Stock-Performance Goals in Executive Compensation Contracts and Management Earnings Guidance. Sean Shun Cao Georgia State University

Transcription

1 Stock-Performance Goals in Executive Compensation Contracts and Management Earnings Guidance Sean Shun Cao Georgia State University Guojin Gong Pennsylvania State University Laura Yue Li University of Illinois at Urbana-Champaign Ha Young Yoon University of Illinois at Urbana-Champaign We thank workshop participants at University of Illinois at Urbana-Champaign and Georgia State University.

2 Stock-Performance Goals in Executive Compensation Contracts and Management Earnings Guidance Abstract We examine whether the incentive to achieve stock-performance goals in executive compensation contracts affects managers discretionary guidance behavior. Results show that when approaching the end of evaluation period, managers facing Right-Below performance, defined as prevailing stock performance below the lowest performance goal by a small margin, are less (more) likely to issue bad (good) news guidance compared with managers facing other performance zones; however, guidance issued by Right-Below firms does not contain more optimistic bias. Such discretionary guidance behavior regarding the timing of good versus bad news magnifies with the monetary benefits from achieving performance goals and weakens with the costs of exercising discretion and the likelihood of being detected. Consistent with the market unable to fully unravel managers discretionary guidance behavior, firms with Right-Below performance have more optimistic analysts forecasts and higher abnormal stock returns near the end of the evaluation period. Shortly after the evaluation period ends, Right-Below firms are more likely to issue bad news guidance and experience declining stock prices. The overall evidence is consistent with managers withholding (accelerating) bad (good) news guidance near the end of the evaluation period to boost the likelihood of achieving stockperformance goals. Keywords: stock-performance goals, executive compensation, management earnings guidance.

3 1. Introduction Stock-based performance evaluation determines an agent s payouts based on her achievement of stock-performance goals during a pre-specified evaluation period. This design feature becomes increasingly popular in executive compensation contracts for large U.S. firms (Core and Packard 2016; Bettis et al. 2016). From 2006 to 2015, the use of stock-based performance evaluation has increased from 21 percent to 50 percent among the largest 750 U.S. public firms and the associated payouts corresponds to 35 percent of total compensation for CEOs. Given the importance of stock-performance goals in evaluating top executives of U.S. firms, we examine the impact of stock-based performance evaluation on managerial incentive to achieve stock-performance goals through discretionary disclosure. Specifically, we focus on managers voluntary earnings guidance behavior and investigate whether managers exert discretion in issuing earnings guidance near the end of evaluation period to manipulate stock performance and boost the chance of achieving stock-performance goals. Prior literature suggests that design features in executive compensation contracts affect corporate reporting decisions (see Healy and Palepu (2001) and Beyer et al. (2010) for recent literature reviews). In this literature, most studies examine accounting-performance goals and how such goals induce managers to manipulate reported earnings (e.g., Healy 1985; Holthausen et al. 1995; Gaver et al. 1995; Larcker et al. 2007; Gong et al. 2016; Bennett et al. 2017). 1 Our study focuses on stock-performance goals and highlights the unique incentive effect on managerial 1 Earlier studies, such as Healy (1985), Holthausen et al. (1995), and Gaver et al. (1995), focus on how the structure of pay-performance relation in bonus plans relates to abnormal accruals. We note that bonus plans mostly employ accounting-performance metrics. In addition, Larcker et al. (2007) find a positive association between the weight of accounting-based compensation and abnormal accruals, suggesting that accounting-based pay provides stronger incentives for earnings manipulation than equity-based pay. More recently, Bennett et al. (2017) find a kink around absolute accounting-performance goals and attribute such kink to discretionary earnings reporting. Gong et al. (2016) focus on relative accounting-performance goals and find that firms strategically timing earnings announcements to facilitate discretionary reporting for the purpose of achieving relative performance goals. 1

4 discretion in voluntary forward-looking disclosure. Stock-based performance evaluation differs substantially from accounting-based performance evaluation and may induce managers to exert discretion in different activities. Under stock-based performance evaluation, most firms have stock performance evaluated at fiscal year end (about 86 percent of grants in our sample), which is several weeks or even months before the release of annual accounting performance. Accordingly, the reporting of annual accounting performance per se (more precisely, the fourth-quarter accounting performance) seldom affects stock-based performance evaluation. In contrast, the issuance of guidance of future earnings can significantly affect market expectation and stock performance towards the end of evaluation period (Cheng and Lo 2006), as managers can exert discretion over the timing and news content of earnings guidance. Therefore, when earnings guidance is more effective in boosting the likelihood of achieving stock-performance goals, managers are more likely to delay (accelerate) bad (good) news earnings guidance near the end of the evaluation period. 2 Moreover, managers may deliberately introduce optimistic biases into earnings guidance to further enhance the likelihood of achieving stock-performance goals. Issuing biased guidance, however, involves higher legal and reputational costs than altering the timing of earnings guidance. We note two counter forces that possibly limit or prevent managers discretionary guidance behavior described above. First, manipulating stock performance is generally more difficult than manipulating self-reported earnings, as stock performance is not only affected by self-reported performance but also by market participants expectations. During our sample period, the SEC requires public firms to provide detailed disclosures about performance goals in proxy statements. It is possible that sophisticated market participants anticipate managers discretionary guidance 2 In this paper, we use delaying bad news interchangeably with withholding bad news (to a later period), as both lead to less bad news towards the end of the evaluation period. 2

5 behavior and more cautiously interpret issued (or the lack of) earnings guidance near the end of the performance evaluation period, discouraging managers from manipulating earnings guidance to achieve stock-performance goals. Second, towards the year end, managers may have contradictory incentives in terms of issuing bad news guidance. In particular, prior evidence suggests that managers may benefit from accelerating, instead of delaying, bad news guidance towards the year end to walk down market expectation and enhance the chance of meeting/beating market expectation (e.g., Richardson et al. 2004). Hence, whether the incentive to achieve stockperformance goals induces managers to opportunistically time the release of earnings guidance awaits empirical investigation. Using a sample of stock-performance based compensation grants between 2006 and 2015, we find that firms with the prevailing stock performance marginally below stock-performance goals (hereafter, Right-Below performance) are less (more) likely to issue bad (good) news earnings guidance during the last quarter of the performance evaluation period, compared with firms having stock performance further below or above stock-performance goals. We fail to find that firms with Right-Below performance provide more optimistic guidance during the last quarter of the evaluation period than firms in other performance zones. These results support our proposition that managers strategically time the release of earnings guidance, but do not impose optimistic biases in issued guidance, towards the end of the performance evaluation period. In cross-sectional analysis, we further examine whether the discretionary disclosure behavior documented above varies with the costs of manipulative disclosure and the benefits of achieving the stock performance goals. We find that the association between Right-Below performance and managers tendency to release less (more) bad (good) news guidance is more pronounced when firms operate in a less litigious environment, when external monitoring is 3

6 relatively weak (proxied by fewer institutional block holders and lower analyst coverage), when discretionary guidance behavior is less likely to be detected by the market (proxied by higher return volatility), and when issuing less (more) bad (good) news guidance is less likely to cause severe market mispricing (proxied by the prevailing market expectation is more pessimistic). In addition, this association is more pronounced when the underlying grants have larger fair value and do not allow the unfulfilled payouts to be rolled over to future periods. Findings from these crosssectional tests reinforce our proposition that managers facing Right-Below performance exert discretion in issuing earnings guidance to boost the likelihood of achieving stock performance goals. We also examine potential implications of the discretionary guidance behavior for realized stock performance, analysts earnings forecasts, and subsequent earnings guidance. Specifically, firms having Right-Below performance experience more favorable analysts forecasts and better stock performance near the end of the evaluation period, compared with firms in other performance zones. However, firms having Right-Below performance experience declining stock performance and a reversal in guidance behavior (i.e., more bad news guidance) shortly after the end of the evaluation period. These findings, together with the cross-sectional results, further corroborate our proposition that the documented guidance behavior for Right-Below firms is driven by managerial incentive to inflate stock performance and achieve stock-performance goals, 3 rather than the natural arrival of earnings news (that are not expected to reverse shortly after) or managers taking greater efforts in the last quarter (which should not vary with costs and benefits of discretionary guidance). 3 In additional analysis, we examine firms using only accounting-performance goals as a falsification test. 4

7 Our study extends prior literature in several ways. First, our findings highlight the distinctive incentive effect arising from stock-based performance evaluation and its implications for managers discretionary disclosure decisions. Prior studies suggest that stock- and accountingperformance goals are both informative about managerial efforts and valuable in performance evaluation (e.g., Holmstrom 1979, 1982; Baiman and Demski 1980; Lambert and Larcker 1987; Sloan 1993). However, accounting-performance goals induce managers to manipulate financial reporting for the purpose of boosting compensation (e.g., Healy 1985; Bennett et al. 2017), raising concerns over the use of accounting-based performance evaluation in executive compensation. Our evidence suggests that stock-based performance evaluation may as well induce managers discretionary behavior, calling caution for the increasing popularity of stock-performance goals. In particular, discretion in issuing forward-looking disclosures, which is protected by Safe Harbor regulations, allows managers to time earnings guidance for the purpose of achieving stockperformance goals. Thus, it is generally more challenging to regulate discretionary guidance behavior than discretionary earnings reporting, as the latter is required to comply with GAAP and subject to more diligent public scrutiny from external auditors and regulatory agencies. Our findings also have implications for the market mechanism and disclosure regulations that potentially discipline managers manipulative behavior induced by compensation design. The evidence suggests that market mechanism, by itself, may not deter manipulative disclosure induced by stock-based performance evaluation. Although SEC requires public firms to disclose details about performance goals in proxy statements, such disclosures are often vague and sometimes deferred until the evaluation period ends. 4 Moreover, Right-Below performance is largely 4 In our sample, for only 23 percent of the grants, performance goals were disclosed in the proxy immediately following the grant date. In untabulated tests, we find that managers having Right-Below performance demonstrate a higher tendency to withhold bad news if performance goals are not disclosed in a timely fashion (i.e., not immediately 5

8 transitory, which further weakens the market s ability to anticipate the impact of Right-Below performance on managers discretionary guidance, leaving room for managerial manipulation. More transparent and timely disclosures of stock-performance goals may facilitate market participants to better understand the implications of compensation design for managerial decisions. Finally, our study adds to the voluntary disclosure literature by documenting a contractdriven incentive for managers discretionary guidance. We answer the call for research on the timing, the specific incentive, and the participants in disclosure manipulation (e.g. Beyer, Cohen, Lys, and Walther 2010). Our findings alert investors to managers strategic guidance in the window leading to the end of stock-performance evaluation period. Existing studies have examined the effects of several executive compensation features (i.e. option grants, length of stock-option vesting period, and size of severance pay) on managers disclosure behavior (Aboody and Kasznik 2000; Cheng et al. 2014; Baginski et al. 2017). Given the significant decline of option grants and the increasing popularity of stock-performance goals in executive compensation over the past decade (Core and Packard 2016; Bettis et al. 2016), our study adds to the literature by providing new evidence of strategic disclosure under an important compensation design feature (i.e., the use of stock-performance goals). The rest of the paper proceeds as follows. Section 2 reviews related literature and develops hypotheses. Section 3 explains key variable measurement and research design issues, with the sample and descriptive statistics presented in Section 4. Section 5 reports empirical results on the discretionary guidance behavior, and Section 6 provides results on the associated consequences. Additional tests are reported in Section 7. We conclude in Section Related Literature and Hypothesis Development following the grant date), although the contrast between the timely and untimely subsamples is insignificant, possibly due to the small sample size of the timely subsample. 6

9 The design of executive compensation contracts has significant implications for managers voluntary disclosure decisions. Prior research demonstrates that certain features of compensation contracts can mitigate agency problems and thus improve voluntary disclosure quality. For instance, equity-based compensation, by aligning incentives between managers and shareholders, can improve the quality of management earnings guidance (Nagar et al. 2003). Moreover, managers are more forthcoming with bad news guidance when they expect to receive severance pay (Baginski et al. 2017) and face longer pay vesting periods (Cheng et al. 2014). Nevertheless, prior studies also document that certain features of compensation contracts can induce managers discretionary guidance behavior for the purpose of inflating payouts. For instance, Aboody and Kasznik (2000) find that managers strategically time earnings guidance around option grant dates (i.e., accelerate bad news and postpone good news) to dampen grant date price and boost the value of granted stock options. More recently, Martin, Seo and Yang (2015) speculate that the compensation committee relies on the prevailing market expectation to set performance goals, which motivates managers to issue pessimistic earnings guidance prior to committee meetings to build excess slack in performance goals. Extending this line of research, we focus on stock-performance goals and examine their implications for voluntary disclosure decisions. Ex ante, managers have strong incentives to achieve stock-performance goals, as failing to meet these goals often results in significant monetary losses in executive pay. In our sample, executive pay linked to stock performance accounts for 34% of total annual compensation, and managers would receive zero payout if actual performance falls below the lowest performance goal. Further, corporate boards may implicitly consider the achievement of performance goals in evaluating managerial ability, and top executives are more likely to be fired if they fail to achieve these goals (Bennett et al. 2017). Hence, 7

10 both monetary incentive and career concern can incentivize managers to exert discretion for the purpose of delivering the stock performance required to achieve stock-performance goals. Figure 1 visually illustrates possible discretionary activities among firms using stock-performance goals. Specifically, we observe a kink surrounding stock-performance goals at the end of the evaluation period, consistent with managers deliberate efforts to achieve stock-performance goals. 5 We propose that managers strategically release or withhold earnings guidance near the end of the evaluation period to increase the chance of achieving stock-performance goals. As the evaluation period approaches the end, managers can better estimate the gap between performance goals and actual performance, which facilitates the planning of discretionary activities. Voluntary disclosure, in particular management earnings guidance, serves as an effective and convenient tool to influence stock performance. Prior evidence shows that management earnings guidance significantly affects market expectations and stock prices (see Hirst et al for literature review). More importantly, managers can exert significant discretion in the timing and the news content of earnings guidance (e.g., Cheng and Lo 2006; Kothari et al.2009; Baginski et al. 2017), partly due to the protection from Safe Harbor regulations. In contrast, managers have limited flexibility to exert discretion in operating and financing decisions near the evaluation period end (e.g., Barth et al.2017). Although managers can manipulate earnings reporting through last-minute adjustments, the last quarter s earnings are usually released after the end of the evaluation period and thus are irrelevant to achieving stock-performance goals. We therefore hypothesize that managers strategically withhold (accelerate) bad (good) news earnings guidance near the end of 5 Interestingly, this kink does not exist at the beginning of the last quarter of the evaluation period. The fact that this kink suddenly arises at the end of the evaluation period highlights managers deliberate efforts to achieve stockperformance goals. 8

11 the performance evaluation period, when doing so can significantly increase the likelihood of achieving stock-performance goals. For managers discretionary guidance to be effective in facilitating the achievement of performance goals, the prevailing stock performance should not fall considerably below the performance goals. We identify four performance zones based on the gap between the prevailing stock performance (measured from the beginning of evaluation period to the beginning of the last quarter in the evaluation period) and the stock performance goal, as follows: 1) Way-Below : prevailing performance is way below performance goal so that discretionary guidance is unlikely to significantly enhance the likelihood of achieving the performance goal. 2) Right-Below : prevailing performance is marginally below performance goal so that discretionary guidance is likely to significantly enhance the likelihood of achieving the performance goal. 3) Above : prevailing performance is above performance goal so that discretionary guidance is unlikely to significantly enhance the likelihood of achieving the performance goal. 4) Way-Above : prevailing performance is way above performance goal so that managers lack the incentive to further increase stock performance. In empirical analyses, we focus on the lowest stock-performance goal that determines whether or not an executive receives a non-zero payout. In executive compensation contracts, the lowest goal is often specified as threshold and sometimes as target or simply benchmark in the absence of threshold. While managers often face multiple performance goals (including threshold, target, and maximum goals), we focus on the lowest performance goal because failing to meet these goals typically results in large monetary losses and causes significant concern over 9

12 managerial ability, thus inducing the strongest incentive for managers to achieve the lowest performance goal (See Figure 3, Zone 1). The above reasoning leads to the following hypothesis, stated in the alternative form: H1: Managers are more likely to withhold (accelerate) bad (good) news near the end of the performance evaluation period when the prevailing stock performance is marginally below the lowest performance goal compared to those in other performance zones. In addition to strategically timing the release of earnings guidance, managers may intentionally introduce optimistic biases into earnings guidance to further boost stock performance. Facing an uncertain environment and protection from "Safe Harbor" regulations, managers can bias earnings guidance to a certain extent, without worrying about being detected and suffering considerable costs. This reasoning leads to the following hypothesis: H2: Management earnings guidance contain more optimistic biases near the end of the performance evaluation period when the prevailing stock performance is marginally below the lowest performance goal compared to those in other performance zones. We note that the discretionary guidance behavior described above may involve significant costs. First, since 2006, the SEC requires public firms to explicitly disclose performance goals in proxy statements, although some firms defer the detailed disclosure until after the realization of stock performance. It is possible that investors partly anticipate discretionary guidance behavior from firms with Right-Below performance, making earnings guidance a less effective tool to boost stock performance and weakening managers incentive to withhold (accelerate) bad (good) news guidance. Second, alternative information sources, such as peer firms disclosures and analysts reports, may reveal the news content in released or withheld earnings guidance, weakening managers incentives to time earnings guidance for the purpose of achieving performance goals. Third, while withholding (accelerating) bad (good) news may improve stock performance, such behavior increases the risk of missing earnings expectations and potentially harms managerial 10

13 reputation (e.g., Richardson et al. 2004). Lastly, untimely or biased guidance potentially exposes managers to greater litigation risk and reputation loss in the long run. We therefore expect that firms with Right-Below performance are more likely to accelerate (withhold) good (bad) news when firms operate in a less litigious environment, when external monitoring is relatively weak, when discretionary guidance behavior is less likely to be detected in the market, and when the involved payouts have larger fair values. We also examine specific design features in stock-based performance evaluation (such as roller-over and interpolation options) and their implications for managers discretionary guidance behavior. 3. Research Design 3.1 Classification of performance zones Whether discretionary issuance of earnings guidance can facilitate managers to achieve stock-performance goals depends upon the deviation between the prevailing stock performance and the performance goal. We measure the prevailing stock performance (TESTRETURN) based on a firm s realized stock performance from the beginning of the evaluation period to the beginning of the last quarter of the evaluation period. We measure performance goal based on the lowest performance goal specified in the proxy statement (TARGETRETURN), defined as the threshold performance or the target performance (in the absence of threshold) in the pay-for-performance scheme. We also measure the maximum performance (MAXRETURN) as the highest stock performance that results in the maximum payout. We then classify each observation into one of the four performance zones based on the relative magnitude of TESTRETURN, TARGETRETURN, and MAXRETURN. Stock-performance goals are specified in either absolute or relative term. Under absolute performance evaluation, performance goals are expressed as a pre-specified level of returns or 11

14 share prices (which are converted into returns in the empirical analysis). 6 Under relative performance evaluation, performance goals are expressed as relative ranks or percentiles against a pre-specified peer group. 7 Hence, we calculate TARGETRETURN (MAXRETURN) as the returns corresponding to the threshold or target (maximum) rank/percentile among the peer group at the beginning of the last quarter of the evaluation period. At the beginning of the final quarter of the performance evaluation period, we classify each grant into one of the four performance zones: WAY-BELOW, RIGHT-BELOW, ABOVE, and WAY- ABOVE, as depicted in Figure 2. Our primary interest is in the RIGHT-BELOW zone, where the realized performance (as of one quarter prior to the evaluation period end) falls below the lowest performance goal, yet close enough to the goal so that discretionary disclosures can potentially help eliminate the gap. We use the market reaction to management guidance (MFRETURN) to gauge the extent to which management guidance can boost stock prices and eliminate the performance gap. Specifically, MFRETURN is defined as the average absolute returns over the [- 2, 2] trading-day window around management earnings forecasts issued in the previous four quarters. When TESTRETURN is lower than TARGETRETURN but the gap is smaller than MFRETURN, we classify the performance zone as RIGHT-BELOW zone. Likewise, the WAY- BELOW zone requires TESTRETURN below TARGETRETURN by more than MFRETURN. 6 Absolute performance goals are expressed in either stock returns or stock prices. From reading proxy disclosures, we note that return-based performance goals are always evaluated at the end of a pre-specified period (ranging from one year to three or more years), but price-based performance goals are sometimes set up as reaching target price for a certain number of days consecutively anytime during the evaluation period. It is possible that target price has already been reached before the beginning of the last quarter of the evaluation period, and thus stock price movements during the last quarter become irrelevant to the ultimate payouts, eliminating any incentive for discretionary voluntary disclosure over the last quarter. In untabluated test, we re-estimate Equation (1) by excluding sample firms that use price-based performance goals and observe similar findings as those reported in Panel B of Table 3. 7 Comparing methods include rank, percentile and percentage points above the benchmark peer performance. About 68% of our sample firms use relative performance goals (some of these firms also use absolute performance goals) in stock-based performance evaluation. 12

15 Moreover, the ABOVE zone requires that TESTRETURN is above TARGETRETURN but below MAXRETURN, and the WAY-ABOVE zone requires TESTRETURN above MAXRETURN. 3.2 Measurement of management earnings guidance behavior We examine two aspects of managers earnings guidance behavior, i.e., the likelihood of issuing good versus bad news guidance and the bias in issued guidance. We define the nature of news by comparing management forecasts (point forecasts or the mid-point of range forecasts) with the prevailing analysts consensus forecasts. Good (Bad) news management guidance walks up (down) market expectation of future earnings and is likely to trigger positive (negative) price movements. To capture the likelihood of issuing good versus bad news guidance, we construct two variables. First, we count the number of good or bad news EPS forecasts issued during the last three months prior to the evaluation period end (G_NEWS (#) or B_NEWS (#), respectively). Second, we construct the percentage of good or bad news EPS forecasts issued during the last three months prior to the performance evaluation period end (G_NEWS (%) or B_NEWS (%), respectively). We include confirming or neutral guidance in the total number of earnings guidance issued when constructing G_NEWS (%) or B_NEWS (%). Although managers can better judge the effectiveness of discretionary disclosure in facilitating the achievement of performance goals when approaching the end of the evaluation period, in reality firms rarely issue earnings guidance near the evaluation period end (in our sample only about 10 percent of firms issued earnings guidance during the last month of any quarter). As most earnings guidance is issued concurrently with previous quarter s earnings release, we choose the three-month period to capture the most active guidance window. In untabulated test, we find qualitatively similar results using management earnings guidance issued during the last month of the evaluation period to construct dependent variables. 13

16 To capture biases in management earnings guidance, we construct the average errors in management EPS forecasts issued during the last three months prior to the end of the performance evaluation period. Management forecast error is calculated as management EPS forecasts (point forecasts or the mid-point of range forecasts) minus I/B/E/S reported actual EPS, scaled by the actual EPS Control for the arrival of earnings news Observed good or bad news earnings guidance is jointly determined by the natural arrival of favorable or unfavorable news and managers discretion over the timing of issuance. In the empirical analysis, we employ five approaches to isolate managerial discretion (over the issuance timing) from the arrival of earnings news. First, we control for firm characteristics that potentially affect the arrival of earnings news, including accounting performance and stock performance, because presumably the arrival of earnings news is correlated with the concurrent performance. Second, we compare the RIGHT-BELOW group with not only the ABOVE and WAY- ABOVE groups, but also with the BELOW group. If RIGHT-BELOW firms guidance behavior is driven by managers incentive to achieve performance goals, RIGHT-BELOW firms are expected to issue more (less) good (bad) news guidance than all the rest of firms (including ABOVE, WAY- ABOVE, and BELOW groups). However, if the arrival of favorable news is positively correlated with the prevailing performance relative to performance goals, we would observe RIGHT-BELOW firms issue more (less) good (bad) news guidance compared than BELOW firms, but less (more) good (bad) news than ABOVE and WAY-ABOVE firms. Third, in cross-sectional analyses, we examine whether the observed issuance behavior for RIGHT-BELOW firms varies with the incentive and costs associated with achieving stock- 8 Results are qualitatively similar if we construct forecast bias based on the last management forecast issued before the end of the evaluation period. 14

17 performance goals. To the extent that the arrival of earnings news is independent of these incentive and costs, evidence from cross-sectional analyses would strengthen our proposition that managerial incentive to achieve stock-performance goals, rather than the arrival of earnings news, contributes to the observed guidance behavior for the RIGHT-BELOW group. 9 Fourth, we examine stock performance and earnings guidance behavior shortly after the end of evaluation period. While we do not expect the arrival of good news to be naturally followed by the arrival of bad news, managers discretion in accelerating (delaying) good (bad) news would lead to a reversal in issued earnings guidance and stock performance after the evaluation period ends. Lastly, we conduct a falsification test based on firms using only accounting-performance goals in setting executive pay. The arrival of performance news should correlate with both accounting performance and stock performance, but accelerating (delaying) good (bad) news guidance should only affect the likelihood of achieving stock-performance goals as opposed to accounting-performance goals, because the latter is determined by realized earnings rather than forecasted earnings. Thus, we do not expect similar findings on RIGHT-BELOW firms if the underlying performance metrics are accounting numbers. 4. The Sample and Descriptive Statistics 4.1 Sample selection We collect stock-performance goals and other detailed features of executive compensation contracts from the IncentiveLab database of Institutional Shareholder Services (ISS), which covers 750 largest firms by market capitalization during our sample period. We obtain management 9 Kothari et al. (2009) measure good versus bad news guidance based on market reactions to earnings guidance. We choose not to follow their approach because the market might partly discern managers incentive to achieve stockperformance goals and thus stock prices may not correctly capture managers discretionary guidance behavior. 15

18 earnings guidance from I/B/E/S Guidance database, and use CRSP monthly files to construct stock returns and stock return volatility. Firms financial information, institutional ownership, and analyst consensus forecasts are collected from COMPUSTAT, 13F filings provided by Thompson Reuters, and I/B/ES, respectively. Our sample period starts in 2006 and ends in Since 2006, the SEC mandates detailed disclosures of performance goals in executive compensation contracts. We start with firms at the intersection of the IncentiveLab, COMPUSTAT and CRSP databases with available target and test period return, yielding 3,303 stock-performance based grants. Following Bennett et al. 2017, we restrict the sample to grants to CEOs because CEOs are the most influential corporate decision makers. We further delete firms that did not provide any earnings guidance over the past two years, since these firms likely adopt a no guidance policy and hence discretionary guidance is largely irrelevant to these firms. After requiring non-missing control variables, we obtain a final sample of 1,196 firm-year-grant observations. 10 Table 1 summarizes the sample selection procedures. 4.2 Descriptive statistics In Table 2, Panel A presents summary statistics for the variables used in the empirical analyses. To mitigate the influence of outliers, we winsorize all continuous variables at 1 and 99 percentiles. Our sample firms are on average large, profitable, with significant institutional ownership and analyst coverage. Panel B presents summary statistics across the four performance zones. In general, accounting performance is better and market capitalization is larger when the prevailing stock performance is higher above the performance goals. As illustrated in Panel C, a majority of firms 10 In the final sample, most firm-years have one single grant that uses stock-performance goals. We conduct a robustness test by only retaining the largest grant for firm-years with two or more grants using stock-performance goals. We obtain qualitatively similarresults. 16

19 stay in the same performance zone from year t to year t+1 (ranging from 60.9% to 71.1%), except for the RIGHT-BELOW zone (17.5%). This evidence suggests that RIGHT-BELOW performance is highly transitory, making it difficult for investors to ex-ante identify suspicious firms and anticipate managerial discretion in earnings guidance issuance. We also note that about 19.37% of unique firms have experienced RIGHT-BELOW performance during the sample period, even though only 6% (= 77/1,196) of observations actually falling into the RIGHT-BELOW zone. 11 Panel D reports Pearson correlations. The four indicators for various performance zones are correlated with several firm characteristics, which justifies a multivariate analysis in subsequent tests. 5. Empirical Results 5.1 Univariate results on management earnings guidance Panel A of Table 3 reports the relative frequency of good versus bad news management earnings guidance across the four performance zones. Firms falling in the RIGHT-BELOW zone have the tendency to issue less (more) unfavorable (favorable) earnings guidance, as evidenced by the lowest (highest) number and percentage of bad news (good news) guidance being issued during the last quarter of the performance evaluation period. The differences in the number and percentage of bad/good news issuance between the RIGHT-BELOW group and the other three groups are mostly significant (i.e., significant in 9 out of 12 comparisons). In particular, in the last quarter of the performance evaluation period, the percentage of bad (good) news management guidance is 40.9 percent (58.72%) for firms with RIGHT-BELOW performance, but greater (less) than 50% for all the other three groups of firms. Interestingly, the WAY-BELOW group does not provide 11 By construction, the RIGHT-BELOW zone has fewer observations compared with the other three zones, because the gap between the prevailing performance and performance goal is required to be smaller than the 5-day market reaction to management earnings guidance. 17

20 more bad news guidance than the ABOVE and WAY-ABOVE groups, suggesting that poor stock performance (relative to performance goals) does not explain the nature of management guidance (i.e., bad news or good news) in the coming quarter. Panel A also shows the average bias in management earnings guidance for the four groups of firms. In the last quarter of the performance evaluation period, the average management forecast errors from RIGHT-BELOW and WAY-ABOVE firms are positive, while those from firms in the WAY-BELOW and the ABOVE groups are negative. However, the differences in the mean forecast errors between the RIGHT-BELOW group and each of the other three groups are not statistically significant. 5.2 Multivariate results on the issuance of management earnings guidance We estimate the following regression model to test H1, with subscripts i, j, and t denoting firm i, grant j, and disclosure period t. B_NEWS(G_NEWS) i,j,t = β 0 + β 1 WAY_BELOW i,j,t + β 2 RIGHT_BELOW i,j,t + β 3 ABOVE i,j,t + X i,j,t + γ ind + τ t + ε i,j,t (1) The dependent variable (B_NEWS or G_NEWS) is the number or the percentage of good or bad news issued by firm i in the last quarter of year t. We use ordinary least squares (OLS) estimation when the percentage of bad or good news guidance is the dependent variable, and Poisson regression estimation when the number of bad or good news guidance is the dependent variable. WAY_BELOW, RIGHT_BELOW, and ABOVE are indicator variables as defined in section 3.1. Since we omit WAY-ABOVE in the model, the coefficients on these indicator variables capture the incremental differences (in managers guidance behavior) compared with the WAY- ABOVE group. Based on H1 (i.e., firms with RIGHT-BELOW performance have the strongest incentive to delay bad news guidance and accelerate good news guidance), we predict β 2 < 0, β 2 18

21 < β 1, and β 2 < β 3 when B_NEWS is the dependent variable, and β 2 > 0, β 2 > β 1, and β 2 > β 3 when G_NEWS is the dependent variable. We control for various firm characteristics that potentially correlate with managers earnings guidance behavior and their incentives to achieve performance goals. First, we include concurrent accounting performance (ROA and ROA), stock performance (TESTRETURN), and earnings surprise (EARNSURP), because firm performance and earnings news may correlate with news arrival and affect the issuance of management guidance. Moreover, as RIGHT-BELOW is defined based on stock performance realized before the previous quarter s earnings release, including EARNSURP can control for potential measurement errors due to the omission of price reaction to EARNSURP in measuring RIGHT-BELOW. 12 Second, we include several proxies for a firm s general information environment and external monitoring strength (i.e., SIZE, ANALYST, INSTOWN, and INSTBLOCK), because larger, well-followed, and better-monitored firms tend to issue more earnings guidance (e.g., Balakrishnan et al. 2014, Ajinkya et al. 2005; Karamanou and Vafeas 2005). Third, we control for litigation risk (LITIGATION) because higher litigation risk may induce more timely disclosure of bad news or warnings (Verrecchia 1983; Skinner 1994, 1997). Fourth, prior studies show that leverage and growth opportunities are associated with greater disclosure (Bamber and Cheon 1998; Lang and Lundholm 1993; Frankel et al. 1999). We thus include debt-to-equity ratio (LEVERAGE) and the book-to-market ratio (BTM). Further, we include liquidity (LIQUIDITY) to control for its association with the prevailing stock performance (TESTRETURN) and voluntary disclosure (Diamond and Verrecchia 1991). Fifth, we include the grant date fair value of the award relative to the executive s total compensation (GRANTSIZE), as 12 Because most management guidance is issued concurrently with previous quarter s earnings release, it is challenging to accurately separate market reaction to earnings surprise from market reaction to news in earnings guidance. We therefore directly include earnings surprise as a control variable. 19

22 greater grant value induces greater incentive to inflate stock price. We also include an indicator variable (OPTIONGRANT) to identify the incidence of option grants in the quarter following the evaluation period end, to account for the incentive to deflate stock price prior to option grant dates (Aboody and Kaznik 2000). Finally, we include working capital accruals (WCACCQ) for the most recent quarter to account for the possibility that managers may manipulate earnings reporting to achieve stock-performance goals. To control for cross-industry and inter-temporal differences in managers earnings guidance behavior, we include industry- and year-fixed effects in the estimation. Panel B of Table 3 reports the multivariate regression results. Columns (1) and (2) present OLS results, where the dependent variable is the percentage of bad or good news guidance, and Columns (3) and (4) present Poisson regression results, where the dependent variable is the number of bad or good news guidance. 13 The OLS and Poisson estimation results generate similar inferences, therefore we focus on OLS estimation results to interpret the findings. As shown under Columns (1) and (2), the coefficients on RIGHT_BELOW are significantly negative and positive, respectively, suggesting that firms with prevailing stock performance right below the performance goal tend to release less bad news and more good news near the end of the evaluation period, compared with firms with prevailing stock performance way above the performance goal. The coefficients on ABOVE and WAY_BELOW are both insignificant, suggesting that WAY-ABOVE group demonstrates similar disclosure behavior as ABOVE and WAY_BELOW groups, despite the potential incentive to dampen stock performance to reserve for future periods. To assess the economic significance, we evaluate the OLS regression results at the sample mean (untabulated). We find that during the last quarter of the evaluation period, the percentage 13 The sample for the OLS regression analysis is smaller than that for the Poisson regression analysis, because we need a non-zero deflator to measure the percentage of guidance news. 20

23 of bad news guidance in the WAY_BELOW, ABOVE, and WAY_ABOVE groups is all above 50 percent (i.e., 51.3 percent, percent and percent respectively), but this percentage is significantly lower at percent for the RIGHT-BELOW group. In contrast, the percentage of good news guidance is percent for the RIGHT-BELOW group, and this percentage is percent, percent, and percent for the WAY_BELOW, ABOVE and WAY_ABOVE groups, respectively. Tests of the coefficient differences between RIGHT-BELOW group and the other groups all yield significance at less than 1 percent to 5 percent levels. In contrast to the RIGHT- BELOW group, we do not observe a significant difference in the nature of earnings guidance between the WAY-BELOW group and the ABOVE (or WAY-ABOVE) group, suggesting that poor performance is unlikely to be the sole driver of the distinctive guidance behavior in the RIGHT- BELOW group. Turning to control variables, we find that changes in accounting performance and most recent earnings surprise are both negatively (positively) related to the percentage of bad (good) news guidance, while stock performance is unrelated to the nature of news in management guidance. Litigation risk is negatively (positively) associated with the percentage of bad (good) news guidance, consistent with the notion that litigation concern hampers bad news disclosure (e.g., Francis et al. 1994). We also find that analyst coverage is positively (negatively) related to the percentage of bad (good) news guidance, but the number of blockholders is negatively (positively) related to the percentage of bad (good) news guidance. Working capital accruals for the most recent quarter, as a rough proxy for earnings management, are positively (negatively) related to the percentage of bad (good) news guidance. 5.3 Multivariate results on the bias in management earnings guidance 21

24 We estimate the following regression model to test H2, with subscripts i, j, and t denoting firm i, grant j, and disclosure period t. MFE i,j,t = β 0 + β 1 WAY_BELOW i,j,t + β 2 RIGHT_BELOW i,j,t + β 3 ABOVE i,j,t + X i,j,t + γ ind + τ t + ε i,j,t (2) The dependent variable (MFE) is the average bias in management earnings guidance by firm i in the last quarter of year t. WAY_BELOW, RIGHT_BELOW, and ABOVE are indicator variables as defined in section 3.1. Same as equation (1), the coefficients on these indicator variables capture the incremental differences (in managers guidance behavior) from the WAY- ABOVE group. Based on H2 (i.e., firms with RIGHT-BELOW performance have the strongest incentive to introduce optimistic biases in management guidance), we predict β 2 > 0, β 2 > β 1, and β 2 > β 3. Table 4 reports the estimation results. We find that the coefficient on RIGHT-BELOW is positive (0.0109), yet statistically insignificant. Combined with the findings in Table 3, the results suggest that the RIGHT-BELOW group strategically withholds negative news to avoid depressing stock performance, but does not impose more optimistic biases on earnings guidance than the other groups. Hence, the strategic timing of favorable versus unfavorable earnings guidance is a distinctive behavior from biasing earnings guidance upwards. 5.4 Cross-sectional tests In Table 5, we examine cross-sectional variations in the earnings guidance behavior along with various legal, informational, and compensation design factors that affect managerial incentive and costs behind the discretionary guidance behavior. For brevity, we only present cross-sectional results with the percentage of bad news as the dependent variable. Inferences remain qualitatively similar when we use the percentage of good news or the number of bad/good news as the dependent variable. We present the cross-sectional tests in the following two subsections. 22

25 Costs of discretionary guidance behavior and its likelihood of being discovered First, prior literature suggests that withholding bad news may lead to sharp decline in subsequent share prices and expose managers to significant litigation risk (e.g., Francis et al. 1994; Skinner 1994, 1997). We thus expect managers to have weaker incentive to withhold bad news in a highly litigious environment. Consistently, Panel A shows a statistically stronger (weaker) relation between RIGHT-BELOW and B_NEWS(%) when firms operate in litigious (non-litigious) industries. Second, a number of studies (e.g., Ajinkya et al. 2005; Karamanou and Vafeas 2005; Balakrishnan et al. 2014) have shown that financial analysts and institutional investors can effectively monitor managers financial disclosure decisions and thus curb their opportunistic disclosure behavior. We thus expect to find a weaker relation between RIGHT-BELOW and B_NEWS(%) when firms have relatively high concentration of institutional ownership (such as institutional block holders) and larger analyst coverage. Panels B and C confirms our expectations. We find that the negative relation between RIGHT-BELOW and B_NEWS(%) is statistically significant in the lowest terciles of institutional block holding and analyst following, while this relation is insignificant in the highest terciles of institutional block holding and analyst following. Third, to the extent that managers discretionary behavior might be detected by the market, managers are unable to inflate stock performance and may face considerable reputation costs. Rogers and Stocken (2005) suggest that high return volatility is associated with the market s difficulty in detecting managers misrepresentation in earnings guidance. Thus, we predict that managers have weaker incentive to strategically time earnings guidance when return volatility is relatively low. Results shown in Panel D are consistent with this prediction, as evidenced by a 23

26 significantly more negative coefficient on RIGHT-BELOW for the highest tercile of return volatility than the lowest tercile. Fourth, managers are less likely to accelerate good news or postpone bad news when concurrent investor sentiment is excessively high, because doing so may lead to even more optimistically biased market expectation and higher potential legal costs. In contrast, accelerating good news or postponing bad news is more feasible given pessimistic market expectation. Supporting this proposition, Panel E shows that the negative relation between RIGHT-BELOW and B_NEWS(%) is significant only when market expectations (proxied by the prevailing analyst consensus forecasts) are pessimistic, and this relation is insignificant when market expectations are optimistic Monetary benefits from discretionary guidance behavior Managers incentives to achieve performance goals via discretionary earnings guidance may also vary with the related monetary benefits and the design features of compensation grants. First, managers are likely to have stronger incentive to boost or support stock performance when they can receive larger payouts from achieving performance goals. Thus, we expect to find a stronger negative relation between RIGHT-BELOW and B_NEWS(%) when the fair value of the related target payouts is relatively high. Consistently, Panel F reports that the negative association between RIGHT-BELOW and B_NEWS(%) is statistically significant when fair values of the grants are relatively large, but insignificant when related payouts are relatively low. Second, we explore several features of compensation arrangements that may affect managerial incentives to achieve performance goals. About 10% of compensation plans allow managers to rollover the grants to the next year if they do not meet the performance goals in the current year. All else equal, managers should have stronger incentives to achieve performance 24

27 goals in the absence of the rollover option. Another relevant feature is whether interpolation is allowed above the lowest performance goal. Since interpolation allows a continuous relation between payouts and stock performance, managers have incentive to deliver better stock performance even after they meet the lowest performance goal. Findings are reported in Panels G and H. The relation between RIGHT-BELOW and B_NEWS(%) is significantly negative only when the compensation plan is not allowed to be rolled over from the current period to the future. Moreover, although RIGHT-BELOW firms appear to withhold (accelerate) bad (good) news guidance regardless of interpolation, they exhibit such behavior to a greater extent when interpolation is allowed above the lowest performance goal and hence exceeding the lowest goal provides potentially higher payoff. Interestingly, the ABOVE group withholds bad news only when there is interpolation above the lowest performance goal. When interpolation is not allowed, the ABOVE group releases more bad news relative to the other groups. Given that ABOVE firms have already passed the lowest performance goal and achieving higher stock performance provides incremental benefits only when there is interpolation, the evidence on the ABOVE group provides further support to our proposition that managers strategically time earnings guidance (in particular, withholding bad news and accelerating good news) for the purpose of achieving performance goals and reaping greater compensation payouts. 6. Consequences of Discretionary Guidance Behavior In this section, we explore various consequences resulting from managers discretionary guidance behavior. Our findings so far are consistent with managers incentive to achieve stockperformance goals inducing discretionary guidance behavior. Such evidence implies that the market cannot completely unravel managers discretionary disclosure behavior. As a result, we expect elevated market expectation regarding future earnings and inflated stock performance 25

28 towards the end of the evaluation period. Consequently, we expected that firms with Right-Below performance are more likely to achieve stock-performance goals. Furthermore, we expect a reversal in management guidance behavior and stock performance shortly after the end of evaluation period. 6.1 Analysts expectation at the end of the evaluation period The underlying premise in H1 is that managers are able to affect market expectation through discretionary guidance. To validate this premise, we analyze the relation between RIGHT- BELOW and analysts consensus forecasts at the end of the performance evaluation period. In Panel A of Table 6, the univariate analysis shows that analysts consensus forecasts are on average pessimistic for the WAY-BELOW, ABOVE, and WAY-ABOVE groups, which is consistent with prior evidence that analysts forecasts are more pessimistic near the end of the year so that managers are more likely to meet or beat the market expectation (e.g., Brown et al. 1985; Richardson et al. 2004). In contrast, for the RIGHT-BELOW group, we observe that analysts consensus forecasts (for both annual and quarterly earnings) are slightly optimistic. In Panel B of Table 6, we conduct a multivariate regression analysis, after controlling for firm size, book-to-market ratio, and most recent earnings news. The coefficient on RIGHT_BELOW is positive and significant when regressed on analysts quarterly or annual consensus forecasts. Hence, compared with the WAY-ABOVE group, the RIGHT-BELOW group has significantly more favorable market expectation as reflected in analysts consensus forecasts. We observe similar evidence when we compare the RIGHT-BELOW group with the WAY-BELOW group or the ABOVE group. These findings are consistent with inflated market expectation induced by discretionary guidance behavior among the RIGHT-BELOW group. 6.2 Stock performance during and after the last quarter of the evaluation period 26

29 To validate that managers discretionary guidance behavior indeed boosts stock performance, we examine stock returns over the last quarter of the performance evaluation period, as well as stock returns subsequent to the evaluation period, for the RIGHT-BELOW group. To the extent that investors cannot fully unravel managers strategic disclosure choices (Verrecchia 1983; Rogers and Stocken 2005), we expect to observe higher returns in the last quarter of the evaluation period and return reversals after the evaluation period for the RIGHT-BELOW group. Table 7 shows market-adjusted stock returns across the four groups of firms. 14 In Panel A, we find the RIGHT-BELOW group exhibits the highest stock returns during the last quarter of the evaluation period (3.28%), compared with the other groups (ranging between -0.23% and 2.34%). After the evaluation period, we observe some evidence of return reversal for the RIGHT-BELOW group, as evidenced by negative stock returns in the subsequent one-month window (-0.13%) and the three-day window around the forthcoming earnings announcements (-0.53%), whereas the other three groups experience either a continuation of positive returns or negative returns of a considerably smaller magnitude. In Panel B, the multivariate regression results are consistent with the univariate results in Panel A, after controlling for common risk factors that affect stock returns. 6.3 Management earnings guidance after the evaluation period To the extent that managers of RIGHT-BELOW firms withhold bad news near the end of performance evaluation period, the amount of bad news may accumulate above the threshold that managers can conceal, especially when actual earnings announcements approach (e.g., Kothari et al. 2009). At the same time, managers may have less good news to release since some have been accelerated to the previous period. 15 Thus, we expect that the RIGHT-BELOW group is more likely 14 Inferences remain the same if we use size-adjusted returns instead of market-adjusted returns, and buy-and-hold returns instead of cumulative returns. 15 Indeed, the number of firms that subsequently issue good news guidance is much smaller than the number of firms that subsequently issue bad news guidance. Hence, we focus on bad news guidance in addressing this issue. 27

30 to release bad news guidance after the end of the evaluation period and up to the forthcoming earnings announcements. Table 8 presents the regression results on the relation between RIGHT-BELOW and the incidence of bad news guidance released during the window between the evaluation period end and the forthcoming earnings announcements. 16 We include the same set of control variables as Equation (1). As expected, we find that the RIGHT-BELOW group is more likely to release bad news guidance than the rest of three groups. Such reversal behavior in bad news guidance triangulates the findings on managers discretionary guidance prior to the end of performance evaluation period. 6.4 The likelihood of achieving stock-performance goals We argue that managers strategically time earnings guidance (by withholding bad news and accelerating good news) to achieve stock-performance goals. To ex-post validate this argument, we examine whether such strategic guidance behavior indeed helps managers to achieve performance goals. For each firm in the RIGHT-BELOW group (i.e., the test sample), we match with a control firm that also uses stock-performance goals but did not provide any management earnings guidance in the past two years (i.e., the control sample). To facilitate the comparison, we require each test firm and its matched control firm to have the closest performance gap between prevailing stock performance and the lowest performance goal. Because these control firms have followed a no guidance history, they are less likely to rely on discretionary disclosure to achieve stock-performance goals. 16 As our sample firms infrequently provide management guidance during the window between the fiscal year end and the forthcoming earnings announcement, we do not use the number or the percentage of bad news guidance as the dependent variable. 28

31 As reported in Panel A of Table 9, the performance gap is statistically indifferent between the test sample and the matched control sample, suggesting that our matching procedures are effective. In Panel B of Table 9, we compare the likelihood of achieving the lowest performance goal between the test sample (i.e., the RIGHT-BELOW group) and the matched control sample. As expected, firms in the RIGHT-BELOW group are significantly more likely to achieve the lowest performance goals than the matched control sample. 7. Supplemental Analyses 7.1 Alternative classification of incentive zones In the main analysis, we identify the RIGHT-BELOW group based on the closeness of the prevailing performance to the lowest performance goal, which is either the performance threshold or the performance target in the absence of a threshold. We acknowledge that managers may have incentives to deliver higher performance even after meeting the lowest performance goal, especially when interpolation is allowed in setting payouts. Consistently, Panel H of Table 4 shows that the ABOVE group also exhibits discretionary guidance behavior under interpolation. As an alternative, we broaden the definitions of incentive zones by incorporating higher performance goals (such as the target and/or the maximum) whenever they are available. Figure 3 illustrates this alternative definition. Specifically, we separately analyze the performance zones that is right below the lowest performance goal (denoted G1), the next performance goal which is usually the target (denoted G2), and the highest performance goal which is usually the maximum (denoted G3). Table 10 reports the results based on the alternative definition of incentive zones. As shown in Panel A, the G1 zone has the lowest (highest) percentage of bad (good) news guidance over the last quarter of the performance evaluation period, compared with the G2 and G3 zones. Hence, the 29

32 lowest performance goals appear to provide the strongest incentives for managers to time earnings guidance. In Panel B, we estimate a multivariate regression using the alternative definitions of incentives zones and the same set of control variables as in Equation (1). Under Column (1), we find a significant negative coefficient on G1, consistent with our earlier findings. Under Column (2), we focus on firms with all three performance goals (threshold, target, and maximum). We find significantly negative coefficients on both G1 and G2, suggesting that the discretionary guidance behavior is not limited to the G1 group, but also exist in the G2 group. 7.2 Falsification test based on accounting-performance goals Our proposition about managers discretionary guidance behavior is largely irrelevant for accounting-performance goals, as accounting-performance goals are evaluated based on realized accounting numbers as opposed to projected accounting numbers (such as earnings guidance). As a result, we do not expect firms using accounting-performance goals to exhibit similar guidance behavior as our sample firms. To minimize measurement noises in comparing actual performance with performance goals, we focus on EPS goals, which represent the most common and clearly defined accounting-performance goals in executive compensation contracts. 17 In Table 11, we re-estimate Equation (1) based on a sample of compensation plans that use EPS instead of stock performance in specifying performance goals. We code RIGHT- BELOW_ACCT based on whether realized EPS (up to the beginning of the last quarter) is close to the lowest performance goal near the end of the evaluation period. Specifically, RIGHT- BELOW_ACCT is equal to 1 if a firm s cumulative EPS achievement up to the beginning of the last quarter is lower than the lowest EPS performance goal but the gap is smaller than the historical average quarterly EPS for respective quarter, and zero otherwise. If our findings on managers 17 Other common accounting-performance goals, such as operating earnings and return on invested capital, are often vaguely defined. 30

33 discretionary guidance behavior is indeed driven by their incentive to achieve stock-performance goals, then we do not expect to observe similar evidence among firms using EPS goals, because withholding (accelerating) bad (good) news guidance per se would not improve EPS performance. As expected, results show that firms do not appear to withhold bad news guidance or accelerate good news guidance when their EPS performance is close to the performance goals. 8. Conclusion We examine the implications of stock-performance goals, a prevalent feature in U.S. executive compensation contracts, for managers voluntary disclosure decisions. The empirical evidence suggests that managers strategically time earnings guidance, via withholding bad news and accelerating good news, near the end of the performance evaluation period, when doing so can significantly enhance the chance of achieving stock-performance goals. Such discretionary guidance behavior varies with environmental factors (such as litigation risk, external monitoring, detection risk, and market sentiment) and features of compensation grants (such as fair value, rollover, and interpolation options). The discretionary guidance behavior temporarily elevates market expectation of future earnings and stock performance at the end of evaluation periods, boosting the chance of achieving stock-performance goals. The overall evidence demonstrates a contract-driven incentive arising from stock-performance evaluation, complementing prior literature on accounting-based performance evaluation and broadening our understanding of the economic consequences of compensation design. 31

34 Reference: Aboody, D., and R. Kasznik CEO stock option awards and the timing of corporate voluntary disclosures. Journal of Accounting and Economics 29: Ajinkya, B. and M. J. Gift Corporate managers' earnings forecasts and symmetrical adjustments of market expectations. Journal of Accounting Research: Ajinkya, B., S. Bhojraj, and P. Sengupta The association between outside directors, institutional investors and the properties of management earnings forecasts. Journal of accounting Research, 43: Baginski, S. P., J. L. Campbell, L. A. Hinson, and D. S. Koo Do career concerns affect the delay of bad news disclosure. The Accounting Review, forthcoming. Baiman, S., and J. S. Demski Economically optimal performance evaluation and control systems. Journal of Accounting Research 18: Balakrishnan, K., M.B. Billings, B. Kelly, and A. Ljungqvist Shaping liquidity: On the causal effects of voluntary disclosure. the Journal of Finance 69: Bamber, L. S., and Y. S. Cheon Discretionary management earnings forecast disclosures: antecedents and outcomes associated with forecast venue and forecast specificity choices. Journal of Accounting Research 36: Barth, M.E., K.H. Gee, D. Israeli, and R. Kasznik Stock Price Management and Share Issuance: Evidence from Equity Warrants. Working paper. Bennett, B., J. C. Bettis, R. Gopalan, and T. Milbourn Compensation goals and firm performance. Journal of Financial Economics. Forthcoming. Bettis, C., J. Bizjak, J. Coles, and S. Kalpathy Performance-vesting provision in executive compensation. Working paper. Beyer, A., D. A. Cohen, T. Z. Lys, and B. R. Walther The financial reporting environment: review of the recent literature. Journal of Accounting and Economics 50: Brown, P., G. Foster, and E. Noreen Security analyst multi-year earnings forecasts and the capital market. Studies in Accounting Research. Cheng, Q., and K. Lo Insider trading and voluntary disclosures. Journal of Accounting Research 44: Cheng, Q., Y. J. Cho, and J. B. Kim Managers pay duration and voluntary disclosures. Singapore Management University School of Accountancy Research Paper

35 Core, J.E A review of the empirical disclosure literature: discussion. Journal of Accounting and Economics, 31: Core, J., and H. Packard Performance vesting provision and executive incentives. Working paper. Diamond, D.W., and R.E. Verrecchia Disclosure, liquidity, and the cost of capital. The Journal of Finance 46: Francis, J., D. Philbrick, and K. Schipper Shareholder litigation and corporate disclosures. Journal of Accounting Research 32: Frankel, R., Johnson, M., and D. J. Skinner An empirical examination of conference calls as a voluntary disclosure medium. Journal of Accounting Research, 37: Gaver, J. J., K. M. Gaver, and J. R. Austin Additional evidence on bonus plans and income management. Journal of Accounting and Economics 19: Gong, G., L. Y. Li, and H. Yin Relative performance evaluation and the timing of earnings release. Working paper. Healy, P The effect of bonus schemes on accounting decisions. Journal of Accounting and Economics 7: Healy, P., K. G. Palepu Information asymmetry, corporate disclosure, and the capital markets: a review of the empirical disclosure literature. Journal of Accounting and Economics 32: Hirst, D. E., L. Koonce, and S. Venkataraman Management earnings forecasts: a review and framework. Accounting Horizons 22: Holmstrom, B Moral hazard and observability. The Bell Journal of Economics 10: Holmstrom, B Moral hazard in teams. The Bell Journal of Economics 13: Holthausen, R. W., D. F. Larcker, and R. G. Sloan Annual bonus schemes and the manipulation of earnings. Journal of Accounting and Economics 19: Karamanou, I., and N. Vafeas The association between corporate boards, audit committees, and management earnings forecasts: an empirical analysis. Journal of Accounting Research 43: Kothari, S. P., S. Shu, and P. D. Wysocki Do managers withhold bad news? Journal of Accounting Research 47:

36 Lambert, R. A., and D. F. Larcker An analysis of the use of accounting and market measures of performance in executive compensation contracts. Journal of Accounting Research 25: Lang, M.H., and Lundholm, R. J Cross-sectional determinants of analyst ratings of corporate disclosures. Journal of Accounting Research, 31: Larcker, D. F., S. A. Richardson, and I. Tuna Corporate governance, accounting outcomes, and organizational performance. The Accounting Review 82: Leuz, C. and P. D. Wysocki The economics of disclosure and financial reporting regulation: Evidence and suggestions for future research. Journal of Accounting Research, 54(2), pp Martin, X., H. Seo, and J. Yang Performance target disclosure and management earnings guidance. Working paper. Nagar, V., D. Nanda, and P. Wysocki Discretionary disclosure and stock-based incentives. Journal of Accounting and Economics 34: Richardson, S., S. H. Teoh, and P. D. Wysocki The walk-down to beatable analyst forecasts: the role of equity issuance and insider trading incentives. Contemporary Accounting Research 21: Rogers, J. L., and P. C. Stocken Credibility of management forecasts. The Accounting Review 80: Skinner, D. J Why firms voluntarily disclose bad news. Journal of Accounting Research 32: Skinner, D. J Earnings disclosures and stockholder lawsuits. Journal of Accounting and Economics 23: Sloan, R. G Accounting earnings and top executive compensation. Journal of Accounting and Economics 16: Verrecchia, R. E Discretionary disclosure. Journal of Accounting Economics 5:

37 Variable Dependent variables B_NEWS (#) B_NEWS (%) BAD G_NEWS (%) G_NEWS (#) GOOD APPENDIX A: Variable Definition Definition Number of bad news management EPS forecasts issued during the last three months prior to the performance evaluation period end. Bad news is defined as negative forecast news, where forecast news is calculated as management EPS forecast minus the mean analyst consensus of all outstanding analyst forecasts not older than 90 days. Number of bad news management EPS forecasts divided by the total number of management EPS forecasts issued during the last three months prior to the performance evaluation period end. Indicator equal to 1 if one or more bad news managent EPS forecasts are issued during the last three months prior to the performance evaluation period end, 0 otherwise. Number of good news management EPS forecasts issued during the last three months prior to the performance evaluation period end. Good news is defined as positive forecast news, where forecast news is calculated as management EPS forecast minus the mean analyst consensus of all outstanding analyst forecasts not older than 90 days. Number of good news management EPS forecasts divided by the total number of management EPS forecasts issued during the last three months prior to the performance evaluation period end. Indicator equal to 1 if one or more good news managent EPS forecasts are issued during the last three months prior to the performance evaluation period end, 0 otherwise. MFE Management forecast bias, calculated as the mean of management forecast errors of all management EPS forecasts issued during the last three months prior to the performance evaluation period end. Management forecast error is calculated as management EPS forecast (the mid-point for range forecasts) minus I/B/E/S reported actual scaled by the actual Independent variables WAY-BELOW Indicator equal to 1 if TESTRETURN, < (TARGETRETURN MFRETURN), 0 otherwise where, TESTRETURN = Buy-and-hold return over the window starting from the beginning of the performance evaluation period and ending 3-months before the end of the performance evaluation period TARGETRETURN = Return to be achieved to earn the target payout. If threshold goal, as well as target goal, is given, threshold goal is used. MFRETURN = Past two-year average market reaction to firm i s management EPS forecasts, calculated as mean of absolute returns during [-2, 2] window around management earnings forecasts issued during the last two years. RIGHT-BELOW Indicator equal to 1 if (TARGETRETURN MFRETURN) <= TESTRETURN < TARGETRETURN, 0 otherwise. ABOVE Indicator equal to 1 if TARGETRETURN <= TESTRETURN < MAXRETURN, 0 otherwise, where MAXRETURN = Return to be achieved to earn the maximum payout WAY-ABOVE Indicator equal to 1 if TESTRETURN >= MAXRETURN, 0 otherwise. Other variables 35

38 AFE(Annual / Quarterly) ANALYST BTM EARNSURP GRANTSIZE HORIZON INSTBLOCK INSTOWN Analyst forecast error, calculated as the mean consensus analyst forecast of annual (quarterly) EPS of year t measured at the end of the performance evaluation period end minus the I/B/E/S reported actual EPS, scaled by the absolute value of the consensus. Number of analysts following firm i in year t, computed as the maximum number of analysts contributing to I/B/E/S consensus in year t. Book to market ratio calculated as book value of equity divided by market value of equity Earnings surprise at earnings announcements announced during the last quarter of year t, calculated as I/B/E/S reported actual EPS minus last analyst forecast consensus prior to the earnings announcement, scaled by the actual EPS. The fair value of the grant measured on the grant date, in dollars, divided by the lag total compensation. Horizon of management forecast, calculated as the number of days between forecast period end date and management forecast date of the longest horizon management forecast. The number of institutional block holders with greater than 5% ownership The percentage of institutional investor holdings INTPL Indicator equal to 1 if payouts are interpolated between performance goals, and 0 otherwise. LEVERAGE Total liabilities divided by book value of equity LIQUIDITY LITIGATION MEET MFRETURN OPTIONGRANT PESS Natural log of share volume divided by the number of shares outstanding Indicator equal to 1 if a firm belongs to a litigious industry. Litigious industries are Bio-Technology (SIC 2833 to 2836); Research, Development, and Testing Services (SIC 8731 to 8734); Computer Hardware (SIC 3570 to 3577); Electronics (SIC 3600 to 3674); Computer Software (SIC 7371 to 7379); Retailing (SIC 5200 to 5961); Communications (SIC 4812, 4813, 4833, 4841 and 4899); and Utilities (SIC 4911, 4922, 4923, 4924, 4931 and 4941). Indicator equal to 1 if TARGETRETURN is achieved at the end of the evaluation period, 0 otherwise where TARGETRETURN = Return to be achieved to earn the target payout. If threshold goal, as well as target goal, is given, threshold goal is used. Past two-year average market reaction to firm i s management EPS forecasts, calculated as mean of absolute returns during the [-2, 2] trading days window around management earnings forecasts issued during the past two years Indicator equal to 1 if there is an option grant in the quarter following evaluation end, 0 otherwise. Indicator equal to 1 if analyst forecast consensus at the beginning of the performance evaluation period is pessimistic, 0 otherwise. Pessimistic analyst forecast consensus is defined as negative analyst forecast error, calculated as the mean consensus analyst forecast of annual EPS of year t measured at the beginning of the last quarter prior to performance evaluation period end minus the I/B/E/S reported actual EPS, scaled by the absolute value of the consensus. POST_BAD RET_DISC RET_EA3 Indicator equal to 1 if bad news guidance is issued between performance evaluation period end and the subsequent earnings announcement, 0 otherwise. Cumulative market-adjusted returns over the three-month window immediately prior to the evaluation period end. Cumulative market-adjusted returns over the three-day window around the subsequent earnings announcement. 36

39 RET_POST ROA ROA ROLLOVER SIZE TESTRETURN VOLATILITY Cumulative market-adjusted returns over the one-month window immediately after the evaluation period end. Return-on-assets, calculated as income before extraordinary items divided by average total assets Change in return-on-assets from the previous year Indicator equal to 1 if the pool under the incentive plan rolls over from one measurement period to another, 0 otherwise. Natural logarithm of market value of equity Buy-and-hold return over the window starting the beginning of the performance evaluation period and ending 3-months before the end of the performance evaluation period. Standard deviation of CRSP daily market-adjusted return over the fiscal year WCACCQ Working capital accruals of the fiscal quarter ending 3-months before the end of the performance evaluation period, calculated as [increase in accounts receivable + increase in inventory + decrease in accounts payable + decrease in income tax payable + net change in other accrued liabilities] / lagged total assets (missing values of inventory, tax payable and other accrued liabilities are replaced with 0) (See Gong, Li, Xie 2009) Variables for falsification test WAY-BELOW_ACCT Indicator equal to 1 if TESTEPS < (TARGETEPS HISTORICALEPS), 0 otherwise where, TESTEPS = Cumulative EPS achievement measured from the beginning of the performance evaluation period to the beginning of the last quarter prior to the performance evaluation period end TARGETEPS = EPS goal to be achieved to earn the target payout HISTORICALEPS = Historical average of quarterly EPS achievement during the same quarter for the past 2 years RIGHT-BELOW_ACCT Indicator equal to 1 if (TARGETEPS HISTORICALEPS) <= TESTEPS < TARGETEPS, 0 otherwise. ABOVE_ACCT WAY-ABOVE_ACCT Indicator equal to 1 if TARGETEPS <= TESTEPS < MAXEPS, 0 otherwise where, MAXEPS = EPS goal to be achieved to earn the maximum payout Indicator equal to 1 if TESTEPS > = MAXEPS, 0 otherwise. 37

40 Figure 1. Distribution of stock returns by three-months before the evaluation period end (orange) and by the end of the evaluation period (transparent). This figure presents the distribution of stock returns as of three-months before the evaluation period end (orange) and as ofthe end of the evaluation period end (transparent). The x-axis represents the distance from the target goal performance, and the y-axis represents the frequency of observations. Extreme values are trimmed at 0.5 and 99.5 percentiles. 38

41 Figure 2. The four performance zones. This figure provides definitions of the four performance zones. The target goal performance is the firm s lowest performance goal and the maximum goal performance is the goal which results in the maximum payout. WAY-BELOW refers to actual performance that is way below the target goal. RIGHT-BELOW refers to actual performance that is right below the target, however within the range of past two-year average market reaction to the firm s management guidance. ABOVE refers to actual performance that is above the target but below the maximum goal and WAY-ABOVE refers to actual performance that is above the maximum goal. Figure 3. Performance goal structure. This figure depicts three types of performance goal structures of executive compensation contracts. G1, G2 and G3 represent the performance zones right below the first (threshold), the second (target) and the third (max) goals respectively. Zone 1 represents the performance zone below the first goal. Zone 2 represents the performance zone above the first goal but below G2, if a second goal exists. Zone 3 represents the performance zone above the second goal but below G3, if a third goal exists. Zone 4 represents the performance zone above the last goal. 39