outside shareholders. This tension is exacerbated by the fact that management also has superior information about the firm. Thus, shareholders use incentives, 6

8 such as linking pay with easily observable performance outcomes instead of difficult to observe effort, to prompt management into behaviors that better align with shareholders interests (Hölmstrom, 1979). This disciplining effect makes executive compensation a governance mechanism. Consistent with this notion, empirical research also demonstrates the deterioration of firm value when executives are overpaid beyond the amount predicted by economic fundamentals (Core, Holthausen, and Larcker, 1999). Stock-based compensation is one such incentive mechanism that has been proposed to mitigate agency conflicts. It may take the form of stock awards (e.g., restricted stock) or stock options, and is typically designed so that the benefit received by management increases monotonically with wealth improvements of the shareholders through an increase in stock price. The strength of stock-based compensation as an incentive derives from the sensitivity of its value to changes in the firm s stock characteristics, similar to that of a marketable derivative. Core and Guay (1999) and Guay (1999) document that boards use these equity awards to align executive and firm interests. Core and Guay (1999) show that firms use equity incentives that align executive pay with stock price performance, while Guay (1999) demonstrates that option awards are designed to encourage risk taking. These incentive measures represent distinct, but related, constructs. We consider each in more detail below. The sensitivity of a manager s firm-related wealth to changes in the firm s stock price, known as delta, is a common proxy for pay-performance sensitivity. 2 Agency theory asserts that increased pay-performance sensitivity should increase the manager s motivation to maximize shareholder wealth (Jensen and Meckling, 1976), and a robust line of research has studied this relationship. Classic research by Coughlan and Schmidt (1985), Jensen and Murphy (1990), and Bizjak et al. (1993) empirically demonstrates the positive correlation 2 Delta is measured as the partial derivative of the value of the equity award with respect to stock price, where equity value is calculated using the Black-Scholes option pricing model (Core and Guay, 1999). It captures the sensitivity of the executive s awards to a 1% change in stock price. Thus, a larger delta implies greater rewards (and thereby incentive) to managers for an increase in the stock price. Delta is largely influenced by restricted stock awards and in-the-money stock options (Core and Guay, 2002). 7

9 between pay-performance sensitivity and shareholder rewards. Other research focuses on consequences of pay-performance sensitivity in specific contexts within accounting and finance. For example, Armstrong et al. (2015) demonstrate that pay-performance sensitivity reduces deviations from optimal tax avoidance. The sensitivity of a manager s firm-related wealth to changes in the stock s volatility, vega, is a proxy for risk taking incentives. 3 Investment in risky projects is a necessary, valueadding activity of the firm, but it increases the variance of potential payoff (Jensen and Meckling, 1976). Theory suggests that managers may have a risk appetite below that of a firm shareholders, making them averse to such risky investment (Jensen and Meckling, 1976). Thus in structuring compensation packages for executives, boards of directors incentivize managers to take value-adding risks in the presence of uncertainty (Prendergast, 2002). Furthermore, by reducing managers insulation from downside consequences, these incentives encourage managers to focus on positive net present value activities not just risk for risk s sake. Guay (1999) documents higher pay-risk sensitivity in firms with high growth opportunities, where risk taking is crucial. Chang et al. (2016) document similar results for new CEOs of financially distressed firms. Research has also linked executive risk taking incentives with research and development and investment, evidence of its effectiveness (Coles, Daniel, and Naveen, 2006; Gormley, Matsa, and Milbourn, 2013). However, prior research has demonstrated that equity-based incentives can also decrease the alignment between management and shareholders. For example, high equity incentives are associated with more aggressive earnings management (Bergstresser and Phillippon, 2006). Further, highly sensitive compensation can have mixed effects on risk taking. On the one hand, the effect of equity-based incentives may exacerbate overconfidence (Cooper, 3 Vega is measured as the partial derivative of the value of the award with respect to stock volatility, and captures the change in value of the award to a 0.01 change in the standard deviation of the stock s returns (Guay, 1999). It represents the increase in the value of the CEO s holdings for an increase in stock volatility, and provides an incentive to engage in risky projects. Vega is largely influenced by stock options (Guay, 1999), particularly if out of the money (Core and Guay, 2002). 8

10 Gulen, and Rau, 2014). In fact, excessively risky behavior was a hallmark of the financial crisis (Bebchuk and Spamann, 2009). On the other hand, it may prompt even greater risk aversion if managers fear losing large proportions of their wealth (Carpenter, 2000; Chen and Ma, 2011). These potentially mixed effects are highlighted in recent research by Billings et al. (2016), who identify a concave relationship between pay-risk sensitivity and return on R&D investment. Stock-based incentives induce discipline and place additional risk on management. It is no surprise then that executives may try to limit their exposure to equity-based incentives, particularly when board monitoring is weak (Yermark, 1997; Collins, Gong, and Li, 2009). For example, Coles et al. (2014) demonstrate that this weak monitoring spills over into compensation design, in that directors appointed by the CEO are associated with higher levels of CEO pay without corresponding increase in pay-performance sensitivity. Greater pay-risk sensitivity may also induce managers to gravitate toward systematic (versus idiosyncratic) risk, which they can then hedge (Armstrong and Vashishtha, 2012) Say-on-pay Votes and Equity-based Compensation We would expect to see shareholders respond significantly to equity-based incentives through their vote in the presence of three conditions. First, shareholders must view these incentives as value-relevant. Second, they must have the capacity to identify and understand these incentives as disclosed in the proxy statement. Third, they must see their vote as consequential in order to generate sufficient motivation to dedicate the resources to evaluating these incentives. We consider each of these conditions in turn in the following sections Is Equity-based Compensation Value-relevant to Shareholders? While theory suggests that equity-based compensation can mitigate agency conflicts, the notion that it will be valued by shareholders is not uniformly supported by existing 9

11 research. The research summarized in previous sections provides evidence for incentive alignment, which suggests that investors should expect equity-based incentives to be included in executive compensation packages. However, Yermack (1995) finds that equity award patterns are not strongly linked to agency risks. Hayes et al. (2012) document a dramatic decline in the use of stock options after favorable accounting treatment was eliminated by FAS 123R, implying stock options pre-fas 123R popularity did not derive solely from its monitoring power. In sum, the empirical evidence is mixed. The indirect evidence discussed above generally identifies stock-based compensation as value-relevant. Nonetheless, given the potential unintended consequences, it is an empirical question whether shareholders will view payperformance sensitivity and pay-risk sensitivity as net beneficial or costly Can Shareholders Effectively Monitor Equity-based Compensation? The effectiveness of mandated say-on-pay has been questioned by the business press, since the vast majority of firms (97% in all years 2011 to 2014) receive majority affirming votes (Carney, 2013; Semler Brossy, 2014;). Critics assert that shareholders are at an informational disadvantage, that they lack the resources to wade through these complex decisions, or that they are biased by sentiment (Bainbridge, 2008). Nonetheless, a growing body of evidence demonstrates that shareholders can serve as effective monitors of executive compensation. The empirical evidence suggests that shareholders have at least a basic understanding of compensation issues. Ertimur et al. (2011) find that shareholders can distinguish between economically-determined and excessive pay when voting on shareholder proposals related to compensation. Specific to say-on-pay, markets tend to react positively to the regulatory passage of the provision, both in the U.S. and with similar provisions implemented abroad (Cai and Walking, 2011; Trottier, 2011; Ferri and Maber, 2013). Al-Issa (2015) and Kimbro and Xu (2016) find evidence that shareholder 10

12 dissatisfaction in the U.K. and U.S., respectively, increases with the excessive portion. Furthermore, Ertimur et al. (2013) demonstrate that proxy advisors like Institutional Shareholder Services (ISS) serve to consolidate the breadth of information contained in compensation disclosures, making it easier for shareholders to process the information. Thus, despite the assertions of skeptics, the evidence to date suggests that shareholders are equipped to interpret compensation packages using the available information, at least with respect to compensation levels Are Shareholders Motivated to Monitor Equity-based Compensation? There is evidence that suggests that shareholders are motivated to take the say-on-pay vote seriously. Thomas and Cotter (2007) document increased shareholder involvement, increased support of shareholder proposals, and increased responsiveness by boards in recent years, demonstrating the growing role and perceived benefit of activism as expressed through voting. Evidence also indicates that say-on-pay does influence board behavior, even though mean voting outcomes are highly favorable. Balsam et al. (2016) document significant changes to compensation packages in anticipation of the first round of votes, including increased proportion of incentive-based pay. Al-Issa (2015) concludes that say-on-pay has been somewhat effective in the U.K. at reducing the excessive portion of compensation. Ferri and Maber (2013) also find evidence supporting the effectiveness of say-on-pay in the U.K. They report that the prevalence of certain excessive pay provisions declined and the sensitivity of CEO pay to poor performance improved after implementation (Ferri and Maber, 2013). Cotter et al. (2012) and Kimbro and Xu (2016) document similar reductions in excessive pay in the U.S. Finally, Correa and Lel (2016) performed a cross-country analysis of nearly forty countries, and determined that countries with say-on-pay laws have lower CEO compensation, lower CEO pay growth rates, and higher pay-performance sensitivity. 11

13 These are important results because they imply that say-on-pay does achieve its intended results, at least to some significant degree, and should encourage shareholders that their efforts are not wasted. Conyon and Sadler (2010), Armstrong et al. (2013), and Iliev and Vitanova (2015), however, find little evidence that voter dissent materially affects future CEO compensation design. In summary, research thus far generally provides encouraging evidence on the potential effectiveness of mandated say-on-pay voting in the U.S. The purpose of our study is to extend this understanding to pay-performance sensitivity and pay-risk sensitivity Hypotheses As noted above, we expect to observe a significant voting reaction by shareholders to equity-based financial incentives under three conditions. First, shareholders must consider equity-based incentives as value-relevant. Given the mix of prior evidence, the value shareholders place on pay-performance sensitivity and pay-risk sensitivity in say-on-pay voting is an empirical question. Nonetheless, there is reason to expect shareholders to approve of these mechanisms in compensation packages. Brickley et al. (1985) and Kato et al. (2005) document positive market responses for performance-based awards, evidenced by positive abnormal returns at the time of grant. Based on the theory and evidence discussed above, it is reasonable to expect shareholders to support greater alignment of incentives with management and therefore to approve of incorporating greater pay-performance sensitivity and risk taking incentives in compensation packages. Second, shareholders must have the ability and resources to process this complex information about compensation buried in the proxy statement. Again, evidence is generally supportive (e.g., Ertimur, Ferri, and Muslu, 2011; Ertimur, Ferri, and Oesch, 2013; Kimbro and Xu, 2016). Although agency problems associated with management or the board can never be entirely eliminated, monitoring efforts by shareholders appear to reduce them. Importantly, the say-on-pay vote and other forms of shareholder activism, such as shareholder 12

14 proposals, are considered relatively low-cost monitoring mechanisms (Ertimur, Ferri, and Muslu, 2011). Finally, shareholders must see their vote as consequential in order to generate sufficient motivation to dedicate the resources to evaluating these incentives. Given the influence executives can exercise over their boards in regards to compensation (Bebchuk and Fried, 2003), it is plausible that shareholders are motivated to vote based on their identification and interpretation of equity-based incentives. Further, there is empirical evidence that boards anticipate shareholder reactions to compensation plans (Balsam, Boone, Liu, and Yin, 2016) and react to shareholders concerns (Ferri and Maber, 2013), which implies that casting a say-on-pay vote is worthwhile. The purpose of our research is to determine whether shareholders are sufficiently adept and inclined to consider the incentive effects of equity-based awards during say-on-pay voting. We therefore study compensation as a governance tool, rather than simply a reward for past effort. Our work also informs the value-relevance of such incentives. Based on prior research on agency theory and shareholder activism, as discussed in the preceding sections, we predict the following relationship between the current year s proposed compensation package and shareholder voting. Hypothesis 1. The equity-based pay-performance sensitivity of the CEO s proposed package is positively associated with the proportion of affirming votes in the shareholder say-on-pay vote. Hypothesis 2. The equity-based pay-risk sensitivity of the CEO s proposed package is positively associated with the proportion of affirming votes in the shareholder say-on-pay vote. 13

15 3. Data and Methodology 3.1. Data Identification Our base sample comes from the ISS Voting Analytics database. We include all firmyear say-on-pay votes since the Act s implementation in January 2011 through December We measure the favorable voting proportion as the ratio of favorable votes to the vote s base (specified by the firm, typically for plus against, or for plus against plus abstain ). We log-transform the proportion of favorable votes to improve the normality of the distribution of the error term, a consequence of the left skew of voting outcomes. 4 Although shareholders cast one vote for the compensation packages for all named executives collectively, not individually, we follow prior research on say-on-pay by focusing on CEO compensation in our main analyses (Al-Issa, 2015; Correa and Lel, 2016; Kimbro and Xu, 2016). We use CEO delta as a proxy of CEO pay-performance sensitivity, and we use CEO vega as a proxy of CEO pay-risk sensitivity. These variables are calculated following Coles et al. (2013). Delta is the change in dollar value of equity incentives of the CEO for a 1% change in stock price. Vega is the change in dollar value of the CEO s incentive options for a 0.01 change in the annualized standard deviation of stock price. We scale these variables by total compensation (Execucomp TDC1 ) to reduce the firm size effect and to improve cross-firm comparability (Edmans, Gabaix, and Landier, 2009; Bushman, Dai, and Zhang, 2016). This expresses the variable as a percent instead of dollar value. Scaled delta and scaled vega are each trimmed at the first and ninety-ninth percentiles to mitigate the vulnerability of these ratios to outliers. Data requirements from Execucomp require us to focus on the S&P We test our hypotheses using OLS regression, where the log of proportion favorable votes 4 We document robustness to alternate specifications of our variables of interest and to alternate model designs in a later section. 14

16 is regressed on the CEO s scaled delta and scaled vega, plus a vector of control variables. Note that company shareholders vote at the annual meeting on the compensation package awarded during the most recent completed fiscal year. For simplicity, we consider this fiscal year end and the annual meeting immediately following to be year t = 0. Although the meeting technically occurs during fiscal year t = 1, voting relates only to executive compensation during fiscal year t = 0. For clarity, a timeline on compensation award, compensation disclosure, and shareholder voting is presented in Figure 1. [Figure 1 about here.] Based on the hypotheses stated earlier, we expect positive coefficients on the scaled equity-based incentive variables. The model (1) is summarized below: ( ) For log Base i,t ( ) Delta = β 0 + β 1 Total Compensation i,t ( ) Vega + β 2 Total Compensation i,t + γ Controls i,t + ε i,t. (1) Control variables, measured at t = 0, are as follows. We control for both the CEO total compensation and one-year change in CEO total compensation, as proxies for the level and excessive portion of compensation. We control for firm size as the log transformation of total assets, as shareholders appear to dissent at smaller firms more frequently than at larger firms (Ertimur, Ferri, and Oesch, 2013). We expect a correlation between voting behavior and profitability, as shareholders are more likely to approve of compensation packages during times of strong financial performance. Therefore, we control for industryadjusted profitability, where profitability is measured as operating income scaled by total assets. Similarly, we expect a correlation between voting behavior and actual stock return 15

17 over the period. Therefore, we control for stock return over the twelve months prior to the vote (Kimbro and Xu, 2016). This also helps control for the degree to which the stock option is in-the-money, as moneyness increases monotonically with stock price. We control for institutional ownership, using the Thompson Reuters Institutional (13f) Holdings Stock Ownership Summary. To control for board independence, we include the percent outside directors, using the GMI Companies dataset. We also control for the proportion of direct insider ownership, estimated using Thompson Reuters Insiders Data, as we expect insiders to vote in favor of their own compensation packages (Marquardt, Myers, and Niu, 2016). We include a dummy variable equal to 1 if the vote recommendation by proxy advisor ISS is against, 0 otherwise (Ertimur, Ferri, and Oesch, 2013). To control for a powerful CEO, we include a dummy variable equal to 1 if the CEO is also the chairperson of the board, 0 otherwise. We further include a dummy variable equal to 1 for financial industry firms, 0 otherwise. This serves two purposes. First, we consider this industry fundamentally different from all others in our study, since we are measuring shareholder response to risk taking incentives immediately subsequent to a financial crisis blamed on risk taking by this same industry (Bebchuk and Spamann, 2009). Second, we consider this a proxy for TARP recipients, since the majority of TARP recipients belonged to this industry (ProPublica, 2015). 5 We also include year fixed effects to the model. We cluster standard errors by year and industry (Petersen, 2009). Definitions of all variables are listed in Appendix A. 5 Firms receiving Troubled Asset Relief Program (TARP) funds were required to hold an advisory say-onpay vote at all annual shareholder meetings until and were subject to executive pay caps until the funding was repaid (Securities and Exchange Commission, 2010). 16

18 4. Results 4.1. Descriptive Statistics Descriptive statistics for our variables of interest are reported in Table 1, Panel A. The observed mean proportion of favorable votes of 90%, and mean ISS against recommendation of 11%, are consistent with prior academic literature (Ertimur, Ferri, and Oesch, 2013). Our measures of pay-performance sensitivity and pay-risk sensitivity are also generally consistent with those reported in prior literature (e.g., Bushman, Dai, and Zhang, 2016). Our sample is more profitable than industry averages, evident by the positive mean industry adjusted profitability. Our sample also had strong stock returns on average of 19%, consistent with market improvements during this time period. In Table 1, Panel B, we report and compare descriptive statistics across quartiles of our dependent variable. Scaled delta is significantly higher in the top quartile of the natural logarithm of shareholder favorable vote than in the bottom quartile, based on tests of differences of the mean (t-statistic = 6.26) and the median (Wilcoxon rank sum z-statistic = 7.34). We do not identify a significant difference in scaled vega across these quartiles, however. Thus, Panel B provides preliminary univariate support for Hypothesis 1, but does not provide significant univariate support for Hypothesis 2. [Table 1 about here.] We also visually inspect the relationship between our dependent variable and independent variables of interest using quartile plots. We plot the mean value for the dependent variable, the natural logarithm of proportion favorable votes, at each quartile of scaled delta and scaled vega in Figure 2, Panels A and B, respectively. Based on our visual inspection, we note a general positive trend across all quartiles. [Figure 2 about here.] 17

19 Pearson correlations are reported in Table 2. Scaled delta and scaled vega are significantly positively correlated at 17%. We consider this a reasonable correlation level, as we expect a positive relationship, but this relatively low correlation also suggests that these variables represent distinct constructs. 6 Scaled delta and scaled vega are each positively correlated with the proportion of favorable votes, as well. The correlation is statistically significant at 5% for scaled delta, but is not statistically significant for scaled vega. [Table 2 about here.] 4.2. Main Multivariate Results Table 3 details the results from our main model. Columns 1 to 3 present the results for scaled delta (pay-performance sensitivity) alone, scaled vega (pay-risk sensitivity) alone, and both together. We include these to verify that, absent controls, our directional predictions are supported. Indeed, we note all positive coefficients on delta and vega in these models. We then add controls in Columns 4 to 6. We focus our discussion on Column 6, which includes both independent variables of interest, all control variables, and two-way clustered standard errors by industry and year (Petersen, 2009). We consider the adjusted R 2 for our model (57%) to be reasonably high, and comparable to prior say-on-pay research (e.g., Ertimur, Ferri, and Oesch, 2013; Kimbro and Xu, 2016). First, we observe a positive coefficient on scaled delta, our proxy for pay-performance sensitivity; the coefficient is significant at the 1% level, supporting Hypothesis 1. This implies that shareholders, on average, vote more favorably for compensation packages that are more sensitive to stock price performance. It appears shareholders tend to approve of compensation packages that align manager wealth improvement with that of the 6 Our observed correlation is smaller than those identified in other research, e.g., Billings et al. (2016) (49%), because we measure our variables as percentages, not dollars, to remove the size effect (Edmans, Gabaix, and Landier, 2009; Bushman, Dai, and Zhang, 2016). 18

20 shareholders. Second, we note a positive coefficient on scaled vega, our proxy for pay-risk sensitivity, which is significant at the 10% level using a two-tailed test (t-statistic = 1.78), supporting Hypothesis 2. This implies that shareholders, on average, vote more favorably for compensation packages that are more sensitive to stock volatility. Shareholders appear to approve of compensation packages that encourage risk taking. Together, these results tell a story of shareholders approving of compensation packages that attempt to mitigate agency costs and that provide incentives to take actions that will benefit shareholders. [Table 3 about here.] It is also noteworthy that our control variables generally align with prior research and intuition. Higher overall levels of compensation garner less approval, while large firms tend to receive a greater proportion of favorable votes. Shareholders tend to approve of compensation packages when the firms report strong accounting and market performance, as evidenced by the highly significant coefficients on profitability and stock return. Further, a corporate governance story is apparent in the results. Higher institutional ownership is associated with lower favorable voting proportions. Lower favorable voting proportions also obtain when a CEO-chairman duality exists. We document an inverse relationship between independent directors and favorable votes, inconsistent with traditional views of outsider board directors as monitors (Denis, 2001), but consistent with more recent evidence of the beneficial roles of insider directors (Masulis and Mobbs, 2011). Consistent with Ertimur et al. (2013), a negative recommendation by ISS is associated with a significantly lower favorable vote proportion. Finally, we find some evidence that shareholders evaluate the financial industry uniquely. The industry is marginally associated with a lower proportion of favorable votes, consistent with targeted criticism following the financial crisis. Our independent variables of interest in Table 3 represent overall levels of payperformance sensitivity and pay-risk sensitivity. However, we also consider it appropriate to 19

21 test the effect of changes in these incentives, as we would expect shareholders to be most interested in the incentive effects of recent awards. For example, Brickley et al. (1985) and Kato et al. (2005) document a significant positive market reaction to firms that increase the incentive features of compensation design. Therefore, we decomposed each incentive into its lagged level (t = 1) and the change during the most recent period (t = 1 to t = 0). The effect of this modeling change is to allow us to estimate separate coefficients for the lagged amount and the change, and, therefore, to better interpret the differential weight placed by voter on these factors. We suggest that this decomposition better captures the impact of the most recent equity awards on shareholder perceptions of the compensation package. We make no other changes to the model. We have included a timeline in Figure 1. Table 4 reports the results of estimating this expanded model. 7 We continue to find results that are consistent with our main analysis. However, interestingly, we also document evidence of differential weights placed by shareholders on the changes and historical levels of equity incentives in their votes. In Columns 1 through 3, the statistical significance of the coefficient of the change in incentive from the prior period to the current period is greater than that of the prior period level. This is true for both pay-performance sensitivity and pay-risk sensitivity. In the fully specified model (Column 6), this conclusion holds for pay-performance sensitivity. Specifically, a positive change in scaled delta is associated with a more favorable voting outcome, statistically significant at 1%. The lagged level of scaled delta is also associated with a more favorable voting outcome, consistent with theory, but this relationship is less significant (10% using a two-tailed test). These results are consistent with shareholders approving of the most recent increase in pay-performance sensitivity. This interpretation 7 Descriptive statistics for the incentive change variables appear reasonable. The mean and median change in scaled delta are and 0.002, respectively. The mean and median change in scaled vega are and 0.000, respectively. We lose approximately 200 observations (3%) due to the requirement of scaled delta and scaled vega in both the current and most recent historical period. 20

22 does not extend to pay-risk sensitivity, however. We find that the lagged level of scaled vega is positively associated with favorable voting outcome, statistically significant at 1%. We do not find that favorable voting outcome is significantly associated at conventional levels with the change in scaled vega in the fully specified model (t-statistic = 1.34). Taken together, the results from Table 4 support our conclusions from Table 3; shareholders reward increased sensitivity of CEO pay to stock performance and risk. Table 4 further shows that shareholders appear to place distinct emphasis on historical levels and recent changes in equity incentives. Our results are consistent with shareholders assigning greater favorable weight to the recent change in pay-performance sensitivity. The evidence does not support a similar interpretation for pay-risk sensitivity, however. [Table 4 about here.] 4.3. Cross-sectional Comparisons In this section, we verify that the interpretations we make above hold in particularly salient settings. We perform additional analyses over cross-sections where we might expect distinct responses to pay-performance sensitivity and pay-risk sensitivity. First, we consider the market-to-book ratio (MTB) as a proxy for growth. Guay (1999) discusses the heightened importance of risk taking incentives in high growth firms. We split our sample at the median MTB, then run separate regressions (identical to Table 3, Column 6) for the high and low portfolios. Results are presented in Table 5, Columns 1 and 2. Scaled delta remains statistically significant only in high MTB firms. This is consistent with the relationship between voting outcome and pay-performance sensitivity being driven by shareholders of high growth firms. Unexpectedly, we do not document evidence that shareholders preference for risk taking incentives varies with growth opportunities. Scaled vega is not statistically significant in either portfolio. Thus, evidence of cross-sectional 21

23 variation based on growth opportunities is mixed. Second, we examine cross-sections of institutional ownership, as we expect institutional investors to have greater resources for more informed voting. We split the sample by median percent institutional ownership, then run separate regressions (identical to Table 3, Column 6) for each group. Results are presented in Table 5, Columns 3 and 4. We find that pay-risk incentives significantly positively associated with favorable votes only in the high institutional ownership group. This is consistent with our expectation of enhanced sophistication of institutions when evaluating compensation. Another notable observation is the difference in adjusted R 2 between these groups. Adjusted R 2 is notably higher for the high institutional ownership group. The voting behavior of the low institutional ownership group (consisting of greater retail investor populations) appears to be less predictable. Nonetheless, the results for low institutional ownership remain encouraging. Although scaled vega loses significance, shareholders of these firms still react significantly to scaled delta. Thus, it appears equitybased incentives are still important to diverse, dispersed shareholder bases. [Table 5 about here.] Finally, we consider cross-sections of performance. Prior research, as well as our own, documents a positive relationship between shareholder voting and performance. Prior research also documents a positive correlation between performance and sentiment (e.g., Talakai, 2016). During periods of negative sentiment, shareholders tend to scrutinize management more closely (Brown, Christensen, Elliott, and Mergenthaler, 2012), but actions to increase pay-performance sensitivity can mollify outrage (Abernethy, Kuang, and Qin, 2015). We expect that when performance is weak, shareholders will place greater emphasis on equity compensation in their votes. We measure two specifications of performance. First, we split the sample by median return on assets, measured as operating income over total assets, then run separate 22

24 regressions (identical to Table 3, Column 6) for each group. Results are presented in Table 5, Columns 5 and 6. Scaled delta is a strong positive predictor of voting outcome in both groups, and the coefficients do not significantly differ across the groups. However, scaled vega is only statistically significant in the low profitability group. In fact, the coefficient for scaled vega is significantly greater for the low profitability group than for the high profitability group using a test of seemingly unrelated regressions. Second, we split the sample by stock returns greater than 0 (gains) and less than 0 (losses) over the twelve months prior to the vote, then run separate regressions (identical to Table 3, Column 6) for each group. Results are presented in Table 5, Columns 7 and 8. Scaled delta is significant at the 1% level in both models. However, these coefficients are not significantly different across the two regressions. Scaled vega is less significant in both separately-estimated cross-sectional models than that obtained using a pooled model. Thus, evidence of the relationship between incentives and voting, conditional on financial or stock performance, is mixed. The results in Table 5 provide complementary evidence to Table 3. We first demonstrate that shareholders tend to approve of equity-based incentives across a broad sample of firms. While the full sample study is informative, theory predicts that agency risks and shareholder response to those risks differ across firms. Table 5 provides evidence that our results are generally stronger in firms where demand for pay-performance sensitivity and pay-risk sensitivity are expected to be higher, as represented by the presence of growth opportunities, the level of sophisticated ownership, and the incidence of weak performance Equity Awards of the CFO and Other Named Executives In our main analyis, we follow prior literature, and focus on the equity-based incentives of the CEO (Ertimur, Ferri, and Oesch, 2013; Al-Issa, 2015; Balsam, Boone, Liu, and Yin, 2016; Correa and Lel, 2016; Kimbro and Xu, 2016), as the CEO is generally the most visible 23

25 member of management. However, the Dodd-Frank Act stipulates that shareholders vote on the compensation packages of all named executives collectively, not individually (Securities and Exchange Commission, 2011). We recognize that all executives are vulnerable to agency conflicts, and equity-based incentives are commonly used to incentivize teams across top management (Bushman, Dai, and Zhang, 2016). We further recognize that the specific agency conflicts faced and the nature of incentives to address them may differ based on the executive s role. Shareholders opinions may differ, as well, based on the executive s role. Therefore in this section, we extend beyond the methods of prior literature by expanding our sample to all named ( Top Five ) executives reported in Execucomp. We consider the CFO of particular interest. Shareholders sensitivity to equity-based incentives of the CFO is unclear ex ante. On the one hand, shareholder may prefer lower equity-based incentives for the CFO to promote a commitment to financial reporting quality (Jiang, Petroni, and Wang, 2010). On the other hand, shareholders may prefer greater equity-based incentives for the CFO. The CFO historically has significant authority over tax planning, and equity-based incentives have been linked to optimal tax planning (Armstrong, Blouin, Jagonlinzer, and Larcker, 2015). In recent years, the CFO has also demonstrated a growing role in strategic planning (Russomanno, 2014), in which incentive alignment may be more critical (Core and Guay, 1999; Guay 1999). Alternatively, it is possible that the incentives of the CFO are simply not important enough to shareholders to trigger any significant response in voting. We estimate scaled delta and scaled vega for each named executive. We keep as separate variables scaled delta and scaled vega of the CEO and CFO. We add control variables for the level and change in CFO total compensation. We aggregate scaled delta and scaled vega of all other named executives in two ways. First, we calculate average scaled delta and average scaled vega, averaged over all non-ceo, non-cfo named executives by firm-year. Second, we calculate summed scaled delta and summed scaled vega, where we sum over all non-ceo, non-cfo named executives in a given firm-year. We regress the log-transformed 24

26 percent favorable vote on this vector, plus controls consistent with our main analysis in Table 3. Results are reported in Table 6. In Column 1, we report the reduced model results, controlling only for CEO incentives. CFO scaled delta is positively related to proportion favorable vote, while CFO scaled vega is negatively related to proportion favorable vote; both are statistically significant at 1%. This is consistent with a preference for pay-performance sensitivity, but a preference against risk taking incentives for the CFO. Columns 2 and 3 add controls, including the two variations of equity-based incentives of the remaining top executive team (averaged and summed, respectively). CEO scaled delta and scaled vega remain highly statistically significant. Moreover, CFO scaled delta is positively related to proportion favorable vote, statistically significant at 1% and 5% in Columns 2 and 3, respectively. The relationship between CFO scaled vega and proportion favorable vote is not significant in either column. We also observe that shareholders reward payperformance sensitivity of remaining top executives. We document a positive relationship between scaled delta of all other named executives and percent favorable vote, statistically significant at 1% and 5% in Columns 2 and 3, respectively. Collectively, these results support an interpretation that pay-performance sensitivity of top management is an important determinant of shareholder say-on-pay votes, beyond just the CEO. We do not document evidence, however, that pay-risk sensitivity of executives other than the CEO significantly influences the voting outcome. [Table 6 about here.] To our knowledge, we are the first to expand our analysis of say-on-pay voting beyond the CEO. The results of this section support our main conclusions. Shareholders appear to incorporate equity-based incentives of management into their voting decisions. Their voting determinants are not limited to just the CEO, however. We demonstrate in Table 25

27 6 that increased pay-performance sensitivities for each of the CEO, CFO, and remaining top executives incrementally explains more favorable voting outcomes. Furthermore, the shareholder response is nuanced and consistent with the distinct roles and agency risks of top management, evidenced by distinct reactions to CEO and CFO pay-risk sensitivity. These results speak to the sophistication of shareholders, even in the presence of a voluminous and complex information Diminishing Returns of Equity-based Incentives Empirical evidence indicates that there are mixed effects of pay-performance sensitivity and pay-risk sensitivity (e.g., Bergstresser and Phillippon, 2006; Armstrong and Vashishtha, 2012). Notably, research suggests that equity-based incentives can have diminishing returns to firm value. For example, Hanlon et al. (2003) document a positive but concave relationship between stock option grants and future operating income. Billings et al. (2016) document a similar relationship between pay-risk sensitivity and return on risky R&D investment. These studies imply that the benefits of equity-based compensation on performance erode when that incentive reaches very high levels. In this section, we consider whether shareholders incorporate possible diminishing returns to equity-based incentives into their votes. If shareholders uniquely adjust their assessment of compensation when equity-based incentives are very high, we would expect the slope of the pay-performance sensitivity and/or pay-risk sensitivity to become less positive when they are at high levels. We address the research question two ways. First following Billings et al. (2016), we include as additional independent variables of interest squared terms for both scaled delta and scaled vega. 8 The data would be consistent with diminishing returns of equity-based incentives if we observe a positive slope on the incentive and a negative slope on its square. Second, we add a dummy 8 We use a quadratic term as a proxy of diminishing returns. We make no theoretical predictions for a peak or optimal equity-based incentive. 26

28 variable equal to one if the equity-based incentive is in the top decile and interact the dummy with the scaled level. Again, diminishing returns imply a positive slope on the incentive and a negative slope on the interaction. Control variables are consistent with Table 3. Results using the quadratic model are presented in Table 7, Columns 1 to 3. In the fully specified model, we document a positive coefficient on scaled delta and a negative coefficient on its quadratic term, both significant at 1%. We do not, however, find a similar pattern for pay-risk sensitivity. Scaled vega and its quadratic term are not statistically significant at conventional levels. Results using the interaction model are presented in Columns 4 to 6, and tell a similar story. Again in the fully specified model, scaled delta remains positive and its interaction with the top decile dummy is negative, both significant at 1%. Scaled vega and its interaction term are not statistically significant. Recent research suggests that the marginal benefits of equity-based incentives are diminishing. Our results provide some support of this argument from the direct perspective of the shareholder. The evidence is consistent with shareholders incorporating the mixed effects of pay-performance sensitivity documented in prior research into their vote. We do not observe such an effect for pay-risk sensitivity. [Table 7 about here.] 4.6. The Effect of Say-on-pay Voting on Subsequent Incentives Regulators appear to advocate greater alignment of manager and shareholder wealth (US Senate Committee on Banking, 2010; Securities and Exchange Commission, 2015). Therefore, to understand the effectiveness of say-on-pay as a mandate, we consider whether shareholder voting influences subsequent incentive design. There is evidence to suggest that boards are responsive to the say-on-pay vote. For example, boards increased the proportion of equity-based compensation in executives pay packages in 2010 in anticipation for the 27

29 first mandatory say-on-pay vote (Balsam, Boone, Liu, and Yin, 2016). Research covering the period subsequent to implementation documents changes in compensation design after dissenting votes, most notably through excessive pay (Cotter, Palmiter, and Thomas, 2012; Ferri and Maber 2013; Al-Issa 2015; Kimbro and Xu 2016). Further, Ferri and Maber (2013), using U.K. data, document that firms with high dissenting votes reduced severance benefits and reduced the practice of retesting (tempering performance targets after expectations for performance worsen), consistent with improved pay-performance sensitivity. To our knowledge, however, prior research has not directly examined the effect of shareholder dissent on equity-based incentives in the U.S. Given regulators objectives with the Act, we consider this an important question to address. To test for an effect of say-on-pay voting outcome on subsequent changes to equity-based incentives, we implement the following model. We measure the change in the CEO s equitybased incentives (change in scaled delta and change in scaled vega) from period t = 0 to t = 1, where t = 0 is the fiscal year for which shareholders vote on compensation and t = 1 is the subsequent fiscal year. We regress this change in incentive on a measure of shareholder satisfaction at the annual meeting at t = 0 using two proxies. First, we use the continuous variable from our main analysis, the natural logarithm of the percent favorable vote. If the board of directors increases pay-performance sensitivity and pay-risk sensitivity of the CEO in response to low favorability, we would expect the coefficient on this variable to be negative. Second, we create a dummy variable equal to 1 if the percent favorable vote is less than the tenth percentile. If the board of directors increases pay-performance sensitivity and pay-risk sensitivity of the CEO when shareholder dissent is especially high, we would expect the coefficient to be positive. Controls are other variables expected to relate to the change in equity-based incentives. We control for an ISS against recommendation at the same time as the shareholder vote (t = 0), as firms may be more likely to adjust compensation if ISS disapproves. We control for 28

30 the change in natural logarithm of total assets, the change in industry-adjusted profitability, the change in market-to-book ratio, and the change in the CEO s total compensation, all measured from t = 0 to t = 1. We control for the stock return over the fiscal year t = 1, as change in compensation may be related to stock performance. Finally we control for the governance environment during t = 1, measured through percent independent directors and a dummy for a dual CEO-chairman, as firms may be less responsive to shareholder votes if governance is weak. Recall a timeline is presented in Figure 1. We include year fixed effects and cluster standard errors by industry and year (Petersen, 2009). Due to the nature of this research design, we lose one year of data (i.e., votes taking place in 2014). Results are reported in Table 8. In Columns 1 and 2, we document evidence consistent with the board increasing pay-performance sensitivity in response to sizable shareholder dissent. We document a negative relationship between the percent favorable vote and subsequent change in scaled delta, statistically significant at the 5% level. We document a positive relationship between a low favorability dummy and subsequent change in scaled delta, statistically significant at 1%. These results are consistent with increased payperformance sensitivity when shareholder satisfaction is low. We do not document similar results for pay-risk sensitivity. The change in scaled vega is not significantly related to the percent favorability or the low favorability dummy at conventional levels. Thus, evidence is mixed. While boards appear to adjust pay-performance sensitivity after shareholder dissent, we do not find evidence to support the notion that they adjust pay-risk sensitivity. [Table 8 about here.] 4.7. Robustness Tests We perform the following series of untabulated tests to verify the robustness of our conclusions to alternate variable definitions and model designs. Robustness tests modify our 29

31 primary model (Table 3, Column 6). Our conclusions regarding pay-performance sensitivity are very robust. Our conclusions regarding pay-risk sensitivity are more sensitive. First, we consider alternate specifications of our dependent variable. We test the robustness of our results to an untransformed proportion favorable vote, an arcsine transformation, and an aligned rank transformation. Our conclusions regarding scaled delta are unaffected. Scaled vega remains statistically significant for the untransformed proportional vote, less significant for arcsine transformation (t-statistic = 1.43), and statistically insignificant at conventional levels for the aligned rank transformation. Second, we consider the following alternate specifications of our explanatory variables of interest. We test the robustness of our results to a model wherein delta and vega are unscaled and log-transformed and a model wherein delta and vega are unscaled and untransformed. The alternate specifications of delta remain statistically significant. Unscaled, log-transformed vega is statistically insignificant at conventional levels; unscaled, untransformed vega is also insignificant. Third, we consider the following modifications to our model. Following Balsam et al. (2016), we confirm in an untabulated test that our results are robust to a tobit model, with interpretations unchanged. We run our model with financial firms excluded entirely. Our results are qualitatively similar, although our results for scaled vega are less significant (t-statistic = 1.51). of our control for total compensation. We consider the following alternate specification We separately identify the components of total compensation cash, equity, and other allowing the coefficients on levels of compensation to vary by compensation type. (Execucomp variables SALARY We include in the cash component salary and bonus and BONUS). We include in the equity component restricted stock grants and option grants (Execucomp variables STOCK AWARDS FV and OPTION AWARDS FV ). We include all other components of TDC1 in the other component (Execucomp variables NONEQ INCENT, DEFER RPT AS COMP TOT, and 30

32 OTHCOMP). Our conclusions remain unchanged. We also consider an alternate specification of our control for excess compensation, where we include a dummy variable equal to 1 if an observation is above the median of excessive compensation, where excessive compensation is defined by Al-Issa (2015), and 0 otherwise; our conclusions are unchanged. Fourth, we verify the robustness of our conclusions to the inclusion of several additional control variables. We control for compensation mix, defined by Balsam et al. (2016) and trimmed at the first and ninety-ninth percentiles, to verify that our equity-based incentive variables capture distinct and incrementally significant constructs. Our results of this robustness test are qualitatively similar, although our results for scaled vega are less significant (t-statistic = 1.37). Next, we control for say-on-pay vote frequency, as the Act mandates that shareholders may elect to hold votes every one, two, or three years (Securities and Exchange Commission, 2011). We find that the vast majority of firms in our sample hold annual say-on-pay votes (over 94% of firm-year observations and over 90% of unique firms). Nonetheless, as a robustness test, we rerun the model with an additional control variable for firm vote frequency equal to 1 for firms with biennial or triennial votes, and 0 otherwise. Our conclusions are unchanged. We also add controls for other firm-specific factors that theoretically may influence shareholder perception of compensation. We add a dummy variable equal to 1 if there was CEO turnover, identified via Execucomp, during the current year. We also add a dummy variable equal to 1 if there was significant M&A activity, identified via Compustat footnotes, during the current year. We further add control variables for shareholder turnout (total votes cast as a percent of shares outstanding) as a measure of an active shareholder body and the standard deviation of monthly stock returns during the twelve months immediately prior to the annual meeting as a measure of risk. After controlling for each of these factors, our results are qualitatively similar, and our inferences are unchanged. 31

33 5. Conclusion In this study, we examine the relationship between equity-based incentives and shareholder voting outcomes on say-on-pay votes, as mandated by the Dodd-Frank Act. We find that higher CEO pay-performance sensitivity is associated with more favorable voting outcomes. We find similar results for CEO pay-risk sensitivity. We also note several observations from additional analyses, particularly regarding payperformance sensitivity. We document that this relationship is heavily influenced by the most recent change in the change in pay-performance sensitivity. We document some evidence that our results are stronger in settings where demand for these incentives is expected to be higher: high growth firms, firms with sophisticated shareholder bases, and firms with weak performance. We also expand our sample to executives beyond the CEO, and find that pay-performance sensitivity of the CFO and other named executives incrementally explain shareholder favorable voting outcomes. We find a concave relationship between shareholder voting and pay-performance sensitivity, implying that shareholders approve of the use of equity-based incentives, but this approval tapers off at high levels. Finally, we document evidence consistent with the interpretation that firms increase pay-performances sensitivity of the CEO in response to a significantly unfavorable vote. Our study supports the notion that the Dodd-Frank Act gives shareholders a significant voice in the design of compensation packages, and, by extension, a role in aligning shareholder and manager interests. Thus, our results should be of interest to regulators. Notably, the Dodd-Frank Act also mandated disclosure of pay-performance sensitivity in narrative form in the annual proxy statement, with the objective to help shareholders in their say-on-pay voting decision and to improve pay-performance sensitivity economy-wide (Securities and Exchange Commission, 2015). 9 This mandate implies and our results corroborate that shareholders 9 At the time of writing, the SEC had proposed regulations implementing the Act s mandate and was processing public comments (Securities and Exchange Commission, 2015). 32

34 view the equity-based incentives at the center of this disclosure mandate as value-relevant. Furthermore, our findings over the changes of equity-based incentives after the say-on-pay vote imply that shareholders can serve as effective monitors of pay-performance sensitivity. This speaks to the effectiveness of mandated say-on-pay. However, our contribution is not limited to a test of the say-on-pay regulation. We document evidence that shareholders, on average, approve of equity-based incentives. While there is a long line of literature devoted to these incentives, we are the first to our knowledge to directly examine shareholder opinion of the matter. Prior research has demonstrated mixed evidence regarding the consequences of equity-based incentives (Jensen and Murphy, 1990; Bergstresser and Phillippon, 2006). Our results indicate that shareholders view the incentives as important and valuable. Our analysis of diminishing returns to equity incentives, however, implies that shareholders understand the tradeoff between the benefits and potential costs of incentives accompanying equity-based compensation. We also contribute to the broad line of research on shareholder activism. Our findings support prior academic research that shareholders are neither apathetic nor uninformed in their monitoring role (Ertimur, Ferri, and Muslu, 2011). Instead, our results support the notion that shareholders want a voice in the design and administration of executive compensation, and express their approval or disapproval of compensation through their vote. Our findings also support prior research that shareholder voting can effect governance changes (Ferri, 2012). In sum, the U.S. say-on-pay mandate represents a unique setting to evaluate shareholder activism over executive compensation. The mandatory nature of the say-on-pay vote vastly increases the sample size and reduces the endogeneity associated with firms choice to hold a vote. Thus, the setting provides power and generalizability. The design of executive compensation contracts is complex. We recommend research regarding shareholder perceptions of other aspects of compensation design, as well as the long-term effectiveness 33

35 of say-on-pay, among other voting mechanisms, to align shareholders and managers. For example, our study may also encourage further research into more complex and nuanced measures of equity-based incentives, including deviations from optimal pay-performance sensitivity and pay-risk sensitivity (Core and Guay, 1999) or dispersion of incentives across executive teams (Bushman, Dai, and Zhang, 2016). To our knowledge, we are the first sayon-pay researchers to extend our results to the entire named executive team. Our results may encourage additional study of shareholder voting regarding non-ceo executives. Thus, the field remains ripe for future research. 34

36 Appendix A. Variable Definitions This appendix details the variable construction for our empirical analyses. Compustat/Execucomp variable names (Xpressfeed format) are in Italics. All other data sources are given in brackets. Voting Base = The share count to which for votes are compared, determined by the company. Voting base may be For + Against, For + Against + Abstain, or Outstanding Shares as of the meeting date [ISS Voting Analytics]. For / Voting Base = Number of for votes Voting Base. We log-transform this ratio for analysis [ISS Voting Analytics]. Delta = We follow Coles et al. (2013) to calculate the change in the executive s wealth for a 1% change in the firm s stock price. Data is sourced from Execucomp. For our analysis, we scale delta by TDC1. Vega = We follow Coles et al. (2013) to calculate the change in the executive s wealth for a 0.01 change in the firm s stock volatility. Data is sourced from Execucomp. For our analysis, we scale vega by TDC1. Total Compensation = TDC1, expressed in millions of dollars. Total Compensation = Change in total compensation from fiscal year t = 1 to fiscal year t = 0. Log (Assets) = Natural logarithm of total assets (AT ). Profitability = Operating income before depreciation (OIBDP) Total Assets (AT ). Industry Adjusted Profitability = Profitability less industry mean profitability, where industry mean is calculated by 4-digit SIC code across the entire Compustat universe. Meeting Cumulative Return = 12 month return [monthly CRSP], ending on the vote meeting date [ISS Voting Analytics]. Institutional Ownership = Shares owned by institutions as of the most recent report prior 35

37 to the meeting date [Thomson Reuters 13F Institutional Holdings database] Outstanding Shares as of the meeting date [ISS Voting Analytics], multiplied by 100. Independent Directors on Board = Percent of outside directors on the board [GMI Ratings]. Insider Ownership = Total shares owned by insiders [Thomson Reuters Insider Filing database] Outstanding Shares as of the meeting date [ISS Voting Analytics]. ISS Against Dummy = A dummy variable which is equal to 1 if ISS recommends against and 0 otherwise [ISS Voting Analytics]. CEO-Chairperson Dummy = A dummy variable which is equal to 1 if CEO is also the chairperson of the board of directors and 0 otherwise [GMI Ratings]. Financial Firm Dummy = A dummy variable which is equal to 1 if SIC code is and 0 otherwise. Market-to-Book = [Stock Price (PRCC F ) Shares Outstanding (CSHO) + Total Debt (DLTT+DLC ) + Preferred Stock Liquidating Value (PSTKL) Deferred Taxes and Investment Tax Credits (TXDITC )] Total Assets (AT ). High Delta Dummy = A dummy variable that is equal to 1 if scaled delta is in the top decile and 0 otherwise. High Vega Dummy = A dummy variable that is equal to 1 if scaled vega is in the top decile and 0 otherwise. Low Favor Dummy = A dummy variable that is equal to 1 if percent favorable vote is in the bottom decile and 0 otherwise. Fiscal Year Cumulative Return = 12 month return [monthly CRSP] ending on fiscal year end date. Excessive Compensation = Following Al-Issa (2015), the residual of a regression of the natural logarithm of total compensation (TDC1 ) on the natural logarithm of tenure [Execucomp], the nature logarithm of lagged sales (SALE), annual returns over both the 36

38 current fiscal year and lagged fiscal year [CRSP], income before extraordinary items (IBC) divided by average total assets (AT ) for both the current and lagged fiscal year, lagged MTB (defined above), and industry and time fixed effects, estimated across the entire Compustat universe. Vote Frequency Dummy = A dummy variable that is equal to 1 if the firm reported a say-on-pay frequency of two or three years and 0 otherwise [ISS Voting Analytics]. Compensation Mix = Following Balsam et al. (2016), [Total Direct Compensation (TDC1 ) Salary (SALARY ) All Other Compensation (OTHCOMP) Change in Pension Value and Nonqualified Deferred Compensation Earnings (PENSION CHG)] Salary (SALARY ). CEO Turnover Dummy = A dummy variable that is equal to 1 if the CEO identified in Execucomp is a different individual than that of the prior fiscal year. M&A Dummy = A dummy variable that is equal to 1 if Compustat SALE FN is coded AA or AB. Shareholder Turnout = [Number of votes for + number of votes against ] Outstanding Shares as of the meeting date [ISS Voting Analytics]. Standard Deviation of Stock Returns = Standard deviation of monthly stock returns over the 12 month period ending the date of the annual meeting [CRSP]. 37

39 References [1] Abernethy, M. A., Kuang, Y. F., and Qin, B. (2015). The Influence of CEO Power on Compensation Contract Design. The Accounting Review, 90(4), [2] Al-Issa, W. M. (2015). Boards Response to Shareholders Dissatisfaction: The Case of Shareholders Say on Pay in the UK. European Accounting Review, 24(4), [3] Armstrong, C. S., and Vashishtha, R. (2012). Executive stock options, differential risktaking incentives, and firm value. Journal of Financial Economics, 104, [4] Armstrong, C. S., Gow, I. D., and Larcker, D. F. (2013). The Efficacy of Shareholder Voting: Evidence from Equity Compensation Plans. Journal of Accounting Research, 51(5), [5] Armstrong, C. S., Blouin, J. L., Jagolinzer, A. D., and Larcker, D. F. (2015). Corporate Governance, Incentives, and Tax Avoidance. Journal of Accounting and Economics, 60(1), [6] Bainbridge, S. M. (2008). Remarks on Say on Pay: An Unjustified Incursion on Director Authority. UCLA School of Law Law and Economics Research Paper Series(08-06). [7] Balsam, S., Boone, J., Liu, H., and Yin, J. (2016). The impact of say-on-pay on executive compensation. Journal of Accounting and Public Policy, 35, [8] Bebchuk, L. A., and Fried, J. M. (2003). Executive Compensation as an Agency Problem. National Bureau of Economic Research(9813). [9] Bebchuk, L. A., and Spamann, H. (2009). Regulating Bankers Pay. The Harvard John M. Olin Discussion Paper Series(641). [10] Bebchuk, L. A., Grinstein, Y., and Peyer, U. (2010). Lucky CEOs and Lucky Directors. The Journal of Finance, LXV(6). [11] Bergstresser, D., and Phillippon, T. (2006). CEO incentives and earnings management. Journal of Financial Economics, 80, [12] Billings, B. K., Moon, J. R., Morton, R. M., and Wallace, D. M. (2016). Risk without Rewards? An Analysis of Risk-Taking Incentives and the Value of Risky Investment. Working Paper. 38

40 [13] Bizjak, J. M., Brickley, J. A., and Coles, J. L. (1993). Stock-based incentive compensation and investment behavior. Journal of Accounting and Economics, 16, [14] Brickley, J. A., Bhagat, S., and Lease, R. C. (1985). The Impact of Long-range Managerial Compensation Plans on Shareholder Wealth. Journal of Accounting and Economics, 7, [15] Brown, N. C., Christensen, T. E., Elliott, W. B., and Mergenthaler, R. D. (2012). Investor Sentiment and Pro Forma Earnings Disclosures. Journal of Accounting Research, 50(1), [16] Bushman, R., Dai, Z., and Zhang, W. (2016). Management Team Incentive Dispersion and Firm Performance. The Accounting Review, 91(1), [17] Cai, J., and Walking, R. A. (2011). Shareholders Say on Pay: Does It Create Value? Journal of Financial and Quantitative Analysis, 46, [18] Carney, J. (2013, 07 03). Why Say on Pay Failed and Why That s a Good Thing. Retrieved from CNBC: [19] Carpenter, J. N. (2000). Does option compensation increase managerial risk appetite? The Journal of Finance, 55(5), [20] Chang, W.-J., Hayes, R. M., and Hillegeist, S. A. (2016). Financial Distress Risk and New CEO Compensation. Management Science, 62(2), [21] Chen, Y.-R., and Ma, Y. (2011). Revisiting the risk-taking effect of executive stock options on firm performance. Journal of Business Research, 64, [22] Coles, J. L., Daniel, N. D., and Naveen, L. (2006). Managerial Incentives and Risk- Taking. Journal of Financial Economics, 79, [23] Coles, J. L., Daniel, N. D., and Naveen, L. (2013). Calculation of Compensation Incentives and Firm-related Wealth using Execucomp: Data, Program, and Explanation. Working Paper. [24] Coles, J. L., Daniel, N. D., and Naveen, L. (2014). Co-opted Boards. The Review of Financial Studies, 27(6),

41 [25] Collins, D. W., Gong, G., and Li, H. (2009). Corporate Governance and Backdating of Executive Stock Options. Comtemporary Accounting Research, 26(2), [26] Conyon, M., and Sadler, G. (2010). Shareholder Voting and Directors Remuneration Report Legislation: Say on Pay in the UK. Corporate Governance: An International Review, 18(4), [27] Cooper, M. J., Gulen, H., and Rau, P. R. (2014). Performance for pay? The relation between CEO incentive compensation and future stock price performance. Working Paper. [28] Core, J. E., Holthausen, R. W., and Larcker, D. F. (1999). Corporate governance, chief executive officer compensation, and firm performance. Journal of Financial Economics, 51, [29] Core, J., and Guay, W. (1999). The use of equity grants to manage optimal equity incentive levels. Journal of Accounting and Economics, 28, [30] Core, J., and Guay, W. (2002). Estimating the Value of Employee Stock Option Portfolios and Their Sensitivities to Price and Volatility. Journal of Accounting Research, 30(3), [31] Correa, R., and Lel, U. (2016). Say on pay laws, executive compensation, pay slice, and firm valuation around the world. Journal of Financial Economics, 122, [32] Cotter, J. F., Palmiter, A. R., and Thomas, R. S. (2012). The First Year of Say-on-Pay Under Dodd Frank: An Empirical Analysis and Look Forward. Vanderbilt Law and Economics Research Paper(12-32). [33] Coughlan, A. T., and Schmidt, R. M. (1985). Executive Compensation, Management Turnover, and Firm Performance. Journal of Accounting and Economics, 7, [34] Denis, D. K. (2001). Twenty-five years of corporate governance research... and counting. Review of Financial Economics, 10, [35] Edmans, A., Gabaix, X., and Landier, A. (2009). A Multiplicative Model of Optimal CEO Incentives in Market Equilibrium. The Review of Financial Studies, 22(12),

42 [36] Ertimur, Y., Ferri, F., and Muslu, V. (2011). Shareholder Activism and CEO Pay. The Review of Financial Studies, 24, [37] Ertimur, Y., Ferri, F., and Oesch, D. (2013). Shareholder Votes and Proxy Advisors: Evidence from Say on Pay. Journal of Accounting Research, 51(5), [38] Ferri, F. (2012). Low-Cost Shareholder Activism: A Review of the Evidence. In C. A. Hill and B. H. McDonnell, Research Handbook on the Economics of Corporate Law (pp ). Edward Elgar. [39] Ferri, F., and Maber, D. A. (2013). Say on Pay Votes and CEO Compensation: Evidence from the UK. Review of Finance, 17, [40] Financial Accounting Standards Board. (2010). Conceptual Framework for Financial Reporting. Statement of Financial Accounting Concepts, No. 8: Chapter 1, The Objective of General Purpose Financial Reporting. Financial Accounting Foundation. [41] Gormley, T. A., Matsa, D. A., and Milbourn, T. (2013). CEO compensation and corporate risk: Evidence from a natural experiment. Journal of Accounting and Economics, 56, [42] Guay, W. (1999). The sensitivity of CEO wealth to equity risk: an analysis of the magnitude and determinants. Journal of Financial Economics, 53, [43] Hanlon, M., Rajgopal, S., and Shevlin, T. (2003). Are executive stock options associated with future earnings? Journal of Accounting and Economics, 36, [44] Hayes, R. M., Lemmon, M., and Qiu, M. (2012). Stock options and managerial incentives for risk taking: Evidence form FAS 123R. Journal of Financial Economics, 105, [45] Hölmstrom, B. (1979). Moral Hazard and Observability. The Bell Journal of Economics, 10(1), [46] Iliev, P., and Vitanova, S. (2015). The Effect of the Say-on-Pay Vote in the U.S. Working Paper. [47] Jensen, M. C., and Meckling, W. H. (1976). Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure. Journal of Financial Economics, 3,

43 [48] Jensen, M. C., and Murphy, K. J. (1990). Performance Pay and Top-Management Incentives. Journal of Political Economy, 98(2), [49] Jiang, J. X., Petroni, K. R., and Wang I. Y. (2010). CFOs and CEOs: Who have the most influence on earnings management? Journal of Financial Economics, 96, [50] Kato, H. K., Lemmon, M., Luo, M., and Schallheim, J. (2005). An empirical examination of the costs and benefits of executive stock options: Evidence from Japan. Journal of Financial Economics, 78, [51] Kimbro, M. B., and Xu, D. (2016). Shareholders have a say in executive compensation: Evidence from say-on-pay in the United States. Journal of Accounting and Public Policy, 35, [52] Marquardt, B. B., Myers, B., and Niu, X. (2016). Strategic Voting and Insider Ownership. Working Paper. [53] Masulis, R. W., and Mobbs, S. (2011). All All Inside Directors the Same? Evidence from the External Directorship Market. The Journal of Finance, LXVI(3). [54] Murphy, K. J. (2013). Executive Compensation: Where We Are, and How We Got There. In G. Constantinides, M. Harris, and R. Stulz, Handbook of the Economics of Finance (pp. Chapter 4: ). Elsevier Science Noth Holland. [55] Petersen, M. A. (2009). Estimating Standard Errors in Finance Panel Data Sets: Comparing Approaches. The Review of Finanical Studies, 22(1), [56] Prendergast, C. (2002). The Tenuous Trade-Off between Risk and Incentives. The Journal of Political Economy, 110(5), [57] ProPublica. (2015, 4 20). Bailout Tracker. Retrieved from org/bailout/list/simple [58] Russomanno, D. (2014). Changes in Internal Controls and Auditor Effort around the initiation of a Chief Accounting Officer (CAO) Position. Dissertation. Pennsylvania State University. [59] Securities and Exchange Commission. (2010). Shareholder Approval of Executive Compensation of TARP. 17 CFR PART 240 ( ). 42

44 [60] Securities and Exchange Commission. (2011). Shareholder Approval of Executive Compensation and Golden Parachute Compensation. 17 CFR Parts 229, 240, 249 ( , ). [61] Securities and Exchange Commission. (2015). Pay versus Performance. 17 CFR Parts 229, 240, 249 ( ). [62] Semler Brossy. (2014, 09 10) Say on Pay Results. Retrieved from semlerbrossy.com/ [63] Talakai, J. (2016) The Impact of Sentiment on Investor Information Gathering and Price Discovery: Evidence from Mergers and Acquisitions. Working Paper. [64] Thomas, R. S., and Cotter, J. F. (2007). Shareholder proposals in the new millennium: Shareholder support, board response, and market reaction. Journal of Corporate Finance, 13, [65] Trottier, K. (2011). What Does the Market Say About Say-on-Pay? A Look at the Canadian Bank Experience. Working Paper. [66] US Senate Committee on Banking, H. a. (2010, 07 01). Brief Summary of the Dodd- Frank Wall Street Reform and Consumer Protection Act. Retrieved from banking.senate.gov/ [67] Yermack, D. (1995). Do corporations award CEO stock options effectively? Journal of Financial Economics, 39, [68] Yermack, D. (1997). Good Timing: CEO Stock Option Awards and Company News Announcements. The Journal of Finance, LII(2). 43

45 Say-on-Pay vote (t = 1) Say-on-Pay vote (t = 0) Say-on-Pay vote (t = 1) Fiscal year end (t = 1) Fiscal year end (t = 0) Fiscal year end (t = 1) Fig. 1. Timeline. This figure presents a visual depiction of the timeline of events involved in our empirical analyses, and is included to ensure clarity. We compute our independent variables of interest, scaled delta and scaled vega, from the data in the compensation disclosures reported in the proxy statement immediately after fiscal year end. Shareholders vote on this compensation at the annual shareholder meeting shortly after the proxy statement release. We consider the compensation data for a given fiscal year and the shareholder vote over this same data to occur at t = 0. 44

46 Panel A: Pay-performance Sensitivity (Scaled Delta) Panel B: Pay-risk Sensitivity (Scaled Vega) Fig. 2. Quartile Plots of Proportion Favorable Votes and Equity-Based Incentives. This figure presents visual depictions of the univariate relationship between proportion favorable votes and pay-performance sensitivity. In Panel A, the x-axis divides the data into four quartiles (where 1 is the lowest and 4 is the highest) of scaled delta, CEO wealth delta scaled by CEO total compensation. In Panel B, the x-axis divides the data into four quartiles (where 1 is the lowest and 4 is the highest) of scaled vega, CEO wealth vega scaled by CEO total compensation. The y-axis in both panels is the mean of our dependent variable of interest, the natural logarithm of the proportion of votes in favor, at the given quartile plus 0.15 to shift all observations above 0 to ease interpretation. 45