Implications of CEO overconfidence in mergers and acquisitions

(1)

Implications of CEO overconfidence in mergers and acquisitions

Friso Hoving 10381813

University of Amsterdam Faculty of Economy and Business Msc Finance, specialization Asset Management

Final draft Master Thesis June 2017

Thesis supervisor: Dr. Caskurlu Second thesis supervisor:

(2)

Statement of originality

I, Friso Hoving, hereby declare that this thesis is my own work and to the best of my knowledge it contains no material previously published or written by another person, nor material which to a substantial extent has been accepted for the award of any other degree or diploma at any educational institution, except where due acknowledgement is made in the thesis.

(3)

Abstract

This research examines the effect of CEO overconfidence in the context of mergers and acquisitions. The final sample contains 1821 different mergers and acquisitions between U.S. publicly listed firms, performed by 1128 different CEO’s between 1993 and 2016. The measure of overconfidence is determined by observing the holding behavior of executive stock options in the personal portfolio of CEO’s. The empirical results show an increase in M&A activity if the CEO of the acquiring firm is overconfident. No significant relationship is found between acquiring CEO overconfidence and the preference for cash as method of payment in mergers and acquisitions or the probability that a merger or acquisition is diversifying. Furthermore, the results show a significant relationship between CEO

overconfidence and the reaction of the market following the announcement of a mergers or acquisition. A distinction is made between deals where only the acquiring CEO is

overconfident and deals where both CEO’s are overconfident. The former leads to a negative reaction of the market, whereas the latter leads to a positive reaction of the market.

Keywords: CEO overconfidence, mergers and acquisitions, M&A activity, method of

(4)

1. Introduction

The position of Chief Executive Officer (CEO) comes with many obligations and

responsibilities. They are always in charge of the firm’s operations and their role in high-level negotiations is indispensable. Besides, administering such a high-profile position attracts attention from society. This might instigate some of the CEO’s dispositions and biases. One of these dispositions is overconfidence. ‘Guts before glory’ might be the motto for many overconfident CEO’s as this attitude could encourage ambition and innovation. However, overconfidence could also distort the CEO’s view on reality, which could have negative implications for the firm they are managing. Therefore, it is important to ascertain the origin of overconfidence. Are CEO’s born overconfident or do they develop

overconfidence throughout experience? One possible explanation for overconfidence that is provided by previous behavioral economic literature states that the origin of overconfidence can be ascribed to the self-attribution bias (Billett & Qian, 2008). Self-attribution learns individuals to be overconfident instead of being critical towards their own actions. The self-attribution bias states that CEO’s tend to overestimate their own contribution regarding good outcomes, whereas they tend to overestimate external factors or bad luck when it comes to bad outcomes (Billet & Qian, 2008).

The implementation of psychological findings, such as overconfidence, in financial models is not a new phenomenon. Initially many economics were skeptical, until certain puzzles in the financial markets, which could not be worked out by the classical economic theory, were explained by including the concept of overconfidence amongst investors. This research will focus on the role of CEO overconfidence in the process of mergers and acquisitions. The primary economic motive behind mergers and acquisitions is to create growth and gain synergies, which means that the combined enterprise value after the merger or acquisition is greater than each separate company would have been on its own (Hoberg & Phillips, 2010). Other motives to undertake a merger or acquisition are both economies of scale and economies of scope, as well as the desire to increase market power (Sudarsanam, Holl & Salami, 1996). If it is assumed that CEO’s are rational, they will always act in the interest of shareholders. Consequently, CEO’s will only undertake a merger or acquisition if this maximizes shareholder value as the announcement of a merger or acquisition usually affects the stock price of both acquiring and target firm. However, previous literature shows that mergers and acquisitions not always result in positive shareholder gains and might even

(6)

be value-destroying for the acquiring firm (Loughran & Vijh, 1997). Overconfidence plays an important role in this process as significant evidence is stated for the hypothesis that CEO’s constantly overestimate their own skills relative to others, which results in over optimism about the outcomes of their decisions. This hypothesis is formalized by the assumption that overconfident CEO’s overestimate the expected returns of their corporate decisions (Malmendier & Tate, 2005). One of the first researchers who examined the effect of overconfidence in mergers and acquisitions was Roll (1986). In his research the hubris hypothesis is introduced, stating that overconfident CEO’s of acquiring firms on average overpay for their targets. Malmendier and Tate (2008) find similar results. They state that as a results of overestimating their capacity to generate profitable returns, overconfident CEO’s overpay for target firms and even undertake mergers and acquisitions which destroy value for their own shareholders. This effect is amplified if more internal funds are available.

Furthermore, CEO’s wish for increased compensation produces a self-interested motivation to acquire since the post-acquisition compensation of the CEO generally

increases. This motivation is largest when CEO’s are relatively underpaid compared to their peers, by which overall M&A activity will increase (Datta, Iskandar-Datta & Raman, 2001). Managerial overconfidence also affects acquisitiveness, whereas a positive relationship is found between CEO overconfidence and the odds of making an acquisition, which is largest when the acquisition is diversifying (Malmendier & Tate, 2008). This research will attempt to verify the effects of CEO overconfidence on different features of the M&A process, such as overall acquisitiveness, the acquiring firm’s performance, the method of payment and whether a deal is classified as diversifying. Moreover, this research will add to the existing literature by examining the effect which CEO overconfidence of the target firm might have on the market reaction after an merger or acquisition announcement. Prior research on target CEO overconfidence is limited as most research is focused on overconfidence of the

acquiring CEO. Essentially, this research aims to answer the following research question:

Is there a relationship between CEO overconfidence on both the market reaction after the announcement of a merger or acquisitions and on corporate decisions made during the deal process?

In this research, announcements of mergers and acquisitions of U.S. public firms between 1993 and 2016 are examined. After restricting this M&A data to some extent and merging it with data on firm – and CEO characteristics, a final sample of 1821 different mergers and

(7)

acquisitions remains. Numerous approaches to quantify CEO overconfidence are posited in previous literature. Hayward and Hambrick (1997) names four drivers of CEO

overconfidence, which are recent organizational success, media appraisal, CEO’s self-importance and weak board vigilance. However, the contribution of the CEO in recent organizational success is generally overestimated as such success could often more

objectively be contributed to other factors. Favorable media coverage on the CEO will foster the overestimation of their own competence and is closely linked with third driver of self-importance. Malmendier and Tate (2005) introduce a measure of overconfidence based on the option holding behavior of the CEO. They focus on detailed information about the CEO’s transactions in the personal portfolio containing their own firm’s stock and options. CEO’s who choose to hold these assets longer than would be rational to a certain threshold are considered overconfident. This measure uncovers a demonstrable confidence in the firm’s prospects as the CEO rationally chooses to expose himself to the firm’s idiosyncratic risk. More specifically, a CEO is categorized as overconfident when he voluntarily holds on to stock options beyond the vesting period in which exercise of the stock options is permitted (Hirshleifer, Low & Teoh, 2012). In this research the measure for CEO overconfidence will be based on the approach of Malmendier and Tate (2005). However, the measures will be adapted to the final dataset following the approach of Campbell, Gallmeyer, Johnson, Rutherford and Stanley (2011).

The remainder of this paper is organized as follows: section 2 presents the literature review in which the theoretical background and subsequent hypotheses will be stated. Section 3 presents the research design and the construction of variables is described. Section 4

discusses the data used for this research. Section 5 discusses the results of the empirical research. This section also includes robustness checks to verify the robustness of the results. At last, section 6 concludes and the implications of this research are discussed.

(8)

2. Literature review

In this section an overview of the related literature with regard to the concept of overconfidence is provided. At first, the role of overconfidence in finance is outlined. Thereafter, prior research on the effects of CEO overconfidence in mergers and acquisitions is discussed and distinct hypotheses are stated.

2.1 Overconfidence in finance

The effects of overconfidence on financial markets is accounted for in many experimental studies and formal economic models. One of those models is the prospect theory by Kahneman and Tversky (1979) and the subsequent disposition effect, which states that investors tend to hold losing investments too long and sell winning investments too soon (Odean, 1998). Overconfident traders unconsciously overestimate the precision of their signal and they overvalue the information they are receiving compared to the information of rivals. In other words, overconfident traders overestimate the efficacy of their own information and are willing to base their decisions on this information (Odean, 1998). Moreover,

overconfidence can be seen as a natural explanation for the patterns of excessive trading in financial markets. Overconfident investors might even trade when the cost of trading is offset by the expected gains of trading (Odean, 1999).

Although overconfidence originates from an individual basis, it is widely detected on an aggregate level in the long-run. Some researchers state that traders learn over time and they allow the self-attribution bias, where people tend to attribute success to their own actions and failure is attributed to external factors, to change over time (Gervais & Odean, 2001). More experienced traders might realize over time that previous success is to be contributed to other provable factors and subsequently their overconfidence will decline. On the contrary, rational traders might learn to become overconfident over time. Due to the underestimation of risk and the overestimation of their own capabilities, overconfident traders consequently outperform rational traders in exploiting misvaluations caused by noise trading, whereby overconfident traders will continue to survive in the long-run (Hirshleifer & Luo, 2001). Research on the effects of overconfidence in corporate finance issues has developed more recently and is mainly focused on overconfidence of the CEO regarding corporate decision making.

(9)

2.2 CEO overconfidence in mergers and acquisitions

2.2.1 CEO overconfidence and M&A activity

Roll (1986) already established that CEO overconfidence significantly affects the premium that is paid by the acquiring firm in a merger or acquisition. Overconfident CEO’s seem to hold biased beliefs regarding their own capabilities and as a result they overestimate the chances of a profitable outcome in the future. Therefore, overconfident CEO’s are most likely willing to take more risks and as a result they would be more eager to undertake a merger or acquisition compared to rational CEO’s. Malmendier and Tate (2008) directly link this overconfidence to M&A activity. They examine the effect of CEO overconfidence on acquisitiveness. In order to measure acquisitiveness they construct a binary variable, which equals one if a firm announced at least one successful merger or acquisition in a particular firm year, or zero otherwise. They conclude that the odds of performing an acquisition are 65% higher if the CEO is classified as overconfident.

Most previous literature provides insights in the effect of managerial overconfidence in the context of U.S. mergers. However, in more recent research conducted by Ferris, Jayaraman and Sabherwal (2013), it is examined whether the results of Malmendier and Tate (2008) also holds when investigating international mergers and acquisitions. Besides, they focus on how these mergers by overconfident managers are performed. Since managerial overconfidence might be influenced by different national cultures, an analysis is conducted on the effects of overconfidence on the amount of merger bids performed by a CEO. Ferris et al. (2013) find that overconfident CEO’s tend to engage in more offers than rational CEO’s. In general, acquisitions may appear overly attractive for CEO’s and overconfident CEO’s tend to be more acquisitive unconditionally (Malmendier & Tate, 2008). Based on the previous literature, the first hypothesis is stated as follows:

Hypothesis 1: Acquiring CEO overconfidence increases the probability of undertaking a merger or acquisition.

(10)

2.2.2 Method of payment in mergers and acquisitions

The preferences and beliefs of a CEO play a big role in all types of corporate decision making. Making the right corporate investment decisions is important for firms since these decisions can often alter the firm’s future value. One of the first influential theories on corporate investment was introduced by Modigliani & Miller (1958). Their theory states that the value of the firm is not affected by the capital structure. More specifically, the main findings of Modigliani and Miller tell us that the method of financing an investment is irrelevant to determine whether the investment is worthwhile.

Before undertaking any merger or acquisition it also matters which type of funding will be used in the transaction. One of the main decisions the CEO needs to make is how to finance the deal. In general, the acquiring firm chooses to pay the target firm with either cash, stock or a combination of both cash and stock. This decision matters for both the acquirer’s stockholders as for the target’s stockholders. For instance, for the acquiring firm the choice of acquiring method could alter the financial structure of the firm. Besides, it is important to know whether the choice of payment in the acquisition might add or destroy firm value. Prior research states that acquisitions paid with cash yield higher positive abnormal returns around the announcement of an acquisition compared to acquisitions paid by stock (Travlos, 1987). Both announcement returns as long-term returns are lower for all stock acquisitions.

However, Heron and Lie (2002) find that changes in operating performance do not differ following a stock acquisition or a cash acquisition. One potential explanation they posit is that investors might be overly optimistic about the long-term future growth opportunities when a stock acquisition is announced. This change in the investor’s expectations will affect stock prices but subsequent operating performance might not substantially change.

Another widely acknowledged framework with regard to the type of funding to use when a favorable investment opportunity arises, is the pecking order theory, first introduced by Myers and Majluf (1984). This theory includes the effect of asymmetric information between mangers and shareholders in the choice for the optimal capital structure. The manager has three options to raise capital in order to advance in an investment opportunity. He can raise external funds, either by issuing equity or debt, or rely on internal funds, such as cash. It is assumed that mangers know more about the fundamental value of the firm

compared to the shareholders. The issuing of equity is least preferred to managers since outside investors will think the manager only issues equity because he thinks the firm is overvalued and he wants to take advantage of it. Hence, a negative signal will spread to the

(11)

market about the fundamental value of the firm. The issuing of debt might increase the risk of financial distress or even bankruptcy. However, it also spreads a positive signal to the market that the firm is solvent enough to pay interests and it signals that the investment opportunity is profitable. Nevertheless, the pecking order theory predicts that managers will always prefer the use of internal funds since the implications of the aforementioned asymmetric

information will be minimized.

Malmendier, Tate and Yan (2007) illustrate that capital structure decisions could be improved once managerial characteristics, such as CEO overconfidence, are accounted for. Their findings indicate that overconfident CEO’s are less likely to issue equity as a method of financing an investment opportunity compared to rational CEO’s. Instead, overconfident CEO’s prefer the issuing of debt over equity which results in an increase in leverage on the long-term. Although, overconfident CEO’s are more conservative in issuing risky debt, whereas they rely even more on using internal funds as method of financing investment opportunities. Malmendier and Tate (2005) showed that overconfidence entices a CEO’s preference for internal over external financing when raising funds to make investments. They find a significant positive relationship between the sensitivity of investment to cash flow and CEO overconfidence, mattering most for firms that are equity dependent. In more recent research, a positive relationship between CEO overconfidence and cash being used as method of payment in mergers and acquisitions is posited by Malmendier and Tate (2008), although this result is only statistically significant with the extra assumption that the firm should be unlikely to be overvalued. Nevertheless, overconfident CEO’s most likely tend to overvalue their own firm compared to what the market concludes. Following the pecking order theory, it is expected that an overconfident CEO will prefer using cash as method of payment in an merger or acquisition. Therefore, the second hypothesis is stated as:

Hypothesis 2: Acquiring CEO overconfidence is positively related with the probability of a deal being financed by cash.

2.2.3 Diversifying mergers and acquisition

Diversification is a corporate strategy of entering a new industry in which a firm is currently not active. By diversifying the management of a firm might expect an increase in economic value through cohesion with their current business activities. Diversification may include the development of new products, but also undertaking a merger or acquisition with firms that

(12)

are active in different industries. Two sets of arguments are widely brought forward to illustrate why firms choose to diversify (Servaes, 1996). The first argument claims that firms diversify to increase shareholder wealth, whereas the second argument identifies

diversification as an result of the agency problems between managers and shareholders. Furthermore, Servaes (1996) examines the value of diversification when many firms started to diversify during the conglomerate merger wave of the late 1960’s and early 1970’s. His main finding is that diversified firms are valued at a discount compared to single segment firms and this discount is declining over time. Firms with high insider ownership were focused instead of diversified when this diversification discount was large, but as the discount decreased, these firms started to diversify due to manager’s private benefits from diversification.

The research of Morck, Shleifer and Vishny (1990) focusses on the relationship between diversification as an acquisition strategy with the quality of the acquisition, measured as the announcement period return of the acquiring firm. They find that these returns are systematically lower or even negative when firms undertake a diversifying acquisition. Berger and Ofek (1995) study the effects of diversification by comparing the value of a diversified firm with the value of the firm’s different segments separately,

operating as independent firms. The gain or loss in value from diversification is measured as excess value and their empirical evidence reveals significant negative excess value,

indicating that diversification destroys firm value. However, the assumption that these conglomerate of segments can be compared to stand-alone firms should be carefully reassessed (Graham, Lemmon & Wolf, 2002). In fact, Graham et al. (2002) confirm the findings of Berger and Ofek by describing a same consistent decline in excess value.

Although, they state this decline in excess value can be assigned directly to the acquirement of yet discounted target firms, even when the diversification itself is not value-destroying. The research of Schoar (2002) adds to the literature by showing that diversification is not a profitable strategy. The net effect on productivity in acquiring another firm is negative. This means that diversifying acquisitions are not always optimal for both the acquiring and target firm. Besides, Schoar states that a higher productive efficiency, as a result of diversification, does not necessarily create more value for shareholders.

Malmendier and Tate (2008) already established that overconfident CEO’s tend to overestimate their skills to generate returns and therefore undertake more mergers and acquisitions, even if these are value-destroying. Furthermore, they find that the

(13)

diversifying. Besides, most prior researches provides empirical evidence that the stock of a diversifying firm is discounted and diversifying mergers might destroy value for the

shareholders of the acquiring firm. Therefore, rational CEO’s are most likely less tempted to undertake a diversifying merger or acquisition. This is confirmed by Malmendier and Tate, who find a significant positive relationship between their proxy for overconfidence and the probability that a deal is diversifying. Thus, the third hypothesis is stated as follows:

Hypothesis 3: Acquiring CEO overconfidence is positively related with the probability of a deal being classified as diversifying.

2.2.4 Market reaction to CEO overconfidence

Moeller, Schlingemann and Stulz (2004) find that acquiring firms generally experience shareholder wealth losses regardless of which method of finance is used. Moreover, public large firms suffer greater wealth losses compared to public small firms at the announcement of an acquisition. These larger firms tend to offer larger acquisition premiums and even undertake acquisitions with negative synergy gains, which is in line with the managerial hubris hypothesis of corporate takeovers stated by Roll (1986). Brown and Sarma (2007) further explored the wealth effects of acquisitions. An acquisition is regarded value

enhancing for both the target firm as the acquiring firm when synergies are created. However, they state that there is a consensus in prior research indicating that acquisitions only seem to benefit the shareholders of target firms, whereas acquisitions are value neutral at best for shareholders of the acquiring firm.

Following the standard bidding theory it is assumed that all bidders are conscious of the winner’s curse. Therefore, they adjust their bid downwards relative to the initial

estimation of the target’s value. Although, when CEO’s of bidding firms are affected by overconfidence, this standard bidding theory does not hold, indicating that they do not entirely incorporate the winner’s curse (Roll, 1986). The research of Doukas and Petmezas (2007) uses a dataset of mostly private acquisitions in the United Kingdom. This allows them to analyze whether the effects of overconfidence are robust outside the USA. Doukas and Petmezas also conclude that overconfident CEO’s realize lower announcement returns compared to rational CEO’s and the long-term performance amongst overconfident CEO’s is poor. In order to determine whether a merger or acquisition is value creating or value

(14)

reaction delivers the most statistically convincing evidence (Andrade, Mitchell & Stafford, 2001).

Andrade et al. (2001) applied two commonly used event windows, which are the three-day event window immediately surrounding the day of the announcement, and a longer event window beginning various days before the announcement and ending when the deal is closed. Their results indicate that the shareholders of target firms are the winners in these transactions. The average three-day cumulative abnormal return for the target firms is significantly large and increasing over the extended event window. However, the average three-day cumulative abnormal return for acquiring firms is negative and becomes even more negative over the extended event window. Although, these latter results are not statistically significant at any conventional significance level, therefore it is difficult to state that shareholders of the acquiring firm are necessarily on the losing side of the transaction. An extension of the event window accommodates subsequent adjustments by the market, however unrelated factors other than the acquisition announcement will also influence the result (Hayward & Hambrick, 1997).

Malmendier and Tate (2008) find that overconfident CEO’s tend to engage in more deals than other CEO’s, lowering the average deal quality. Furthermore, as stated in prior research, overconfident CEO’s might overpay for the target, even in value-destroying deals due to the CEO’s excessive optimistic believes and overestimation of their own skills. The market reaction to M&A bids should consequently be lower when the acquiring firm is managed by an overconfident CEO. The results of Malmendier and Tate confirm this

hypothesis by showing a market reaction after a takeover announcement of -90 basis points if the CEO is classified as overconfident against -12 basis points if the CEO is not considered overconfident. Based on the prior research, a negative market reaction is expected after the acquisition announcement by an overconfident CEO, leading to the fourth hypothesis:

Hypothesis 4: Acquiring CEO overconfidence negatively affects the cumulative abnormal return of the acquirer’s stock surrounding an acquisition announcement.

2.2.5 Overconfident CEO’s of target firms

Nowadays the role of overconfidence in various financial models is widely acknowledged by many researchers. Overconfidence of a CEO mainly affects the decision-making process regarding investment opportunities. Since the introduction of Roll’s hubris hypothesis,

(15)

research focusing on the role of CEO overconfidence in M&A activity is extended. Although, many of this prior research is focused on the CEO’s overconfidence of the acquiring firm. After all, these CEO’s are the ones responsible whether to acquire the target firm. However, as in any negotiation process, two entities are involved. Above all, Malmendier and Tate (2008) conducted research on the effect of CEO overconfidence in the process of mergers and acquisitions. In their conducted research, they also focus on overconfidence of the CEO of the acquiring firm and its subsequent implications. Although, in one of their earlier

working papers they state that allowing CEO’s of target firms to be overconfident might yield relevant comparative statistics (Malmendier & Tate, 2004). Similar to the overconfidence of the acquiring CEO, the overconfident CEO of the target firm will most likely believe they are able to create at least as much value their selves as opposed to the acquiring firm after

acquisition. As a result, acquisitions of target firms managed by overconfident CEO’s are more prone to be classified as hostile and these CEO’s are more likely to reject a bid, even when this bid is beneficial to their shareholders. The inclusion of CEO overconfidence on the target’s side of the deal might also affect the market reaction to the announcement of such a deal. Empirical research on this matter is scarce, albeit Malmendier and Tate (2004) mention this as a further extension to their research. This research will contribute to the existing literature by including the effects of target CEO overconfidence and the interaction between overconfidence of both acquiring and target CEO.

One of the few researches that examines the role of CEO overconfidence of the target firm is conducted by John, Liu and Taffler (2011). More specifically, they empirically test the effect of CEO overconfidence, for both acquiring CEO’s and target CEO’s, as well as the interaction between both biases on both the premium paid and the firm performance of the acquiring firm. They find that the premium paid by the acquiring firm is between 7% and 9% higher when the CEO’s of both the acquiring firm and the target firm are overconfident, compared to merely one of the two or neither of the two being overconfident. Besides, the market’s reaction to a deal announcement between two firms both managed by overconfident CEO’s is more severe than deals in which either one or no overconfident manager is

involved. John et al. (2011) find an announcement abnormal return of the acquiring firm which is between 10% and 12% more negative in the occurrence of both CEO’s being overconfident. These results confirm the thought that CEO overconfidence of the target firm, particularly in interaction with CEO overconfidence of the acquiring firm, is influential for both characteristics of a deal as for the quality of an acquisition. Thus, based on the limited previous literature available, the effect on the market reaction following an announcement of

(16)

a deal when both CEO’s are overconfident will be examined, as an extension of the fourth hypothesis. Hence, the fifth and last hypothesis is stated as follows:

Hypothesis 5: The simultaneous effect of acquirer and target CEO overconfidence negatively affects the cumulative abnormal return of the acquirer’s stock surrounding an acquisition announcement even more than compared to only one or none of the CEO’s being

(17)

3. Methodology

The previous section provided an overview of the theoretical and empirical background in the finance literature regarding CEO overconfidence. Subsequently, hypotheses are developed on the potential implications of CEO overconfidence related to the M&A process. In this section the methodology used to test the hypotheses will be described. First, the construction of the measure of overconfidence will be outlined. Then, the computation of all dependent variables and control variables will be stated. At last, the different regression models will be presented.

3.1 CEO Overconfidence measure

In this research the measure of CEO overconfidence will be constructed based on the option holding behavior of the CEO. This measure is based on the research conducted by

Malmendier and Tate (2005) and altered following the approach of Campbell et al. (2011). In general, CEO’s are rewarded with large quantities of stock and option grants of their own firm. These options cannot be traded and short-selling of this stock is often restricted. Therefore, for risk-averse, undiversified CEO’s it would be optimal to exercise these stock options early once the option is abundantly in-the-money (Hall & Murphy, 2002). Based on this assumption the option-based overconfidence measures can be constructed.

Due to the inability to hedge themselves by trading the granted options or short-selling stock, CEO’s can be considered undiversified. Therefore, they are exposed to the idiosyncratic risk within their firm’s stock. In order to decrease this exposure CEO’s could minimize their stock option holdings by exercising the options immediately after the vesting period is over. Only an overconfident CEO is willing to hold on to his options beyond this vesting period as he is convinced the stock value will rise even more under his management. This means CEO’s must trade off the costs of this underdiversification against the potential value of holding on to the options (Malmendier & Tate, 2005). This tradeoff balances around a threshold price, which best can be defined as the stock price at which the CEO is indifferent between exercising early or holding the option for another period (Hall & Murphy, 2002). Hall and Murphy (2002) state that the optimal point for exercising an option depends on the executive’s individual wealth, degree of risk aversion and level of diversification. These drivers are included in their model and based on calibrations of this model, Malmendier and Tate (2005) state that CEO’s are overconfident if they hold stock options from which the stock price exceeds the exercise price by more than 67 % (i.e., 67% in-the-money).

(18)

The research of Malemendier and Tate is based on a proprietary dataset with detailed data on stock option holdings and exercising. Instead, in this research ExecuComp data is used, which is available for a large number of executives over a long period of time. Hence, in this research CEO overconfidence will be measured following the approached used by Campbell et al. (2011) and Hirshleifer et al. (2012). The ExecuComp data does not contain specific exercise prices for each option grant, so the average exercise price of the aggregated options is computed based on the approximation method stated by Core and Guay (2002).

At first, the realizable value for each option is calculated by dividing the total realizable value of all unexercised exercisable options on the number of these unexercised exercisable options. Only data of exercisable options is used since the choice of holding on to options which are in fact exercisable is of interest for this research to quantify

overconfidence. The average strike price will be calculated by subtracting the realizable value per option from the stock price at the end of the fiscal year. The average percentage

moneyness of the options will be denoted by dividing the realizable value for each option by the estimated average strike price.

CEO’s who hold on to their stock options after being valued at 67% in-the-money or higher are considered overconfident. However, the sample is restricted to CEO’s who show this option holding behavior at least twice during the sample period. This restriction ensures that all CEO’s had the opportunity to be classified as overconfident and it reduces the chance that CEO’s are assigned this classification disproportionally for times the stock of their firm is performing well (Malmendier & Tate, 2005). Although, the overconfidence classification is yet allocated the first time the CEO exhibits the option holding behavior after exceeding the threshold of 67% in-the-money. At last, a binary variable is constructed equaling one when this threshold is exceeded at least twice during the sample period, or equaling zero otherwise. The measure for target CEO overconfidence is constructed following the same approach, albeit based on the option holding behavior of CEO’s managing the target firms in the final sample.

High overconfidence measure

In the subsample of overconfident CEO’s a distinction is made between overconfident CEO’s and CEO’s with relative high overconfidence. The approach of Campbell et al. (2011) is followed to determine whether a CEO can be classified as highly overconfident. They require CEO’s with relative high overconfidence to hold stock options that are more than 100% in-the-money. Besides, the same condition is applied that the option holding behavior after

(19)

exceeding the 100% moneyness threshold has to occur at least twice. A binary variable is constructed equaling one when this threshold is exceeded at least twice during the sample period, or zero otherwise.

Low confidence measure

As opposed to the measure for high overconfidence, Campbell et al. (2011) define CEO’s with low confidence as those who exercise stock options that are less than 30% in-the-money and do not hold on to remaining exercisable stock options that are greater than 30% in-the-money. The latter is determined by calculating the average percentage moneyness of

exercised options. This follows from the aforementioned approach of calculating the average percentage moneyness, albeit applied to exercised options instead of unexercised exercisable options. A binary variable is constructed equaling one when both thresholds are not exceeded at least twice during the sample period, or zero otherwise.

3.2 Dependent variables

For the first hypothesis, a binary variable is created equaling one if a firm announced at least one successful merger or acquisition in a particular firm year, or zero otherwise. The second hypothesis tests whether CEO overconfidence has any effect on the method of payment used to finance a deal. In the final sample all deals are financed by either cash, stock, or a

combination of cash and stock. A binary variable is created equaling one when a deal is completely financed by cash, or zero otherwise. Similarly, a binary variable is constructed to classify an acquisition as diversifying in order to test the third hypothesis. An acquisition is defined as diversifying if the acquiring firm acquires a firm, which is active in another

industry. Following Malmendier and Tate (2008) firms can be subdivided in 48 Fama-French industry groups. Each firm is assigned to an industry group based on its four-digit Standard Industrial Classification (SIC) code. However, the SIC codes are constructed in a ranked structure where the first two digits indicate the industry sector in which the firm is operating. Subsequently, a binary variable is constructed, which equals one if the target firm is

operating in another industry than the acquiring firm, based on the first two digits of the SIC code, or zero otherwise. All of these dependent variables are binary variable. Therefore, multiple logit regressions will be conducted. These logistic regressions describe the relationship between the dependent binary variable with the independent variables. The different regression models will be described in more detail later in this section.

(20)

Market reaction to the announcement of mergers and acquisitions

In order to test the last two hypotheses on the effect of CEO overconfidence on the

immediate market reaction after an M&A announcement, the stock return of the acquiring firm will be examined. It is assumed that investors are rational and the market is considered to be efficient. This means all information is available for the entire market and stock prices will be fairly priced. More specifically, investors immediately act on newly available information, whereby the stock price will be adjusted. An event study will be conducted to measure the cumulative abnormal returns (CAR) of the acquiring firm, capturing the effect of the announcement on stock return. The event will be the announcement of a merger or

acquisition and a three-day event window is chosen to measure the cumulative abnormal return. Stock prices should be adjusted perfectly to capture the effect of this announcement. Abnormal returns are measured by taking the difference between the actual return and the expected normal return, estimated by the capital asset pricing model (CAPM). This is illustrated by the following formula:

𝐴𝑅𝑖𝑡 = 𝑅𝑖𝑡− (𝛼𝑖 + 𝛽𝑖𝑅𝑀𝑡)

Where 𝐴𝑅_𝑖𝑡 is the abnormal return for stock i at time t. 𝑅_𝑖𝑡 is the actual return of stock i at time t. The expected return of stock i is estimated by the CAPM model denoted by (𝛼𝑖 +

𝛽_𝑖𝑅_𝑀𝑡). In this equation, 𝑅_𝑀𝑡 denotes the return on the market at time t and 𝛽_𝑖 denotes the co-movement of stock i with the market. The expected return is estimated over an estimation period of 252 trading days prior to the announcement. A gap of 42 days is kept between the end of this estimation period and the beginning of the event window. If the estimation period and the event window would overlap, the normal return model parameters would be

estimated based on returns affected by the event, resulting in biased outcomes.

Cumulative abnormal returns are then computed by cumulating the abnormal returns of firm i over the three-day event window. The formula for calculating cumulative abnormal returns is as follows:

𝐶𝐴𝑅_{𝑖(𝜏1,𝜏2)} = ∑ 𝐴𝑅_𝑖𝑡

𝜏2

(21)

The cumulative abnormal return of each acquiring firm i in the sample over the three-day event window, from one day prior to the announcement (𝜏1) until one day after the

announcement (𝜏2), is calculated by cumulating the daily abnormal returns of this firm. This cumulative abnormal return will be used as dependent variable to test hypothesis (4) and (5). Additionally, the cumulative abnormal returns with alternative event windows of five days and eleven days are constructed to check for robustness of the results.

3.3 Control variables

In this research multiple regressions are conducted in order to test the hypotheses stated. Based on related prior research, several control variables are included to capture the distinctive effect of CEO overconfidence on announcement returns, method of payment, diversification and acquisitiveness. The control variables are divided into three subsections, which are deal specific control variables, firm specific control variables and CEO specific control variables. This section will briefly describe all control variables and explains how they are constructed. Moreover, the list of all variables and how they are constructed is reported in table 9, which is included in the appendix.

Deal specific control variables

The created independent variables for method of payment and diversification will also be included as control variables when measuring the effect of CEO overconfidence on the cumulative abnormal return of the acquiring firm. Additionally, deal size is included as a control variable, measured as the value paid in the transaction to acquire the target firm. Subsequently, the natural logarithm of this value is taken. Martin (1996) expected a negative relationship between the deal size and the use of cash to finance a deal. However, in his research on the effects of transaction value on the use of cash, stock or a combination of both stock and cash, the results were ambiguous (Martin, 1996).

CEO specific control variables

First of all, the effect of gender is controlled for by creating a binary variable equaling one if the acquiring CEO is a male and equaling zero if the acquiring CEO is female. Huang and Kisgen (2013) state that male executives more often engage in acquisitions and

announcement returns are approximately 2% lower than announcement returns of acquisitions undertaken by female executives. Following Hirshleifer et al. (2012), stock

(22)

ownership of the CEO is included to control for the proportion of stock the CEO holds when performing the merger or acquisition. This variable is measured as the percentage of shares owned by the CEO, excluding options. If this proportion of stock is high, the CEO is more likely to act in the interest of the shareholders and he is less likely to be overconfident. Additionally, CEO tenure is included to control for the number of years the CEO has held office. Some observations specifying the date a CEO is retired is either missing or incorrect in Compustat, resulting in negative values of CEO tenure. Those observations are excluded from the final sample. CEO tenure is included since it is likely that CEO’s, who are new in office, are more eager to prove themselves. Therefore, they might undertake more mergers and acquisitions. At last, the proportion of fixed compensation might also influence CEO overconfidence. This variable is constructed as the proportion of salary to total compensation. Theory predicts that CEO overconfidence is associated with a higher proportion of fixed compensation since the total CEO compensation is less dependent on his own efforts. Controlling for compensation decreases the probability that the overconfidence measure includes compensation incentives instead of measuring actual overconfidence (Schrand & Zechman, 2012).

Firm specific control variables

The first firm specific variable that is controlled for is firm size. Moeller et al. (2004) showed that the announcement returns for the acquiring firm are approximately 2% higher for relative small acquiring firms, irrespective of the method of payment, and this effect remains over time. Therefore, this variable is included in the regression analyses, measured as the natural logarithm of total assets at the beginning of the year in which the merger or acquisition is announced. Investment opportunities are controlled for including Tobin’s Q as a control variable. Following Malmendier and Tate (2008), Tobin’s Q is measured as the ratio of market value of asset to the book value of assets. The book value of assets is denoted by total assets. However, the market value of assets is defined as total assets plus market equity, subtracted by book equity. Firms with lower Tobin’s Q tend to acquire more, implying that acquisitions might be a substitution for profitable investment opportunities (Malmendier & Tate, 2008). Although, Hirschleifer et al. (2012) state that Tobin’s Q is positively related to CEO overconfidence. Besides, a higher Tobin’s Q of the acquiring firm results in higher abnormal stock returns for both acquiring as target firm, especially when Tobin’s Q of the target firm is low (Servaes, 1991). This variable is trimmed at the 1% level to decrease the positive skewness of the distribution. Internal levels of resources are controlled for by

(23)

including the control variable cash flow. This variable is constructed as income excluding extraordinary items plus depreciation and is also trimmed at the 1% level to exclude outliers. Additionally, cash flow is normalized by capital at the beginning of the year. Malmendier and Tate (2005) already stated a significant positive relationship between the sensitivity of

investment to cash flow and CEO overconfidence. Moreover, the more internal funds available, the lower are financial constraints. This will most likely lead to an increase in M&A activity and a higher probability of the deal being financed with cash. Similarly, the leverage ratio of the acquiring firm is included as a control variable. An increase in the leverage ratio will most likely negatively affect the availability of cash. Besides, a high leverage ratio could financially constraint firms, whereas these firms might not be able to undertake mergers and acquisitions. Thus, a negative relationship between leverage ratio and acquisitiveness as well as the deal being financed by cash is expected. Leverage is measured as total liabilities over total assets. Furthermore, Moeller et al. (2004) include the ratio of total debt over firm’s market value to control for its effect on cumulative abnormal returns.

At last, a proxy for corporate governance is included, measured by board

independence. Following the approach of Brown and Sarma (2007), the proxy for effective corporate governance is measured by the proportion of independent directors on the entire board of directors. Directors are considered independent if they are not currently employed at the firm. Abnormal returns should be negatively related to the proportion of outsiders on the board, but positively related to the proportion ownership held by outsiders on the board (Subrahmanyam, Rangan & Rosenstein, 1997).

3.4 Research design

In the previous section the measurement of CEO overconfidence and the construction of the other dependent variables are defined. Moreover, the construction of all control variables is explained. These variables will serve in multiple regression models in order to test the stated hypotheses. The first hypothesis predicts that acquiring CEO overconfidence will yield a positive effect on overall M&A activity. To test this hypothesis a logit regression is conducted since the dependent variable is a binary variable. In a logistic regression the regression model allows the dependent variable to be categorical. Hypothesis (1) is tested using the following model:

(24)

(1,2,3) Pr{𝑌𝑖𝑡 = 1|𝑂𝐶𝑖𝑡, 𝑋𝑖𝑡} = 𝐺(𝛽1+ 𝛽2𝑂𝐶𝑖𝑡+ 𝑋𝑖𝑡′𝐵) (1)

Where Pr denotes the probability distribution of variable 𝑌_𝑖𝑡 equaling one. Variable 𝑌_𝑖𝑡 is a binary variable and will equal one if a firm announced at least one successful merger or acquisition in a particular firm year. G is assumed to be the logistic distribution and the variable 𝑂𝐶_𝑖𝑡 is the measure for overconfidence based on the option holding behavior of the acquiring CEO. 𝑋_𝑖𝑡 is a set of control variables. Hypothesis (2) is tested by using the same logistic framework shown in equation (1). Although, this hypothesis states that acquiring CEO overconfidence will positively affect the probability of a deal being completely financed by cash, so 𝑌𝑖𝑡 is a binary variable equaling one if a deal is completely financed by

cash, or zero otherwise. The third hypothesis suggests that acquiring CEO overconfidence will positively affect the probability of a merger or acquisition being classified as

diversifying. Hypothesis (3) is also tested with the same logistic framework shown in

equation (1), with the dependent variable 𝑌𝑖𝑡 denoting one if the acquiring firm and the target

firm are operating in a different industry, based on the two-digit SIC codes, and zero

otherwise. In all three regressions a different set of control variables (𝑋_𝑖𝑡) is included, which will be illustrated in the regression output in section 5. Moreover, the odds ratio of all independent variables are presented. These ratios will represent the odds that an outcome of the dependent variable Y will occur given a certain exposure to dependent variable X relative to the odds of the outcome in absence of that exposure. To test for a positive effect of the dependent variable on the independent variable, the reported ratios are compared to 1. Odds ratios can be computed as follows:

𝑂𝑑𝑑𝑠 𝑟𝑎𝑡𝑖𝑜 = 𝑒𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 ₍₂₎

Hypothesis (4) and hypothesis (5) are tested by performing the an Ordinary Least Squares (OLS) regression. The regression model is stated below:

(4, 5) 𝐶𝐴𝑅_𝑖𝑡 = 𝛼 + 𝛽₁𝑂𝐶 𝐴𝑐𝑞𝑢𝑖𝑟𝑒𝑟 + 𝛽₂𝑂𝐶 𝑇𝑎𝑟𝑔𝑒𝑡 + 𝛽₃𝑂𝐶 𝐴𝑐𝑞𝑢𝑖𝑟𝑒𝑟 ∗ 𝑂𝐶 𝑇𝑎𝑟𝑔𝑒𝑡 + 𝛽₄𝐹𝑖𝑟𝑚 𝑆𝑖𝑧𝑒 + 𝛽₅𝐶𝑎𝑠ℎ 𝐹𝑙𝑜𝑤 + 𝛽₆𝑇𝑜𝑏𝑖𝑛′_{𝑠 𝑄 + 𝛽} 7𝐶𝑎𝑠ℎ + 𝛽8𝐷𝑖𝑣𝑒𝑟𝑠 + 𝜀 (3)

(25)

In equation (3) the effect of acquiring CEO overconfidence on the market reaction is examined, controlled for target CEO overconfidence and firm characteristics such as firm size, cash flow and Tobin’s Q. Moreover, the two binary variables Cash and Divers are included, capturing the method of payment of the deal and whether the deal is considered diversifying. Additionally, an interaction variable is constructed between acquirer CEO overconfidence and target CEO overconfidence, allowing to examine the effect of

simultaneous overconfidence of both acquiring as target CEO on the response of the market. This will also change the interpretation of 𝛽1, whereas this variable now only captures the

effect of acquiring CEO overconfidence in the situation the target CEO is not classified as overconfident. Additionally, the proxy for corporate governance as well as the CEO’s gender and his proportion of fixed salary to total compensation are added to check for robustness of the results. This leads to the following regression:

(4, 5) 𝐶𝐴𝑅_𝑖𝑡 = 𝛼 + 𝛽₁𝑂𝐶 𝐴𝑐𝑞𝑢𝑖𝑟𝑒𝑟 + 𝛽₂𝑂𝐶 𝑇𝑎𝑟𝑔𝑒𝑡 + 𝛽₃𝑂𝐶 𝐴𝑐𝑞𝑢𝑖𝑟𝑒𝑟 ∗ 𝑂𝐶 𝑇𝑎𝑟𝑔𝑒𝑡 + 𝛽4𝐹𝑖𝑟𝑚 𝑆𝑖𝑧𝑒 + 𝛽5𝐶𝑎𝑠ℎ 𝐹𝑙𝑜𝑤 + 𝛽6𝑇𝑜𝑏𝑖𝑛′𝑠 𝑄 + 𝛽7𝐶𝑎𝑠ℎ + 𝛽₈𝐷𝑖𝑣𝑒𝑟𝑠 + 𝛽₉𝐵𝑜𝑎𝑟𝑑 𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑐𝑒 + 𝛽₁₀𝐺𝑒𝑛𝑑𝑒𝑟 + 𝛽₁₁𝐹𝑖𝑥𝑒𝑑 𝐶𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛 + 𝜀 (4)

The cumulative abnormal returns over a five-day event window and a eleven-day event window will be used as dependent variables in the same regression models of equation (3) and equation (4) as a check for robustness of the results.

(26)

4. Data

4.1 Sample selection

This section describes how the data is collected and subsequently how the sample is

constructed. In the next section summary statistics are included to provide an overview of the final sample. The final dataset contains data on mergers and acquisitions which are obtained from the Thomson One (former SDC Platinum) database. The sample consists of mergers and acquisitions announced between 1993 and 2016. All acquiring and target firms are US public firms since most information is available for these firms. Besides, it simplifies the merging with other databases such as CRSP and Compustat. Additional restrictions on selecting the data in the database of Thomson One are applied. First of all, the transaction must be

completed and the deal value must be greater than $10 million. Besides, acquisition deals in which the deal value is relatively small will be excluded. Announcements in which the deal value is smaller than 5% of the total equity of the acquiring firm, measured as the common equity value over the last twelve months, are considered small as these deals are too small to be affected by the CEO’s non-standard beliefs, such as overconfidence. Additionally, all mergers and acquisitions in which the acquiring firm is aiming to own less than 50 % of the shares after the deal are excluded from the dataset. At last, financial and utility firms are excluded from the sample. Firms with SIC codes in the range of 4900 – 4999 and 6000 – 6999 are considered financial and utility firms, respectively. These firms operate in a special regulatory environment, whereas the corporate governance structure and operational

characteristics are significantly different from non-financial firms (Subrahmanyam et al., 1997).

In the Thomson One database several variables are obtained in order to construct dependent variables and deal specific control variables. These variables include the deal announcement date, percentage of shares aimed to acquire, percentage of shares acquired, status of the deal, value of the transaction, ratio of the value of transaction to common equity and the proportion of the transaction being financed by either cash or stock. Additionally, the firm names, 6-digt CUSIP codes and industry SIC codes of both acquiring firms and target firms are collected. An overview on transforming the initial dataset into the final M&A dataset is presented in table 1. In order to measure CEO overconfidence, information on option based grants per CEO is retrieved from the ExecuComp database using Wharton

(27)

Research Data Services (WRDS). This data is specified as annual data and covers the same range of years between 1993 and 2016. Variables which are included are the number of shares acquired upon option exercise as well as the value realized upon option exercise. Additionally, all data needed for the construction of the CEO specific control variables is also extracted from this database. The most important variables which are included are the age and gender of the CEO, the date he started and left as CEO, the percentage of total shares owned and the CEO’s salary and total compensation scheme.

The closing price of each fiscal year is required for the measurement of moneyness and subsequent overconfidence. This variable is retrieved from the CRSP database alongside variables for the construction of the firm specific control variables. These variables are retrieved during the period between 1992 and 2016, because the computation of firm size requires information on a firm’s total assets one year prior to the year of interest. The most important firm specific control variables which are included are total assets, total liabilities, depreciation and amortization, net income, common shares outstanding and capital

expenditures. For the computation of Tobin’s Q additional variables are required, including the preferred stock liquidation value, stockholder’s equity and the deferred taxes and investment tax credit.

The next step is to merge the different datasets into one final dataset. However, all the databases use different unique identifiers. The Thomson One database identifies firms by 6-digit CUSIP codes, whereas the CUSIP codes in the CRSP database could have changed over time. However, the CRSP database also contains historical CUSIP codes (NCUSIP), which correspond to a certain date. The historical CUSIP codes consist of 8 digits, so first they are converted into 6-digit CUSIP codes. The data on mergers and acquisitions is now merged with the CRSP database on the correct CUSIP codes corresponding with the announcement dates of all mergers and acquisitions. As a result the PERMNO codes of the acquiring firms is retrieved. The PERMNO codes are transformed into GVKEY codes by using the

CRSP/Compustat merged linking table in WRDS since the Compustat database uses GVKEY to identify firms. The CEO specific data is now merged with the combined dataset of mergers and acquisitions and firm specifics by merging on corresponding GVKEY codes.

In addition, the data required for the computation of board independence is retrieved from the Institutional Shareholder Services (ISS) database. This data is merged into the final dataset by merging on 6-digit CUSIP codes of the acquiring firms. However, not every firm-year observation includes information on board independence since the ISS database only contains data from 1996 until the present. Besides, data on the composition of the board of

(28)

directors is missing for some firm years. At last, the cumulative abnormal returns are computed by taking the difference of the actual return and the estimated return and are cumulated over the three-day event window surrounding the announcement date of a merger or acquisition. Additionally, the cumulative abnormal returns over event windows of five days and eleven days are computed to examine the robustness of the results.

4.2 Descriptive statistics

In this section several tables are presented to provide a clear overview of the variables examined and subsamples of interest. Table 1 provide information on the restrictions applied to the merger and acquisition dataset. Initially, data on all mergers and acquisitions is

retrieved from the Thomson One database. After applying restrictions on this data a final sample of 3306 different mergers and acquisitions remains.

Table 1: Thomson One deal restrictions and sample size.

This table presents the restrictions applied to the merger and acquisition data retrieved from the Thomson One database. The final subsample consist of 3306 different mergers and acquisitions.

Variable Restriction Explanation Sample size

Announcement date Between 01/01/1993 – 31/12/2016 n/a Acquirer nation Include United States of America 267714 Target nation Include United States of America 224085

Acquirer status Include Public 109313

Public status Include Public 30630

Deal value ($) Minimum 10.000.000 20499

Deal status Include Completed 7965

Deal value (%) Minimum 5% of equity value acquirer 7958 Aimed to acquire (%) Minimum 50% of target shares 6008 Acquire SIC code Exclude 4000-4999 & 6000-6999 3306

Subsequently, this dataset is merged with the CEO specific data and the firm specific data. Observations containing missing values in any of these datasets are removed from the final sample. At last, the data on board independence is merged into the final sample and

(29)

mergers and acquisitions performed by 1128 different CEO’s in the timespan between 1993 and 2016. Table 10 in the appendix shows the distribution of all mergers and acquisitions over the sample years and what proportion of these deals is undertaken by an overconfident CEO. Besides, table 10includes the average deal value per sample year and both the proportion of deals which are completely financed by cash as the proportion of deals which are diversifying. Furthermore, the average cumulative abnormal return per year is presented.

In table 2 the summary statistics of all variables are presented. Panel A of this table shows the summary statistics on CEO overconfidence, where the high overconfidence measure, the low overconfidence measure and the average percentage moneyness is also included. Between the mean and median of moneyness is an absolute difference of 0.39 which means that the final sample consists of a relative small group of CEO’s, who’s option grants are very deep in-the-money. Besides, 57 % of all deals is undertaken by an

overconfident CEO and a distinction is made between highly overconfident CEO’s and CEO’s with low confidence levels. In the final sample, 43 % of the CEO’s is considered highly overconfident, whereas 6% of the acquiring CEO’s has a low confidence level. CEO’s of target firms are almost evenly overconfident compared to acquiring CEO’s, that is 56 % is overconfident on average. After merging the different datasets only 486 observations of target CEO overconfidence was matched with the target firms. The course of overconfidence amongst acquiring and target CEO’s over the sample years is presented in figure 1in the appendix.

Panel B presents the deal specific variables, where the cash and diversifying binary variables are used as dependent variables. The mean of the cash approaches half of the total sample, with 52 % of the deals being financed completely by cash. This is similar to the percentage of cash used as method of payment stated by Moeller et al. (2004). However, in their sample 40% of all deals is considered a pure cash deal. Furthermore, 30 % of all deals is considered a diversifying merger or acquisition. The mean of deal value is significantly higher than the median implying that the sample consists of relatively much small deals and a couple of very large deals. The largest amount paid in a transaction by the acquiring firm is more than $164 billion. The percentage of target shares acquired on average is close to the percentage of target shares aimed to acquire. In panel C variables regarding CEO specifics are presented. The ratio of fixed salary to the CEO’s total compensation scheme is 0.21 on average, whereas some CEO’s are only compensated by variable payments instead of a fixed salary. Furthermore, 98 % of all CEO’s is male and the average tenure is about 11 year,

(30)

which is significantly higher than the average of 8.5 years found by Malmendier and Tate (2008).

Table 2: Summary statistics

Panel A shows the summary statistics of the variables regarding CEO overconfidence. Moneyness is calculated as the average percentage moneyness derived following the approach of Campbell et al. (2011). Acquirer overconfidence is a binary variable equaling one for all years an acquiring CEO fails to exercise his option grants which are at least 67% in-the-money and zero otherwise. Similarly, target overconfidence is constructed for target CEO’s. High overconfidence is included as a robustness check. Panel B shows the deal specific variables. Cash and diversifying are both dependent binary variables. Cash equals one if a deal is completely financed by cash, or zero otherwise. The diversifying binary variable equals one if the target firm operates in another industry than the acquiring firm, based on the two-digit SIC codes, or zero otherwise. The other variables are included as control variables, where deal value is denoted in millions of dollars. The variables in panel C and panel D are also control variables, where salary and total compensation are denoted in thousands of dollars and tenure in years. Cash flow is denoted in millions of dollars and firm size is the natural logarithm of total assets. Panel E shows the summary statistics on the cumulative abnormal returns of the acquirer. The five-day and eleven-five-day event windows are included for robustness checks.

Variable Observations Mean Median S.D. Minimum Maximum

Panel A: CEO Overconfidence

Moneyness 1821 0.94 0.55 1.12 0.00 6.24 Acquirer overconfidence 1821 0.57 1.00 0.50 0.00 1.00 High overconfidence 1821 0.43 0.00 0.50 0.00 1.00 Low overconfidence 1821 0.06 0.00 0.24 0.00 1.00 Target overconfidence 486 0.56 1.00 0.50 0.00 1.00

Panel B: Deal specific variables

Cash (binary) 1821 0.52 1.00 0.50 0.00 1.00 Diversifying (binary) 1821 0.30 0.00 0.46 0.00 1.00 Deal value 1821 1670 341.57 5893 10.00 164747 Shares aimed to acquire (%) 1424 99.05 100.0 5.29 50.25 100.00 Shares acquired (%)

Panel C: CEO specific variables

1542 92.03 100.0 24.76 0.21 100.00 Salary 1821 847.48 795.04 475.60 0.00 8100 Total compensation 1816 8516 5053 11649 42.61 134458 Fixed compensation (%) 1816 0.21 0.16 0.17 0.00 1.00 Ownership (%) 962 2.28 0.48 5.09 0.00 50.68 Gender 1821 0.98 1.00 0.13 0.00 1.00 Tenure

Panel D: Firm specific vaiables

1224 11.12 9.51 7.76 0.05 43.44 Firm size 1804 8.05 7.96 1.67 3.22 12.91 Cash flow 1786 0.38 0.27 0.89 -12.32 14.33 Tobin’s Q 1584 2.11 1.75 1.15 0.65 7.07 Leverage 1814 0.53 0.53 0.20 0.04 1.79 Independent board 826 0.67 0.70 0.17 0.11 0.92

Panel E: Cum. Abnormal Returns

Acquirer CAR [-1,1] 1821 -0.01 -0.00 0.07 -0.39 0.31 Acquirer CAR [-2,2] 1821 -0.01 -0.00 0.08 -0.45 0.64 Acquirer CAR [-5,5] 1821 -0.01 -0.00 0.09 -0.74 0.46

(31)

Panel D of table 2 state the firm specific variables which is controlled for in the various regression models. Cash flow is denoted as income before extraordinary items plus depreciation, normalized by capital at the beginning of the year. The average ratio of 0.38 and the median of 0.27 are similar to the results stated by Malmendier and Tate (2008). The average and median of Tobin’s Q are 2.11 and 1.75, respectively. Values higher than 1 indicates that the firm’s market value is higher than the replacement cost of its recorded assets, which implies that the market overvalues that firm. The mean of 2.11 is similar to the average Tobin’s Q in the sample of Moeller et al. (2004). The independent board variable controls for corporate governance and is measured as the proportion of independent, outside directors on the board of directors. The mean of 0.67 indicates that in the final sample two-third of the directors on the board of directors of the acquiring firm are not employed at the firm at the time a merger of acquisition is announced. In panel E of table 2 the summary statistics of the cumulative abnormal returns over the different event windows is presented. These variables serve as dependent variables in the regressions testing the market’s response to the announcement of a merger or acquisition. On average, the cumulative abnormal returns are negative for each of the event windows as prior research would suggest (Malmendier & Tate, 2008; Hayward & Hambrick, 1997). The spread between negative cumulative abnormal returns and positive cumulative abnormal returns is increasing when the event window is extended.

(32)

5. Results

In this section the empirical findings of the different regression models are presented and discussed. First, the results testing the first hypothesis on the effect of acquiring CEO overconfidence on acquisitiveness are presented. Thereafter, the results testing the second hypothesis on implications of acquiring CEO overconfidence on the method of payment in mergers and acquisitions are reported. Afterwards, the results testing the third hypothesis on the effect of acquiring CEO overconfidence on the probability a deal is classified as

diversifying are presented. At last, the results testing whether overconfidence of both

acquiring and target CEO’s affects the market reaction to a deal announcement are reported. The results will be discussed on their statistical and economical significance. Moreover, the results will be compared to results stated in previous literature. Additionally, robustness checks are performed at the end of this section.

5.1 CEO overconfidence and M&A activity

The first hypothesis states that acquiring CEO overconfidence should have a positive effect on overall acquisitiveness. Table 3 presents the results of a logistic regression, where the dependent variable is a binary variable equaling one if a firm announced at least one successful merger or acquisition in a particular firm year, or zero otherwise. The first three columns control for common firm specific variables, whereas the last three columns are included to check whether the effect of overconfidence on acquisitiveness is also robust to CEO specific control variables. Furthermore, in regression (2) and (5) year fixed effects are included and regression (3) and (6) also include industry fixed effects to control for the effect of time-invariant industry trends. Based on regression (1) of table 3, hypothesis (1) should be accepted. Acquiring CEO overconfidence shows a positive relationship with the probability a successful deal takes place, significant at a 1% significance level. Since this is a logistic regression analysis, the odds ratios are provided in brackets. The coefficients in a logistic regression are a logarithmic transformation of the odds ratio. The odds ratio denotes the odds of an overconfident CEO to be more acquisitive as opposed to a rational non-overconfident CEO. The odds ratio of 1.320 means that the odds of an overconfident CEO undertaking a successful merger or acquisition is 32% higher than the odds a non-overconfident CEO does so. This is consistent with the findings of Malmendier and Tate (2008).

Implications of CEO overconfidence in mergers and acquisitions