What Does “Say on Pay” Say about Audit Risk? Free

On July 21, 2010 the Dodd-Frank Wall Street Reform and Consumer Protection Act (U.S. House of Representatives 2010, hereafter Dodd-Frank) was signed into law. Under the “say-on-pay” provision (U.S. House of Representatives 2010, Sec. 953), shareholders have the opportunity to vote on the company's executive compensation plan. Thirty-seven of the firms in the Russell 3000 received a majority of no votes on their compensation plan in 2011 (the first year of voting). The number of firms receiving a negative majority vote has grown in the following years. In both 2012 and 2013, 57 firms in the Russell 3000 received a majority of no votes. Through mid-2014, 53 firms have had a majority of the shareholder vote go against their compensation plan.¹

Ertimur, Ferri, and Oesch (2013) show that the combination of high executive pay and poor performance is associated with the proportion of negative proxy advisor recommendations and negative votes that the compensation plan receives. The Dodd-Frank say-on-pay requirement thus makes transparent the growing pressure on top management to perform when pay is high. The growing number of majority negative votes since 2011 suggests that this pressure continues to increase since the say-on-pay requirement became law.

In this paper, we explore how the Dodd-Frank say-on-pay requirement impacts the audit environment. Our analysis reveals the differences in audit risk between the rejected (negative vote greater than 50 percent) and control companies in the five-year period before the vote. Rejected firms had more restatements, weaker internal controls, and higher audit fees than control firms. In addition, half of the restatements occurred after the say-on-pay vote. Thus, the vote itself might have raised questions about the quality of earnings and the risky audit environment. Increased audit fees in the year prior to the vote suggest that the auditors may have sensed an increase in risk. The shareholder negative voting results support that suspicion, and suggest that auditors should consider say-on-pay results as part of their formal risk assessments.

Based on the empirical results, we discuss how the audit environment could change in the Dodd-Frank era. Arguably, the pressure to push the financial reporting envelope may get even greater. Consider that even though shareholder say-on-pay votes are nonbinding, companies are required to report the results of say-on-pay votes via SEC Form 8-K filings. In years subsequent to say-on-pay votes, companies must also close the loop by disclosing in their proxy statements how the board responded to say-on-pay votes.

While the determination of executive compensation is not a function of the audit committee or the external auditor, an implication highlighted by the say-on-pay vote is that the pressure to justify CEO pay can influence internal controls and financial statement reporting. This is not a new lesson. In fact, it was one of the major reasons for strengthening of the audit function via the Sarbanes-Oxley Act. Prior empirical studies document a positive relation between the intensity of stock-based incentives and financial fraud (see, e.g., Cheng and Warfield 2005; Bergstresser and Philippon 2006; Burns and Kedia 2006; Johnson, Ryan, and Tian 2009). The bottom-line implication for auditors is that the combination of the high-pressure pay-performance incentives and poor controls leads to a high audit risk environment.

Dodd-Frank represents one of the most comprehensive financial regulatory reform measures in U.S. history. Passed in response to the 2008 financial crisis, the Dodd-Frank Act aims to promote U.S. financial stability and corporate transparency. A key theme of the Act is to better empower shareholders in the governance process by focusing on executive compensation and corporate governance practices at public companies. Specific provisions of the law include enhanced executive compensation disclosures, mandatory say-on-pay votes by shareholders, and mandatory followup on the part of management. Dodd-Frank, therefore, aims to improve communications between shareholders and the board of directors. Mandated compensation disclosures must provide clear and concise information to shareholders, while the say-on-pay vote provides shareholders a mechanism to convey their judgments to the board.

Dodd-Frank requires that companies include a resolution in their proxy statements asking shareholders, in a nonbinding vote, for approval of the compensation of their named executive officers. Also required is a separate nonbinding vote on whether the executive compensation vote should take place every one, two, or three years. The frequency of the executive compensation vote must be reaffirmed at least every six years.

Although shareholder say-on-pay votes are nonbinding, there is a closing of the loop requirement for management. First, companies are required to report the results of say-on-pay votes using Form 8-K filings with the SEC. Second, in years subsequent to say-on-pay votes, companies must close the loop by disclosing in their proxy statements how the board responded to say-on-pay votes. Proxy advisor Institutional Shareholder Services (ISS) has indicated that it would recommend that shareholders vote against all members of the board of directors (except new nominees that would be evaluated on a case-by-case basis) if the board adopted an executive compensation say-on-pay voting schedule that occurs less frequently than recommended by the majority of its shareholders (ISS 2012).

Dodd-Frank requires companies to compare executive compensation to the company's financial performance. Companies must disclose the median annual total compensation of all employees (except the CEO), the annual total compensation of the CEO, and the ratio of the median annual total compensation of all employees (excluding the CEO) to the CEO's total annual compensation (U. S. House of Representatives 2010, Sec. 953). While Dodd-Frank does not dictate executive compensation per se, proxy advisors such as ISS have emphasized that executive pay should be related to company performance. Logic thus suggests that the combination of high executive pay and poor company performance would increase the chances for a negative proxy advisor recommendation and a subsequent negative shareholder vote.

While Dodd-Frank formalized the requirement for a say-on-pay vote, the pressure for shareholder input on executive pay has been growing over time. In the five-year period before Dodd-Frank became law, a number of significant events increased the pressure to justify executive pay. In 2006, for example, the Securities and Exchange Commission increased disclosure requirements for executive compensation. In 2007, the U.S. House passed the Shareholder Vote on Executive Compensation Act (U.S. House of Representatives 2007).² During this timeframe, there were also proxy votes at several large firms (e.g., Merck) that involved giving shareholders a nonbinding advisory role in executive pay. ISS reported nearly 40 percent average support for such proposals at six annual meetings in 2007. Large institutional shareholders (e.g., California State Teachers' Retirement System [CalSTRS]) were also becoming much more vocal about executive pay issues during this timeframe.³

The passage of Dodd-Frank is thus the culmination of pay-for-performance pressure that has been growing for a number of years. This paper uses the experiment provided by the first round of say-on-pay voting to examine audit risk in a sample of rejected and control firms. Since the vote will be required going forward, we also discuss how formal votes and recommendations from proxy advisors (such as ISS) will impact audit risk in the future.

We begin by identifying the 36 companies that received a majority of “no” votes during the first round of say-on-pay voting in January through June 2011.⁴ We obtain the list of sample firms from CompensationStandards.com (2011). The percentage of no votes received by rejected companies ranged from a high of 68 percent at Helix Energy Solutions Group to a slight negative majority of 50.9 percent at Tutor Perini Corporation. Our study is based on the analysis of these 36 companies and a matched control sample that did not receive a negative majority of say-on-pay votes.⁵ We select the control group as follows: First, we retrieve all firms in Compustat's database with the same two-digit Standard Industrial Classification (SIC) as the rejected company. We then partition all companies in each two-digit SIC code, excluding the rejected companies, into quartiles based on net sales revenue. We choose the match firm by selecting the median company from the same net sales revenue quartile and two-digit SIC as the rejected company. We repeat this procedure each year to ensure a relevant control group across time.

For the 72 sample companies (36 rejected companies and their matches), we collect company specific financial and market performance data each year from 2006 through 2010.⁶ We use Standard & Poor's Compustat database as our source for financial performance and executive compensation data. We rely on company proxy statements to hand collect missing executive compensation data. We gather financial restatement data, data on the internal control environment, and audit fees from Audit Analytics. We gather stock return data from the Center for Research on Security Prices (CRSP). Consistent with the approach used by ISS, we compute total shareholder returns (TSR) for each sample company (change in stock price for the year, plus cash dividends, divided by the stock price at the beginning of the year). The comparison of TSRs between the rejected and control companies does not formally control for risk to the degree a multifactor pricing model would, but it serves to illustrate the potential impact of metrics used by influential advisors such as ISS.⁷

Table 1 provides the description and data sources for our financial statement, market performance, and executive compensation variables.

Table 2 provides statistical comparisons of the rejected companies and the control group. The results show that rejected companies generally had poorer financial performance (ROA) and stock performance (TSR) in the five-year period (2006–2010) prior to the vote, although the five-year averages are not statistically different. In the year prior to the vote, 2010, the accounting and stock performance differences are statistically different. Mean ROA for rejected companies was 1.67 percent versus 4.47 percent for matched companies (p = 0.040). Similarly, rejected companies had a mean TSR of −1.29 percent, while the control sample had a positive TSR of almost 17 percent (p = 0.015).⁸

Table 2 also provides information on executive compensation from 2006–2010. Total and equity-based average CEO compensation is significantly higher for the rejected companies than for the control companies over the 2006–2010 period (p-values = 0.026 and 0.021, respectively). For the year 2010, CEO total compensation for the rejected companies averages about $9.7M compared to $4.6M for the control group (p-value = 0.001). Average equity compensation for CEOs of rejected companies was $5.7 million in 2010, more than double that of the control group at only $2.2 million (p-value = 0.003).

Thirty-one of the 36 companies in our study that received a majority of negative say-on-pay votes had a negative recommendation from ISS (Noked 2011). Ertimur et al. (2013) show that proxy advisor recommendations influence say-on-pay votes.⁹ However, Ertimur et al. (2013) also find that proxy advisor recommendations are far from uniform. Just over 11 percent of the 1,275 firms they sample from the S&P 1500 receive a negative ISS recommendation, while nearly 22 percent receive a negative recommendation from the proxy advisor, Glass Lewis & Co. Conditional on having one negative recommendation, the two proxy advisors agree only 17.9 percent of the time. So while proxy advisors' recommendations influence voting, they also appear to be based on differing methodologies and assessments (Ertimur et al. 2013).

We now turn to examining the audit environment in the five-year period prior to the vote. Table 3 shows a comparison of the number and magnitude of restatements for both the rejected and control firms. We include all restatements that involve changing financial results during the period from 2006 through 2010. Eight of the 36 rejected firms have at least one restatement (two of the eight have multiple restatements). In total, the rejected firms have ten income-reducing restatements that result in a cumulative reduction in net income of over $1B. In contrast, one of the control firms has a restatement that reduces net income by about $5M. The inference from these results is that rejected firms, on average, had poorer financial statement quality and a bias toward overstating income.

In addition, half of the restatements occurred after the say-on-pay vote in 2011. Thus, the vote itself might have raised questions about the quality of earnings and the risky audit environment.

Table 4 shows that five of the rejected firms had material internal control weaknesses that span a total of nine years during the 2006–2010 period.¹⁰ In contrast, two control firms had material internal control weaknesses for one year each, a total of two years during the same period. Perhaps even more revealing is the relative nature and severity of the internal control weaknesses shown in Table 4. Eighty percent of the rejected firms that reported internal control weaknesses had issues regarding the adequacy of accounting resources and accounting personnel competency. Some of the more glaring weaknesses included concerns about senior management competency, ethical issues with personnel, nonroutine transaction controls, and multiple year-end adjustments. Consistent with the restatement results in Table 3, the internal control findings illustrate that rejected firms had weaker control over financial reporting.

Table 5 provides a comparison of the audit fees for the rejected and control companies. The audit fees for the rejected companies generally increase over the 2006–2010 period, while the audit fees for the control group decrease in both 2009 and 2010. Although not statistically different, beginning in 2008 and continuing through 2010, the rejected firm sample has higher mean audit fees than their matched control group (Panel B, Table 5). Interestingly, in 2010 (the year before the vote), average audit fees for the rejected firms reached their highest point over the five-year period, $2.6 million, while the average audit fees for the control group fell to their lowest point, $1.8 million. In 2010, the difference is marginally statistically different (p = 0.102).

Table 6 examines the percentage change in audit fees for the rejected and control companies. In a matched comparison with the control group (Panel B, Table 6), rejected companies have a larger percentage change in audit fees than their matched control group in each one-year period ranging from 2006 through 2010. While the percentage change comparisons for the two groups are only statistically different in the 2009 to 2010 period, the p-value monotonically declines over the entire period. In 2010, the average percentage change for rejected companies is over 16 percent higher than the matched control group and the difference is statistically different (p = 0.035). Similar to Stanley (2011), our findings are consistent with the hypothesis that auditors raised their fees due to their perception of higher business and/or audit risk.

We examine the audit environment of 36 firms that received a majority of negative votes on their executive compensation in the first round of Dodd-Frank voting in 2011. Compared to a matched control group, the rejected firms had generally poorer financial and market performance but higher executive compensation in the five years before the vote (2006–2010). In 2010, the poor performance and high pay results become even greater and statistically different.

Rejected firms also had relatively poor financial statement quality in the period (2006–2010) before the vote. We find that seven of the 36 rejected firms had a total of ten income-reducing financial restatements during that period, compared to only one restatement for the control group. In addition, rejected firms had much weaker internal control environments (as measured by the total number of years of material internal control weaknesses and the severity of the nature of those weaknesses), as well as higher audit fees in the period prior to the vote. In 2010, the change in audit fees for rejected firms is significantly higher than the change in the control group audit fees.

The findings suggest that auditors should use a negative vote as an input to their risk assessments. Half of the restatements occur after the say-on-pay vote. Rejected firms also tend to have weak internal controls in high-risk areas. The increase in audit fees in the year prior to the vote is an indication that auditors may have sensed increased risk; but, shareholder voting results support a more formal use of the vote by auditors as part of their assessment of the audit risk environment.

Our findings should be interpreted and generalized with caution as they are based on a small sample with a majority of negative say-on-pay votes. But our empirical evidence does suggest that Dodd-Frank mandated say-on-pay voting does say something meaningful about audit risk. Our findings thus provide some insight on an interesting consequence of the say-on-pay rule that we hope motivates additional research on the topic.