The Effect of the Say-on-Pay Vote in the U.S.

Peter Iliev and Svetla Vitanova

February 27, 2014

ABSTRACT

We use a natural quasi-experiment to isolate the causal effect of holding advisory shareholder votes on executive compensation as mandated by the Dodd-Frank Act. Firms with a public float under $75 million did not have to hold a Say-on-Pay vote. Focusing on firms around the cutoff, we find a negative market reaction to the exemption from the Say-on-Pay rule. We find that the regulation increased both the level of CEO pay and the sensitivity of pay to performance. Firms did not avoid or postpone compliance, and did not change their compensation structure. We also document an increase in shareholder support for directors among firms that were required to hold a Say-on-Pay vote.

Keywords: CEO Compensation, Shareholder Voting, Corporate Governance, Say-on-Pay JEL Classification: G30, G38

Peter Iliev (email: pgi1@psu.edu, phone: 814-863-5456, address: 348 Business Building, University Park, PA) and Svetla Vitanova are from Smeal College of Business at the Pennsylvania State University. We thank our team of dedicated research assistants: Thomas Corsale, Robert Chatt, and Christopher Maduforo. We wish to thank David Haushalter, Jared Williams, Matthew Gustafson, Michelle Lowry, Laura Field, Chris Muscarella, Yelena Larkin, Bernie Black, Katherine Litvak, Kristian Rydqvist, Henry Hansmann, and seminar participants at the Pennsylvania State University, the New York Accounting and Finance Forum, and the Conference on Empirical Legal Studies for their helpful comments and discussions. Peter Iliev acknowledges financial support from the Smeal Research Grant Program.

1

“Public outrage over high executive pay, regulatory reforms to give shareholders greater power and more scrutiny of how they use that power have created a climate in which even the most profitable companies can be targets of activism if their corporate governance is not up to scratch.”

The Economist, “Shareholders at the Gates”, March 9, 2013

Following the 2008 financial crisis, a controversial Say-on-Pay requirement was added to the

Dodd–Frank Wall Street Reform and Consumer Protection Act of 2010, mandating regular

shareholder votes on executive compensation. The new rule signals a change in the direction of

financial regulation. Rather than mandating enhanced disclosure and independent boards of

directors, this new approach relies on direct shareholder input into the inner working of public

companies. It is hard to anticipate the potential effect of the new regulation because shareholders

rarely had the chance to vote on CEO compensation.

In a congressional testimony Kaplan (2007) points out that CEO pay is largely determined by

market forces and that proposed bills on mandatory shareholder compensation votes will impose

costs while having no significant benefits. This “efficient contracting” view contends that the

observed level and composition of CEO pay is a competitive equilibrium outcome in the market

for managerial talent. In that case, the mandated shareholder voting should not reduce CEO pay.

On the contrary, it might increase pay because it introduces additional risks to the CEO position

through the mandated periodic votes on CEO pay.

In contrast, arguments based on the “managerial power” view maintain that both the level and

composition of CEO pay is determined by captured boards that cater to rent-seeking CEOs. If

this is the case, bypassing the boards by having a direct shareholder vote could lower inflated

CEO pay. In a different congressional testimony Bebchuk (2007) argues that the open and public

nature of such a vote would force the board to implement shareholder demands. These opposing

views are often shared in the extensive theoretical and empirical literature on executive

compensation (see Bertrand (2009), Frydman and Jenter (2010), Murphy (2013)). Ultimately, the

2

impact of the mandatory Say-on-Pay vote is an empirical question.

Identifying the effect of any regulation is inherently difficult. The new Say-on-Pay rule was

implemented by the Securities and Exchange Commission (SEC) together with a flurry of Dodd–

Frank-mandated regulations, including rules about the recovery of executive compensation,

hedging by employees, and a mandated review of the use of compensation consultants. Studies

relying on the change in CEO compensation levels around the passage of the new rule will also

be affected by factors determining CEO pay that are not driven by the new regulation, such as

changes in the level of real economic activity or trends in the labor market for corporate talent.

Ideally, the researcher would conduct an experiment in which some firms are randomly assigned

to comply with the new rule while others do not have to comply with the rule. In such an

experiment, the researcher can compare the firms that have to comply to similar non-affected

firms, and thus can attribute the differences in firm outcomes to the new regulation.

Fortunately, something close to the ideal experiment exists. Although the majority of U.S.

companies were required to hold their first Say-on-Pay vote at their annual shareholder meetings

in 2011, a group of firms was given a two-year break by the SEC. The SEC exempted from the

Say-on-Pay and frequency votes “smaller reporting companies” that did not exceed $75 million

in public float in the preceding years. In this paper, we compare the outcomes of the firms above

the cutoff that had to hold a Say-on-Pay vote to firms below the cutoff that did not have to

comply with the new regulation. This is a valid empirical design because the exact eligibility rule

is not related to firm fundamentals. By looking at firms in the neighborhood of the SEC

enforcement region, we can compare the outcomes of similar firms relying on a difference-in-

differences approach. Some firms self-selected into holding a Say-on-Pay vote even if they were

not in the group required to do so. To address this issue, we use an instrumental variables

3

approach. Finally, the results in this paper quantify the impact of a two-year delay in the

introduction of Say-on-Pay. The results probably underestimate the true effect of a permanent

break if firms did not expect further rule extensions.

We quantify the impact of the new regulation on CEO pay levels, CEO pay composition,

CEO severance payments, and pay sensitivity to returns. We also isolate the differential reaction

to the announcement of the Say-on-Pay break for small firms, providing the first direct evidence

for the valuation effect of the new rule. We further investigate firms’ reactions to the new

regulation and its effect on shareholder voting for directors—arguably the only regular

shareholder vote with substantial impact on firm policy. In all tests we focus on a sample of

firms in the neighborhood of the $75 million rule cutoff.

An event study around the announcement of the SEC rule documents a negative 1.5% three-

day return among the firms that were exempt from the new rule relative to firms that were not.

Although this magnitude is moderate, this result suggests that shareholders value the option of

continually voting on executive compensation. We also find that firms that held a Say-on-Pay

vote also received more shareholder support for their directors. Delegating the decision to the

shareholders reduced pressure on directors who no longer share the sole responsibility for setting

CEO pay.

We find that the Say-on-Pay increased the level of CEO pay by 14%. Simultaneously with

the rise in pay we find a significant increase in the sensitivity of pay to performance after the

introduction of the new regulation. Our evidence does not support the notion that shareholder

votes disciplined the firms into reducing CEO pay. On the contrary, our results are consistent

with CEO pay increase to compensate managers for the higher sensitivity of pay to performance

and the extra risk introduced by regular shareholder votes. Consistent with these results, there is

4

no evidence that firms actively avoided compliance. We find that a significant number of firms

decided to hold a vote even though they were not required to do so.

Our paper is the first to look at the overall valuation effect of the Say-on-Pay rule by

comparing the event reaction to the rule announcement among firms that were required to

comply with the rule to that of firms that did not have to comply. The evidence in this paper

provides an important causal inference that contributes to the ongoing debate about the

effectiveness of shareholder votes on executive compensation.1

The paper proceeds as follows. Section I introduces previous work on the subject and the

details of the new SEC rule. Section II discusses the identification strategy and Section III

describes the data. Section IV reports results from event studies and quantifies the effect on

director elections. Section V presents the effect of the Say-on-Pay vote on executive

compensation and the incentives of firms to voluntary comply with the new rule. Section VI

concludes.

1 This paper adds to a growing literature that tests the effects of financial regulations using exogenous shocks (Greenstone, Oyer, and Vissing-Jorgensen (2006), Atanasov, Black, Ciccotello and Gyoshev (2010), Iliev (2010), Becker and Strömberg (2012)).

5

I. The Setting

Prior to the recent Dodd-Frank mandate, shareholders in the United States could propose

votes to approve executive compensation, but such votes were rare.2 Ertimur, Muslu, and Ferri

(2011) document 134 “Vote-No” campaigns and 1,198 non-binding shareholder proposals during

the 1997–2007 period, with the overwhelming majority of compensation-related proposals being

to link pay to performance, approve golden parachutes, or tie executive compensation to social

criteria. They document that activists target firms with high CEO pay and high excess CEO pay,

and that voting significantly reduce overall CEO pay.3 The evidence in the paper is suggestive of

the potential effect of a general across-the-board mandatory vote on executive compensation.

However, the fact that the rare previous Say-on-Pay votes were initiated by shareholder activism

in a subset of firms with troubled pay practices makes any inference about the potential effect of

the new mandatory Say-on-Pay vote problematic.

The infrequent and activist-driven nature of compensation votes changed with the signing of

the Dodd-Frank Act in July 2010. Section 951 of the new law requires that shareholders approve

the company’s executive compensation in a non-binding vote occurring at least once every three

years. Furthermore, shareholders must hold an additional vote the first year and every six years

thereafter to determine the frequency of the compensation votes: every one, two, or three years.

The new law left most of the implementation details of the Say-on-Pay rule up to the Securities

and Exchange Commission (SEC).

2 Say-on-Pay votes are not restricted to the U.S. In a study of the 2002 UK say-on-pay regulation, Ferri and Maber (2013) find that firms with excessive CEO pay realize positive returns. They also show that firms that experienced shareholder dissent changed their pay practices to increase their sensitivity to poor performance, although they find a statistically insignificant increase in the levels of CEO pay. On the other hand, Correa and Lel (2013) study changes in CEO pay pre and post the introduction of Say-on-Pay votes in a sample of countries, documenting a statistically significant 6.1% decrease in CEO pay in the post period. They confirm that Say-on-Pay laws increase sensitivity to performance. Yermack (2010) provides an overview of the shareholder voting literature. 3 Gillan and Starks (2000) document that market reactions to shareholder proposals vary according to the issue and the identity of the sponsor.

6

In January 2011, the SEC announced its final rule implementing the new Say-on-Pay

regulation. Firms with annual meetings after January 21, 2011 had to hold a mandatory Say-on-

Pay vote as well as a vote regarding the Say-on-Pay frequency. Although shareholder votes in

the United States are non-binding, no board can easily ignore a majority verdict against its CEO

package. The SEC rule on Say-on-Pay requires that firms must disclose in reports filed with the

SEC whether and how they considered the voting results in setting executive pay. Moreover,

Institutional Shareholder Services Inc., the biggest proxy access consultant in the United States,

adopted a key change to its proxy voting policies. Beginning in 2014, ISS will recommend

voting against directors if they “fail to act” on shareholder proposals that receive a majority vote

cast in the previous year. The second biggest proxy access consultant, Glass, Lewis & Co., will

scrutinize the board’s response on issues for which more than 25% of shareholders oppose

management.4 If indeed compensation is not set optimally, it is reasonable to expect Say-on-Pay

votes to have a substantial impact on CEO compensation.

The SEC gave a two-year break to firms that qualified for “smaller reporting company”

status and indicated that it might consider making this break permanent. Firms maintain this

status as long as their public float (the part of equity not held by management or large

shareholders, as reported on the first page of the company’s 10K filing) is less than $75 million

as of the last business day of its second fiscal quarter.5 The smaller reporting companies held

their first Say-on-Pay votes in 2013.6 However, when the new rule was proposed, all indications

4 See “ISS, Glass Lewis, and the 2013 Proxy Season,” by John F. Olson, Gibson, Dunn & Crutcher LLP and Georgetown Law Center, February 2013. 5 Companies with a less than a $75 million public equity float (or, if the float cannot be calculated, those having revenues less than $50 million) maintain their smaller reporting company status. Once a company crosses these thresholds, it loses this status. A company can re-qualify for the smaller reporting company status if its public float falls below $50 million (or, if the float cannot be calculated, it has revenues less than $40 million) in subsequent years (SEC Final Rule33-8876). 6 See “Smaller Reporting Companies Subject to Say-on-Pay Rules in the 2013 Proxy Season,” Wiggins and Dana, Advisory, February 2013.

7

were that the break might turn into a permanent solution, as implied by a similar break given

when implementing Section 404 of the Sarbanes-Oxley Act that was extended multiple times.

Commissioner Kathleen Casey questioned whether Dodd-Frank even mandates a say-on-pay

vote for all public firms, while Troy Paredes (also a republican commissioner) insisted that the

commission did not do enough for smaller public companies. Democratic commissioner Elisse

Walter commented that “We will have time to gain experience with the rule to see if further

adjustments are necessary for small businesses.” 7 Further, the Wall Street Journal story

reporting the break for small firms quoted an SEC official commenting that the commission

might consider a permanent exemption for smaller companies.8 Appendix A provides details

about the regulation.

This study quantifies the effect of the new rule by comparing the outcomes of firms that had

to comply with the rule because they crossed the $75 million threshold to firms that were just

below the threshold. The strength of our approach is that it can isolate the effect of the Say-on-

Pay rule from competing effects of the Dodd-Frank Act and any other contemporaneous events.

The cost of this approach is that it looks at relatively small firms, and the results might not apply

to larger firms. However, previous research shows that the CEO pay increase has not been

confined to the largest firms in the United States (Frydman and Jenter 2010). Second, small firms

suffer more from asymmetric information and captured boards, and thus arguably could reap

higher benefits from using the votes of their shareholders to reset CEO pay to an optimal

contract. Third, there are many more small firms than large firms, and thus small firms present

an interesting setting of their own.

7 “SEC approves investor say on CEO pay,” by Ronald D. Orol, MarketWatch, January 25, 2011. 8 “SEC, in Split Vote, Adopts 'Say on Pay' Rule,” by Jessica Holzer, The Wall Street Journal, January 25, 2011.

8

II. Identification

We study firms near the $75 million threshold to reduce the potential bias from

unobservable factors (such as firm investment opportunities) that might be correlated with the

assignment rule and with the outcomes of interest (such as CEO pay, earnings, and returns).9 The

strength of this specification is that once we control for the known differences between the

public float of the firms, the estimated difference between the firms required to hold Say-on-Pay

votes and the firms not required to hold Say-on-Pay votes can be attributed directly to the new

regulation. We start our analysis with a difference-in-difference approach in which we compare

the difference between these two sets of firms before and after implementation of the new rule,

allowing for time and firm fixed effects and controlling for firm performance and governance.

We define the post period including the year preceding the Say-on-Pay vote because firms might

change their executive compensation in anticipation of the vote. Further, shareholders might be

voting based on the previous year’s compensation because, when the vote is held, that would be

the only year with reported actual CEO pay on file. 10

Because some firms decided to hold a Say-on-Pay vote when they were not formally

required to, we further specify instrumental variables tests in which we rely on the firms’ voting

status. This approach relies on the $75 million cutoff specified by the SEC to predict the actual

observed voting. Intuitively, there should be no reason why firms above and below the $75

million public float cutoff would suddenly have significantly different compensation packages in

2011, other than through the effect of the new Say-on-Pay rule.

9 We focus on firms that reported a public float between $40 and $110 million in the second quarter of 2009—the year preceding the Say-on-Pay rule. Results are similar if we focus on a tighter, $50 to $100 band of firms. 10 The actual Say-on-Pay proposal is generic, with the following SEC suggested wording: “The compensation paid to the company’s named executive officers, as disclosed pursuant to Item 402 of Regulation S-K, including the Compensation Discussion and Analysis, compensation tables and narrative discussion, is hereby APPROVED.” SEC Final Rule 33-9178.

9

Next, to assess the validity of our research design, we discuss whether firms anticipated the

rule, the choice of some firm to self-treat, and the existence of confounding events.

A. Rule Anticipation

Table 1 shows the timeline of the new rule. The final SEC rule mandated a Say-on-Pay vote

for all companies that did not meet the “smaller reporting company” criteria and held their

annual meetings after January 21, 2011. Companies had to hold a Say-on-Pay vote if they

crossed the $75 million public float threshold in any year prior to 2011. While the originally

proposed Say-on-Pay stated that it would not exempt small issuers from the rule, the SEC sought

public comments on whether the rule would unduly burden small companies (SEC Proposed

Rule 33-9153). 11 The final rule (SEC Final Rule 33-9178), announced on January 25, 2011,

stipulated that the SEC changed its initial position after receiving a number of comment letters

urging a delay in compliance for small firms.

Our approach assumes that the firms did not manipulate their public floats before the

passage of Dodd-Frank. If firms could perfectly control their stock prices and chose to

manipulate them in anticipation of the new rule, then results comparing voting and non-voting

firms could be biased. A firm’s ability to manipulate its stock returns and filing status was, in

fact, rather limited during the turbulent stock markets between 2008 and 2011. The empirical

evidence does not support that firms actively managed their public floats. Figure 1 reports the

overall public float distribution for all public firms in the United States and the public float

distribution near the $75 million cutoff in the year before the rule was implemented. The figure

11 From SEC Final Rule 33-9178: “Should we fully, partially, or conditionally exempt smaller reporting companies or some other category of smaller companies from any or all of our proposed rules? If so, which ones? Are any of our proposed rules unduly burdensome to smaller reporting companies and if so, how are they unduly burdensome?”

10

does not show a clustering of firms just under $75 million.

B. Voluntary Adoption of Say-on-Pay Voting

A simple approach to implementing the difference-in-differences design is to compare the

outcomes of firms that held a Say-on-Pay vote with firms that did not hold a vote. This approach

assumes that firms cannot avoid treatment and that firms do not select to be treated. Indeed, we

do not find evidence that firms actively avoided holding a Say-on-Pay vote. First, firms did not

know about the small firm break in the new rule before it was too late. The break for small firms

was announced together with the final rule and became effective immediately, while the smaller

reporting company status was determined in the years preceding the vote. Second, firms did not

change their annual meeting dates strategically to postpone holding a Say-on-Pay vote. Figure 2

tabulates the annual votes held near the rule effective date in the year of and in the year

preceding the new regulation. If firms were actively trying to avoid the new rule, we would

expect a significant number of firms to hold their annual votes just before the effective date. We

do not see an increase in annual meetings in November or December, nor a corresponding drop

in January meetings. Finally, firms that were supposed to hold a vote complied with the rule,

suggesting that firms could not evade the rule by using any loopholes.

However, we find that 45 firms chose to hold a Say-on-Pay vote even though they qualified

for the “smaller reporting company” status (out of 192 such firms) and thus were not required to

do so. This self-selection into treatment might bias our results if we compare firms that held a

vote to firms that did not hold a vote because firms might have strategically decided whether

they wanted to hold the vote. An alternative simple approach of performing a difference-in-

differences estimation based on the actual status of the firm and not on whether they had a vote is

also problematic. If the Say-on-Pay vote has effect on the variables of interest, then our control

11

group of firms that were not required to hold a vote will include firms that experienced treatment

(by holding a vote), and therefore we will underestimate the true treatment effect.

Therefore, we provide evidence based on an instrumental variables approach using the

firms’ SEC decision rule determining the firms’ reporting status as an instrument for the firm

holding a Say-on-Pay vote (Angrist, Imbens and Rubin (1996)). The firms’ reporting status is

based on the position of their public float relative to the $75 million cutoff. This instrument is a

strong predictor of having a Say-on-Pay vote, but it should not predict shareholder pay once we

control for the small size differences between the firms. Conceptually, one would not expect

firms to experience a relative jump in their compensation practices when they cross the threshold

for reasons other than holding the Say-on-Pay vote. This approach produces unbiased estimates

in the case where some firms select into treatment for endogenous reasons and experience a

treatment effect.

C. Contemporaneous Regulations

An important assumption behind our approach is that changes in outcomes between affected

and unaffected companies in 2010 can be attributed to the Say-on-Pay rule and not to another

event that happened at the same margin in 2010. In other words, there should be no important

events that augmented the effect of the SEC rule for firms just above and just below the $75

million cutoff in 2010. The SEC has used the same cutoff before to enforce different rules.

Starting in 2003, accelerated filers had to produce an annual report 15 days earlier than smaller

reporting companies.12 The SEC gave smaller reporting companies a break from complying with

Section 404 of the Sarbanes-Oxley Act. This break was partially reversed in 2010, when they

12 Smaller reporting companies were classified as non-accelerated filers prior to 2007. The non-accelerated filer definition used the same $75 million public float cut-off.

12

were required to file Management Reports under Section 404 but smaller reporting companies

still do not have to file auditor attestation of their Section 404–related Management Report.

However, these rules were implemented before 2011 and should not lead to a marked change in

CEO pay in 2011. Therefore, the difference-in-difference framework isolates the change in CEO

pay produced by the 2011 Say-on-Pay rule. Finally, the firm fixed effects in all estimations

adjust for the different outcomes between the groups that might stem from past regulations.

Another potentially important governance provision in the Dodd-Frank Act was also

scheduled by the SEC to take effect only for firms that did not qualify for the “smaller reporting

company” status. On August 25, 2010, the SEC adopted new measures to facilitate director

nominations by shareholders.13 Under this rule, shareholders that own at least 3 percent of the

company shares would have been able to have their nominees included on the firm proxies. The

application of this rule for smaller reporting companies was deferred for three years, creating a

confounding event for studying the effect of the similarly implemented Say-on-Pay rule.

However, this rule was struck down by the U.S. Circuit Court of Appeals on July 22, 2011. The

court issued a harsh verdict related to the ruling, implying that the SEC did not analyze the

excessive costs of contested board elections enough. The SEC backed down from the new rule,

indicating that it would not re-open the issue in the immediate future.14 We expect that the failed

attempt to implement a proxy access rule at the same threshold will have small effect on CEO

pay because the rule would need time to effectively change the director pool. In robustness

checks we verify that our effects still hold if we just compare the cross sections of 2009 (well

before the proxy access rule) to 2012 (well after the failed proxy access rule). Thus, we conclude

13 See “SEC Adopts New Measures to Facilitate Director Nominations by Shareholders,” SEC Press Release 2010-155 made public on August 25, 2010 and the corresponding SEC Final Rule 33-9136. 14 See “Court Deals Blow to SEC, Activists,” The Wall Street Journal, July 23, 2011 and “Statement by SEC Chairman Mary L. Schapiro on Proxy Access Litigation” SEC Press Release 2011-179 made public on September 6, 2011.

13

that studying the impact of the Say-on-Pay rule by comparing the effects on smaller reporting

companies relative to slightly larger companies is a set-up not confounded by significant

competing effects.

III. Sample and Descriptive Statistics

Firms that did not qualify as a “smaller reporting company” had to comply with the new

Say-on-Pay rule and hold an advisory compensation vote. The “smaller reporting company”

status is determined by a firm’s public float. (The public float is the part of equity not held by

management or large shareholders, measured at the end of the firm’s second fiscal quarter.) Once

a firm’s public float exceeds $75 million, it no longer qualifies for the compliance break. Both

the firm status and the exact public float are reported annually on the first page of firms’ 10K

filings.

To ensure that our sample includes all firms that were near the rule cutoff, we collect the

public float of all firms in Compustat from 10K filings with EDGAR in 2009. If we were to

simply use the market value of equity instead of public floats to identify firms that were near the

cutoff, we would get a significant error rate because public float is smaller than the total value of

equity. For example, 47% of the firms in our sample with a public float below $75 million would

be considered above the cutoff because their market value of equity exceeds $75 million.

Moreover, because the rule is based on the history of public floats over time, a proper definition

of treated and non-treated firms has to rely on the exact “smaller reporting company” status as

reported in the firms’ 10K filings. Therefore, after identifying a sample of firms near the cut-off

we collect their “smaller reporting company” status from their 10Ks for the years of our study.

Figure 1 presents the distribution of the public float of all firms in 2009. We select a sample

14

of firms with a public float between $40 and $110 million in 2009, the year before the rule

became effective.15 We exclude firms that received money from the Troubled Asset Relief

Program (TARP) because these firms came under substantial pressure from the American

Recovery and Reinvestment Act of 2009 to limit or even abolish incentive compensation and

Golden Parachutes.16 We hand-collect data on shareholder voting, CEO compensation, and firm

governance, because this data is only readily available for large firms. For the firms in our

sample, we collect the results of their annual shareholder meetings from 10K and 10Q filings

prior to 2010 and from 8-K filings after 2010.17 We also collect data about CEO pay, CEO

characteristics, and management ownership from the firms’ annual proxy statements on form

DEF14. These data are further combined with data from Compustat and CRSP, resulting in a

final sample of 496 firms with four cross-sections of data over the 2008 to 2011 period.

Table 2 reports the descriptive statistics. The firms in our sample have an average value of

assets of $513 million. The mean public float of $152 million over the 2008 to 2011 period is

about twice as larger as the 2009 mean value because firm returns fall sharply in the beginning of

the sample and then rise sharply towards the end. The volatility in firm returns effectively

prevents firms from managing the market value of their publicly traded equity (their public

float). The mean return on assets is negative while the stock market return is positive, consistent

with the economic crisis and the subsequent rebound in stock valuations during the sample

period. The table also reports a variety of CEO and firm governance characteristics collected

from the companies’ definitive proxy statements. The average CEO pay is about $1.1 million

with a significant cash component (58%). The CEOs are promised $2.4 million in change-of- 15 Our results are similar if we selected a tighter range of public floats $50 and $100. We chose to collect data for a range of $35 million around the rule cut-off in order to increase the power of our tests. 16 See “Summary of Executive Compensation Limits on TARP Recipients,” CBA Regulatory Compliance Bulletin, February 23, 2009. 17 The SEC changed the rules for reporting the results from annual shareholder votes from Forms 10-Q and 10-K to Form 8-K, effective February 28, 2010 (SEC Final Rule 33-9089).

15

control payments. Consistent with the voting literature, the average support for directors from

shareholder votes is 92%.

The identification design in this paper assumes that firms that comply with the new rule and

firms that do not comply were similar before the rule was implemented but potentially different

in their outcomes after the rule took effect. Table 3 compares the characteristics of the firms that

held Say-on-Pay votes in 2011 with the firms that did not hold such votes. The characteristics are

measured two years before the passage of the Say-on-Pay rule. Inevitably, the firms that

complied with the new vote are larger than the firms that did not. The two groups of firms have

similar sales growth, returns, and ROA in the two years before the new law. Moreover, their

CEO characteristics are the same with the exception of CEO age and tenure—the non-affected

firms tend to have slightly older CEOs that have been in place for two more years. Important

corporate governance characteristics like Staggered Board, Management Ownership, CEO

Ownership, and CEO is Chairman are similar across both groups. To alleviate the concern that

the two sets of firms are not comparable in all characteristics, we use firm fixed effects, and

explicitly control for firm size, firm returns, and all the other CEO characteristics reported in

Table 3. The difference-in-difference approach ignores the differences in levels between the two

groups and instead isolates the differential change in outcomes around the new rule. We further

use instrumental variable (IV) estimation to predict treatment based on the SEC rule eligibility.

IV. The Effect of Say-on-Pay on Firm Returns and Firm Voting

We start our analysis by reporting the market reaction to giving a break for small companies.

Then, we present the vote outcomes from the first proxy voting season, and we test the idea that

holding an annual Say-on-Pay vote can relieve directors from the pressure of setting CEO

16

compensation.

A. Event Study

Ultimately, the effect of Say-on-Pay should be judged based on the overall valuation effect

of the new rule. Although not perfect, event studies can reveal the expected net effect of the new

rule to the firms’ shareholders. A few recent papers have estimated event studies around the

announcements of the Dodd-Frank Act provisions, thereby exploring the cross-sectional

properties of firm returns (Cai and Walkling (2011), Larcker, Ormazabal, and Taylor (2011)).

While these studies suggest that the event’s effect is a function of firm characteristics, they

cannot provide a direct measure of the overall effect of the regulation. We instead test for the

differential impact between firms that must comply with the rule and those that do not have to

comply. This set-up has the natural advantage of providing a treatment and control group, which

allows us to directly measure the overall effect of the Say-on-Pay mandate. It also eliminates the

effect of competing events that affected all firms at the time of the announcement.

Table 4 reports the results from our event studies of the Say-on-Pay rule announcements.

While the Dodd-Frank Act mandates Say-on-Pay votes, the law leaves it up to the SEC to

implement the mandate with a separate rule. The first event reported in Table 4 is the October

18, 2010 SEC press release announcing the Say-on-Pay rule proposal. We do not expect to see a

significantly differential reaction to the announcement among firms that had to comply with the

rule versus those that did not because the SEC explicitly indicated in this press release that the

proposed rule would not exempt small firms (SEC Press Release 2010-198). The commission

believed that the new rule would not disproportionately burden small issuers and that investors

had the same interest in voting for small issuers as for larger ones. The SEC, however, asked for

comments on the question “Should we fully, partially, or conditionally exempt smaller reporting

17

companies or some other category of smaller companies from any or all of our proposed rules?”

opening up the possibility of an exemption (SEC Proposed Rule No. 33-9153)

We use a market model and a four common risk factors model identified in Fama and

French (1993) and Jegadeesh and Titman (1993). Consistent with the idea that the proposed rules

did not significantly affect smaller reporting companies, we do not see a statistically significant

difference in the three-day event window abnormal returns between the affected and not affected

firms. However, we observe an economically meaningful negative -0.81% return for the small

firms that can be potentially exempted from the new rule, suggesting negative news.

The main event of interest is the SEC press release announcing the adoption of the final Say-

on-Pay rule on January 25, 2011. Importantly, this final rule changed the earlier SEC position

that small firms would not get an exemption. This change in the SEC’s position allows us to

clearly identify the market’s assessment of the net effect of the Say-on-Pay vote. The stock

market reaction to this event provides a close approximation of the valuation impact of the Say-

on-Pay vote because the SEC announcement about the small company break was largely

unanticipated (Schwert (1981)).

To test for the aggregate effect of the announcement, we form a long-short portfolio. This

portfolio buys the firms not required to hold a vote and sells the firms required to hold a vote.

The long-short portfolio has a negative three-day cumulative abnormal return of -1.5%18. The

observed negative reaction of the market is consistent with the finding that the Say-on-Pay votes

did not force firms to change their CEO compensation away from the optimal level. On the

contrary, it is likely that the Say-on-Pay voting mandate opened an important new

communication channel between shareholders and firms, especially considering that the majority

18 A potential worry about the event study results is that small company returns might suffer from stale price bias due to non-trading. We verified that all stocks used in the event study have positive trading volumes during the event study estimation window.

18

of firms chose to stage these votes annually.

B. Vote Outcomes

The new Say-on-Pay regulation became effective for shareholder elections held after

January 21, 2011. The early experience with Say-on-Pay in the United States has been one of

near unanimous support. For example, the Wall Street Journal reports that in a sample of 2,532

companies, the Say-on-Pay vote passed in 98.5% of the cases.19 However, the article also reports

that in 39 instances the executive pay was rejected and that was used as ammunition for lawsuits.

Surprisingly, proxy advising firms did not have a decisive influence on these early results as ISS

was recommending a “No” vote for a much larger set of 298 companies.20

Figure 3 graphs the distribution of the Say-on-Pay and frequency votes for the firms in our

sample. Consistent with the overall U.S. experience, just two firms out of the 350 firms that had

a Say-on-Pay vote in our sample received a negative vote. However, 30 more firms received less

than 75% support, which is the level that triggers a Glass Lewis investigation into the board’s

response to the vote.

The Say-on-Pay regulation also mandates a separate vote on the frequency of the advisory

compensation vote. The second histogram in Figure 3 shows that shareholders overwhelmingly

support annual Say-on-Pay votes. While the Say-on-Pay votes tend to be in favor of the current

executive compensation, the very act of holding annual votes can have a profound effect on the

level and structure of CEO pay. First, a failed vote can lead to significant additional costs to

19 “A ‘Yes’ on Say on Pay,” The Wall Street Journal, July 8, 2011. 20 We do not have the proxy advisory firms’ recommendations for the votes in our sample. The ISS Voting Analytics database covers Russell 3000 firms, which are by definition larger than our firms. Therefore, we cannot attribute any effects to proxy firm recommendations, but we can extrapolate the overall unconditional effect of the new regulation. Larcker, McCall, and Ormazabal (2012) find that some firms change their compensation programs to avoid negative proxy advisor recommendations. Ertimur, Ferri, and Oesch (2012) explore the impact of the two leading proxy advisory firms (ISS and Glass Lewis & Co.) on the Say-on-Pay voting outcomes.

19

change the executive compensation structure of the firm. Second, a Say-on-Pay vote can be

viewed as a general “confidence” vote for the management; hence, anything less than a high

level of support could have important implications for the firm. Thus, one might expect CEO

compensation packages to change both before and after the vote even when most votes show

strong level of support for the CEO pay package. In the next section we test if the new rule

affected CEO pay in the United States.

C. Director Voting

The new Say-on-Pay rule changed the implicit duties of directors. Even though directors are

still ultimately responsible for setting CEO pay and monitoring CEO performance, they now

have direct input from shareholders on these issues. The new mandatory votes might alleviate

some of the pressure on directors, leading to a better relationship with their shareholders and

potentially to higher support for them in director elections. Say-on-Pay votes in overseas markets

have indeed often resulted in improved dialogue between shareholders and directors.21 To test

this hypothesis, we compare the average support in director elections among firms that held a

Say-on-Pay vote relative to firms that did not. The new regulation mandated a Say-on-Pay vote

during the first shareholder meeting that included director elections; thus, these two votes were

held simultaneously. We focus on director votes because Cai, Garner, and Walkling (2009) and

Fischer, Gramlich, Miller, and White (2009) show that even small differences in director votes

have a substantial effect on firm policies, leading to lower “abnormal” CEO compensation and to

a higher probability of removing poison pills and classified boards.

Table 5 reports the results. We estimate that firms that held a Say-on-Pay vote experienced

2.2% more support for directors relative to firms that did not hold such vote. These results are

21 “Say on Pay: Results From Overseas,” by Stephen Deane, The Corporate Board, July/August 2007.

20

similar if we just compare the firms that held a say-on-pay vote to those that did not, and when

we instrument the treatment with the assigned treatment based on the SEC rule. Even though

2.2% seems economically small, the average support for directors is 92%, suggesting a 27% drop

in negative votes.

V. The Effect of Say-on-Pay on CEO Compensation

Executive compensation in the United States has steadily increased over time for firms of all

sizes (Frydman and Jenter (2010)). Because the new SEC regulation seeks shareholder approval

of CEO pay, we next test the effect of the new rule on both the level and structure of CEO pay.

A. The Evolution of CEO Pay

We first examine the evolution of CEO pay over the sample period. Some firms decided to

hold a Say-on-Pay and frequency vote even though they were not required to do so. Therefore,

we split the group of firms that held a Say-on-Pay vote into firms that were required to hold the

vote (mandatory voting firms), firms that were not required to hold a vote but decided to hold

one (voluntary voting firms), and firms that were not required to hold a vote and did not hold a

vote (non-voting firms). These results are reported in Figure 4. The results indicate an increase in

the pay of the firms that were required to hold a vote and do not support the notion that the new

rule decreased total CEO compensation.

Figure 5 reports the average equity ratio and cash ratio for the three groups of firms (the

mandatory voting firms, the voluntary voting firms, and the non-voting firms). The cash

compensation ratio is defined as the ratio of salary and bonus to total CEO compensation, and

the equity compensation ratio is the ratio of stock awards and option awards to total CEO

21

compensation. We do not observe any significant changes in the time series of mandatory voting

firms’ ratios relative to the other two groups, suggesting that the types of pay in the CEO

contracts were not changed significantly. The time series evidence, however, is not conditional

on firm performance, CEO characteristics, and the overall level of firm compensation. Next, we

document the effect of the new rule with a difference-in-difference approach, controlling for

CEO characteristics, firm-specific performance and governance characteristics, and firm fixed

effects. We further instrument the compliance using the SEC rule to isolate the causal effect of

the new rule.

B. The Effect of the Rule on CEO Pay

In our first two regressions, we rely on the actual treatment status: we are comparing the

effect for firms that held a Say-on-Pay and a frequency vote with firms that did not hold these

votes. Table 6 reports the results from these regressions of the logarithm of total CEO pay on a

Say-on-Pay vote dummy. We also control for a variety of CEO characteristics, governance

measures as well as firm and year fixed effects to control for different levels of pay across firms

and over time. Firms that held a Say-on-Pay vote experienced a statistically significant 9.8%

increase in the level of their compensation in the year before and the year after the rule came into

effect. It is important to note that the estimated effects in the OLS model with and without

control variables are the same. This confirms our assertion that the treatment and control groups

of firms have similar characteristics. If we instrument the actual treatment, we find even bigger

increase of 14.1%. The difference between the OLS and IV estimates comes from the group of

firms that comply voluntarily with the new rule. These firms did not increase their pay (as

suggested by Figure 3). Including them in the treated group biases the OLS estimate downward.

The IV estimate corrects this bias using the firm status as an instrument.

22

Overall, the evidence in Table 6 does not support the idea that the Say-on-Pay votes will

have the intended consequence of lowering executive pay. The evidence is consistent with

alternative interpretations. The new rule mandates frequent shareholder votes on executive

compensation, which creates additional risks for a CEO. In an optimal contract setting, CEOs

should be paid additionally for the new risks of receiving a no confidence vote. An alternative

explanation that the board and management is using the Say-on-Pay votes to rubber stamp even

larger compensations, is also possible.

To further examine these results we eliminate extreme CEO pay observations (Murphy

(2013)). A few extreme CEO pay observations can drive the results of CEO pay studies.22 To

confirm that our results are indeed not due to extreme values, we winsorize the top and bottom

5% of the total pay within each treatment group and filing year. We repeat the estimations in

Table 6 with the winsorized variables and the results are qualitatively similar. As an alternative

robustness check, we limit the sample of firms to those that had a public float between $60

million and $90 million in the year preceding the new regulation to assure even closer match in

firm size. Our results remain the same. We also confirm that our results remain consistent if we

include linear and higher-order polynomials of the public float as well as asymmetric effects of

the public float above and below the cutoff rule.

C. The Effect of the Rule on Pay Performance Sensitivity

Next, we ask if the increase in the overall compensation is associated with a higher

sensitivity of pay to performance. Financial research has focused on the sensitivity of CEO pay

to firm performance as a key measure of the effectiveness of the CEO pay contract (Bertrand

22 We remove CBS Corp from our sample because its executive compensation of $69.9 million in 2011 was 64 standard deviations away from the sample mean.

23

and Mullainathan( 2001) and Garvey and Milbourn (2006)). If shareholder votes lead to stronger

link between pay and performance, then the new reform can enhance firm value even if it does

not affect the level of CEO pay. Table 7 reports the results of regressing the logarithm of total

CEO pay on lagged firm returns interacted with pre and post regulation dummy. We report

results that condition on the raw return (models 1 and 2) and on the industry adjusted return

(models 3 and 4). We find that CEO pay sensitivity to raw performance and abnormal

performance increases after the new Say-on-Pay rule is in effect. These results suggest that the

new rule ties better pay to performance and can improve the CEO contract. The improved

sensitivity is consistent with the abnormal positive returns from the event study in Table 4 and

with the findings in Ferri and Maber (2013).

Table 8 tests if the different components of CEO pay changed after the Say-on-Pay vote.

Consistent with the graphical evidence in figure 5, we do not find significant changes in the ratio

of cash or equity compensation to total compensation, or in the levels of the promised

termination and change of control payments. Given that the optimal ratios are theoretically and

empirically hard to pin down, it is not surprising that allowing the shareholders to vote on the

overall CEO pay package does not significantly affect them.

D. Voluntary Voting

We next examine why some firms chose to hold voluntary Say-on-Pay and frequency votes

although they were not required to do so under the new SEC rule. On one hand, those firms

might be eager to implement the new rules and receive feedback from their shareholder before

the rule becomes mandatory for all firms. On the other hand, firms might strategically accelerate

their Say-on-Pay compliance because they want to receive pay approval at better terms.

Therefore, we test whether firms that chose voluntarily to hold a Say-on-Pay vote experienced

24

high stock returns ahead of the votes or worse returns after the vote. Firms that suspect that their

future returns will be low might be tempted to put executive compensation up for a vote while

their shareholders are satisfied with the firm’s current performance.

Figure 6 reports buy-and-hold returns for the voluntary voting firms, the mandatory voting

firms, and the firms that did not hold a Say-on-Pay vote. We choose to report raw buy-and-hold

returns and buy-and-hold returns relative to a small firm benchmark (the Russell 2000 Index)

because these measures closely mimics the experience of actual investors and likely affects the

Say-on-Pay vote. The results are strongly supportive of the hypothesis of timing the vote. Firms

that voluntarily chose to put a Say-on-Pay vote and a frequency vote on their ballots experienced

a significant drop in their buy-and-hold returns in the months after the election date. We interpret

this evidence as a short-term opportunistic move on the part of management. Given that

shareholders are likely to judge management based on stock performance, holding a vote right

before bad performance gives the management breathing room to turn the firm around before the

next vote. Moreover, holding the initial vote at favorable terms implies that the frequency vote is

more likely to result in a three-year Say-on-Pay voting interval rather than an annual voting. This

conjecture is supported by the actual outcome of the frequency votes. In the mandatory voting

group, 65% of firms had a majority vote in support of holding annual CEO pay votes. In

comparison, only 50% of votes in the voluntary voting group had majority support for annual

votes. Furthermore, Table 9 reports that this difference in support for annual votes is statistically

significant when we control for firm governance, firm size, and firm performance.

25

VI. Conclusion

This paper uses the exogenous variation generated by the SEC implementation of the rule

requiring Say-on-Pay and frequency votes for U.S. firms to identify the causal effect of this

regulation. We compare the outcomes of firms around the new rule compliance cutoff before and

after the rule took effect. The paper’s most significant contribution is its ability to disentangle the

effect of the new Say-on-Pay rule from the effects of other contemporaneous events. We find

that compliance with the Say-on-Pay rule did not cause a decrease in levels of CEO pay and did

not changed pay composition. On the contrary, firms that had to comply with the new rule

experienced both an increase in the total CEO pay and higher pay to performance sensitivity.

Further, the market reacted negatively to the announcement of the unexpected break for small

firms, consistent with a positive net effect from the new rule. Finally, firms that had to comply

with the Say-on-Pay rule experienced a significant increase in support for their directors during

shareholder elections. Taken together, these results are consistent with the view that the new

Say-on-Pay rule did not have the intended effect of curbing CEO compensation.

26

REFERENCES

Angrist, Joshua D., Guido W. Imbens, and Donald B. Rubin, 1996, Identification of Causal Effects Using Instrumental Variables , Journal of the American Statistical Association 91, 444-455.

Atanasov, Vladimir A., Bernard S. Black, Conrad S. Ciccotello, and Stanley B. Gyoshev, 2010, How Does Law Affect Finance? An Examination of Equity Tunneling in Bulgaria, Journal of Financial Economics 96, 155-173.

Bebchuk, Lucian A., 2007, Written testimony submitted before the Committee on Financial Services, United States House of Representatives, Hearing on Empowering Shareholders on Executive Compensation, March 8.

Becker, Bo, and Per Johan Strömberg, 2012, Fiduciary Duties and Equity-Debt Holder Conflicts, Review of Financial Studies 25, 1931–1969.

Bertrand, Marianne, 2009, CEOs, Annual Review of Economics, April 2009, 1, 121-150.

Bertrand, Marianne and Sendhil Mullainathan, 2001, Are CEOS Rewarded for Luck? The Ones without Principals Are., The Quarterly Journal of Economics, 116, pp. 901-932.

Cai, Jie, and Ralph A. Walkling, 2011, Shareholders’ Say on Pay: Does it Create Value? Journal of Financial and Quantitative Analysis 46, 299–339.

Cai, Jie, Jacqueline L. Garner, and Ralph A. Walkling, 2009, Electing Directors, Journal of Finance 64, 2389 – 2420.

Correa, Ricardo and Ugur Lel, 2013, The Effects of Say-on-Pay Laws on Executive Compensation, CEO Pay Slice, and Firm Value around the World, Working paper.

Ertimur, Yonca, Volkan Muslu, and Fabrizio Ferri , 2011, Shareholder Activism and CEO Pay, Review of Financial Studies 24, 535-592.

Ertimur, Yonca, Fabrizio Ferri, and David Oesch, 2012, Shareholder Votes and Proxy Advisors: Evidence from Say on Pay, Journal of Accounting Research 51, 951-996.

Fama, Eugene F., and Kenneth R. French, 1993, Common Risk Factors in the Returns on Stocks and Bonds, Journal of Financial Economics 33, 3–56.

Ferri, Fabrizio, and David A. Maber, 2013, Say on Pay Votes and CEO Compensation: Evidence from the UK, Review of Finance 17, 527-563.

Fischer, Paul E., Jeffrey D. Gramlich, Brian P. Miller, and Hal D. White, 2009, Investor Perceptions of Board Performance: Evidence From Uncontested Director Elections, Journal of Accounting and Economics 48, 172-189.

Frydman, Carola, and Dirk Jenter, 2010, CEO Compensation, NBER Working Paper No. 16585.

Garvey, Gerald T., and Todd T. Milbourn, 2006, Asymmetric benchmarking in compensation: Executives are rewarded for good luck but not penalized for bad, Journal of Financial Economics 81, 197–225.

27

Gillan, Stuart L., and Laura T. Starks, 2000, Corporate Governance Proposals and Shareholder Activism: The Role of Institutional Investors, Journal of Financial Economics 57, 275-305.

Greenstone, Michael, Paul Oyer, and Annette Vissing-Jorgensen, 2006, Mandated Disclosure, Stock Returns, and the 1964 Securities Act, The Quarterly Journal of Economics 121, 399-460.

Iliev, Peter, 2010, The Effect of SOX Section 404: Costs, Earnings Quality and Stock Prices, Journal of Finance 65, 1163-1196.

Jegadeesh, Narasimhan, and Sheridan Titman, 1993, Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency, Journal of Finance 48, 65–91.

Kaplan, Steven N., 2007, Testimony of Steven N. Kaplan on "Empowering Shareholders on Executive Compensation" and H.R. 1257, the “Shareholder Vote on Executive Compensation Act” before the Committee on Financial Services United States House of Representatives, March 8, 2007.

Larcker, David F., Allan L. McCall, and Gaizka Ormazabal, 2012, The Economic Consequences of Proxy Advisor Say-on-Pay Voting Policies, SSRN Working Paper 2101453.

Larcker, David F., Gaizka Ormazabal, and Daniel J. Taylor, 2011, The Market Reaction to Corporate Governance Regulation, Journal of Financial Economics 101, 431-448.

MacKinley, A. Craig, 1997, Event Studies in Economics and Finance, Journal of Economic Literature 35, 13–39.

Murphy, Kevin, 2013, Executive Compensation: Where We Are, and How We Got There, in George Constantinides, Milton Harris, and René Stulz (eds.), Handbook of the Economics of Finance, Elsevier Science North Holland.

Schwert, G. William, 1981, Using Financial Data to Measure Effects of Regulation, Journal of Law and Economics 24, 121–158.

Yermack, David, 2010, Shareholder Voting and Corporate Governance, Annual Review of Financial Economics 2, 103-125.

28

Appendix A

Rules and Regulations Defining the Say-on-Pay Vote “DODD-FRANK WALL STREET REFORM AND CONSUMER PROTECTION ACT” ‘SEC. 14A. SHAREHOLDER APPROVAL OF EXECUTIVE COMPENSATION. ‘‘(a) SEPARATE RESOLUTION REQUIRED.— ‘‘(1) IN GENERAL.—Not less frequently than once every 3 years, a proxy or consent or authorization for an annual or other meeting of the shareholders for which the proxy solicitation rules of the Commission require compensation disclosure shall include a separate resolution subject to shareholder vote to approve the compensation of executives, as disclosed pursuant to section 229.402 of title 17, Code of Federal Regulations, or any successor thereto. ‘‘(2) FREQUENCY OF VOTE.—Not less frequently than once every 6 years, a proxy or consent or authorization for an annual or other meeting of the shareholders for which the proxy solicitation rules of the Commission require compensation disclosure shall include a separate resolution subject to shareholder vote to determine whether votes on the resolutions required under paragraph (1) will occur every 1, 2, or 3 years.”

From SEC Proposed Rule 33-9153 on SHAREHOLDER APPROVAL OF EXECUTIVE COMPENSATION AND GOLDEN PARACHUTE COMPENSATION. (Oct. 18, 2010) E. Treatment of Smaller Companies Our proposed rules would not exempt small issuers from the requirements of Sections 14A(a) and 14A(b). We believe the shareholder advisory votes and additional disclosure required by Section 14A and our proposed rules would be significant for investors in all issuers, including smaller reporting companies.132 As a result, the proposed rules discussed above will all apply to smaller reporting companies, with the exception of our proposed amendment to Item 402(b) of Regulation S-K, as smaller reporting companies are not required to provide a CD&A. We do not believe that smaller reporting companies should be exempt from the say-on-pay vote, frequency of say-on-pay votes and golden parachute disclosure and vote because we believe investors have the same interest in voting on the compensation of smaller reporting companies and in clear and simple disclosure of golden parachute compensation in connection with mergers and similar transactions as they have for other issuers. From SEC Final Rule 33-9178 on SHAREHOLDER APPROVAL OF EXECUTIVE COMPENSATION AND GOLDEN PARACHUTE COMPENSATION. (Jan. 25, 2011) SUMMARY: We are adopting amendments to our rules to implement the provisions of the Dodd-Frank Wall Street Reform and Consumer Protection Act relating to shareholder approval of executive compensation and “golden parachute” compensation arrangements. Section 951 of the Dodd-Frank Act amends the Securities Exchange Act of 1934 by adding Section 14A, which requires companies to conduct a separate shareholder advisory vote to approve the compensation of executives, as disclosed pursuant to Item 402 of Regulation S-K or any successor to Item 402. Section 14A also requires companies to conduct a separate shareholder advisory vote to determine how often an issuer will conduct a shareholder advisory vote on executive compensation. In addition, Section 14A requires companies soliciting votes to approve merger or acquisition transactions to provide disclosure of certain “golden parachute” compensation arrangements and, in certain circumstances, to conduct a separate shareholder advisory vote to approve the golden parachute compensation arrangements.

29

In addition to their non-binding status, none of the shareholder votes required pursuant to Section 14A is to be construed “as overruling a decision by such issuer or board of directors.” Section 14A(a)(3) requires that both the initial shareholder vote on executive compensation and the initial vote on the frequency of votes on executive compensation be included in proxy statements “for the first annual or other meeting of the shareholders occurring after the end of the 6-month period beginning on the date of enactment” of the Act.33 Thus, the statute requires separate resolutions subject to shareholder vote to approve executive compensation and to approve the frequency of say-on-pay votes for proxy statements relating to an issuer’s first annual or other meeting of the shareholders occurring on or after January 21, 2011, whether or not the Commission has adopted rules to implement Section 14A(a). Many commentators agreed with our proposed approach not to exempt smaller reporting companies from Rule 14a-21(a) and Exchange Act Section 14A(a)(1). Some commentators did suggest that smaller reporting companies should be exempt from the say-on-pay vote or required to conduct a say-on-pay vote on a triennial basis beginning in 2013.5 Companies that qualify as “smaller reporting companies” (as defined in 17 CFR 240.12b-2) as of January 21, 2011, including newly public companies that qualify as smaller reporting companies after January 21, 2011, will not be subject to Exchange Act Section 14A(a) and Rule 14a-21(a) and (b) until the first annual or other meeting of shareholders at which directors will be elected and for which the rules of the Commission require executive compensation disclosure pursuant to Item 402 of Regulation S-K (17 CFR 229.402) occurring on or after January 21, 2013. From SEC Final Rule 33-8876, SMALLER REPORTING COMPANY REGULATORY RELIEF AND SIMPLIFICATION (Jan 4, 2008) As adopted, a smaller reporting company is defined as a company that meets all of the following criteria: is not an investment company, an asset-backed issuer, or the majority-owned subsidiary of a parent that was not a smaller reporting company; had a public float of less than $75 million as of the last business day of its most recently completed second fiscal quarter; and in the case of an issuer whose public float was zero, had annual revenues of less than $50 million during its most recently completed fiscal year for which audited financial statements are available on the date of the filing. As adopted, once an issuer fails to qualify for smaller company status, it will remain unqualified unless it determines that its public float, as calculated in accordance with the definition, was less than $50 million as of the last business day of its second fiscal quarter.

30

Figure 1. The Distribution of the Public Float in 2009.

The figures depict the public float distribution of all Compustat firms in 2009. The top figure shows the 0 to $1billion range, and the bottom figure focuses on the $25 million to $125 million range.

31

Figure2. Annual Meetings Distribution around the Rule Effective Date.

The figures depict the distribution of the dates of the annual shareholders meetings. The top figure shows the distribution in the year before the rule was enforced. The bottom figure shows the distribution in the year the rule was enforced. All firms that held meetings after January 21, 2011 and were not “smaller reporting company” had to hold a Say-on-Pay and frequency vote.

32

Figure 3. The Say-on-Pay Vote Results.

This figure depicts the distribution of support for the Say-on-Pay vote and the proposal for annual frequency of future Say-on-Pay votes at annual meetings held in 2011 for 350 firms. The Support for CEO Pay variable is the percentage of “For” votes divided by the percentage of “For,” “Against,” and “Abstain” votes cast on the mandatory advisory vote to approve the CEO compensation. The Support for Annual Vote variable is the percentage of votes that support an annual vote on CEO pay divided by votes cast in support of annual, once every two years, and once every three years votes.

33

Figure 4. The Evolution of CEO Pay.

The figures depict the average CEO compensation over time. Firms are split into Mandatory Voting firms (firms that had to hold a Say-on-Pay vote in 2011 based on their public float), Voluntary Voting firms (firms that did not have to hold a Say-on-Pay vote in 2011 but voluntarily chose to do so), and Non-Voting firms (firms that did not hold a Say-on-Pay vote in 2011).

34

Figure 5. The Evolution of Equity and Cash Compensation.

The figures depict the average Equity Compensation Ratio (top) and Cash Compensation Ratio (bottom) over time. Cash Compensation Ratio is the ratio of salary and bonus to total compensation. Equity Compensation Ratio is the ratio of stock awards and option awards to total compensation. Firms are split into Mandatory Voting firms (firms that had to hold a Say-on-Pay vote in 2011 based on their public float), Voluntary Voting firms (firms that did not have to hold a Say-on-Pay vote in 2011 but voluntarily chose to do so), and Non-Voting firms (firms that did not hold a Say-on-Pay vote in 2011).

35

Figure 6. Returns around the Say-on-Pay Vote.

The figures depict the Buy-and-Hold Return (top) and Buy-and-Hold Return over Russell 2000 (bottom) for the three portfolios of firms over the event time. Time 0 corresponds to the company’s annual election data in 2011. The three portfolios are based on Mandatory Voting firms (firms that had to hold a Say-on-Pay vote in 2011 based on their public float), Voluntary Voting firms (firms that did not have to hold a Say-on-Pay vote in 2011 but voluntarily chose to do so), and Non-Voting firms (firms that did not hold a Say-on-Pay vote in 2011).

36

Table 1. Rule Timeline.

The table lists the important dates related to the Say-on-Pay rule , and the timeline for a firm whose fiscal year end is in December. The Dodd-Frank Act was passed and signed into law on July 21st, 2010. For the purposes of our statistical analysis, fiscal 2008 and 2009 years are the pre-treatment period, and 2010 and 2011 fiscal years are the treatment period. We include 2010 in the treatment period to account for firms changing their compensation in anticipation of the new rule. In robustness checks, we have dropped data from 2010 and achieved similar results.

Date Regulatory Event Firm Event (firm with December fiscal year end) Jun 30, 2008 Firm calculates its public float as of the last day of

the second fiscal quarter. Dec 31, 2008

Fiscal year end. Firm reports its public float, checks a box whether it is a “smaller reporting company”. CEO pay for the fiscal year is reported in the firm proxy (in April).

Jun 30, 2009 Firm calculates its public float. Jun 10, 2009 Secretary Geithner states that

the Administration supports "Say on Pay" legislation

Dec 31, 2009

Fiscal year end. CEO pay reported in the firm proxy (in April).

Jun 30, 2010 Firm calculates its public float. Jul 21, 2010 President Obama signs the

Dodd–Frank Wall Street Reform and Consumer Protection Act.

Oct 18, 2010

SEC Proposes Rules on "Say on Pay" (Press Release 2010-198)

Dec 31, 2010

Fiscal year end. CEO pay reported in the firm proxy (in April).

Jan 25, 2011 SEC Final Rule (Press Release 2011-25). Rule becomes effective immediately.

May 2011 If not qualified for the small company exemption, the firm holds the first mandatory advisory vote on executive compensation, generally known as "say-on-pay" votes, and an advisory vote on the desired frequency of say-on-pay votes.

Dec 31, 2011

Fiscal year end. CEO pay reported in the firm proxy (in April).

37

Table 2. Descriptive Statistics

The sample consists of 496 firms over four years. Total Assets is the company total assets (AT) from Compustat. Sales Growth is the log of the ratio of current sales (SALE) to lagged sales. Firm Return is the 12 month return over the previous fiscal year. ROA is the return on assets computed as income before extraordinary items (IBC), divided by the book value of assets (AT). Public Float is the part of equity not held by management or large shareholders, as reported on the first page of the company’s 10-K filing. Staggered Board is a dummy variable equal to one if all board members are not up for reelection the same time. Management Ownership is the total stock ownership of directors and executive officers from the company proxy filing. CEO characteristics (Chairman, Ownership, Age and Tenure) are collected from company proxy filings. First Year CEO is a dummy variable equal to one in the first year of the CEO’s tenure. Total Compensation is the CEO total compensation field as reported on the company proxy statement. Cash Compensation Ratio is the ratio of salary and bonus to total compensation. Equity Compensation Ratio is the ratio of stock awards and option awards to total compensation. Termination and Change of Control are the potential payments upon termination or change in control. Director Vote is the support for directors at the company annual meeting: the average across all directors of the number of “For” votes as a percentage of “For” and “Withheld” votes.

Variable Units Mean Std. Dev. Observ. SourceTotal Assets millions 513 1,051 1,957 CompustatSales Growth rate 0.10 0.84 1,955 CompustatFirm Return % 0.15 0.97 1,907 CRSPROA ratio -0.15 0.59 1,973 CompustatPublic Float millions 125 193 1,957 10-KStaggered Board dummy 0.48 0.50 1,978 DEF 14Management Ownership % 19.96 18.91 1,978 DEF 14CEO is Chairman dummy 0.41 0.49 1,978 DEF 14CEO Ownership % 6.39 11.98 1,978 DEF 14CEO Age years 56.05 8.49 1,978 DEF 14CEO Tenure years 7.91 8.46 1,978 DEF 14First Year CEO dummy 0.10 0.29 1,978 DEF 14Total Compensation $ 1,110,000 1,080,000 1,973 DEF 14Cash Compensation Ratio ratio 0.58 0.33 1,973 DEF 14Equity Compensation ratio 0.25 0.33 1,973 DEF 14Termination $ 1,620,000 1,790,000 1,261 DEF 14Change of Control $ 2,390,000 2,640,000 1,318 DEF 14Director Vote % 92.09 9.82 1,453 8-K

38

Table 3. Treated Versus Non-Treated Firms in 2008 and 2009

This table reports the characteristics of firms that held a Say-on-Pay vote in 2011 and firms that do not hold Say-on-Pay vote in 2011 in the two years preceding the passage of the Say-on-Pay SEC regulation: 2008 and 2009. See Table 1 for variable definitions.

Say-on-Pay Vote No Say-on-Pay Vote Variable Mean Std. Dev. Observ. Mean Std. Dev. Observ. Difference p-values

Total Assets 576 1,222 692 413 741 289 162 0.036

Sales Growth 0.16 0.84 679 0.20 1.01 284 -0.04 0.487

Firm Return -0.23 0.61 646 -0.22 0.53 272 0.00 0.945

ROA -0.16 0.57 692 -0.19 0.64 289 0.03 0.458

Staggered Board 0.50 0.50 695 0.46 0.50 291 0.04 0.267

Mngmt Ownership 20.98 20.18 695 20.37 16.18 291 0.61 0.648

CEO is Chairman 0.43 0.50 695 0.39 0.49 291 0.05 0.162

CEO Ownership 6.94 13.23 695 6.28 10.08 291 0.66 0.449

CEO Age 55.18 8.46 695 56.62 8.36 291 -1.44 0.015

CEO Tenure 7.01 7.80 695 9.05 9.71 291 -2.04 0.001

First Year CEO 0.13 0.34 695 0.10 0.30 291 0.03 0.154

39

Table 4. Event Study Estimations

The table shows the event study results. The dependent variable is the equal-weighted portfolio that buys all companies that were not affected by the new rule and sells all companies that were affected by the rule at the event date. The estimations are based on a 60-day estimation window immediately before the event window. We estimate the following models: Rit=αi+βi1·MKTRFt+εit (market model), and Rit=αi+βi1·MKTRFt+ βi2·SMBt+ βi3·HMLt+ βi4·UMDt+εit (four factor model), E(εit)=0, var(εit)=σ

2, for the 60-day estimation window. Rit is the portfolio return. MKTRFt , SMBt, HMLt, and UMDt are the return on the market, the Fama-French size, book-to-market, and momentum factors. We use the predicted normal portfolio returns for the three-day event window to calculate cumulative abnormal returns (MacKinley (1997)). Two-sided t-statistics are reported in brackets. ∗, ∗∗, and ∗∗∗ denote two-sided statistical significance at the 10%, 5%, and 1% levels, respectively. (#) Event Date, Event Description

Market Model Four Factor Model

(1) Oct. 18, 2010 SEC Proposes Rules on Say-on-Pay and Proxy Vote Reporting, SEC Press Release 2010-198

-0.63% [-0.71]

-0.81 [-1.31]

(2) Jan. 25, 2011 SEC Adopts Rules for Say-on-Pay as Required Under Dodd-Frank Act, SEC Press Release

-2.13*** [-2.73]

-1.45** [-2.10]

40

Table 5. Director Voting

The table shows OLS and IV regressions in which the dependent variable is the average shareholder support for all directors at the company meeting: the average of the number of “For” votes as a percentage of “For” and “Withhold” votes. Say-on-Pay Vote is a dummy variable equal to one in the year of the Say-on-Pay vote if the firm held a Say-on-Pay vote. Say-on-Pay Vote is instrumented by Mandatory Voting - the company treatment status based on the company public float position relative to the $75 million cut-off. We also report the first stage coefficient on the instrument (Mandatory Voting) and the partial F-test of the instrument. All other variables are summarized in Table 2. The numbers in parentheses are robust standard errors clustered on the firm level. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.

OLS OLS IV IV

Say-on-Pay Vote 1.9790** (0.917)

2.1127** (0.945)

2.0555* (1.158)

2.2474* (1.177)

Log(Total Assets)

-2.3530* (1.372)

-2.3664* (1.353)

Sales Growth

0.2999 (0.275)

0.2997 (0.274)

Firm Return

0.5517* (0.286)

0.5492* (0.284)

ROA

1.0181 (1.028)

1.0204 (1.019)

Staggered Board

-0.0514 (1.528)

-0.0362 (1.532)

CEO is Chairman

0.7101 (1.634)

0.7076 (1.628)

CEO Ownership

0.0260 (0.046)

0.0257 (0.046)

Management Ownership

0.0182 (0.027)

0.0185 (0.027)

Firm Fixed Effects Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Adjusted R-squared 0.244 0.246 0.240 0.242 Observations 1,433 1,433 1,433 1,433 First-Stage Regression, Voted instrumented by Mandatory Voting, including first-stage controls, public float, and fixed effects

Mandatory Voting 0.7699*** (0.038)

0.7626*** (0.038)

First Stage R-squared 0.872 0.874 Partial F-test 398.44 398.00

41

Table 6. Total Compensation Regressions

The table shows the results of panel regressions, where the dependent variable is logarithm of total CEO compensation. The sample consists of data on 488 firms between 2008 and 2011. Say-on-Pay Vote is a dummy variable equal to one in 2010 and 2011 if the firm had a Say-on-Pay vote. Say-on-Pay Vote is instrumented by Mandatory Voting - the company treatment status based on the company public float position relative to the $75 million cut-off. All other variables are summarized in Table 2. The numbers in parentheses are robust standard errors clustered on the firm level . *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.

OLS OLS IV IV

Say-on-Pay Vote 0.0953** (0.048)

0.0981** (0.049)

0.1429** (0.061)

0.1410** (0.063)

Log(Total Assets)

0.0104 (0.081)

0.0092 (0.081)

Sales Growth

-0.0013 (0.016)

-0.0012 (0.016)

Firm Return

0.0314* (0.018)

0.0303* (0.018)

ROA

0.0325 (0.042)

0.0306 (0.042)

Staggered Board

0.2128 (0.159)

0.2159 (0.159)

CEO is Chairman

0.4477*** (0.153)

0.4520*** (0.152)

CEO Ownership

-0.0014 (0.006)

-0.0012 (0.006)

Management Ownership

0.0009 (0.003)

0.0009 (0.003)

Firm Fixed Effects Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Adjusted R-squared 0.712 0.721 0.713 0.722 Observations 1,692 1,692 1,692 1,692 First-Stage Regression, Voted instrumented by Mandatory Voting, including first-stage controls, public float, and fixed effects

Mandatory Voting 0.7731*** (0.038)

0.7690*** (0.038)

First Stage R-squared 0.906 0.907 Partial F-test 421.38 402.02

42

Table 7. Sensitivity of Pay to Performance

The table shows the results of panel regressions, where the dependent variable is logarithm of total CEO compensation. The sample consists of data on 488 firms between 2008 and 2011. Post is a dummy variable equal to in the post treatment period if the firm had a Say-on-Pay vote. Pre is a dummy variable equal to one minus the Post dummy. The Pre and Post dummies are interacted with the raw firm returns (models [1] and [2]), and with the industry adjusted returns using the Fama-French 49 industry classification (models [3] and [4]). Models [2] and [4] control for Log(Total Assets), Sales growth, ROA, Staggered Board, CEO Chairman, CEO Ownership, Management Ownership, CEO Age, CEO tenure, and Public Float (summarized in Table 2). The numbers in parentheses are robust standard errors clustered at the firm level. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.

(1) (2) (3) (4)

Return*Pre 0.0018 (0.028)

0.0050 (0.028)

Return*Post 0.0526*** (0.020)

0.0555*** (0.019)

Adj. Return* Pre

-0.0045 (0.028)

-0.0009 (0.028)

Adj. Return*Post

0.0524** (0.021)

0.0552*** (0.020)

Industry Return* Pre

0.0817 (0.050)

0.0807 (0.051)

Industry Return*Post

0.0606 (0.084)

0.0649 (0.084)

Additional Controls No Yes No Yes Firm Fixed Effects Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Adjusted R-squared 0.714 0.722 0.714 0.722 Observations 1,692 1,692 1,692 1,692

43

Table 8. Compensation Ratios

The table shows the results of panel regressions, where the dependent variables are ratios of the CEO compensation components to the total CEO compensation. Equity Compensation Ratio is the ratio of stock awards and option awards to total compensation. Log(Termination) and Log(Change of Control) are the natural logarithm of the pre-disclosed payments upon termination or change in control. The sample consists of data on 488 firms between 2008 and 2011. Say-on-Pay Vote is a dummy variable equal to one in 2010 and 2011 if the firm had a Say-on-Pay vote. Say-on-Pay Vote is instrumented by Mandatory Voting - the company treatment status based on the company public float position relative to the $75 million cut-off. All other variables are summarized in Table 2. The numbers in parentheses are robust standard errors clustered at the firm level. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.

Cash Ratio Equity Ratio Termination (log)

Change of Control (log)

Say-on-Pay Vote -0.0064 (0.036)

0.0027 (0.036)

0.0182 (0.080)

0.0451 (0.108)

Log(Total Assets) 0.0023 (0.027)

-0.0068 (0.021)

-0.1381 (0.134)

-0.1082 (0.140)

Sales Growth 0.0015 (0.006)

-0.0004 (0.006)

-0.0174 (0.021)

-0.0193 (0.021)

Firm Return -0.0068 (0.006)

0.0116 (0.007)

0.0325* (0.017)

0.0292 (0.018)

ROA 0.0001 (0.018)

0.0112 (0.016)

0.0836 (0.054)

0.0429 (0.050)

Staggered Board 0.0147 (0.057)

-0.0411 (0.068)

0.0343 (0.191)

-0.1962 (0.206)

CEO is Chairman -0.2160* (0.119)

0.1997* (0.121)

-0.0142 (0.319)

-0.0363 (0.351)

CEO Ownership 0.0040** (0.002)

-0.0019 (0.002)

0.0049 (0.005)

0.0072 (0.005)

Management Ownership

0.0002 (0.001)

-0.0004 (0.001)

-0.0036 (0.003)

-0.0104*** (0.003)

Firm Fixed Effects Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Adjusted R-squared 0.337 0.289 0.805 0.779 Observations 1,692 1,692 1,116 1,177 First-Stage Regression, Voted instrumented by Mandatory Voting, including first-stage controls, public float, and fixed effects

Mandatory Voting 0.7690*** (0.038)

0.7690*** (0.038)

0.783*** (0.054)

0.770*** (0.055)

First Stage R-squared 0.907 0.907 0.935 0.930 Partial F-test 402.02 402.02 208.04 197.65

44

Table 9. Support for Annual Votes

The table shows the results of a cross-sectional regression of firms that held a Say-on-Pay vote, where the dependent variable is the percentage support for holding annual Say-on-Pay votes. Voluntary Say-on-Pay Vote is a dummy variable equal to one if a firm was not required to hold a vote and held a voluntary Say-on-Pay vote. The sample consists of data on 349 firms that held elections in 2011. All other variables are summarized in Table 2. The numbers in parentheses are robust standard errors. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.

(1) (2) (3)

Voluntary Say-on-Pay Vote -14.5323*** (5.417)

-11.8379** (5.472)

-9.4122* (5.379)

Staggered Board

1.9978 (3.169)

2.8546 (3.154)

CEO is Chairman

-2.0814 (3.388)

-3.8450 (3.472)

CEO Ownership

-0.2603 (0.200)

-0.2408 (0.197)

Management Ownership

-0.2001* (0.111)

-0.2220** (0.112)

Log(Total Assets)

2.9411** (1.226)

Sales Growth

-3.4324* (1.860)

Firm Return

2.2962 (1.461)

ROA

6.9587* (3.972)

Public Float

0.0002 (0.019)

Adjusted R-squared 0.022 0.060 0.083

Observations 349 349 349