Mergers and Acquisitions: The relationship between M&A activity
and irrelevant peak prices
University of Amsterdam Msc. Business Economics, Finance
Master Thesis Supervisor: Florian Peters
June 2014
Robert van Huizen 5732530
1
Abstract
This paper falls into the rapidly grown field of behavioural finance and found significant evidence that irrelevant peak prices are related to the level of merger activity and the probability of a merger. Market gaps are negative predictors for the total level of quarterly merger activity. Individual gaps are negative related with the probability of a merger in a specific quarter. The negative relationship between individual gaps and the probability of a merger is very differential for cash and stock acquisitions. The negative relation between the market gap and the level of merger activity is tested on two recessions. Furthermore, this paper tried to find a new anchor in form of the market peak prices.
2
1. Introduction………..………...3
2. Related Literature………..………..………...5
1. Related Literature………..…….5
2. Merger Activities and Waves…………..………...9
3. Other biases that influence merger activity……….….…10
3. Hypotheses………..……12
4. Data and Descriptive Statistics ………..………..…..16
1. Data Collection………..……….16 2. Descriptive Statistics ………..………..…...19 5. Methodology………..…..20 6. Results.……….………..…22 7. Robustness Checks.…………..………..……….28 8. Conclusion………...30 References……….32 Appendix……….35
3
1. Introduction
Mergers and Acquisitions (hereafter, M&A) are the largest transactions that take place in a modern capitalist economy. It is unsurprising that M&A activity is among the most frequently studied phenomena in financial research. In a review, from the online Journal of McKinsey & Comp.,
Goedhart et al. (2006) show that firms have improved their performance in M&A over the past decades. The review remarks that, based on percentages of shareholder value creation, a ten-year peak is reached. However, a large proportion of the literature suggests that there is plenty of room for improvement of shareholders value creation. For example, a significant percentage of acquirers are overpaying for the target and three out of four M&As fails to achieve their financial and strategic goals (Marks et al., 2001). A standard textbook story emphasizes synergies and economics of scale as the most important value creators in a merger or acquisition. The synergy advantages are also described as the “2+2=5” effect (Cartwright et al., 1993). The offer price starts with an estimation of these synergies, the economics of scale and the increased combined value of the new corporate structure. After all of these estimations occur, the bidding firm sets an objective and specific offer price. But, in practice, a large number of assumptions must be made to come to these estimations. Boards can bluff, other bidders may emerge, or the selling party know important information that is unknown to the acquirer; these real-life considerations make the offer price a more subjective estimation (Baker et al., 2012).
Daniel Kahneman and Amos Tversky are one of the first important writers who connected psychology and economics. After this, a wave of academics started writing papers about several behavioral biases. Roll (1986) was one of them and determined the “hubris” hypothesis. He showed the existence of hubris by professional managers who are making decisions about the acquisitions. Roll’s hypothesis was the starting point for academics to document the overconfidence bias in M&A. This is one of the many theories and biases that exist in an M&A process. Generally these theories and biases are focused on the bidder perspectives and leave the target perspective aside. This paper will emphasize both sides during the M&A. An M&A process consist of a number of stages. Viguerie et al. (2007) fragmented the M&A process in three different steps. The first step in the M&A process is the preliminary due diligence, second is the bidding phase, and the last step is the final phase of closing payment term. Thereafter there is the holding period. Targets are actually in the holding period while bidders are in the bidding phase. This paper will highlight in particular the behavioral biases in the last stages and will show that when the current stock price of the target is near their recent peak price the chance of a merger will increase. The reference point of view in mergers will be used to explain the main hypothesis. The reference point of view was primarily established in an article by Malcolm Baker, Xin Pan and Jeffrey Wurgler in 2012. Baker et al. (2012) found that several
4 important aspects of mergers are shaped by psychological reference points. For instance, recent peak prices help to explain the bidder’s offer price and to explain merger waves. The reference points in this article are recent peak prices of stocks and indexes. These peak prices are in line with the peak prices in the article of Baker et al. (2012). The 1-year high of a stock price is easy to obtain and will be the bases for the independent variables.
The intensity of merger activities fluctuates over time and is determined by several of motivations. These motivations are mainly financial and have been researched very frequently. Where Baker et al. (2012) already shaded some light on the relation between the anchoring biases and the phenomena of merger waves, this paper will further investigates the influence of irrelevant peak prices on the concentration and clustering of merger activities. It is important to know that mainly irrelevant peak prices have influences on merger activity. The reference point of view explains other views of different articles on M&A. While it is obvious that no single theory will entirely explain the existence of merger activity and merger clustering, the reference point of view theory will fill in some of the open questions.
This research will extend the Baker et al. (2012) study on merger waves. In the first place this study amplifies that individual gaps of targets are predictors of the timing of a merger. Other market peak prices will be used to test if those peak prices are as well predictors for merger waves.
Furthermore, an investigation will be conducted regarding the consistency of the reference point of view after a huge decline in the stock market. Two recessions will be investigated to test if the reference point of view still holds after a period of decreasing stock prices and indexes. While investigating the individual merger activity, a distinction will be made between cash and stock acquisitions. The last test is about a potential anchor or reference point, and will be checked to see whether these new anchor is related with the timing of a merger.
The structure of this paper will be divided into another seven chapters. Chapter 2 will give a guide throughout the most important literature related to the theory which is necessary to
understand the different biases and the underlying thoughts for this paper; furthermore, Chapter 2 provides some other biases which may emerge in a bidding process. The third chapter is used to describe the hypotheses. Chapter 4 will present the methodology that will be used to come to the results for the research questions. Data collection and description are discussed in Chapter 5, while Chapter 6 will lay out the empirical results and Chapter 7 will check the data for robustness. Finally Chapter 8 concludes this research paper.
5
2. Literature Review
It is important to explain and understand the underlying theory and behavioral biases. This section will present the literature that is related to the research and the hypotheses named later. First a review will be determined with a number of theories, biases and effects. Finally there will be shed some light to the merger activity, merger waves and to other biases that can emerge in the bidding process. This is useful and helps to understand the principals of merger activity and the region where the reference point of view is established.
2.1 Related Literature
Kahneman and Tversky (1974) found that the anchoring and adjustment bias on an initial belief or numerical estimate is an original cognitive bias. In their research, they established that even though an initial thought is randomly produced, any changes away from this the initial thought are almost always insufficient.In their work they use a simple numerical equation to show the anchoring bias (see Box I, Example 1 and 2).The anchoring and adjustment bias can be strongly linked with two other behavioral biases, overattribution bias (Corneille et al., 1996 and Quattrone, 1982) and
hindsight bias (Hell et al., 1996). Overattribution bias occurs when people determine and adjust their human behavior by the causal factor whichever is most available to them. They try to find reasons for their own personal behavior. Hindsight bias occurs when people claim that they know more about things then they actually knew (Hell et al., 1996). Additional works documented on the anchoring bias are an article of Neale et al. (1987) which related the anchoring bias with real estate valuation, consumers spending judgment which was the main findings in the paper from Hoch et al. (1998), estimations of altered risks was determined by Anderson et al. (1989) and duty motivation
emphasized in a paper by Switzer et al. (1991). Switzer et al. (1991) examines anchoring biases upon judgments of future effort, future performance and on actual judgment in time and effort. Specialists of respected auditing firms are mostly determined as highly objective. But even these specialists have anchoring biases (Example 3).The true value of stocks is more difficult to define or form than, for instance, real estate valuation. Nonetheless in the stock markets, anchoring bias vastly exists. (Shiller, 1990).
Kahneman and Tversky expanded their research and emphasize the Prospect Theory in 1979. The Prospect Theory explains that around the reference point there will arise an S-shaped utility function which is concave in the domain of gains and convex in the domain of losses. This indicates that people are risk-seeking in the domain of losses and risk-averse in the domain of gains. Another important aspect of the Prospect Theory is loss aversion. Loss Aversion is the fact that people have a steeper utility function in the domain of losses than the utility function in the domain of gains, this
6 means that losses hurt more than gains of the same magnitude makes you happier (Kahneman and Tversky, 1979). The reason behind this is that choices with a risky forecast are showing several pervasive effects that are inconsistent with the basic utility theory. People are underweighting probable outcomes with a higher expected return in comparison with certain outcomes (See Box II
for an example).
One important effect, which often is linked to the Prospect Theory, is the disposition effect. Shefrin and Statman (1985) was the first paper to make this relation. Weber et al. (1998), Odean (1998) and Grinblatt et al. (2005) are three research papers who came with more evidence for the disposition effect and argued the relation between the Prospect Theory differently. The arguments
Box I:
Example 1: Anchoring and adjustment bias (Kahneman et al., 1979)
Different groups of students where ask to do one of the following simple expressions
8 x 7 x 6 x 5 x 4 x 3 x 2 x 1 or 1 x 2 x 3 x 4 x 5 x 6 x 7 x 8
The median of the first groups who did the first equation was 2.250 and the median for the second groups who made the second equation was 512. The correct answer is 40.320
In case of this example with a different initial value or staring point lead to a different estimation of the outcome
Example 2: Anchoring and adjustment bias (Kahneman et al., 1974)
Another experimentation, applicants observed the spinning of a roulette-type wheel decorated with numbers from 1 to 100. Then the applicants were asked what percentage of African countries were members of the United Nations. The random numbers generated by the wheel had
influences to the answers on the answers of the applicants.
For example, when the roulette wheel number was 10, the median answer was 25%; when the spun number was 65, the median answer was 45%.
Example 3: Anchoring and adjustment bias (Biddle et al., 1981)
Specialists of the four largest or “biggest” auditing firms where asked to appraise the fraud rate for approximately a 1000 firms. The group was split up into two separate groups and they were linked to different kind of anchors. The first group was asked if 10 firms of the approximately 1000 firms committed fraud, the second group were asked if fraud was expected in 200 of the 1000 firms committed fraud. Next the separate groups were asked again to estimate the real fraud rate in the sample. The group, who was linked to the anchor of 200 firms committed fraud out of 1000, had a significantly higher fraud rate then the other group, who were linked to the anchor of 10 of the 1000 firms committed fraud. So the different groups of specialist in the four largest auditing firms clearly failed to adjust their final estimation on the fraud rate to their initial anchor.
7 between these several explanations of the connection between the Prospect Theory and the
disposition effect vary slightly from paper to paper, but the essence is always the same. The
disposition effect is the tendency of investors to sell shares whose price has increased, while keeping assets that have dropped in value. Investors are less willing to recognize losses, but are more willing to recognize gains. Odean (1998) discards other theories for the investor’s bias of less willing to recognize losses and more willing to recognize gains, like portfolio rebalancing and lower trading costs related with low-priced stocks. Grinblatt et al. (2001) found the same outcomes for investors from Finland and Shapira et al. (2001) the same results for Israel. The fact that investors on the stock market are unwilling to sell stocks who are in the domain of losses relative to those stocks who are in the winning domain is specifically shocking because capital gains are linked with tax cost and capital losses are linked with tax benefit.
The above explained behavioral biases are the bases for the reference point of view (Baker et al., 2012) in mergers. Baker et al. (2012) proposed in their article a reference point view of mergers, this holds that salient but mostly unrelated reference point stock prices of the target underwrite M&A activities through both the prices and the types and quantities of firms traded, this idea is based on the prospect theory. Baker et al. (2012) concluded that a variety of recent peak prices help to explain the bidder’s offer price, the probability of a deal success and the bidder’s announced returns. A fourth phenomenon, which is explained with the reference point prices are aggregate merger activity or merger waves. They found that the 52-week high peak prices is a negative predictor of merger activity, if market prices are ten percent point below the 52-week high peak price then the aggregate merger activity drops by eighteen percent point relative to their trends.
Most of the biases emerge from the perspective of the bidders. The reference point of view explains different aspects from the view of the bidders as well, but generally on the perspective of targets in the activity of mergers. In the bidding process several psychological biases can emerge for bidders and for targets. A bidder needs to set a price and several behaviour aspects may affect this bid price. Irrelevant peak prices of the last period may become heavily anchors. When the
adjustments of these anchors are not sufficiently, this anchor affects the bidding price. It may even lead to not bidding at all. A bidder finds it easier to pay a recent peak price with a low premium than vice versa. From the perspective of the target it is the other way around. The shareholders from the targets need to approve in case of a merger. As shown by Odean (1998) these shareholders are less willing to recognize losses and more willing to recognize gains. These gains and losses are occurred by their reference points. And what Baker et al. (2012) already studied and what this paper will further study is that merger activity is related to that reference point and these reference point are the peak prices of those companies who are the targets.
8 Genesove et al. (2001) did a quite similar study about the reference point dependence. They used data from downtown Boston’s housing market between 1990 and the end of 1997. They used the purchased price as a plausible reference point. Obviously, dwellings sell quicker and mainly above asking price when the market is in an upward slope and vice versa when the housing market is in a burst. Genesove et al. (2001) found that there is a lower probability of sale, a higher asking price and a longer time on the market when in loss region. The results were quite the same as in mergers. The main finding in their article was that they revealed that loss aversion affects seller behaviour in the market of dwellings in real estate.
2.2 Merger Activity and Waves
Motivations for a particular merger can be financial or non-financial. Certainly the most mentioned explanation for a merger is the microeconomic explanation, economies of scale. The acquirer is always trying to find an optimum balance. By improving its scale, the acquirer wants to reduce the average costs per unit (this is a result of spreading the fixed costs over a larger number of production). As mentioned in the introduction, an additional reason, which found in several standard textbook, is the advantages of synergies. These synergies arises when recourses of the target gives
Box II:
Example 3: Utility function convex/concave
From Kahneman et al. (1979) experimented a convex utility function in the domain of losses and a concave function in the domain of gains. They illustrated this by these kinds of tests:
You have been given 1000 euro and you have to choose between the following options:
1) gain 1000 euro with 50% chance or 2) 500 euro with 100% certainty The outcome was that 16% choose option 1 and 84% choose option 2
Then you have been given 2000 euro and you have to choose between the following options:
1) loss 1000 euro with 50% chance or 2) loss 500 with 100% certainty The outcome was that 69% choose option 1 and 31% choose option 2
This indicates that people are risk averse in the domain of gains and risk seeking in the domain of losses
Note that the final outcome of tests A and B are the same (option 1: 2000, 50% or 1000, 50% and option 2: 1500 100% certain).
9 financial benefits the acquirer. Another neoclassical explanation of merger activity is the industry shock theory which is mainly established by Mitchell et al. in 1996. The writers explain the
appearance of merger clustering (merger waves) by an economics shock in a particular industry and argue that M&A is a tool to restructure the industry as a response of the shock. Other academics who support the idea that merger waves are explained by an economics shock in a particular industry are Blair et al. in 1993. Blair et al. (1993) investigated the relation between financial restructuring and free cash flow during the 1980s, and found sufficient evidence that this relation exist, especially for the first part of the 1980s. Harford (2004) suggest an important component for a merger wave to occur, this component is capital liquidity. Capital liquidity causes capital reallocation, and will boost merger activity when there is an industry shock.
A more behavioural hypothesis to clarify merger activities begins by the managerial hubris hypothesis (Roll, 1986). Roll’s hypothesis proposes that managements are taking a value maximizing approach, but miscalculate the value of the takeover. During trends in merging, managers might not put enough effort in execution adequate valuation and basically become caught up in the market trend. Without the necessary due diligence, the managers could overvalue benefits, undervalue the disadvantages and overrate their capability to manage the bigger firm. Roll’s theory formed the bases for the explanation of merger waves which was determined by Rhodes-Kropf et al. (2003). In the paper of Rhodes-Kropf et al. (2003) they show that rational managerial behavior and uncertainty about misevaluation would lead to a correlation between market performance and merger waves. They overestimate synergies, given uncertainty in the markets about true company value during stock market booms.
The Q-theory of investments expects that a firm should increase their investment rate in line with their Q (market-to-book ratio). Jovanovic et al. (2002) supports this and linked the Q-theory with merger activity. Jovanovic et al. (2002) found that a particular firms M&A activity responds by their Q. Another important aspect of their research was that the Q explains some aspects of merger waves. The writers found that several merger waves were a responds to profitable reallocation opportunities. More recent article of Shleifer et al. (2003) supports the Q-theory and argues that we see waves in merger activity because of stock overvaluation. Therefore, overvalued companies are more likely to acquire other companies, either because these companies can take benefit of their overvalued acquisition exchange and can finance the acquisition better. Companies who are undervalued, or relatively less over valued are more likely to become targets then overvalued companies. Verter (2002) found evidence of higher levels of merger activity in higher-valuation markets and that this correlation is mainly caused by mergers financed by stocks. Rhodes-Kropf et al. (2004) also support the Q-theory and showed that merger waves and market-to-book ratios, relative to its true estimate valuations, are positively correlated. Higher valued companies, in terms of
10 market-to-book ratio, are more likely to be bidders comparing to lower valued companies who are more likely to be targets.
The discussion in the sector above emphasized that there are two key explanations which can explain merger waves. First, the neoclassical findings by Shleifer et al. (2003) and by Rhodes-Kropf et al. (2004) and second the behavioral explanation by Mitchell et al., (1996). A third explanation can be added, also a behavior explanation, and is based on the lemming effect. The lemming effect linked with M&A behavior describes a situation where decision makers copy the markets trends of M&A behavior. This behavior is mostly due by decision makers who are afraid of missing on opportunities by remaining in a stabilized industry (Lovallo et al., 2010). Some more recent papers propose that merger waves can be caused by jealous CEOs. For instance Goel et al. (2008) emphasizes in their article that CEOs are jealous of peer CEOs their compensation. These compensations are determined by the company’s value, therefore the jealousy of the CEOs can give mergers waves a boost, because a way to increase company value and output is to merge and acquire other companies.
2.3 Other biases that influences merger activity
The following section is created to show the other biases that may emerge in an M&A process. These biases will not have a substantial influence on the further research.
A bias which was first examined by Capen et al. (1971) is the winners curse. The winners curse may emerge when bidders get a noisy signal, eventually the bidder with the highest signal places the highest bid and wins. If you know you have won means that you had the highest signal, probably too high, so winning may be bad in this case. Another bias, a bit more social bias, is the sunflower effect. This effect shows that lower scaled employees are in the uncertainty to disapprove the CEO’s policies because they are not want to bring their future career in danger. (Sibonny et al., 2006).
Irrational escalation of commitment bias and the sunk-cost fallacy can be named together. Commitment bias causes people to increase the commitment of resources and therefore take more risk after a negative result of their investments (Staw, 1976). The bias is especially large for people who were personally responsible for the negative investments, and are unwilling to commit their mistakes and failures. Managers can frequently continue in bid even when the price exceeds the original value, because the manager wants to defend their bid (Smit et al., 2010). It is clear that this arises more when the process of the bid is very costly. That is why sunk cost fallacy is related to commitment bias, sunk cost fallacy is where increased investments are caused by previous
cumulative investments. The last bias that will be discussed is company pressures. Like the anchoring adjustment, company pressure is a bias judgment of managers. The recourses and liquidity is not
11 endless, definitely not for acquisition projects. So the internal competition between managers cheers the managers to make highly optimistic appraisals of the targets in order to get the funding from the partners to set up a bid (Kahneman et al., 2003).
12
3. Hypotheses
The key focus of this chapter is the creation of the hypotheses. A combination of different theories and literature will be used to develop these hypotheses. First, the main hypothesis and thereafter the other hypotheses will be determined.
Main Hypothesis: The gap between recent peak prices and current stock prices is a negative predictor of merger activities.
The gap is the difference between an irrelevant peak price at a certain date and the current price at that certain date. The Prospect theory (Kahneman et al., 1979) states that people are risk-seeking in the domain of losses and risk-averse in the domain of gains. Further, the disposition effect (Odean, 1998) declares that shareholders are less willing to sell if they are in a domain of losses comparing to the domain of gains. When the gap is high, the current stock price is decreased in the near past. Consequently, when the reference point of the shareholders is the recent peak price (for instance, the 1-year high), they are facing losses and are not willing to sell, at least until some acquirer is paying the 1-year high. On the other hand, the acquirer is more willing to pay the recent peak price if it is not far from the current price. Therefore, based on the literature and theory, the expectation of the main hypothesis is that the gap is a negative predictor. This is consistent with the findings from Baker et al. (2012), who showed that when the market gap between several peak levels and the current level of the market index grows, the merger activity drops. Baker et al. (2012) used the 52-week high market index and the current market index to measure a gap.
To answer the main hypothesis, five hypotheses will be formed. While studying the main hypothesis and obtaining the data it was clear that the market gap, which Baker et al. (2012) used, was slightly to modest. Consequently this paper distinction the gap in a market gap and an individual gap. The market gap is in line with Baker et al. (2012) and the individual gap will be estimated for each company who are a target in a certain period. Hypothesis 1 will use the market gap, hypothesis two will use the individual gap. After these two hypotheses are investigated, the reference point of view is further been tested in the third hypothesis. An analysis will be conducted regarding two recoveries in two different time periods. The analyses will check whether the reference point of view holds after a recession. In the fourth hypothesis a sample distinction will be made between
acquisitions financed by cash and by issuing stocks. Hypothesis 5 will be another new hypothesis where the first and second hypotheses are combined. It will be investigated if individual targets have, beside the individual reference point, a second anchor in form of the market peak prices.
13
Hypothesis 1: The market gap has a negative effect on the level of merger activity.
The market gap is the differences between an irrelevant peak of the market index and the current level of the market index. As discussed earlier, this hypothesis was already tested by Baker et al. (2012) with a time window from 1973 until 2007. They found a negative relationship between the market and the aggregate merger activity. Their research used the 52-week high of the market index to estimate the market gap. This paper uses, in line with Baker et al. (2012), the 1-year high and extends the research with the half year high and the two year high peaks of the market index. From the findings of the literature and the discussion in the main hypothesis, the expectation is that the gap, using these different peak prices, will have a negative effect on the level of merger activities.
Hypothesis 2: The individual gap is a negative predictor for the timing of merger activity.
The individual gap is the differences between an irrelevant peak price at a certain date of a particular stock and the current price of that stock at that certain date. As mentioned earlier, by studying the data, it was noted that the individual gap around the date of a merger was particularly low. Figure 1 emphasises this observation. From the more neoclassical literature, the probability of an individual merger has a strong relation with the firm-level characteristics of each individual target and the market conditions. But from a behaviour perspective, targets think the price is fair when their stock is near their reference point (peak prices) and on the other hand bidders will easier pay the gap or premium. In accordance with previous hypothesis, the expectation is that the individual gap is a negative predictor of an individual merger. Actually this hypothesis indicates that the individual gap is a negative predictor for the timing of the merger activity. This is partially due to the fact that only targets are selected in the sample.
Hypothesis 3: The recovery of merger activity is more rapid than the decline of merger activity after a burst of the stock market.
The stock markets are facing huge losses after a burst or during financial distress. The steep decline of the stock market will almost hit every company in the market. When this decline has ended, the recovery of the stock market begins. From the data that is used for this paper, it can be shown that the recovery is more slowly than the decline was in the first place. Looking at the current financial crisis, the time that the stock market was declining was less than one year versus the recovery, which is still not finished. This trend is closely seen after the Internet bubble. Prospect theory and disposition effect expects that shareholders will not sell after the decline and thus hold their current stocks. The gap between the peak prices and their current prices will increase significant for almost every stock in the market. Consequently, the expectation based on the main literature,
14 which also is used to explain hypothesis 1 and 2, is that the increased gap will lead to a downfall of mergers. After the decline in 2008, M&A activity declined by 38 percent in the United States and comparable amounts in other parts of the world (Duett, 2010). After one year, the differences between the peak prices (1-year high peak price) and the current stock price are vastly diminished. This is caused by the fact that the peak price is the 1-year high; therefore after one year the peak is mostly around the current price. Therefore, when the reference point of view for mergers holds, a more rapid recovery of the merger activities is expected. Bidders are more likely to pay the recent peak prices due to the declining premium that they have to pay and the target shareholders are more likely to believe they can get a fair value for their companies (Corkery, 2010).
Hypothesis 4: In stock acquisitions, the individual gap and Mean Tobin’s Q are negative predictors for
the timing of a merger. In cash acquisitions, the individual gap is a negative predictor for the timing of a merger.
An acquisition can be financed by issuing stocks, cash, or other methods. When an
acquisition is financed by stocks, the expectation is that the Mean Tobin’s Q is highly related with the timing of the merger (Verter, 2002). Again from Shleifer (2003), when the companies in the market are on average more overvalued, they are more likely to acquire other companies. The acquirers are paying more easily the premiums (arising from the individual gaps) when their stocks are overvalued because these overvalued companies can take benefit of their overvalued acquisition exchange and can finance these acquisitions better. Because the premiums can be paid more easily, the effect of the individual gap will diminish. Cash acquisitions are not related with the Mean Tobin’s Q. When cash is used to take over another company, the bidder will use a smaller part of stocks or even issue no stock at all, and the effect of the Mean Tobin’s Q will not have any significant influences on the timing of an individual merger. The expectation is that, in cash acquisitions, the timing of a merger is highly related to the individual gap, and in stock acquisitions, the timing of a merger is more related to the Mean Tobin’s Q and less to the individual gap compared with cash acquisitions.
Hypothesis 5: The market gap is a negative predictor for the timing of an individual merger
This hypothesis will be an extension of the second hypothesis by adding a market gap (the difference between the 1-year high price of the market and the current price of the market). The expectations of the hypotheses 1, 2, and 3 are mainly based on the literature review, namely the anchoring biased and the Prospect Theory (Kahneman et al., 1979). According to hypothesis 2, a negative relation between The individual gap (1 yr high) and the timing of a merger is expected for the target. In this hypothesis, the expectation is that targets can likewise have a reference point or an anchor based on The market gap (1 yr high). The shareholders feel that they are losing if they sell
15 when the market is below its 1-year high; this effect is conditional on the individual gap. Even if the firm’s stock price remains constant, a declining stock market will make targets unwilling to sell. They will anchor their beliefs on the 1-year high market value. Therefore, a negative relationship between the market gap and the timing of an individual merger is prospected. The 1-year high level of the market index acts as a second anchor or a reference point for the target.
16
4. Data and Descriptive Statistics
This chapter primarily contains and describes the data that will be used in this paper. The first section deals with the data collection and describes the databases. The second section shows the most important descriptive statistics.
4.1 Data Collection
In order to measure the models predictive performance a number of datasets need to be merged and variables need to be estimated. In total, three different databases are used. All mergers activities are obtained between January 1, 1984 and December 31, 2013. This research starts from 1984 because, from that time, the total number of M&As has far exceeded the number of M&As in the 1960s and 1970s. M&A activities in 1960 are particularly known as the conglomerates merger activities. From 1980, an increase in M&A activities was exposed between companies of diverse sizes and industries (Tetenbaum, 1999).The new data (compared with Baker et al., 2012) from 2007 till 2013 is interesting, as the market gap increased significantly after the burst in 2007. Thomson Reuters database covers information about all merger activity; this research only obtains merger activity that is concentrated in the United States. The Center for Research in Security Prices (CRSP) is used to collect information about the securities in the AMEX, NASDAQ and NYSE stock markets. This data is used to obtain the peak prices and current prices for each individual target. Moreover, CRSP has a stock market index; this index is based on the AMEX, NASDAQ and NYSE stock markets. The market data is obtained for the market peak prices. The time window will be the same as the Thomson Reuters time window. Furthermore, the Merged Compustat database facilitates all firm level characteristics about companies in the United States, corresponding data of the balance sheets and income statements. This data is primarily used to determine the most important control
variables. From Merged Compustat, data will be obtained with again the similar time window (1984-2013). To define the two recessions and their recovery, the National Bureau of Economic Research (NBER) is consulted. The NBER is a non-profit research organization that provides information about economic business cycles and recessions. From 1984 until 2013 there were three recessions; two of these three recessions will be used to test the third hypothesis.Part of the information provided by the NBER are the peaks and the troughs of the recessions. The first recessions had its through in the first quarter of 1991. The second recession is well known as the Internet bubble and burst. From the peak in the first quarter of 2001, the recessions began and had a duration of eight months, ending in the second quarter of 2002. From June 26, 2002, a trial of three years is obtained. The last recession between 1984 and 2013 began March 3, 2009. From this date on, another trial of three years will be taken to analyse the a-symmetric recoveries between the stock market and the merger activity.
17 Because of the fast recovery of the first recessions, only the last two recessions will be used in this research. The latter part of Section 4.1 will determine the dependent variables, independent variables and the control variables.
Merger Activity
The dependent variable, Merger Activity, is derived using data from the Thomson Reuters database. All merger activity, public and private targets, in the United States are selected in the sample. The total merger activity in each quarter from 1984 until 2013 is calculated (Merger Activity contains 120 observations). A merger is selected in the sample when the acquirer owns more than fifty percent of the total stocks after the transaction. In total, 108.989 private and public mergers were selected from Thomson Reuters. Aggregate merger activity needs to be detrended. Detrending is done by dividing the total merger activity for each quarter by an average of quarterly merger activity with a trial of two years, starting one year prior to the specific quarter. Another aspect that has to be taken into account is the normalization for the total companies in the index. The number of companies in the CRSP market index, which consists of three stock markets AMEX, NASDAQ, and NYSE, fluctuates between 5.478 and 9.104 in the time window. This study will normalize the
detrended quarterly merger activity by multiplying the total quarterly mergers with a percentage of companies in that particular quarter, because a higher number of companies will certainly causes more merger activity.
Dummy Merger
Equivalent to the dependent variable Merger Activity, the dependent variable Dummy Merger is obtained from Thomson Reuters database. Dummy Merger is significantly different from Merger Activity, because only public targets that have been merged are selected. Two other requirements are that the acquirers own more than fifty percent of the target’s outstanding shares after the transaction and the target has to be settled in the US. A total of 8.972 targets are selected for this dataset. Every observation contains a “0” if there is no merger, and “1” if there is a merger in a specific quarter for the target. Some targets have more than one time a “1” in their time window; this indicates that the target company has merged more than one time.
The market gap (1 yr high)
This independent variable, The market gap (1 yr high), is derived from the CRSP database. CRSP has data for an index that contains the AMEX, NASDAQ and NYSE. The market gap (1 yr high) is estimated by subtracting the total value of the index (CRSP: totval) from the highest total value of the past year. When a merger or takeover take place there is, prior to the announcement date, always
18 some rumours. There are some individuals that have more information, about that specific merger, than others. Therefore, the total value lags by one month because these noisy signals and rumours need to be taken into account.
The individual gap (1 yr high)
This independent variable The individual gap (1 yr high) is calculated by using the CRSP database. By observing the highest trading price during the day (CRSP: askhi) a 1 yr high price is determined for each observation. The individual gap (1 yr high) is calculated through dividing the 1 yr high price by the price (CRSP: prc), because the prices need to be controlled for rumours and noisy signals the one-month lag is used. The individual gap (1 yr high), is winsorized at the top and the bottom 1%.
Tobin’s Q
The Tobin’s Q is a firm characteristic control variable estimated for the targets each year. The data is collected from the Merged Compustat database. Each company only updates their annual fundamentals once in a year, so only one month in a year has an observation that contains a Tobin’s Q. The best prediction for the other months is around the date when the firm-level characteristics are published. Therefore this research will provide the five previous months as well as the next six months with the Tobin’s Q. The Tobin’s Q is defined as the market value divided by the total assets (Compustat: AT). The market value is estimated as the total assets minus book equity plus market equity. The book equity is calculated by the total shareholders’ equity (Compustat: SEQ) plus deferred taxes and investment tax credit (Compustat: TXDITC) minus the redemption value of preferred stock (Compustat: PSRKRV). The market equity is estimated by multiplying the price
(Compustat: PRC) by the outstanding shares (Compustat: SHROUT). The Tobin’s Q is winsorized at the top and the bottom 1%.
Mean Tobin’s Q
From Maksimovic et al. (2001) and Jovanovic et al. (2001) we know that merger activity will be affected by the market-to-book ratio (Tobin’s Q). The annual Tobin’s Q for each individual company is estimated by collecting data from Merged Compustat. Each company announced their annual statements at a different month, so it is possible to determine a mean quarterly Tobin’s Q. There are significantly more observations in the last quarter of the year (i.e., October, November and December) as compared with the other quarters (65% of the observations in the last quarter, the remaining observations are fairly distributed over the quarters). Nonetheless there were sufficient observations to estimate a legitimately Mean Tobin’s Q for each quarter in the sample.
19
Other Variables
Other gaps, such as The market gap (2 yr high) and The market gap (0,5 yr high) are
calculated in the same way as The market (1 yr high) (Hypothesis 1), only these gaps using other peak prices. The Return on Assets is calculated by Net income (Compustat: ni) divided by total assets (Compustat: at). The Earning Price ratio is defined by dividing the earnings before interest and tax (Compustat: EBIT) by the market equity (calculations aforementioned). The Return on Assets and the Earning Price ratio are two control variables used in all the hypotheses. These variables are
winsorized at the top and bottom 1%. Furthermore, the SIC codes for the companies are downgraded to their first numbers, to divide them in major groups, for instance agriculture and construction industry.
4.2 Descriptive Statistics
An overview of the collected dataset is presented in the following table:
[ Table 1 ]
In Table 1 an overview is presented of the different variables used in this paper. A total of twelve variables are obtained, including two dependent variables, four independent variables and six control variables. The total number of observations differs for almost each variable. It fluctuates between 120 and 247.251 observations. The market gap (1 yr high), reported in Table 1, varies between 1,00 and 2,35. The individual gap (1 yr high) fluctuates between 1,00 and 6,83 and the individual gap (1 yr high) also has quite some variations, as shown by the standard deviation.
20
5. Methodology
To investigate the causal effect between the market gap and the aggregate merger activity the following OLS regression (Hypothesis 1) will be used:
(1)
A control variable for the Market-to-Book ratio (Mean Tobin’s Q) will be added, because the last 125 years of business history indicate that periods of high Market-to-Book ratios will coincide with a high level of merger activity (Shleifer, 2003 and Jovanovic et al., 2001). Two other control variables, Return on Assets and Earning Price ratio, are added to the regression. Firms are more attractive if their ROA and E/P ratios are higher, so a positive relation with the total merger activity is expected.
To investigate the causal effect between the individual gap and the timing of the merger activity, the following probit model (hypothesis 2) will be executed:
| (2)
From a neo-classic perspective to merger waves, Mitchell and Muherin (1996) documented clear clustering of waves within industries and that these clustering where caused by numerous of technological, economic, or regulatory shocks in the industry. The industry fixed effect mentioned in the equation will control for this an industry shock. An individual Tobin’s Q is added because when the target’s market-to-book value is low the target is fairly prices or even under-valued by the market. So when the targets are undervalued the chance for being acquired is more likely than when they are overvalued (Rhodes-Kropf et al., 2004). Higher valued companies, in terms of market-to-book ratio, are more likely to be bidders comparing with lower valued companies who are more likely to be targets. Therefore the expectation is, compared with the Mean Tobin’s Q, that the individual Tobin’s Q has a negative relation with the timing of a merger. For the same reason and with the similar expectations, as in regression 1, three control variables are added, Mean Tobin’s Q ROA and E/P ratio. The variable year is included for controlling for time trends.
To investigate whether the recovery of the merger activity, after the trough of the
recessions, is more rapid than the recovery of the market (hypothesis 3), an analysis will be prepared to look for the asymmetric recovery of the Merger index and the Market index. The Merger Index controls for the trend of the past three months. Graphs of two recessions and associated recoveries will be determined and analysed.
21 To study the fourth hypothesis, the following probit model will be used:
| (3)
Hypothesis 4 emphasize that the cash and stock acquisitions have separate relations
recording to the individual gap and the mean Tobin’s Q. This model will therefore be divided into two subsets. The first sample only includes mergers financed by a minimum of 65% of stocks. The second sample will only contain mergers financed by a minimum of 65% of cash. The probit model in
Equation (4) is used for these two samples separately. Further this probit model uses all the control variables of Equations (2 and 3). The variable year is included for controlling for time trends.
For hypothesis 5, the chance that individual firms will be merged depends on the market gap and their individual gaps, an extended probit model of Hypothesis 2 will be used:
|
(4)
Compared with the second probit model (hypothesis 2), this model adds a new variables. The independent variable The market gap (1 yr high) which is also used in the regression for hypothesis 1. The variable year is included for controlling for time trends. The OLS regression used to test
hypothesis 1 uses time-series data. The probit models for hypotheses 2, 4 and 5 uses cross-sectional time series data.
22
6. Results
This part of the paper will show the results of the regressions and probit models. The results are displayed in tables and an exclusive analysis will be specified.
Hypothesis 1: The market gap has a negative effect on the level of merger activities
Table 2 displays the results of Equation (1) in the methodology:
[ Table 2 ]
As can be observed in Table 2, in the OLS regression without control variables (Columns 1, 4 and, 7), the findings are that the market gap coefficients are negative related with the level of Merger Activity, these effects are significant at the 1% level. Adding control variables (Columns 2, 3, 5, 6, 8, and 9) decrease the magnitude of the effect but increases the R-squared.
Baker et al. (2012) found that when market prices are 10% below their 52-week high, the merger rate falls by 18% relative to its trend. The results in this paper support this negative effect, but the effect is slightly smaller. Column 1 shows that if the The market gap (1 yr high) increases by 1%, the Merger Activity will decrease with 0,88% from its trend. The trend is the average of quarterly merger activity with a trial of two years, starting one year prior to the specific quarter. Accordingly if The market gap (1 yr high) increases by 10%, the Merger Activity will decrease by 8,85% from its trend. Colum 2 shows the results of the OLS regressions with the Mean Tobin’s Q. If The market gap (1 yr high), in Column 2, increases by 1%, the Merger Activity will decrease with 0,74% from its trend. The independent variable, The market gap (1 yr high), is significant at the 1% level. Colum 3 adds two more control variables, Return on Asset and Earning-Price Ratio. Adding all of the control variables increases the R-square, from 0,2343 to 0,2966. Observing the results in Column 3, if The market gap (1 yr high) increases with 10%, the Merger Activity will decrease by 7,25% relative to its trend. The market gap (1 yr high) is again significant at a level of 1%.
Columns (4, 5, 6) and (7, 8, 9) have the same interpretation as the first three columns, only the market gaps are estimated with different peak prices. Consistent with the hypothesis and the expectations, The market gap (2 yr high) and The market gap (0,5 yr high) are negative predictors for Merger Activity compared to the trend of Merger Activity. The market gaps have respectively a negative effect of between 0,91-0,66% and 0,84-0,75% on Merger Activity. The independent
variables are significant at the 1% level. Comparing Columns 3, 6, and 9, The market gap (2 yr high) in Column 3 has the highest effect on the level of Merger activity at -0,75%, as well as the highest R-square.
23
Hypothesis 2: The individual gap is a negative predictor for the timing of merger activities
The following table shows the result of Equation (2) in the methodology:
[ Table 3 ]
Observing the outcomes in Table 3 shows that The individual gap (1 yr high) is a negative predictor of the probability that a merger emerge in a specific quarter. The individual gap (1 yr high) is significant at the 1% level in all the columns. In Column 1, the probability of the merger is predicted by only The individual gap (1 yr high). For Column 3, the same probit model is used but with a new control variable: Tobin’s Q. In Column 5, the probability of the merger is predicted by The individual gap (1 yr high), Tobin’s Q, Return on Assets, E/P ratio and the Mean Tobin’s Q. Columns 2, 4, and 6 are the similar probit model as respectively Columns 1, 3, and 5 but with dummy variables controlling for industry fixed effects. Each probit model are controlling for the time trend by including the year variable.
All the coefficients in Table 3 are the marginal effects on the DummyMerger. Besides the dummy variables for the industries, all variables are continuous variables. A marginal effect for continuous variables measures the instantaneous rate of change. They provide a decent
approximation to the amount of change in the dependent variable that will be produced by a one unit change in independent variable. In this model the change of the probability of a merger in a specific quarter caused by a 1% change in The individual gap (1 yr high). The other variables are calculated at the multivariate point of means.
In Column 1, the negative marginal effect of The individual gap (1 yr high) on the Dummy Merger is 0,34%. Column 3 adds the control variable Tobins’s Q but changes the magnitude of The individual gap (1 yr high) slightly, which is, in contrast with Column 5, where the probit model adds two more control variables. The negative marginal effect of The individual gap (1 yr high) on the Dummy Merger changes to 0,57%. For instance, if the 1-year high peak price is 30% above its current price, the probability of a merger in that quarter is decreasing with 17,10%.
The results of the probit model in Columns 2, 4, and 6 are identically built up as the Columns 1, 3, and 5. The only difference, as mentioned earlier, is that Columns 2, 4, and 6 are controlling for industry fixed effect. Controlling for industry fixed effect has no major effects in the probit models. The marginal effect and the t-statistics are approximately the same. Column 6 shows that the highest negative marginal effect of The individual gap (1 yr high) on the Dummy Merger . The negative marginal effect is 0,57%. All the results obtained in this table are consistent and supports hypothesis 2.
24
Hypothesis 3: The recovery of merger activity is more rapid than the decline of merger activity after a burst of the stock market
Two graphs of each of the burst/recoveries are determined in Figure 1 and 2. The two graphs, which are about the Internet bubble/burst/recovery and the Housing
bubble/burst/recovery, will have a time window that contains the burst and the recovery.
The Internet bubble/burst/recovery
[ Figure 1 ]
The recovery of the market and the merger index are approximately with the same speed or coefficient. Figure 1 shows that the merger index is recovered earlier than the market index. The recovery of the merger index is approximately three months earlier: this is in line with hypothesis 3. The findings in Figure 1 are definitely not significant and therefore cannot be used to determine that Hypothesis 3 is consistent.
The Housing bubble/burst/recovery
[ Figure 2 ]
The recovery of the merger index is significantly slower than the market index. But in line with the Internet recovery, the merger index is recovered earlier than the market index. The recovery of the merger index is again approximately three months earlier. This is again somewhat in line with hypothesis 3, and indicates that the earlier recovery is due to the reference point of view. The findings in Figure 1 and 2 are definitely not significant and therefore cannot be used to support Hypothesis 3.
Hypothesis 4: In stock acquisitions, the individual gap and Mean Tobin’s Q are negative predictors for
the timing of a merger. In cash acquisitions, the individual gap is a negative predictor for the timing of a merger.
The following table shows the results of Equation (3), estimated for the subset of mergers financed by stocks, in the methodology:
25 The results in Table 4 show the marginal effect of the coefficients related to the timing of the merger activity. These marginal effects can be interpreted in the same way as explained in the Results section for Hypothesis 2. All acquisitions in this sample are financed by stocks (at least 65% of the total acquisition price), which accounts for a total of 2.211 mergers. Column 1 shows that the individual gap is still a negative predictor for the timing of the merger. It is significant at the 5% level. The magnitude of the negative effect is less than when the sample consist all the targets. Column 3 is the same probit model but adds the control variables, individual Tobin’s Q and market mean Tobin’s Q. Adding these control variables causes the individual gap to be insignificant even at the 10% level. The results in Column 3 show that the only predictor of the timing of the merger activity, when the acquisition is financed by stocks, is the Mean Tobin’s Q. The Mean Tobin’s Q is a positive predictor for the timing of the merger in this sample. Column 3 shows that the positive marginal effect is 2,23%. Column 5 adds two more control variables, ROA and the E/P ratio. These control variables are firm characteristics where the expectation is that they are positively related with the attractiveness of the target. The Mean Tobin’s Q is significant at the 1% level and the individual gap (1 yr high) is significant at the 5% level. The magnitude of the negative effect of the individual gap is again much less than when the sample consist all the mergers. Columns 2, 4, and 6 are the same probit models as respectively Columns 1, 3, and 5 but control for industry fixed effects. The differences between these columns are very small. The results displayed by Table 4 supports the hypothesis.
The following table shows the results of Equation (3), estimated for the subset of mergers financed by cash, in the methodology:
[ Table 5 ]
The results in Table 5 display the marginal effect of the coefficients related to the timing of the merger activity. Again, these marginal effects can be interpreted in the same way as explained in the Results section for Hypothesis 2. All acquisitions in this sample are financed by cash (at least 65% of the total amount): a total of 3.367 mergers are obtained. Column 1 shows that the individual gap is a negative predictor for the timing of the merger. It is significant at the 1% level. The magnitude of the negative effect is much higher than when the sample consist all the mergers. Column 3 is the same probit model but adds the control variables, individual Tobin’s Q and the market mean Tobin’s Q. The Mean Tobin’s Q is significant at the 5% level but the magnitude of the effect is significantly less compared with the acquisition financed by stocks. Again, the individual gap is significant at the level of 1%. The negative marginal effect of the individual gap on the probability of the timing of the
26 merger activity is 0,84%. Column 5 adds the two variables about the firm characteristics of the targets. The negative marginal effect of the individual gap increases to 1,15%. The individual gap is significant at the level of 1%. The Mean Tobin’s Q is still significant at the 5% level but again the magnitude of the effect is significantly less compared with the acquisition financed by stocks.
Columns 2, 4, and 6 are the same models as Columns 1, 3, and 5 but control for industry fixed effects. The results in this table partly support the hypothesis. The Mean Tobin’s also partially predict the timing of the merger, all though the magnitude of the effect is much less than compared with the sample with stock acquisitions.
Hypothesis 5: The market gap is a negative predictor for the timing of an individual merger
The following table is the result of the Equation (4) determined in the methodology:
[ Table 6 ]
As can be observed in Table 6, Column 1, The market gap (1 yr high) is a negative predictor of the probability that a merger emerge in a specific quarter. In Column 1, the probability of the merger is predicted by The individual gap (1 yr high) only. For Column 3, the same probit model is used but with two additional control variables: The individual gap (1 yr high) and the Tobin’s Q. In Column 5, the probability of the merger is predicted by The market gap (1 yr high), The individual gap (1 yr high), Tobin’s Q, Return on Assets, E/P ratio. Even in this probit model, The market gap (1 yr high) is a significant negative predictor at the 1% level. Column 7 adds the Mean Tobin’s Q. This causes the market gap (1 yr high) to be insignificant at the 10% level. Columns 2, 4, 6 and 8 are a similar probit model as respectively Columns 1, 3, 5 and 7, but with dummy variables controlling for industry fixed effects. Each probit model are controlling for the time trend by including the year variable.
The results in Table 6 show the marginal effect of the coefficients related to the timing of the merger activity. These marginal effects can be interpreted in the same way as explained in the results section for Hypothesis 2 and similar as the other probit models are interpreted. Column 1 shows that The negative marginal effect of The market gap (1 yr high) on the Dummy Merger is 0,89%. Column 3 adds the control variables, The individual gap (1 yr high) and the Tobins’s Q and the negative
marginal effect changed to 1,13%. In Column 5, where the model adds two more control variable, the marginal effect is unchanged compared with Column 3. Column 7 adds the control variable Mean Tobin’s Q to the probit model. The market gap (1 yr high) is not significant at the 10% level. The negative relationship between the market gap and the timing of an individual merger is caused by the bidders. The market gap is highly correlated with the mean Tobin’s Q. The correlation is -0,45. When the market gap is large, the value of the stocks of potential bidders are dropped and therefore
27 more often undervalued. Undervalued companies are less likely to acquire other companies, either because these companies disadvantages of their undervalued acquisition exchange or they cannot finance the acquisition easily (Shleifer et al., 2003). The findings in these columns are inconsistent with hypothesis five and are not supporting that the market gap acts as a second anchor for the targets
Columns 2, 4, 6, and 8 are the same probit model as respectively Columns 1, 3, 5, and 7 only these columns are controlling for the industry fixed effect. The coefficient does not change
28
7. Robustness Checks
In order to evaluate the robustness of the results, additional tests will be done to see if similar results could be attained. A re-evaluation of the regression and probit models will be determined. Regressions and probit models will allow other independent variables, alternative control variables and other time windows.
In the first re-evaluation regression (1) will allow a few other independent variables. Four new peak prices (3, 9, 18, and 30 months high) are used to estimate the market gaps. These new market gaps are then performed similar as the OLS regressions of Equation (1). As can be observed in Figure 4, before 1994 the total merger activity is low comparing with after 1994. Therefore, the sample will be split into three different time windows. The first sample has a time window of 1984-1993, the second of 1994-2003 and the third 2004-2013.
[ Table 7 ]
The results, when looking at Table 7, are somewhat the same as the outcomes reported in the Results section. The four different market gaps are all negative and significant at the 1% level. The subsample, containing the time window 1994-2003, is negative and significant at the 1% level. The other two time windows are also negative, but not significant. This is caused by a reduced amount of observations.
The second robustness check will add three new firm characteristics, EBITDA/assets, Sales/assets and Liabilities/assets, to re-evaluate the probit model in Equation (2). These variables are winsorized at the top and the bottom 1%. The EBITDA ratio measures profitability of a firm in relation to its total assets. The Sales ratio shows the sales or revenues in relation with the total assets. The Liabilities ratio is a financial leverage measurement. The capital structure considerations are important in merger decisions and targets have systematically lower levels of leverage (Stevens, 1973).
[ Table 8 ]
Table 8 is a replica of the Table 3 but adds the three new firm characteristics in Columns 7 and 8. The EBITDA ratio and the Liability ratio have both a positive effect at the 1% level. Using the three alternative firm characteristics gives results that do not deviate much from the
29 Furthermore, the cash and stock acquisitions are selected when the acquisition price is 65% financed by respectively cash and stocks. In the robustness check for Equation (3) the selection will be based on 35% and 100%. The last robustness check can be linked to the Cash and stock robustness checks. Because it has been found that the market gap is not significant in probit model (4), Columns 7 and 8, the market gap is not a second anchor for the targets. This section will be used to test whether this is also observed for the subset of mergers that are financed by only cash. Cash acquisitions are not highly related to the Mean Tobin’s Q, as shown in the results.
[ Table 9 ] & [ Table 10 ]
As can be observed in Tables 9, the coefficients are somewhat similar to the coefficients displayed in Table 4. Columns 1, 3, 5, and 7 are showing the results with a subset of mergers
financed by 35% of stocks or cash. The number of observations does not differ much from the subset of mergers financed by 65% of stocks or cash, this indicates that mergers are almost always heavily financed by stocks or heavily financed by cash. The number of observation does differ for the subset of mergers financed completely by stocks or cash Columns 2, 4, 6, and 8. Nonetheless, the signs and magnitudes of the estimated probit model coefficients are not changing much comparing with the results in Table 4. Table 10 shows the coefficients of the probit model in Equation (4) using a new subset, this subset contains only mergers financed by cash. The marker 1 yr high is, even in this new circumstance, not an anchor or reference point for the targets.
The results in tables 7 through 10 displays the evidence that the estimated regression and probit models coefficients can be reliably interpreted as the true causal effects of the associated dependent variables.
30
8. Conclusion
This research investigated the relationship between peak prices and merger activity for American companies in the period from 1984 until 2013. Two main datasets are used: firstly, all merger activities (public and private) are included in the dataset and secondly, the dataset contains only mergers where the targets are public trading companies. A total of 8.972 public targets have been merged during this time period. To analyse and explain the behavioural biases and theories behind the relation between irrelevant peak prices and merger activity, a literature review has been constructed. Where obviously the anchoring biases, the Prospect Theory (Kahneman et al., 1974 and 1979) and the disposition effect (Odean, 1998) established in the theoretical framework, are the most important theories. These three theories along with the reference point of view in mergers (Baker et al., 2012) are forming the bases of this review. Other possible reasons that influence merger activity are emphasized and explained in the literature review and mostly added as control variables in the equations. This paper did not found any evidence to contradict these theories. Accordingly, five hypotheses have been formulated to explain the relation between the gaps (estimated with the peak prices and the current prices) and merger activity. To investigate this relation, this paper conducted different models and found significant evidence that the gap is negative related with merger activity.
First, this research found significant results that the market gap is a negative predictor for the level of merger activity. It can be shown that when the market prices are 30% below their 1-year high, the merger rate falls by 26,4% relative to its trend. This effect is less than the effect that Baker et al. (2012) found, the differences can be explained by the diverse time window (Baker et al. (2012) that uses data from 1973 -through 2007). It can be shown in the robustness checks that the last period (2004 - 2013) has the lowest effect. Nonetheless, the results are significant and in line with the findings in the main literature.
Second, a study is provided about the individual merger. In this second study, the main results are that the individual gap is a significant negative predictor for the probability of a merger. This paper conducted a dataset with all public targets and obtained all the individual gaps of these targets, from this point the research reduced the sample to quarterly data. The significant results shows that the negative marginal effect on the probability of a merger is 5,7%, in a specific quarter, when the individual gap is increased with 10%. Bidders are more likely to pay the recent peak prices due to the declining premium. And the targets are more likely to believe they can get a fair value, because the current stock price is near their reference point (Corkery, 2010).
31 The third study tries to test the reference point of view on two recessions is somewhat too modest. Considering the results, it cannot be said that the reference point of view hold or will not hold in these recessions.
Fourth, a distinction is made between stock and cash acquisitions. This paper found
significant evidence that stock and cash acquisitions differs in relation to the effect of the individual gap on the probability of a merger. It can be said that the probability of a merger financed by stocks is heavily predicted by the Mean Tobin’s Q (marginal effect of 2,22%) and is not predicted by the individual gap (marginal effect of 0,24% and not significant). While, the probability of a merger financed by cash, is heavily predicted by the individual gap (marginal effect of 1,15%) and the magnitude of the effect of the Mean Tobin’s Q is diminished (marginal effect of 0,41%).
In the last research, this paper made an attempt to find a second anchor for the targets in form of the 1-year high of the market index. From the results there is no significant evidence that the market gap is related with the timing of a merger in the sample of this research.
The aforementioned results support the reference point of view in mergers, conducted by Baker et al. (2012). These findings can in the future be applied to other aspects and circumstances. Future research can in the first place further investigate if other anchors or reference points might have effect on the probability of a merger. Thereafter, this research can easily be expanded to other countries in addition to the United States, in order to test if the results are consistent. Additionally, the results provided by this paper are constrained for a certain time period (1984-2013). By
expanding the time period and consequently increasing the number of observations will significantly amplify the validity of the results (for instance, Baker et al., (2012) obtained their data from 1974).
It is understandable that no single theory will entirely explain the existents of merger activity and merger clustering: nonetheless, the reference point of view theory will fill in some of the open questions.
