V. Results
5.2 Stock performance
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
30
non-monotonic same as the effect on industry-adjusted ROA.
From Table 9, the excessive overconfident and excessively diffident CEOs are adversely affected firms’ ROE. The stricter measure of excessively overconfident CEOs measured by Net stock purchases excluding stock dividends after pledge tells the coefficient of top 10% is -0.0312 significantly at 1% which suggests that ROE in excessively overconfident CEO firms is 148% less than in moderate overconfident CEO firms. The reasonable measure of excessive diffident CEOs measured by Net stock purchases tells the coefficient of bottom 10% is -0.01 significantly at 1%, which suggests that ROE in excessively diffident CEO firms is 47% less than in moderate overconfident CEO firms. Same as Q and ROA that adverse effect on ROE is bigger in excessively overconfident CEO firms than excessively diffident CEO firms.
From the above three measures of firm profitability, the excessively overconfident CEOs do harm the companies and the adverse effect enlarges when approaching the extreme cutoffs. Like Goel and Thakor (2008)’s prediction, drastic overconfident CEOs would overinvest in the projects no matter the positive or negative NPV by overestimate their ability and future firm value under their leadership. In the same way, excessive diffident CEOs also decrease firm profitability.
However, there is weaker effect of excessively diffident CEOs on firm performance, indicating that the underinvestment problem may not obvious or the results are positively biased for liquidity-motivated sales need.
5.2 Stock performance
To understand the performance led by excessively overconfident CEOs and excessively diffident CEOs, two methods Fama-French portfolio and Fama-Macbeth Regression are used in the paper ever applied by Fahlenbrach (2007).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
31
First, Performance-attribution regression is applied by examining monthly excess return over risk free rate of three equal-weighted portfolios, excessive overconfident CEO portfolio, moderate overconfident CEO portfolio, and excessively diffident CEO portfolio, in the sample from 1992 to 2010. The method to classify CEO confident level is the same as that to proxy firm valuation and described in Appendix A and then assigned firm having one specific confident level CEO measured by lagged one year to the portfolio of that confident level CEO. The portfolio is reset annually.
Former literature contributes that several equity characteristics of firms have been certified to be significant in forecasting future returns, including market factor, size, book-to-market ratio and past returns (Fama and French (1992), Fama and French (1993), and Carhart (1997)). Thus I estimate above four factor model and focus on the estimated intercept coefficient as the abnormal return to demonstrate whether there is a CEO characteristic factor to influence firm stock performance. The four factor premium data is collected from multiple equity factor database of TEJ and calculated by the method used by Fama and French (1993), and Carhart (1997).
Market factor (RMRF) is the market premium calculated by gross market return over risk-free rate per month, size factor (SMB) the market capitalization premium per month, book-to-market factor (HML) the book-to-market ratio premium per month, and past return factor (Momentum) the annual stock return premium before past two months per month. The dependent variable Excess Return (ER) is the monthly gross return in excess of the risk-free rate, one month time deposit saving rate, from equal-weighted investment portfolios.
The simplified result is shown in Table 10 with columns 1 to 4 describing the results of three confidence level portfolios measured by Net Stock Purchases, columns 5 to 8 describing that of three confidence level portfolios measured by Net Stock
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
32
Purchases after pledge, columns 9 to 12 describing that measured by Net Stock Purchases excluding stock dividends, and columns 13 to 16 describing that measured by Net Stock Purchases excluding stock dividends after pledge. As the results shown in Table 10, monthly alphas in excessive overconfident CEO portfolio are significantly smaller than that in moderate overconfident CEO portfolio, and go down though non-monotonic when approaching the extreme cutoff of CEO confidence measure by Net stock purchase and Net stock purchase after pledge. However, alphas in excessively diffident CEO portfolio are bigger than that in moderate overconfident CEO portfolio but insignificant and also go down non-monotonically. The results from above confidence measure indicate the excessively overconfident CEOs have adverse impact on firm stock performance but the influence of excessively diffident CEOs is not obvious. Moreover, the monthly alphas in moderate overconfident CEO portfolio are significant manifesting that there are other factors to influence portfolio performance.
After considering stock dividends in measuring CEO confident level, monthly alphas in excessively overconfident CEO portfolio and in excessively diffident CEO portfolio are significantly smaller than that in moderate overconfident CEO portfolio and part of those in moderate overconfident CEO portfolio becomes insignificant, indicating excessively overconfident and excessively diffident CEOs do have some adverse impact on firm stock performance. However, the coefficients in excessively overconfident CEO portfolio are negative but down to zero against to my hypothesis that the adverse influence would be bigger when approaching extreme cutoff of excessively overconfident level measure. The coefficients in excessively diffident CEO portfolio are negative to large scale but then lessened when approaching to extreme cutoff corresponding to my prediction in some manner.
From Table 10, the excessive overconfident and excessively diffident CEOs
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
33
adversely affect firm stock performance, especially when considering contingent factors to influence the measure of CEO’s confidence level. However, the trend is not going down monotonically when approaching to extreme cutoff of CEO confidence level. Moreover, significant monthly alpha in moderate overconfident CEO portfolio tells that there are definitely other factors to influence the returns of these three portfolios.
To find out the possible influencing factors other than above factors on firm stock performance and understand the performance led by excessively overconfident CEOs and excessively diffident CEOs, the Fama-Macbeth regression of annual stock return is then conducted for from 1992 to 2010. The Fama-Macbeth regression is followed:
The dependent variable is the annual stock compound gross return (R) of firm for year t. The excessively overconfident CEO and excessive diffident CEO dummies are assigned to the firms by lagged one year. Beta is the systematic risk of a company regressed by annual individual stock excess returns and market excess returns. Excess return is the gross return per year less risk-free rate, one year time deposit saving rate averaged from five major Taiwanese financial institutions. Market return is the return of TSEC weighted index. The book-to-market ratio is calculated by book value of common stock in the end of prior one year plus deferred taxes over market value of common stock in the end of prior year. The market value (MV) is calculated by stock price in the end of last fiscal year multiplying shares outstanding, the return (Return) is compound gross return in the end of last year. All above are proxies as four factors
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
34
of Fama-French three factors model and one momentum and plus the dummy measured confidence level the results are shown in columns 1 to 4 in Table 10.
Then I add related independent variables to the extensive list of characteristics in (Fahlenbrach (2007), Brennan, Chordia, and Subrahmanyam (1998), and Gompers and Metrick (2001)) including Institutional Ownership (IOW) in last fiscal year, Dividend Yield (DY) the ratio of dividends in previous fiscal year to market capitalization of the year, Firm Age (FA) the number of years passed from the year of the firm founded to the last fiscal year, and CEO Ownership (COW) calculated in last year summarized in columns 5 to 8. Same as Fahlenbrach (2007), cross-sectional Fama-Macbeth (1973) regressions is used by estimating an equal weighted cross-sectional regression of total return on explanatory control variables and my confidence dummy first and then count the mean of each year as coefficients and a time-series standard deviation adjusted by Newey-West method of 18 year coefficients.
To understand the performance led by excessively overconfident CEOs and excessively diffident CEOs, the regression of annual stock return is conducted for from 1992 to 2010 ever applied by Fahlenbrach (2007). The dependent variable is the annual stock compound gross return (R) of firm for year t. The excessively overconfident CEO and excessive diffident CEO dummies are assigned to the firms by lagged one year. Beta is the systematic risk of a company regressed by annual individual stock excess returns and market excess returns. Excess return is the gross return per year less risk-free rate, one year time deposit saving rate averaged from five major Taiwanese financial institutions. Market return is the return of TSEC weighted index. The book-to-market ratio is calculated by book value of common stock in the end of prior one year plus deferred taxes over market value of common stock in the end of prior year. The market value (MV) is calculated by stock price in the end of
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
35
last fiscal year multiplying shares outstanding, the return (Return) is compound gross return in the end of last year. All above are proxies as four factors of Fama-French three factors model and one momentum and plus the dummy measured confidence level the results are shown in columns 1 to 4 in Table 11. Then I add related independent variables to the extensive list of characteristics in (Fahlenbrach (2007), Brennan, Chordia, and Subrahmanyam (1998), and Gompers and Metrick (2001)) including Institutional Ownership (IOW) in last fiscal year, Dividend Yield (DY) the ratio of dividends in previous fiscal year to market capitalization of the year, Firm Age (FA) the number of years passed from the year of the firm founded to the last fiscal year, and CEO Ownership (COW) calculated in last year summarized in columns 5 to 8. Same as Fahlenbrach (2007), cross-sectional Fama-Macbeth (1973) regressions is used by estimating an equal weighted cross-sectional regression of total return on explanatory control variables and my confidence dummy first and then count the mean of each year as coefficients and a time-series standard deviation adjusted by Newey-West method of 18 year coefficients.
Table 11 presents the results with panel A describing the confidence measured by Net Stock Purchases, Panel B that by Net Stock Purchases after Pledge, Panel C by Net Stock Purchases excluding stock dividends, and Panel D by Net Stock Purchases excluding stock dividends after pledge. As shown in Table 11, the trend of excessive overconfident CEOs and excessively diffident CEOs on stock performance is non-monotonic. The coefficients of excessive overconfident CEO went down to negative first then up when approaching to the extreme cutoff in every panel.
However, only negative coefficients are significantly downward in excessive overconfident CEOs measured as Net Stock Purchases excluding stock dividends after pledge indicating the excessively overconfident CEOs have adverse influence on the stock performance: The coefficient of top 40% to top 30% is -0.0167 and -0.0218
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
36
significant at 10%, which suggests that stock return in excessively overconfident CEO firms migrates from 68% less to 89% less than in moderate overconfident CEO firms.
For excessively diffident CEOs, the effect is also non-monotonic to the extreme cutoff. Same as of excessive overconfident CEOs, only negative coefficients are significant in excessive diffident CEO measured as Net Stock Purchases excluding stock dividends after pledge indicating that excessively diffident CEOs have adverse influence on stock performance: The coefficient of bottom 30% is -0.0276 significant at 1%, which suggests that stock return in excessively diffident CEO firms is 112%
less than in moderate overconfident CEO firms.
From Table 11, I found out that excessive overconfident CEOs and excessive diffident CEOs have adverse impact on stock performance but not that significant as profitability Q, ROA, and ROE. The pattern is not that clear and insignificant, may because the CEOs characteristic overconfidence and diffidence is not easily sensed by investors or the personal characteristic of CEOs is not the main driver to influence the stock performance. As we know, the stock return has been discovered that that is affected primarily by concurrent systematic risk, company size, book-to-market effect, and momentum. From the regression results, I found that dividend yield and book-to-market have significant influence power on firm stock performance.
Dividend yield is positive significantly at 1% and persistent in all regressions indicating that how many dividends paid would influence the investors’ demand on the stock and future price increase. Book-to-Market are also positive significant at 5%
after controlling firm and CEO characteristics showing that return of value stock is higher than of growth stock correspond to our knowledge that the price of value-oriented stock would reverse after long term depression.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
37