Master Thesis:
CEO Gender and Managerial Earnings Forecast
Student name: Bodine Mooijekind Student number: 10677410
Education: MSc Accountancy & Control, variant Control
Institution: University of Amsterdam, Faculty of Economics and Business
Supervisor: Dr. B. Qin
Date of final version: 18 June, 2015 Word count: 12,737 words
Statement of Originality
This document is written by student, Bodine Mooijekind, 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.
Abstract Purpose
The purpose of this study is to examine the relationship between CEO gender and managerial earnings forecast issuance, bias, and related analyst revisions.
Research design and methodology
This is a quantitative research that examines CEOs from U.S. firms. Data is gathered through databases (ExecuComp, I/B/E/S, Insider Filing Data Feed, and Compustat), and manually through Yahoo using search words.
Findings
This study examines four areas relating managerial earnings forecast and gender: forecast issuance, optimism and pessimism bias, equity ownership, and analyst revisions. I find results that suggest differences between managerial earnings forecast and gender, however regression analysis proved that these differences are insignificant. I can interpret this in two ways: firstly, based on theoretical interpretations, it is possible that there are no differences between female and male CEOs, because female CEOs act like their male counterparts. Therefore, indicating that there is no relationship between gender and managerial earnings forecasts. Secondly, based on empirical interpretations, it is possible that I could not find significant results, because of empirical issues (see limitations). Also, I find no relationship between managerial earnings forecast and analyst revisions, indicating that analyst do not react to the forecasted earnings.
Limitations
The limitations of my study are to be found in the small sample selection. Specifically, in the small sample of female CEOs, small sample range, excluded forecast indicators, the focus on CEOs, and manually collected of data. In addition, limitations are to be found in the limited time available for research.
Originality
This paper contributes empirical to the limited managerial earnings forecast literature, the extensive gender-based literature, and the less extensive analyst forecast literature. In
earnings forecast, and the analyst perspective, if analysts react differently if the managerial earnings forecast is issued by a male or female CEO. Also, this paper has societal value by helping managers, analyst, and other relevant actors by understanding and interpreting
managerial earnings forecast and analyst forecast (revisions). I found no relationship between managerial earnings forecast and gender. Therefore, my study provides evidence that there is gender equality in a CEO population, which provides new insight into beliefs about gender differences. Finally, in contrast to the expectations and findings of prior literature I found no relationship between analyst revisions and managerial earnings forecast.
Key words
Managerial earnings forecast research, managerial earnings forecast bias, managerial earnings forecast issuance, analyst revisions, gender-based research, psychological characteristics, Chief Executive Officer, expectation management, communication, equity ownership
Table of contents
Statement of Originality ... 2
Abstract ... 3
1 Introduction ... 6
2 Literature review and hypotheses development ... 11
2.1 Managerial earnings forecasts as valuable tool... 11
2.2 Biasing of managerial earnings forecasts... 12
2.2.1 First category ‘Expectation management’ ... 13
2.2.2 Second category ‘Communication’ ... 13
2.2.3 Third category ‘Others’ ... 13
2.3 Gender differences and biasing managerial earnings forecasts ... 14
2.3.1 Psychological characteristics ... 15
2.3.2 First category ‘Expectation management’ ... 15
2.3.3 Second category ‘Communication’ ... 16
2.3.4 Third category ‘Others’: Equity ownership ... 16
2.3.5 Psychological characteristics and gender ... 17
2.4 Hypotheses development ... 19
3 Research methodology ... 21
3.1 Population and Sample selection ... 21
3.2 Statistical models and variable definitions... 22
3.3 Regression Specifications ... 28
4 Results ... 29
5 Discussion ... 36
Appendix A Variable definitions ... 38
6 1 Introduction
For many years now, gender differences are of interest to many researchers. These researchers (e.g., Powell and Ansic, 1997; Jianakoplos and Bernasek, 1998; Byrnes et al., 1999; Schubert, 2006; Francis et al., 2009; Cuadrado et al., 2012; Croson and Gneezy, 2009) show that
significant gender differences exist in e.g., conservatism, leadership style, risk and lose averseness, and ethical behavior. Also, gender differences is receiving considerable attention in the work place. To illustrate, the 2009 KMPG survey of 955 companies reveals that more than 61% of these companies have a diversity policy in place to monitor gender, age,
race/ethnicity, and disability (KPMG, 2009).
In addition, a research from Grant Thorton IBR 2014 shows that the proportion of females running businesses across the world has risen to 12% up from 10% in 2013 and just 5% in 2012. So recent years show a strong rise in females at top positions (Grant Thornton, 2014). This rise in females in top positions gave reason to extent the gender differences research to examine gender differences in leadership positions. For example Peni and
Vähämaa (2010) examine female executives and earnings management. They find evidence to suggest that firms with female CFOs are associated with income-decreasing discretionary accruals, thereby implying that female CFOs are following more conservative earnings management. In addition, Dezsö and Ross (2012) argue that female representation in top management brings informational and social diversity benefits to the top management team, enriches the behaviors exhibited by managers throughout the firm, and motivates females in middle management. The result should be improved managerial task performance, and thus better firm performance.
A subject also discussed in prior literature is the managerial earnings forecast. According to Healy and Palepu (2001) managerial earnings forecasts are one of the primary vehicles in the US through which managers voluntarily disclose private information to outside
stakeholders. These managers and investors see benefits for issuing managerial earnings forecast. To explain the benefits, Hirst et al. (2008) state that managerial forecasts represent one of the key voluntary disclosure mechanisms by which managers establish or alter market earnings expectations, preempt litigation concerns, and influence their reputation for
transparent and accurate reporting. As benefit for investors, Trueman (1986) state that managerial earnings forecasts give investors a more favorable assessment of the manager’s ability to anticipate economic environment changes and to adjust production plans
accordingly. Forecast release can thereby translate into a higher firm market value. Also, managerial earnings forecasts help to reduce information asymmetry between investors and
7 managers, which reduces agency costs, and that is beneficial for the investors/owners’ wealth of the related company (Diamond and Verrecchia, 1991; Coller and Yohn, 1997). Thus, managerial earnings forecasts are considered both beneficial for the firm and for the users.
However, there is discussion about the credibility of the managerial earnings forecasts. For example, Conference Board (2003) reports that nearly 40% of investors rate corporate profit projections as not credible. Warfield (2005) finds one of the reasons for such
skepticism. He finds that managers often face incentives that increase the likelihood that they will intentionally misstate actual earnings to meet their forecasts. However, Hirst et al. (2007) state that not all managers with such incentives provide biased earnings forecasts and
subsequently manage earnings to those forecasts.
Managers face psychological and economic motivations to optimistically bias earnings forecasts. For example, Chambers and Windschitl (2004) and Malmendier and Tate (2005a) suggest people are inherently optimistic and Hribar and Yang (2015) show that this optimism extends to management earnings forecasts. Rogers and Stocken (2005) and Hutton et al. (2003) find economic motivations for managers to optimistically bias forecasts. Ironically, Matsumoto (2002) and Cotter et al. (2006) find that managers also have incentives to establish beatable expectations by pessimistically biasing forecasts.
The previous paragraphs showed the importance of managerial earnings forecast in the financial market and the existence of incentives for CEOs to issue upward-bias or downward-bias managerial earnings forecasts. Therefore, it is important to shed more light on what influence managerial earnings forecasting, to better understand and interpret the earnings numbers. Specifically, I find the gender effect important to highlight, because like previously mentioned prior literature documents that gender differences exist in other areas and there is a rise in female CEOs. So the gender effect is of increasing importance. Relevantly, literature describe specific characteristics of a CEO, which can influence how and if a CEO
intentionally or unintentionally bias their managerial earnings forecasts (e.g., illegal behavior with regards to intentional misstate forecasted earnings (Miller, 2009) or overconfident
behavior which result in upward biased forecasts (Hribar and Yang, 2015)). Therefore, for my study I examine whether there exists a gender difference in forecasting of earnings within an CEO population. This lead to the following research question for my thesis: “is there a difference in issuance and biasing of managerial earnings forecast between female and male CEOs, and does analysts react differently to a managerial earnings forecast issued by a male or female CEO?”
8 I answer my research question through four hypotheses that test different areas of the relationship between managerial earnings forecast and gender: forecast issuance, optimism and pessimism bias, equity ownership, and analyst revisions. Data is gathered through databases (ExecuComp, I/B/E/S, Insider Filing Data Feed and Compustat), and manually through Yahoo using search words for CEOs that from U.S. firms. From the population of 2,442 observations I took a sample of 500 observations. I used this sample to test my hypotheses.
I find results that suggest differences between managerial earnings forecast and gender, however regression analysis proved that these differences are insignificant. I can interpret this in two ways: firstly, based on theoretical interpretations, it is possible that there are no differences between female and male CEOs, because female CEOs act like their male counterparts. Therefore, indicating that there is no relationship between gender and
managerial earnings forecasts. Secondly, based on empirical interpretations, it is possible that I could not find significant results, because of my sample size. In addition, I find no
relationship between managerial earnings forecast and analyst revisions, indicating that analyst do not react to the forecasted earnings.
My study has empirical and societal value. Firstly, my study empirical contributes to the extensive gender-based literature and the limited managerial earnings forecast literature. Like I previously mentioned prior literature acknowledged that significant gender-based differences exist e.g., in conservatism, leadership style, risk and lose averseness, and ethical behavior (e.g., Powell and Ansic, 1997; Jianakoplos and Bernasek, 1998; Byrnes et al., 1999; Schubert, 2006; Francis et al., 2009; Cuadrado et al., 2012; Croson and Gneezy, 2009). In addition, prior literature examines some effects of these gender differences on for example firm financial performance, and quality of board of directors, and earnings management (e.g., Dezsö and Ross, 2012; Adams and Ferreira, 2009). However, my study contributes to the gender-based literature , because in contrast with these above mentioned studies I find proof for gender equality in a CEO population.
Furthermore, several studies have focused on the purpose and benefits, credibility, and reasons for biasing managerial earnings forecasts (e.g., Healy and Palepu, 2001; Hirst et al., 2008; Chambers and Windschitl, 2004). In particular, prior literature finds the importance of managerial earnings forecasts, because it is a tool frequently used in the financial market, and it is one of the primary vehicles through which managers voluntarily disclose private
information to outside stakeholders (e.g., Hilary and Hsu, 2011; Healy and Palepu, 2001). However, to my knowledge, there is no research done that explicitly focus on examining if the
9 gender of the CEO has an impact on the biasing and issuance of managerial earnings forecast. Although, there is one study, the study of Huang and Kisgen (2013), that touch upon the relationship between CEO gender and managerial earnings forecasts. They researched whether male executives are overconfident by examining earnings forecasts. They find that earnings forecast made by firms with male executives have significantly narrower
bands/range than those with female executives. They use this finding specifically to find differences of overconfidence between male and female executives. However, they researched only a small part, the managerial earnings forecast precision (i.e. bands), of how CEO gender can influence managerial earnings forecasting. And prior literature finds several other
possible reasons, next to overconfidence, why female or male CEOs might influence managerial earnings forecasting (e.g., biasing of the managerial earnings forecast can be influenced by equity ownership incentives (e.g., Hutton et al., 2003; Aboody and Kasznik, 2000)). Because of the importance of managerial earnings forecast, the existence of gender differences in other areas and presumable in managerial earnings forecasting, and the lack of research in this area, I found it important to examine the relationship between CEO gender and managerial earnings forecasts more in-depth. My study found, in contrast with Huang and Kisgen (2013), no evidence for an association between managerial earnings forecast and gender.
In addition, my study contributes empirically to the analyst forecast literature. Prior literature finds that analysts react to managerial earnings forecasts (e.g., Waymire, 1986; Jennings, 1987). However, to my knowledge, there is no research done that examines the relationship between CEO gender and analyst revisions after managerial earnings forecast issuance. Therefore, in my study I examined whether analyst revisions are influenced by the gender of the CEO. In other words, are analysts aware or do they presume that there is a gender difference between male or female CEOs in managerial earnings forecasting or not. In this way my study looks to both the CEO and analyst perspective related to managerial
earnings forecast, which gives a more broad and clear picture for users of managerial earnings forecasts and analyst forecast about what influence the managerial earnings forecast and analysts forecasts. In particular, the relationship between gender and managerial earnings forecast and if analysts are aware of the gender differences and incorporate these in their forecasts. I contribute to the analyst forecast literature, because I found evidence that proved that analysts do not react to managerial earnings forecast issuance, which is in contrast with the studies of Waymire (1986) and Jennings (1987).
10 Secondly, the societal value of my study is demonstrated by the usefulness of my findings to better understand and interpreted the managerial earnings forecast and analyst forecast (revisions) for managers, analysts, and other relevant users of managerial earnings forecast and analyst forecasts. The findings of this research may be useful to better understand the effects of the gender of the CEO. If there is a difference this might be useful for users of managerial earnings forecasts to keep the gender of the CEO in mind by reviewing the
managerial earnings forecast and analyst forecasts revisions (e.g., which gender is sensitive to act upon incentives to intentional bias managerial earnings forecasts, and does gender specific differences in character of the CEO lead to intentional or unintentional managerial earnings bias). Furthermore, if there is a gender difference than it might be one of the reasons why there is a rise in female CEOs. However, contrasting the expectations I found no differences in gender, therefore I found evidence for gender equality. As result, my study raises questions about beliefs of gender differences.
The remainder of the paper is organized as follows. In chapter 2, I discuss the theoretical foundation of my analysis and develop my research hypotheses. In chapter 3, I describe the research methodology. I present the results of the analysis in chapter 4. Finally, I conclude the paper in chapter 5.
11 2 Literature review and hypotheses development
In this chapter I describe the relevant literature and hypotheses for my research. In paragraph 2.1 I explain why a managerial earnings forecast is a valuable tool for issuers and users. This explanation contributes to my research, because firstly it shows what motives managers to issue managerial earnings forecasts, and secondly it shows that the forecasting of earnings has influence in the financial market. In paragraph 2.2 I go further into managerial earnings forecast biasing. For my research it is important to know the reasons why CEOs bias their managerial earnings forecasts, and why not, in order to define the gender-based differences. Finally, in paragraph 2.3 I go further into the relationship between gender differences and managerial earnings forecasts biasing. In this paragraph I go further into the question if the gender of the CEO has an impact on biasing of managerial earnings forecasts based on literature. As result this leads to my hypotheses.
2.1 Managerial earnings forecasts as valuable tool
King et al. (1990) define management earnings forecasts as voluntary disclosures that provide information about expected earnings for a particular firm prior to the expected reporting date. Although the forecasts may be the consensus of the whole management team, Lin et al. (2005) assume that the CEO has the final say in the team, and thus bear the responsibility of what is stating in their managerial earnings forecasts.
Prior literature examines the motivations for issuing managerial earnings forecasts. For example, Hirst et al. (2008) state that managerial earnings forecasts represent one of the key voluntary disclosure mechanisms by which managers establish or alter market earnings expectations, preempt litigation concerns, and influence their reputation for transparent and accurate reporting. Furthermore, Ajinkya and Gift (1984) find that many of the motivations managers have for issuing earnings forecasts are congruent with those of shareholders. They assume that the issuing of forecasts is largely driven by stock price considerations, with managers issuing forecasts and analysts and investors demanding them to reduce the
asymmetry in information between managers and analysts and investors. Lower information asymmetry is viewed as desirable, because it is associated with higher liquidity (Diamond and Verrecchia, 1991; Coller and Yohn, 1997), and lower cost of capital (Leuz and Verrecchia, 2000).
In addition, prior literature documents the influence of management earnings forecasts in the financial market. For example, Pownall et al. (1993) find that they have been shown to affect stock prices. In addition, Baginski and Hassell (1990) give evidence that they affect
12 analysts’ forecasts. Also many other studies (e.g., Waymire, 1986; Jennings, 1987) find that analysts update their forecasts in response to firms’ earnings forecasts, suggesting the
important role of management forecasts in financial market. The study of Cotter et al. (2006) confirm this, they find that approximately 60 percent of analysts revise their forecasts within five days of managerial earnings forecasts. Also, Trueman (1986) state that managerial earnings forecasts give investors a more favorable assessment of the manager’s ability to anticipate economic environment changes and to adjust production plans accordingly. Forecast release can thereby translate into a higher firm market value.
Following Yongtae and Myung (2012), the above mentioned incentives for issuing managerial earnings forecasts can be summarized in three categories: ‘Expectations management’ to meet or beat market expectations at the time of the actual earnings announcement, ‘Communication’ to convey credible earnings information to analysts and investors, and ‘Others’ (e.g., incentives related to equity ownership).
To conclude, prior literature gives evidence that managerial earnings forecast is an influential tool for managers, and is a useful tool for analysts and investors (financial market). Thus, I can assume that the numbers of the managerial earnings forecast are considered valuable to the financial market. Given the importance of managerial earnings forecast I believe that more insight in what impacts the accuracy and biasing of the managerial forecast is necessary. In the following paragraph I explain the reasons why CEOs are motivated to bias their managerial earnings forecast.
2.2 Biasing of managerial earnings forecasts
Kato et al. (2009) explain that a management can take one of three possible approaches to setting forecasts: firstly, managers can set forecasts in an unbiased manner, based on their best estimates of earrings for the period. Secondly, managers can make downward-biased
(pessimistic) forecasts. Finally, managers can make upward-biased (optimistic) forecasts. In the case of upward-biased forecasts the expected earnings exceed the actual earnings, and for downward-biased forecasts the actual earnings exceed the expected earnings.
Based on the three categories for issuing managerial earnings forecast, mentioned in paragraph 2.1, I describe below the relevant reasons for (un)biasing managerial earnings forecasts.
13 2.2.1 First category ‘Expectation management’
Based on the first category ‘Expectation management’ CEOs issue managerial earnings forecasts to meet or beat market expectations. The market typically rewards firms when actual results meet or exceed analysts’ expectations (Matsumoto, 2002; Cotter et al., 2006; Bartov et al., 2002), and CEOs are likely to attempt to avoid negative earnings surprises, because such surprises generally lead to negative price revisions (Skinner and Sloan, 2002). Thus, CEOs have a strong incentive to avoid negative earnings surprises by either managing earnings upward or guiding earnings expectations downward (Matsumoto, 2002). Kim and Song state that analysts rely heavily on management forecasts in forming their expectations. Also other prior studies show that management forecasts influence analyst forecasts (e.g., Baginski and Hassell, 1990; Cotter et al., 2006;Waymire, 1986; Jennings, 1987). Therefore, CEOs can establish beatable expectations to meet or exceed analysts’ expectations by downward-bias their managerial earnings forecasts (Cotter et al., 2006).
2.2.2 Second category ‘Communication’
Prior literature suggests several motivations for CEOs to communicate with investors through managerial earnings forecasts. For example, Chen et al. (2011) show an increase in analysts’ forecast dispersion following firm’s decision to stop forecasting earnings. To the extent that dispersion in analysts’ earnings forecasts leads to mispricing of shares. Thus, CEOs have an incentive to issue managerial earnings forecasts to reduce forecast dispersion, thereby mitigating the mispricing (Diether et al., 2002). In addition, CEOs can issue managerial earnings forecasts to communicate private earnings information to investors and analysts (Yongtae and Myung, 2012). Owners of a company can request a managerial earnings forecasts to reduce information asymmetry, and thereby reduce agency costs (Coller and Yohn, 1997; Diamond and Verrecchia, 1991). Frankel et al. (1995), Coller and Yohn (1997), and Noe (1999) assume that managers communicate private information with investors
through managerial earnings forecasts and that managers have an incentive to disclose truthful information (e.g., legal liability and reputation costs). Unbiased disclosure is associated with a firm’s endeavor to reduce information asymmetry (e.g., Ajinkya et al., 2005). To conclude, managerial earnings forecasts can be seen as a tool for managers to convey relevant and credible (unbiased) information to the market.
2.2.3 Third category ‘Others’
Next to most important and identified reasons, expectation management and communication, research indicates ‘other’ reasons for CEOs to issue and biasing managerial earnings forecasts
14 than expectation management and communication. For example, researchers (e.g., Hutton et al., 2003; Aboody and Kasznik, 2000) show that equity ownership influence managerial earnings forecast bias. Below I describe this reason more in-depth.
Equity ownership
Equity ownership can be an incentive for CEOs for issuing/biasing their managerial earnings forecasts. Hutton et al. (2003) find that CEOs may intentionally optimistically bias their forecast when their wealth depends on the forecast, such as when they hold exercisable in-the-money. To illustrate, CEOs release good news (upward-biased) forecasts to increase their firms’ stock prices. If these CEOs hold options and their firms’ stock price rises, this will result in a higher cash-out by exercise of these options (in-the-money options). Also, a higher stock price result in higher shareholders’ value, which is beneficial for the shareholders. On the other side, Aboody and Kasznik (2000) identify a CEO issue bad-news earnings forecasts around stock option award periods to temporarily depress stock prices and take advantage of a lower strike price on CEOs’ option grants.
This behavior, to act upon equity ownership incentives, can be seen as insider trading (Weihong, 2010; Rogers and Stocken, 2005). Rogers and Stocken (2005) state that CEOs in anticipation of transacting in their firm’s shares, may wish to release forecasts that are
deliberately misleading to profitably exploit the stock mispricing these forecast might induce. In particular, they find that CEOs have incentives to issue optimistic forecast in anticipation of disposing of stock or options and pessimistic forecasts in anticipation of acquiring stock or options. Such behavior violates the U.S. securities laws governing the release of forward-looking statements and insider trading (Arshadi, 1998).
For my research I will give more insight if gender is one of the things that have an impact on managerial earnings forecast biases. In the following paragraph I explain how gender might impact biasing of managerial earnings forecasts.
2.3 Gender differences and biasing managerial earnings forecasts
I this paragraph I link managerial earnings forecasts bias with psychological characteristics that can differ for each gender based on the three previous identified categories. In addition, I go further into how these psychological characteristics differ for males and females according to prior literature. This results in the hypotheses for my study.
15 2.3.1 Psychological characteristics
Psychological characteristics are important for my study, because it is examined that these are the main issues, next to physical differences, that differ for males and females (e.g., Powell and Ansic, 1997; Jianakoplos and Bernasek, 1998; Byrnes et al., 1999; Schubert, 2006; Francis et al., 2009; Cuadrado et al., 2012; Croson and Gneezy, 2009). I will also use the psychological characteristics to describe the link between gender and the three categories in following subparagraphs. The first psychological prior research identified is overconfidence. A large body of research suggests that people are inherently optimistic about what they can achieve (e.g., Chambers and Windschitl, 2004; Malmendier and Tate, 2005a). Hribar and Yang (2015) show that this optimism extends to management earnings forecasts. They state that the overconfidence of a CEO is positively related to optimistically biasing of managerial earnings forecasts. On the other side, less confident and more risk averse CEOs are likely to downward-bias their managerial earnings forecasts. For example, Hutton (2003) find that CEOs who forecast bad news are likely to be concerned about whether they will be blamed for their firms’ poor earnings performance, or are concerned about litigation risk (if they do not disclose bad news upfront, they can get sued by investors when the bad news comes out later on (Donelson et al., 2012)). Below I link the tree categories with the relevant
psychological characteristics.
2.3.2 First category ‘Expectation management’
CEOs can establish beatable expectations to meet or exceed analysts’ expectations by downward-bias their managerial earnings forecasts (Cotter et al., 2006).
Psychological characteristics and expectation management
Firstly, if a CEO will establish beatable expectations depend on how confident he/she is about the firm future earnings. If the CEO is overconfident about his own and firm performance than he/she will not be motivated to establish beatable expectations, because he/she is confident that the firm will meet or even beat the market expectations. Overconfident CEOs have the tendency to overestimate their ability (‘unrealistic optimism’) (Langer, 1975, Kross et al., 2011). Another psychological characteristic that can have an impact is risks averseness, risk averse CEOs are more likely to downwardly bias their forecasts, because they do not want to take the risk that they will not meet analysts’ expectations (Matsumoto, 2002; Cotter et al., 2006).
16 2.3.3 Second category ‘Communication’
Managerial earnings forecasts can be seen as a tool for managers to convey relevant and credible (unbiased) information to the market (Frankel et al.. 1995; Coller and Yohn,1997; Noe,1999).
Psychological characteristics and communication
Firstly, risk averse CEOs are more likely to disclose bad news upfront, so they cannot get sued by investors when the bad news comes out later on (Donelson et al., 2012). In addition, they are less likely to intentional misstate earnings, because there is the risk that the
misstatements become public which can result in reputation and legal liability costs. Bad news disclosure can be seen as downward-bias managerial earnings forecasts (Kross et al., 2011) . Secondly, CEOs that are overconfident have unrealistic expectations. The information that they disclose is based on these expectations, so the information will be truthful but is unrealistic. Finally, there are other incentives why CEOs do not issue unbiased managerial earnings forecasts, see expectations management and equity ownership.
2.3.4 Third category ‘Others’: Equity ownership
If CEOs have equity of their company than they have an incentive to bias their managerial earnings forecast downwards before option grants and upwards before they exercises options (i.e. insider trading activities).
Gender and equity ownership
Mohan and Ruggiero (2003) and Jordan et al. (2007) found that women in top management positions receive less compensation than their male counterpart. In addition, Hersch (1998) found that women prefer stable, less risky compensation, but when they get risky (variable compensation as stock options), they receive less than men doing the same functions.
However, Khan and Vieito (2013) argue that because females are more risk averse that equity based compensation can be used as an incentives to female CEOs to take risks, but Khan and Vieito do not find evidence that boards are attending to the risk aversion differences between male and female CEOs when they design the compensation packages. So I assume that female CEOs have less equity compensation in relation to male CEOs. In addition, the study of Beams et al. (2003) about insider trading activities find that there are differences between the way men and woman perceive the deterrence variables. Specifically, woman were found to believe their likelihood of getting caught was greater and were more concerned about their
17 relationships with their peers than men. Therefore, I presume that female CEOs are less likely to engage in insider selling activities than male CEOs. In other words, I presume that male CEOs are most likely to act upon the incentive for biasing managerial earnings forecasts (upwards before exercise of option and downwards before grant option) than female CEOs.
2.3.5 Psychological characteristics and gender
In the first and second category, expectation management and communication, I have already mentioned the relevant psychological characteristics that have an impact on managerial earnings forecasts bias. These are: overconfidence, and risk aversion. In this subparagraph I will describe the gender differences for these characteristics.
Gender differences in overconfidence, optimism, and pessimism
Optimism is a part of overconfidence, and pessimism is the opposite of optimism. That is the reason why I specifically investigate gender differences in overconfidence in this
subparagraph.
A large body of psychology literature shows overconfidence as a characteristic of human beings. This literature defines overconfident human beings as: they have the tendency to overestimate their ability (‘unrealistic optimism’), they manifest unrealistically positive self-evaluations (‘better-than-average effect’), they think that they have more control over events than can objectively be true (‘illusion of control’), and they believe that their
knowledge is more precise than it really is (‘miscalibration’) (e.g., Langer, 1975; Hardies et al., 2011; Griffin and Brenner, 2004; Glaser et al., 2004; DellaVigna, 2009).
Overconfidence is likely to occur to those who hold power (Petit and Bollaert, 2012), therefore overconfidence of a human being in a CEO position is likely to occur, because a CEO holds a certain amount of power over the related firm. Other reasons why CEO is likely to be overconfident is, because individuals are the most optimistic about outcomes which they believe are under their control (Langer, 1975), and individuals are more prone to overestimate outcomes to which they are highly committed (Weinstein, 1980). Top corporate managers are likely to satisfy both of above mentioned conditions, because a CEO has the ultimate say about his/hers firm’s big strategic decisions (control over outcome), and a large portion of CEO compensation (stocks and options) depends on how well the company is doing (highly committed to outcome of their corporate decisions) (Malmendier and Tate, 2005b).
A number of studies have documented a gender difference in overconfidence, with men being more overconfident than women in a wide variety of domains. For example,
18 Johnson et al. (2006) find that unprovoked attacks and wars are more often initiated by men than by women, because men are more overconfident about their expectations of success in conflict. In addition, prior literature finds that men are also more overconfident than women in a wide variety of domains related to mathematics, science, and technology (Niederle and Vesterlund, 2007; O’Laughlin and Brubaker, 1998; Pajares and Miller, 1994). According to the status characteristics theory routine assumptions are made about which gender has superior skills and abilities. For example, men are assumed to be better than woman at math (Berger et al., 1977), this can be an indicator why men are more overconfident than woman in domains related to mathematics, science, and technology. Finally, the gender difference in overconfidence in financial areas appears to be quite robust. It has, for example, been shown that women are substantially less confident than men in their investment decisions (Estes and Hosseini, 1988), and it has been suggested that men trade more than women in financial markets because of overconfidence (Barber and Odean, 2001).
To summarize, the above mentioned literature documents evidence that men are more overconfident than woman in different areas. Thus, I presume that male CEOs are more overconfident than female CEOs.
Gender differences in risk aversion
A number of studies show that woman are more risk averse than men in multiple areas. For example, Hersch (1996) finds that women make safer choices than men when it comes to making risky consumer decisions, such as: smoking behavior, seat-belt use, preventive dental care, and having regular blood pressure checked. In addition, Pacula (1997) observers that women exhibit greater risk aversion to drug use than men. Furthermore, Jianakoplos and Bernasek (1998) find, by using U.S. data of household holdings of risky assets, that single women exhibit relatively more risk aversion in financial decision making than single men.
This finding is also documented in the literature for financial matters. For example, Bernasek and Shwiff (2001) find that gender is the most significant factor explaining the allocation of the defined contribution pension to stock. They find men allocate more of their pension fund to stock than women.
To summarize, the above mentioned literature documents evidence that woman are more risk averse than men in different areas. Thus, I presume that female CEOs are more risk averse than male CEOs.
19
2.4 Hypotheses development
This paragraph shows how I develop my hypotheses based on the assumptions made in previous paragraphs.
Summary assumptions form literature review
The table below illustrates the assumptions, which I extract from the previous paragraphs, that I need to in order to develop my hypotheses.
TABLE 1
Summary literature review: assumptions for managerial earnings forecast bias Expectation
management
Communication Others: Equity ownership Female CEOs
Effect for managerial earnings forecast bias*
Less
(over)confident, more risk averse
More risk averse, less
(over)confident
Less equity ownership and less likely to engage in insider selling
activities
- - Less related to grant
option
Male CEOs
Effect for managerial earnings forecast bias*
More
(over)confident, less risk averse
Less risk averse, more
(over)confident
More equity ownership and more likely to engage in insider selling activities
+ + Before grant option:
- * - : downward-biased
+: upward-biased
Forecast issuance
From the literature review I conclude that male CEOs are more overconfident and less risk averse than female CEOs. Hribar and Yang (2015) find that overconfident CEOs, arguable male CEOs, are more likely to issue managerial earnings forecasts. Also, issuance of managerial earnings forecast can be seen as a risk-taking activity, because there is a chance that realized earnings will fall short of the forecasted earnings, which generally lead to negative price revisions (Skinner and Sloan, 2002). Following this reasoning, my first hypothesis is:
20
H1 (forecast issuance): Male CEOs (female CEOs) are more (less) likely to issue a management earnings forecast.
Optimism/pessimism bias
Following the literature review I presume that gender of the CEO has an impact on
managerial earnings forecast biases. Table 1 shows that male CEOs are more overconfidence and less risk averse than female CEOs. This given leads to the following hypothesis:
H2 (Optimism and pessimism bias): Male CEOs (female CEOs) are positively associated with upward-biased (downward-biased) managerial earnings forecasts.
Equity ownership
Table 1 shows that male CEOs have more equity ownership and are more likely to engage in insider trading activities, therefore are more likely to downward-bias their forecast before grant option. This given leads to the following hypothesis:
H3 (equity ownership): Male CEOs (female CEOs) are more (less) likely to downward-bias before grant option.
Analyst reactions
This last hypothesis examines if analysts react differently if the managerial earnings forecast is issued by a male or female CEO. Prior literature finds that analysts react less to forecasts issued by overconfident CEOs (e.g., Hilary et al, 2011; Wong et al., 2014). Therefore, I presume that analysts react less to forecasts issued by male CEOs, because they are more overconfident than female CEOs. This given leads to the following hypothesis:
H4 (analyst reactions): Analysts react less to managerial earnings forecasts issued by male CEOs than issued by female CEOs.
21 3 Research methodology
In this chapter I explain the research methodology. Firstly, I describe the population and sample selection. Secondly, I illustrate the statistical model for each hypothesis. Finally, I describe the regressions specifications.
3.1 Population and Sample selection
I gathered data from ExecuComp, I/B/E/S, Insider Filing Data Feed, Compustat, and manually.
Firstly, I used the ExecuComp database for data about executives working by US firms between 2010-2014, resulting in 44,549 observations. Then, I removed all the non-CEO executives, because my research is based on a CEO population. This resulted in a deletion of 36,769 observations. Secondly, I used the I/B/E/S database to collected data about analyst forecasts and realized earnings. Thirdly, I collected data about option grant from the Insider Filing Data Feed for my remaining CEOs. Furthermore, I used Compustat to collect data about e.g., share prices and gross profit. The use of Compustat, I/B/E/S, and other data unavailability resulting in a deletion of 112 observations, because of data unavailability. Finally, I hand-collected the managerial earnings forecast data.
I narrowed my population range to 2013-2014, because of data availability regarding the hand-collection of the managerial earnings forecast data. Also, I already mentioned in the introduction that there was a rise in female CEOs in these years, so it is more relevant to examine recent data. The narrowing of the range resulted in a deletion of 5,226 observations.
The final number of observations in the population is 2,442 among 1,782 different companies. The data collection of the population is summarized in table 2. From the population of 2,442 I took a sample of 500 observations among 447 different companies. These 500 observations will be used to test hypothesis 1.
I used the 500 observations to hand-collected the managerial earnings forecast data. Therefore, I entered the following search words in ‘Google’: ‘EPS’, ‘guidance’, ‘company name/ticker’, ‘outlook’, and ‘fiscal year’. I limited my search to EPS (Earnings Per Share) forecasts, because this is the largely used forecast indicator, and I/B/E/S data is almost entirely based on EPS. From the 500 observations I found 217 managerial earnings forecasts among 204 different companies for fiscal year 2013 and 2014. These 217 observations will be used for hypothesis 2, 3, and 4. The data collection of the sample is summarized in table 3.
22 TABLE 2
Population selection
Observations
Initial population of all executives (ExecuCom) 44,549
Less:
Non-CEOs (ExecuCom) -36,769
Data unavailability -112
Observations for fiscal years 2010-2012 -5,226
Total removed observations -42,107
Usable amount of observations 2,442
Number of different companies 1,782
TABLE 3 Sample selection
Observations
Initial sample 2,442
Sample (hypothesis 1) 500
Number of different companies 447
Observations of managerial earnings forecast issuance (hypothesis 2, 3, and 4) 217
Number of different companies 204
3.2 Statistical models and variable definitions
Below I show the statistical models of the hypothesis, which include the variable definitions. The variable definitions are summarized in appendix A.
23 Hypothesis 1 (Issuance)
Figure 1 illustrates the statistical model of hypothesis 1.
FIGURE 1
Statistical model hypothesis 1
* GENDER= 1 if the CEO is male and 0 if the CEO is female.
** ISSUANCE=1 if the firm issued at least one forecast, and 0 if otherwise.
*** Following prior literature (e.g., Ajinkya et al., 2005; Kross et al., 2011), I used several independent variables to control for other possible determinants of the properties of management forecasts:
LMVAL = log of the market value of a firm’s common equity at the beginning of the fiscal
period. The prior literature provides evidence supporting the positive association between firm size and management earnings forecasts (e.g., Kasznik and Lev, 1995).
AUDIT = 1 if the company is audited by one of the Big 4 auditors (Auditor-number: 1-7), and
0 otherwise. Auditor reputation could also be a factor in disclosure decisions. Thus, prior research indicates that firms using Big 4 auditors tend to have more disclosure (Lang and Lundholm, 1993).
NUMEST = number of analysts following the firm. Prior research (Lang and Lundholm, 1993
and 1996) documents a positive association between corporate disclosure quality and the number of analysts following a firm.
24
LITIGATE = 1 for all firms with high litigation risk: in the biotechnology (SIC Code: 2833–
2836 and 8731– 8734), computers (3570–3577 and 7370–7374), electronics (3600–3674), and retail (5200–5961) industries, and 0 otherwise (based on Francis et al., 1994). If litigation risk is high, timely bad news disclosures can help firms reduce litigation costs (e.g., Skinner, 1994 and Field et al., 2005).
MTB = ratio of market value to book value of common equity at the beginning of the fiscal
period (market value of equity divided by the book value of equity at the beginning of the fiscal period). I used MTB as a proxy for proprietary costs (Bamber and Cheon, 1998). The amount of proprietary information costs reduce (increase) firms’ motivation to disclose good news (bad news) information (Verrecchia, 1983).
LOSS = 1 if the firm reported (gross profit) losses in the current period, and 0 otherwise. Prior
research suggests that earnings are less value relevant for loss-making firms (Hayn, 1995) and that meeting or beating financial analyst expectations is less important for these firms
(Degeorge et al., 1999). In combination with expectation management a firm is thus less likely to issue a managerial earnings forecasts if the they report a loss.
Hypothesis 2 (Optimism/pessimism bias)
Figure 2 illustrates the statistical model of hypothesis 2.
FIGURE 2
25 * GENDER= 1 if the CEO is male, and 0 if the CEO is female.
** BIAS=[( EPS of managerial earnings forecast − actual EPS)/price at the beginning of
the fiscal period]. If BIAS > 0 than is the earnings forecast optimistically biased, and if BIAS < 0 than is the earnings forecast pessimistically biased. If a company issued a EPS forecast between ranges then I used the mean of the range.
*** I used the following control variables:
LMVAL = log of the market value of a firm’s common equity at the beginning of the fiscal
period. The prior literature provides evidence supporting the positive association between firm size and management earnings forecasts (e.g., Kasznik and Lev, 1995).
AUDIT = 1 if the company is audited by one of the Big 4 auditors (Auditor-number: 1-7), and
0 otherwise. Prior research indicates that firms using Big 4 auditors tend to have better and accurate disclosure (Lang and Lundholm, 1993).
NUMEST = number of analysts following the firm. Prior research (Lang and Lundholm, 1993
and 1996) documents that CEOs tend to issue downward guidance when analyst forecasts are more rather than less optimistic (Cotter et al., 2006).
LITIGATE = 1 for all firms with high litigation risk: in the biotechnology (SIC Code: 2833–
2836 and 8731– 8734), computers (3570–3577 and 7370–7374), electronics (3600–3674), and retail (5200–5961) industries, and 0 otherwise (based on Francis et al., 1994). If litigation risk is high, timely bad news disclosures can help firms reduce litigation costs (e.g., Skinner, 1994 and Field et al., 2005). Rogers and Stocken (2005) find some evidence that managers of firms with higher litigation risk issue less optimistic earning forecast.
MTB = ratio of market value to book value of common equity at the beginning of the fiscal
period (market value of equity divided by the book value of equity at the beginning of the fiscal period). I use MTB as a proxy for proprietary costs (Bamber and Cheon, 1998).
Proprietary information costs reduce (increase) firms’ motivation to disclose good news (bad news) information (Verrecchia, 1983).
LOSS = 1 if the firm reported (gross profit) losses in the current period, and 0 otherwise. Prior
26 that meeting or beating financial analyst expectations is less important for these firms
(Degeorge et al., 1999).
HORIZON = number of days between the forecast date and the fiscal period-end date divided
by 365 (91.25 for quarter). Prior work uses this measure to proxy for greater earnings
uncertainty and the unobservable precision of managers’ beliefs (Baginski and Hassell, 1997). I control for forecast horizon (HORIZON), because its length affects forecast bias, managers likely have weaker incentives for downward guidance over longer than shorter forecast horizons ( Bartov et al., 2002; Richardson et al., 2004). In addition, several studies (e.g., Johnson et al., 2001; Ajinkya et al., 2005) find that management forecasts are less optimistic when they are issued closer to the end of forecast period.
Hypothesis 3 (Equity ownership)
Figure 3 illustrates the statistical model of hypothesis 3.
FIGURE 3
Statistical model hypothesis 3
* TIME_GRANT= 1 if a forecast is made ahead of option grant (<1,5 month ahead) , and
27 ** BIAS=[(EPS of managerial earnings forecast − actual EPS)/price at the beginning of
the fiscal period]. If BIAS > 0 than is the earnings forecast optimistically biased, and if BIAS < 0 than is the earnings forecast pessimistically biased.
*** See control variables of hypothesis 2.
**** GENDER= 1 if the CEO is male and 0 if the CEO is female.
Hypothesis 4 (Analyst reactions)
Figure 4 illustrates the statistical model of hypothesis 4.
FIGURE 4
Statistical model hypothesis 4
* EPS_MEF= the EPS (Earnings Per Share) of managerial earnings forecast.
** ANALYST_REVISIONS= the analyst forecast revision (the revised consensus analyst
forecast less the pre-existing consensus analyst forecast).The revised consensus analyst forecast is the updated consensus forecast following the management earnings forecast (within 15 days). If there is not a revised analyst forecast,
ANALYST_REVISIONS is zero. The pre-existing consensus analyst forecast is the most recent consensus before the management earnings forecast.
*** I used the same control variables as hypotheses 2 and 3. Because these control variable are related do managerial earnings forecast and prior literature links forecast revisions to managerial earnings forecasts (e.g., Waymire, 1986; Jennings, 1987). Also, the size
28 of the company and the number of analysts covering it will dictate the size of the pool from which the estimate is derived. Furthermore, Brown (2001) documents substantial differences between the analyst forecast errors of loss and profit firms. Analysts have greater problems forecasting earnings for loss firms. It is therefore likely that
management’s ability to forecast earnings would be similarly circumscribed for firms making losses.
**** GENDER= 1 if the CEO is male and 0 if the CEO is female.
3.3 Regression Specifications
The specifications of the various (probit) regressions are as follows: Hypothesis 1 (Issuance):
ISSUANCE= β0 + β1 GENDER + β2 LMVAL + β3 AUDIT + β4 NUMEST + β5 LITIGATE + β6 MTB+ β7 LOSS + ε
Hypothesis 2 (Optimism/pessimism bias):
BIAS= β0 + β1 GENDER + β2 LMVAL + β3 AUDIT + β4 NUMEST + β5 LITIGATE + β6 MTB + β7 LOSS + β7 HORIZON + ε
Hypothesis 3 (Equity ownership):
BIAS= β0 + β1 TIME_GRANT * GENDER + β2 GENDER + β3 TIME_GRANT + β4 LMVAL + β5 AUDIT + β6 NUMEST + β7 LITIGATE + β8 MTB + β9 LOSS + β10 HORIZON + ε
Hypothesis 4 (Analyst reactions):
ANALYST_REVISIONS = β0 + β1 EPS_MEF * GENDER+ β2 GENDER + β3 EPS_MEF + β4 LMVAL + β5 AUDIT + β6 NUMEST + β7 LITIGATE + β8 MTB + β9 LOSS + β10 HORIZON + ε
29 4 Results
In this chapter I describe the empirical results from my study. First, I start with the descriptive statistics. Then, I explain the results from the regressions for each hypothesis. All regressions are tested for multicollineairity using the variance inflation factor, as result I detected no multicollinearity issues.
Descriptive statistics
Table 4 shows the descriptive statistics for the variables. All the mean variables are positive. Some results I like to highlight are: firstly, the mean of ISSUANCE indicates that more than the half of the sample does not issue a managerial earnings forecast. Furthermore, the mean of GENDER indicates that there are more male CEOs in the sample, which is in accordance with the presumptions that there are more male CEOs. In addition, the mean of BIAS indicates that in total there are more optimistically biased managerial earnings forecasts, and together with more male CEOs in the sample, leads to my presumption that male CEOs issue more
optimistically biased managerial earnings forecasts.
TABLE 4 Descriptive statistics
Variable Mean
Standard
Deviation Minimum Maximum
25th percentile 50th percentile (median) 75th percentile ISSUANCE 0.4340 0.4961 0 1 0 0 1 GENDER 0.8200 0.3846 0 1 1 1 1 BIAS 0.0006 0.0133 -0.0693 0.0713 -0.0024 -0.0007 0.0008 TIME_GRANT 0.1060 0.3085 0 1 0 0 0 TIME_GRANT * GENDER 0.0922 0.2899 0 1 0 0 0 ANALYST_REVISIONS 0.0080 0.1222 -0.2300 1.44 0 0 0 EPS_MEF 2.80 2.13 -0.2200 16.70 1.34 2.55 3.78 EPS_MEF * GENDER 2.26 1.90 -0.2300 8.22 0.6000 2.05 3.38 Control variables LMVAL 8.21 1.65 1.16 13.29 7.11 8.07 9.21 AUDIT 0.9140 0.2806 0 1 1 1 1 NUMEST 4.41 3.54 0 19 2 4 7 LITIGATE 0.258 0.4380 0 1 0 0 1 LOSS 0.012 0.1090 0 1 0 0 0 MTB 6.01 77.88 -524.36 1542.22 1.65 2.56 4.38 HORIZON 0.6588 0.5839 0.0438 3.40 0.3945 0.5918 0.8493
30 Hypothesis 1 (Issuance)
Table 5 describes the results of tabulate the variables ISSUANCE and GENDER.
TABLE 5
ISSUANCE and GENDER GENDER
ISSUANCE Female Male Total
0 58 225 283
1 32 185 217
Total 90 410 500
Based on table 5, I find that 43.40% (217/500) of the total sample issue at least on managerial earnings forecast in the sample period. From this 43.40%, 14.75 % are female and 85.25% are male. That is 35.55% (32/90) of all the female CEOs and 45.12% of all the male CEOs in the sample. To summarize, tabulate indicates that in this sample, in accordance with my
hypothesis 1, male CEOs do issue more forecasts. However, I need to check if this difference is significant. Because my dependent variable (ISSUANCE) is a dummy variable I use probit regression analysis to check significance.
Table 6 summarizes the results of the first probit regression model. The
constant (Intercept) indicates the value of ISSUANCE in the probit regression equation.
TABLE 6 First regression model
I Intercept -2.51 (0.000)* GENDER 0.2623 (0.107) Control variables LMVAL 0.2657 (0.000)* AUDIT 0.1369 (0.552) NUMEST - 0.035 (0.144) LITIGATE 0.1171 (0.401) MTB 0.0129
31 (0.213) LOSS - 0.4873 (0.426) Number of observations 446 Pseudo R-squared 0.0846 * Significant at 0.05 level
Based on the regression model, I find that the differences between ISSUANCE and GENDER is insignificant at a 0.05 level (0.107>0.05). I do find that ISSUANCE is positively associated with LMVAL (i.e. firm size). Thus, the issuance of managerial earnings forecast can be explained by the size of the company instead of the gender of the CEO. In other words, how bigger the size of the company how greater the likelihood that the company will issue a managerial earnings forecast.
To conclude, tabulating ISSUANCE and GENDER shows that male CEOs do issue more managerial earnings forecast in my sample (9.68% more than female CEOs). However, I cannot prove that this result is significant. Therefore, I do not have evidence for my first hypothesis. I find no evidence that male CEOs (female CEOs) are more (less) likely to issue a management earnings forecast
Hypothesis 2 (Optimism/pessimism bias)
Table 7 describes the results of tabulate the variables GENDER and BIAS.
TABLE 7 GENDER and BIAS
GENDER BIAS<0 BIAS = 0 BIAS>0 Total
Male 111 (60%) 6 (3.24%) 68 (36.76%) 185 (100%) Female 21 (65.63%) 0 (0%) 11 (34.38%) 32 (100%) Total 132 6 79 217
Based on table 7, I find that female CEOs issue 5.63% (68.63-60) more pessimistically biased forecast than male CEOs. However, male CEOs issue 2.38% (36.76-34.37) more
optimistically biased forecast than female CEOs. To summarize, tabulate indicates that in this sample, in accordance with my hypothesis 2, male CEOs do issue more optimistically biased
32 forecasts and females more pessimistically. However, I need to check if this difference is significant. Below I use regression analysis to check significance.
Table 8 column I summarizes the results of the second regression model. The R-squared indicate that approximately 7.54% of variance in BIAS is explained by this model. The constant (Intercept) indicates the value of BIAS in the regression equation.
TABLE 8
Second regression model (I) and third regression model (II)
II III Intercept 0.0045 0.0054 (0.547) (0.474) GENDER -0.0010 -0.0007 (0.677) (0.802) TIME_GRANT 0.0096 (0.241) TIME_GRANT * GENDER -0.0050 (0.564) Control variables LMVAL -0.0008 -0.0011 (0.374) (0.241) AUDIT 0.0047 0.0052 (0.240) (0.195) NUMEST -0.0003 -0.0003 (0.360) (0.470) LITIGATE -0.003 -0.0033 (0.030)* (0.100) MTB 0.0000 -0.0000 (0.274) (0.401) LOSS 0.0078 0.0073 (0.566) (0.589) HORIZON 0.0034 0.0026 (0.040)* (0.134) Number of observations 211 209 R-squared 0.0754 0.0918 * Significant at 0.05 level
Based on the regression model, I find that the differences between BIAS and GENDER is insignificant at a 0.05 level (0.677>0.05). I do find that BIAS is associated with LITIGATE and HORIZON. LITIGATE is negatively associated with BIAS. In other words, companies
33 that are associated with litigation are more likely to issue pessimistically biased managerial earnings forecast. HORIZON is positively associated with BIAS. In other words, forecast with a long forecast horizon are associated with optimistically biased managerial earnings forecasts.
To conclude, I do find a difference in biasing of forecasting between male and female CEOs (female CEOs issues more pessimistically biased forecast and male CEOs more optimistically biased forecasts). However, this difference is not significant. Therefore, I do not have evidence for hypothesis 2. I find no proof that male CEOs (female CEOs) are positively associated with upward-biased (downward-biased) managerial earnings forecasts.
Hypothesis 3 (Equity ownership)
Table 9 describes the results of tabulate the variables GENDER and TIME_GRANT.
TABLE 9
TIME_GRANT and GENDER GENDER
TIME_GRANT Female Male Total
0 29 165 194
1 3 20 23
Total 32 185 217
Based on table 9, I find that only 10.60% (23/217) of the CEOs, that issue at least one managerial earnings forecast, also receive options in fiscal year 2013 or 2014. From this 10.60% is 13.04% (3/23) female and 86.96 % (20/23) male. That is 9.38% (3/32) of all the female CEOs and 10.81% (20/185) of all male CEOs of the sample. This is in accordance with my presumption that male CEOs receive more equity than female CEOs.
I use regression analysis to test if TIME_GRANT is associated with BIAS. Table 8 column II summarizes the results of the third regression model. Based on this table, I do not find an association with TIME_GRANT and BIAS at 0.05 level (0.241>0.05). Therefore, I cannot explain the relationship between TIME_GRANT and BIAS through GENDER, because there is no relationship.
To summarize, I do not find an association with BIAS and TIME_GRANT, which is in contrast with presumption that receiving of options is related to managerial forecast bias. Therefore, the GENDER cannot explain the relationship between BIAS and TIME_GRANT, because I cannot find a relationship. To conclude, I do not have evidence for hypothesis 3. I
34 find no evidence that male CEOs (female CEOs) are more (less) likely to downward-bias before grant option.
Hypothesis 4 (Analyst reactions)
Table 10 describes the results of tabulate the variables ANALYSTREVISIONS and GENDER.
TABLE 10
ANALYSTREVISIONS and GENDER Analyst revisions No revisions Total Male 63 122 185 Female 11 21 32 Total 74 143 217
Based on table 10, I find that 34.10% (74/217) analyst consensus forecasts are revised after the company issued a managerial earnings forecast. So in accordance with my stated presumptions analyst do react to managerial earnings forecast issuance. From this 34.10%, 85.14% (63/74) are reactions to forecast issued by male CEOs and 34.38% (11/74) to
forecasts issued by female CEOs. That is 9.38% (11/32) for all the female CEOs and 10.99% (63/185) for all the male CEOs in the sample. This in contrast with my presumption that analyst react less to managerial earnings forecast issued by male CEOs. However, I need to check if this difference is significant. Below I use regression analysis to check significance.
I use regression analysis to test if there is an association between
ANALYST_REVISIONS and the EPS of the managerial earnings forecasts (EPS_MEF). Table 11 summarizes the results of the fourth regression model. The R-squared indicate that approximately 3.96% of variance in ANALYST_REVISIONS is explained by this model. The constant (Intercept) indicates the value of ANALYST_REVISIONS in the regression
35 TABLE 11
Fourth regression model
IV Intercept 0.006 (0.938) GENDER -0.0057 (0.694) EPS_MEF 0.0013 (0.851) EPS_MEF * GENDER -0.0034 (0.694) Control variables LMVAL -0.0033 (0.732) AUDIT 0.0153 (0.695) NUMEST 0.0052 (0.150) LITIGATE -0.0152 (0.413) MTB -0.000 (0.849) LOSS 0.2697 (0.042)* HORIZON 0.0044 (0.787) Number of observations 209 R-squared 0.0396 * Significant at 0.05 level
Based on this table, I do not find an association between ANALYST_REVISIONS and EPS_MEF at 0.05 level (0.851>0.05). Indicating that analyst do not react to managerial earnings forecast issuance, which is in contrast with the expectations. Therefore, I cannot explain the relationship between ANALYST_REVISIONS and EPS_MEF through GENDER, because there is no relationship. However, I find that LOSS is positively associated with ANALYST_REVISIONS.
To conclude, I do not have evidence for hypothesis 4. I find no proof that analysts react less to managerial earnings forecasts issued by male CEOs than issued by female CEOs.
36 5 Discussion
Research showed the importance of managerial earnings forecast in the financial market and the existence of incentives for CEOs to issue upward-bias or downward-bias managerial earnings forecasts. Therefore, it is important to shed more light on what influence managerial earnings forecasting, to better understand and interpret the earnings numbers. Specifically, I found the gender effect important to highlight, because like previously mentioned prior literature documents that gender differences exist in other areas and there is a rise in female CEOs.
Therefore, for my study I examined whether there exists a gender difference in
forecasting of earnings within an CEO population. This led to the following research question for my thesis: “is there a difference in issuance and biasing of managerial earnings forecast between female and male CEOs, and does analysts react differently to a managerial earnings forecast issued by a male or female CEO?” I examined four areas relating managerial
earnings forecast and gender: forecast issuance, optimism and pessimism bias, equity ownership, and analyst revisions.
My results suggest that there is no association between the gender of the CEO and managerial earnings forecast. Moreover, I do not find significant differences between gender and forecast issuance. On the other hand, I find a positive relationship between firm size and issuance of managerial earnings forecasts. Also, I do not find significant differences between gender and managerial earnings forecast bias. However, I do find that managerial earnings forecast bias is negatively associated with litigation and positively associated with forecast horizon. In addition, I do not find an association with managerial earnings forecast and equity ownership (options grants), which is in contrast with prior literature that find results that equity ownership is related to managerial earnings forecast bias. Therefore, I cannot explain the relationship between managerial earnings forecast bias and equity ownership through gender.
Furthermore, my results suggest that there is no association between analyst revisions and forecasted EPS of managerial earnings forecasts. Indicating that analysts do not react to managerial earnings forecast issuance. Therefore, I cannot explain the relationship between analyst revisions and managerial earnings forecasts through gender. However, I do find that reported loss of a company is associated with analyst revisions.
To summarize, I do find results that suggest differences, however regression analysis proved that these differences are insignificant. I can interpret this in two ways: firstly, based on theoretical interpretations, it is possible that there are no differences between female and
37 male CEOs, because female CEOs act like their male counterparts. Prior literature (e.g.
Hardies et al., 2012; Kumar, 2010; Nekby et al., 2008; Gomez-Mejia, 1983; Smith and Rogers, 2000) confirms this, they find that CEOs are not a random sample of males and females but a well-selected subpopulation. Which could mean that I found evidence to
demonstrate gender equality. Secondly, based on empirical interpretations, it is possible that I could not find significant results, because my sample was too small.
My study has empirical and societal value. Firstly, my study empirical contributes to the extensive gender-based literature, the limited managerial earnings forecast literature and the analyst forecast literature. Secondly, the societal value of my study is demonstrated by the usefulness of my findings to better understand and interpreted the managerial earnings
forecast and analyst forecast (revisions) for managers, analysts, and other relevant users of managerial earnings forecast and analyst forecasts. Also, to get some more insight into gender equality.
The limitations of my study are to be found in the small sample selection. Specifically, in the small sample of female CEOs, the small sample range, the excluded forecast indicators, the focus on CEOs, and the manually collection of data. In addition, the limited time available for research.
I suggest for future research to examine which of my interpretations are right. In other words, is there indeed gender equality in forecasting of earnings. Thus, I suggest that future research will resolve my empirical issues to rule out my empirical based interpretation. For example, by using I/B/E/S guidance database instead of hand-collecting data, focusing on CEO and CFO to involve more females in the population, and included other forecast
indicators (like FFO and revenue). Also, It might be interesting to examine for future research what else can influence managerial earnings forecasts to better understand and interpret the forecasted numbers.
