The impact of CEO overconfidence on mergers and acquisitions : an exploration on the positive roles of managerial overconfidence

(1)

U

NIVERSITY OF

A

MSTERDAM

Amsterdam Business School

MSc Finance, Asset Management track

Master Thesis

The Impact of CEO Overconfidence on Mergers and

Acquisitions:

An Exploration on the Positive Roles of Managerial Overconfidence

Student: Shufan Lei

Student Number: 11411449

Supervisor: Dr. Florian Peters

Date: July 2017

Abstract

This paper serves as a contribution to the growing literature investigating the relationship between CEO overconfidence and Mergers and Acquisitions. The sample of the study contains 2565 completed M&A announcements made by 1013 observed CEOs in 974 U.S. public firms during 2006 and 2016. We proposed an updated continuous measure of CEO overconfidence using the detailed package level stock option information from the Compustat Executive Compensation database. The empirical results confirmed the acquisitiveness of the overconfident CEO and revealed a non-monotonic relationship between CEO overconfidence and the short-term cumulative abnormal returns of the acquirers in M&A for firms with an efficient or independent board.

(2)

Statement of Originality

This document is written by Shufan Lei, 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

(3)

I. Introduction

Mergers and acquisitions (M&A) are transactions that enable companies to change their business conditions through combining or acquiring the ownerships among business entities. The Wall Street Journal reported that 2015 was the biggest M&A year ever. In 2015, the total transaction value of announced mergers and acquisition in the United States broke a new record in M&A history, leading to an increased value of 12% over the previous year (Wall Street Journal, 2015).

Within the realm of mergers and acquisitions, the available literature has determined three main motives for the acquisitions activities. Along with the synergy gains and the agency conflicts, the managerial hubris is also a driving force behind the emergence of mergers and acquisitions (Bruner, 2004). The hubris hypothesis serves to explain the effect of the individual managerial traits on corporate takeover activities (Roll, 1986). However, as quoted from Bruner (2004), “The hubris hypothesis for M&A activity says too much and too

little.” Thus, the behavior finance remains a somewhat nascent field of research. Under the

traditional corporate finance studies, we generally assume managers and investors to be rational, however, this assumption has been relatively disproved in the recent years. In the last decade, we have seen a growing concern in the field of corporate behavior in finance regarding the less-than-rational behavior of managers or investors.

Besides driving the urge to conduct mergers and acquisitions, the hubris theory (Roll, 1986) also provides a possible explanation to the value loss in corporate takeovers. Amongst the literature of mergers and acquisitions, the issue of the value creation for the shareholders has been long discussed. In the study of value creation of M&A, Andrade, Mitchell, and Stanfford (2001) and Moeller, Schlingeman, and Stulz (2005) documented the significant value loss of the acquirers in M&A transactions, and hubris theories have proved once more invaluable in explaining the observed phenomenon. Overconfidence or hubristic managers tend to over estimate the returns and underestimate the associated risks, thus leading to overpayment of the target firms and causing value-destructive acquisitions. Malmendier and

(5)

Tate’s research (2005, 2008) has served to both confirm and extend the hubris theory. The research revealed a significant negative market reaction to M&A announcements made by overconfident CEOs as compared to non-overconfident CEOs.

Previous research on the negative effects of CEO overconfidence gives rise to questions as to why firms still employ these overconfident managers and enable them to make investments and financial decisions. It has been postulated that instead of engaging in solely undesirable decisional biases, overconfident CEOs may also exhibit some positive contributions to firms’ value. Prime examples include Steve Jobs and Mark Zuckerburg, who are often described as overconfident CEOs, are at the same time considered to be pioneers in their fields. A striking example of this overconfident behavior can be seen in the year of 2004 that Facebook acquired WhatsApp for $19 billion. The decision, which was once deemed to be overly optimistic, may indeed be one of the best Facebook purchases ever. In theory, Goel and Thakor (2008) and Gervais et al. (2011) argue that managerial overconfidence aids in resolving the issue of underinvestment. CEOs with a moderate level of overconfidence are less conservative and invest at the first-best level. It also seems to be the consensus that an extremely high level of overconfidence leads to overinvestment and decrease shareholders’ wealth. Campbell et al. (2011) conducted empirical research to first demonstrate the different effects of low, moderate and high levels of CEO overconfidence. The results indicate that, for companies with good corporate governance, relatively low and high levels of overconfident CEOs are more likely to experience forced turnover, while moderately overconfident CEOs are instead not. Furthermore, moderately overconfident CEOs are more innovative and perform better in the competitive scenario (Hirshleifer et al. 2012). They are more diligent in learning about the projects (Gervais, Heaton, and Odean, 2011) and are not necessarily detrimental to the firm value. Prior theoretical studies have suggested a non-monotonic relationship between CEO overconfidence and shareholder wealth. As such, we can assume that there exists an optimal level of executive overconfidence, by which investors would respond positively to the mergers and acquisitions announcements made by such CEOs.

(6)

In order to provide a better understanding of the positive roles of managerial overconfidence and to capture its non-monotonic effect in the M&A context, this research employed a comprehensive sample of 2565 completed M&A announcements which were collected from 974 U.S. public firms with 1013 observed CEOs during 2006 and 2016. Followed the CEO’s option-holding behavior, we proposed an updated measure of CEO overconfidence using the detailed package level stock option information from ExecuComp. The acquisitions deals are extracted from Thomson One. Information concerning the board of directors was obtained from the Institutional Shareholder Services database, and further accounting data was constructed using Compustat data. A short-term event study was also employed to measure the M&A performance of acquiring firms. We then estimated the linear multivariate regression and quadratic regressions to explore the effect of CEO overconfidence on the M&A frequencies and the value creations. In term of the endogeneity issue, we prudently include relevant control variables to reduce the omitted variable bias of the regressions. The correlation matrix and the VIF test were performed to detect the collinearity among the independent variables. In addition, the fixed effect regressions were estimated to account for heterogeneity, and standard errors are robust to heteroscedasticity. In general, our empirical results were robust and coherent within this study.

Consistent with previous literature (see Roll, 1986; Bruner, 2004; Moeller et al., 2005; Malmendier and Tate, 2005; 2008), we found overconfident CEOs made significantly more mergers and acquisitions than rational CEOs. However, in contrast with the prior findings of Malmendier and Tate (2008), we found a concave relationship between the acquirers’ average cumulative abnormal return and the in-the-money stock options held by the CEOs. This evidence indicates that, in the case of mergers and acquisitions, investors’ wealth increases with the level of CEOs’ overconfidence up to a moderate point and decreases with the level of overconfidence beyond a certain threshold. Most strikingly, the significant hump-shaped relationship between CEO overconfidence and market reactions can be only observed in firms with good governance. The result is robust under the different corporate governance measures. In general, the results support the hypothesis that for firms with good governance, executive

(7)

overconfidence improves shareholders wealth through mergers and acquisitions decisions, but in other case destroys firms’ value with the excessive levels of optimism.

This paper serves as a contribution to the growing literature investigating the relationship between CEO overconfidence and M&A. The research sheds light on the positive role of CEO overconfidence and its effects on mergers and acquisitions. Furthermore, it provides a possible explanation to the paradox of firms consistently hiring overconfident CEOs. Specifically, we updated the overconfidence measures based on a more accessible and detailed data option and examined the most recent effect of managerial overconfidence on M&A. Our analysis also complements the works Malmendier and Tate (2005, 2008) by exploring the positive effects of CEO overconfidence and extends the theory of Goel and Thakor (2008), Gervais et al. (2011), and Campbell et al. (2011) into a mergers and acquisitions context. As such, to the best of our knowledge, this paper is the first to document the nonlinear relationship between the CEO overconfidence and the market reactions of the acquirers in M&A.

The remainder of the paper is organized as follows: Section 2 presents the relevant literature reviews and proposed the research hypotheses. Section 3 discusses the research methods. Section 4 gives the description of the data collection and demonstrates an overview of the summary statistics. Section 5 and Section 6 show the empirical results and the robustness check. Section 7 concludes and comment on the limitations of the study as well as the direction for the future research.

(8)

II. Literature Review and Hypotheses

In this chapter, literature relevant to the research will be reviewed. Firstly, the value creation of the mergers and acquisitions will be discussed. Following this, an explanation of the overconfidence concept from a psychological viewpoint will be offered. To this end, the hubris theory and the overconfidence in the content of corporate finance will be discussed. This section will be concluded by an assessment of the measures of overconfidence and the hypotheses derived from the literature.

2.1. M&A and Value Creation

Mergers and acquisitions transactions are conducted when two or more entities transfer or combine their ownerships. Numerous studies have documented the economic and financial significances of M&A transactions. Firms conduct M&As to increase the market power, expand the businesses, and benefit from the economies of scale and/or scope. All this, along with the expectation of synergy effects, self-serving attempts to correct the agency problem and the managerial hubris also drive the occurrence of M&As (Bruner, 2004).

The value creation for the shareholders and the distribution of the gains among participants are the primary concerns of M&A research. To this end, a short-term event study has been widely adopted as the most reliable evidence to study the value creation or destruction of the shareholders in mergers (Andrade, Mitchell, and Stanfford, 2001). In the study of value creation in M&A, Andrade et al. (2001) conducted the study on 4300 mergers in the USA during 1973 and 1998. The result indicated the average cumulative abnormal return over a three-day event window for the combination of both participants was significantly positive (1.8% on average). This may suggest the significant value creation of M&A transactions, however, it is worth noting that returns are disproportionally distributed among the targets and acquirers. Within the study, the three-day aggregated abnormal return for the target firms is 16%, while this average abnormal returns for acquirers is -0.7%. Moeller Schlingeman and Stulz (2005) also verified the value loss of acquirers’ shareholders

(9)

when the M&A was announced. They examined public firms who made M&As from 1981 to 2001 and found that the shareholders of the acquiring firms lost an average of $25.2 million in M&As. The result remains stable over the different time period. Furthermore, Andrade et al. (2001) and Moeller et al. (2004) also noted the effects of financing payment and the firms’ size on the value creation. In general, from an acquirers’ perspective, financing with equity generated a lower return to the acquiring firms than financing with cash in M&As (Andrade et al. 2011). As for the size effect, the average announcement return of small acquirers is around two percent higher than larger acquirers (Moeller et al. 2005).

The destruction of value for the bidding firms indicates the overpayment for the targets in M&A due to the overvaluation. One of the explanations for the excessive payment is the misevaluation due to the inefficient stock market (Shleifer and Vishny, 2003). This theory contrasts with the hubris theory, in which it is hypothesized that the markets are efficient while the managers are irrational. Amongst all the personal characteristics of the managers that might influence the stock reaction, managerial overconfidence seems able to play a role in explaining the relevant business decisions. In this research, the hubris theory will be used as a baseline to construct the research of the effect of managerial behaviors on M&A performances. To further understand the hubris theory, the overconfidence concept will be discussed first from the psychological point of view.

2.2. Overconfidence Definition

Overconfidence can be defined as a miscalibration of probabilities. Individuals tend to believe themselves to be superior to average. A survey conducted by Svenson (1981) indicated that individuals exhibited a tendency to believe themselves as being the better driver of the group. This phenomenon also refers to “better than average effect”. In another case, an individual shows an over-estimation of their predictions. People overvalue their information and predict in a narrow confidence interval. Alpert and Raiffa (1982) provided strong evidence of this over-precision behavior. In their experiments, people were asked to give the estimation interval for several quantities based on their own knowledge. The report found that the

(10)

estimated ranges that participants predicted are much narrow than the realized ranges. Only 33.4% of the observation fell within the 50% estimated confident interval. This is also known as the overconfidence effect. That is to say, overconfident people overestimate the accuracy of the information they have, and undervalue the risk of the outcomes. The overconfidence bias is different from optimism bias. Optimism is related to the tendency in individuals’ belief that the odds of favorable results will happen in the future. They believe that the results are more likely to be favorable than they actually were. Weinstein (1980) carried the experiments to understand individuals’ beliefs of themselves in terms of positive and negative events in the future. The results demonstrate that people tend to believe that superior situations are more likely to occur to them than inferior situations. Thus, optimism is often defined as the overestimation of the mean. Ben-David et al. (2007) distinguished between the overconfidence and optimism and investigated their effects on corporate policies. In this paper, they defined overconfidence as a general miscalibration in beliefs, and optimism as overestimation of the mean of expected returns.

Even though the terms of overconfidence and optimism are treated as distinct concepts, in the recent finance literature regarding the managerial behavior, these two terms are often used interchangeably. Malmendier and Tate (2005, 2008) give overconfidence and optimism different definitions, in which they argue that “overconfidence” represents the overestimation of outcomes related to own abilities and “optimism” represents the overestimation of exogenous outcomes. For the purposes of this research, we will follow the definition from Malmendier and Tate (2005, 2008), unless otherwise indicated.

2.3. Managerial Overconfidence and Mergers & Acquisitions

In the studies of CEO’s personal traits in effects of the firm investment decisions, the hubris hypothesis was first recognized by Roll (1986) to explain M&A activity in managerial psychology (Bruner, 2004). The research explains the takeover phenomenon in which bidding firms overpay the target firms on account of the managerial hubris. The hubris theory explains the common phenomenon in M&A in which target firms seem to gain, while acquirers tend to

(11)

lose. An overconfident CEO tends to overestimate their own managerial abilities and the synergy gains created from the M&A transactions (Roll, 1986). Due to the auction-like transaction of mergers, the bid goes to the firm that provides the highest price. Since the estimated price of the overconfident manager is higher than the intrinsic value of the bid, the winner overpays for the deal and thus loses money (Thaler, 1988). This phenomenon is also referred as the winner’s curse. After proposing the hubris theory, Hayward and Hambrick (1997) examined the relationship of CEO’s hubris and the premium paid in acquisitions. The study found the strong positive relationship between four CEO overconfidence indicators and the takeover premium. Moreover, the effect of CEO hubris on the magnitude of acquisition premium paid by acquirers is strengthened if the hubris CEO lack of sufficient monitor from the board of the directors.

Building on hubris theory (Roll, 1986), Malmendier and Tate (2005) argued that managerial overconfidence might cause the distortion of corporate investments due to the sensitivity to cash flow. As an overconfident CEO tends to overestimate the firm value and believes the market undervalues their firm stocks, they consider the equity finance overly costly. As a result, overconfident CEOs tend to overestimate the return, leading to overinvest when cash is abundant and cut the budget when firms are cash constrained. To further understand the effects that CEO overconfidence has on value creation for the acquirers in M&As, Malmendier and Tate (2008) investigated 477 public firms in U.S. from 1980 to 1994 and analyzed the market reaction of the bidding firms when the acquisitions were announced. The study first used the “LongHolder” of the option as a proxy for CEO overconfidence and concluded that overconfident CEOs conducted acquisitions more frequently in firms with abundant cash and created less value than those more pragmatic CEOs. The odds of making the acquisitions by “LongHolder” CEOs are 0.195, which is 1.65 times of the odds for the rational CEO (0.118). The research also found a significantly more negative effect (-90 basis points) of the overconfident executive on the acquirer’s stock market response to the M&A announcements than the effect (-12 basis points) of the non-overconfident executive. Furthermore, “LongHolder” CEOs are particularly likely to diversify their acquisitions. In

(12)

addition, Malmendier and Tate (2008) employed the media portrayal to measure the CEOs’ personality, thereby confirming the results. While the existing empirical research of CEO overconfidence and M&A are mostly conducted within the U.S. market, Ferris, Jayaraman, and Sabherwal (2013) extended the research of Malmendier and Tate (2008) to an international horizon and concluded that CEO overconfidence is, in fact, an international phenomenon.

2.4. Positive Role of Managerial Overconfidence

As previously mentioned, early research has documented the negative effects of managerial overconfidence. This brings us to the paradox of firms consistently hiring overconfident CEOs emerges. Instead of simply making discretionary investments, as in the case of more rational CEOs, overconfident CEOs may also contribute some positive contributions to firms’ value. That being said, the existing studies have shown that the benefits for shareholders of having an overconfident CEO are only confined to moderate levels of overconfidence.

Goel and Thakor (2008) were convinced that overconfident CEOs could benefit shareholders because they are less risk-averse. Intuitively, we expect the rational CEOs to be unbiased and preferred by firms, whilst, a rational CEOs might underinvest and pass over valuable investment opportunities. The authors proposed that the underinvestment of rational CEOs reduces firm value. With a moderate level of overconfidence, CEOs will exhibit a lower cutoff of the signal in terms of assessing investment portfolios. Consequently, they tend to underestimate the risk and overestimate the precision of the information; this may lead to the acceptance of riskier but value-enhancing portfolios and as a result, enhance firm value. The value improvements by overconfident CEOs will be more pronounced in riskier industries, however, excessive overconfidence will eventually lead to an overinvestment problem and result in the value-destroying investments. A similar study conducted by Gervais, Heaton, and Odean (2011) confirmed and completed these findings. As hubristic CEOs are more prone to overestimate the precision of the personal information and presume a favorable outcome than reality, they will put more effort to learn about the risky project. Thus,

(13)

overconfident managers can be beneficial to firms in solving the agency problem, however, an extreme level of overconfidence deteriorates the value creation. In line with Goel and Thakor (2008) and Gervais et al. (2007), Hirshleifer et al. (2012) believed that, besides the value erosion and shareholder’s value destruction caused by the CEO’s optimism in M&A as discussed in prior research, overconfident CEOs are more innovative and respond to the investment opportunities more effectively. They found that firms with overconfident CEOs pursue risky projects. Moreover, it has been proved that overconfident CEOs are superb at scouting for investments with growth opportunity, and further translating into firm value. Thus, they do not necessarily destroy firm values. The results indicate that overconfident CEOs tend to perform better in the risky and growth industries, hence providing an explanation as to why firms hire overconfident managers. The research of Galasso and Simcoe (2011), also effectively demonstrated that CEO overconfidence could increase firms’ innovation, making them more attracted to riskier and more innovative companies.

Campbell et al. (2011) were the first to conduct the empirical research to document the effects of different CEOs overconfidence levels. They used executive compensation data from Execucomp database during 1992 and 2005 to study the CEO overconfidence and forced turnover rate. They hypothesized that those CEOs with a moderate level of overconfidence invest at the optimal level that improves shareholder value. The hypotheses are grounded in the theoretical work of Goel and Thakor (2008) that overconfidence can be compensated by the risk aversion. But with sufficient overconfidence, the risk-averse CEOs will engage in the empire-building behavior. As a result, relatively low or high levels of overconfident CEOs have a higher tendency of being fired than moderately leveled overconfident CEOs. This research found a significant positive relationship between the high and low optimism and the probability of forced turnover in the firms with good governance.

However, it is worth noting that the benefits of the biased managerial executive are highly associated with the characteristics of the industry, firm, the board of the directors, and the personal traits of the manager (Lobão, 2016). Goel and Thakor (2008) seem convinced that the effectiveness of the board is crucial in the effect of overconfident CEO. For instance,

(14)

the concave relation between overconfidence and firm value can be observed only if the board acts in the best interest of shareholders (Campbell et al., 2011). Other recent empirical studies of CEO overconfidence emphasize the relations between CEO overconfidence and corporate financial policies (Malmendier, Tate, and Yan 2011), the link between overconfident CEO and the corporate debt maturity structure (Huang et al., 2016), and its subsequent impact on compensation structure (Humphery-Jenner et al., 2016) etc.

2.5. Measures of CEO Overconfidence

A key factor of the research is the question of how to measure the CEO overconfidence. Within the available literature, the majority associates the measure of overconfidence with the idiosyncratic risks that the CEOs expose themselves to. A risk-averse CEO will diversify their portfolios to hedge the risk of firm’s performance. As CEOs are commonly compensated with firm’s stock options, their decisions towards with regard to their own firm’s stock options will reflect their beliefs in the company. Malmendier and Tate (2005) proposed that unbiased CEOs would exercise the option, as soon the funds are available in order to mitigate risk. Unlikely, overconfident CEOs overestimate their abilities and hold onto the belief that the firm is undervalued. Because of the high expectation of future returns, overconfident CEOs are inclined to hold the in-the-money options longer, even until the expiration date. Based on these arguments, Malmendier and Tate (2005) first constructed two measures of managerial overconfidence based on CEO’s personal portfolio decisions: Holder67 and LongHolder. The Holder67 measure says that the CEOs are overconfident if they held the vested options with the moneyness above 67% at least twice during the sample period. The “LongHolder” categorizes CEOs as overconfident if they ever held options over the duration, provided that option is at least 40% in the money. The third measure used in Malmendier and Tate (2005) is the Net Buyer. Overconfidence CEOs are not only holding the options longer, they may even exhibit the tendency to increase the holdings of firm’s stock. The validity of Net Buyer measures is established on this assumption. Malmendier and Tate (2008) applied the same measures of the LongHolder and Holder 67 in their study into the effects of CEO

(15)

overconfidence in acquisitions. In addition, they introduced the media portrayal measure. They characterized the CEO’s traits based on the descriptions in journals and other publications. If the press used the words “confident”, “optimistic”, “idealistic” etc. the CEO was defined as overconfident. A more thorough discussion of the measurement can be found in Campbell et al. (2011). The study summarized the previously proposed proxies of CEO overconfidence and adopted four sets of optimism measures: the stock option holding measure, the media-based measure, the net stock purchases measure, and the measure based on firm investment level. Additionally, in the study of overconfidence and early-life experience, Malmendier, Tate, and Yan (2011) extended the analysis to 2007 based on data from Conpustat’s Execucomp dataset. Other measures include the forecast error documented by Lin et al. (2005) and self-attrition measures proposed by Li (2010).

In this research, we adopted the data from compustat’s Execucomp and constructed an updated calculation of the option moneyness based on the formula as proposed by Campbell et al. (2011). The details of the formulations of the overconfident variables will be discussed in the methodology section.

2.6. Hypotheses Development

This research is devoting to investigating the effects of overconfident CEOs in M&A. Despite numerous prior studies having addressed the effect of executive overconfidence on the company’s investment decisions and performances, there is a relatively limited amount of research, which discusses the positive effects of CEO overconfidence. This study aims to add some insights to the subject by taking the positive role of CEO overconfidence into consideration. As discussed in the literature reviews, the positive role of overconfidence will be mainly confined to a moderate confidence level (Goel and Thakor, 2008; Gervais, Heaton, and Odean, 2011; Campbell et al., 2011; Galasso and Simcoe, 2011; Hirshleifer, Low, and Teoh, 2012). In addition, mergers and acquisitions have always been a popular issue of research. With the application of event study methodology, the effect of CEO overconfidence on corporation’s M&A performance can be further verified. Also, an updated measure of the

(16)

overconfidence using detailed package level stock option information from ExecuComp will be constructed. And it enables us to first examine the proposed effect of CEO overconfidence on M&A intuitively.

Since overconfident CEOs have the tendency to overestimate the synergy effect within mergers and acquisitions, they are more likely to undertake M&A deals than non-biased CEOs. They tend to overvalue the investment opportunity and have a low threshold for the signal to accept the deals (Brown and Sarma, 2007; Goel and Thakor, 2008). However, the overconfident CEO perceives external financing as unduly costly (Malmendier and Tate, 2005, 2008). In firms with relative constraint internal cash flow, an overconfident CEO might pass over the acquisitions and invest less. As a result, these two counter effects of CEO overconfidence within the domain of M&A frequency is ambiguous. Empirically, Malmendier and Tate (2008) proved that overconfident CEOs have a higher probability of conducting M&As than rational CEOs. Furthermore, according to Billett, Matthew and Qian (2008), due to the self-attrition bias, acquirers who became more overconfidence after making successful acquisitions are more likely to make sequent acquisitions afterward. This evidence demonstrates the tendency for overconfident CEOs to conduct M&A more frequently than non-overconfident CEOs. Therefore, we hypothesize that overconfident CEOs are more acquisitiveness than the rational counterpart.

Hypothesis 1: Overconfident CEOs conduct more mergers and acquisitions than rational

CEOs.

To continue developing the research based on the first hypothesis, the research will examine the market reactions to the M&A announcements that were conducted by the overconfident CEOs. Due to the tendency of overestimating the mergers returns, overconfident CEOs might overpay the targets and lead to value destructive investments. The phenomenon is in accordance with the winner curse. Thus, we will then test the value creation of the overconfident CEOs in M&A. Based on the hubris theory; we would presume a

(17)

negative reaction of the acquirers’ stock prices to the M&A announcements made by overconfident CEOs.

Hypothesis 2: CEO overconfidence affects the cumulative abnormal returns of the acquirers’

stock surrounding an acquisition announcement negatively.

Overconfident CEOs are more willing to devote themselves to challenging and risky projects. They can respond more quickly to investment opportunities (Hirshleifer, Low, and Teoh, 2012). In light of the fact that overconfident CEOs perform more outstanding in the competitive scenario (Hirshleifer et al. 2012) and are more committed to learning about the projects (Gervais, Heaton, and Odean, 2007), this would seem to indicate that they might not be detrimental to firm value. Moreover, when compared to rational CEOs, moderately overconfident CEOs are less conservative and as a result invest at the first-best choice (Goel and Thakor, 2008; Campbell et al., 2011). That being said, CEOs with a middle-level overconfidence can improve the investment levels, firm value, and will be less likely to encounter a forced turnover. Nevertheless, extremely high levels of overconfidence often lead to overinvestment and decrease the shareholders’ benefits. The earlier studies suggested a non-monotonic relationship between CEO overconfidence and shareholder wealth. As such, we can assume that investors’ wealth will increase with the level of CEOs’ overconfident up to a moderate level and thereby decrease when the level of overconfidence surpasses a certain threshold. Hence, the benefits received by the shareholders grow with diminishing rate and appear a concaved relationship between the acquirers’ stock returns and the levels of CEO overconfidence. Nevertheless, the positive effects of overconfident CEOs to the value of shareholders by mitigating the underinvestment and agency costs are held with the precondition that the CEOs are well monitored and the boards act in the interest of shareholders (Campbell et al., 2011). To see if this mechanism is tenable in the content of the M&A, we form the following hypothesis:

Hypothesis 3: For firms with the good governance, a moderate level of CEO overconfidence

(18)

positively, while a low or a high level of CEO overconfidence affects the returns negatively.

III. Methodology

As the primary objective of the research, this study is centered on exploring the relationship between CEO overconfidence and acquiring firms’ M&A performance. To fulfill this objective, in this section, the event study methodology will be discussed first, followed by the discussion of the overconfidence measures. Afterward, the variables to form the regressions analysis will be constructed. In the end, the regressions will be developed in order to test the proposed hypotheses.

3.1. Event Study Methodology

To explore the short-term effect of CEO overconfidence on acquirers’ stock price in mergers and acquisitions, an event study will be adopted to test the market reaction of the stock returns around the announcement date of M&A. In the event study, the market returns and firms’ stock returns were used primarily in the estimation window so as to predict the expected return in the event window, after which the abnormal return for stock prices of firms in the research sample was calculated. The event study methodology followed the guidance from the “Event Studies in Economics and Finance” (A. Craig Mackinlay, 1997) and undertook the market model analytical procedure.

The market model, which was used to calculate return, is presented as follow:

𝑅_!,! = 𝛼_! + 𝛽_!𝑅_!,! + 𝜀_!,! (1) 𝐸 𝜀_!,! = 0 𝑣𝑎𝑟 𝜀_!,! = 𝜎_!!_!

𝑅!,!= The return of security i for period t

R_!,!= The return of the S&P 500 Index market portfolio for period t α! = The intercept coefficient for the market model

(19)

𝜀_!,! = The residual error term on security I for period t

As for the measuring and analyzing abnormal returns, the results is calculated through the following equation:

𝐴𝑅_!,! = 𝑅_!,!− 𝛼_! − 𝛽_!𝑅_!,! (2) τ = Event time τ; 𝜏 = 0 as the event date

AR_!,! = The abnormal return of security i for event time τ

The abnormal return is the disturbance term of the market model calculated from the equation (1). The conditional variance of the abnormal return can be formulated as follow:

𝜎! _𝐴𝑅 !,! = 𝜎!!!+ 1 𝐿_! 1 + 𝑅!,!− 𝜇! ! 𝜎_!! (3)

In this research, the estimation window is L1 = 130 days, and the event window is L2 = 21

days in length respectively, 10 days proceed and succeed the announcement date. Due to the sufficiently large estimation window, the calculation can be simplified by assuming that the second component of the variance equals zero.

To aggregate the observed abnormal return, for an individual cumulated abnormal return over (𝜏1, 𝜏2) T1 < 𝜏1 ≤ 𝜏2 ≤ T2:

𝐶𝐴𝑅_! 𝜏_!, 𝜏_! = !! 𝐴𝑅_!,!

!!!! (4) For each event period, τ = T_!+ 1, … , T_!, the average abnormal return is calculated through equation (2) for all 11 days. And for all N firms, the Average abnormal return on day τ:

𝐴𝑅 ! = 1 𝑁 𝐴𝑅!,! ! !!! (5)

Finally, the average abnormal returns can also be aggregate over the event window to form cumulative average abnormal returns:

𝐶𝐴𝑅 𝜏_!, 𝜏_! = 𝐴𝑅_! !!

!!!!

(20)

𝑉𝑎𝑟 𝐶𝐴𝑅 𝜏_!, 𝜏_! = 𝑣𝑎𝑟 𝐴𝑅_! !!

!!!!

(7)

Therefore the significant of the abnormal returns can be test using:

𝜃 = 𝐶𝐴𝑅 𝜏!, 𝜏!

𝑉𝑎𝑟 𝐶𝐴𝑅 𝜏_!, 𝜏_! !/! ∼ 𝑁 0,1 (8)

3.2. CEO Overconfidence Measures

To quantify CEO overconfidence, for the media-based measure, the hand pick of the information will be required. The magazines, news, articles, and the journal of the CEO’s portrayal can be found in LexisNexis, however, as the data collection is both tedious and somewhat biased, as such this measure was not utilized within this research. Additional earnings per share forecast also serve as a valid measure of CEO confidence and the data is more accessible than the press publications. Earnings forecasts and the actualities can download from I/B/E/S through WRDS. In this research, the option-based measure was mainly adopted. For the option-based measure, Execucomp provided data of U.S. executive compensation. This data alongside the CEOs’ option compensation can be used to formulate the measurement.

Grounded on the assumptions made by Malmendier and Tate (2005), we can construct the proxy of CEO overconfidence using the executives’ decisions about their companies’ stock options. The measure is valid due to the features of the option compensation. As the executives are banned from short selling their own firm’s stock options, holding company shares and options exposes them to the firm-related risk. Therefore, Malmendier and Tate (2005,2008) proposed that rational and risk-averse CEOs should exercise their companies’ stock options once the options are vested and in the money, while, overconfident CEOs tend to hold in-the-money options much longer. Additionally, Malmendier and Tate imposed the threshold on the moneyness of the option that CEOs were holding. The "LongHolder” measure can only contain information regarding CEOs overconfidence wherein the option is in the money.

(21)

Malmendier and Tate (2005, 2008) constructed their overconfidence measures based on detailed option information from 1980 to 1994, which were not available in Compustat Executive compensation database until 2006. Early research, like that of Campbell et al. (2011) and Hirshleifer, low and Teoh (2012), constructed the measure of overconfidence using the average moneyness of the option as the approximation. Given that the detail package-level exercise price and the duration of the options are not available at that time, the options’ moneyness was calculated by first generating the average exercise price of the option. In this section, the detail of how to construct the measure of the overconfidence based on the prior approximation measure will be elaborated first, followed by an updated of the measure based on the detailed option information

3.2.1 Overconfidence measures – Pre-2006

The moneyness of the option is the ratio of the current stock price over the strike price of the corresponding stock option (Malmendier and Tate, 2015). In order to construct the measure, the average strike price will be calculated first. In Wharton Research Data Services, the value of the in-the-money vested option at the fiscal year end in the pre-2006 format is estimated by multiplying the difference between the close price of the underlying stock at the year-end and the exercise price of the option with the total number of unexercised options. Therefore, we can derive the average intrinsic value of the option, which is the underlying stock’s price minus the strike price of the call option, by doing the reverse calculation. After having the average intrinsic value of the option, we can obtain the average exercise price of the option by subtracting the intrinsic value of the option from the stock price. Hence, the average moneyness of the options is the stock price at the fiscal year end divides the estimated average exercise price of the options (WRDS, 2017). The formulas to generate the overconfidence measures using the aggregated average options information can be found in Appendix I.

3.2.2 Overconfidence measures – Post-2006

From 2006, the Financial Accounting Standards Board implemented the FASB 158. The new financial accounting standard FAS123R required to report of the employee’s stock options in

(22)

the financial statements. With the detailed package level of information beginning from 2006, we can update the measure by using the most recent data. The detailed information about the stock options is contained in the Compustat Executive Compensation - Outstanding Equity Awards database.

Before constructing the moneyness of the option based on the updated data, a comparison of the data contained in the Annual Compensation and Outstanding Equity Awards tables were made first to give an overview of the variables among the datasets. For simplification, we assume that in each year a CEO holds j different vested options packages with the strike price of 𝑘_! and the number of the option 𝑛_!. Additionally, the aggregate total number of the option that the CEO holds is N and the aggregated value of the option is V. The stock price at the year-end is assumed to equal to p. The corresponding variables in the Execucomp database are described in Table 1.

To check the consistency among pre- and post- FAS123 regime data, the following variables are constructed from the OutstandingAwards table to match the variables in the AnnComp table. 𝑛! ! !!! = N ……….……….…..………(1) [𝑛!∗ (𝑝 − 𝑘!) ! !!! ] = V………..……….………(2) [ !! !! ! !!! ∗ 𝑝 − 𝑘! ] ! !!! = ! !……….………(3) [ !! !! ! !!! 𝑘_!] ! !!! = 𝑝 − ! !………..………(4) ! [ !! !! ! !!! !!] ! !!! − 1 = !!!! [!!!!] ! !!! ! !! ! !!! − 1 = ! !! !_!− 1 ……...……….………(5)

The left sides of the equations are calculated based on new detailed data delivered in the Outstanding Equity Award table, and the right sides of the equations are computed based on aggregated data that are reported in AnnComp table. Equation (1) and (2) indicates that, the sum of the number of vested unexercised options held by the executive at fiscal year-end equal to the aggregate number of unexercised options, and the estimated aggregate value of

(23)

in-the-money equal to the sum of intrinsic value (𝑝 − 𝑘_!) times the number of vested unexercised options. In order to align with the definition, we need to compute the moneyness of in-the-money options. However, with the updated dataset, the data contained the number of the options that were held by the CEOs, regardless of the condition of the moneyness. Therefore we adjusted the intrinsic value of options that are out-of-money to zero, which means we assumed that the strike price of such options equal to the underlying stock price. Equations (3) to (5) are the simplified notations to calculate the average moneyness of the options. In equation (3), the number weighted of the intrinsic value of each option equals to the average realizable value per option; in equation (4), the number weighted strike price of each option matches the estimated strike price based on the aggregate data; and in equation (5), the stock price over the number weighted strike price equal to the weighted moneyness per option, which both fit the average percentage moneyness of option derived from the early measurements.

In previous literature (Campbell et al., 2011; Malmendier and Tate 2005, 2008), the overconfidence measures were treated as the discrete variables. Given the availability of the moneyness of the option, we are able to use the magnitude of the option moneyness percentage and further explore the curvature relationship between the level of CEO overconfidence and market reactions to acquisitions announcements.

3.3. Control Variables

Before applying the option-based overconfidence measure to test the effect of CEO overconfidence on M&A, it is crucial to carefully control the stock returns (Malmendier and Tate, 2015). Hence, we imposed several control variables to control for the deals, firms, and CEOs characteristics respectively.

3.3.1 Deals controls variables

To control for the M&A frequency and the M&A return, we first introduced the deals control variables.

(24)

- Relative Merger Size

The prior mergers and acquisitions literature does not reach a consensus concerning the effect of the relative merger size, the effect is however constantly significant. In Asquith et al. (1983), the relative merger size is positively related to the bidders’ abnormal return, while the coefficient is negative in Travlos (1987). Following the measure of Malmendier and Tate (2008) and Moeller et al. (2004), the relative merger size is generated as the fraction of transaction value over the acquirers’ market value 4 weeks prior to the announcement date. It helps to eliminate the possible effects on the acquirers’ cumulative abnormal returns in the regressions.

- Deal Attitude

As defined in the Thomson One, deal attitudes are classified into friendly, hostile, neutral and unsolicited. Friendly transactions are offers that the boards recommend. Offers that are officially rejected but the acquirers persist acquiring are defined as hostile. And the neutral attitude indicates that the attitude of the board is not applicable. As for the unsolicited, they are offers that have not yet given a recommendation. In this study, the deal attitude is a dummy variable that equals to 1 if the deal attitude is “friendly” and equals to 0 if otherwise.

- Relativeness

The relativeness is the measure for diversifying M&A. Based on the early studies, Sicherman and Pettway (1987) documented the significant positive cumulative stock returns of the bidder in related acquisitions, and the CAR around the event date are significantly more negative for diversifying deals. Morck, Schleifer, and Vishny (1990) also suggested that the diversification destroys the value of M&A. Thus, the control variable for diversifying M&A is constructed as a binary with 1 signifies the within-industry deals. If the targets and the acquirers have the different first 2-digit SIC codes, then we defined the deal as diversifying and assigned the variable of 0.

- Cash Financing

The payment method of the M&A deals serves as another important control. Andrade et al. (2001) and Heron & Lie (2002) found the negative stock returns of the acquirer firm at the

(25)

M&A announcement date when the deals are paid with equity, and the stock returns are zero or slightly positive when they are financed with cash. This is due to the fact that issuing equity signals the overvaluation of the shares and therefore leading to the negative market reaction to acquirers’ stock returns (Myers and Majluf, 1998). Based on the variable provided in Thomson one, the cash financing control variable was created. The cash payment variable equal to 1 if the consideration offered in the transaction is CASH, EARNOUT or LIABILITIES, and it equal to 0 if otherwise.

-Target public status

Additionally, due to the significant bargaining power of public firms over the private firms, acquiring the private target firms normally generate a positive effect on shareholder’s return. The empirical research made by Chang (1998) and Hansen and Lott (1996) both found the higher cumulative average abnormal returns of the acquirers buying the private targets over the public targets.

3.3.2 Firms controls variables

As for firm’s controls, all the variables are defined as the value at the end of the latest fiscal year prior to the M&A announcement.

- Firm size

Firm size, as a standard M&A control, is included as firms control in this research. As mentioned in the literature review, Moeller, Schilingemann, and Stulz (2004) found that small acquirers achieved higher gains from acquisitions than larger firms. The firm size is calculated by taking the natural logarithm of the total assets.

- Tobin’s Q

Tobin’s Q is the value of the market value of the assets over the book value of the assets. It is a measure of economic efficiency. However, there were contradictive views about the effect of Tobin’s Q on stock market reaction. Specifically, Lang, Stulz, Walking (1991) documented the positive effects and Moeller et al. (2004) revealed the negative effect. The formula to calculated the tobin’s Q is presented in appendix II.

(26)

Following the measure of Malmendier and Tate (2008), the cash flow is the income before extraordinary items plus the depreciation and amortization and divided by the total assets at the beginning of the year. The cash flow measures the availability of internal financing. Additionally, early research (Malmendier and Tate, 2005; 2008) found that overconfident CEOs are cash flow sensitive and as a result tend to make more M&As and thereby generate more negative stock returns when cash is abundant.

- Capital Expenditures

Capital expenditures (CAPEX) represent the capitals of the firm that are used to acquire or upgrade the property, plant, and equipment, without the amounts increased from mergers and acquisitions. It represents the investment activities of a firm. The variable is calculated as the capital expenditures at the beginning of the year over the total vale of tangible fixed property (Malmendier and Tate, 2008).

- Leverage

Firms’ leverage also has the effect on firm’s value. It can be computed as the book value of debts over the market value of total assets. Firms can benefit from the tax shield by having the debt, while it may also increase the default risk of the company.

3.3.3 CEOs controls variables

The last category of the control variables is the CEOs controls. The CEOs control variables controlled for the CEOs’ personal characteristics, the CEOs’ compensation, and the cooperate governance.

- Stock ownership

According to the studies made by Malmendier and Tate (2008) and Hirshleifer et al. (2010), the percentage of the share ownership can be derived as the fraction of shares and options owned by the CEO over the total number of shares outstanding. The percentage of stock ownership controlled for the CEOs’ incentive. Berger, Ofek and Yermack (1997) suggested that the stock ownership could measure the alignments of the CEO and the shareholders. The higher the share owned by the CEO, the better the CEO performs in the interest of the shareholder.

(27)

- Vested options

The vested options to control the CEOs compensation were also obtained. To capture the non-linear effect of vested option on the cumulative abnormal return, the quadratic vested options term is added as well. The firm value increase with the vested option of the CEO at the low level due to the effective incentive improves. While, with excessive compensation, the CEO may destroy the firm value, and provoke the negative reaction on cumulative abnormal return.

- Efficient board

Campbell et al. (2011) predicted that CEOs with a moderate level of overconfidence are more welcomed by the shareholders if the boards act in the investors’ interests. Along with Malmendier and Tate (2008), the efficient board variable is a binary variable that equal to 1 if the board size is efficient (number of the board members is between 4 and 12). This variable measures the corporate governance. A small board gives more power to the CEOs and a larger board results in the inefficiency of the corporate governance.

- Age and gender

Since overconfidence is related to the CEOs’ personal characteristics, it is also necessary to control for its effects. Barber and Odean (2001) researched the effect of gender on overconfidence and concluded that males are more likely to be overconfident than females. Consequently, the dummy variable of CEO gender was added to the set of controls. The gender variable equals to 1 if the gender of the CEO is male. In addition, Barber and Oden (2001) and Levi et al (2014) suggested that younger individuals are more overconfidence and more willing to undertake mergers and acquisitions. Furthermore, the early-life experience can also affect corporate financial decision (Malmendier, Tate, and Yan, 2011). As such, the ages of the CEOs previous to the M&A announcement was included to control for the CEOs’ characteristic.

(28)

3.4. Regression Models

To test the hypothesis 1, the OLS regression with the dependent variable of M&A frequent will be conducted. The dependent variable is the number of M&As conducted by CEO per company during the sample period of 2006 to 2016. The results aim to verify the tendency of overconfident CEOs conducting multiple acquisitions than their rational counterparts. The control variables include the firm’s size controls, the growth and investment opportunity controls, CEO incentive controls, and corporate governance controls. The year fixed effect is included to control the effect of M&A waves (Brown and Sarma, 2007).

𝑁! = 𝛼 + 𝛽₁𝑂𝐶𝑖+ 𝛽₂𝑆𝑖𝑧𝑒𝑖,𝑡+ 𝛽₃𝑄𝑖,𝑡+ 𝛽4𝐶𝑎𝑠ℎ 𝑓𝑙𝑜𝑤𝑖,𝑡+ 𝛽5𝑆𝑡𝑜𝑐𝑘𝑜𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝𝑖,𝑡 + 𝛽₆𝑉𝑒𝑠𝑡𝑒𝑑 𝑜𝑝𝑡𝑖𝑜𝑛𝑠_𝑖,𝑡+ 𝛽₇𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 𝑏𝑜𝑎𝑟𝑑 + 𝜀

The second hypothesis was tested by formulating regression with the dependent variable of bidders cumulative abnormal returns around the M&A announcements dates. The regression examines the general effect of overconfident CEOs on value creations from the M&As. The cumulative abnormal returns over an event window (-1, +1) will be tested first, and the returns defined over a five-day and eleven-day period will be conducted in the robustness check. Predictors included measures of CEO overconfidence and other standard control variables. A preliminary regression model is presented as follows:

𝐶𝐴𝑅_!,! = 𝛼 + 𝛽_!𝑂𝐶_!,!+ 𝛽_!𝑆𝑖𝑧𝑒_!,!+ 𝛽_!𝑄_!,!+ 𝛽_!𝐶𝑎𝑠ℎ 𝐹𝑙𝑜𝑤_!,! + 𝛽_!𝐶𝑎𝑝𝐸𝑥𝑝𝑒𝑛𝑠𝑒_!,! + 𝛽!𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒!,!+ 𝛽!𝑀𝑒𝑟𝑔𝑒𝑟 𝑆𝑖𝑧𝑒!,!+ 𝛽!𝐶𝑎𝑠ℎ 𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑛𝑔!,! + 𝛽_!𝑅𝑒𝑙𝑎𝑡𝑒𝑑𝑛𝑒𝑠𝑠_!,!+ 𝛽_!"𝐴𝑡𝑡𝑖𝑡𝑢𝑑𝑒_!,!+ 𝛽_!!𝑃𝑟𝑖𝑣𝑎𝑡𝑒 𝑇𝑎𝑟𝑔𝑒𝑡_!,! + 𝛽_!"𝑉𝑒𝑠𝑡𝑒𝑑 𝑜𝑝𝑡𝑖𝑜𝑛𝑠_!,!+ 𝛽_!"𝑉𝑒𝑠𝑡𝑒𝑑 𝑜𝑝𝑡𝑖𝑜𝑛𝑠_!,!! _{+ 𝛽} !"𝑆𝑡𝑜𝑐𝑘 𝑜𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝!,! + 𝛽_!"𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡𝑏𝑜𝑎𝑟𝑑_!,!+ 𝛽_!"𝑎𝑔𝑒_!,!+ 𝛽_!"𝑔𝑒𝑛𝑑𝑒𝑟_! + 𝜀

To capture the possible nonlinear effect of the degree of CEO overconfidence and acquirers’ stock returns, the squared moneyness of the option measure will be included. The quadratic regression is presented as follows. Since shareholders would benefit from the overconfident CEO up to a moderate level, the shareholder’s wealth increases with the level

(29)

of CEO overconfidence with a diminishing rate. Accordingly, we can predict a hump-shaped relation between the Overconfidence and the acquirer’s cumulative abnormal return.

𝐶𝐴𝑅_!,! = 𝛼 + 𝛽_!𝑂𝐶_!,!+ 𝛽_!𝑂𝐶_!,!!_{+ 𝛽} !𝑆𝑖𝑧𝑒!,!+ 𝛽!𝑄!,!+ 𝛽!𝐶𝑎𝑠ℎ 𝐹𝑙𝑜𝑤!,! + 𝛽_!𝐶𝑎𝑝𝐸𝑥𝑝𝑒𝑛𝑠𝑒_!,!+ 𝛽_!𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒_!,! + 𝛽_!𝑀𝑒𝑟𝑔𝑒𝑟 𝑆𝑖𝑧𝑒_!,! + 𝛽_!𝐶𝑎𝑠ℎ 𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑛𝑔_!,!+ 𝛽_!"𝑅𝑒𝑙𝑎𝑡𝑒𝑑𝑛𝑒𝑠𝑠_!,!+ 𝛽_!!𝐴𝑡𝑡𝑖𝑡𝑢𝑑𝑒_!,! + 𝛽!"𝑃𝑟𝑖𝑣𝑎𝑡𝑒 𝑇𝑎𝑟𝑔𝑒𝑡!,!+ 𝛽!"𝑉𝑒𝑠𝑡𝑒𝑑 𝑜𝑝𝑡𝑖𝑜𝑛𝑠!,!+ 𝛽!"𝑉𝑒𝑠𝑡𝑒𝑑 𝑜𝑝𝑡𝑖𝑜𝑛𝑠!,!! + 𝛽_!"𝑆𝑡𝑜𝑐𝑘 𝑜𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝_!,!+ 𝛽_!"𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡𝑏𝑜𝑎𝑟𝑑_!,!+ 𝛽_!"𝑎𝑔𝑒_!,!+ β_!"gender_! + ε

Mergers and acquisitions happen in waves over the different time period and influence differently across the industries (Bruner, 2004). To eliminate the unobserved heterogeneity, in the last two regressions both year-fixed and industry-fixed effects are introduced into the analysis. The time fixed effect controls for omitted variables that affect the dependent variable differently over time but are invariant among the entities, and the industry fixed effect absorbs the impact of omitted industries characteristics that are constant among years (Stock and Watson, 2015). The standard errors were clustered to account for heteroskedasticity.

(30)

IV. Data and Descriptive Statistics

4.1. Data Collection

To construct the empirical research, the data collection and the data sources that used in this study will be discussed in this section. The data sources include the Thomson one, Compustat Executive Compensation, Datastream, Compustat Fundamentals Annual and the Institutional Shareholder Services (ISS) database. M&A related information was obtained from SDC Thomson ONE, and the CEO compensation data are extracted from the ExecuComp. To construct the event study, the historical daily return around the announcement data for the sample firms as well as the market index are provided by DataStream. Other firms related information was constructed using Compustat (accounting data) and ISS (board related data).

4.1.1 M&A data

This study focused on the listed U.S. firms that made M&A announcement from 01/01/2006 to 12/31/2016. In the first step, the initial data selected from the Thomson One - Mergers and Acquisitions database based on the following selection criteria.

Search criteria for the M&A announcements: a. Acquirer nation is United States of America b. Acquirer public status is public

c. Announcement date between 01/01/2006 to 12/31/2016

d. The acquirers controlled less than 50% of the targets’ shares prior to the transaction and owned above 50% of the shares after the transaction.

e. The deal is completed.

f. The deal value is more than 10$ million as disclosed in Thomson One.

In order to acquire sufficient information, the sample of the study is composed of public firms which made acquisitions in the sample time period. Because in the Execucomp database, the detailed option information is only available after 2006, the time span was defined as

(31)

2006 onward. Moreover, the financial firms (SIC code 6000 to 6999) and the utilities firms (SIC codes 4900 to 4949) are excluded from our sample. In total, 7986 M&A announcement made by 3000 public firms were identified based on the aforementioned criteria. Additionally, some M&A related variables as well as firm identifiers, which will be used to merger the data among different databases, are retrieved from Thomson One. The variables include announcement dates, acquirers’ identifiers (CUSIP and DataStream code), acquirers and targets SIC codes, deal attitude, transaction value, acquirers’ market value before announcement and deal payments.

4.1.2 CEO compensation data

With 3000 firms obtained from the Thomson One in the first step, a list of sample firms’ CUSIP codes can be gathered. Before collecting the corresponding firms’ CEOs compensation information, I converted the 6-digit CUSIP that was used in the Thomson One to 8-digit CUSIP code in order to be accepted in the WRDS ExecuComp database. To construct the overconfidence measure mentioned in the methodology, both Annual Compensation table and the Outstanding Equity Awards table were used. The Annual Compensation table contained the aggregated compensation data from 1992 to 2017. The Outstanding Equity Award Dataset contains the detail outstanding stock option held by the officer and applied current FASB 158 accounting standard. The database only contained the data from 2006 onward. As the Executive Compensation databases only include the S&P 1500 firms and firms that were once included in the S&P 1500 firms, our sample firms deducted to 1018 in the Annual Compensation table, and it further deducted to 1002 firms after merged with data from Outstanding Equity Award table. The intermediate CEO compensation sample covered 1002 firms and 1774 CEOs during the time period from 2006 to 2016. Moreover, variables that used to formulate the overconfidence measures, other CEO related control variables are also acquired from the ExecuComp. For example, the number of common shares outstanding, shares owned by the executive, age, and gender of the executives are obtained accordingly.

(32)

4.1.3 Accounting data and board size

Subsequently, to construct the firm control variables, the financial accounting data at the last fiscal year end prior to the announcement year (2005 to 2015) were extracted from the Compustat Monthly updates - fundamentals Annual database. The gvkey identifiers, which were obtained from the Compustat ExecuComp, gave the convenience to collect data and match the variables among the datasets. The accounting data were used to construct the firm control variables including the firm size, Tobin’s Q, Cash Flow and the Leverage of the acquirer firms.

The board size information that was previously used to form the board efficiency measure was obtained from Institutional Shareholder Services (ISS) in WRDS. However, the ISS data set is divided into two parts, with the Directors starting from 2007 and the Directs Legacy ending in 2006. Since our board data window covers a time period from 2005 to 2015, it was necessary obtain the variables twice and with different digits of the identifier. The 9-digit CUSIP Identifiers were used in Directors and 6-digit CUSIP were used in Directors Legacy. The CUSIP converter in WRDS was used to convert the CUSIPs -8 to CUSIPs-9.

4.1.4 Event study data

After the accounting data and the board information being collected, I matched them with the M&A data. Up until this point, it generated a matched 3700 M&A deals made by 1001 firms. The last step was to collect the acquirer firms’ stock returns and market index from the DataStream and calculate the cumulative abnormal return over the three-day and five-day event window. Following the previously discussed event study methodology, the average abnormal return were obtained for an event window of t = [-10, 10] with the estimation window [-160，-31]. As for the processing of the stock price, due to public holidays, there are some data missing on various days in the time series for both adjusted closing prices of S&P 500 index and the firm’s daily stock prices, in such case, the data will be filled with the next trading day’s price. Due to the lack of sufficient stock information, ten firms were excluded

(33)

from the sample and the sample contained 3641 announcements of M&A made by 991 firms from 2006 to 2016.

4.1.5 Missing data and outliers

In the end, we merged the collected data from the various datasets into a final combined dataset. Observations with missing data were excluded from the final sample so as to have a clean dataset to test for the regressions. After normality test of the variables, variables with severe outliers were winsorized at 2% level. Explicitly, variables that were winsorized include cash flow, Tobin’s Q, merger size, CAPEX, leverage, vested option and stock ownership. Moreover, to set the panel data, the duplicates of CEO and announcement date items were deleted as well. The final matched data have 2565 observations and contained 974 U.S. public firms with 1013 observed CEOs.

4.2. Summary Statistics

Before conducting the regression analysis, the detail descriptive statistics of the research sample will be discussed. The summary statistics will present an overview of the data and give some initial insights for the research.

Figure 1 presents the value and number of M&A announcements made by the sample firms in each year from 2006 to 2016. The bar chart indicates the number of deals by years and the line graph plots the total value of the transactions over years. We saw a larger proportion of the M&A announcements took place in the first 2 years during the sample time span. Consistent with the sixth M&A wave (Alexandridis, Mavrovitis, and Travlos, 2012), the deal number increased in 2006 and peaked in 2007. In 2008 and 2009, the merge frequency plunged due to the financial crisis. Besides, the number M&A announcements surged again in 2014, which imply for the seventh wave (IMAA, 2017). However, the current trend in 2016 indicates the downward tendency in the number of M&A in the U.S. In terms of total transaction value, we only included deals with the transaction value over 10 million dollars in the sample. The total transaction values in 2015 reached 700 billion dollars, which achieved