A. Impact of the Feedback Effect on Corporate Policy

To understand how stock prices affect corporate decisions, previous research examines the real effects of the stock market on various corporate policies such as investment decisions (Baker, Stein, and Wurgler (Reference Baker, Stein and Wurgler2003), Chen, Goldstein, and Jiang (Reference Chen, Goldstein and Jiang2007), Campello and Graham (Reference Campello and Graham2013), Edmans, Jayaraman, and Schneemeier (Reference Edmans, Jayaraman and Schneemeier2017), Bird et al. (Reference Bird, Karolyi, Ruchti and Truong2021), and Jayaraman and Wu (Reference Jayaraman and Wu2020)), takeovers (Edmans, Goldstein, and Jiang (Reference Edmans, Goldstein and Jiang2012), Baker, Pan, and Wurgler (Reference Baker, Pan and Wurgler2012)), financing decisions (Dittmar, Duchin, and Zhang (Reference Dittmar, Duchin and Zhang2020), Hovakimian and Hu (Reference Hovakimian and Hu2020)), and voluntary disclosure decisions (Zuo (Reference Zuo2016)). This is because stock prices provide valuable information to which CEOs do not otherwise have access. We use the terms “learning” and “feedback” interchangeably because, in a broader sense, market prices can confirm underlying information, even if decision-makers do not learn new information. Prior studies analyze the effects of feedback on corporate policies, and we extend this literature by examining the impact of feedback on board governance decisions, specifically the decision to set CEO pay.

Recent studies employ abnormal stock returns surrounding a specific corporate event to measure private investor information not known to managers. This method ensures that stock-price changes reflect investors’ private information about a particular issue, which improves the accuracy of the information managers can learn from stock prices. For instance, Jayaraman and Wu (Reference Jayaraman and Wu2020) show that managers increase (decrease) annual capital expenditures in response to positive (negative) stock-market reactions to capital expenditure forecasts. Previous research indicates that the market reacts positively to news related to SOP that enhances corporate governance (Cai and Walkling (Reference Cai and Walkling2011), Illiev and Vitanova (Reference Illiev and Vitanova2019)).

B. Spillover Effects of Shareholder Activism

Although prior literature examines the spillover effects of shareholder activism, little research explores the spillover effects of a mandated SOP vote.Footnote ¹ One notable exception is the study by Denis, Jochem, and Rajamani (Reference Denis, Jochem and Rajamani2020), who investigate a sample of S&P 1500 firms and their compensation benchmarking peers that receive weak SOP voting support (less than the 10th percentile of the vote distribution for Russell 3000 firms) from 2011 to 2012. Their findings suggest that nontarget firms reduce executive pay following their peers’ voting failure. The authors identify two channels through which this effect occurs: First, boards directly learn governance information from peer firms’ voting outcomes; second, when compensation peers reduce CEO pay, focal firms’ boards follow suit to maintain the compensation benchmarking.

C. Hypothesis Development

In the context of the SEC’s SOP rule, if a spillover effect exists, the voting failure of a compensation benchmarking peer should lead to dissatisfaction among the focal firm’s shareholders about executive pay based on an improper benchmark and possible overstatement. This argument aligns with Gleason, Jenkins, and Johnson (Reference Gleason, Jenkins and Johnson2008), who observe negative market reactions among the peers of restated firms due to investor suspicions about financial reporting quality. Dissatisfied investors may sell the focal firm’s shares upon hearing about their compensation peers’ SOP failure. The focal firm’s abnormal return around compensation peers’ SOP news should be lower when the latter’s voting failure is more severe, as investors are more likely to follow news about other firms they consider important. The spillover effect on the focal firm should be stronger when a peer firm experiences a more substantial abnormal return, as it partially reveals the event’s severity. Hence, we present our first hypothesis as follows:

Prior research suggests that managers can learn private information from investors. Specifically, when a focal firm experiences a negative market response to its compensation peers’ SOP voting failure, it indicates shareholder dissatisfaction with the focal firm’s existing CEO compensation. This private information can help the board appropriately adjust the CEO compensation.

We argue that boards take note of shareholder responses to peers’ SOP news and use this information to inform future decisions. Therefore, if a firm experiences a more negative abnormal return following its compensation peers’ SOP news and does not take any corrective action, the firm is more likely to face a voting failure and severe reputational damage. As a result, we expect that these firms to be more motivated to repair their reputations and mitigate the risk of future voting failures by improving their executive pay practices, such as substantially decreasing CEO pay. Thus, we present our second hypothesis as follows:

We investigate three factors that affect the feedback effect. First, the feedback effect’s strength depends on the relative power of the CEO and the board. When the board is more powerful than the CEO, a negative CAR should be more effective in reducing excessive CEO pay because the CEO is more likely to accept a pay cut. Additionally, board members may find it beneficial to establish a reputation for being shareholder-friendly (by cutting CEO pay) rather than management-friendly (by not cutting CEO pay), as both reputations exist in the board labor market (Levit and Malendo (Reference Levit and Malenko2016)).

Second, the feedback effect depends on third-party intermediaries’ attitudes toward the CEO pay and the availability of alternative information channels. Board members are more likely to respond to investor sentiment communicated through stock prices, and they are more likely to reduce CEO pay when third-party intermediaries such as proxy advisory firms express concerns.

Third, we examine the interaction between professional compensation consultants and information from stock prices. If consulting opinions and stock prices are substitutes, board members are more likely to rely on stock price when the consultant cannot provide reliable advice. We predict that the feedback effect is stronger for firms that hire smaller compensation consultant firms (i.e., with fewer clients) because these firms are less likely to provide reliable advice. Hence, we present our third hypothesis, as follows:

Investors reward (or punish) directors who learn information about their opinions on CEO pay from stock prices. In particular, if directors cut CEO pay after adverse returns (around peers’ SOP failure events), it indicates learning from investor opinions revealed in stock prices. Consequently, investors should perceive such directors as more capable and worthy of higher support in future director elections. We argue that the relation between CEO pay cuts and favorable voting focuses mainly on directors on compensation committees (the effect may exist to a lesser extent for directors not on compensation committees because some investors may not have information to distinguish these two types of directors).

This prediction is consistent with Coles and Hoi (Reference Coles and Hoi2003), who find that directors who demonstrate shareholder-friendly behavior and monitor management are more likely to gain additional directorships. The prediction is also consistent with Fos and Tsoutsoura (Reference Fos and Tsoutsoura2014), who find that proxy contests have a significant adverse impact on incumbent directors. Moreover, anecdotal evidence supports our prediction, as some institutional investors explicitly state their intention to vote for directors who take into account investor concerns about executive pay.Footnote ² Hence, we present our fourth hypothesis as follows:

A. Data and Sample

We obtain data from Incentive Lab on the composition of the peer groups used for compensation benchmarking. The vendor provides detailed data on executive compensation for S&P 500 firms and for the majority of S&P 400 firms. To attract and retain competent managers, firms typically benchmark executive pay against comparable peers that are competitors for labor-market talent (Faulkender and Yang (Reference Faulkender and Yang2010), Bizjak, Lemmon, and Nguyen (Reference Bizjak, Lemmon and Nguyen2011), and Albuquerque, DeFranco, and Verdi (Reference Albuquerque, DeFranco and Verdi2013)). Starting in 2006, the SEC requires public firms to disclose in the CD&A, “any benchmarking of total compensation or any material element of compensation, identifying the benchmark and, if applicable, its components (including component companies)” (SEC final rules 33-8732a: Executive Compensation and Related Person Disclosure). According to Equilar’s (2016) report, more than 95% of S&P 500 firms disclose the use of compensation peers in determining executive pay. Appendix A provides an example of Apple, Inc.’s disclosure regarding its compensation benchmarking peer group. We utilize these actual self-selected peer firms, instead of hypothetical peers constructed based on certain firm characteristics (e.g., firm size and industry membership), to document directly the feedback effect of SOP voting. Compensation committees typically consider two types of peers in a compensation benchmark: compensation peers, which are used for setting overall CEO pay, and bonus peers, which determine performance-based awards (Francis et al. (Reference Francis, Hasan, Mani and Ye2016), Edmans, Gabaix, and Jenter (Reference Edmans, Grabaix, Jenter, Hermalin and Weisbach2017), and Bizjak et al. (Reference Bizjak, Kalpathy, Li and Young2022)). We focus on compensation peers in our article because relatively few firms report bonus peers (Francis et al. (Reference Francis, Hasan, Mani and Ye2016)).

We identify firms that fail in SOP voting (i.e., negative votes exceed 50%) from Semler Brossy’s annual say-on-pay reports. An SOP vote failure indicates that investors are highly dissatisfied with existing CEO pay. We obtain financial accounting data from Compustat, executive compensation data from ExecuComp, stock-price data from CRSP, corporate-governance information from Institutional Investors Services (ISS), and institutional shareholdings from Thomson Reuters Institutional (13F) Holdings.

Panel A of Table 1 reports the sample-selection process. We start with 7,432 firm-year observations representing 1,157 unique focal firms covered by Incentive Lab during the proxy seasons of 2011–2018. We first exclude 364 observations that are missing data needed for the empirical analyses. We then exclude 135 observations in which the focal firms fail in the SOP voting because the market reaction to the focal firms’ voting failure may confound the spillover effect from the peer failures.Footnote ³ We further exclude 5,582 observations (1,088 focal firms) without peers that fail in the voting. Hence, the final sample consists of 1,351 firm-event observations (686 focal firms) with at least one peer that experiences voting failure (85.6% of the focal firms have only one SOP-failed peer, 12.1% have two peers that fail in the SOP voting, and the remaining 2.3% have more than two peers with the voting failure). For the analyses on pay changes from the pre- to post-event years, each firm-event observation corresponds to two firm-year observations (i.e., one observation for the pre-event year and another for the post-event year). Accordingly, there are 2,702 firm-year observations in the main sample.

Table 1Sample Selection and Summary Statistics

B. Summary Statistics

Panel B of Table 1 reports the descriptive statistics for the variables used in the main empirical analyses. The mean of the focal firms’ total CEO pay (CEOPAY) is $8.6 million, and the natural logarithm of the focal firms’ CEO compensation (LN CEOPAY) has a mean of 8.717. The mean of focal firms’ market reactions to their peers’ voting failure (CAR) in the full sample is close to 0, suggesting that news of compensation peers’ SOP voting failures does not affect all focal firms.

Regarding event severity, the mean percentage of negative votes the SOP-failed peers receive is 61.7%, and the mean number (percentage) of SOP-failed peers is 1 (7.9%) in the focal firms’ compensation peer group (with a mean group size of 22). Regarding SOP-failed peers’ own market reactions to the event, the mean 3-day CAR is −0.3%. On average, the common institutional ownership between the focal firm and its SOP-failed peer (COMINST PCT) is 52.6%. For 1.0% of the focal firms, at least one director has a board seat at the SOP-failed peers (INTERLOCK), and 1.9% of the focal firms have an industry peer that is not the compensation peer that fails in the SOP voting (INDPR FAILSOP).

Relative to compensation peers, 44.9% of focal firms have CEO compensation above the median pay for the peer group (CEOPAY ABOVE). The mean firm size (SIZE) is 8.837, equivalent to approximately $6.9 billion of market capitalization. The means of the market-adjusted annual stock returns and the industry-adjusted return on assets are 0.026 (0.060) and 1.097 (0.654) for the current (prior) year, respectively.Footnote ⁴ CEO tenure is 8.1 years on average. The mean CEO ownership is 1.3%, and the mean institutional ownership is 60.0%. The mean percentage of favorable votes is 76.2%, and 9.2% of the focal firms experience a negative voting recommendation by ISS.

With respect to peer firms’ characteristics and labor-market conditions, the mean PR CEOPAY is 9.298, which is equivalent to approximately $10.8 million of CEO compensation paid by the SOP-failed peers; these peers’ average stock performance (PR RET) is 7.1%, and accounting performance (PR ROA) is 4.0% on average. For 37.8% of the focal firms, at least one of the top five executives moves from or to the industry of their SOP-failed peers. In terms of similarity of CEO capability, the mean difference in the focal firms’ CEOs’ stock performance relative to their SOP-failed peers is 20.5%.Footnote ⁵ Also, 7.2% of the focal firms experience CEO turnover, and 9.4% of the focal firms have a CEO who is 65 years old or older.

To explore the similarities and differences among the focal firms in our sample and the other firms covered by Incentive Lab, we compare the main variables for the firms with SOP-failed peers (1,351 firm-year observations in the final sample) and the firms without any peers that fail in the voting (5,582 firm-year observations, which we exclude from the final sample).

We report the results in Appendix C. We find that for CEO-related variables, the two groups of focal firms have similar levels of CEO pay (LN CEOPAY) and their compensation peers’ (CPs’) pay (PR MEDPAY) is also similar. However, the CP-SOP-failure group (in the first 2 columns) has higher co-opted boards and longer CEO tenure than the remaining firms, implying that the directors in the former group are less effective in identifying proper CPs than are directors in the latter group. We find that for firm fundamentals, the CP-SOP-failure group has similar firm size, stock returns, and institutional ownership, but lower ROA, book-to-market ratio, and leverage compared to the remaining firms. For voting-related variables, we find no significant differences between the two groups.

Since differences in CEO-level or firm-level variables can explain the occurrence of SOP-failed CPs and the compensation of focal firms, we control for these variables in our main empirical tests of focal firms’ compensation.

Figure 1 illustrates the framework for this analysis. Note that there are two elements of voting failure for compensation peers: the voting failure itself and the market response to the failure announcement. Both factors may lead to market responses for the focal firm and compensation adjustments. We test whether a focal firm’s stock returns in response to such an event have a feedback effect on its board’s compensation-reduction decision, beyond the effects from compensation peers’ voting results.

Figure 1Framework of the Feedback Effect Triggered by Peers’ SOP Voting Failure

Figure 1 plots the channels of learning for focal firm directors after a firm’s compensation peers announce an SOP voting failure. The peer firm’s SOP voting failure may cause a negative market response toward the focal firm. The focal firm’s CAR provides an additional source of information for the firm’s directors regarding shareholder opinion about the event. The focal firm may cut CEO pay after its directors observe the compensation peer firm’s SOP voting failure and the focal firm’s own CAR around the peer’s event.

C. Market Responses to Compensation Peers’ Voting Failure

In this section, we show that the response to the news is heterogeneous. First, we find that focal firms are more likely to experience a lower CAR when their compensation peers’ SOP failure is severe, suggesting that the focal firms’ investors observe the severe SOP failure of compensation peers and then convey their dissatisfaction with it. We view a compensation peer’s SOP failure as severe if the peer receives a high percentage of negative votes or if a large fraction of peer firms experiences SOP voting failures. We also find that focal firms’ CAR is lower when the compensation peer in question experiences a lower CAR around its SOP voting failure. These results suggest that the spillover effect is stronger when the market views the event as more salient and more negative.

In Hypothesis 1, we posit that the immediate market reaction reflects shareholder dissatisfaction with the focal firm’s inappropriate benchmarking and the consequently biased executive pay. We test Hypothesis 1 with the following OLS regression:

(1) $$ {CAR}_{i,t}={\alpha}_0+{\alpha}_1 PR\hskip0.32em {SEVERITY}_{i,t}+{\alpha}_2 PR\hskip0.32em {CAR}_{i,t}+ Controls. $$

The dependent variable, CAR, captures the immediate market reaction to the peer event. To construct this variable, we first calculate the focal firm’s 3-day cumulative abnormal market-adjusted returns (CAR) from 1 day before to 1 day after the SOP-failed peer’s annual shareholder meeting. If a focal firm has more than 1 SOP-failed peer, we first calculate its CAR around the announcement date of each peer’s SOP failure. Then, we use the firm’s mean CAR, measured over these multiple-event periods as the CAR in year t. Footnote ⁶ A more negative market reaction would reflect stronger shareholder dissatisfaction with executive pay that is based on a potentially biased compensation benchmark.

The first independent variable of interest, PR SEVERITY, measures the extent of the event severity for a focal firm and is determined by two components. The first is the percentage of negative (i.e., dissenting) votes the SOP-failed peer receives (PR NEGVOTE). If a firm has more than one SOP-failed peer in 1 year, we use the mean PR NEGVOTE. The second component is the proportion of SOP-failed peers in the focal firm’s compensation benchmarking group (PR FAILPCT), which is equal to the number of SOP-failed peers (PR FAILNUM) divided by the peer group size (PR TOTNUM). We then define PR SEVERITY as a dummy variable equal to 1 if either PR NEGVOTE or PR FAILPCT is in the top quintile of the sample distribution, and 0 otherwise. The second independent variable, PR CAR, indicates the extent of an SOP-failed peer’s market reaction to its own voting failure, which is measured by the 3-day CAR surrounding its annual shareholders’ meeting date. Hypothesis 1 predicts a negative coefficient on PR SEVERITY and a positive coefficient on PR CAR in equation (1).

We include the following control variables, which could affect focal firms’ market reactions to their peers’ events. It is plausible that the spillover effect could go through the channels of common institutional investors (COMINST), shared directors (INTERLOCK), or other industry peers that experience voting failure (INDPR FAILSOP). We also control for a set of focal-firm and peer-firm characteristics (e.g., firm size, growth, profitability, and stock performance). Last, we control for firm and year fixed effects.

Table 2 reports the results for the regression based on equation (1). In column 1, the regression result shows a negative coefficient on PR SEVERITY (the coefficient = −0.004 and t = −2.02) and a positive coefficient on PR CAR (coefficient = 0.188 and t = 6.46), consistent with Hypothesis 1. The magnitude of market reaction is not negligible, as more severe voting failure decreases a focal firm’s 3-day CAR by 0.40%. Moreover, as SOP-failed peer CAR decreases by 1%, focal firm CAR decreases by 0.19% accordingly. In column 2, we include two components of PR SEVERITY (i.e., PR NEGVOTE, which is the percentage of negative votes the SOP-failed peer receives, and PR FAILPCT, which is the percentage of SOP-failed peers in the compensation benchmarking group). We find a significantly negative coefficient on PR NEGVOTE (coefficient = −0.025 and t = −2.39), suggesting that the focal firms have a more negative CAR when peers’ SOP voting receives more dissenting (i.e., negative) votes. Instead, the proportion of the focal firm’s compensation peers that fail in the voting (PR FAILPCT) does not drive the market reaction. The result is consistent with the notion that a focal firm’s negative market reaction mostly reflects shareholder dissatisfaction with its pay practices when its compensation peer experiences stronger investor opposition in the SOP voting. Our finding supports the view that investors regard SOP as a value-creating mechanism (Cai and Walkling (Reference Cai and Walkling2011), Ferri and Maber (Reference Ferri and Maber2013), Cuñat, Giné, and Guadalupe (Reference Cuñat, Giné and Guadalupe2016), and Iliev and Vitanova (Reference Illiev and Vitanova2019)). The control variables have weak effects on CAR, except for the proxy advisor’s recommendation to vote against CEO pay (ISSAGAINST).

Table 2Sensitivity of Focal Firms’ Prices to the Severity of Peers’ SOP-Failure Events

Overall, the results suggest that the severity of the compensation peers’ SOP voting failure and the extent of the peers’ own market reactions mainly determine the market responses to the focal firms (around peers’ SOP events). The results ensure the validity of using market reactions around the peer event to capture investor dissatisfaction with focal firms’ compensation benchmarking. It is notable that the adjusted R ² is low (13.37%), suggesting that the market reactions may also reflect news other than the severity of the peer event on adjacent dates.

D. Subsequent CEO Pay Reductions due to the Feedback Effect

We use a difference-in-differences design to test the feedback effect (i.e., focal firms learn from the market and subsequently cut CEO pay), as developed in Hypothesis 2:

(2) $$ {\displaystyle \begin{array}{c} LN\hskip0.32em {CEOPAY}_{i,t}={\beta}_0+{\beta}_1{POST}_{i,t}\hskip0.32em \ast \hskip0.32em {NEGCAR}_{\mathrm{i},\mathrm{t}-1}+{\beta}_2{NEGCAR}_{i,t-1}\\ {}+{\beta}_3{POST}_{i,t}+{\beta}_4{POST}_{i,t}\hskip0.32em \ast \hskip0.32em PR\hskip0.32em {SEVERITY}_{\mathrm{i},\mathrm{t}-1}\\ {}+{\beta}_5 PR\hskip0.32em {SEVERITY}_{i,t-1}+{\beta}_6{POST}_{i,t}\hskip0.32em \ast \hskip0.32em PR\hskip0.32em {NEGCAR}_{i,t-1}\\ {}+{\beta}_7 PR\hskip0.32em {NEGCAR}_{i,t-1}+ Controls.\end{array}} $$

The dependent variable, LN CEOPAY, is the natural logarithm of total focal firm CEO compensation, measured in the most recent year after a compensation peer’s SOP failure. We use the dummy variable, NEGCAR, as the independent variable to capture a strong versus weak negative market reaction for the focal firms. That is, NEGCAR equals 1 if the CAR is below the annual sample median, and 0 otherwise. We define the post-event year (POST = 1) as the first year when a focal firm can adjust CEO pay subsequent to its peer’s voting failure. To be the post-event year, we require that the focal firm’s fiscal year end is at least 6 months after its peer’s SOP event because it takes time for the focal firm’s directors to adjust their compensation design (Denis, Jochem, and Rajamani (Reference Denis, Jochem and Rajamani2020)). The pre-event year (POST = 0) refers to the year prior to the post-event year. Such identification of the focal firm’s pre- and post-event periods provides a more precise way to capture the board’s pay-cut decision in response to the peer’s voting failure. In a robustness test reported in Section III.G.3, we use the 1-year window as a relatively rough way of identifying the timing of the board’s action.Footnote ⁷ If the feedback effect exists, we expect the coefficient on POST * NEGCAR to be negative.

We include PR SEVERITY (the severity of the peer event), which captures the spillover effect through the learning-from-peers and compensation-targeting channels (Denis, Jochem, and Rajamani (Reference Denis, Jochem and Rajamani2020)), as well as its interaction term with POST. If these channels coexist with the learning-from-prices channel in our study, the coefficient on POST * PR SEVERITY would be negative. In addition, we also include PR NEGCAR (a dummy variable that equals 1 if SOP-failed peers’ CAR is negative, and 0 otherwise) and its interaction with POST to test the possibility that the focal firm’s board also learns from its SOP-failed peer’s market reaction.

We control for a comprehensive set of variables that could affect focal firms’ CEO pay. First, it is plausible that common institutional investors, shared directors, or other industry peers that experience voting failure may affect CEO pay. We, therefore, control for common institutional ownership (COMINST, which equals 1 if the percentage of the focal firm and the SOP-failed peer’s shares owned by a common institutional investor is above the annual sample median, and 0 otherwise), board interlock (INTERLOCK, which equals 1 if a focal firm director also serves on the board of the SOP-failed peer, and 0 otherwise), and an industry-peer effect (INDPR FAILSOP, which equals 1 if the focal firm has an SOP-failed industry peer that is not the compensation peer, and 0 otherwise).

Second, following prior studies (e.g., Core, Guay, and Larcker (Reference Core, Guay and Larcker2008)), we control for focal firm characteristics and CEO characteristics, including the focal firm’s lagged CEO pay (lagged LN CEOPAY), its CEO pay relative to its compensation peers (CEOPAY ABOVE, which equals 1 if the focal firm’s CEO pay is above the median for the compensation peer group, and 0 otherwise), median peer pay (PR MEDPAY, the median of the natural logarithm of compensation peers’ CEO pay), firm size (SIZE, the natural logarithm of the market value of equity), growth opportunity (BTM, the book-to-market ratio), leverage (LEV, the leverage ratio), stock performance (RET, the annual market-adjusted stock returns), accounting profitability (ROA, the industry-adjusted return on assets), sales (SALE, the natural logarithm of sales revenue), CEO tenure (LN TENURE, the natural logarithm of the number of years the CEO has served in that role), CEO ownership (CEOOWN, the percentage of shares the CEO owns), and institutional ownership (INST, the percentage of shares institutional investors own).

Third, we further control for the focal firm’s own SOP voting outcome (SOPVOTE, the percentage of favorable votes the focal firm receives), the proxy advisor’s voting recommendation (ISSAGAINST, which equals 1 if ISS recommends voting against the focal firm, and 0 otherwise), and SOP voting frequency (SOPFREQ, which equals 1 if the focal firm votes on SOP frequently, and 0 otherwise).

Fourth, we control for the key independent variables in Denis, Jochem, and Rajamani (Reference Denis, Jochem and Rajamani2020), who suggest that peer firm characteristics and labor-market conditions inform the pay decisions a focal firm’s board makes in response to its peer’s SOP voting failure. Hence, we further control for the peer’s CEO pay (PR CEOPAY, the natural logarithm of the SOP-failed peer’s total CEO compensation), the peer’s CEO pay relative to other compensation peers (PR PAYABOVE, which equals 1 if the SOP-failed peer’s CEO pay is above the median for the compensation peer group, and 0 otherwise), and the peer’s performance (PR RET, the annual stock returns of the SOP-failed peer; and PR ROA, the return on assets of the SOP-failed peer). We also control for labor market commonality and mobility, including TALENTFLOW, which equals 1 if at least one of the top five executives moves between the focal firm and the industry of the SOP-failed peer in the most recent 5 years, and 0 otherwise; DIFF CEORET, the difference in CEO capability, measured by the mean market-adjusted annual stock returns for the most recent 3 years of the CEO’s tenure, between the focal firm and its SOP-failed peers; TURNOVER, which equals 1 if the focal firm experiences CEO turnover, and 0 otherwise; and RETIRE, which equals 1 if the focal firm’s CEO is at least 65 years old, and 0 otherwise (Jenter and Kanaan (Reference Jenter and Kanaan2015), Cai, Xu, and Yang (Reference Cai, Xu and Yang2021)). Last, we control for firm and year fixed effects.

Table 3 reports the regression results for testing the feedback effect on the board’s pay-cut decision, as developed in Hypothesis 2. We first conduct the regressions of LN CEOPAY (the natural logarithm of the focal firm’s total CEO pay) on NEGCAR (the dummy variable for a strong versus weak negative market reaction to the peer event), as specified in equation (2), without including other learning channels (i.e., the event severity and the SOP-failed peer’s market reaction). In column 1, the coefficients on the interaction term POST * NEGCAR are significantly negative (coefficient = −0.134 and t = −3.70). The results show that focal firms experiencing a negative market reaction to their peers’ voting failure reduce CEO pay by more than 12.54% (1 – e^−0.134) in the subsequent year relative to those with no adverse market-spillover effect. Given the sample mean of CEO pay is $8.62 million, the economic magnitude of our effect is around $1.08 million (12.54% * $8.62 million). This result is consistent with Hypothesis 2, which suggests that focal firms learn from the market to adjust potentially biased CEO compensation downward.Footnote ⁸

Table 3Feedback Effect and Focal Firms’ CEO Pay – A Difference-in-Differences Analysis

We then analyze whether the board learns from other channels, including the severity of the peer event (PR SEVERITY) and the peer’s negative market reaction (PR NEGCAR). In column 2, we control for PR SEVERITY and its interaction with the post-event dummy, which captures the effect through the channel in Denis et al. (Reference Denis, Jochem and Rajamani2020), without considering focal firms’ market reactions surrounding the event (NEGCAR). The coefficient on the interaction term POST * PR SEVERITY is negative and significant (coefficient = −0.089 and t = −2.25). The results corroborate Denis, Jochem, and Rajamani (Reference Denis, Jochem and Rajamani2020), who find that focal firms with SOP voting failures subsequently cut CEO pay. For the peer’s negative CAR, the insignificant coefficient on POST * PR NEGCAR suggests that the focal firm’s board does not learn from the market reaction of an SOP-failed peer when deciding to adjust CEO pay.

In column 3, we include the market-spillover variable NEGCAR, together with proxies for other learning channels in the regression. The coefficient on POST * NEGCAR continues to be significantly negative (−0.130, t-value is −3.55), and the coefficient on POST * PR SEVERITY also remains significant (−0.083, t-value is −2.09).Footnote ⁹ The results suggest that focal firms’ CAR (e.g., NEGCAR) contains information beyond peers’ SOP voting outcome and that it plays a more important role in the board’s compensation policymaking than does peer voting outcome (e.g., PR SEVERITY). In other words, the board learns more from its own firm’s prices than it does from peer voting failures. These results indicate that more than any other learning channels, the feedback effect has an incremental impact on CEO pay reduction at focal firms. The overall findings thus support Hypothesis 2.

Our study shows that peers’ SOP failures immediately affect focal firm stock prices. The spillover effect takes place before firms reduce CEO compensation, and it offers a potential explanation for why focal firms adjust CEO compensation. We add to this literature by offering a new explanation for the spillover effect that complements Denis et al. (Reference Denis, Jochem and Rajamani2020). We differentiate between the sources of information (i.e., between learning from prices and learning from peers’ voting outcomes). We also offer direct evidence that directors benefit from the learning (i.e., higher investor support in future elections). Our study sheds light on the black box of directors’ decision-making processes.

With respect to control variables, CEO compensation is higher for larger firms (SIZE), those with better stock performance (RET), and CEOs with longer tenure (LN TENURE). Moreover, CEOs at retirement age (RETIRE) or who switch jobs between focal firms and SOP-failed peers (TALENTFLOW) receive lower compensation. The results also show a positive coefficient of lagged SOPVOTE and a negative coefficient of lagged ISSAGAINST, although not statistically significant. The signs of the coefficients on lagged SOPVOTE and lagged ISSAGAINST imply that focal firms with weaker shareholder support in prior SOP voting or that receive a negative recommendation for SOP proposals from the proxy advisor in the previous year have lower CEO pay in the next year.Footnote ¹⁰

E. Cross-Sectional Analyses

In this section, we investigate how the feedback effect varies with CEOs’ relative power on the board (proxied by CEO-chair duality, CEOs who are not newly appointed, directors hired after a CEO takes office, and busy boards), ISS’s voting recommendation, and compensation consultants. Table 4 presents the regression results for these cross-sectional tests as in Hypothesis 3.

Table 4Cross-Sectional Analyses

F. Placebo Tests

We conduct two placebo tests to address the concern that the market feedback effect on focal firms’ pay reduction could be simply statistical significance obtained by chance. First, the main findings could be related to news unrelated to the peer event in adjacent periods. In view of this, we artificially choose a set of nonevent dates (i.e., 10–50 days before and after the actual event dates) and compute the 3-day CAR surrounding these pseudo dates. Similarly, we define the dummy variable NEGCAR PSEUDO as 1 if the pseudo-CAR is below the annual sample median (i.e., a more negative CAR), and 0 otherwise. We then run the regression by replacing NEGCAR (based on the actual event dates) with NEGCAR PSEUDO (based on the pseudo-event dates). Panel A of Table 5 shows that none of the coefficients on the interaction term POST * NEGCAR PSEUDO is significant in any of the tests based on the pseudo-events.

Table 5Placebo Tests

Second, we identify 182 focal firms-year without SOP-failed compensation peers but with SOP-failed industry peers that are not compensation peers (excluding focal firm-year observations if any of its compensation peers has an SOP failure in the year). We use 1-digit SIC to define industries (similar results if we rely on a reduced sample using 2-digit or 3-digit SIC). There are 364 focal firm-year observations in the regression after retaining pre- and post-event observations around each identified focal firm-year. As reported in Panel B of Table 5, the insignificant coefficient on POST * NEGCAR PSEUDO suggests no evidence of pay reduction for focal firms following the industry peers’ voting failure. The result indicates that focal firms’ directors do not respond by cutting CEO pay for these pseudo-events related to industry peers.Footnote ¹³

Overall, the results from the placebo tests indicate that boards of focal firms do not respond by cutting CEO pay in the above pseudo-events, lending credibility to the evidence of the feedback effect documented in the main analyses.

G. Robustness Tests

We conduct a set of tests to ensure the robustness of the main results, including i) an alternative sample consisting of event firms and propensity-score-matched (PSM) nonevent firms; ii) another alternative sample excluding event firms whose shareholders’ meeting dates overlap with SOP-failed peers’ meeting dates; iii) how focal firms’ market reactions around peers’ SOP events affect excessive CEO compensation; iv) an alternative window of market reactions from 2 days before to 2 days after the peer event; v) the aggregate instead of the average CAR over multiple peer events during a year; vi) the residual CAR estimated from equation (1); vii) an alternative measure of CAR based on returns on failed SOP dates minus returns on successful SOP dates; viii) another alternative measure of CAR based on returns on failed SOP dates minus returns on other dates in a year; and ix) an alternative timeline for determining pre- and post-event years. The results of these robustness tests are reported in Table 6.

Table 6Robustness Tests