3 Empirical Studies
3.2 Empirical Results
Our empirical studies test the relation between the forecast induced stock price trends, man-agement compensation structure (Prediction 2) and managerial overconfidence (Prediction 4), and how the choices of price trend and managerial overconfidence affect the manager’s effort decision (Predictions 3 and 5). Finally we test whether the choice of stock price trend will affect the companies’ stock volatility.
We test our theoretical predictions by estimating equations (*) to (&&&), whiel controlling for important factors that have systemic effects on hte corss-sectional variation in firm perfor-mance. Table *** and ** present the regression results on the choice of price stock trends for good and bad news, respectively.
Table 4: Effects of Compensation Structures and Managerial Overconfidence on the Choices of Stock Price Trends (Good News)
The table presents the results of regressions of stock Price Trends on compensation structures and CEO overconfidence for good news. Managers’ choices of stock price trends can be either Preignition or Inoculation, whose definitions vary with the type of information. Dummy variable Preignition (good) is equal to 1 if the manager’s choice is Preignition for good news. There are two groups of compensations: stock option awards (Stock_awards, Option_awards) and non-equity incentive compensations (Salary, Bonus, Noneq_incent). Overconfident is an indicator variable equal to 1 for all years after the CEO holds options that are at least 67% in the money. Standard errors are corrected for clustering of observations at the firm level (t-statistics are in parentheses).
*, **, and *** measure significance at the 10%, 5%, and 1% level, respectively.
Dependent variable:
Preignition (good)
Probit Regression (N=1157) Logistic Regression (N=1157) Coefficient
Estimate (p-value) Coefficient
Estimate (p-value)
Intercept 0.0342888 0.926 0.0567302 0.925 Overconfident 0.1763768** 0.032 0.2903778** 0.031
Age -0.0042693 0.390 -0.0064358 0.424
Salary_t 0.9664556* 0.073 1.5061840* 0.087 Bonus_t -0.0175667 0.936 -0.0467393 0.892 Stock_awards_t -0.0177601 0.763 -0.0237464 0.808 Option_awards_t -0.2917570*** 0.004 -0.4734700*** 0.006 Noneq_incent_t 0.1173594 0.206 0.1940484 0.206 Pension_chg_t 0.0193263 0.839 0.0416334 0.791 Othcomp_t 0.2862570* 0.074 0.4574256* 0.081 Size_t 0.1723750** 0.033 0.2868193** 0.033 Ln(Sales_t) -0.1553991** 0.044 -0.2611051** 0.040 Growth of sales_t-1 -1.6590000*** 0.000 -2.8996580*** 0.000 Ppeemp_t -0.0007884** 0.041 -0.0014275 0.108 Leverage_t -0.3015026* 0.093 -0.5102244* 0.098 Cashholding_t 0.0072789 0.231 0.0114533 0.262 Tobinq_t 0.0997410*** 0.000 0.1680110*** 0.000
Pseudo R2 0.0749 0.0758
Table 5: Effects of Compensation Structures and Managerial Overconfidence on the Choices of Stock Price Trends (Bad News)
The table presents the results of regressions of stock Price Trends on compensation structures and CEO overconfidence for bad news. Managers’ choices of stock price trends can be either Preignition or Inoculation, whose definitions vary with the type of information. Dummy variable Preignition (bad) is equal to 1 if the manager’s choice is Preignition for bad news. There are two groups of compensations: stock option awards (Stock_awards, Option_awards) and non-equity incentive compensations (Salary, Bonus, Noneq_incent). Overconfident is an indicator variable equal to 1 for all years after the CEO holds options that are at least 67% in the money. Standard errors are corrected for clustering of observations at the firm level (t-statistics are in parentheses).
*, **, and *** measure significance at the 10%, 5%, and 1% level, respectively.
Dependent variable:
Preignition (bad)
Probit Regressions (N=458) Logistic Regression (N=458) Coefficient
Estimate
(p-value) Coefficient
Estimate (p-value)
Intercept 0.0543674 0.925 0.0633984 0.947 Overconfident -0.3505896** 0.013 -0.6089210*** 0.009
Age -0.0036122 0.663 -0.0052754 0.698
Salary_t 2.1427230** 0.023 3.6720450** 0.022 Bonus_t 0.7001144 0.291 1.3693920 0.308 Stock_awards_t 0.3761224** 0.013 0.6788836** 0.019 Option_awards_t 0.2177069 0.303 0.4183311 0.260 Noneq_incent_t -1.6516460*** 0.000 -3.0145020*** 0.000 Pension_chg_t -0.7146580*** 0.003 -1.1560100*** 0.005 Othcomp_t -0.1995368 0.344 -0.3698495 0.296 Size_t -0.3192763** 0.019 -0.5318035** 0.029 Ln(Sales_t) 0.2461747* 0.062 0.4096380* 0.069 Growth of sales_t-1 1.1797060*** 0.002 2.0198890*** 0.004
Ppeemp_t 0.0003002 0.268 0.0005816 0.302 Leverage_t 0.4918430** 0.016 0.8281066** 0.020 Cashholding_t 0.0445601** 0.035 0.0748601** 0.033 Tobinq_t -0.0561584** 0.032 -0.1066195 0.104
Pseudo R2 0.1321 0.1358
A. Stock Price Trends and Compensation Structures
()= 0+ 1 + 02+ 03+ (9) First, Prediction 2 says that their impacts of compensation strucutres on the choice of price trends vary with the type of informaiton. Prediction 2 involves with the joint effects from compensaton structures and information type. First, the type of information affects the manager’s perceived future earnings (Φ1−), and hence the the implicit loss in performance
bonus will be higher with managerial overconfidence. For good news, the manager’s perceived future earnings is greater than average, hence the implicit loss in performance bonus in the net benefit of increasing 4 is lower than average for good news. Hence, ceteris paribus, the incentive of increasing 4 is greater than average for good news. Contrarily, the incentive of increasing 4 is lower than average for bad news.
Second, the compensation strucutre will change the proportion of the positive benefit from stock option award and the negative effect in performance bonus. Explicitly, when the propor-tion of stock oppropor-tion awards increases, the positive benefit of increasing 4 will inrease. When the proportion of non-equity incentive compensation incrases, the negative effect of increasing 4 will inrease.
Overall, both good news and increasing the proportion of non-equity incentive compensations have negative effects on the net benefit of increasing 4 hence there is more incentive to decrease 4; Contrarily, both bad news and increasing the the proportion of stock option awards have postiive effects on the net benefit of increasing 4 hence there is more incentive to increase 4
The depedent variables in both Table *** and ** are the preignition dummy, which takes value 1 if 4 0 for good news, and 4 0 for bad news. Table *** shows that, for good news, when the proportion of stock option awards increases, managers are more likely to use innoculation (in this case, 4 0) as their dislcosure strategies. The coefficients of option awards are significantly negative at 001 for both probit and logistic regressions (-0.291757 and -0.473470, repsectivley). This suggests that in this case, the positive effect of stock option awards dominates the negative impact of good news. Managers are less likely to use preignition, or alternatively, they are more likely to use innoculation when there is good news.
Table ** shows that, for bad news, when the proportion of stock option awards increases, managers are more likely to use preignition (in this case, 4 0) as their dislcosure strategies.
The coefficients of stock awards are significantly positive at 005 for both probit and logistic regressions (0.3761224 and 0.6788836, repsectivley). Since both the increase of stock awards and bad news will increase the net benefit of increasing 4 managers are more likely to use preignition (in this case, 4 0) when there is bad news.
Table ** also shows that when the proportion of non-equity incentive compensations in-creases, the coefficients of Noneq_incent and Pension-chg are both significantly negative at
001 for probit regressions (-1.651646 and -0.714658, respectively) and for logistic regres-sions (-3.014502 and -1.15601, respectively). As explained, both Noneq_incent and Pension-chg will increase the implicit cost of increasing 4 hence decrease the net benefit of increasing 4
Therefore managers are less willing to use preignition as disclosure strategies. Alternatively, for bad news, managers are more willing to use innoculation as the non-equity incentive compen-sations increases.
B. Stock Price Trends and Managerial Overconfidence
Prediction 4 says that their impacts of managerial overconfidence on the choice of price trends also vary with the type of informaiton.
As explained, price trends affects three directions: benefit from stock option award, implicit loss in performance bonus, and explicit cost of price manipulation Prediction 4 involves with the joint effects from managerial compensation and information type. First, as described, the incentive of increasing 4 is greater than average for good news. Contrarily, the incentive of increasing 4 is lower than average for bad news.
Second, the effect of managerial overconfidence is similar to that of good news, i.e., over-confidence will increase the implicit loss in performance bonus and therefore, decrease the net benefit of increasing 4
Overall, both good news and managerial overconfidence will decrease the net benefit of increasing 4 hence there is more incentive to decrease 4 When bad news is combined with managerial overconfidence, the results depends on the two conflicting impacts on the implicit cost in performance bonus.
Table 4 shows that, for good news, managerial overconfidence will decrease the benefit of increasing 4 hence managers are more likely to use preignition (in this case, 4 0). The coefficients of Overconfident are significantly positive at 005 for both probit and logistic regressions (0.1763768 and 0.2903778, repsectivley). Since both good news and managerial
overconfidence will decrease the net benefit of increasing 4 managers are less likely to use innoculation (in this case, 4 0), or alternatively, they are more likely to use preignition when there is good news.
Table 6: Effects of Stock Price Trends and Managerial Overconfidence on Managers’ Effort Decisions (Good and Bad News)
The table presents the results of regressions of Managers’ Efforts on stock price trends and managerial overconfidence for good and bad news. We measure Managers’
Effort by sales growth, as effort is a key determinant for growth. The choices of stock price trends can be either Preignition or Inoculation, whose definitions vary with the type of information. Dummy variable Preignition (good/bad) is equal to 1 if the manager’s choice is Preignition for good/bad news. Overconfident is an indicator variable equal to 1 for all years after the CEO holds options that are at least 67% in the money. Standard errors are corrected for clustering of observations at the firm level (t-statistics are in parentheses). *, **, and *** measure significance at the 10%, 5%, and 1% level, respectively.
Dependent variable:
Managers’ Efforts
OLS (Good news, N=1157) OLS (Bad news, N=458) Coefficient
Estimate
(p-value) Coefficient
Estimate (p-value)
Intercept 0.0718450* 0.070 -0.1167838 0.230 Preignition (good) -0.0398579*** 0.000
Preignition (bad) 0.0665518*** 0.000 Overconfident 0.0474517*** 0.000 0.0003311 0.986
Age 0.0003294 0.604 -0.0012539 0.353
Salary_t -0.1279101** 0.023 0.0694351 0.556 Bonus_t 0.0126215 0.395 0.0143054 0.875 Stock_awards_t 0.0048293 0.299 0.0057369 0.602 Option_awards_t -0.0177552* 0.100 -0.0295798 0.204 Noneq_incent_t 0.0288426*** 0.000 0.0851294 0.162 Pension_chg_t -0.0171201** 0.011 -0.0235044 0.378 Othcomp_t 0.0049320 0.735 0.0185129 0.522 Size_t 0.0332593*** 0.000 -0.0096676 0.626
Ln(Sales_t) -0.0304217*** 0.000 0.0167200 0.354 Growth of sales_t-1 0.1379784*** 0.000 0.1177120 0.173 Ppeemp_t -0.0000343 0.399 -0.0000773** 0.042 Leverage_t -0.0391637** 0.013 -0.0450453* 0.098 Cashholding_t 0.0020927*** 0.003 0.0034745 0.203 Tobinq_t 0.0110799*** 0.000 0.0142381** 0.014
R-squared 0.2167 0.1884
Table ** shows that, for bad news, the coefficients of Overconfident are significantly negative at least 005 for both probit and logistic regressions (-0.3505896 and -0.608921, repsectiv-ley). Managerial overconfidence will increase the implict cost of performance bonus, and this impact is greather than that of bad news, which will decrease the implicit cost of performance bonus. Hence managerial overconfidence will decrease the net benefit of increasing 4 and hence managers are less likely to use preignition (in this case, 4 0) when there is bad news.
C. Managers’ Effort Decisions and Stock Price Trends
Predictions 3 and 5 suggest how manager’s effort decision will change with the choices of price trend and managerial overconfidence. (Predictions 3 and 5). We test our theoretical predictions by estimating equations (&) to (&&), while controlling for important factors that have systemic effects on hte corss-sectional variation in firm performance. Table & and &&
present the regression results on managers’ effort decisions for good and bad news, respectively.
Since the earnings are determined by the firms’ future perspect and managers’ effort levels, we use the growth in sales as a proxy for effort levels.
Prediction 3 says that managers’ effort decisions will change with the choice of stock price trends, but the impacts vary with the type of information. Explicitly, for both good or bad news, the manager’s effort level can increase or decrease with the probability of using preignition, depending on the relative sizes of and 4 When managers are free to manipulate both the effort levels and the stock price trends, equation (5) shows the overall trade-off faced by the managers.
If the manager shirks (i.e., (Φ1−− 22) 0) or overworks (i.e, (Φ1− − 22) 0)
then we need to have 4 {4¡
(32 − 1) −24¢
} 0 and 0 to satisfy the first order condition in equation (3).
First, the type of information will affect the managers’ perceived earnings Φ1− thus changing the incentives of using effort levels and stock price trends to keep the maximisation condition satisfied. When there is good news, the managers’ perceived earnings will be greater than expectation, and the net benefit of increasing is higher than expectation. Ceteris paribus, the manager should increase effort to lower down this net benefit. On the other hand, when there is bad news, the managers’ perceived earnings will be lower than expectation and the net
benefit of increasing is lower than expectation. Ceteris paribus, the manager should decrease effort to increase this net benefit.
However, the overall net benefits of 4 and can be lower or higher than expectation, depending on the relative size of and 4 If the net benefit of 4 is large enough to dominate the impacts from good and bad news, then managers’ effort decisions can move in an opposite direction from those only with information effects. Hence, Prediction 3 suggests that managers’
effort levels can increase or decrease with the probability of using preignition.
Table & shows that, for good news, managers’ effort levels will decrease with the probability of using preignition. The coefficients of the dummy variable Preignition (good) is significantly negative at 001 for the OLS regression (-0.0398579). Using preignition (in this case, 4 0) decreases the net benefit of 4 Despite that there is a positive impact from good news, the overal net benefits of 4 and is lower than expectation, hence managers need to reduce effort to pull up the net benefit to keep the maximisation condition satisfied.
Table & also shows that, for bad news, managers’ effort levels will increase with the probabil-ity of using preignition. The coefficients of the dummy variable Preignition (bad) is significantly positive at 001 for the OLS regression (0.0665518). Using preignition (in this case, 4 0) increases the net benefit from 4 Despite that there is a negative impact from bad news, the overal net benefits of 4 and is higher than expectation, hence managers need to increase effort to lower down the net benefit to keep the maximisation condition satisfied.
Since managers are free to manipulate both the effort levels and the stock price trends, we will later demonstrate the results from simultaneous regressions in the robust tests.
D. Managers’ Effort Decisions and Managerial Overconfidence
Propostion 5 suggests that overconfidence has different impacts on managers’ effort decisions for differernt type of information. For good news, managers’ effort level will increase with managerial overconfidence; For bad news, managers’ effort level can increase or decrease with managerial overconfidence.
Both the type of information and managerial overconfidence will affect the managers’
per-ceived earnings Φ1− thus changing the incentives of using efforts and stock price trends to keep the maximisation condition satisfied. When good news is combined with managerial overconfidence, the managers’ perceived earnings will be greater than expectation, and the net benefit of increasing is higher than expectation. However, when there is bad news, the nega-tive effect on managers’ perceived earnings is conteracted by the posinega-tive effect from managerial overconfidence.
Table & shows that, for good news, managers’ effort levels will increase with managerial overconfidence, but for bad news, the impact of overconfidence is not significant. The coefficients of the dummy variable Overconfident is significantly positive at 001 for the OLS regression for good news (0.0474517). The coefficients of the dummy variable Overconfident is positive but not significant for the OLS regression for bad news (0.0003311). The negative effect from bad news is counteracted by the positive positive effect from managerial overconfidence, and hence the overall effect is minor and insignificant.
Table 7: Effects of Stock Price Trends on Stock Return Volatility (Good and Bad News)
The table presents the results of regressions of Stock Return Volatility on stock price trends for good and bad news. Stock return volatility is the standard deviation of daily stock returns over the fiscal year, in percentage. The choices of stock price trends can be either Preignition or Inoculation, whose definitions vary with the type of information. Dummy variable Preignition (good/bad) is equal to 1 if the manager’s choice is Preignition for good/bad news. Overconfident is an indicator variable equal to 1 for all years after the CEO holds options that are at least 67% in the money.
Standard errors are corrected for clustering of observations at the firm level (t-statistics are in parentheses). *, **, and *** measure significance at the 10%, 5%, and 1% level, respectively.
Estimate (p-value) Coefficient
Estimate (p-value) Intercept 0.8159992*** 0.000 0.6812437*** 0.000 Preignition (good) -0.1038262*** 0.000
Preignition (bad) 0.2071553*** 0.000 Overconfident 0.0280835*** 0.002 -0.0439241** 0.023
Age -0.0003165 0.557 0.0011686 0.398
Salary_t -0.0105024 0.871 0.0331235 0.835 Bonus_t 0.0131501 0.528 0.0063723 0.768 Stock_awards_t 0.0025930 0.653 -0.0178481 0.484 Option_awards_t 0.0159844 0.143 -0.0698914** 0.026 Noneq_incent_t -0.0118249 0.166 -0.0703874 0.126 Pension_chg_t -0.0072981 0.415 -0.0515331* 0.068 Othcomp_t 0.0039125 0.828 -0.0051050 0.781 Size_t -0.0434816*** 0.000 -0.0949386*** 0.000
Ln(Sales_t) -0.0028378 0.730 0.0437030* 0.061 Growth of sales_t-1 0.1871094*** 0.000 0.2458475*** 0.001 Ppeemp_t 0.0000856*** 0.007 0.0000813 0.185 Leverage_t 0.0595214*** 0.008 0.2130688*** 0.000 Cashholding_t 0.0006517 0.170 -0.0003584 0.897 Tobinq_t -0.0121098*** 0.000 -0.0221514*** 0.001 R-squared 0.3323 0.4240
E. Stock Price Trend and Stock Volatility
Table % shows the relation between stock price trends and stock volatility. The linkage is that stock price trends can affect the uninformed investors’ perceptions about future earnings.
Since the literature has documented that investors will overreact to bad news (Veronesi, 1999),
we expect that the downward price trend will cause more volatility to stock return. In other words, stock volatility is negatively related to the choice of preignition for good news and positively related to the choice of preignition for bad news.
Table % justifies our predictions. For good news, choosing preignition can decrease stock volatility for good news and choosing preignition will increase stock volatility for bad news. The coefficients of the dummy variable Overconfident is significantly negative at 001 for the OLS regression for good news (-0.1038262), and he coefficients of the dummy variable Overconfident is significantly positive at 001 for the OLS regression for bad news (0.2071553).
Table 8 Simultaneous Probit Regressions of Stock Price Trends and Managers’ Effort Decisions (Good news)
The table presents the results of simultaneous probit regressions on stock price trends and managers’ efforts on compensation structures and managerial overconfidence for good news. The choices of stock price trends can be either Preignition or Inoculation, whose definitions vary with the type of information. Dummy variable Preignition (good) is equal to 1 if the manager’s choice is Preignition for good news. We measure Managers’
Effort by sales growth, as effort is a key determinant for growth. There are two groups of compensations: stock option awards (Stock_awards, Option_awards) and non-equity incentive compensations (Salary, Bonus, Noneq_incent). Overconfident is an indicator variable equal to 1 for all years after the CEO holds options that are at least 67% in the money. The Instrument variable for the regression on Preignition (good) is a dummy variable whose value is 1 if there was good news in the previous period and 0 for otherwise.
The instrument variable for the regression on Managers’ Efforts is the medium of industry growth, where industry is classified by the first two codes in SIC. Standard errors are corrected for clustering of observations at the firm level (t-statistics are in parentheses).
*, **, and *** measure significance at the 10%, 5%, and 1% level, respectively.
Simultaneous Probit Regressions ( Good news, N=1157) Dependent variable:
Preignition (good)
Dependent variable:
Managers’ Efforts Coefficient
Estimate (p-value) Coefficient
Estimate (p-value)
Preignition (good) 0.0040664 0.954
Managers’ Efforts -6.7225010*** 0.000
Intercept 0.3631134 0.272 0.0715542 0.168 Overconfident 0.4206310*** 0.000 0.0507970*** 0.000
Age -0.0002113 0.962 0.0003121 0.627
Salary_t -0.3219751 0.510 -0.1515321** 0.014
Bonus_t 0.0769992 0.700 0.0109925 0.365
Stock_awards_t 0.0210331 0.682 0.0056377 0.210 Option_awards_t -0.2776943*** 0.001 -0.0049637 0.688 Noneq_incent_t 0.2597348*** 0.001 0.0226206*** 0.008 Pension_chg_t -0.1051668 0.192 -0.0143288** 0.024 Othcomp_t 0.1925195 0.247 -0.0057128 0.708 Size_t 0.3181629*** 0.000 0.0329807*** 0.000 Ln(Sales_t) -0.2894067*** 0.000 -0.0314047*** 0.000 Ppeemp_t -0.0006789*** 0.003 -0.0000172 0.610 Leverage_t -0.4312517*** 0.003 -0.0394289** 0.012 Cashholding_t 0.0182954*** 0.001 0.0021408*** 0.000 Tobinq_t 0.1301000*** 0.000 0.0105855*** 0.000