The incentives of CEO option duration for management earnings forecasts

!

!

!

! ! M.Sc!Business!Economics!/!Finance,!2015! !

Master!Thesis!

The!Incentives!of!CEO!Option!Duration!for!

Management!Earnings!Forecasts! !

Student name: Luxi Zheng

Student number: 10838902

Supervisor: Dr. Florian Peters

!

!

!

Statement of Originality

This document is written by Luxi Zheng who declares to take full

responsibility for the contents of this document.

I declare that the text and the work presented in this document is

original and that no sources other than those mentioned in the text

and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the

supervision of completion of the work, not for the contents.

! ! ! ! ! ! ! ! ! ! ! ! !

CONTENT!

1. Introduction!

This paper examines the relationship between CEOs option durations and management earnings forecast errors during 2006 and 2013. Numerous past academic studies extensively studied economic implications of CEOs’ compensations and their incentives on managerial performance. Most of them focus on the relation between executives’ compensation and their firm performance (e.g. see Jensen, M.C. et.al. (1990); Smith et.al, (2008)); the link between compensation and corporate governance (e.g. see A. Bebchuk, M. Fried and I. Walker (2009)); and the effect of executives’ compensation on earnings forecasts (e.g. see Kanagaretnam, J.Lobo and Mathieu (2005)).

While impacts of various CEO pay terms on the incentives are well understood, the role of pay duration has not been fully explored. For example, R. Gopalan et al. suggested that longer pay duration is associated with more growth opportunities, more long-term assets, greater R&D intensity, less risky and better stock performance. Whereas some critics argue that short-term compensation increases the chance of CEOs to act in their own self-interested and behave myopically.

Firms typically impose restrictions on sale of executive equity grants with durations that range from a few months to several years.1 _{This paper is motivated by strong}

concerns about implications of CEO pay duration for the accuracy of management earnings forecasts. Due to lack of restricted stock data, I switch to focus on the effect of CEO option duration instead of total CEO pay duration.

I collect management forecast earnings per share from the Institutional Brokers Estimate System (I/B/E/S) for the year 2006-20132

. I measure the management earnings forecast error as the difference between forecast earnings per share and actual earnings per share, normalized by the stock price 7 days before the actual announced date.

I collect data on option vesting schedules from ExecuComp database, which is based on the table “Outstanding Equity Awards”. To ensure the integrity of data, I only retain CEOs options in the firms that can be matched with those reported in I/B/E/S database. Besides, I document variations in CEOs option durations and provide evidence that forecast errors decrease as CEOs option durations increase. The shorter option duration, the greater incentive for CEO to manipulate the stock price. In other words, as the option duration decreases, forecast EPS is deviating from the actual one, resulting in more biased forecasts. On the other hand, welfares of CEOs with short !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

1_{! Brian!Cadman,!Tjomme!Rusticus,!and!Jayanthi!Sunder!(2009).!}

2_{! The!Outstanding!Equity!Awards!File!provides!option!data!from!2006!onwards,!and!the!data!of!2014!is!} incomplete.! !

option durations are linked with short-term value of their companies, thus CEOs have more incentives to overestimate earnings, which is benefiting from selling overvalued stocks. While CEOs with long option durations are less likely to exercise their options before the vesting date, in other words, they are less sensitive to the short-term change in the stock prices, motivating them to report more accurate EPS.

According to the reasoning, I predict that CEOs with short option durations have negative effects on management earnings forecast errors. The intuition is that CEOs with short option durations would stimulate CEOs to behave myopia, diverting their efforts to extraction of inefficient private benefits at the expense of companies’ long-term value.

To test this competing hypothesis, I begin by examining management earnings forecast errors as a function of CEOs option durations and other control variables. I proxy CEOs option durations in one year as sum of the weighted average number of exercisable options multiply by time to maturity. And the management earnings forecast errors served as the difference between management forecast earnings per share and actual earnings per share, scaled by the stock price 7 days before the actual EPS announced date. I find a moderate evidence that the vesting periods of CEOs options are negatively associated with management earnings forecast errors, suggesting that long vesting options granted as a part of CEO compensation would reduce management earnings forecast errors and help CEO to make better management decisions. My inference is controlling for four closely related alternative explanations related to: (i) firms with poor performance and low financial quality have greater incentives to provide optimistic earning forecasts, (ii) rich information environment, low book to market ratio, low leverage effectively decreasing the probability of management forecast bias, (iii) uncertainty in economics environment driving both management and financial risks, and thus increasing management earnings forecast errors, and (iv) litigation industries environment (e.g. Bio-Technology, Computer Hardware, Computer Software, Electronics and Retailing) associated with CEOs to intentionally bias voluntarily disclosed information. My findings contribute to prior study along several dimensions. First, I investigate the role of option duration on the incentives. I focus on option grants because stock option grants are the primary component of stock-based compensation of CEOs. Second, I pay close attention to the group of CEOs option vesting periods rather than other executives’ total compensation, because CEOs who act as the primary shareholders are more likely to use their managerial discretion to manipulate stock prices and maximize their benefits. As a consequence, an overestimated earnings forecast is stronger associated with CEO’s behavior than other managers. Third, unlikely other recent studies3

that use information disclosures as an instrument to !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

capture the CEO behavior on forecasting. I document significant variation in the vesting schedule of CEO options and find evidence that option vesting is an important contract feature that company use to avoid management fraud.

In addition to the economic determinants described above, I present the empirical evidence on the relationship between CEOs option durations and management earnings forecast errors for the firms over the period 2006-2013, testing my prediction using a sample of 9,351 firm-year observations that collected from the intersection of ExecuComp, IBES, and CRSP. Since management earnings forecasts are not exactly synchronized between the forecasts file and the actuals file, I winsorize the potential outliers of forecast errors at the top and bottom 2.5 percentiles. CEOs option durations in the sample range from immediate vesting to cliff vesting on 11 years later.

The rest of this paper is structured as follows. Section 2 reviews some related literature and develops the hypotheses. Section 3 describes data and variable definitions. The empirical results and possible explanations are discussed in Section 4. Finally, Section 5 concludes the paper.

2. Literature!Review!and!Hypothesis!development!

2.1!Literature!Review!

Recent research has explored the relationship between CEO compensation and stock performance. Fich and Shivdasani (2005) described positive returns for firms when CEO accepts stock-based compensation. Lilienfeld-Toal, Ulf von and Ruenzi, (2013) found the CEO ownership act as a significant role in incentive payments, which confirmed the hypothesis that high ownership firms outperform low ownership firms. Other studies are linked to whether executive pay induces the manager to increase forecast accuracy. Cheng and Warfield (2004) indicated the implications of executive stock option compensation for the accuracy and bias in analysts’ earnings forecasts. They mentioned that forecast accuracy decreases as the proportion of total

compensation received by CEOs from stock options increases. Kanagaretnam, J.Lobo and Mathieu (2005) found analysts’ earnings forecasts accuracy decrease and

forecasts optimism increase as the proportion of executive compensation from stock options increases. Therefore, if the stock option pay is relatively large compared to the other compensations, the longer pay duration is expected to decrease the accuracy of the forecasts. In addition, Brisley (2006) found that although stock option grants are a beneficial way of paying CEO, but the major downside is that stock options with short vestings may exacerbate the risk-seeking behavior that induced executives to a blind pursuit of profitable projects but foregoing risks.

In recent debates, equity-based compensation has grown up to be a big part of overall employees’ compensation, especially for the top executives. One of the most

important reasons is to align the interests of managers (especially CEOs) with the interests of stockholders. As mentioned by Damodaran (2005), as publicly traded firms have matured and become larger, interests between shareholders (who owned the firm) and managers (who run the firm) have diverged. In a seminal work, Jensen and Meckling (1976)argued that managers acting in their best interests used to take measures to destroy stockholders’ value. In addition, Byard and Li (2005) identified the effects of “timing opportunism” on CEO and director compensation, which refers to the “opportunistic practice” where CEOs increase their value of stock options by lowering their companies’ stock prices.

The accounting literature also illustrated the impact of compensation on earnings management. Healy (1985) confirmed that managers manipulate earnings in order to increase their benefits. Cheng and Warfield (2003) concluded that the incidence of reported earnings that deviate from analyst forecasts is significantly higher for CEOs with a larger proportion of stock-based compensation.

!

With respect to the duration of CEO compensation, some of the critics pointed out executive pay duration should be linked to firm’s long-term value. Bebchuk and Fried (2010) argued that short-term pay arrangements are likely to have driven the excessive risk-taking behavior exhibited by bank executives before the financial crisis, and expanding the horizontal of pay duration is critical to addressing CEO interests in aligning with long-term value of shareholders. Other analysts share the same views, for example, “longer pay duration can improve the interest alignment between shareholders and managers” (Gopalan et al. 2013); “a longer pay duration can induce managers to be more forthcoming with bad news, because disclosures of bad news can help firms improve their investment efficiency, thereby benefiting managers in the long run” (Cheng, Cho and Kim, (2014)).

In addition, other literature indicated the optimal duration of executive compensation. In the article of Gopanlan, Milbourn, Song, and V. Thakor (2013), they demonstrated three features of optimal duration: First, the optimal duration is decreasing in the extent of mispricing of the firm's stock. Second, the optimal duration is longer in firms with poorer corporate governance. Third, CEOs with short durations are more likely to engage in myopic investment behavior, and this relationship is stronger when the extent of stock mispricing is larger.

2.2!Hypothesis!Development!

I develop my hypothesis by discussing the variance of the CEOs option durations and how does it influence the management earnings forecast errors. In my paper, I

consider inefficiency in management forecasts is characterized as CEOs are too confident about firms’ future performance. Given the potential for short-term mispricing of firms’ stocks, giving CEOs short vestings allow them to benefit from selling overvalued stocks or buying undervalued stocks. However, long vestings are primarily driven by the desire to restrict risk-taking CEOs from selling their equity holdings and their incentives to take risks, which also potentially in alignment with those of shareholders’ interests. In the following paragraphs, I will discuss why CEOs with long option durations are more likely to have less forecast errors than those with short option durations.

First, I choose to examine CEOs option durations instead of other forms of compensation because stock options represent the largest component of CEOs compensation and secondly, since my focus is on the time-based option grants, I would like to exclude any option grants with performance vesting conditions. Gerakos, Goodman, Ittner, and Larcker (2007) found that option grants with performance vesting conditions are relatively rare during the sample period, and performance-based equity grants are more common for restricted stocks than stock options.

Second, prior research suggested that CEO option duration is decreasing in the extent of mispricing of firm’s stock4

. On the other hand, vesting periods restrict executives from exercising their stock options and selling their underlying stocks in a specific period of time. In other words, CEOs with long option durations are less likely to exercise their options and sell their stocks in short-term. As such, by restricting CEOs from exercising their stock options, long vesting periods “extend the horizon” over which CEOs must hold their options before they can benefit from increased stock prices.5

Thus, CEOs with long vestings are less sensitive to the short run stock prices change and have weaker incentives to make overconfidence forecasts than those with short option durations. At this point, firms with long vesting periods help to align the interests of CEOs with those of shareholders.

Third, CEOs benefit from options only if they can exercise their options and sell the underlying stocks. Whereas vesting periods restrict CEOs from benefiting their options until the options have been fully vested. These impose CEOs to take additional risks if CEOs are more flexible to exercise their options and selling the stocks. These risks include a blind pursuit of profitable projects, manipulate stock prices and ultimately lead to substantial biases between the management earnings forecasts and actuals. Therefore, compared with firms prefer long term durations of options, CEOs enjoy with short vesting terms of stock options, all else equal.

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

4_{! Gopalan,!Milbourn,!Song!and!Thakor!(2010).!} 5_{! Cadman,!Rusticus!and!Sunder!(2009).!}

The above discussion suggests that CEOs with long-term option durations are often associated with less management earnings forecast errors. Thus, the hypothesis of my paper can be generated as follows:

H1: CEOs with short option durations are more likely to report more forecast errors than CEOs with long option durations.

3. Data!and!Variable!Definitions!

3.1!Sample!Selection

!

Management forecasts of earnings per share are obtained from the Institutional Brokers Estimate System (I/B/E/S). I merge the forecasts EPS with the actual EPS that obtained from IBES Detail History – Actuals file by using the same firm ticker and forecast period end date. To ensure the quality of data, I restrict the sample to the fiscal year 2006 to 2013. Also, I only consider annual forecasts of earnings per share, The forecast error is defined as the difference between the management forecasts earnings per share and actual earnings per share, normalized by the stock prices reported in 7 days before the actual EPS announced date. The stock price deflator is used to control for the potential spurious relations resulting from the change of macroeconomic conditions like inflation, and the estimate of forecast earnings per share can be used to access optimistic management. The forecast error increases imply the disagreement between forecast earnings per share and actual earnings per share raises. Formally, the forecast error, FE, for firm i, year t is calculated as:

! ! ! ! !"!,! =!"#!,!_!!!"#!,!

!,! ! ! ! ! ! !! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! (1)! ! ! $

The forecasts and actual earnings per share are obtained from I/B/E/S for comparability with the forecast. The deflator Pi,t is the stock price for firm i at 7 days

before the actual EPS announced date reported in the CRSP split adjusted Daily Stock File. Since management forecasts earnings are not exactly synchronized between the forecasts file and the actuals file, I winsorize the potential outliers of forecast error at the top and bottom 2.5 percentiles.

ExecuComp – Outstanding Equity Awards provide informations on the vesting terms of stock options. I compute option duration (DUR) as the sum of the weighted average number of exercisable options relative to total option grants multiple time to

maturity.6

Specifically, the equation of option duration can be written as: !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

6_{! For example, you may have 3 different exercisable options in one year, 25% of option A with 8 years time to} maturity, 30% of option B with 4 years time to maturity, and 15% of option C with 11 years time to maturity. The option duration of this year can be counted as 25%*8+30%*4+15%*11=4.85 years!

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!"#_!,! = #!"#$%&!,! #!"#$%&!,! ! !!! !∗ !!,! ! !!! (2)

Where #Optioni,t refers to the number of exercisable option i during the year t. To

ensure the integrity of the data, I delete observations where number of exercisable options is missing, and retain options only grant to CEOs. Ti,t refers to the time to

maturity of option i granted in the year t. I do not consider any option grants with performance vesting conditions, because performance based vesting awards are more common for restricted stocks and option grants with performance based vesting conditions are relatively rare.7

!

To test my prediction, I collect data on both level and length on CEO stock options from ExecuComp. For each stock option grant, I obtain the number of grant (i.e. number of exercisable securities underlying unexercised options and number of unexercisable securities underlying unexercised options), length of time to maturity (i.e. option expiration date minus the grant date).

Theoretically, CEOs who issue less biased forecasts are likely to have a better understanding of how changing in the economic environment affects firms’ earnings, and would be more effective in managing firms in an uncertainty environment. To ensure pay duration does not capture effects of other variables, I control for a series of variables that prior literatures suggested that might affect management earnings forecast errors, and potentially confound the association between management earnings forecast errors and CEOs option durations.

First, prior research suggested that managers (include CEOs) in firms with poor performance have greater incentives to provide “optimistic earnings forecasts to support market earnings expectations” (Rogers and Stocken, 2005; Rogers and Buskrik, 2006; Gong, Li and Xie, 2009). Firm performance also affects the vesting schedule of CEO equity based grants. Thus, I include return on assets (ROA); Altman’s Z score (AltmanZ); change of earnings (ChgEarn); sales growth rate (growth) and an indicator variable equal to 1 if firm reported losses (LOSS) to control for the potential effects of firm performance on management earnings forecast errors and CEOs option durations.

Second, I add some additional information related to firm general information that might affect CEOs to issue biased management forecasts. I control for firm size (lMV), since larger firms have richer information environment and generally face “greater public scrutiny”, thus have “greater incentives to avoid excessive errors in

management forecasts” (e.g. Duru and Reeb, 2002; Baginski et al., 2002; Gu and Wu, 2003). I also control for firm book to market ratio (btm) since it can be regarded as an investment guide. Artificially high book-to-market ratio is created for the purpose that !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

attracts investors, which intensifies CEOs to forecast optimistically. Moreover, the debt ratio (leverage) qualifies how leveraged a company is, and the degree of risk to the company. When the debt ratio is high, the company has a lot of debts relative to its assets. It is thus carrying more burdens in the sense that principal and interest payments take a significant amount of cash flows. Thus debt ratio (leverage) captures monitoring creditors that may limit a firm ability to overinvest (Masulis et al.

2007).

Third, following prior literature (e.g. Gong, Li and Xie, 2009), I control for the economic uncertainty to mitigate concerns that firm operating in an uncertainty environment finds it easier to generate errors in management earnings forecasts. In contrast, where there is more certainty environment, forecast errors will naturally be smaller. Thus, I control for cash flow volatility (CFOVOL), sales growth volatility (SALEGRVOL), earnings volatility (EARNVOL), and operating cycle (OPCYCLE) to mitigate any potential effects of uncertainty concerns about management earnings forecasts.

Finally, I include an indicator variable identifying litigious industry (litigation). Since the litigation industry environment (e.g. Bio Technology, Computer Hardware, Computer Software, Electronics and Retailing) affects CEOs incentives to “intentionally bias voluntarily disclosed information”8

.

To mitigate the influence of outliers in the data, I winsorize the top and bottom 1 percentiles of regression variables ROA, AltmanZ, IMV, btm and SALEGRVOL. Variables CFOVOL, ChgEarn, EARNVOL and OPCYCLE are winsorized at the top and bottom 2.5 percentiles.

To test my hypothesis, I regress management earnings forecast errors on CEOs option durations and identified other determinants management earnings forecast errors. I use Ordinary Least Squares regression with standard errors adjusted for heteroskedasticity and firm level clustering. The regression can be written as:

!"!,!! = !!+ !!∗ !!"!,!+ !!∗ !"#!,!+ !! ∗ !"#$%&'!,!+!!∗ !ℎ!"#$%!,!+ !!∗

!"##_!,! + !_!∗ !"#_!,! + !_!∗ !!" + !_!∗ !"#"$%&"_!,! + !_!∗ !"#$#%_!,!+ !_!"∗ !"#$%&'(#!,!+ !!!∗ !"#$%&'!,! + !!"∗ !"#$#%&!,!+ !!"∗ !"#"$%#"&'!,! +

!_{!,!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!} (3)$

The variables in Equation (3) are defined in the appendix.

3.2 ! Descriptive!Statistics! !

The EPS forecast errors sample consists of 1,299 firms during the fiscal year 2006-2013. I matched forecast EPS with actual EPS from the IBES database, and split adjusted stock prices from the CRSP database. I exclude firms that cannot be identified in either IBES or CRSP, which result in an initial forecast errors sample of 14,496 observations.

!

I obtain CEOs option grants from ExecuComp – Outstanding Equity Awards, which include 12,974 observations. To ensure consistency between CEOs option durations and EPS forecast errors, I keep 9,351 observations that can be well identified in both IBES and ExecuComp. These variables are subsequently matched with company accounting data from the Compustat database. Finally, I strictly constraint my sample to firms where I can observe at least one CEO option duration and one management earnings forecast error for each firm-year. Compared with entire Compustat firms, my sample firms have a somewhat higher return to assets (ROA), Altman Z score (AltmanZ), market value of equity (MV), cash flow volatility (CFOVOL), operating cycle (OPCYCLE), change of earnings (ChgEarn), but lower book to market ratio (btm) and sales growth rate (growth). The sample, however, very similar to the average COMPUSTAT firms with respect to earnings volatility (EARNVOL) and debt ratio (leverage). The final sample consists of 9,351 firm-year observations.

Table 2 provides descriptive statistics for the variables in the regression. The average management earnings forecast error (FE) is positive at 0.835%, and the median is -0.104% (close to 0). Almost 85% management earnings forecast errors within -1% and +1%, but there are still 10% of forecast errors outlying the sample observations. As also shown in the table, the average option duration is 5.833 years, and the median is 5.922 years. The minimum duration is immediate, and the maximum duration in my sample is 11.5 years.

Table 3 contains the pairwise correlation between management earnings forecast errors, CEOs option durations and other control variables in Equation (1). I find that management earnings forecast errors are negatively correlated with CEOs option durations (Pearson correlation=-0.011), suggesting that CEOs who have long term option vesting period tend to issue less optimistic (pessimistic) forecasts of earnings per share. In addition, correlations between CEOs option durations and other control variables are modest, correlations are mostly smaller than 0.1 in magnitudes, suggesting that the multicollinearity concerns are unlikely to affect the properties of regressions.

Table 4 presents the mean and median of management earnings forecast errors (FE) across 20 quantiles of CEOs option durations (DUR). As shown, the mean value of management earnings forecast errors decreases from 0.392% for the lowest option duration quantile to 0.275% for the highest option duration quantile. The difference in

the mean management earnings forecast errors between the highest and lowest option durations quantile is -0.117%. The negative relationship between management earnings forecast errors and CEOs option durations is also transparent in the Graph 1. While the median value of management earnings forecast errors almost unchanged from the lowest to the highest option duration quantile, and the fitted regression line illustrated in Graph 2 only shows a slight negative slope. In summary, the univariate results in Table 4 and Graph 1 and 2 provide the initial evidence that CEOs with long-term option durations tend to issue less biased EPS forecasts.

4. Empirical!Analysis!

Table 5 provides the univariate regression results from estimating Equation (3). As shown in Column (1), the coefficient on CEOs option durations (DUR) is negative (Coefficient=-0.00984), indicating that CEOs are more likely to issue biased forecasts if they have short vesting options. However, the insignificant coefficient for the CEOs option durations with a t-statistic of -0.61 indicates that the independent variable - CEOs option durations, in itself, does not have a significant effect on management earnings forecast errors.

As robustness checks, I again examine the relationship between management earnings forecast errors (FE) and CEOs option durations (DUR) after controlling for other factors that also affect management earnings forecast errors. As reported in Column (2), the coefficient of CEOs option durations is also negative, but shows an insignificant trend towards significant (Coefficient=-0.0202, t-statistic=-1.32). In Column (3) and Column (4), I report the estimation of management earnings forecast errors (FE) on CEOs option durations (DUR) with firm and year fixed effects and standard errors clustered at the firm level. I find coefficients of CEOs option durations in both Columns are also negative (Coefficient=-0.0220 and Coefficient=-0.0129 separately), while neither of them is significant (t-statistic=-1.39 and t-statistic=-0.53 separately).

Again, I include both firm fixed effects and year fixed effects in Column (5), similar to other specifications, I find an insignificant negative coefficient on CEOs option durations (Coefficient=-0.00826, t-statistics=-0.33).

Furthermore, I observe the coefficient using the year fixed effects estimators [Column (3)] is much similar with the coefficient using the OLS estimators [Column (2)], but coefficients using the firm fixed effects estimators are much smaller [Column (4) and Column (5)]. This would suggest that compared with the year fixed effects, the firm fixed effects have lower power to detect the relation between CEOs option durations and management earnings forecast errors. Because most of year-to-year variations in

CEO option duration occur between firms with very little occurring within firms over the time. This may explain why the t-statistics are higher in Column (2) and Column (3). And the reason of insignificant coefficients of CEOs option durations with the OLS estimators and the fixed effects estimators because of the potential outliers significantly distort the estimation results.

In addition, the R-squared has been largely improved if I take the firm fixed effects into account. With year fixed effects only, the R-squared is 0.180, suggesting that 18% of CEOs option durations variations are explained by the model. While with firm fixed effects only, the R-squared is 0.397, improving by 220%. When I include both year fixed effects and firm fixed effects, the model could explain 41.5% of CEOs option durations variances.

Turning to control variables, I find a strong positive effect on cash flow volatility (CFOVOL), suggesting that firms with more volatile cash flow tend to report more forecast errors. I also observe forecast errors are more biased when firms’ earnings (ChgEarn) change more often. In addition, I find that the management earnings forecasts are more optimistic when firms reported losses (LOSS) in the past year. Furthermore, coefficients of logarithm market value (lMV) and sales growth (growth) are significantly negative, suggesting that CEOs in large firms and in firms with high sales growth rate tend to forecast more conservatively in my sample.

Additionally, after introducing firm fixed effects, I find a significant negative effect on return on assets (ROA), indicating that firms with better past performance tend to more cautious to report management earnings forecast. While book to market ratio (btm) and debt ratio (leverage) only positively significant with OLS estimators and year fixed effects estimators, suggesting that firms have meaningful between firms variations in variables book to market ratio (btm) and debt ratio (leverage).

Previously, I find the some evidences of the negative relation between management earnings forecast errors (FE) and CEOs option durations (DUR), but all of them are statistically insignificant. On the Econometrics front, this weak performance may be attributed to many factors, including errors in variables, model misspecification, sensitivity to outliers, employing an inappropriate method and etc.. In my paper, the inefficiency arises from that the Ordinary Least Square estimations used above focus on the mean as a measure of the independent variable, which is more affected by outliers and other extreme data. In addition, I find the variable - CEOs options durations is highly non-normal distributed, and has strong outliers that can significantly distort estimation results.

To address this issue, I find the quantile regression is more appropriate because (1) it can be used in various distributions, unlike least squares regression is more

appropriate for normal distribution; (2) if the extremely value change, the quantile regression coefficient doesn’t change it value and standard deviation;9 (3) unlike least squares regression, quantile regression helps to alleviate some of the statistical problems due to fat tails or outliers.10 _{I investigate the underlying regression by using}

a quantile regression analysis method, which provides more robust to non-normal errors and outliers than the least square regression, and provide more efficient estimates by considering the impact of covariance on the entire distribution, not merely the mean.

Table 6 provides quantile regression results for Equation (3). As shown, quantile regression method increases the explanatory power of the model. The estimate coefficients for CEOs option durations using the QR estimators [Column (2)] (Coefficient=-0.00902, t-statistic=-1.67) and year fixed effects estimators [Column (3)] (Coefficient=-0.00928, t-statistic=-1.66) are extremely similar in both regressions and significant negative at 10% level, suggesting that firms have meaningful between firms variation in regressions. These results support the conjecture that CEOs with long option durations tend to report less management earnings forecast errors. In contrast, coefficients on CEOs option durations (Dur) are still insignificant when I use firm fixed effects [Column (4) and Column (5)], indicating that year-to-year variations occur very little within firms over the time.

5. Conclusion!

In this paper, I examine the relationship between CEOs option durations and management earnings forecast errors. I hypothesize that CEOs with short option durations tend to report more management earnings forecast errors. Short-term option vestings stimulate CEOs to manipulate stock prices and reported earnings, diverting their efforts to get private benefits at the expense of companies’ long-term value, ultimately resulting in biased forecasts.

To test this hypothesis, I collect management earnings forecasts of 1,299 firms during the fiscal year 2006-2013, merged with the actual EPS and split-adjusted stock prices. To ensure consistency between dependent variable and independent variable, I keep the data that can be well identified in both forecast errors database and option durations database. I restrict my sample to the firms where I can observe at least one CEO option duration and one management earnings forecast error for each firm-year. The final sample consists of 9,351 firm-year observations.

I firstly use the ordinary least squares method to test the estimation. The results provide weak evidence that CEOs with long-term option vestings are likely to issue !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

9_{! Chiang!and!Li!(2012)!} 10_{! Barnes!and!Hughes!(2002)}._!

more management earnings forecast errors, even after controlling for the year fixed effects and firm fixed effects. The reason is that the sample of CEOs option durations is highly non-normal distributed, and has strong outliers and extreme values that can significantly distort estimate results. To address this issue, I then investigate the underlying regression by using quantile regression method. The estimate coefficients of CEOs option durations are statistically significant, supporting the conjecture that CEOs with long option durations tend to report less management earnings forecast errors. For control variables, I observe forecast errors mostly reported by firms with high cash flow volatility, frequent changes in earnings, and firm reported losses in the past years. CEOs in larger firms and firms with higher sales growth rates tend to report more management earnings forecast errors. These results are consistent with previous analysis.

Looking back to my paper, there are several limitations in my estimations. First, because ExecuComp only provide the maturity of stock options, I estimate option duration as sum of the weighted average number of exercisable options relative to total option grants multiple time to maturity. As far as I know, Equiliar database already provides the vesting schedule of stock options, perhaps it is better to use it as a supplement of ExecuComp in later research. Second, because of management earnings forecasts are not exactly synchronized between the forecasts file and the actuals file, the sample of management earnings forecast errors is highly non-normal distributed, and has strong outliers and extreme values. Although the dependent variable is winsorized at the top and bottom 2.5 percentiles, and I use quantile regression later to provide more robust to non-normal errors, but the potential outliers also distort estimated results. Third, the empirical result suggests that the negative relationship between management earnings forecast errors and CEOs option durations is attributed to CEOs’ response to the change of the economic environment in business prospects. However, there might exist some alternate explanations for the same phenomenon and completely rule out competing explanations.

! ! ! ! ! ! ! ! ! ! ! ! !

References!

1. Aswath Damodaran (2005), Employee Stock Options (ESOPs) and Restricted Stock: Valuation Effects and Consequences. Working paper, Stern School of Business. 2. Barnes, M., & Hughes, A.W. (2002). A Quantile Regression Analysis of the Cross

Section of Stock Market Returns. Available at SSRN: http://ssrn.com/abstract=458522 3. Bettis, J.C., J.M. Bizjak, J. Coles, and S. Kalpathy (2007), Equity grants with

performance-based vesting conditions, Working Paper, Arizona State University. 4. Brian Cadman, Tjomme Rusticus, and Jayanthi Sunder (2009), The Effect of Economic

Determinants and Reporting Incentives on Stock Option Vesting Periods, University of Utah, Northwestern University. Available at SSRN:http://ssrn.com/abstract=1545602.

5. Byard, D., Li. Y (2005), The impact of Option-based Compensation on Director Independence. Working paper, Baruch College – CUNY.

6. Cheng, Q., and T. Warfield (2003), Stock Based Compensation, Earning Management to Meet Analysts’ Forecasts, and the Informativeness of Earnings. Working paper,

University of Wisconsin.

7. Thomas C. Chiang and Jiandong Li (2012), Stock Returns and Risk: Evidence from Quantile Regression Analysis. Journal of Risk and Financial Management 5 (2012): 20-58.

8. Dan Bernhardt, Murillo Campello and Edward Kutsoati (1999). Analyst Compensation and Forecast Bias. Tufts University, Department of Economics Working Paper No.

99-09. Available at SSRN: http://ssrn.com/abstract=173888.

9. Demsetz, H. and Villalonga, B. (2001). Ownership Structure and Corporate

Performance. Journal of Corporate Finance, 7, p.209-233.

10. Eliezer M. Fich and Anil Shivdasani (2004), The Impact of Stock-Option Compensation for Outside Directors on Firm Value, Journal of Business, Forthcoming, Drexel

University. Available at SSRN: http://papers.ssrn.com/abstract=491343.

11. Gerakos, J., T. Goodman, C. Ittner, and D. Larcker, 2007. Compensation objectives and non- cash pay: an investigation of benefits and broad-based equity grants. Working Paper, The University of Pennsylvania.

12. Guojin Gong, Laura Yue Li and Hong Xie (2009), The Association between

Management Earnings Forecast Errors and Accruals. The Accounting Review 84(2):

497-530.

13. Healy (1985), The Effect of Bonus Schemes on Accounting Decisions. Journal of

Accounting and Economics 7 (1-3): 85-107.

14. Jensen, M.C., Meckling, W.H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics 3, 305-360.

15. Jensen, M.C. and Murphy K. J. (1990). CEO Incentives – It’s Not How Much You Pay, But How. Harvard Business Review, May-June 1990, 3, p. 138-153.

16. JL Rogers, A Van Buskirk (2006), How Do Managers Deal with Bad News?

Management Forecasts Around Negative Earnings Surprises. Working paper, University of Chicago.

17. Kathleen A. Farrell, and David A. Whildbee (2003). The Impact of Firm Performance Expections on CEO Turnover and Replacement Decisions, JEL classification: G30;

G34, University of Nebraska – Lincoln, Washington State University.

18. Kiridaran (Giri) Kanagaretnam, Robert Mathieu and Gerald J. Lobo (2006). CEO Compensation Mix and Analysts' Forecast Accuracy and Bias. York University - Schulich School of Business, Wilfrid Laurier University and University of Houston - C.T. Bauer College of Business. Available at SSRN: http://ssrn.com/abstract=874075 19. Lilienfeld-Toal, Ulf von and Ruenzi, (2013). CEO Ownership and Stock Market

Performance, and Managerial Discretion. Journal of Finance, Forthcoming. Available at SSRN: http://ssrn.com/abstract=1343179 or

http://dx.doi.org/10.2139/ssrn.1343179.

20. Lucian A. Bebchuk and Jesse M. Fried (2009), Executive Compensation as an Agency Problem, Journal of Economic Perspectives, Vol. 17, pp. 71-92, 2003, Harvard Law and Economics Discussion Paper No. 421, Harvard Law School and Harvard Law School. Available at SSRN: http://papers.ssrn.com/abstract=364220

21. M.H. Franco Wong, and X. Frank Zhang (2009), CEO optimism, Analyst Rationality, and Earnings Forecast Bias, University of Toronto, Yale University. Available at SSRN: http://papers.ssrn.com/abstract=814085

22. Masulis, R., C. Wang, andF. Xie (2007), Corporate governance and acquirer returns. Journal of

Finance 62: 1851-1889.

23. Neil Brisley (2006), Executive Stock Options: Early Exercise Provisions and

Risk-taking Incentives. The Journal of Finance, Volume 61, Issue 5, pages 2487–2509. 24. Qiang Cheng, Young Jun Cho and Jae B. Kim (2014). Managers’ Pay Duration and

Voluntary Disclosures, 2014 Canadian Academic Accounting Association (CAAA)

Annual Conference, Singapore Management University, Singapore Management

University - School of Accountancy. Available at SSRN: http://ssrn.com/abstract=2372814.

25. Qiang Cheng and Terry Warfield (2004), Equity Incentives and Earnings Management,

Accounting Review, April 2005, Singapore Management University and University of

Wisconsin - Wisconsin School of Business. Available at SSRN:

http://ssrn.com/abstract=626848

26. Q Cheng, T Warfield (2003), Stock-based Compensation, Insider Trading, and Earnings Management. Working Paper, University of British Columbia and University of Wisconsin.

27. Radhakrishnan Gopalan, Todd T. Milbourn, Fenghua Song and Anjan V. Thakor (2013). Duration of Executive Compensation. Journal of Finance, Forthcoming. Available at SSRN: http://ssrn.com/abstract=2198397.

28. Radhakrishnan Gopalan, Todd T. Milbourn, Fenghua Song and Anjan V. Thakor (2013). The Optimal Duration of Executive Compensation: Theory and Evidence. AFA 2012

Chicago Meetings Paper, Washington University in Saint Louis - John M. Olin

Business School, Washington University in Saint Louis - Olin Business School,

Saint Louis - John M. Olin School of Business. Available at SSRN: http://ssrn.com/abstract=1656603.

29. Rogers, J. L. and P. C. Stocken (2005), Credibility of management forecasts. The

Appendix!

Table 1 Variable Definitions

The association between management earnings forecast errors and CEOs option durations is tested using the following variables:

Variable Definition Management earnings

forecast error in percent (FE)

difference between median forecast EPS and actual EPS scaled by the stock price 7 days before the actual announced date

Option duration (Dur)

sum of the weighted average number of exercisable options multiply by time to maturity

Return on assets (ROA)

earnings before extraordinary items divided by one year lagged total assets

Altman's Z-score (AltmanZ)

1.2*working capital/total assets - 1.4*retained earnings/total assets + 3.3*operating income/total assets + 0.6*market value of equity/total liabilities + sales/total assets

Litigious industry (litigation)

indicator variable equal to 1 for litigious industries including

Bio-Technology (SIC 2833 to 2836), Computer Hardware (SIC 3570 to 3577), Electronics (SIC 3600 to 3674), Retailing (SIC 5200 to 5961), and Computer Software (SIC 7371 to 7379), and 0 otherwise.

Log market value of

equity (lMV) logarithm of market value of equity

Book to market ratio (btm) book value of equity divided by market value of equity Volatility of operating

cash flow (CFOVOL)

standard deviation of operating cash flows divided by lagged total assets during the past five years, then scaled by average operating cash flows divided by lagged total assets over the same period.

Volatility of sales growth (SALEGRVOL)

standard deviation of sales growth over the past five years, scaled by the average sales growth over the same period.

Change of earnings

(ChgEarn) change of earnings in year t, scaled by year-end close price. Earning volatility

(EARNVOL)

standard deviation of income before extraordinary items scaled by total assets over the past five years.

Firm reported loss (LOSS) indicator variable equal to 1 if firm report loss in year t. Debt ratio (leverage) total liability divided by total assets

Sales growth (growth) change of sales in year t, scaled by total sales in year t-1 Operating cycle

(OPCYCLE)

average accounts receivable divided by sales, plus average inventory divided by cost of good sold.

Table 2

Descriptive Statistics

(1) (2) (3) (4) (5) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20) VARIABLES N mean sd min max p1 p5 p10 p25 p50 p75 p90 p95 p99 Mean Total Forecast error in percent 9,351 0.835 10.75 -86.13 461.2 -3.315 -1.435 -0.917 -0.397 -0.104 0.193 1.494 3.460 15.61 -

Option duration 9,351 5.833 1.946 0 11.50 1.021 2.459 3.194 4.516 5.922 7.180 8.295 9 10 -

Return on assets 9,263 0.066 0.081 -0.895 0.562 -0.216 -0.042 0.003 0.033 0.068 0.103 0.147 0.175 0.252 -0.022

Altman's Z-score 8,746 3.642 3.252 0.239 33.27 0.599 0.966 1.254 1.824 2.667 4.353 6.794 9.129 18.43 7.483

Litigious industries 9,322 0.327 0.469 0 1 0 0 0 0 0 1 1 1 1 -

Market value of equity 9,322 11,427 27,952 1.889 382,421 86.19 279.0 453.7 1,044 3,204 9,510 25,128 45,755 170,869 4416.488

Book to market ratio 9,322 0.482 0.495 -9.086 6.522 -0.0288 0.103 0.164 0.255 0.401 0.614 0.888 1.117 2.101 0.626

Cash flow volatility 8,708 0.454 1.496 -44.05 24.88 -0.364 0.0809 0.107 0.173 0.290 0.530 0.892 1.390 3.987 0.137

Volatility of sales growth 8,782 0.988 32.86 -703.4 995.5 -29.85 -5.335 -2.348 0.381 0.788 1.615 3.775 6.792 34.22 -

Operating cycle 9,193 0.511 1.389 0 22.38 0.027 0.069 0.123 0.205 0.304 0.451 0.673 0.896 9.293 2.091

Earning volatility 8,973 226.9 490.4 1.168 6,897 3.061 6.510 9.162 21.71 63.29 194.7 543.7 1,129 2,277 216.101

Change of earnings 9,292 -3.507 131.6 -6,557 1,582 -131.4 -19.10 -6.928 -0.724 0.295 1.565 6.391 13.93 67.45 -10.486

Firm reported loss 9,322 0.098 0.297 0 1 0 0 0 0 0 0 0 1 1 -

Debt ratio 9,295 0.543 0.203 0.0642 1.877 0.113 0.207 0.280 0.405 0.545 0.676 0.799 0.880 1.024 0.556

Sales growth 9,308 0.082 0.188 -0.875 2.625 -0.363 -0.173 -0.086 0.002 0.070 0.142 0.255 0.362 0.757 0.594

Table 2 presents descriptive statistics for the sample. Column (1) - Column (19) describe the descriptive statistics in the sample between 2006 and 2013. Column (20) shows the mean of variables for the entire COMPUSTAT firms between 2006 and 2013.

! ! !

Table 3

Correlations among Regression Variables

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (1) Forecast errors, in percent 1.000

(2) Option duration -0.011 1.000

(3) Return on assets -0.178 0.009 1.000

(4) Altman Z score -0.036 0.006 0.363 1.000

(5) Litigation 0.007 -0.060 0.056 0.154 1.000

(6) Log market value -0.113 0.078 0.267 -0.139 0.010 1.000

(7) Book to market 0.092 -0.066 -0.302 -0.210 -0.066 -0.319 1.000

(8) Cash flow volatility 0.078 -0.039 -0.034 0.111 0.011 -0.085 0.018 1.000

(9) Sales growth volatility 0.008 -0.020 -0.003 0.002 -0.009 -0.023 0.000 0.006 1.000

(10) Change of earnings -0.074 0.027 0.155 0.005 -0.007 0.045 0.148 0.026 0.000 1.000 (11) Earnings volatility 0.047 0.025 -0.095 -0.116 0.044 0.485 -0.052 -0.021 -0.014 -0.031 1.000 (12) Loss 0.216 0.048 -0.620 -0.079 0.010 -0.267 0.267 0.045 0.008 -0.117 0.084 1.000 (13) Debt ratio 0.055 0.049 -0.221 -0.550 -0.191 0.240 -0.149 -0.027 -0.018 -0.056 0.210 0.062 1.000 (14) Sales Growth -0.110 -0.013 0.208 0.220 0.035 0.017 -0.153 0.095 0.006 0.019 -0.111 -0.158 -0.105 1.000 (15) Operating cycle 0.060 -0.042 -0.088 0.093 -0.080 -0.034 0.066 0.017 0.003 -0.037 0.046 0.052 0.182 0.021 1.000 Table 3 presents Pearson correlations between management earnings forecast errors. CEOs option durations and other control variables for the sample period 2006-2013. Bold figures indicate significance levels at less than 1 percent. All variables are defined in the Table 1.!

! !

Table 4

Univariate Relation between Management Earnings Forecast Errors (FE) and CEOs Option Durations (DUR)

Across CEOs Option Durations Ranks Quantile Ranks of CEOs Option

Durations N Mean Forecast Errors

Median Forecast Errors Lowest 468 0.392 -0.099 2 469 0.136 -0.140 3 466 0.279 -0.110 4 468 0.267 -0.096 5 468 0.034 -0.110 6 467 0.209 -0.112 7 467 0.170 -0.084 8 471 0.171 -0.128 9 464 0.225 -0.055 10 468 0.242 -0.111 11 469 0.186 -0.085 12 468 0.083 -0.126 13 468 0.272 -0.083 14 467 0.172 -0.118 15 466 0.195 -0.088 16 470 0.254 -0.107 17 465 0.179 -0.091 18 469 0.064 -0.176 19 489 0.175 -0.114 Highest 444 0.275 -0.087 Highest-Lowest -0.117 0.012 Total 9351 0.199 -0.104

Table 4 reports the mean and median of management earnings forecast errors across 20 quantiles of CEOs option durations. See Table 1 for variable definitions.

GRAPH 1

Mean Management Earning Forecast Errors across CEOs Option Durations Quantiles

Graph 1 presents mean management earnings forecast errors (FE) across 20 quantiles of CEOs option durations (DUR). FE and DUR are defined in the Table 1. Firm-year observations are assigned in equal numbers to quantile portfolios based on DUR. The sample has 9,351 firm-year observations for which

FE, DUR, and control variables in Equation (3) are available from Compustat, Execucomp, CRSP and IBES database. The sample period is from 2006 to 2013.

GRAPH 2!

Median Management Earning Forecast Errors across CEOs Option Durations Quantiles

!

Graph 2 presents median management earnings forecast errors (FE) across 20 quantiles of CEOs option durations (DUR). FE and DUR are defined in the Table 1. Firm-year observations are assigned in equal numbers to quantile portfolios based on Dur. The sample has 9,351 firm-year observations for which

FE, DUR, and control variables in Equation (3) are available from Compustat, Execucomp, CRSP and IBES databases. The sample period is from 2006 to 2013.

TABLE 5

Regression of Management Earning Forecast Errors on CEOs Option Durations

VARIABLES (1) (2) (3) (4) (5)

Option duration (Dur) -0.00984 -0.0202 -0.0220 -0.0129 -0.00826 (-0.61) (-1.32) (-1.39) (-0.53) (-0.33) Return on assets (ROA), Winsorized fraction .01 -1.768 -1.541 -3.598 -2.953 (-1.75)* (-1.52) (-3.24)*** (-2.74)*** Altman Z score (AltmanZ), Winsorized fraction .01 0.00547 0.00571 -0.000941 0.00359

(0.33) (0.35) (-0.04) (0.16)

Litigation (litigation) 0.0914 0.0822 (1.46) (1.31)

Log market value (lMV), Winsorized fraction .01 -0.0524 -0.0515 -0.284 -0.367 (-1.82)* (-1.77)* (-2.67)*** (-2.92)*** Book to market ratio (btm), Winsorized fraction .01 0.424 0.472 0.197 0.233

(2.09)** (2.31)** (0.73) (0.80) Cash flow volatility (CFOVOL), Winsorized

fraction .025

0.351 0.335 0.311 0.317

(3.30)*** (3.16)*** (2.99)*** (3.08)*** Sales growth volatility (SALESGRVOL),

Winsorized fraction .01

0.00800 0.00646 0.00699 0.00540

(1.57) (1.32) (1.31) (1.07)

Change of earnings (ChgEarn), Winsorized fraction .025

0.0117 0.0110 0.0149 0.0126

TABLE 5 (continued) Earnings volatility (EARNVOL), Winsorized

fraction .025

6.78e-06 3.42e-05 0.000219 0.000210

(0.06) (0.30) (1.08) (1.04)

Loss (LOSS) 0.868 0.861 0.564 0.547

(4.56)*** (4.58)*** (2.57)** (2.58)**

Debt ratio (leverage) 0.640 0.727 0.0672 0.236

(2.80)*** (3.20)*** (0.12) (0.44)

Sales growth (growth) -1.366 -1.063 -1.021 -0.518

(-7.54)*** (-6.05)*** (-4.66)*** (-2.55)** Operating cycle (OPCYCLE), Winsorized

fraction .025 0.0704 0.102 0.537 0.548 (0.54) (0.76) (1.48) (1.58) Constant 0.256 0.103 -0.00629 2.359 2.679 (2.54)** (0.28) (-0.02) (2.06)** (2.17)** Observations 9,351 8,105 8,105 8,105 8,105 R-squared 0.000 0.163 0.180 0.397 0.415

Firm FE NO NO NO YES YES

Year FE NO NO YES NO YES

***"p<0.01,"**"p<0.05,"*"p<0.1"

Table 5 presents univariate regressions of Equation (1). The dependent variable management earnings forecast errors (FE) is multiplied by 100 and winsorized at the top and bottom 2.5 percentiles. To correct cross-correlation amongst grants by firm, the coefficients’ standard errors are clustered at firm level. The Robust t-statistics are reported in parentheses. Variables ROA, AltmanZ, lMV, btm, and SALESGRVOL are winsorized at the top and bottom one percentiles. Variables CFOVOL, ChgEarn, EARNVOL and

OPCYCLE are winsorized at the top and bottom 2.5 percentiles. All variables are defined in Table 1.

Colunm (1) and Column (2) reports results from ordinary least squares regressions of management earnings forecast errors on CEOs option durations (the Δ in CEO option duration for a 1% Δ in management earning forecast errors). Column (3) and Column (4) repeat the regression adding the year fixed effects and the firm fixed effects respectively. Column (5) includes both the year fixed effects and the firm fixed effects. ***, **, and * denote coefficient estimates significantly different from zero at 1%, 5% and 10% confidence interval (two-sided).

Table 6

Regression of Management Earning Forecast Errors on 20 Quantiles of CEOs Option Durations

VARIABLES (1) (2) (3) (4) (5)

20 quantiles of option duration (DUR) -0.00278 -0.00902 -0.00928 -0.00231 -0.00639 (-0.52) (-1.67)* (-1.66)* (-0.28) (-0.67) Return on assets (ROA), Winsorized fraction .01 -1.490 -1.301 -3.102 -3.420 (-1.36) (-1.18) (-2.63)*** (-3.23)*** Altman Z score (AltmanZ), Winsorized fraction .01 -0.00199 -0.00158 -0.00856 0.00576

(-0.11) (-0.09) (-0.36) (0.28) Litigation (litigation) 0.0818 0.0714

(1.22) (1.06)

Log market value (lMV), Winsorized fraction .01 -0.0443 -0.0425 -0.269 -0.332 (-1.53) (-1.45) (-2.54)** (-2.56)** Book to market ratio (btm), Winsorized fraction .01 0.404 0.450 0.226 0.213

(1.88)* (2.08)** (0.80) (0.76) Cash flow volatility (CFOVOL), Winsorized

fraction .025

0.381 0.369 0.331 0.280

(3.39)*** (3.27)*** (3.18)*** (2.47)** Sales growth volatility (SALESGRVOL), Winsorized

fraction .01

0.00679 0.00501 0.00491 0.00805

(1.26) (0.97) (0.87) (1.62)

Change of earnings (ChgEarn), Winsorized fraction .025

0.0119 0.0111 0.0145 0.0130

TABLE 6 (continued) Earnings volatility (EARNVOL), Winsorized

fraction .025

-2.81e-05 -4.53e-06 0.000191 0.000274

(-0.25) (-0.04) (0.98) (1.34)

Loss (LOSS) 0.891 0.879 0.640 0.461

(4.36)*** (4.34)*** (2.70)*** (2.30)**

Debt ratio (leverage) 0.597 0.683 0.129 0.227

(2.45)** (2.82)*** (0.22) (0.45)

Sales growth (growth) -1.403 -1.084 -1.059 -0.481

(-7.38)*** (-5.84)*** (-4.65)*** (-2.29)** Operating cycle (OPCYCLE), Winsorized

fraction .025 0.0297 0.0601 0.352 0.730 (0.21) (0.42) (0.91) (2.16)** Constant 0.228 0.0764 -0.0452 2.197 2.435 (3.46)*** (0.20) (-0.12) (1.97)** (1.94)* Observations 9,351 8,123 8,123 8,123 8,123 R-squared 0.000 0.160 0.178 0.397 0.419

Firm FE NO NO NO YES YES

Year FE NO NO YES NO YES

***"p<0.01,"**"p<0.05,"*"p<0.1"

Table 6 presents quantile regressions of Equation (1). The dependent variable management earnings forecast errors (FE) is multiplied by 100 and winsorized at the top and bottom 2.5 percentiles. The independent variable CEOs option duration (DUR) is converted into 20 quantiles. To correct cross-correlation amongst grants by firm, the coefficients’ standard errors are clustered at firm level. The Robust t-statistics are reported in parentheses. Variables ROA, AltmanZ, lMV, btm, and SALESGRVOL are

winsorized at the top and bottom one percentiles. Variables CFOVOL, ChgEarn, EARNVOL and OPCYCLE are winsorized at the top and bottom 2.5 percentiles. All variables are defined in Table 1.

Colunm (1) and Column (2) reports results from quantile regressions of management earnings forecast errors on 20 quantiles of CEOs option durations (the Δ in CEO option duration for a 1% Δ in management earning forecast errors). Column (3) and Column (4) repeat the regression adding the year fixed effects and the firm fixed effects respectively. Column (5) includes both the year fixed effects and the firm fixed effects. ***, **, and * denote coefficient estimates significantly different from zero at 1%, 5% and 10% confidence interval (two-sided).