Risk arbitrage in takeovers:; investigating risk arbitrage spreads and takeover success prediction in emerging markets

University of Amsterdam, Amsterdam Business School

Msc Finance: Quantitative Finance track

Master Thesis

Risk arbitrage in takeovers; Investigating Risk Arbitrage Spreads

and Takeover Success Prediction in Emerging Markets

Author: Kuiper, Vincent

Student Number: 10715878

July 2018

2

Abstract

This thesis examines whether emerging markets are attractive to perform risk arbitrage. To analyze the risk-reward relationship, I investigate how risk arbitrage spreads (reward) and takeover success probabilities (risk) differ in emerging markets, compared to developed markets. This study has several implications for the hedge fund industry and in general for the research on mergers and acquisitions. It provides an improved specification of the determinants of arbitrage spreads and takeover success prediction. Spreads and takeover success probability are analyzed separately for deals with target firms in emerging and developed markets. This study finds that spreads in emerging markets do not significantly differ from spreads in

developed markets. However, deals in emerging markets are associated with a higher takeover failure probability, compared to developed markets. This means that emerging markets appear to be less attractive for risk arbitrageurs, since the risk associated with the higher deal failure probability is not compensated by a higher arbitrage spread. Furthermore, positive M&A waves are associated with higher probability of deal success, whereas it is not found to significantly affect the spread. A noteworthy finding in this study is that arbitrage spreads are found to be negatively associated with takeover failure probability. To address endogeneity, fixed-effects models have been used and additional robustness checks are done. These tests support the main findings of this thesis.

Keywords: Risk Arbitrage, Merger Arbitrage Spread, Takeover Success Prediction, M&A

3

Statement of Originality

This document is written by Vincent Kuiper 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

4

Table of Contents

1. Introduction ... 5

2. Literature review and research question ... 9

2.1. Takeover Market and M&A waves ... 9

2.2. Risk Arbitrage Spread ... 11

2.3. Determinants of Takeover Success Prediction ... 15

2.4. Emerging and Developed Markets ... 17

3. Methodology ... 19

3.1. Variables ... 19

3.2. Econometrical Models ... 25

3.3. Addressing Endogeneity ... 27

3.4. Relation of significance and signs to hypotheses ... 28

4. Data and Descriptive Statistics ... 29

4.1. Sample Construction ... 29

4.2. Summary Statistics ... 31

4.3. Parametric Tests ... 34

4.4. Correlation Matrices ... 35

5. Results ... 36

5.1. Results OLS Regression Models (1) – (3) ... 36

5.2. Results Logistic Regression Models (4) – (6) ... 40

6. Robustness Checks ... 44

7. Conclusion ... 45

7.1. Summary and Discussion ... 45

7.2. Limitations and suggestions for further research ... 47

APPENDIX I: Observations by country and year ... 49

APPENDIX II: Correlation Matrices ... 52

APPENDIX III: Parametric tests ... 54

APPENDIX IV: Robustness Checks ... 56

APPENDIX V: Findings in previous related studies ... 64

5

1. Introduction

Risk arbitrage is the investment strategy designed to benefit from uncertainties surrounding the outcome of takeover bids (Nguyen, 2009). Since 1990, the assets under management of risk arbitrage hedge funds have grown with 33,934% from $233 million to $79.30 billion, by the end of 2017 (Barclays Hedge 2018). Furthermore, Hedge Fund Research states that risk arbitrage strategies, also known as merger arbitrage strategies, normally have over 75% of positions in announced corporate transactions over a certain market cycle. This shows the upward surge of risk arbitrage strategies and how arbitrageurs use this approach more and more as investing opportunity. Risk arbitrage is an investment strategy that involves buying shares, immediately after an announced tender offer, of the target company and holding it till the deal is consummated. Prior literature consistently finds positive abnormal returns for this arbitrage strategy. (e.g., Jindra & Walkling, 2004; Branch & Yang, 2003; Mitchell & Pulvino 2001). According to Jetley and Ji (2010), the objective of this strategy is to benefit from the so called ‘arbitrage spread’. This arbitrage spread is defined by the percentage difference that the target’s immediate post-announcement share price trades below the initial offer price. The spread1 shows the market’s pricing of the target firm conditional on the existence of the bid (Jindra and Walkling, 2004). The risk for the arbitrageur is that the deal may fail, leading to a decline in the target’s share price and thus a loss. If the deal is completed, the arbitrageurs make a profit that equals the difference between the initial offer price and the target’s post-announcement share price.

For risk arbitrageurs, it is important to know the probability of takeover success, since completion or failure directly affects their profits. Completion of the deal means profits for risk arbitrageurs, whereas deal failure results in declining target’s share price and so losses for risk arbitrageurs. Thus, to have knowledge about the prediction of takeover success is highly beneficial for risk arbitrageurs, since this will help them to identify profitable situations. According to Cornelli and Li (2002), risk arbitrage is the most important element to affect the success rate of takeovers. Therefore, the relationship between the risk arbitrage (spreads) and takeover success prediction is interesting to investigate.

Furthermore, knowledge of different market classifications in terms of development can help risk arbitrageurs even more. Schleifer and Vishny (1997) investigate the limits of arbitrage and suggest further research on why some markets are more attractive for arbitrage than others. For example, Jetley and Ji (2010), state that due to the increasing amount of money and

6

attention for risk arbitrage in developed markets, average spreads have declined over the last decades. In contrast, risk arbitrage in emerging markets has never been examined. This implies that deals in new markets, such as emerging markets, may still contain the large and promising spreads as developed markets contained when the profession of risk arbitrage began four decades ago. Because of the rise of target firms located in emerging markets (EMTFs2

) in takeovers (Deng and Yang, 2015), emerging markets might include new investment opportunities for risk arbitrageurs.

Likewise, with the increasing integration of the world economy, takeovers will involve more often companies from different countries (Erel, Liao & Weisbach, 2010). Duppati and Rao (2015) investigate mergers and acquisitions between mature markets and emerging markets, focusing on USA and India. They state that there are indeed differences between the markets and that it would be interesting to extend the study to other emerging markets. Trade and supply-chain barriers, often associated with emerging markets, can cause uncertainty. This may be reflected in the spread or the takeover success probability (Control Risks, 2017). Additionally, different cultures, difficult market access and less transparent information can cause deals to fail more often in emerging markets, which increases the level of risk for takeovers in emerging markets.

Hence, my research has several implications for the hedge fund industry, for private investors who entrust their money to these hedge funds and for the research in general on the field of M&A. For the hedge fund industry, an improved specification of arbitrage spread and takeover success prediction is provided, together with a suggestion whether hedge funds should perform risk arbitrage in emerging markets. Considering the implications for the research on the field of M&A, an answer is provided to the question whether the generally used deal and firm characteristics affect takeovers differently, comparing emerging and developed markets.

1.1 Objectives

The main objective of this thesis is to show risk arbitrageurs whether it is attractive to invest in emerging markets. Therefore, the first objective of this thesis is to investigate how EMTFs among other variables affect risk arbitrage spreads, following the model of Officer (2007). By performing fixed-effects models, it is examined whether there are both statistical and

economical significant results.

7

To complete our main objective, the second objective of this thesis is to investigate how EMTFs among other variables (e.g., arbitrage spread) affect takeover success probability. Following the logistic regression model of Branch, Wang and Yang (2008), different effects on takeover failure probability are found when EMTFs are involved.

These two objectives together form the fundament of this research, since both results will affect the attraction of emerging markets for risk arbitrageurs. Hence, the main research question of this thesis is: Is it attractive for risk arbitrageurs to perform risk arbitrage in emerging

markets?

As a sub-analysis, this thesis also examines different effects of M&A waves. It is investigated how periods prior to crises (positive wave) and periods of crises (negative wave) influence the attraction of performing risk arbitrage.

1.2. Contribution

Some research has been done on risk arbitrage spread and on takeover success prediction separately. However, literature on the relationship between risk arbitrage spread and takeover success probability is scarce. Moreover, the extant research lacks investigating the effects of different market classifications on this spread and takeover success prediction. There is a gap in analyzing this topic from a comparative approach regarding emerging and developed markets. This thesis adds to the existing literature by an improved and innovative specification on the causal relationship between risk arbitrage spreads and takeover success prediction, by comparing emerging markets and developed markets. It follows the structure of papers like Jindra and Walkling (2004) and Officer (2007), by focusing first on arbitrage spread and then focusing on the main analysis of the paper. Prior research has suggested multiple times that the relationship between the spread and takeover success is very interesting to investigate (e.g., Jindra and Walkling 2004, Cornelli and Li 2002, Officer 2007), but often lacks investing it. The investigation of M&A waves can be seen as a sub-analysis. By controlling for these waves, broad macro economical effects on risk arbitrage spread and takeover success prediction are tested. Therefore, in this thesis, answers to the following sub questions will be provided:

1.

2.

8

4.

5.

Trying to achieve the thesis’ first objective, the model of Officer (2007) will be used as fundament to analyze the determinants of risk arbitrage spread, together with additions to analyze the effect of M&A waves and EMTFs. Furthermore, two additional regressions are done, where sample is split up in EMTFs and DMTFs involved. Also country fixed effects are added and robustness checks are done to address for endogeneity problems. It shows that the traditional predictors for arbitrage spreads in developed markets have less predictive power for spreads in emerging markets.

The multivariate logistic regression model of Branch, Wang and Yang (2008) will be used as fundament to identify the effect of EMTFs on takeover success prediction. Previous literature is followed to include the most powerful predictor variables (e.g., Walkling, 1985; Hoffmeister & Dyl, 1981; Schwert, 2000). The additional tests done follow the same structure as mentioned above: the sample is split up in EMTFs and DMTFs. Additionally, country and time fixed effects are included as well.

Hence, for risk arbitrageurs a high-risk high-reward possibility in emerging markets is expected, which should be reflected in wider spreads and higher takeover failure probabilities. This study finds that indeed takeovers in emerging markets are more likely to fail, but this is not compensated by a significant wider spread. Hence, emerging markets do not appear to be attractive for risk arbitrageurs. Furthermore, deal and firm characteristics explain less how spreads and takeover failure probability are formed in emerging markets, compared to

developed markets. This confirms the prediction that other factors such as cultural aspects are more important in emerging markets.

A noteworthy finding is that spreads are found to be negatively associated with takeover failure probability. This finding is robust for different methods of time intervals and

subsamples. This finding undermines the predictive power of the market regarding takeover outcome. Additionally I find that positive waves insignificantly affect arbitrage spreads, but significantly decrease the probability of deal failure. For risk arbitrageurs, this means that takeovers that occur in positive waves are beneficial, since they are associated with lower risk and equal rewards.

9

The outline of this thesis is as follows. In section 2 I will present the literature review. Section 3 describes the methodology in detail to show how I elaborate on the research design. Furthermore endogeneity issues are discussed. Section 4 describes how I obtained the data and descriptive statistics are presented. Section 5 shows the empirical results, following section 6 where I discuss robustness checks. In section 7, a summary and discussion of this thesis is presented. Furthermore, limitations to my research and suggestions for further research are described. This thesis ends with multiple appendices and the list of references.

2. Literature review

This section will elaborate on prior research relating to the topic of risk arbitrage (spreads), takeover success prediction and the distinction between emerging markets and developed markets. The section is structured as follows. First, in section 2.1, takeovers and M&A waves are concisely introduced. Subsequently, in section 2.2, prior research on risk arbitrage spreads is discussed. By following the main theories in the existing literature, the determinants of the spread are explained. Then, in section 2.3, previous literature on takeover success prediction is discussed. Section 2.4 offers a description about the main differences between takeovers in emerging and developed markets. These differences will show the importance of this research and why risk arbitrageurs should analyze emerging markets differently from developed markets.

2.1. Takeover market and M&A waves

This section will concisely introduce the takeover market and clarify how we define it. Understanding the takeover market and M&A waves directly affects how the preliminary and final samples in this study are constructed. The following paragraphs explain how takeovers and M&A waves are defined and how this influences the remainder of this thesis.

According to Manne (1965), the takeover market can be defined as a market where firms aim to compete with each other to take control of other firms’ corporate resources. Having influential power on a company’s management can be named as corporate control. Jensen and Ruback (1983) review much of the historical prior literature on this market. When firms believe they can use other firms’ resources to create more value, they can obtain control of these resources by doing acquisitions: when the bidding company acquires the target company, the target company’s control rights are transported to the board of directors of the acquiring company. In this way, they can obtain control of the corporate resources of a firm, by

10

acquiring the company itself.

Furthermore, they state that competing management teams can be seen as the primary activist entities, whereas stockholders play a passive, but primarily important, judicial role. Stockholders often just go for the highest bid, without being loyal to present managers. Arbitrageurs enter the process after the offer announcement, by valuing offers and providing liquidity in the stock trading.

In addition, they mention that takeovers can occur in different methods: mergers, tender offers or proxy contests. Mergers are negotiated directly with the target’s management team, whereas tender offers skip this step and offer their proposition directly to the target’s

shareholders. These shareholders then decide individually whether they accept the proposition and so tender their shares for sale to the acquiring company. Proxy contests occur when a rebellious group tries to obtain power in the board of directors. In the remainder of this paper, we acknowledge these three forms of takeovers. However, an acquisition of 5% of the shares when you had no shares prior to the acquisition is considered an investment instead of a takeover, since there is no real gain in control. Therefore, in the remainder of this paper, we mention a deal a takeover when prior to the announcement the acquiring party had less than 50% of the target’s shares and after the acquisition more than 50% of the target’s shares.

Over the last three decades, takeovers occur in so called waves. Positive waves can be defined as a period of increasing M&A activity in number and value. Negative waves are associated with periods of decreasing M&A activity. As figure 1 shows, positive waves often occur before financial crises. In the period between 1993 and 1999 and 2003-2007, we see a huge increase in M&A activity in number and value. Following theory, both periods resulted in crises (Tech Crisis in 2000, Credit Crisis in 2008). When the crises began, positive waves shifted to negative waves, leading to a period of years with decreasing value in M&A number and value. Hence, the M&A waves imply whether takeovers occur during crises or before crises.

11

Figure 1: Number & Value of M&A Worldwide

Kastrinaki and Stoneman (2012) investigate the determinants of M&A activity. They state that exogenous macroeconomic factors, average level of takeovers per industry, pre-emption effects and firm size have significant effects. Hence, including M&A waves in our models control for mostly macroeconomic factors, that could affect takeover success probabilities and so arbitrage spreads.

2.2 Risk arbitrage spread

In this section, risk arbitrage spread is explained in detail. By discussing its definition and main theories about it, this section tries to emphasize the importance for risk arbitrageurs of

understanding historical and statistical relationships considering the spread. This will be done based on prior research. Furthermore, the risk profile of risk arbitrage will be discussed. Section 2.2.1 and 2.2.2 explain the concept of risk arbitrage and what possible methods there are to calculate the spread, which are important features since different definitions and

quantifications could lead to different results. Finally, the determinants of the arbitrage spread are discussed in detail in section 2.2.3. By discussing these determinants, it is easier to

understand why spreads in emerging markets might differ from spreads in developed markets.

2.2.1. Definition

According to the Encyclopedia of Finance, risk arbitrage is defined as: ‘’Speculation on perceived mispriced securities, usually in connection with merger and acquisition targets.’’ In this paper, we focus solely on risk arbitrage in the merger and acquisition area.

12

Risk arbitrage in M&A’s differs from pure arbitrage, which is risk-free and requires no capital. Pure arbitrage includes buying and selling securities at different markets to benefit from price discrepancies. However, risk arbitrageurs incur a risk that the deal fails, or will be withdrawn. Furthermore, risk arbitrage is an event-driven investment strategy, in the way that it benefits from pricing inefficiencies after corporate events, such as takeovers (Barclays Hedge, 2008).

According to Weston et al. (2004), risk arbitrage can be defined as buying shares of takeover target companies after an offer is publicly announced, and holding this stock until the deal is consummated. This type of investing exists, since after an announced takeover bid, the target’s share price normally trades at a lower price than the initial offer price, due to the risk of deal failure. If the takeover eventually is consummated, the risk arbitrageur captures the

difference between the target’s stock price and the offer price, also known as the risk arbitrage spread. If the target’s share price trades at a higher price than the initial offer price, the market expects an upward revision by the initial bidder or an improved offer by other bidders.

2.2.2. Different quantifications

There are multiple ways to quantify the risk arbitrage spread. Different calculations of spreads may give different final results in this research. Hence, the different methods of calculation of the spread are discussed in this section. First, there is a distinction between cash offers and stock offers.

For cash offers, the arbitrage opportunity and construction of the spread are explained in the previous paragraph. For stock offers, the bidding firm exchanges its own stocks for the target’s stock at the agreed exchange ratio. This indicates that the initial offer price, which in the cash offer is equal to the amount of cash paid per share, now equals to the acquirer’s stock price multiplied by the exchange ratio. Furthermore, in stock offers, arbitrageurs do not only buy the target’s stock, but also short sell the stock of the bidding firm (Barclays Hedge).

Equally, when the payment contains both cash and stock, (e.g., 100 dollars cash and 5 shares of the bidding firm for every share of the target) the arbitrage implies buying one share of the target firm and short selling 5 shares of the bidding firm (Jetley and Ji, 2010).

Prior literature calculates the spread in slightly different ways. Some papers (e.g., Officer 2007, Branch and Wang 2009) state it as a percentage: Initial offer price minus target’s stock price, divided by the offer price. Some literature (e.g., Mitchell and Pulvino 2001, Jetley and Ji 2010) calculates the spread slightly different: Initial offer price minus target’s stock price, divided by target’s stock price. The percentage outcome is the exact percentage the

13

target’s share price has to increase, to equal the initial offer price. Other literature, (e.g., Jetley and Ji 2004) takes the integer of initial offer price minus target’s share price. In the remainder of this paper, we follow the method of Mitchell and Pulvino (2001) and Jetley and Ji (2010). The quantification by Officer (2007) and Branch and Wang (2009) has been used as a robustness check in section 6.

2.2.3. Determinants of risk arbitrage spread

Previous research investigates what factors affect risk arbitrage spreads. The following paragraphs describe what determinants these are, based on economic reasoning. This might explain any difference in spreads in emerging and developed markets. It can appear that certain variables affect spreads in developed market significantly, whereas they do not have any effect on spreads in emerging markets. This would indicate that spreads in emerging markets are determined by different factors than the commonly used factors in developed markets.

Brown and Raymond (1986) were the first to completely focus on investigating risk arbitrage spread and its predictive value for takeover success. However, they only investigated the observable value of the spread, without doing any deeper analysis on the spread. They did conclude, just as Samuelson and Rosenthal (1986), that the target’s stock price reaction to offer announcements is a useful indicator for takeover success prediction. Cornelli and Li (2002) find that risk arbitrageurs can act as large shareholders after the offer announcement and so highly influence the post-announcement target’s stock price.

Jindra and Walkling (2004) analyze the risk arbitrage spread and its effect on revision returns and offer duration. First they emphasized on the importance of understanding predictors of the spread, similar to the structure of this thesis, by doing a sub-analysis on the spread. By doing an OLS regression, they find that bid premium, friendly offers, the presence of rumors and target’s stock prices trading below $5 have a significant positive effect on the spread. Furthermore, they find that high takeover activity, target’s pre-offer stock price run up and the level of toehold (percentage of the target’s shares that the bidder has prior to the

announcement), have a significant negative effect on the spread. Their main conclusion was that the risk arbitrage spread is significantly negatively related to price revision and

significantly positively related to offer duration. Moreover, they find that price revision and offer duration are highly related to successful takeovers, which makes it interesting for future research to measure the direct effect of the spread on takeover success prediction.

Hsieh and Walkling (2005) examine arbitrage holdings in acquisitions. They find that the change in arbitrage holdings is positively related to takeover success. Furthermore, they

14

state that risk arbitrageurs significantly affect bid premium, arbitrage returns and the

probability of takeover success, in line with Cornelli and Li (2002) and Gomes (2001). They mention risk arbitrage as a major predictor of takeover success and conclude that risk

arbitrageurs play a critical role in takeover markets. This conclusion shows the direct link between risk arbitrage and takeover outcome.

Officer (2007) examines the performance based arbitrage hypothesis, introduced by Shleifer and Vishny (1997). He investigates comovement in arbitrage spreads and its response to large arbitrage losses. Officer (2007) first runs a regression on the spread, to investigate which variables significantly affect the spread. This is similar to the structure of Jindra and Walkling (2004) and one of the six regressions done in this thesis.

Although the structure of this analysis on the spread is used in this thesis, Officer’s (2007) research purpose clearly differs from this thesis. This thesis is interested in the question whether performing risk arbitrage in emerging markets is attractive, which requires additional tests on takeover success prediction as well. Officer’s (2007) main purpose is to examine the performance based arbitrage hypothesis, which differs evidently. However, Officer (2007) shows the importance of understanding the spread and its determinants, which is important in this thesis as well. He finds that duration of the deal, bid premium and cash offers significantly positively affect the spread. Takeover completion, price revision, post-bid competition, both firms operating in the same industry, level of toehold, target’s market capitalization and hostile offers significantly negatively affect the spread. His findings on the effect of cash offers and hostile offers are contrary to the findings of Jetley and Ji (2010).

Jetley and Ji (2010) investigate the average decline in risk arbitrage spreads over the last 15 years. They state that this decline can be linked to the increased amount of assets under management of risk arbitrage hedge funds, in addition to the decline in total returns of these hedge funds. Furthermore, decreasing transaction costs and changes in risk may explain some of the decline in the spread. Besides, they find that changes in deal characteristics over the last 15 years may explain the decline. Higher bid premiums, hostile offers and deals with collar agreement significantly positively affect the spread. On the other hand, increased trading in the target’s stock after takeover announcement, cash offers and the target’s market capitalization significantly negatively affect the spread. They conclude that their findings suggest that some of the average decline in the spread is probably permanent. This study is interesting for this thesis, since they examine why spreads in general decline. The findings can be linked to emerging markets and help to the prediction whether risk arbitrage in emerging markets is attractive.

15

Andries and Virlan (2017) have written the most recent paper about this topic. They measured whether arbitrage spreads reflect the level of risk in risk arbitrage strategies with Polish targets. They mention it would be an exciting debate whether deals in emerging markets differ from developed markets, analyzing their effect on risk arbitrage spread and takeover success prediction.

Overall, deal spreads reflect the risk over takeover completion in multiple ways. For successful deals, the spread narrows towards the resolution date. For failed deals, the spread increased towards the termination date. The spread reacts to all different risk factors that threaten the chance of takeover success. Furthermore, when measuring spreads, takeover success probability has to be measured as well. High spreads with high takeover success probability are mostly profitable for risk arbitrageurs, whereas low spreads and low takeover success probability relationships should be avoided for these arbitrageurs. The spread seems to be directly linked to takeover success, which will be explained in the following section.

2.3 Determinants of takeover success prediction

Much research has been done on takeover success prediction. However, prior research lacks investigating deal success prediction in emerging markets. Takeover success probability reflects the risk in takeovers for risk arbitrageurs, since deal success results in profits and deal failure leads to losses. According to Hazen (1987), risk arbitrageurs lost their ability to

determine the success of takeovers, since short tendering and hedged tendering are prohibited by the SEC. Therefore, natural market powers determine takeover outcome. Hence,

understanding what variables affect this probability of success could lead to further knowledge and research on the prediction of takeover success in emerging markets. This section discusses what variables affect takeover success probability, based on economic reasoning and prior research.

Hoffmeister and Dyl (1981) investigate how firms undertaking a tender cash offer can measure the probability of takeover success. Their first model intended to test what variables affected takeover success. They find that the target’s size and target resistance strongly

negatively affect the outcome. On the other hand, Walkling (1985) develops a model to test the prediction of tender offer outcomes, applying both linear and logistic regressions. He finds that prior research utilizes models that contain specification errors, which lead to some insignificant results. When controlling for this misspecification, he finds equal results for the coefficient estimates and predictive outcomes in both linear and logistic regressions. He finds that bid premium, solicitation fees and level of toehold significantly positively affect takeover success

16

prediction. Target resistance significantly negatively affects offer success. He suggests that further research should analyze this managerial power more extensively.

As mentioned in 2.2.3, Brown and Raymond (1986) were the first to focus on risk arbitrage spread and its predictive value for takeover success. First they find that successful takeovers imply significantly different estimated probabilities compared to failed takeovers. This means that the research on takeover prediction gives reliable results. Also, they find clear results that risk arbitrage differentiates between successful and failed takeovers, indicating that it can be used as a predictor for the success rate of takeovers. Their suggestion for future research is to analyze the post announcement target stock price more, since this price contains predictive power for the success rate of takeovers. This is in line with Samuelson and

Rosenthal (1986).

Schwert (2000) examines hostility in takeovers. He finds that most hostile offers reflect strategic choices made by the acquirer to maximize their wealth after a potential deal.

Therefore, he suggests that further research should investigate the effect of hostile offers on takeover success prediction. He predicts hostility to be associated with decreasing probability of takeover success, since it meets resistance from the target.

Hutson and Kearney (2005) analyze the interaction between target and acquirer stocks during takeover bids. They find that target stocks significantly move differently after a takeover announcement, which imply merger arbitrage opportunities. Furthermore, they find interaction between target and bidding firms. They state that this can be explained since both stock price movements reflect the probability of takeover success.

Branch, Wang and Yang (2008) compare two different models for takeover success prediction: the traditional logistic regression model and the artificial neural network

technology. They find that both models are sufficient to predict consistent outcomes. This outcome is important for the remainder of this thesis, since it confirms the utilization of the logistic regression model. Furthermore, they find four dominating variables that affect the outcome of a takeover bid: arbitrage spread (following Samuelson and Rosenthal 1986, Brown and Raymond 1986), target resistance, deal structure and transaction size. Arbitrage spread, stock offer and transaction size significantly negatively affect the probability of success, whereas target resistance is significantly positively associated with the chance of deal success. This is in line with prior literature (e.g. Branch and Yang 2008, Schwert 2000, Kallio 2013).

17

2.4. Emerging and Developed markets

This section discusses previous literature on differences between emerging and developed markets. Understanding these differences contributes to understanding the results of the empirical part of this study. Multiple studies have shown the increasing presence of emerging markets in international investment (Deng 2012, Wang et al. 2012, Deng and Yang 2015). However, it has not been studied whether deals in emerging markets fail more often than deals in developed markets and whether this is compensated for arbitrageurs with higher spreads. This section sheds a light on the differences between the markets and tries to explain how these differences might affect spreads and takeover success probabilities.

Umber (2010) states that takeovers in foreign markets include information asymmetries, more agency conflicts, higher additional costs (language and currency differences) and higher transaction costs. This is interesting, since Jetley and Ji (2010) state that the arbitrage spread reflects transaction costs. Direct transaction costs are linked to the trading costs of risk arbitrage, whereas indirect costs are associated with the price impact of trades.

Burns and Liebenberg (2011) investigate whether takeovers affect emerging markets differently compared to developed markets. They do this by analyzing the response of the target’s rival firms. They state that country, industry and cultural factors matter more in emerging markets and that individual firm characteristics are more important for developed markets.

Erel, Rose and Weisbach (2012) emphasize that due to increasing integration of all economic markets across the world, more takeovers in emerging markets occur. By doing a multivariate regression, they find that geography does affect takeover characteristics. Their research shows that cross-border takeovers occur more often in a nearby country than far away. They also find that the level of economic development and accounting quality is associated with a higher chance of being an acquirer rather than a target. Furthermore, country-specific characteristics, like currency appreciation and macroeconomic functioning have a direct effect on how acquiring firms value the takeover.

Ahern, Daminelli and Fracassi (2015) investigate the effect of cultural values on takeovers. Their findings indicate that the volume of successful takeovers decreases when cultural differences increase. Especially the fundamentals of national culture (individualism, hierarchy and trust) play a major role in the level of merger volume and synergy gains, according to their research. They also state that cultural borders can increase costly frictions, which decrease the likelihood of deal success.

18

Fletcher (2004) links the cultural importance in takeovers mentioned by Ahern et al. (2015) to the difference between emerging and developed markets, which make both papers interesting for this thesis. He mentions that acquiring firms should focus on cultural identity instead of geographical boundaries, when they seek takeovers in new markets. Furthermore, he states that cultural characteristics in emerging markets differ often from developed markets in terms of familiarity, time, space, consumption patterns, communication and negotiation.

Hence, takeovers in emerging markets differ notably from developed markets. As section 2.2.3 concluded, the risk arbitrage spread and deal outcome probability reflect the risk-reward relationship in risk arbitrage. Risk arbitrageurs should seek for low risk and high risk-reward markets, which would benefit them the most in terms of profit. Due to several factors in

emerging markets such as economic development, trade and supply-chain barriers and cultural identity, takeovers in emerging markets seem to include more risks that could lead to deal failure more often. This leads to the following hypothesis:

H1-A: Takeovers with targets in emerging markets are associated with higher probability of deal failure than deals in developed markets.

According to the basic theory of risk arbitrage, arbitrage spreads reflect the risk of deal completion. This means that high spreads are associated with less confidence in takeover success, whereas low spreads are related to high confidence in takeover success. Hence, the higher expected probability of deal failure in emerging markets should be reflected in higher spreads as well, to compensate for the higher risk. Therefore, the following hypotheses is generated:

H1-B: Takeovers with targets in emerging markets have higher arbitrage spreads than in developed markets.

Also, these studies show that takeover outcome and corresponding risk levels in

emerging markets are influenced by different factors than for developed markets. Factors such as cultural aspects, economic development and are found to play an important role in takeovers in emerging markets. This could indicate that the traditional, commonly used variables used to predict spreads and takeover outcome are not reliable predictors for takeovers in emerging markets. Therefore, the following and final hypothesis is generated:

H2: The traditional predictors for arbitrage spread and takeover outcome have less predictive value for emerging markets

As a sub-analysis, this thesis examines the effect of M&A waves on spreads and takeover success probability. Following Kastrinaki and Stoneman (2012), it is predicted that macro economic factors affect takeover outcome and so arbitrage spreads. Positive M&A

19

waves are expected to be associated with higher confidence in takeovers and higher success rates. Hence, hypothesis H3 is generated:

H3: Takeovers that occur during positive M&A waves, have lower spreads and lower takeover failure probability, compared to negative waves

In the following section the methodology is described, which operationalizes hypotheses H1-H3.

3. Methodology

This section describes the methodology and discusses the data used in this thesis. First, in section 3.1., the variables tested are discussed. Subsequently, in section 3.2., the econometric models that we test will be described. Section 3.3., gives an explanation why the research design of this paper allows for testing causal effects. Then, in section 3.4., the relation of the significance and signs of the final outcomes to the hypotheses that are tested are

clarified. By following this methodology, together with following our sample that is described in section 4, one can replicate the results of this paper.

3.1. Variables

The following section will describe and define all variables tested in our models. First, in section 3.1.1, the dependent variables are defined. Then, in section 3.1.2, the independent variables are discussed. Additionally, the expected outcome of the estimates is presented, based on economic reasoning.

3.1.1. Dependent variables: Risk arbitrage spread and takeover failure probability

Since the empirical part is split up in two parts, we have two dependent variables. The first one that is examined is:

(1) spread: the percentage difference that the target’s immediate post-announcement share price trades below the initial offer price, also known as the risk or merger arbitrage spread.

For part one the effect on the arbitrage spread is investigated. As mentioned in section 2, the arbitrage spread differs slightly for cash offers in comparison with stock offers. For cash deals, where the bidding firm only offers cash to the target) we define the arbitrage spread following Mitchell and Pulvino (2001) and Jetley and Ji (2010):

!"#$%&_!"#!,! = !!""#$ − !!"#$%!,!!! !_!"#$%!,!!!

20

Where:

!"#$%&_!"#!,! = The arbitrage spread on trading day t, for cash offers !_!""#$ = The initial offer price in cash, per share of the target firm,

!_!"#$%!,!!!= The closing target′s stock price, 1 day after the offer announcement

For stock offers (where the bidding firm exchanges its own stocks for the target’s stock at the agreed exchange ratio), the arbitrage spread is defined as:

!"#$%&!"#$%,! =

(!!"##$%,!∗ !") − !!"#$%!,!!! !_!"#$%!,!!!

Where:

!"#$%&_!"#$%,! = The arbitrage spread on trading day t, for stock offers

!_!"##$%,! = The closing acquirer’s stock price at the day of announcement

!" = The exchange ratio. This is the number of the acquirer’s shares offered for one share of the target company.

!!"#$%!,!!! = The closing target’s stock price, 1 day after the offer announcement.

In the second part of the empirical research, the effect on the following variable is investigated: (2) Failed: Dummy variable indicating takeover failure.

It takes the value of a dummy variable; 1 if the deal is withdrawn or rejected, 0 if accepted or consummated. Thomson One states whether the deals are completed or withdrawn. By

formulating a logistic regression, which is further explained in section 3.2, it is investigate how the independent variables affect the probability of takeover success.

3.1.2. Independent variables

Based on prior literature discussed in section 2, the following independent variables are

identified to affect risk arbitrage spread. First, the explanatory variables and its expected effects are discussed. Then, the control variables are presented. Primarily based on Officer (2007), the following variables can be considered as explanatory variables. It includes some variables that are rather unknown at the moment of measurement of the spread. However, Officer (2007) states he takes the simple approach of proxying for these ex-ante expectations with ex-post outcomes, identical to Jindra and Walkling (2004):

21

(1) Logduration: The logarithm of the duration of a takeover process

The variable Logduration measures the number of days between announcement and resolution of the takeover. For completed deals, duration is calculated by counting the days between ‘date announced’ and ‘date effective’. For failed deals, duration is calculated by counting the days between days announced and date withdrawn. Officer (2007) and Jindra and Walkling (2004) find a significant positive effect of duration on spread. Jindra and Walkling (2004) state that a higher duration implies higher holding costs, which has to be compensated with a higher return and so spreads are expected to increase with the duration of the offer. Hence, I expect duration to be positively associated with the spread.

(2) Status: Completion or failure of the deal

The variable status is a dummy variable, which has a value of 1 if the deals is completed and 0 if otherwise. Officer (2007) finds a negative significant effect. This is in line with theory, since the spread reflects the risk of completion of the deal. A completed deal normally implies lower risk compared to withdrawn deals and so should decrease the spread. However, the inclusion of this variable is rather debatable, since the final outcome is known after the time the spread is measured. Yet, status will be included if it improves the model, just as Officer (2007) and Jindra and Walkling (2004) has done. The opposite of this variable, a dummy variable for failed deals, is also generated. This dummy is called failed and will be used in regression models (4) – (6). Hence, I expect final deal completion to be negatively associated with the spread.

(3) Competition: Whether there are multiple bidders involved

Competition measures if multiple firms are involved on the bidding side during a takeover. It

takes a value of 1 if the number of bidders is more than 1, and equals 0 if there is only 1 bidder. In accordance with variable status it is debatable, since at the moment of the first bid it is hard to predict whether multiple bidding firms will be involved. However, if it improves our model, it will be included. Officer (2007) finds a negative relationship between competition and spread. This is in line with theory, since post-bid competition is expected after the initial offer, the final price of the target’s share will be higher at deal completion than the initial offer. This means the target’s stock price will increase and so the spread should narrow, since spread is based on the initial offer price instead of a revised offer. Hence, I expect competition to

negatively affect the spread. Also, according to Walkling (1985), the presence of multiple bids increases the possibility that revision will occur. Therefore, competition is expected to decrease takeover success probability.

22

(4) Bidpremium: bid premium: Percentage difference between initial offer price and pre-announcement target’s share price.

Bidpremium is the percentage bid premium. Broadly following Officer (2007) and Jennings and

Mazzeo (1993), bid premium is calculated by dividing the initial offer price per share by the target’s share price 4 weeks before the offer announcement. It is winsorized with cuts 1 and 99, to control for outliers. Jennings and Mazzeo (1993) state that larger bid premiums will

discourage competing firms from making an offer and decreases the probability of another bid, so larger spreads are expected. He finds significant results that are in line with this theory. In accordance, Officer (2007) finds significant positive effects as well. Hence, I expect

bidpremium to positively affect the spread. Although it is expected that bidpremium positively

affects the spread, high bid premiums imply more attractiveness and normally reduce target resistance, which could increase takeover success probability. Also, following Walkling

(1985), higher bid premiums result in an increased amount of shares being tendered, increasing the probability of deal success. Therefore, it is expected that bidpremium positively affect takeover success probability.

(5): Pricerevision: change in offer price between announcement and resolution.

Price revision is calculated as the percentage change between the initial offer price and the final offer before resolution. Officer (2007) finds that spreads narrow if price revision is expected, since a higher second bid mostly causes the target’s stock price to rise further towards the initial offer price. Jindra and Walkling (2004) agree. Hence, I expect pricerevision to negatively affect the spread.

(6) EM: Dummy variable indicating EMTFs

This variable is added to the model to investigate the effect of EMTFs on the spread. It has a value of 1 if the target firm is located in an emerging market and 0 if otherwise. Following prior literature on emerging markets discussed in section 2.4., it can be expected that deals with EMTFs involved increase the spread. Several reasons can be found in section 2.4., but a

comprehensive argument is that deals with EMTFs have higher risk and so the spread should increase. Hence, I expect a positive effect of EMTFs on the spread.

This variable is also used in one of the regressions measuring the effect on takeover success probability. It can be expected that due to the increased amount of risk in EMTFs, less

takeovers get consummated. Hence, I expect a negative effect of EMTFs on takeover success. (7) Sought: Percentage of target’s shares sought by the bidding firm

Sought can be defined as the percentage of outstanding equity sought by the bidding firm.

23

takeover success. The idea behind it is that when the bidding firm seeks a very high amount of equity, current large shareholders do not allow for this to happen. Therefore, I expect sought to negatively affect the takeover success probability.

(8) Friendly: Dummy variable indicating friendly deals

The friendly variable contains a value of 1 if the bid is considered friendly and has a value of 0 if otherwise. It can be seen as the opposite of hostile, so expectations can be related to this as well. Branch, Wang and Yang (2008) find that friendly takeovers are more likely to succeed than hostile offers. The reason for this is that if target firms resist, they may use defense mechanisms that more often lead to a failed deal. A friendly offer means less resistance that contributes to the chance of failure. This is in line with previous findings (Branch, Wang and Yang 2008, Kallio 2013). Hence, I expect friendly offers to positively affect the takeover success probability.

(9) Spread: The percentage difference between the 1-day post announcement target’s share price and the initial offer price.

Spread is the most interesting variable for this thesis. Following prior literature (e.g.,

Samuelson and Rosenthal 1986, Brown and Raymond 1986, Branch, Wang and Yang 2008), the spread can be seen as the compensation for risk arbitrageurs. This means the higher the spread, the higher the risk involved, which means the probability of deal success declines. Therefore, I expect the spread to negatively affect takeover success probability

(10) Runup: This variable measures the target’s stock price increase in the period before the deal announcement.

It is defined by the percentage difference between the target’s stock price 4 weeks prior to the announcement and its price 1 day prior to the announcement. According to prior literature (e.g., Branch and Wang 2009, Jindra and Walkling 2004), increasing stock price prior to deal

announcements can indicate speculative activity, which could imply increasing probability of deal success. Hence, I expect runup to positively affect takeover success probability.

3.1.3. Control variables

The remaining variables are used as control variables in the first three models that measure effects on the spread. These variables control for the experience of bidding firms, whether both acquiring and target firm operate in the same industry, whether the bidding forms holds a toehold, whether the deal is hostile, for the size of the target company and for the method of payment:

24

Experience is a dummy variable that has a value of 1 if the acquiring firm has done any other acquisition in the previous years. This indicates the bidding firm has experience with doing acquisitions that could improve the chance of success. Cuypers, Cuypers and Martin (2016) find that experience advantage plays an important role in acquisitions and contributes to the reliability of bidding firms, which means the spread should decline and takeover success probability should increase.Hence, I expect experience to negatively affect the spread.

(12) Sameindustry: Dummy variable for both firms operating in same industry. Potential synergy gains cause many firms to acquire other firms in the same industry. Officer (2007) finds that takeovers with both firms operating in the same industry decrease the spread. Flanagan, D’Mello and O’Shaugnessy (1998) find that similar industry classifications increase the probability of takeover success, which is in line with the expectation of a lower expected spread. Hence, I expect sameindustry to negatively affect the spread.

(13) Toehold: percentage of the target’s shares that the bidder has prior to the announcement.

Toeholds increase the bargaining power of bidding firms, since they increase their influence on the managerial team (Jindra and Walkling, 2004). Furthermore, toeholds are valuable if another bidder offers a higher bid. Overall, the bidding firm benefits from having a toehold, which by theory should increase the probability takeover success but should not necessarily decrease spreads. For the same reason that is expected that competition increases spread, it can be expected that high levels of toehold increase spread because the possibility of price revision declines with high levels of toehold. Hence, following this theory I expect toehold to have a positive effect on the spread.

(14) Hostile: Dummy variable for hostile or unsolicited deals

The variable hostile has a value of 1 if there is a hostile deal and has a value of 0 if otherwise. It can be seen as the opposite of the variable friendly. Prior research consistently find that hostile bids lead to a lower takeover success probability. As mentioned before, due to the fact that target managements have numerous tricks and defense mechanisms to prevent being taken over, like poison pills, staggered board and seeking a white knight. Therefore, I expect hostile to positively affect the spread. On the other hand, Jennings and Mazzeo (1993) state that hostile offers are often associated with multiple bidders involved, leading to bid revision and so

smaller spreads could be seen as well.

(15) LogtargetMVE The logarithm of the target’s market capitalization.

The target’s market capitalization is calculated following prior literature as the target’s stock price four weeks prior to the announcement, multiplied by the shares outstanding prior to the

25

announcement. Larger companies often have larger existing shareholder value, which make it harder to acquire these companies. Officer (2007) finds negative effect of logtargetMVE on spread, but many papers (e.g., Hoffmeister and Dyl 1981, Mitchell and Pulvino 2001, Branch and Yang 2003) find that increasing logtargetMVE strongly decreases the takeover success prediction. Based on theory, bigger takeovers are harder to acquire and do not affect any other variables discussed in this section, so I expect logtargetMVE to positively affect the spread and negatively affect takeover success probability.

(16) Cash: Dummy variable for cash offers

Cash is a dummy variable that has a value of 1 if deals are financed with cash only. Previous

literature finds contrary effects of cash deals on the spread. Jindra and Walkling (2004) and Officer (2007) find a positive effect, whereas Jetley and Ji (2010) find cash deals to negatively affect the spread. The main theory behind it is that cash financed takeovers are less complicated and have more certainty than stock offers and is associated with less risk during a takeover. Therefore, I expect cash to negatively affect the spread and positively affect takeover success probability.

(17) Waves: Dummy variable indicating takeover date is in positive wave or not

This dummy variable controls for time, in the way that its value is 1 if the takeover takes place in a positive wave and 0 if it occurs in a negative wave. Following figure 1, the years that belong to the positive waves are: 1993-1999 and 2003-2007. I expect that positive waves, which imply a better economic state, are associated with higher takeover success probability and lower arbitrage spreads

3.2. Econometrical models

As explained in the previous section, the methodology consists out of two parts. This also means that two econometrical models are used, which both perform three regressions to statistically test our hypotheses. The setup of the first model follows Officer (2007) that tests the effect of several variables on the arbitrage spread. Testing additional variables can be done unlimitedly, if it improves the model and does not cause multicollinearity. The model that is used is described in equation (1):

1 !"#$%&!"!#$= !!+ !!∗ !"# !"#$%&'( + !!∗ !"#"$% + !!∗ !"#$% !"#$%$&' + !!

∗ !"#$%&' !"#$%&'&'"( + !!∗ !"#$%&'"(' + !!∗ !"#ℎ + !!∗ !"#$ !"#$%&'( + !! ∗ !"#$%&' + !_!∗ !"#ℎ!"# + !_!"∗ !"#$%&$'($ + !_!!∗ !"# !"#$%&'()_!!! + !_!"∗ !" + !

26

First, we perform this test to see how all variables affect the spread. Next, we split up the sample in only EMTFs and only DMTFs, to check how the predictor variables respond to this. Two more regressions are performed on these subsamples, in case the full sample is not representative enough. The data that is used can be defined as panel data, since we have

multiple observations for several variables for multiple moments. Therefore, the regressions are adjusted for this, by adding fixed effects. Fixed effects are chosen instead of random effects, after performing a Hausman test that indicates fixed effects are more appropriate for this dataset. Assuming cultural and other country-specific effects of countries influence all

takeovers with EMTFs and DMTFs, countries are clustered. Hence, we perform two tests with country fixed effects, for DMTFs separately from EMTFs. This shows clearly how the

predictor variables affect the spread with EMTFs involved, compared to the situation where only DMTFs are involved. The two equations are the following:

2 !"#$%&!"#$%= !!+ !!∗ !"# !"#$%&'( + !!∗ !"#"$% + !!∗ !"#$% !"#$%$&' + !!

∗ !"#$%&' !"#$%&'&'"( + !!∗ !"#$%&'"(' + !!∗ !"#ℎ + !!∗ !"#$ !"#$%&'( + !! ∗ !"#$%&' + !_!∗ !"#ℎ!"# + !_!"∗ !"#$%&$'($ + !_!!∗ !"# !"#$%&'()_!!! + !_!"∗ !"#$% + !!+ !

3 !"#$%&_!"#$%= !_!+ !_!∗ !"# !"#$%&'( + !_!∗ !"#"$% + !_!∗ !"#$% !"#$%$&' + !_!

∗ !"#$%&' !"#$%&'&'"( + !_!∗ !"#$%&'"(' + !_!∗ !"#ℎ + !_!∗ !"#$ !"#$%&'( + !_! ∗ !"#$%&' + !!∗ !"#ℎ!"# + !!"∗ !"#$%&$'($ + !!!∗ !"# !"#$%&'()!!! + !!"∗ !"#$% + !_!+ !

Where:

!_! = Country fixed effects ! = Error term

The second econometric model in this thesis is based on the model of Branch and Wang (2009). It is a logistic regression model that investigates the effect of several variables on the takeover failure probability. This results in the following equation:

4 !"#$%$&'&() !" !"#$ !"#$%&'_!"!#$

= 1

1 + !

(!(!!!!!∗!"#"$!!!∗!"#!"#$%!&'(!!!!!!∗!"#$%&'(!!!∗!"#$%&!!!∗!"#$!!!!!∗!"#!!!!∗!"#$%&'"('! !!∗!"!!!∗!"#$%!!!"∗!"#$%&'&'"(!!!

27

Following equations (2) and (3), we adjust equation (4) also to only EMTFs and DMTs. This makes another two equations, which are identical to (4) but only the sample has changed: 5 !"#$%$&'&() !" !"#$ !"#$%&'_!"#$% =

= 1

1 + !

(!(!!!!!∗!"#"$!!!∗!"#!"#$%!&'(!!!!!!∗!"#$%&'(!!!∗!"#$%&!!!∗!"#$!!!!!∗!"#!!!!∗!"#$%&'"(' !!!∗!"#$%&'&'"(!!!!!!

6 !"#$%$&'&() !" !"#$ !"#$%&'_!"#$%

= 1

1 + !

(!(!!!!!∗!"#"$!!!∗!"#!"#$%!&'(!!!!!!∗!"#$%&'(!!!∗!"#$%&!!!∗!"#$!!!!!∗!"#!!!!∗!"#$%&'"(' !!!∗!"#$%&'&'"(!!!!!!

Where:

!_! = Time fixed-effects

Using the Hosmer and Lemeshow (HL) test, a goodness of fit test is done for these prediction models. This is explained in section 4.3.

3.3. Addressing Endogeneity

As the main research purpose of this thesis is to test the causal effect of spread and EMTFs on takeover success probability, additional tests are performed to add to the causal interpretation of the results. Several robustness checks have been done which are explained in section 6. The results are tested for different time intervals, subsamples and methods of calculation.

These checks are done to address for possible endogeneity problems. When endogeneity is present, the results can be stated as biased and inconsistent. Endogeneity issues can be

caused by two problems: omitted variable bias and reverse causality. Omitted variable bias means that there is correlation between the independent variable and the error term in the model. Hence, factors outside the model affect the independent variable. Reverse causality means that changes in the dependent variable result in changes in the independent variable, instead of the opposite way.

These endogeneity problems cause biased and inconsistent estimates. There are two types of solutions for these problems: Control for omitted variable bias by an IV regression and performing quasi-natural experiments. However, it is often hard to recognize which variables are omitted and experiments often require small samples. Furthermore, IV estimates are often considered weak and are in that case inconsistent.

28

An alternative solution is adding random/fixed effects, which offers a solution to the endogeneity problem without requiring an IV regression. A Hausman test is done in section 4.3 to test whether random or fixed effects are most suitable. The test shows that fixed effects fit best in the models used in the methodology. The central motivation for using a fixed effects model is to correct the omitted variable bias (Chi, 2005). He states that although fixed effects will not completely capture changes in unobservable heterogeneity, fixed effects significantly alleviates the endogeneity problem caused by omitted variables. Fixed effects are due to omitted variables that are specific to cross-sectional units or to time periods (Egger, 2000).

In this thesis, country and time-fixed effects are used since it is assumed that country-specific factors affect the time-invariant variables. By adding these fixed effects, the omitted variable bias is eliminated. Reverse causality does not appear to be a problem in the models that are used.

Egger (2000) finds that for certain samples a fixed country and time effects model can be the most suitable. He motivates that samples that are affected by geographical, historical or political contexts, should use country and time fixed effects model. Hence, this method is used in this thesis as well, to control for the endogeneity problem. Furthermore, prior literature on risk arbitrage spreads and takeover success prediction use fixed effects models as well to test causal effects, instead of IV regressions or quasi-natural experiments. In this way, the

methodology adds to the causal interpretation of the results and helps to address the endogeneity problem.

However, fixed effects estimates also have its limitations. For example, it is more sensitive to measurement error than IV estimates. IV estimates control for measurement error (Hersch and Stratton, 1997). Furthermore, if variables do not vary much within the chosen group, the effect of these variables cannot be assessed. In this case, the estimates are inconsistent. Nevertheless, these drawbacks of the fixed-effects model are accepted in this thesis. For this research, the fixed-effects model appears to give the most reliable results. This is based on prior research on risk arbitrage and takeover success prediction, in addition to literature about endogeneity and solutions to endogeneity problems.

3.4. Relation of significance and signs to hypotheses

The interpretation of our results is important to understand, since this reveals how we can relate the outcomes to our hypotheses that are tested. For all equations we use the corresponding p-values of the independent variables to analyze its significance. P-p-values smaller than 0.10 are considered as reasonable significant, whereas p-values smaller than 0.05 or 0.01 are considered

29

as significant and highly significant. For both models, the fixed-effect model and the logistic regression model, positive and negative signs relate to the definition of the independent variable.

For equations 1-3 this means that if the dummy variables equal 1, the spread increases with the coefficient corresponding to the dummy variable. For the continuous variables, it implies that for a one-unit increase in the independent variable, the spread increases with the corresponding coefficient value. The only exception is !"# !"#$%&'()!!! . Since this is a logarithm, the outcome can be interpreted as follows: an increase in the target market

capitalization of one percent is expected to result in an increase in spread by (!!!/100) units of the spread.

This also means that if spread has a positive and significant estimate in equation (4), this implies that a higher spread is associated with a significant higher probability of deal failure. The values for the coefficients can be interpreted in different ways. The odds of deal failure equal the exponential value of the coefficient. This means that a coefficient value of y for bidpremium can be interpreted as an increase in deal failure probability of !!%, for a 1% increase in bid premium. An extra column will be provided in the output table of the logistic regression models, which shows the odds ratios instead of only the coefficient estimates.

4. Data and descriptive statistics

The data used for this research is mostly obtained from database Thomson One, formerly known as SDC Platinum. Additional information to calculate the target’s market value of equity is retrieved from database CRSP.

To distinguish between emerging markets and developed markets, I use official criteria from the International Monetary Fund (IMF) and FTSE Group. The IMF assigns 23 countries the status of ‘emerging market’ and FTSE Group assigns 24 countries as ‘developed market’. The list with classified countries for both emerging and developed markets can be found in Appendix I, table 3. The preliminary sample consists out of 17,771 takeover announcements. The criteria are written in table 1:

30

Table 1: Preliminary sample construction

Request Operator Description Hits

Database Include All M&A's n/a

Target Nation Include

All countries classified as emerging market or

developed market 1,076,555 Tender Offer Flag

(Y/N) Equals Yes 24,389

Date Announced Between 01/01/1993 and 01/01/2018 19,526

Target Public Status Is Public 17,771

Furthermore, the following variables are obtained from Thomson One to generate all the variables that are used in this research: Acquiror name, Acquiror Nation, Target Name, Target

Nation, target CUSIP, Date Announced, Date Effective, Date Withdrawn, Status, Initial offer price, Acquiror’s price per share prior to deal announcement, Exchange ratio, Target closing price 4 weeks prior to announcement date, Target closing price 1 day prior to announcement date, Target closing price 1 day after announcement date, percentage change offer price, number of bidders, Consideration structure, Acquiror’s industry, Target’s industry, percentage owned after transaction, percentage acquired, deal attitude, Target Ticker Symbol, percentage shares sought, cross-border indicator.

Furthermore, by using STATA, missing values are excluded in addition to observations that do not match in my research. Deals where the acquirer had more than 50% of the target’s shares before the announcement or less than 50% after the announcement. The risk arbitrage spread, duration, bid premium and emerging market dummy were created manually through STATA with the obtained data. For cash offers, the spread can be calculated by: initial offer price minus target’s stock price, divided by target’s stock price. For stock offers, the spread can be calculated by: (acquirer’s stock price multiplied by the exchange ratio), or the initial offer value, minus the target’s stock price, all divided by the target’s stock price. After dropping all missing or uninteresting values, we end up with a sample of 5,142 observations, as shown in table 2:

Table 2: Final sample construction Activity Number of observations Preliminary sample 17,771

Excluding missing values independent variables 10,080

Excluding deals for which toehold > 50% 6,898