經理人自信程度對價值股與成長股投資績效的影響 - 政大學術集成

全文

(1)國立政治大學財務管理研究所碩士論文. 立. 政治大. ‧ 國. 學. The Effect of CEO Confidence Level on the Investment. ‧. Performance of Value and Growth Stocks. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 指導教授：周冠男博士研究生：游詩婷撰. 中華民國一百零三年六月.

(2) 致. 謝. 就讀政大財管所這兩年受到很多人的幫助，首先最要感謝的就是周冠男老師，雖然我們一直都是以視訊會議方式討論論文，真正見到面的機會很少，但老師總是很認真地指導我們，幫助我們，無論我們遇到甚麼困難，老師也都是用鼓勵的方式，真的很開心我的指導教授是周冠男老師。另外也特別感謝盧敬植老師，在我對難解的 SAS coding. 政治大. 無助時，幫助我順利跑出 data，不用再靠著人工輸入，非常感謝。. 立. ‧ 國. 學. 在同儕當中也要好好感謝我一直以來的好朋友們，無論是同師們的慧. ‧. 慈、宜均、茹蘋，還是總是一起同組的季穎、俊宏，還是總是很貼心的子淇，又或者是總愛開玩笑的別組財個組員們，因為有我們在一起，. y. Nat. er. io. sit. 才能開開心心度過這兩年，因為大家在一起煩惱、聊天、討論、前進，. n. 才能順利把論文完成。 a. iv l C n hengchi U. 最後也要特別感謝我一直以來的好朋友，啟揚和孟璇，雖然我們在完全不同的領域，實質上的幫助給不了，但心靈上有你們陪伴我，就讓我有繼續前進的動力。. 游詩婷謹識於國立政治大學財務管理研究所中華民國一百零三年六月 ii.

(3) Abstract Empirical literatures show that value stocks outperform growth stocks consistently. Recent researches also document that only the moderately overconfident CEOs make optimal investment decisions and increase the firm value. Thus, in this paper, we argue that value premium may not hold in some particular firm characteristic if value and growth firms are subdivided into firms managed with. 政治大. different CEO confidence level. Using a sample of U.S firms over 1997-2013 from. 立. ExecuComp, we find, in growth firms, only the firms managed by excessively. ‧ 國. 學. overconfident CEOs results in poor performances. Besides, the value firms managed. ‧. by excessively diffident CEOs earn the highest return; however, the growth firms earn. Nat. io. sit. y. a positive return when the CEOs are moderately overconfident. We show that firms. n. al. er. should hire CEOs matched their specific characteristics to increase firm value.. Ch. engchi. i n U. v. Keywords: excessively overconfident, moderately overconfident, diffident, CEOs, value premium. i. excessively.

(4) Table of Contents List of Tables ................................................................................................................ iii I. Introduction ................................................................................................................ 1 II. Literature Review and Hypotheses ........................................................................... 5 III. Methodology ............................................................................................................ 8 III.1 Value and Growth Stocks Measure ................................................................ 8 III.2 CEO Confidence Level Measure ................................................................... 8 III.3 Empirical Approach ..................................................................................... 10. 政治大 IV.1 Sample .......................................................................................................... 12 立. IV. Sample and Summary Statistic............................................................................... 12. ‧ 國. 學. IV.2 Summary Statistic ........................................................................................ 13 V. Empirical Result ...................................................................................................... 21. ‧. VI. Conclusion ............................................................................................................. 29. n. al. er. io. sit. y. Nat. VII. Reference .............................................................................................................. 31. Ch. engchi. ii. i n U. v.

(5) List of Tables Table1. Hypotheses summary ........................................................................................ 7 Table2. Sample Frequency of CEO Confidence Level ................................................ 15 Table3. Summary Statistics of Stock Performance of Different Classifications of firms ...................................................................................................................................... 16 Panel A. Summary Statistics of Value and Growth portfolios (before combine with CEO confidence level) ................................................................................. 16. 政治大 levels .................................................................................................................... 17 立. Panel B. Summary Statistics of portfolios of three different CEO confidence. ‧ 國. 學. Panel C. Summary Statistics of Value and Growth portfolios (after combine with CEO confidence level) ......................................................................................... 18. ‧. Panel D. Summary Statistics of portfolios intersected by book to market ratio and. sit. y. Nat. CEO confidence level .......................................................................................... 19. n. al. er. io. Table4. Summary of Regression Results of Stock Performance of Different. i n U. v. Classifications of firms ................................................................................................ 24. Ch. engchi. Panel A. Summary of Regression results of Value and Growth portfolios (before combine with CEO confidence level) .................................................................. 24 Panel B. Summary of regression results of Value and Growth portfolios (after combine with CEO confidence level) .................................................................. 25 Panel C. Summary of regression results of portfolios of three different CEO confidence levels .................................................................................................. 26 Panel D. Summary of regression results of portfolios intersected by book to market ratio and CEO confidence level ............................................................... 27. iii.

(6) I. Introduction. Behavioral finance is an important topic in the recent literatures. Many studies try to use behavioral finance to examine market anomalies that cannot be completely explained by modern finance. As Stambaugh et al. (2012) use investor sentiment to examine anomalies and demonstrate that those anomalies are significant only when. 政治大. investor sentiment is high. Following the concept of Stambaugh, we argue that value. 立. premium may not hold in some particular firm characteristic if value and growth firms. ‧ 國. 學. are subdivided into firms managed with different CEO confidence level.. ‧. Thus, our main purpose in this paper is to find how the CEO’s confidence level. Nat. io. sit. y. influences the investment performance of value and growth firms. Besides, most of. er. papers investigate the relation between corporate policy and CEO confidence level;. al. n. v i n C the however, few papers investigate of different types of CEO h estock n gperformance chi U. confidence level. Thus, another main objective in this study is to test which confidence level of CEOs is most suitable to overall companies. Several studies show that value stocks seem to have a value premium. Whether using the earning to price ratio(Basu 1977), the book to market ratio (Rosenberg et al. 1985) or the cash flow to price ratio (Chan et al. 1991) to form portfolios, value portfolios outperform growth portfolios consistently. To our best knowledge, mainly 1.

(7) existing studies demonstrate this market anomaly in two points of view, which are time varying risk and investors’ overreaction. However, none of papers examine the relation between CEO confidence level and value premium. Overconfident CEOs tend to overestimate the returns to their investment project (Heaton 2002) and result in a suboptimal investment behavior (Malmendier and Tate 2005). The poor performance of these bad investments may lead to a decrease of stock price in the long run. Goel. 政治大. and Thakor (2008) certify that excessively overconfident CEOs tend to experience. 立. overinvestment problem and excessively diffident CEOs tend to encounter. ‧ 國. 學. underinvestment problem. Only the moderately overconfident CEOs make optimal. ‧. investment decisions and increase the firm value. Thus, we hypothesize that firms. Nat. io. sit. y. managed by moderately overconfident CEOs earn the highest stock returns among. er. three CEO confidence levels (excessively overconfident, moderately overconfident. al. n. v i n C hare classified by CEO’s and excessively diffident) which e n g c h i U stock option data in this paper. Besides, we also hypothesize that firms managed by excessively overconfident CEOs earn the lowest stock return among three CEO confidence levels. We assume value and growth firms also follow the above CEO pattern, thus, we hypothesize the value firms managed by moderately overconfident CEOs earn the greatest stock returns in six intersections of two book to market ratio and three CEO confidence level and the growth firms managed by excessively overconfident CEOs earn the 2.

(8) lowest stock return in six intersections. We build value weighted portfolios and use the portfolio monthly returns as the dependent variable in the Fama-French four factors model but exclude the HML factor to test whether there is a significantly abnormal return in any of the groups. Empirical results show that CEOs confidence level can affect firm’s stock performance. Our result is consistent with previous studies which propose that. 政治大. moderately overconfident CEO increase firm value by mitigating the underinvestment. 立. problem and lead to a better stock performance. Our results also suggest that only. ‧ 國. 學. growth firms managed by excessively overconfident CEOs earn a negatively. ‧. abnormal return although value firms still earn a higher return than growth firm in. Nat. io. sit. y. each of CEO’s confidence level. Moreover, the value firms managed by excessively. er. diffident CEOs earn significantly positive returns; however, the growth firms earn. al. n. v i n positive returns when the CEOsCare This result imply that h moderately i U e n g c hoverconfident. the value firms should hire CEOs who are more conservative, on the contrary, growth firms should hire CEOs who are more willing to take risk in order to get a good match. The paper proceeds as follows. Section II reviews previous researches in value premium and CEO overconfidence, and build research hypotheses we concern in this paper. Section III presents the research methodologies and measures we applied in 3.

(9) this study. Section IV describes the sample data used in this analyses and the summary statistic of our dataset. Section V provides the empirical results of our hypotheses. Section VI summarizes the result and describes the main implication of our result.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 4. i n U. v.

(10) II. Literature Review and Hypotheses. These days, many researches examine the phenomena in modern finance by focusing on the personal characteristics of the top decision maker in firms rather than firm level characteristics. Heaton (2002) is the precursor examining and discovering the CEOs with optimistic personal characteristics lead to a distortion of investment. 政治大. decisions. Heaton document that optimistic managers may decline positive net present. 立. value projects that must be financed externally because they believe capital market. ‧ 國. 學. undervalue their firm’s securities. Besides, optimistic managers tend to overvalue. ‧. their corporate projects and may invest in negative net present value projects even. Nat. io. sit. y. when they are loyal to shareholders. Malmendier and Tate (2005) examine the relation. er. between overconfidence CEOs and investment decision by constructing measures of. al. n. v i n personalCportfolio of CEOs h e n gdecisions chi U. overconfidence use. also show that. overconfidence managers overinvest when they have abundant internal funds, but curtail investment when they requite external financing. Later, Campbell et al. (2011) follow the approach of Malmendier and Tate but refining the measures to subdivide CEO confidence level into three groups, excessively overconfident, moderately overconfident and excessively diffident. They document that CEO with relatively low or high optimism face a higher probability of forced turnover than moderately 5.

(11) optimistic CEOs face. The result is consistent with Goel and Thakor (2008) show that CEO’s overconfidence enhances firm value up to a point by solving the underinvestment problem with risk averse CEO. However, the effect is nonmonotonic and an excessively overconfidence tend to make value-destroying investments. Thus, the board should only remain moderately overconfident CEO and fire both excessively diffident and excessively overconfident CEO. Based on the former. 政治大. empirical studies presented managerial overconfidence can account for corporate. 立. investment distortions, we argue that these value-destroying investment decision will. ‧ 國. 學. result in a poor stock performance in the long run. Thus our first and second. ‧. hypotheses in alternative form are. Nat. io. sit. y. H1: The firms managed by excessively overconfident CEOs will result in the worst. er. return performance whether in value firms or growth firms.. al. n. v i n C h overconfidentU CEOs will result in the best H2: The firms managed by moderately engchi return performance whether in value firms or growth firms. In other word, not all of growth firms do bad performances. A number of studies show that diversify firms trade at discount compared with single segment firm (Lang and Stulz 1994). However, after Panayiotis C. Andreou et al. (2011) examine the relation between managerial overconfidence and the diversification discount, they show that the diversification discount is concentrated 6.

(12) exclusively in companies managed by overconfidence managers. Moreover, the discount is greater and persists for several years only for firm managed by overconfidence managers. Following this concept, we examine the relation of CEO confidence and value premium. We reclassify firms into six intersections of two book to market ratio and three CEO confidence level. We argue that not all of the growth stocks tend to underperform value stocks. Only the growth firms managed by. 政治大. excessively overconfident CEOs will have poor stock performances. We also argue. 立. that not all of the value stocks could have a better stock performance, only the value. ‧ 國. 學. firms managed by the moderately overconfident CEOs could really add firm value.. ‧. Thus our third hypothesis in alternative form is. Nat. io. sit. y. H3: The value firms managed by moderately overconfident CEOs will result in the. er. largest return among the six groups and the growth firms managed by excessively. al. n. v i n Cthe overconfident CEOs will result in U the six groups. h esmallest h i among n g creturn Table1 Hypotheses summary. This table summarizes the main hypotheses described in this paper. We subdivide value (growth) firms into three groups by three CEO confidence levels and predict the stock returns of each group. Excessively overconfident Moderately overconfident. Value. Growth. relatively low the highest. the lowest relatively high. 7.

(13) III. Methodology. III.1 Value and Growth Stocks Measure The extant literatures on value and growth stock consider a number of indicators as the book to market ratio, the cash flow to price ratio, the earning to price ratio and the dividend to price ratio. Chan and Lakonishok (2004) report that the book to. 治政 market ratio has been extensively used in academic studies 大 and the book to market 立 ‧ 國. 學. ratio also has been showed to be as good as other measures to identify value and growth stocks. (Lakonishok et al. 1994) Thus, in this paper, we use the book to market. ‧. ratio as our measure of value and growth stocks. Following Fama and French (1995),. sit. y. Nat. io. n. al. er. book equity is book value of stock equity plus deferred taxes and investment tax. i n U. v. credit minus book value of preferred stock; market equity is stock price times share. Ch. engchi. outstanding. We omit firm with negative book equity. Stocks are sorted to value and growth stocks based on the breakpoints for the top 30% and the bottom 30% of book to market ratio.. III.2 CEO Confidence Level Measure Measuring confidence level empirical exist some difficulty because it is not easy to observe directly. The extant literature on CEO confidence employs a number of 8.

(14) different measures. Malmendier and Tate (2005) construct measures of CEO confidence level based on option exercising and net stock purchasing behavior of the CEO. Malmendier and Tate (2008) develop a measure based on portrayal of CEO in the press coverage in leading business publication. Schrand and Zechman (2012) use measures based on firm level investing and financing activities that prior research found to be related to managerial overconfidence, including excess investment, debt. 政治大. to equity ratio, the use of convertible debt or preferred stock, dividend policy. As the. 立. CEO data limit, we base our measures on CEO’s stock option holding and exercising. ‧ 國. 學. decision following the approach by Malmendier and Tate (2005) but change the cutoff. ‧. of option moneyness to 100% (Campbell et al. 2011) in order to identify CEOs with. Nat. io. sit. y. relative high optimism. That is if CEOs tend to hold options too long (even deeper. n. al. therefore we classify them. er. than 100% in the money), they would like to have even greater overestimation error,. v i n asCexcessively h e n g coverconfident h i U CEOs.. Also following. Campbell et al. (2011), we define excessively diffident CEOs as who exercise options less than 30% in the money. The remainders who hold or exercise options with option moneyness between 30% to 100% are classified as moderately overconfident CEOs. However, different from previous researches which examine the relationship between CEO overconfidence and corporate policy, we study the relationship between CEO overconfidence and stock performance. There may exist a problem to classify CEOs 9.

(15) confidence level and examine the stock performance in the corresponding period. That is the more increase in stock price will result in both larger option moneyness and higher investment return and therefore tend to lead a conclusion that the firms managed by excessively overconfident CEOs will perform better. In order to avoid the endogenous problem, we use the moneyness in the preceding five years to identify the current CEO confidence level. Therefore, we assume the CEO confidence level may. 政治大. change overtime. Moreover, we require that the CEO exhibit the option. 立. holding/exercising behavior at these two times during the five year and omit the CEO. ‧ 國. 學. who have all of his options out of money or have no options or have tenure less than. ‧. five years.. y. Nat. n. er. io. al. sit. III.3 Empirical Approach. i n U. v. The major approach in this paper is to put each type of portfolio in the Fama. Ch. engchi. French four factors model to examine whether there is a significantly abnormal return and to compare the level of abnormal return between each group. Our analysis consists of two sections. First, we test the investment performance of firm managed by different CEO confidence level and the value (growth) firm respectively. Second, we incorporate CEO confidence level into value (growth) firm and test the effect of CEO confidence level on value (growth) firm. Following Fama and French (1995), we use the book to market ratio and the 10.

(16) CEO confidence level in year T-1 to divide our sample into different type of groups (value, growth, excessively overconfidence, moderately overconfidence, excessively diffident, and six intersections of two book to market ratio and three CEO confidence level groups). Then we use the market capitalization at the end of year T-1 as the weight to form portfolios at the end of each July. Finally, we use the following regression to test our hypotheses. 𝑅𝑖,𝑡 − 𝑅𝑓,𝑡 = 𝛼 + 𝛽1 MKTt + 𝛽2 SMBt + 𝛽3 MOMt + μt. 治政 Here we exclude the HML factor because Hueng et al.大 (2013) document that loadings 立 ‧ 國. 學. on the HML factor cause the Fama-French three factor model to provide a downward bias of the performance of value portfolio.. ‧. n. er. io. sit. y. Nat. al. Ch. engchi. 11. i n U. v.

(17) IV. Sample and Summary Statistic. IV.1 Sample In order to measure the CEO confidence level, we collect the CEO’s option data from ExecuComp. We also collect the common equity, preferred stock, deferred taxes and investment tax credit, common share outstanding and stock price from Compustat. 治政 to calculate the book to market ratio. To compare the 大investment performance we 立 ‧ 國. 學. obtain monthly return from Center for Research in Security Prices and the Fama-French factors data from Fama-French’s website. As the CEO option data limit,. ‧. our sample period in this paper is 1997 to 2013. (The CEO data is available from. sit. y. Nat. io. n. al. er. 1992, however, we use the data in year T-5 to year T-1 to identify the confidence level. i n U. v. in year T. Thus our portfolios start from 1997.) We exclude financial firms (SIC code. Ch. engchi. 6000-6999) and utilities (SIC code 4900-4999) and any missing data from the sample. There are 7,056 total CEO-year observations and after we combine the CEO and value (growth) data, only 3,370 CEO-year observations remain. Additionally, in order to check our result is consistent overtime, we separate the sample into two periods to do a robust check. Thus, all of the following examinations will contain three time periods, full sample, 1997-2004 and 2005-2013. Table 2 presents the frequency of three CEO confidence levels in our sample 12.

(18) from 1997 to 2012. As shown in table 2, most of CEOs are moderately overconfident in our sample and we get little CEOs with excessively diffident character. The table 2 also shows that moderate overconfidence is the most common CEO character in both value firms and growth firms; however, there is a large proportion of CEOs with excessively overconfident character in growth firms, especially when we compare to value firms. We argue that may be the reason why growth firms tend to have a poor stock performance.. 立. 政治大. IV.2 Summary Statistic. ‧ 國. 學. Table 3 presents the summary statistics of different classification of portfolios on. ‧. monthly stock returns in this paper and it also shows if there is a significant difference. sit. y. Nat. io. n. al. er. between any two portfolios. The table contains four panels. Panel A shows that the. i n U. v. value portfolio of full sample keeps outperforming growth portfolio despite the. Ch. engchi. difference is not significant. Panel B shows that the average monthly return in moderately overconfident CEOs portfolio is the largest and the average monthly return in excessively overconfident CEOs portfolio is the smallest. The result is robust in each period and consistent with Goel and Thakor’s prediction. After we classify value (growth) firms with CEO confidence level, we check if the value portfolio still outperform growth portfolio. As panel C shown, the value portfolio still gets a larger means of return than growth portfolio in each of sample period. However, comparing 13.

(19) to panel A, the difference becomes smaller. Panel D shows that value firms still earn a higher monthly stock return than growth firms no matter in what CEO confidence level. Nevertheless, comparing to the mean in panel C, only growth firm managed by excessively overconfident CEO earns a poor stock return, the means of stock return in growth firms managed by moderately overconfident CEO and excessively diffident CEO is larger than the mean of growth firm which is not classified by CEO. 政治大. confidence level in panel C. The result is consistent with our second hypothesis that. 立. not all of growth firms do bad performances.. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 14. i n U. v.

(20) Table 2 Sample Frequency of CEO Confidence Level This table shows the frequency of three different CEO confidence level observations measured by CEO option data and the six intersections of two book to market ratio and three CEO confidence level groups for the sample from 1997 to 2012. The unit is a CEO-firm-year data. Value Growth Portfolio Excessively Moderately Excessively Excessively Moderately Excessively Excessively Moderately Excessively year Overconfident Overconfident Diffident Overconfident Overconfident Diffident Overconfident Overconfident Diffident. 101 93 111 1918. 308 337 394 4274. 19 29 34 308. al. 18 16 8 265. Ch. 49 97 126 120 106 106 57. 2 4 7 11 12 5 5. engchi 90 77 81 1217. 15. y. 2010 2011 2012 Total. 11 19 23 26 33 32 18. 56 68 56 56. sit. 18 23 24 24 22 16 13. 1 2 3 5 3 0. er. 237 277 329 357 317 292 281. ‧ 國. 112 127 150 143 156 164 115. 立. 政 16治大 56. ‧. 2003 2004 2005 2006 2007 2008 2009. 5 10 11 15 13 7. 學. 7 13 17 16 14 19. n. 77 191 237 232 207 201. io. 32 94 134 132 134 120. Nat. 1997 1998 1999 2000 2001 2002. i n U. v. 8 9 9 86. 17 43 68 60 70 61. 24 50 43 25 27 40. 3 5 5 5 5 5. 59 51 56 52 55 56 47. 57 41 46 70 68 66 85. 4 6 5 4 3 0 3. 31 29 53 808. 82 81 112 917. 5 8 11 77.

(21) Table 3 Summary Statistics of Stock Performance of Different Classifications of firms The table presents the summary statistics of different classification of portfolios on monthly stock returns in this paper. This table also shows if there is any significant difference in any two classifications by using the independent sample t-test. Our sample period is 1997 to 2013. In order. 政治大. to do a robust check, the table will contain three time period, 1997 to 2013, 1997 to 2004 and 2005 to 2013. This table contains four panels. Panel A is statistics of value and growth portfolio of full sample before combine with CEO confidence level, panel B of portfolios of firms managed by three different CEO confidence levels, panel C of value and growth portfolios after combine with CEO confidence level, panel D of portfolios of six intersections of two book to market ratio and three CEO confidence level.. 立. ‧ 國. 學. Value Growth Value. 192 192 96. 0.9621 0.5897 1.1155. Growth. 96. 0.4707. Ch 5.4618. Value Growth. 96 96. 0.8087 0.7088. 5.5609 4.6472. Maximum (%). al. -22.325 -18.061 -13.3738. e n0.5574 gchi -22.325 -18.061. 16. Mean Difference (value-growth) (t-value). 13.6662 10.7513 10.7113. 0.3724 (0.72). iv n U -14.1596. 0.6447 (0.88). 13.6662 10.7513. 0.1000 (0.14). er. 5.1288 5.059 4.6814. sit. y. Std Dev (%) Minimum (%). n. 2005-2013. Mean (%). io. 1997-2004. Number. Nat. 1997-2013. ‧. Panel A. Summary Statistics of Value and Growth portfolios (before combine with CEO confidence level).

(22) Table 3 Panel B. Summary Statistics of portfolios of three different CEO confidence levels Mean Number. Mean (%). Std Dev (%). Minimum (%). Maximum (%). Difference(2)-(1) Difference(2)-(3) (t-value)(%). 5.071. -20.141. Moderately overconfident(2). 192. 0.8463. 4.1995. -16.343. Excessively diffident(3). 192. 0.7553. 4.6567. -15.76. 12.9161. Excessively overconfident(1). 96. 0.5419. 4.987. -14.7675. 10.9653. Moderately overconfident(2). 96. 0.8801. 4.1465. -11.2384. 9.577. Excessively diffident(3). 96. 0.7523. -14.0043. 12.9161. 96. 0.5804. 5.1798. -20.141. 12.6584. 96. 0.8126. 4.2732. -16.343. 10.7368. 96. 0.7583. 4.7194. -15.76. 9.5764. Difference(3)-(1) (t-value)(%). 12.6584. 政治10.7368 大. 0.2852 (0.60). 0.0910 (0.20). 0.1941 (0.39). 0.3381. 0.2103. (0.51). 0.1278 (0.20). 0.2322. 0.0543. 0.1779. (0.34). (0.08). (0.25). (0.30). er. sit. y. ‧. al. n. overconfident(2) Excessively diffident(3). 4.618. io. Excessively overconfident(1) Moderately. 立. Nat. 2005-2013. 0.5612. (t-value)(%). Mean. 學. 1997-2004. 192. ‧ 國. 1997-2013. Excessively overconfident(1). Mean. Ch. engchi. 17. i n U. v.

(23) Table 3 Panel C. Summary Statistics of Value and Growth portfolios (after combine with CEO confidence level) Number. 治政 -23.6506 -18.1288 大. Value Growth Value. 192 192 96. 0.788 0.558 0.7431. Growth. 96. 0.5139. 4.6681. -12.4546. Value Growth. 96 96. 0.8329 0.6021. 5.5211 4.1798. -23.6506 -18.1288. 立. Maximum (%) 14.5724 10.8917 14.5724. -13.04. 0.2308 (0.33). y. n. engchi. 18. i n U. v. 0.2292. 12.7599 10.1168. sit. io. Ch. 0.2300 (0.47). (0.34). ‧. Nat. al. Mean Difference (value-growth) (t-value). 10.8917. er. 2005-2013. 5.0615 4.4193 4.5847. Minimum (%). 學. 1997-2004. Std Dev (%). ‧ 國. 1997-2013. Mean (%).

(24) Table 3 Panel D. Summary Statistics of portfolios intersected by book to market ratio and CEO confidence level Excessively overconfident. Moderately overconfident. Value. Growth. Value. 192 0.7304 6.5161 -24.3447 19.9751. 192 0.4951 5.2577 -18.245 15.3749. 1997-2013. n. al. Ch. engchi. 0.0562 (0.07). i n U. 0.2169 (0.32). 19. 96 0.6062 4.6836 -11.5552 12.4308. sit. 96 0.8232 4.6926 -12.6851 15.7352. er. 96 0.5064 5.3502 -14.0996 10.6905. y. (0.41). io. Mean Difference (t-value). 96 0.5626 6.5347 -14.8092 19.9751. 4.3944 -18.0721 12.4308 0.197. Nat. 1997-2004 Number Mean (%) Std Dev (%) Minimum (%) Maximum (%). (0.39). ‧ 國. (t-value). 4.9651 -21.6497 15.7352. Value. Growth. 180 1.4739 7.1983 -38.8798 29.8722. 180 0.7994 4.6094 -15.7588 13.6448 0.6745. ‧. 0.2352. 立. 政 192 治大 192 0.8347 0.6376. 學. Number Mean (%) Std Dev (%) Minimum (%) Maximum (%) Mean Difference. Growth. Excessively diffident. v. (1.06) 84 1.9048 6.5592 -16.2534 18.7287. 96 0.654 5.1739 -15.7588 13.6448 1.2508 (1.41).

(25) Table 3 Panel D (continued) 96 0.4839. 6.5273 -24.3447 12.5899. 5.1917 -18.245 15.3749. 立. 96 0.669. 96 1.0968. 84 0.9654. 10.7322. 7.7289 -38.8798 29.8722. 3.8895 -10.0465 12.8598. 政5.2481治大 4.1092 -21.6497 -18.0721 13.4809 0.1771 (0.26). ‧. io. sit. y. Nat. n. al. er. 0.4142 (0.49). 96 0.8461. 學. Std Dev (%) Minimum (%) Maximum (%) Mean Difference (t-value). 96 0.8981. ‧ 國. 2005-2013 Number Mean (%). Ch. engchi. 20. i n U. v. 0.1314 (0.15).

(26) V. Empirical Result. To understand the effect of CEO confidence level on value (growth) firms and the stock performance of each group in our sample, we use the Fama-French four factors model but exclude the HML factor to test whether there is a significantly abnormal return or a significant difference between any two groups. Table 4 presents. 政治大. the Summary regression results of different classification of portfolios on monthly. 立. stock returns in this paper. The table contains four panels. First, as panel A showed,. ‧ 國. 學. the value portfolio of full sample earns a significantly positive abnormal return; on the. ‧. contrary, the growth portfolio of full sample earns a negative abnormal return. The. Nat. io. sit. y. return between value and growth portfolio is also significantly different indicating. er. there is a value premium in our sample which is consistent with previous researches.. al. n. v i n However, as panel B showed, C after the data without CEO option data and h we e nomit gchi U classify the value and growth firms again, the value premium becomes insignificant. This may be the result of sample bias. Second, the panel C indicates that the portfolio of firm managed by moderately overconfident CEO keeps earning a significantly positive abnormal return and the portfolio of firm managed by excessively overconfident CEO earns a negative abnormal return. The result is consistent with Goel and Thakor (2008) who 21.

(27) documented that CEO should take risk moderately to diminish the underinvestment problem and increase the firm value. However, the portfolio of firm managed by excessively diffident CEO also earns a positive abnormal return despite the return is not significant. Additionally, only the return between moderately overconfident CEO and excessively overconfident CEO is significant different. The return between moderately overconfident CEO and excessively diffident CEO is not significantly. 政治大. different. Thus, we argue the conclusion in Goel and Thakor (2008) stated that both. 立. excessively overconfident CEO and excessively diffident CEO decrease the firm. ‧ 國. 學. value and both of them should be fired by the board. Our results show that the stock. ‧. performances of firms managed by excessively diffident CEOs are still acceptable.. Nat. io. sit. y. Finally, we examine the effect of CEO confidence level on value (growth) firm.. er. As the panel D shown, comparing to the alpha in panel B, a portfolio managed by. al. n. v i n Cearns excessively overconfident CEOs abnormal return whether in value or h e na lower gchi U growth firms. This result is consistent with our first hypotheses and Malmendier and. Tate’s (2005) prediction that overconfident CEOs tend to have suboptimal investment decisions and lead to a worse stock performance. The panel D also shows that despite value portfolio still earn a higher abnormal return than growth portfolio in every CEO confidence level, only the portfolio of growth firm managed by excessively overconfident CEO earns a negative abnormal return. The result is same as our second 22.

(28) and third hypotheses that growth portfolio can still earn a high abnormal return if we exclude those firms managed by excessively overconfident CEOs. However, the inconsistent result is that the value portfolio in CEOs with excessive diffidence earns a significantly positive abnormal and the return is the highest in six intersections even comparing to the value portfolio in CEO with moderate overconfidence. This result shows that value firms should hire excessively diffident CEO to get a better. 政治大. performance. As Zhang (2005) documented, value firm experience less flexibility in. 立. cutting capital because of the costly reversibility and countercyclical price of risk.. ‧ 國. 學. Thus, a CEO who is more conservative and careful in his decisions will be more. ‧. suitable for a value firm. On the contrary, Serfling (2012) demonstrate that growth. Nat. io. sit. y. firms encounter more underinvestment problems. He construct an investment strategy. er. with buying portfolio of firms managed by younger CEOs who are more willing to. al. n. v i n take risk and selling portfolio C of firms by older CEOs who are more risk h e nmanaged gchi U. averse. He found the strategy only generate a significantly positive abnormal return in high growth firms which is consistent with our result that growth firm should hire CEO who is willing to take a risk (moderate overconfidence) but not take too much risk(excessively overconfidence).. 23.

(29) Table 4 Summary of Regression Results of Stock Performance of Different Classifications of firms The table presents the regression result of different classification of portfolios on monthly stock returns in this paper. Our sample period is 1997 to 2013. In order to do a robust check, the table will contain three time period, 1997 to 2013, 1997 to 2004 and 2005 to 2013. The regression specification underlying the results reported in this table is: 𝑅𝑖,𝑡 − 𝑅𝑓,𝑡 = 𝛼 + 𝛽1 MKTt + 𝛽2 SMBt + 𝛽3 MOMt + μt ≠. 政治大 This table also shows if there is any significant difference in any two classifications by using the following regression specification: 立 𝑅 − 𝑅 = 𝛼 + 𝛽 MKT + 𝛽 SMB + 𝛽 MOM + μ , i ≠ j 𝑖,𝑡. 𝑗,𝑡. 1. t. 2. t. 3. t. t. ‧ 國. 學. Where 𝑅𝑖,𝑡 and 𝑅𝑗,𝑡 is the portfolio return of the different classification in this paper. This table will only show 𝛼 in the regression model. ‧. because our main purpose is to see whether our portfolio can earn an abnormal return. The table contains four panels. Panel A is results of value and growth portfolio of full sample before combine with CEO confidence level, panel B of portfolios of firms managed by three. sit. y. Nat. different CEO confidence levels, panel C of value and growth portfolios after combine with CEO confidence level, panel D of portfolios of six intersections of two book to market ratio and three CEO confidence level.. al. (2). n. (1). er. io. Panel A. Summary of Regression results of Value and Growth portfolios (before combine with CEO confidence level) (3). iv. 1997-2013. 1997-2004. 2005-2013. n U i e n g1997-2004 1997-2013 2005-2013 ch. 0.39892**. 0.70198**. 0.07967. -0.04517. -0.15049. 0.08513. 0.44409*. 0.85247**. -0.00546. (2.13). (2.30). (0.48). (-0.51). (-1.11). (0.76). (1.84). (2.10). (-0.03). Number. 192. 96. 96. 192. 96. 96. 192. 96. 96. Adj-R2. 75%. 61%. 92%. 94%. 94%. 95%. 2%. 15%. 16%. Value. Intercept. Ch. Growth. 24. Difference(value-growth) 1997-2013. 1997-2004. 2005-2013.

(30) Table 4 Panel B. Summary of regression results of Value and Growth portfolios (after combine with CEO confidence level) (1). (2). (3). Value. Growth. Difference. 1997-2004. 2005-2013. 0.29174. 0.40300. 0.14263. (1.21). (1.04). Number. 192. Adj-R2. 57%. 1997-2013. 1997-2004. 2005-2013. 0.04849. 0.03979. 0.24128. 0.35452. 0.10284. (0.66). (0.45). (0.27). (0.30). (0.88). (0.80). (0.39). 96. 96. 192. 96. 96. 192. 96. 96. 34%. 85%. 88%. 86%. 90%. 1%. 12%. 14%. ‧. 0.05046. ‧ 國 io. sit. y. Nat. n. al. er. Intercept. (Value-Growth). 學. 1997-2013. 政治大立1997-2013 1997-2004 2005-2013. Ch. engchi. 25. i n U. v.

(31) Table 4 Panel C. Summary of regression results of portfolios of three different CEO confidence levels (1). (2). moderate治 overconfidence 政大2005-2013 1997-2013 1997-2004 立 0.39654*** 0.56753** 0.21993**. 0.01929. -0.09463. (-0.22). (0.10). (-0.53). (2.72). (2.25). Number. 192. 96. 96. 192. Adj-R2. 87%. 87%. 89%. 77%. Nat. 1997-2004. 2005-2013. 0.30793. 0.41438. 0.18588. (2.26). (1.26). (1.02). (0.74). 96. 96. 192. 96. 96. 66%. 95%. 48%. 29%. 73%. Difference(2)-(3) (t-value). er. io. Difference(2)-(1) (t-value). 1997-2013. 1997-2004 2005-2013 a1997-2013 iv l C 0.08860 0.15315 0.03405 n hengchi U (0.41) (0.44) (0.13). 1997-2004. 2005-2013. 0.42542**. 0.54824*. 0.31456. (2.24). (1.70). (1.54). Number. 192. 96. 96. 192. 96. Adj-R2. 16%. 19%. 10%. 1%. 3%. n. 1997-2013 Alpha. y. -0.02888. Alpha. ‧ 國. 2005-2013. ‧. 1997-2004. 學. 1997-2013. excessive diffidence. 26. Difference(3)-(1) (t-value). sit. Excessive overconfidence. (3). 1997-2013. 1997-2004. 2005-2013. 0.33682. 0.39509. 0.28051. (1.21). (0.86). (0.89). 96. 192. 96. 96. 1%. 12%. 17%. 6%.

(32) Table 4 Panel D. Summary of regression results of portfolios intersected by book to market ratio and CEO confidence level 1997-2013 Number. (2). Value+ moderately overconfident. (3). Value+ excessively diffident. (4). Growth+ excessively overconfident. 0.34003 (1.39) 1.11622** (2.52) -0.08593 (-0.51). (5). Growth+ moderately overconfident. (6). Growth+ excessively diffident. 192. ‧ 國. 立. 0.24151 (1.35) 0.34075 (1.14). 0.31457 (0.54). 治 0.45888 政 192 54.60% 大 (1.13) 180 192 192. Nat. 1.76322** (2.64) -0.02811 80.50% (-0.11) 33.40%. io. 0.33435 (1.06) 0.36886 28.20% (0.77) 68.70%. n. Ch. engchi. 27. i n U. v. 2005-2013 Adj-R. 2. 96. 24.80%. 96. 30.70%. 84 96. er. 180. al. 41.10%. 96. y. 0.21782 (0.60). Number. ‧. Value+ Excessively overconfident. Intercept. 學. (1). Adj-R. sit. Intercept. 1997-2004 2. 96. Intercept 0.17442 (0.47). 0.16898 (0.79) 0.37246 17.20% (0.71) -0.17720 80.60% (-0.79) 0.12727 (0.84) 0.19371 19.90% (0.59) 57.50%. Number. Adj-R2. 96. 68.90%. 96. 84.40%. 96. 56.30%. 96. 82.30%. 96. 87.20%. 84. 46.00%.

(33) Table 4 Panel D (continued) 1997-2013. (9). Difference (3)-(6) Difference (2)-(1). (11). Difference (2)-(3). Difference (5)-(4). (14). Difference (5)-(6). (15). Difference (6)-(4). 0.44719 (1.21). 0.03%. 180. 7.24%. 180. 3.96%. 192. al. n. (13). 0.70%. 180 180. (1.75) 0.14431 (0.25) -1.10589* (-1.93) 1.39941** (1.99) 0.36246 (0.90) -0.03451 (-0.06). 16.8%. 13.20%. 96. 7.20%. 84. 0.30% 2.07%. 96 84 84. i n C h2.57% engchi U 13.01%. 28. 0.39697 (0.67). Adj-R. 96. 7.40% 治 (0.26) 政大 1.28559*. 192. io. Difference (3)-(1). (-1.80) 0.87931* (1.83) 0.32744 (1.42) -0.07579 (-0.23). Nat. (12). 立. 168. 0.34268 (0.54) 0.12452. 1.70%. ‧. (10). 192. (0.36) 0.82988* (1.88) 0.12221 (0.35) -0.68240*. Number. y. Difference (2)-(5). 192. Intercept. sit. (8). 0.30375 (0.76) 0.09852. ‧ 國. Difference (1)-(4). Adj-R. 2005-2013 2. 學. (7). Number. 96. er. Intercept. 1997-2004 2. v. 96 96. 10.12%. Intercept 0.35162 (0.81) 0.04171 (0.16) 0.23561 (0.48) -0.00544 (-0.01) -0.20347. (-0.40) 0.19804 11.20% (0.31) 0.30447 17.99% (1.31) -0.07184 2.24% (-0.21) 14.62%. 0.34042 (0.81). Number. Adj-R2. 96. 3.60%. 96. 14.30%. 84. 4.20%. 96. 4.93%. 96. 7.39%. 96. 4.93%. 96. 14.8%. 84. 1.60%. 84. 9.33%.

(34) VI. Conclusion The main goal of this paper is to find how the CEOs’ confidence level influences the value premium. Our analysis consists of two sections. In the first section, we separate the value-growth effect and CEOs’ confidence level and test the performance of portfolios respectively. We find the value stock portfolios experience significantly. 政治大. positive abnormal returns and there is a value premium in the long term which is. 立. consistent with the previous studies. We also find the portfolio of firms managed by. ‧ 國. 學. moderately overconfident CEOs experience significantly positive abnormal return.. ‧. The result is robust and implies CEOs should take risk moderately to increase firm. io. sit. y. Nat. value.. er. In the second section, we incorporate the value-growth effect with CEOs’. al. n. v i n C hinto six groups and confidence level. We classify firms e n g c h i U test the performance of each. portfolio. We find only the portfolio of growth firms managed by excessively overconfident CEOs experience a negative return. The result shows that the poor performance of growth stocks is result from the firms managed by excessively overconfident CEOs. Besides, we also find the portfolio of value firms managed by excessively diffident CEOs earns significantly positive return, however, the portfolio of growth firms earns a positive return when the CEOs are moderately overconfidence. 29.

(35) The result implies firms should hire CEOs matched with their specific characteristics. For value firms, they are permanent in mature and encounter more risk to reduce assets in place. Thus the CEOs should be more conservative when they make decisions. For growth firms, they tend to experience more underinvestment problem than others. Thus the CEOs should take more risk to capture the growth opportunities.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 30. i n U. v.

(36) VII. Reference Basu, S. 1977. 'Investment performance of common-stocks in relation to their price-earning ratios - test of efficent market hyothesis.' Journal of Finance, 32:3, 663-82. Campbell, T. C., Gallmeyer, M., Johnson, S. A., Rutherford, J. & Stanley, B. W. 2011. 'CEO optimism and forced turnover.' Journal of Financial Economics, 101:3, 695-712. Chan, L. K. C., Hamao, Y. & Lakonishok, J. 1991. 'Fundamentals and Stock Returns in Japan.' Journal of Finance, 46:5, 1739-64. Fama, E. F. & French, K. R. 1995. 'Size and Book-to-Market Factors in Earnings and Returns.' Journal of Finance, 50:1, 131-55. Goel, A. M. & Thakor, A. V. 2008. 'Overconfidence, CEO Selection, and Corporate Governance.' Journal of Finance, 63:6, 2737-84.. 立. 政治大. ‧ 國. 學. ‧. Heaton, J. B. 2002. 'Managerial Optimism and Corporate Finance.' Financial Management, 31:2. Hueng, C. J., Pettengill, G. & Chang, G. 2013. 'Comparing Value and Growth Mutual Performance: Bias from the Fama-French HML Factor.' Academy of Economics and Finance Journal:4, 75-86.. y. Nat. sit. n. al. er. io. Lakonishok, J., Shleifer, A. & Vishny, R. W. 1994. 'Contrarian Investment, Extrapolation, and Risk.' Journal of Finance, 49:5, 1541-78. Lang, L. H. P. & Stulz, R. M. 1994. 'Tobin's q, Corporate Diversification, and Firm Performance.' Journal of Political Economy, 102:6, 1248-80. Malmendier, U. & Tate, G. 2005. 'CEO overconfidence and corporate investment.' Journal of Finance, 60:6, 2661-700. Malmendier, U. & Tate, G. 2008. 'Who makes acquisitions? CEO overconfidence and the market's reaction.' Journal of Financial Economics, 89:1, 20-43. Panayiotis C. Andreou, John A. Doukas & Louca*, C. 2011. 'Corporate Diversification and Managerial Overconfidence.'. Ch. engchi. i n U. v. Rosenberg, B., Reid, K. & Lanstein, R. 1985. 'Persuasive Evidencee of Market Inefficiency.' Journal of Portfolio Management, 11:3, 9-16. Schrand, C. M. & Zechman, S. L. C. 2012. 'Executive overconfidence and the slippery slope to financial misreporting.' Journal of Accounting and Economics, 53:1–2, 311-29. Serfling, M. A. 2012. 'CEO Age, Underinvestment, and Agency Costs.' Working paper. 31.

(37) Stambaugh, R. F., Yu, J. & Yuan, Y. 2012. 'The short of it: Investor sentiment and anomalies.' Journal of Financial Economics, 104:2, 288-302. Zhang, L. U. 2005. 'The Value Premium.' The Journal of Finance, 60:1, 67-103.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 32. i n U. v.

(38)