Compensation Committee Meeting and Management Earnings Guidance

Xiumin Martin Olin School of Business Washington University [email protected]

Hojun Seo Olin School of Business Washington University [email protected]

Jun Yang Kelley School of Business Indiana University [email protected]

September 2015

We thank Richard Frankel for helpful comments. We also thank workshop participants at Washington University in St. Louis. We are grateful for financial support from the Olin Business School and the Kelley School of Business.

Compensation Committee Meeting and Management Earnings Guidance

Abstract: Corporate boards determine performance metric for CEOs’ annual incentive plans at compensation committee meetings at the beginning of a fiscal year. We study whether management issues downward-biased earnings guidance right before these meetings with the objective of influencing these performance metric. We find that management earnings guidance issued immediately before the meetings tends to be lower than the prevailing consensus analyst forecasts relative to management earnings guidance issued outside this window. This downward bias is only present when the performance metric is linked to earnings-per-share (EPS). We do not observe the downward bias when revenue serves as the performance metric. In addition, the downward-biased management earnings guidance prior to compensation committee meetings is more pronounced when the prevailing consensus analyst forecast is much higher than the firm’s target EPS, projected target EPS, or the actual EPS in the current period. We also observe stronger downward-biased management earnings guidance prior to compensation committee meetings when institutional ownership is more concentrated or when the CEO serves as a chairman of the board. Last, we find that analysts downwardly revise their earnings forecasts after observing management earnings guidance, both of which are issued before the compensation committee meetings. Taken together, our findings suggest that managers have incentives to issue pessimistic earnings guidance before compensation committee meetings and that analyst earnings forecasts might serve as an anchor for the compensation committee to defend its choice of the performance metric under shareholder pressures. Keywords: Annual Incentive Plan (bonus contracts), Performance Targets, Management Earnings Guidance JEL Codes: G34, M41, M52

I. Introduction Annual incentive plans (bonus plans) have been widely used in incentivizing corporate executives to improve short-term performance in the U.S. (Murphy 1999).

Bonus payout

accounts for approximately twenty percent of total annual pay for CEOs of the S&P 500 companies in recent years (Kim and Yang 2012). Kim and Yang document that at the time of bonus plan approval, performance targets used for determining the bonus payout amount (e.g., EPS target) are set lower than the prevailing consensus analyst forecasts, making it easier for firms to achieve the performance goals ex post. However, little is understood with respect to what is based on for corporate boards to determine performance goals. Even much less is known about the ramification of this process. Starting in 2006, SEC enhanced executive compensation disclosure requirements for public companies, which is intended to provide investors with a clearer and more complete picture of the compensation earned by a company’s named executive officers. In particular, the rules require disclosures about awards granted to each named executive officer. The awards are under an incentive plan or otherwise contingent on the achievement of performance goals, including estimated future (minimum, target, and maximum) payouts for (threshold, target and stretch) performance for both equity incentive plans and annual incentive plans.1 We argue that the enhanced disclosure rules by the SEC allow shareholders to scrutinize executive compensation structures more rigorously, which might exert pressures on corporate boards to defend their choice of performance goals specified in annual incentive plans. In particular, we posit that corporate boards rely on external benchmarks to determine the performance goals in

1

More details on the enhanced disclosure requirements are available at https://www.sec.gov/rules/final/2006/338732a.pdf

1

annual incentive plans in response to enhanced shareholder scrutiny. This in turn gives rise to incentives for corporate managers to guide the external benchmarks downward in order to achieve lower performance goals ex post. When earnings-per-share (EPS) serves as the performance measure in annual incentive plans, the natural external benchmark in consideration is the prevailing consensus analyst earnings forecasts. Consistent with this expectation, Kim and Yang (2012) find that performance targets based on the EPS measure in annual incentive plans (i.e., EPS targets) are similar to but lower than the prevailing consensus analyst forecasts. In addition, Kim and Yang document that the EPS target in annual incentive plans are indistinguishable from corporate-issued earnings guidance on the approval date of the plan.2 If indeed corporate boards rely on the prevailing consensus forecasts to determine the performance target, we expect managers to issue more pessimistic earnings guidance in the period immediately before compensation committee meetings at which performance goals are set, compared to management earnings guidance issued in other periods. The benefits from providing pessimistic earnings guidance are that they downwardly guide security analysts’ earnings forecasts which could be used as the benchmark for corporate boards to set the performance goal. Using 4,515 firm-guidance-year observations for the sample period of 2006 to 2012, we conduct various tests to examine whether managers release more pessimistic earnings guidance in the 90 days before the approval date of annual incentive plans by compensation committees (event period, hereafter). Following prior research, we measure management forecast bias as the difference between management earnings forecast and the prevailing consensus analyst forecasts (Ajinkya and Gift 1984; Baginski et al. 1993; Kothari et al. 2009; Rogers and Van Buskirk 2013). Consistent with our expectation, we find that management earnings guidance in the event period 2

As discussed in the next section,

2

is more pessimistic, particularly when the performance measure in annual incentive plans contains EPS but not revenue measure. In contrast, when the performance measure in annual incentive plans contains revenue but not EPS measure, we find no systematic bias in management earnings guidance issued in the event period relative to those issued in other periods. In addition, we find that managers issue more pessimistic earnings guidance in the event period when the prevailing consensus analyst forecasts during the event period is more optimistic. Next, we conduct cross-sectional tests to examine whether management issues more pessimistic earnings guidance to affect their performance targets when shareholder activism is more intensive or when the CEO is more powerful. We find that managers are more likely to issue pessimistic earnings guidance in the event period when institutional ownership is more concentrated or when the CEO of the firm also serves as the chairman of the board. These results suggest that corporate boards tend to rely on external benchmarks such as the prevailing consensus analyst forecasts to set the performance targets in annual incentive plans in the presence of shareholder activism and/or when CEOs are influential on board decisions. To check our maintained assumption that managers issue pessimistic earnings guidance in order to affect the prevailing consensus analyst forecasts, we examine revisions of the consensus analyst forecasts in response to the management earnings guidance issued in the event period. We find that analysts indeed downwardly revise their forecasts more (revisions occur before the approval date of annual incentive plans) following the issuance of more pessimistic management earnings guidance. We also find that the downward analyst forecast revisions are greater when the prevailing consensus analyst forecasts are more optimistic in the event period. These results support our maintained assumption that analysts react to pessimistic management earnings guidance issued in the event period.

3

Though we argue that under shareholder pressures corporate boards are inclined to rely on the consensus analyst forecasts to determine performance goals, it is plausible that compensation committees also use management earnings guidance as a direct input. We find that the prevailing consensus analyst forecasts are incrementally positively associated with the EPS target over and above management earnings guidance but not vice versa. Furthermore, the effect of management earnings guidance on the EPS target becomes indistinguishable from zero after controlling for the prevailing consensus analyst forecasts in the EPS target regression model. From these results we conclude that compensation committees likely rely on the prevailing consensus forecasts to determine performance goals although we can’t completely rule out the possibility that compensation committees might also use management earnings guidance as a direct input. Furthermore, we demonstrate that more pessimistic management earnings guidance issued in the event period is associated with lower concurrent EPS target relative to the EPS target and the actual EPS performance in the prior period. This evidence supports our maintained assumption that management issues more pessimistic earnings guidance during the event period to low ball the performance target. Finally, we conduct a falsification test to validate our argument that managers are more likely to issue pessimistic earnings guidance after 2006 because of the enhanced disclosure requirements on executive compensation. We specifically assign pseudo bonus plan approval dates to the period before 2006 using typical bonus plan approval dates and performance target metrics in annual incentive plans after 2006. We then identify management earnings guidance in the “event period” similarly to our main analysis. Our results do not hold in this falsification test, which supports our argument that the findings from this study are mainly driven by the enhanced

4

executive compensation disclosure requirements, which facilitate shareholder activists to be more involved in the process of determining executive compensation. Our paper makes three key contributions to the literature. First, our study provides evidence that corporate boards rely on external benchmarks to determine the performance goals in CEO annual incentive plans. Second, our study shows that managers behave opportunistically via issuance of management earnings guidance to influence the determination process of performance goals. In this sense our paper is related with Cheng and Lo (2005), Brockman, Khurana, and Martin (2008), Lennox and Ge (2011), and Dimitrov and Jain (2013).These studies show that management earnings guidance is used opportunistically to facilitate insider trading, corporate repurchases, demonstration of good performance, and M&A transactions. Third, our findings suggest that the pressure for corporate boards to rely on external benchmarks comes from shareholders. The cooperation between corporate boards and managers in the process of setting performance goals likely reflects agency costs. Ironically, our findings imply that shareholder activism might inadvertently promote managerial opportunism. The rest of the study is organized as follows. In the next section, we discuss relevant literature and develop our hypothesis. In Section 3, we describe the empirical design. Sample construction is discussed in Section 4. We present the empirical results in Section 5 and conclude the study in Section 6.

II. Background and Hypothesis Development Background in CEO annual incentive plan Annual incentive plans for top executives are designed to improve firms’ short-term performance. “Virtually every for-profit company offers a bonus plan covering its top executives

5

and paid annually based on a single-year’s performance” (Murphy 1999). A typical annual incentive plan contains performance measures, the corresponding performance goals, and the structure of the pay-for-performance relation. A typical bonus plan exhibit nonlinear relation between pay and performance because no bonus is paid until a threshold performance (usually expressed as a percentage of the performance target) is achieved, and a “minimum bonus” (usually expressed as a percentage of the target bonus) is paid at the threshold performance. Target bonuses are paid for achieving the performance target 3 . A cap is typically placed on bonuses paid and the cap is commonly expressed as a percentage of the performance target). Companies use a variety of financial and non-financial performance measures. However, almost all companies rely on some measure of accounting profits. The commonly used accounting measures consist of EPS, revenues, net income, pre-tax income, operating profits (EBIT), or economic value added. Most performance targets for accounting-profit performance measures are based on a single criterion. These criteria include budgets, prior-year performance, board discretion, peer-group comparisons, timeless standards, cost of capital, and a combination of these criteria. Murphy (1999) argues that internal performance metrics are problematic if executives can participate in setting performance standards. Similarly, Anderson, Dekker, and Sedatole (2008) suggest that the benefits of pay-for-performance will be attenuated if managers are given the opportunity to influence performance goals. Interestingly, Kim and Yang (2012) show that firms tend to set performance targets of CEO annual incentive plans below the prevailing consensus analyst forecasts to make these performance goals more achievable. This paper provides the mechanism through which downward-biased earnings guidance issued by the management helps lower the consensus analyst forecasts prior to the approval date of annual incentive plans, which in turn lowers the performance targets in CEO annual incentive plans. 3

We use performance target and target performance goal interchangeably in this paper.

6

The nonlinearity of pay-performance relationship in bonus plans motivates management to distort firms’ financial reporting. Prior literature shows that firms tend to manipulate earnings upward when performance falls just short of the threshold performance goal, and downward when performance substantially exceeds the stretch performance goal at which no additional payout is awarded (Healy 1985; Gaver, Gaver, and Austin 1995; Holthausen, Larcker, and Sloan1995). Using distributional analyses, Kim, Martin, and Yang (2015) show that firms tend to just beat the target performance goals of CEO annual incentive plans. Literature in management forecasts Firms provide voluntary disclosures for various reasons. Evidence shows that management uses voluntary disclosures to reduce information asymmetry between firms and investors (Ajinkya and Gift 1984; Ajinkya, Bhojraj, and Sengupta 2005; Karamanou and Vafeas 2005; Lennox and Park 2006; Hui, Matsunaga, and Morse 2009), manage earnings expectations (Mastsumoto 2002; Cotter, Tuna, and Wysocki 2006), reduce litigation costs (Skinner 1994; Kasznik and Lev 1995; Skinner 1997; Soffer, Thiagarajan, and Walther 2000; Field, Lowry, and Shu 2005, Rogers and Van Buskirk 2009), or affect stock-based compensation (Noe 1999; Aboody and Kasznik 2000; Cheng and Lo 2006; Brockman et al. 2008). In this study, we focus on the role of management earnings guidance in affecting managers’ bonus compensation. In particular, we rely on the findings from prior research that managers have the incentive and ability to guide securities analysts. For example, Richardson, Teoh, and Wysocki (2004) find that analyst forecasts are more optimistic for firms whose managements wish to sell shares on the firms’ behalf or they personally own. Cotter et al. (2006) demonstrate that firms may employ management guidance to mitigate over-optimism in analysts’ forecasts. Feng and McVay (2010)

7

document that analysts, wishing to curry favor with management, weight management earnings guidance more heavily than predicted when revising their short-term earnings forecasts. Hypothesis development As discussed previously, performance standards in executive annual incentive plans are commonly determined at the beginning of a fiscal year at a compensation committee meeting. To set the performance targets, compensation committees may rely on some external benchmarks that are readily available, relevant, and unbiased such as earnings forecasts issued by security analysts. The reason that compensation committees rely on these benchmarks might have two folds. One is to obtain some reference points. The other is that compensation committees might find it in its own interest to refer to external benchmarks if doing so makes it easier to defend its choice of performance targets in front of shareholders. This is because there have been various changes in compensation-related shareholder activism since the 1990s. For example, union pension funds have become the dominant proponent replacing individual shareholders after 2002. They introduced new types of proposals calling for enhanced shareholder voting rights on CEO pay, more transparent reporting, and a tighter link between pay and performance. As a result, the frequency of and the voting support for compensation-related shareholder proposals have increased significantly (Gillan and Starks 2007; Ertimur, Ferri, and Muslu 2010). Moreover, shareholders’ Just-Vote-No campaigns meant to obtain changes in executive pay practice became more frequent. In some high-profile cases (e.g., Home Depot, Pfizer), the campaigns might contribute to the ouster of the CEO and board members. Regardless of what reason it might be, we expect that managers tend to issue pessimistic earnings forecasts prior to compensation committee meetings to lower performance targets if they believe that compensation committees rely on analyst forecasts to set the performance target and that

8

management earnings guidance has a direct impact on analyst forecasts. Based on this line of reasoning, our main hypothesis is stated as follows: Hypothesis: Management earnings guidance is more pessimistic during the period prior to compensation committee meetings than those issued in other periods.

III. Research Design In our hypothesis, we predict that more pessimistic management earnings guidance is issued right before the approval date of annual incentive plans. To examine this hypothesis, we follow prior research (Brockman et al. 2008) and use the following regression model. MF Biast =

α + β1 Event EPS targett + β2 Sizet-1 + β3 Market-to-Bookt-1 + β4 ABRETt-1

(1)

+ β5 ∆ROAt-1 + β6 EarnVolt-1 + β7 Losst-1 + β8 Litigation Riskt + β9 Horizont + εt

The dependent variable, MF Biast, is defined as the management EPS forecast less the consensus analyst forecast three days before the date of management forecast divided by stock price three days before the date of management forecast. We multiply this variable by 100 to ease the interpretation of economic magnitude. We use all quantitative management annual earnings forecasts. If the management forecast is a range forecast, we choose the mid-point of the forecast (e.g., Rogers and Van Buskirk 2013; Bergman and Roychowdhury 2007). If the management forecast is a one-sided range forecast (e.g., it specifies a maximum or a minimum value), we calculate management forecast bias only if the mean consensus is above (below) the minimum (maximum) value; otherwise, the management forecast is assumed to contain no bias (Bergman and Roychowdhury 2007). We calculate the daily EPS consensus analyst forecasts using the IBES unadjusted detail file. We specifically calculate the daily consensus based on individual analyst forecasts, which are required to be reported within the 90-day window 9

immediately preceding that specific date to ensure that our daily consensus is not based on stale analyst forecasts. We exclude individual analyst forecasts if IBES excludes the forecasts from calculating IBES-reported EPS consensus. In addition, following Rogers and Van Buskirk (2013), we adjust daily consensus when we calculate management forecast bias if a management earnings forecast is issued concurrently with the earnings announcement (i.e., bundled management earnings forecasts).4 Our main independent variable, Event EPS targett, is an indicator variable equal to one for the latest management earning forecast issued during the 90-day window (event window) before the approval date of annual incentive plans for firms with an EPS performance target but without a revenue performance target in bonus plans in period t, zero otherwise.5 The purpose of identifying firms disclosing the EPS or revenue performance targets has two folds [or four folds??]. First, we are certain about the specific performance measures that firms use. Second, we know the exact target value. Both facilitate us to design a placebo test that aims to isolate the compensation motivated disclosure strategy that only applies to EPS being the performance measure. Furthermore, the exact performance target value allows us to conduct a cross-sectional analysis by identifying when management has the strongest incentive to use disclosure strategy.

4

Rogers and Van Buskirk (2013) argue that when bundled management forecasts are compared to the prevailing analyst expectations, the prevailing analyst expectations need to be adjusted assuming that analysts update earnings forecasts for subsequent periods using realized earnings at earnings announcement. By using the analyst forecast revision model in Rogers and Van Buskirk (2013) we specifically calculate management earnings bias for bundled forecasts by comparing prevailing analyst expectations with management earnings forecasts conditional on the current period’s earnings surprise 5 We identify the event management forecasts by using the approval date of bonus plans. If the approval date of bonus plans is missing (i.e., even after we fill the date by using the grant date of bonus plans), we calculate the typical approval date of annual incentive plan for each firm during our sample period and identify the event management earnings forecast. If the firm has disclosed performance targets but we still are not able to identify the event management forecast for that period, we assume the first management forecast during the first fiscal quarter as the event management forecast because the compensation committee meetings determining annual incentive plans are typically held during the first fiscal quarter. We find that the mean difference between the current period end and the approval date of bonus plan is 317 days and the mean difference between the current period end and the event management forecast is 342 days confirming this assumption. We check that all results are qualitatively similar without this assumption.

10

Last, the exact performance target also allows us to examine if management disclosure strategy is successful ex post by comparing the target EPS goal with the prior year target EPS goal or prior year actual EPS, both of which serve as a benchmark representing what the target EPS goal would be without the influence of management disclosure. We choose 90 days as the event window because management commonly issues earnings forecasts on a quarterly basis. If multiple identical management forecasts are issued during the event period, we select the first management forecast because subsequent management forecasts in the string are more likely to convey less or no information, and consequently will have little impact on analysts’ decision to revise forecasts. We also include a comprehensive set of control variables. We include firm size to control the overall information environment of a firm. The Sizet-1 variable is defined as the natural logarithm of the market value of equity as of the fiscal quarter preceding the date of management forecast. We use the market-to-book ratio to control growth options and proprietary costs (Bamber and Cheon 1998). The Market-to-Bookt-1 variable is measured as the market-to-book ratio [of equity?] as of the fiscal quarter preceding the date of management forecasts. The ABRETt-1 variable is measured as cumulative abnormal returns measured as the excess firm returns over the CRSP value-weighted returns during the three months ending two days before the issuance of a management forecast. The ABRETt-1 variable controls the effect of momentum and performance on earnings forecasts (Brockman et al. 2008). We also include return on assets to further control the effect of performance on management forecast news. The ∆ROAt-1 variable is measured as the change in annual return on assets as of the fiscal year preceding the date of management forecasts. We include two variables to control the difficulty in predicting earnings. The Losst-1 variable is an indicator equal to one if the firm reported losses in the previous period

11

[quarter or year??], zero otherwise. Earnings forecasts are more likely to be difficult for analysts and managers for loss firms, affecting management forecast news. The EarnVolt-1 variable is measured as standard deviation of quarterly earnings scaled by lagged total assets over 12 quarters preceding the date of management forecasts. If earnings are volatile, then managers and analysts are less likely to provide precise forecasts. We include a proxy for litigation risk, Litigation Riskt, because prior research suggests that litigation risk is an important motive for managers to voluntarily disclosure forward-looking information (e.g., Francis et al. 1994). The Litigation Riskt variable is an indicator equal to one for firms in one of the biotechnology (SIC 2833–2836 and 8731–8734), computers (3570–3577 and 7370–7374), electronics (3600–3674), and retail (5200–5961) industries, and zero otherwise. Lastly, we include forecast horizon as a control variable because prior research documents that forecast biases depend on forecast horizon (Rogers and Stocken 2005; Bergman and Roychowdhury 2007). The Horizont variable is measured as the difference between the date of a management forecast and the date of fiscal period end divided by 365. We use ordinary least square (OLS) regression model with standard errors clustered at the firm level. We include firm and year fixed effects to control for firm-invariant and timeinvariant factors that might affect management forecast bias. We also include fiscal quarter indicators to account for different management incentives to issue biased forecasts in different fiscal quarters (e.g., Matsumoto, 2002).

IV. Data and Descriptive Statistics Plan details are disclosed in the firm’s proxy filing starting in 2006, with improved coverage over the next six years of our sample period. According to Kim and Yang (2012), the 12

average and median number of performance measures used in the annual incentive plans is about three, of which two are quantitative measures. The top two measures for firms with multiple performance measures are EPS and revenue. For each firm-year with disclosed information, we identify whether the firm uses EPS (or EPS growth) or revenue as performance measures in the annual incentive plan. We further collect information on the corresponding performance targets if available. If there are multiple annual incentive plans at a firm, we use the one for the CEO. According to the tax rule of 162(m), to be tax deductible, the formulas for performancebased pay in the annual and long-term incentive plans have to be approved in the first 90 days of a fiscal year. Committee approval dates are collected from the Grants of Plan-Based Awards table in a firm’s proxy filings or from 8-K filings. Some companies have the same plan approval date every year. This helps us extrapolate the approval dates of some of the annual incentive plans prior to 2006.7 In Appendix B, we provide several examples of proxy filings in which we collect the approval date of annual incentive plans. Information below is extracted from the annual incentive plan of Colgate-Palmolive in 2007: 8 “Bonus payouts for a particular year are determined … [omitted by authors] by a formula based on the level of growth achieved the prior year in Base Business Earnings-Per-Share or the applicable division’s net sales and net profit after tax. The P&O Committee has discretion to adjust the calculated awards downward, but not upward. For 2007, in order for Named Officers

7

For some other companies, we fill in the approval dates of the annual incentive plan using the approval or grant date of the long-term incentive plan if the approval occurred in the first quarter and these incentive plans tend to be approved on the same day historically. We verify these assumptions by asking a compensation consultant and an expert at ISS Corporate Services (Risk Metrics) and both confirm that our assumptions are reasonable. 8 Some companies state that they do not disclose performance targets because the information is confidential in a competitive environment. For example, Affiliated Computer Services states the following on page 26 of its fiscal 2007 proxy statement. “We have not disclosed target levels with respect to specific quantitative or qualitative performance-related factors considered by the Compensation Committee because disclosure of the specific performance goals would give our competitors information that could be leveraged for competitive advantage which would result in competitive harm to the Company.” Other companies simply do not disclose any information on performance targets.

13

with corporate-wide responsibilities to earn bonuses at the top end of their range, Base Business Earnings-Per-Share had to grow by 11.0% above the 2006 Base Business Earnings-Per-Share.” We merge the approval dates data with COMPUSTAT for financial data, IBES Guidance database for management earnings forecasts, IBES database for analyst earnings forecasts and IBES-reported actual EPS data, and CRSP for stock return data. Our final sample consists of 1,021 firm-year observations with available EPS or revenue performance targets in annual incentive plans, and the sample period ranges from 2006 to 2012. For 1,021 firm-year observations, we identify 4,515 management annual earnings forecasts with available financial data. Following prior studies (e.g., Rogers and Stocken 2005; Boone and White 2014), we drop pre-earnings announcement forecasts (i.e., earnings forecasts issued after fiscal period but before the earnings announcement date) because pre-earnings announcement forecasts are commonly issued to avoid negative earnings surprises and deter litigation and thus they are a part of earnings announcement strategies (e.g., Kasznik and Lev 1995; Soffer et al. 2000). The number of observations in any particular test varies depending on the availability of data necessary for each test. All continuous variables are winsorized at 1st and 99th percentiles to mitigate the influence of extreme observations. Table 1 presents descriptive statistics. Approximately 6.0% (2.6%) of management earnings forecasts during our sample period are the latest management earnings forecasts issued within the 90-day window before the approval date of executive annual incentive plans for firms with an EPS (revenue) performance target but without revenue (EPS) performance target in the plan. 4.4% of management earnings forecasts are the latest management earnings forecasts issued within the 90-day event period for firms with both EPS and revenue performance targets in the annual incentive plans.

14

Figure 1 presents the timeline of events. In our sample, the mean difference between the current period end for which management issues earnings forecasts and the approval date of bonus plan is 317 days (standard deviation of 15.86 days) and the mean difference between the current period end and the date of latest management forecasts issued during the event period is 342 days (standard deviation of 26.8 days), suggesting that the compensation committee meetings are typically held in the first fiscal quarter of each period and the latest management earnings forecasts in the event period are issued on average 25 days before the bonus plan approval date. Table 2 reports correlations. We first note that the MF Biast variable is significantly negatively correlated with the Event EPS targett variable (-0.05), but not significantly correlated with the Event EPS REV targett or the Event REV targett variable, providing preliminary evidence that management EPS forecasts during the event window are generally pessimistic for firms where the EPS performance is the only measure used in their CEO annual incentive plan; however, management EPS forecasts do not demonstrate pessimism when revenue performance is the metric considered in the incentive plan, regardless of whether EPS performance is jointly used or not. We also note that the Event EPS targett variable is significantly negatively correlated with analyst forecast revisions, AF Revisiont, (-0.08), providing initial evidence consistent with our argument that managers issue pessimistic earnings guidance in the event window to walk down the consensus analyst forecast. We note that the AF Revisiont variable is positively correlated with management forecast bias (0.22), implying that analysts revise forecasts in the direction consistent with management bias.

V. Empirical results

15

Main analysis Table 3 reports results of testing our main Hypothesis. In column (1), our main variable of interest, Event EPS targett, is loaded negatively and statistically significant at the one percent level. This evidence is consistent with our expectation that when managers’ bonus is directly linked to the firm’s EPS, management earnings guidance issued in the event period is more pessimistic relative to those issued in other periods. From economic perspective, management earnings guidance issued before the approval date is .204% lower proportional to stock price than guidance issued in other periods, which represents 13.8% of the standard deviation of the MF Biast variable. Regarding control variables, we find that the Market-to-Bookt-1 variable is positively associated with management forecast bias, suggesting that firms with high growth options are more likely to provide optimistic management forecasts. The Litigation Riskt variable is negatively associated with management forecast optimism, indicating that management is more likely to be pessimistic in forecasting earnings when litigation risk is high (Francis et al. 1994; Skinner 1994). Next, we examine whether earnings guidance issued in the event period by managers whose bonus is linked to both EPS and revenue or linked to revenue only also demonstrates pessimistic tendency. If it is management intention to lower performance target metrics that leads to forecast pessimism, then we would expect to find weaker results for firms whose managers’ bonus is linked to both EPS and revenue because the additional linkage to revenue might dilute managers’ incentive to spin on the earnings forecasting wheel. In addition, we expect to find no results for firms whose managers’ bonus is only linked to revenue because management earnings guidance is unlikely to have any influence on revenue target. Hence, we introduce two additional variables in columns (2) and (3): the Event EPS REV targett variable in column (2) is an

16

indicator variable equal to one for the latest management earning forecast issued during the event period for firms that disclosed both EPS and revenue performance targets in annual incentive plans in period t, zero otherwise; the Event REV targett variable in column (3) is an indicator variable equal to one for the latest management earning forecast issued during the event period for firms that disclosed revenue but not EPS performance target in annual incentive plans in period t, zero otherwise. In column (2), the Event EPS REV targett variable is loaded negatively but is statistically insignificant, which is consistent with our expectation. When we move to column (3), we find the coefficient on the Event REV targett variable is positive and insignificant. In column (4), when we consider all three variables simultaneously, we find that the Event EPS REV targett variable becomes negative and significant at the ten percent level and that all other results remain robust. Overall, we find results consistent with our hypothesis that management issues more pessimistic earnings guidance right before the approval date of annual incentive plans relative to earnings guidance issued in other periods. The additional analyses also provide further confirmation that pessimistic forecasts issued in the event period appear to be driven by management desire to set lower performance target metric that is easier to achieve.

Cross-sectional analyses Analyst forecast optimism To provide additional supports to our argument, we conduct cross-sectional analyses in this section. First, we test whether our results vary with the optimism of consensus analyst forecast in the event period. If compensation committee refers to analyst forecasts in setting performance targets, which incentivizes managers to guide analyst earnings forecasts

17

downwardly to lower performance targets, then we expect management earnings forecasts issued in the event period to be more pessimistic when the prevailing consensus analyst forecasts are more optimistic. To examine this prediction, we use an indicator variable, the AF Biast variable, which measures the optimism of the prevailing consensus analyst forecasts in the event period and interact this variable with the Event EPS targett variable. We expect to find a significantly negative coefficient on this interaction term. Table 4 reports the results. In column (1), we define the AF Biast variable as an indicator variable equal to one if the prevailing consensus analyst forecast three days before the management forecast issuance date less the actual EPS in period t scaled by stock price is greater than the sample median, and zero otherwise. Consistent with our expectation, we find significantly negative coefficient on the interaction term, Event EPS targett × AF Biast, suggesting that management has greater incentives to issue pessimistic earnings guidance in the event period when the prevailing consensus analyst forecast is more optimistic relative to the actual EPS in period t. In column (2), we replace the actual EPS in period t with the projected EPS based on the actual EPS and EPS growth rate in period t-1 in defining the AF Biast variable. In column (3), we use the EPS target in annual incentive plans instead of the actual EPS in period t to define the AF Biast variable. In both columns, we continue to find a significantly negative coefficient on the interaction terms. Overall, our cross-sectional analysis yields further supports to our hypothesis that managers walk down analyst forecasts before the compensation committee meetings.

Shareholder pressures

18

To disentangle the two non-mutually exclusive reasons with respect to why compensation committees rely on analyst earnings forecasts in setting the performance target metric, we conduct further analysis to see if our results vary with shareholder pressure. Smith (1996) finds that the level of institutional holdings is positively associated with the probability of being targeted by shareholder activism. Gillan and Starks (2002) demonstrate union funds play an important role in shareholder activism, which submit 40% of the proposals in 2004 and 2005. If the presence of large institutional investors constitutes a threat of shareholder activism, we would expect corporate boards are more likely to rely on analyst forecasts to set the performance targets in executive bonus plans in these instances. Therefore, managers in firms with greater concentration of institutional ownership are more likely to guide consensus analyst forecasts downward. We use the INSTOWN TOP5t-1 variable to measure the concentration of institutional ownership. The INSTOWN TOP5t-1 variable is an indicator equal to one if the sum of top five institutional ownership for firm i at the end of fiscal quarter preceding the issuance date of management forecast exceeds the sample median, and zero otherwise. In column (1) of Table 5, we interact the INSTOWN TOP5t-1 variable with the Event EPS targett variable in equation (1). We find that the coefficient on Event EPS targett is negative but statistically insignificant while the coefficient on the interaction term, Event EPS targett × INSTOWN TOP5t-1, is a significantly negative, at the five percent level. This evidence is consistent with the argument that managers’ incentives to issue pessimistic earnings guidance in the event period mainly arise from corporate boards relying on analyst earnings forecasts for setting performance target when boards face greater pressures from shareholders. Economically, when a firm moves from the group with low institutional ownership concentration to the group with high institutional ownership

19

concentration, management earnings forecasts are .311% lower proportional to stock price during the event period, which represents 21% of the standard deviation of the MF Biast variable.

CEO power The result from column (1) of Table 5 suggests that shareholder pressures might motivate corporate boards to rely on external benchmarks for bonus contracts, which incentivizes managers to engage in issuing pessimistic earnings guidance. However, in the absence of opportunities (or in the presence of strong governance), managers might not be able to do so since opportunistic guidance allows manager to reap bigger bonuses. Thus, we examine whether earnings guidance is more pessimistic in the event window when firms’ internal corporate governance is weaker. In Table 5 column (2) we use the CEO duality variable to proxy for weak internal governance, which gives CEO the opportunity to engage in pessimistic earnings guidance before annual compensation committee meetings. More specifically, the CEO Dualityt variable is an indicator equal to one if the CEO is also the chairperson of the board, and zero otherwise (Core, Holthausen, and Larcker 1999). We interact this variable with the Event EPS targett variable. We find a negative coefficient on the Event EPS target variable. However, it is statistically indistinguishable from zero. The coefficient on the interaction term, Event EPS target × CEO Duality, is negative and statistically significant at the five percent level. From economic perspective, management EPS forecast is .25% lower proportional to stock price during the event period. This represents 17% of the standard deviation of the MF Biast variable. These results imply that managers’ issuance of pessimistic earnings guidance before compensation meeting reflects agency costs. Taken together, the findings from Table 5 suggest that shareholder

20

activism might not be effective and sometimes it might create perverse incentives for managers to extract rents.

VI. Discussion and additional analysis Analyst forecast revisions post management earnings guidance An imbedded assumption in developing our hypothesis is that managers can walk down analyst forecasts through earnings guidance. Though prior research provides ample supporting evidence, we nevertheless check the validity of this assumption in this section. To test this assumption, we use the following regression model. AF Revisiont =

α + β1 Event EPS targett + β2 AF Biast + β3 Event EPS targett × AF Biast

(2)

+ β4 Sizet-1 + β5 Market-to-Bookt-1 + β6 ABRETt-1 + β7 ∆ROAt-1 + β8 EarnVolt-1 + β9 Losst + β10 Litigation Riskt + β11 Horizont + εt

The dependent variable, AF Revisiont, is defined as the mean consensus analyst forecasts three days after the date of management forecast minus the mean consensus analyst forecasts three days before the issuance date of management forecast scaled by the absolute value of the latter. All other variables are defined previously. In this test, we also use the AF Biast variable introduced in Table 4 to examine whether analysts revise their forecasts down even more after observing more pessimistic earnings guidance issued by the management in the event period when the prevailing consensus analyst forecasts before management guidance is more optimistic. Table 6 reports the results. In column (1), we find that analysts revise their forecasts downward in response to management earnings guidance issued in the event period as evidenced by the significantly negative coefficient on the Event EPS targett variable at the one percent level. This result validates our assumption that earnings guidance issued in the event window walks 21

down analyst forecasts presumably because the event window guidance is more pessimistic. In column (2), we use the AF Biast variable measured based on the actual EPS, and interact it with the Event EPS targett variable. The interaction term, Event EPS targett × AF Biast, has a significantly negative coefficient at the one percent level, suggesting that analysts revise down their forecasts more in the event period when the prevailing consensus forecasts before the guidance are more optimistic. We find consistent results in column (3) and (4) when using the other two variables measuring analyst optimism based on either projected EPS in column (3) or EPS target in column (4). Overall, we find supporting evidence for our maintained assumption.

Do corporate boards use analyst forecasts as the benchmark? Another imbedded assumption underlying our hypothesis is that corporate boards refer to analyst forecasts to determine the EPS-based CEO performance target in annual incentive plans. However, it is also plausible that boards might obtain reference directly from management earnings forecasts, which will yield the same results as those documented in Table 3. However, arguably earnings forecasts issued by management might appear to be less objective compared with that provided by security analysts. This might undermine boards’ ability to defend the performance metric used in annual incentive plans in front of shareholder activists. To distinguish which benchmark is likely to be relied on by boards, we regress the EPS target figure in the annual incentive plan on earnings forecasted by the manager and earnings forecasted by security analysts separately. Management earnings forecast is measured immediately before compensation committee meetings and the consensus analyst forecast is measured three-days after the date of management earnings forecasts but before compensation committee meetings. We employ the following regression model using firm-year observations.

22

EPS targett =

α + β1 Management Forecastt + β2 Analyst Forecastt + β3 Sizet-1

(3)

+ β4 Market-to-Bookt-1 + β5 ABRETt-1 + β6 ∆ROAt-1 + β7 EarnVolt-1 + β8 Losst + β9 Litigation Riskt + εt

The EPS targett variable is the EPS target in the annual incentive plan for period t scaled by stock price. The Management Forecastt variable is the management earnings forecast issued right before the approval date of annual incentive plan divided by stock price. The Analyst Forecastt variable is the mean consensus analyst forecasts three days after the date of management forecast in the event period scaled by stock price. All other variables are defined previously. Table 7 reports the estimation results. In column (1) of Table 7, we first include the Management Forecastt variable along with control variables in the model. We find a positive and significant coefficient on the Management Forecastt variable, suggesting that corporate boards might rely on management earnings forecasts to set the EPS targets of CEO annual incentive plans. In column (2), we include the Analyst Forecastt variable in the model and find a positive and significant coefficient, suggesting that corporate boards might rely on consensus analyst forecasts to set the EPS target. It is notable that the adjusted R2 increases from 0.528 in column (1) to 0.713 in column (2). Therefore, consensus analyst forecasts seem to have more explanatory power for the variance of the EPS target used in bonus plans. Moreover, the coefficient estimate of Analyst Forecastt is 0.552 while that of Management Forecastt is only 0.117. In column (3), we include both the Management Forecastt variable and the Analyst Forecastt variable in the regression. The results show that management forecasts are positive but statistically indistinguishable from zero. In contrast, consensus analyst forecasts are incrementally significant over and above management earnings forecasts. Taken together, the results from Table 7 suggest that it is more likely that corporate boards rely on analyst forecasts 23

to set the performance target in CEO annual incentive plans. These results further validate our assumption.

Do managers benefit from pessimistic earnings guidance? In developing our argument we assume that performance target is set lower when management issues pessimistic earnings guidance before compensation meetings. In this section, we check whether this indeed is the case. In Table 8, we provide the results testing whether management guidance issued in the event window is associated with the relative easiness at which managers achieve the actual EPS target. We measure the dependent variable of the relative easiness, DIFF-Target-Past EPSt (DIFF-Target-Past EPS targett), by the difference between EPS target in current period and the actual EPS in period t-1 (the EPS target in period t1) divided by stock price in column (1) (in column (2)). The higher the values, the more difficult is to achieve the EPS target. In column (1), we find a significantly positive coefficient on the MF Biast variable at the five percent level. This evidence supports our assumption that pessimistic earnings guidance issued before compensation meetings makes EPS target easier to achieve. In column (2), we continue to find significantly positive coefficient on MF Biast.

Differentiating board naïve reliance from shareholder pressure to rely on external benchmark The results from prior analysis based on the institutional ownership concentration suggest that shareholder pressure leads corporate boards to rely on analyst forecasts to set performance target in executive annual incentive plans. To further differentiate this explanation from that corporate boards naively rely on external benchmarks in setting the performance target we conduct analysis based on the period before 2006. As noted in Section II, companies are required

24

to disclose details of compensation structure after 2006, where performance target and compensation committee meetings are also disclosed in a company’s annual proxy filing. If the latter explanation is true, we would expect the results to hold before 2006 because the reliance on external benchmarks by boards in setting performance metric should not depend on whether performance metric is publicly disclosed or not. If the former explanation is true, however, we would expect the results only to hold after 2006 because the expanded public disclosures on executive compensation structure provide shareholder activists the information, thus the enhanced ability to be involved in the executive compensation process. Table 9 reports the results for the period before 2006. Before 2006, we do not observe performance targets in annual incentive plans because they were not disclosed. Therefore, we use the sample firms that have relatively constant compensation policy regarding the choice of performance metrics during the post-2006 sample period. In doing so we assume that these firms have the same policy before 2006 as that after 2006.9 Because we do not observe the actual date of compensation committee meetings before 2006, we assume that these meeting dates do not vary significantly across years and hypothetically define compensation committee meeting dates before 2006 based on the typical compensation committee meeting dates for each firm after 2006.10 We then identify the latest management earnings guidance issued before the hypothetical compensation committee meeting dates, Pseudo Event EPS targett, in the same way as our main analysis. In Table 9, we find a negative coefficient on the Pseudo Event EPS targett variable but

9

Our final sample consist of 269 unique firms during the post-2006 sample period. We find that 86% (= 231 firms / 269 firms) of our sample firms has not changed the compensation policy regarding the choice of performance target metrics in the post-2006 period, indicating that the choice is sticky. As a robustness check, we use all sample firms (i.e., 269 firms) and assume the first year’s performance target metrics during the post-2006 sample period as the performance target metrics before 2006, and find qualitatively similar results. 10 We find that the average of each firm’s standard deviation of meeting dates during the post-2006 sample period is 8.92 days, indicating that each firm’s approval dates do not vary significantly. We further check the sensitivity of our results assigning each firm’s first or the last compensation committee meeting date during the post-2006 sample period as the hypothetical meeting dates before 2006. We find similar results.

25

it is statistically insignificant. Furthermore, the absolute value of the coefficient estimate is much smaller than that reported in Table 3. Therefore, our combined evidence suggests that corporate board accedes to shareholder pressure rather than acts naïvely to rely on external benchmarks in setting performance target.

VII.

Conclusion This study examines whether managers issue pessimistic earnings guidance right before

the approval date of executive annual incentive plans. We find that indeed management guidance issued before the approval date is more pessimistic than that issued in other periods. In addition, the pessimistic management earnings forecasts are more pronounced when the prevailing consensus analyst forecasts before the guidance issuance are more optimistic. Moreover, we show that analysts revise their forecasts downward during the event period presumably in response to the pessimistic guidance issued by the management. Finally, we show that analyst forecasts rather than management forecasts are likely to be referenced by boards in setting performance target and pessimistic management guidance indeed lowers the actual performance target in bonus plans. Our study highlights the opportunism in issuing management guidance to the private benefits of management. More interestingly, we find that managers’ guidance is more pessimistic when institutional ownership is more concentrated or when the CEO is also a chairperson of the corporate board. The opportunism of management-issued earnings forecasts seems to result from shareholder activism, which intends to mitigate such opportunism in the first place. Our study demonstrates that strong internal governance such as separation of CEO role from board

26

chairman is pivotal to curb such opportunism. Thus, internal governance seems to be more effective than external monitors to alleviate agency costs in our setting.

27

Reference Aboody, D., & Kasznik, R. (2000). CEO stock option awards and the timing of corporate voluntary disclosures. Journal of Accounting and Economics, 29(1), 73-100. Ajinkya, B., Bhojraj, S., & Sengupta, P. (2005). The association between outside directors, institutional investors and the properties of management earnings forecasts. Journal of Accounting Research, 43(3), 343-376. Ajinkya, B. B., & Gift, M. J. (1984). Corporate managers' earnings forecasts and symmetrical adjustments of market expectations. Journal of Accounting Research, 22(2), 425-444. Anderson, S. W., Dekker, H. C., & Sedatole, K. L. (2008). An empirical examination of negotiated goals and performance-to-goal following the introduction of an incentive bonus plan, Working Paper 37. Baginski, S. P., Conrad, E. J., & Hassell, J. M. (1993). The effects of management forecast precision on equity pricing and on the assessment of earnings uncertainty. The Accounting Review, 913-927. Bamber, L. S., & Cheon, Y. S. (1998). Discretionary management earnings forecast disclosures: Antecedents and outcomes associated with forecast venue and forecast specificity choices. Journal of Accounting Research, 68(4), 167-190. Bergman, N. K., & Roychowdhury, S. (2008). Investor sentiment and corporate disclosure. Journal of Accounting Research, 46(5), 1057-1083. Boone, A. L., & White, J. T. (2014). The effect of institutional ownership on firm transparency and information production. Journal of Financial Economics, forthcoming Brockman, P., Khurana, I. K., & Martin, X. (2008). Voluntary disclosures around share repurchases. Journal of Financial Economics, 89(1), 175-191. Cheng, Q., & Lo, K. (2006). Insider trading and voluntary disclosures. Journal of Accounting Research, 44(5), 815-848. Core, J., Holthausen, R., & Larcker, D. (1999). Corporate governance, Chief Executive Officer Compensation, and firm performance. Journal of Financial Economics 51, 371-406. Cotter, J., Tuna, I., & Wysocki, P. D. (2006). Expectations management and beatable targets: How do analysts react to explicit earnings guidance? Contemporary Accounting Research, 23(3), 593-624. Dimitrov, V., & Jain, P. (2011). It’s showtime: Do managers report better news before annual shareholder meetings? Journal of Accounting Research, 49(5), 1193-1221. 28

Ertimur, Y., Ferri, F., & Muslu, V. (2010). Shareholder activism and CEO pay. Review of Financial Studies, 113, 1-58. Faulkender, M., & Yang, J. (2010). Inside the black box: The role and composition of compensation peer groups. Journal of Financial Economics, 96(2), 257-270. Feng, M., & McVay, S. (2010). Analysts' incentives to overweight management guidance when revising their short-term earnings forecasts. The Accounting Review, 85(5), 1617-1646. Field, L., Lowry, M., & Shu, S. (2005). Does disclosure deter or trigger litigation? Journal of Accounting and Economics, 39(3), 487-507. Francis, J., Philbrick, D., & Schipper, K. (1994). Shareholder litigation and corporate disclosures. Journal of Accounting Research, 32(2), 137-164. Gaver, J. J., Gaver, K. M., & Austin, J. R. (1995). Additional evidence on bonus plans and income management. Journal of Accounting and Economics, 19(1), 3-28. Gillan, S. L., & Starks, L. T. (2002). Institutional investors, corporate ownership and corporate governance: Global perspectives (No. 2002/09). WIDER Discussion Papers/World Institute for Development Economics (UNU-WIDER). Gillan, S. L., & Starks, L. T. (2007). The evolution of shareholder activism in the United States. Journal of Applied Corporate Finance, 19(1), 55-73. Hartzell, J. C., & Starks, L. T. (2003). Institutional investors and executive compensation. The Journal of Finance, 58(6), 2351-2374. Holthausen, R. W., Larcker, D. F., & Sloan, R. G. (1995). Annual bonus schemes and the manipulation of earnings. Journal of Accounting and Economics, 19(1), 29-74. Healy, P. M. (1985). The effect of bonus schemes on accounting decisions. Journal of Accounting and Economics, 7(1), 85-107. Hui, K. W., Matsunaga, S., & Morse, D. (2009). The impact of conservatism on management earnings forecasts. Journal of Accounting and Economics, 47(3), 192-207. Karamanou, I., & Vafeas, N. (2005). The association between corporate boards, audit committees, and management earnings forecasts: An empirical analysis. Journal of Accounting Research, 43(3), 453-486. Kasznik, R., & Lev, B. (1995). To warn or not to warn: Management disclosures in the face of an earnings surprise. The Accounting Review, 70(1), 113-134. Kim, D. S., Martin, X., & Yang, J. (2015) Beating the target: Performance management around the annual incentive target, Working Paper. 29

Kim, D. S., & Yang, J. (2012, December). Behind the scenes: Performance target setting of annual incentive plans. In AFA 2010 Atlanta Meetings Paper. Kothari, S. P., Shu, S., & Wysocki, P. D. (2009). Do managers withhold bad news? Journal of Accounting Research, 47(1), 241-276. Lennox, C. S., & Park, C. W. (2006). The informativeness of earnings and management's issuance of earnings forecasts. Journal of Accounting and Economics, 42(3), 439-458. Lennox, C, & Ge, R. (2011). Do acquirers disclose good news or withhold bad news when they finance their acquisitions using equity? Review of Accounting Studies, 16(1), 183-217. Matsumoto, D. A. (2002). Management's incentives to avoid negative earnings surprises. The Accounting Review, 77(3), 483-514. Murphy, K. J. (1999). Executive compensation. Handbook of labor economics, 3, 2485-2563. Noe, C. F. (1999). Voluntary disclosures and insider transactions. Journal of Accounting and Economics, 27(3), 305-326. Richardson, S., Teoh, S. H., & Wysocki, P. D. (2004). The walk‐down to beatable analyst forecasts: The role of equity issuance and insider trading incentives. Contemporary Accounting Research, 21(4), 885-924. Rogers, J. L., & Stocken, P. C. (2005). Credibility of management forecasts. The Accounting Review, 80(4), 1233-1260. Rogers, J. L., & Van Buskirk, A. (2009). Shareholder litigation and changes in disclosure behavior. Journal of Accounting and Economics, 47(1), 136-156. Rogers, J. L., & Van Buskirk, A. (2013). Bundled forecasts in empirical accounting research. Journal of Accounting and Economics, 55(1), 43-65. Skinner, D. J. (1994). Why firms voluntarily disclose bad news. Journal of accounting research, 38-60.Skinner, D. J. (1997). Earnings disclosures and stockholder lawsuits. Journal of Accounting and Economics, 23(3), 249-282. Smith, M. P. (1996). Shareholder activism by institutional investors: Evidence from CalPERS. The Journal of Finance, 51(1), 227-252. Soffer, L. C., Thiagarajan, S. R., & Walther, B. R. (2000). Earnings preannouncement strategies. Review of Accounting Studies, 5(1), 5-26.

30

Appendix A Variable Definitions Variables

Descriptions

Event EPS targett

Event EPS targett is an indicator variable equal to one for the latest management earning forecast issued during the 90 day period prior to the approval date of annual incentive plans for firms with EPS performance target but without revenue performance target, zero otherwise. Event EPS REV targett is an indicator variable equal to one for the latest management earning forecast issued during the 90 day period prior to the approval date of annual incentive plans for firms with both EPS and revenue performance targets, zero otherwise. Event REV targett is an indicator variable equal to one for the latest management earning forecast issued during the 90 day period prior to the approval date of annual incentive plans for firms with revenue performance target but without EPS performance target, zero otherwise. MF Biast is measured as the management quantitative earnings forecast less the prevailing mean consensus analyst forecast 3 days before the date of management forecast scaled by stock price 3 days before the date of management forecast, and multiplied by 100. If the management forecast is the range forecast, we choose the mid-point of the forecast. If the management forecast is a one-sided range forecast (e.g., it specifies a maximum or a minimum value), we calculate forecast news only if the mean consensus analyst forecast is above (below) the maximum (minimum) value of the one-sided range forecast; otherwise, the management forecast is assumed to contain no news. Sizet-1 is measured as the natural logarithm of firm i’s market value of equity at the end of fiscal quarter preceding the date of management forecast. Market-to-Booktt-1 is measured by the market-to-book ratio at the end of fiscal quarter preceding the date of the management forecast. ABRETt-1 is defined as cumulative abnormal returns measured as the excess firm returns over the CRSP value-weighted returns during the three months ending 2 days before the issuance of a management forecast. ∆ROAt-1 is measured as changes in return on assets in period t-1. Return on assets are calculated as income before extraordinary items divided by lagged total assets. EarnVolt-1 is measured as standard deviation of quarterly earnings scaled by lagged total assets over 12 quarters preceding the date of management forecast. Litigation Riskt is an indicator variable equal to one for firms in the biotechnology (SIC 2833–2836 and 8731–8734), computers (3570–3577 and 7370–7374), electronics (3600–3674), and retail (5200–5961) industries, zero otherwise.

Event EPS REV targett

Event REV targett

MF Biast

Sizet-1

Market-to-Bookt-1 ABRETt-1

∆ROAt-1

EarnVolt-1

Litigation Riskt

31

Losst-1 Horizont AF Revisiont

INSTOWN TOP5t-1

CEO Dualityt

Losst-1 is an indicator variable equal to one if the firm reported losses in period t-1, zero otherwise Horizont is measured as the difference between the date of fiscal year-end and the date of management forecast divided by 365. AF Revisiont is defined as the mean consensus analyst forecast three days after the date of management forecast less the mean consensus analyst forecast three days prior to the date of management forecast scaled by the absolute value of the latter. INSTOWN TOP5t-1 is measured as the sum of top five institutional ownership at the end of fiscal quarter preceding the date of management forecast. CEO Dualityt is an indicator variable equal to one if a CEO serves to the chairperson of the board, zero otherwise.

32

Appendix B Examples Example 1. Mar 25, 2010 Hess Corporation proxy statement

33

Example 2. Jan 27, 2011 Air Products and Chemicals, Inc. proxy statement

34

Example 3.May 28, 2009 Affiliated Computer Services, Inc. proxy statement

35

Figure 1 Timeline

The figure represents the timeline of events. Compensation committee meetings determining annual incentive plans are typically held during the first quarter of the fiscal year. We identify the latest management earning guidance issued during the 90-day window (the event period) before the approval date of annual incentive plans for firms with either an EPS performance target or a performance revenue target in bonus plans. Prevailing analyst expectations to gauge the management forecast bias is measured three days prior to the date of latest management earnings guidance issued in the event period.

36

Table 1 Descriptive Statistics Variables

N

Mean

Std

Q1

Median

Q3

Event EPS targett Event EPS REV targett Event REV targett MF Biast Sizet-1 Market-to-Bookt-1 ABRETt-1 ∆ROAt-1 EarnVolt-1 Losst-1 Litigation Riskt Horizont AF Revisiont INSTOWN TOP5t-1 CEO Dualityt

4,515 4,515 4,515 4,515 4,515 4,515 4,515 4,515 4,515 4,515 4,515 4,515 4,458 4,031 3,294

0.060 0.044 0.026 -0.197 9.530 1.988 0.004 0.001 0.010 0.036 0.233 0.628 -0.017 0.256 0.648

0.238 0.205 0.159 1.483 0.915 1.051 0.108 0.036 0.014 0.185 0.423 0.412 0.034 0.071 0.478

-0.299 8.947 1.256 -0.062 -0.011 0.003 0.317 -0.028 0.207 -

-0.092 9.447 1.678 0.004 0.001 0.005 0.650 -0.006 0.243 -

0.004 10.084 2.359 0.070 0.011 0.010 0.886 0.000 0.294 -

This table reports descriptive statistics for the sample with available information. The sample period ranges from 2006 to 2012. All continuous variables are winsorized at the 1 and 99 th percentiles. All variables are defined in the Appendix.

37

Table 2 Correlations Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15)

Event EPS targett Event EPS REV targett Event REV targett MF Biast Sizet-1 Market-to-Bookt-1 ABRETt-1 ∆ROAt-1 EarnVolt-1 Losst-1 Litigation Riskt Horizont AF Revisiont INSTOWN TOP5t-1 CEO Dualityt

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

(13)

(14)

(15)

-0.05 -0.04 -0.08 -0.07 -0.09 0.00 -0.01 -0.06 0.02 -0.07 0.30 -0.08 -0.01 0.02

-0.05 -0.04 -0.06 0.01 0.09 0.00 0.01 0.05 0.00 0.05 0.25 -0.08 0.01 -0.02

-0.04 -0.04 -0.04 -0.01 0.05 0.01 0.01 0.03 -0.01 0.10 0.19 -0.07 0.04 -0.07

-0.05 -0.03 0.01 0.00 -0.02 -0.09 0.04 -0.05 -0.03 -0.03 -0.11 0.61 -0.10 -0.05

-0.07 0.03 0.00 0.01 0.25 0.03 0.06 -0.04 -0.13 0.12 0.04 0.05 -0.33 0.20

-0.06 0.05 0.04 -0.05 0.17 0.07 0.10 0.36 -0.15 0.21 0.00 -0.04 0.08 0.00

0.00 0.00 0.01 -0.05 0.02 0.05 0.05 0.03 0.00 0.00 0.00 -0.12 0.02 -0.01

0.00 0.01 0.00 0.04 0.02 0.04 0.02 0.03 -0.21 0.05 0.01 0.06 0.03 -0.01

-0.03 0.03 0.00 -0.04 -0.04 0.27 0.01 0.12 0.25 0.24 0.00 -0.08 0.14 -0.02

0.02 0.00 -0.01 -0.03 -0.14 -0.08 0.00 -0.30 0.36 0.01 -0.01 -0.03 0.15 -0.10

-0.07 0.05 0.10 -0.07 0.11 0.20 0.00 0.02 0.25 0.01 -0.02 -0.05 0.07 -0.04

0.21 0.16 0.12 -0.03 0.03 -0.02 0.00 0.03 0.03 0.00 -0.04 -0.11 -0.05 0.04

-0.09 -0.08 -0.08 0.22 0.08 -0.01 -0.12 0.07 -0.05 -0.07 -0.04 -0.08 -0.05 0.02

0.01 0.00 0.00 -0.11 -0.38 0.04 0.01 0.02 0.18 0.11 0.05 -0.04 -0.07 -0.14

0.02 -0.02 -0.07 0.02 0.19 -0.02 0.00 0.00 -0.01 -0.10 -0.04 0.05 0.02 -0.21 -

This table reports Pearson (Above) / Spearman (Below) correlations of main variables. Sample period ranges from 2006 to 2012. All continuous variables are winsorized at 1st and 99th percentiles. Significance level at 5% is bolded. All variables are defined in the Appendix.

38

Table 3 Management EPS forecast bias prior to the approval date of AIP MF Biast Independent Variables Event EPS targett Event EPS REV targett Event REV targett Sizet-1 Market-to-Bookt-1 ABRETt-1 ∆ROAt-1 EarnVolt-1 Losst-1 Litigation Riskt Horizont Constant

Firm FE Year FE Fiscal Quarter Indicator Number of observations Adjusted R squared Coefficient difference tests 𝛽1 (Event EPS targett) = 𝛽2 (Event EPS REV targett) 𝛽1 (Event EPS targett) = 𝛽3 (Event REV targett) 𝛽2 (Event EPS REV targett) = 𝛽3 (Event REV targett)

(1)

(2)

(3)

(4)

-0.204*** (-3.414) 0.117 (0.791) 0.179* (1.849) -0.156 (-1.289) 0.202 (0.497) 1.873 (0.864) -0.148 (-1.311) -0.255** (-2.043) 0.085 (1.234) -2.331* (-1.777)

-0.049 (-0.853) 0.113 (0.768) 0.180* (1.879) -0.156 (-1.287) 0.211 (0.519) 1.976 (0.914) -0.152 (-1.339) -0.259** (-2.056) 0.073 (1.097) -2.329* (-1.772)

0.007 (0.110) 0.114 (0.768) 0.180* (1.881) -0.155 (-1.284) 0.203 (0.496) 1.973 (0.910) -0.152 (-1.340) -0.258** (-2.034) 0.072 (1.076) -2.339* (-1.778)

-0.222*** (-3.819) -0.099* (-1.743) -0.050 (-0.779) 0.117 (0.787) 0.178* (1.841) -0.157 (-1.296) 0.220 (0.550) 1.866 (0.869) -0.146 (-1.305) -0.255** (-2.064) 0.089 (1.294) -2.303* (-1.750)

Yes Yes Yes 4,515 0.819

Yes Yes Yes 4,515 0.818 Coeff. Diff. -0.123 -0.172 -0.0492

Yes Yes Yes 4,515 0.818 F-Stat 2.152 4.155 0.418

Yes Yes Yes 4,515 0.819 p-Value 0.144 0.043 0.518

This table present the estimation results from the following regression model. MF Biast

=

α + β1 INDVARt + β2 Sizet-1 + β3 Market-to-Bookt-1 + β4 ABRETt-1 + β5 ∆ROAt-1 + β6 EarnVolt-1 + β7 Losst-1 + β8 Litigation Riskt + β9 Horizont + ԑt

INDVARt represents Event EPS targett, Event EPS REV targett, or Event REV targett in column (1), column (2), and column (3), respectively. In column (4), all three variables are included in the regression simultaneously. All other variables are defined in the Appendix and all continuous variables are winsorized at the 1 and 99 th percentiles. Standard errors are clustered by firm. Robust t-statistics are in parentheses. *, **, and *** represent significance level at the 10%, 5%, and 1%, respectively.

39

Table 4 Cross-sectional variations: Analyst forecast bias MF Biast Variables Event EPS targett AF Biast Event EPS targett × AF Biast Sizet-1 Market-to-Bookt-1 ABRETt-1 ∆ROAt-1 EarnVolt-1 Losst-1 Litigation Riskt Horizont Constant

Firm FE Year FE Fiscal Quarter Indicator Number of observations Adjusted R squared

(1)

(2)

(3)

Actual EPS -0.030 (-0.526) -0.163*** (-2.944) -0.328*** (-3.604) 0.121 (0.752) 0.164* (1.753) -0.154 (-1.140) 0.211 (0.463) 0.832 (0.234) -0.126 (-0.922) -0.325** (-2.348) 0.102 (1.300) -2.307 (-1.604)

Projected EPS -0.120* (-1.683) -0.079 (-1.295) -0.169* (-1.723) 0.090 (0.562) 0.178* (1.944) -0.079 (-0.582) 0.172 (0.344) 0.357 (0.097) -0.067 (-0.556) -0.295** (-2.286) 0.106 (1.378) -2.089 (-1.471)

Target EPS 0.051 (0.811) -0.204*** (-3.017) -0.460*** (-5.116) 0.117 (0.560) 0.218* (1.712) -0.039 (-0.246) -0.158 (-0.328) 1.228 (0.288) -0.141 (-1.046) -0.321** (-2.306) 0.134 (1.632) -2.543 (-1.374)

Yes Yes Yes 3,714 0.832

Yes Yes Yes 3,714 0.829

Yes Yes Yes 3,079 0.821

This table present the estimation results from the following regression model. MF Biast

=

α + β1 Event EPS targett + β2 AF Biast + β3 Event EPS targett × AF Biast + β4 Sizet-1 + β5 Market-to-Bookt-1 + β6 ABRETt-1 + β7 ∆ROAt-1 + β8 EarnVolt-1 + β9 Losst-1 + β10 Litigation Riskt + β11 Horizont + ԑt

AF Biast is an indicator variable equal to one if the mean consensus analyst forecast 3 days prior to the date of management forecast less the benchmark EPS figure divided by stock price is above the sample median, zero otherwise. The top row denotes the benchmark EPS figures. In column (1) the benchmark is IBES-reported actual EPS in period t. In column (2), the benchmark is IBES-reported actual EPS in period t-1 multiplied by EPS growth rate during period t-1. In column (3), the benchmark is the EPS target in annual incentive plans in the current period. All other variables are defined in the Appendix and all continuous variables are winsorized at the 1 and 99th percentiles. Standard errors are clustered by firm. Robust t-statistics are in parentheses. *, **, and *** represent significance level at the 10%, 5%, and 1%, respectively.

40

Table 5 Cross-sectional variations: Shareholder pressures and CEO power MF Biast Variables Event EPS targett INSTOWN TOP5t-1 Event EPS targett × INSTOWN TOP5t-1 CEO Dualityt Event EPS targett × CEO Dualityt Sizet-1 Market-to-Bookt-1 ABRETt-1 ∆ROAt-1 EarnVolt-1 Losst-1 Litigation Riskt Horizont Constant Firm FE Year FE Fiscal Quarter Indicator Number of observations Adjusted R squared

(1)

(2)

-0.041 (-0.471) 0.007 (0.077) -0.311** (-2.273) 0.111 (0.710) 0.187* (1.673) -0.111 (-0.785) 0.338 (0.760) 1.922 (0.813) -0.161 (-1.334) -0.289** (-2.047) 0.084 (1.116) -2.327* (-1.678)

-0.063 (-0.772) 0.106 (1.285) -0.250** (-2.155) 0.199 (1.352) 0.125 (1.466) -0.218 (-1.460) 0.444 (0.668) -1.544 (-0.555) -0.190 (-1.451) -0.349** (-2.583) 0.071 (0.875) -3.026** (-2.246)

Yes Yes Yes 4,031 0.826

Yes Yes Yes 3,294 0.826

This table present the estimation results from the following regression model. MF Biast

= α + β1 Event EPS targett + β2 INSTOWN TOP5t-1 (or CEO Dualityt) + β3 Event EPS targett × INSTOWN TOP5t-1 (or CEO Dualityt) + β4 Sizet-1 + β5 Market-to-Bookt-1 + β6 ABRETt-1 + β7 ∆ROAt-1 + β8 EarnVolt-1 + β9 Losst-1 + β10 Litigation Riskt + β11 Horizont + ԑt

The INSTOWN TOP5t-1 variable is an indicator variable equal to one if the sum of top five institutional ownership is above the sample median, zero otherwise. The CEO Dualityt variable is an indicator variable equal to one if the CEO is to serve the board chairperson, zero otherwise. All other variables are defined in the Appendix and all continuous variables are winsorized at the 1 and 99th percentiles. Standard errors are clustered by firm. Robust tstatistics are in parentheses. *, **, and *** represent significance level at the 10%, 5%, and 1%, respectively.

41

Table 6 The effect of pessimistic management earnings guidance on analyst forecast revisions Variables

AF Revisiont (2) (3) Actual EPS Projected EPS -0.002 -0.004 (-0.686) (-1.541) -0.010*** -0.001 (-6.196) (-0.680) -0.015*** -0.012** (-3.311) (-2.572) 0.008* 0.008 (1.959) (1.605) -0.006*** -0.005** (-3.311) (-2.596) -0.046*** -0.042*** (-6.897) (-6.292) 0.025 0.028 (1.007) (1.088) 0.081 0.057 (0.848) (0.588) -0.004 -0.002 (-0.525) (-0.228) 0.009*** 0.009*** (3.854) (3.493) -0.004* -0.004 (-1.893) (-1.571) -0.074* -0.072* (-1.905) (-1.681)

(1)

Event EPS targett AF Biast Event EPS targett × AF Biast Sizet-1 Market-to-Bookt-1 ABRETt-1 ∆ROAt-1 EarnVolt-1 Losst-1 Litigation Riskt Horizont Constant Firm FE Year FE Fiscal Quarter Indicator Number of observations Pseudo R-squared

-0.010*** (-3.624) 0.004 (1.015) -0.003* (-1.658) -0.045*** (-7.138) 0.047** (2.253) -0.010 (-0.147) 0.004 (0.614) 0.011*** (4.026) -0.004* (-1.793) -0.044 (-1.159) Yes Yes Yes 4,458 0.136

Yes Yes Yes 3,671 0.162

Yes Yes Yes 3,671 0.145

(4) Target EPS 0.002 (0.692) -0.006*** (-3.981) -0.025*** (-5.807) 0.010** (2.027) -0.006*** (-2.663) -0.041*** (-6.002) 0.006 (0.239) 0.076 (0.707) -0.003 (-0.279) 0.008*** (3.461) -0.002 (-0.791) -0.093** (-2.075) Yes Yes Yes 3,036 0.172

This table present the estimation results from the following regression model. AF Revisiont

= α + β1 Event EPS targett + β2 AF Biast + β3 Event EPS targett × AF Biast + β4 Sizet-1 + β5 Market-to-Bookt-1 + β6 ABRETt-1 + β7 ∆ROAt-1 + β8 EarnVolt-1 + β9 Losst-1 + β10 Litigation Riskt + β11 Horizont + ԑt

The dependent variable, AF Revisiont, is measured as the mean consensus analyst forecast 3 days after the date of management forecast less the mean consensus analyst forecast 3 days prior to the date of management forecast divided by the absolute value of the latter. AF Biast is an indicator variable equal to one if the mean consensus analyst forecast 3 days prior to the date of management forecast less the benchmark EPS figure divided by stock price is above the sample median, zero otherwise. The top row denotes the benchmark EPS figures. In column (2) the benchmark is IBES-reported actual EPS in period t. In column (3), the benchmark is IBES-reported actual EPS in period t-1 multiplied by EPS growth rate during period t-1. In column (4), the benchmark is the EPS target in annual incentive plans in the current period. All other variables are defined in the Appendix and all continuous variables are winsorized at the 1 and 99th percentiles. Standard errors are clustered by firm. Robust t-statistics are in parentheses. *, **, and *** represent significance level at the 10%, 5%, and 1%, respectively.

42

Table 7 Determinants of EPS performance targets in AIP

Variables

(1)

Management Forecastt Analyst Forecastt Sizet-1 Market-to-Bookt-1 ABRETt-1 ∆ROAt-1 EarnVolt-1 Losst-1 Litigation Riskt Horizont Constant

Industry FE Year FE Number of observations Adjusted R squared

EPS targett (2)

(3)

0.117** (2.181) -0.002 (-1.482) -0.008*** (-5.204) -0.049*** (-4.426) -0.007 (-0.308) 0.010 (0.159) 0.000 (0.117) 0.001 (0.382) 0.010 (0.781) 0.059** (2.486)

0.552*** (4.033) -0.001 (-0.557) -0.006*** (-5.303) -0.030*** (-2.684) 0.040* (1.922) -0.106 (-1.640) 0.008** (2.554) 0.005 (1.425) -0.000 (-0.024) 0.038* (1.711)

0.016 (0.872) 0.538*** (4.020) -0.001 (-0.601) -0.006*** (-5.166) -0.030*** (-2.740) 0.040* (1.941) -0.105 (-1.617) 0.008** (2.553) 0.005 (1.416) 0.000 (0.023) 0.037 (1.648)

Yes Yes 469 0.528

Yes Yes 469 0.713

Yes Yes 469 0.713

This table present the estimation results from the following regression model. EPS targett

=

α + β1 Management Forecastt + β2 Analyst Forecastt + β3 Sizet-1 + β4 Market-to-Bookt-1 + β5 ABRETt-1 + β6 ∆ROAt-1 + β7 EarnVolt-1 + β8 Losst-1 + β9 Litigation Riskt + β10 Horizont + ԑt

EPS targett is measured as the EPS target in bonus plans divided by stock price. Management Forecastt is defined as the management earnings forecast divided by stock price 3 days prior to the date of management forecast. Analyst Forecastt is defined as the mean consensus analyst forecast 3 days after the date of management forecast divided by stock price. Other variables are defined in the Appendix and all continuous variables are winsorized at the 1 and 99th percentiles. Standard errors are clustered by firm. Robust t-statistics are in parentheses. *, **, and *** represent significance level at the 10%, 5%, and 1%, respectively.

43

Table 8 The effect of pessimistic earnings guidance on performance targets

Variables MF Biast Sizet-1 Market-to-Bookt-1 ABRETt-1 ∆ROAt-1 EarnVolt-1 Losst-1 Litigation Riskt Horizont Constant

Industry FE Year FE Number of observations R squared

DIFF-Target-Past EPS (1)

DIFF-Target-Past EPS Target (2)

0.003** (2.289) 0.003* (1.896) -0.003 (-1.028) -0.005 (-0.494) 0.109** (2.035) -0.152 (-0.813) 0.014* (1.799) 0.011 (1.247) 0.008 (0.801) -0.023 (-0.967)

0.003** (2.378) 0.004* (1.977) 0.003 (1.080) 0.024 (1.285) 0.192*** (4.808) -0.058 (-0.442) -0.002 (-0.208) 0.003 (0.595) 0.010 (0.842) -0.054** (-2.023)

Yes Yes 469 0.181

Yes Yes 343 0.312

This table present the estimation results from the following regression model. DEPVARt

=

α + β1 MF Biast + β2 Sizet-1 + β3 Market-to-Bookt-1 + β4 ABRETt-1 + β5 ∆ROAt-1 + β6 EarnVolt-1 + β7 Losst-1 + β8 Litigation Riskt + β9 Horizont + ԑt

DEPVARt is either the DIFF-Target-Past EPSt variable in column (1) or the DIFF-Target Past EPS Targett variable in column (2). DIFF-Target-Past EPSt is measured as the EPS performance target in period t less IBESreported actual EPS in period t-1 divided by stock price. DIFF-Target-Past EPS Targett is measured as the EPS performance target in period t less the EPS performance target in period t-1 divided by stock price. Other variables are defined in the Appendix and all continuous variables are winsorized at the 1 and 99 th percentiles. Standard errors are clustered by firm. Robust t-statistics are in parentheses. *, **, and *** represent significance level at the 10%, 5%, and 1%, respectively.

44

Table 9 Falsification test: Management earnings guidance before 2006 Variables Pseudo Event EPS targett Pseudo Event EPS REV targett Pseudo Event REV targett Sizet-1 Market-to-Bookt-1 ABRETt-1 ∆ROAt-1 EarnVolt-1 Losst-1 Litigationt Horizont Constant

MF Bias (3)

(1)

(2)

-0.025 (-0.288) 0.392 (0.999) 0.028 (0.479) 0.323 (0.869) 1.483 (0.893) 17.125** (2.259) 0.404 (1.017) -0.078 (-0.543) -0.279*** (-3.246) -4.581 (-1.364)

-0.001 (-0.006) 0.391 (0.999) 0.028 (0.480) 0.324 (0.872) 1.482 (0.892) 17.147** (2.261) 0.404 (1.015) -0.078 (-0.542) -0.281*** (-3.315) -4.581 (-1.364)

-0.028 (-0.223) 0.392 (0.998) 0.028 (0.480) 0.324 (0.873) 1.481 (0.892) 17.150** (2.262) 0.403 (1.014) -0.078 (-0.544) -0.280*** (-3.299) -4.581 (-1.364)

-0.030 (-0.337) -0.012 (-0.112) -0.036 (-0.274) 0.392 (0.998) 0.028 (0.478) 0.323 (0.869) 1.482 (0.892) 17.122** (2.260) 0.404 (1.016) -0.078 (-0.543) -0.276*** (-3.059) -4.580 (-1.364)

Yes Yes Yes 3,449 0.907

Yes Yes Yes 3,449 0.907

Yes Yes Yes 3,449 0.907

Yes Yes Yes 3,449 0.907

Firm FE Year FE Fiscal Quarter Indicator Number of observations Adjusted R squared

(4)

This table present the estimation results from the following regression model. MF Biast

=

α + β1 Pseudo INDVARt + β2 Sizet-1 + β3 Market-to-Bookt-1 + β4 ABRETt-1 + β5 ∆ROAt-1 + β6 EarnVolt-1 + β7 Losst-1 + β8 Litigation Riskt + β9 Horizont + ԑt

Pseudo INDVARt represents Pseudo Event EPS targett, Pseudo Event EPS REV targett, or Pseudo Event REV targett in column (1), column (2), and column (3), respectively. In column (4), all three variables are included in the regression simultaneously. The sample period in this table ranges from 1999 to 2005 (i.e., pre-period). Since we do not observe the performance targets before 2006, we use the sample firms that have not changed their performance targets during the sample period 2006-2012 and assume that those firms have the same performance targets before 2006. We also hypothetically assign the approval date of bonus plans based on the typical approval date for each firm in the sample period 2006-2012 because we do not observe the approval date of bonus plans before 2006. All other variables are defined in the Appendix and all continuous variables are winsorized at the 1 and 99th percentiles. Standard errors are clustered by firm. Robust t-statistics are in parentheses. *, **, and *** represent significance level at the 10%, 5%, and 1%, respectively.

45