different during a financial crisis?
Master Thesis
January 2018
Name: Thomas Colenbrander
Student number: S2370778
Supervisor: Victoria Purice
Double Degree Master: MSc International Financial Management Rijksuniversiteit Groningen
MSc Business and Economics
Uppsala Universität
Abstract
This thesis researches what effect the economic crisis of ’08-’09 had on the returns of M&A deals, and whether the type of diversification in the deal affected this relationship. In order to do this, a sample of 898 M&A deals announced between 2006 and 2011 is constructed. With this data an event study is executed, followed by an OLS regression. Both the event study and the regression find a significant negative effect of the crisis period on the abnormal returns of M&A deals, while no significant moderating effect of both cross-border and industrial diversification is found. The results show that it is important for other research on M&A returns to look at the state of the economic cycle at the point in time the M&A deal was announced, as this could influence the outcomes of these studies. Furthermore, the results should discourage firms to participate in M&A deals during crisis periods, as these will be value destroying compared to the same deals during non-crisis periods.
Field Key Words: mergers & acquisitions, financial crisis, cross-border diversification,
industrial diversification, event study, regression analysis.
1. Introduction
Does it make sense to have so much research on Mergers and acquisitions (M&A)? As the topic is complex and the results are contradicting, generalizability seems to be a problem. However, M&A is a hot topic. This is evident looking at the amount of M&A deals completed over the years. While only 2675 deal transactions were completed in 1985, they peaked in 2016 with 49,039 transactions (IMAA, n.d.). Firms nowadays see M&A deals as very good investments because of the possibility of capturing the synergies between the acquirer and target firm, and because of the possibility of increasing market power (Andrade, Mitchell and Stafford, 2001).
Because of this increased interest by firms, and the corresponding rise in amount of M&A deals, M&A is also a hot topic in research. M&A decisions are one of the main big strategic decisions firms have to make nowadays, and having a lot of different research could help to gain a greater understanding on these deals. This should in the end lead to more generalizable results, which is where this thesis wants to contribute.
M&A deals are very complex, as a lot of factors come into play before the deal can be completed. Both parties in the deal, the acquirer and the target company, want to create the best deal for themselves, and they have to interact with their environment while creating this deal.
The complexity of the deals also makes research on M&A rather difficult, which results in a dissimilarity in conclusions regarding M&A returns. When looking at 41 different articles about M&A announcement returns of the acquiring firm, 13 articles reported a negative return, 17 articles reported a positive return and 14 articles reported an insignificant return (Bruner, 2002).
This shows that a lot of factors have to be dealt with when trying to research M&A returns, as these differences in reported results could be caused by a difference in deal characteristics, environments or other situations during these deals.
There are many deal characteristics of M&A influencing returns, like the ownership status of the target, the deal value and the usage of cash or equity in payment of the deal (Fuller, Netter and Stegemoller, 2002; Moeller, Schlingemann and Stulz, 2004; Masulis, Wang and Xie, 2007).
Another example of this is the diversification effect. Theory shows that both cross-border diversification as well as industrial diversification have a negative effect on the M&A returns, due to an increase in agency costs (Berger and Ofek, 1995; Denis, Denis and Yost, 2002;
Moeller and Schlingemann, 2005).
Next to the deal characteristics, the environment can also influence the M&A returns. One of
the most far-reaching changes in the economic environment is an economic crisis, like the credit
crunch in ’08-’09. As M&A deals have to interact with their environment, it can be expected
that such a big economic crisis should influence M&A deals. This is supported by literature, as research shows that less M&A deals happened during the ’08-‘09 economic crisis, because of firms having less money to invest (Grave, Vardiabasis and Yavas, 2012). Research also shows that the return on M&A deals in the banking sector declined around announcements during the
’08-‘09 financial crisis, due to the extra systematic uncertainty during this crisis (Beltratti and Paladino, 2013). Even though the banking sector has very different operations compared to all other industries, this result is not yet widely tested in literature for other sectors. However, this is important to research, as Beltratti and Paladino (2013) explain: “If M&As carried out during normal times behave differently than M&As carried out during crises, then empirical studies should be careful in considering all observations together”. In other words, could crisis periods be part of the reason for the dissimilarity in the results of the literature?
Besides, not only the returns themselves could be affected by a crisis, as crises can affect other elements of M&A deals too. Research found that there was a higher premium in excess value associated with cross-border diversified firms during the financial crisis (Chang, Kogut and Yang, 2016). Furthermore, research also found that multi-segment firms were valued higher during the financial crisis than outside of the crisis, as these industrial diversified firms had lower risk (Grave et al., 2012). Even though theory showed earlier concluded that diversification lowered the returns of M&A deals, the so-called diversification discount, the lower risk of these kind of deals might turn that discount into a premium during a crisis. This could lead to an extra inconsistency in the results.
All in all, this research wants to contribute to the literature of M&A activities, by trying to create a deeper understanding of the fluctuation in acquirer’s abnormal returns. As Beltratti and Paladino (2013) explained, crisis and non-crisis periods taken together could bias the results, if the returns differ between these periods. This would hold for a difference in other variables affecting M&A returns as well, like the effect of diversification. Therefore, this thesis would like to answer the following research question: How does the financial crisis of ‘08-‘09 influence the acquirer’s abnormal return in M&A deals, and what effect do cross-border and industrial diversification have on this?
The main results of this thesis, consisting of an event study and an OLS regression, show that
the abnormal returns of M&A deals during crisis times are indeed significantly lower than the
returns of deals taking place outside of the financial crisis. Furthermore, no significant
moderating effect of both cross-border diversification as well as industrial diversification is
found. These results hold at the robustness analysis.
The rest of this thesis is structured as follows. Section 2 will cover the theoretical background, which will summarize the findings of literature on the important topics concerning the research question, and state the consequently constructed hypotheses. Section 3 will contain the overview of the data and the explanation of the methodology, which will explain the two models executed to test the research question. Section 4 will cover the results of both models, and the discussion of these results. This is followed by the conclusions of this thesis in section 5, after which the sixth and final section will explain the limitations of the research and state recommendations for further research.
2. Theoretical background
This section reviews the existing literature on M&A returns, diversification and the effects of economic crises on these two. By using the findings of these papers, the hypotheses are constructed.
2.1 Motives for M&A deals
This theoretical background starts with a short introduction into mergers and acquisitions, showing the increase in interest of M&A deals and explaining the motives for firms to take part in these deals. In mergers and acquisitions, two firms are joining each other to start a cooperation. While a merger is most of the times a deal where the size of the two firms is almost equal, acquisitions have one firm that is a substantially bigger than the other (Goergen and Renneboog, 2004). As already shown in the introduction, M&A is a very hot topic nowadays.
This is evident looking at the growth in amount of M&A deals over the years. While there were only 2000 deals at the start of the 1980’s, this amount already grew to 30,000 deals in 1998.
Moreover, the amount of M&A deals keeps increasing, as it peaked in 2016 with 49,078 deals (IMAA, n.d.). What causes this immense interest in M&A deals?
First of all, firms see M&A deals as very good investment opportunities. According to both Mukherjee, Kiymaz, and Baker (2004) and Kiymaz and Baker (2008), the most important reason why firms think M&A is a good investment, is the possibility for the acquirer to capture the synergies between itself and the target, for example by creating bigger economies of scale.
Another reason is the possibility to increase market power, like by taking over competitors to create oligopolies or monopolies (Andrade et al, 2001).
Market power does also play a role in the idea of Öberg and Holtström (2006), who explains
that M&A deals might be contagious. They find that in a supply chain, whenever a customer
undertakes an M&A deal, the supplier does the same not much later. This phenomenon is also
found the other way around, indicating that the goal might be to restore the power balance that was before. If a firm loses bargaining power, this could hurt the performance of that firm.
Another motive for M&A can be the personal motives of the management. The manager could for instance want to manage the biggest firm possible, which can be accomplished through M&A deals. This is called empire building (Jensen, 1986). Another personal interest might be lowering the risk of bankruptcy through M&A, which increases job security (Berger and Ofek, 1995). These personal motives do also play a role in the effect of diversification, which will be covered later on this section.
There is a lot of theory about why firms participate in M&A deals, but, except for the personal motives of the management, these theories always begin with the motive to perform better as a firm after the deal. However, are M&A deals doing well in creating value for the acquirer? And what effects could influence this performance? That will be the main topic of interest in the rest of this thesis.
2.2 M&A performance during non-crisis periods
In order to be able to compare M&A performance between crisis and non-crisis periods, it is
important to know how M&A returns behave during non-crisis periods first. Most articles use
the abnormal returns in their research when looking at the performance of M&A deals. The
abnormal return is the return earned on top of the normally expected return, and therefore
caused by the announcement of the deal. Research shows that in less distressed economic times,
the acquirer’s abnormal return is insignificantly different from zero. This is caused by the
premium paid by the acquirer, in combination with potential positive synergies gained by the
acquirer. (Shah and Arora, 2014). This view is not shared by everyone, as Moeller,
Schlingemann and Stulz (2005) find a negative abnormal return for the acquirers in their
sample, just as Kaplan and Weisbach (1992). On the other hand, Cai, Song and Walking (2011)
conclude that acquirer’s returns are on average positive. This result is supported by Asquith et
al. (1983) as well as by Georgen and Renneboog (2004), who also find positive abnormal
returns for the acquirer. All in all, there is a great diversity in results looking at M&A acquirer
returns. This dissimilarity is researched by Bruner (2002). In an overview of theory looking at
the acquirer’s abnormal return it is found that from the 41 studies reviewed, 13 studies showed
negative returns, 17 studies showed positive returns and 14 studies showed returns insignificant
from zero. These results show that M&A deals are complex and diverse, as already discussed
in the introduction. Altogether it is assumed in this thesis that during non-crisis times, the
acquirer of the M&A deal has on average an abnormal return around zero. Potential positive or
negative returns are then caused by other factors influencing the M&A deal, like for instance diversification.
2.3 M&A returns during a financial crisis
The financial crisis of ’08-’09 had a large impact on many economies. The risk of default for firms grew, increasing the need for reserves. In the meantime, banks were also performing worse, decreasing the amount of loans firms were able to get. This would normally cut the budget for M&A deals, which is generally not a core activity of a firm. This idea is supported in literature, which finds that during more difficult economic times, the deal price and the amount of M&A deals dropped. This effect holds the other way around for more favourable economic times too (Grave et al., 2012). That does not directly mean that the returns of these M&A deals during the crisis reacted the same way, or reacted differently at all during the financial crisis. However, it shows that crisis times have an influence on M&A deals in general.
Besides, there are reasons to believe that M&A returns are different during the financial crisis.
Several factors might influence the returns in crisis times, for instance the increased systematic uncertainty in crisis times or lower prices that have to be paid for firms in financial distress. In the banking sector, the extra systematic uncertainty turned out to result in lower acquirer’s abnormal returns at announcement date during the financial crisis of ’08-’09. As the outcomes of M&A deals are by themselves already uncertain, these deals turned out to be less attractive during the hard times banks had during the crisis (Beltratti and Paladino, 2013). This might be an indication that the ’08-’09 financial crisis had a negative effect on announcement returns in general, even though one has to keep in mind that the banking sector is fundamentally different than any other sector. In addition, research shows that during so-called “hot” merger periods, the announcement returns are higher. These hot merger periods are periods when the stock markets are doing better, a period during the peak of an economic cycle (Rosen, 2006). It is likely that this would also hold for the lows of an economic cycle, the financial crisis. However, what happens with the announcement returns during this financial crisis is not yet widely researched in literature for all sectors.
The negative effect during crisis times is inconsistent with the view of Hotchkiss and Mooradian
(1998), who claim that that taking over financially distressed firms is at such a discount, that it
has a positive outcome for the acquirer, resulting in a positive return. This could also have been
the case in the financial crisis, when a lot of firms were in financial distress. However, other
research shows that these cheap sales of a firm in distress do not actually exist, as the premium
paid for a distressed firm during the crisis is not different from the premium paid for a firm not
in distress (Ang and Mauck, 2011). Consequently, taking over firms in distress cannot increase the announcement returns of M&A deals during the financial crisis.
Concluding on the acquirer’s abnormal returns of M&A deals, theory is not completely clear on the results. During non-crisis periods, it seems that there are no significant abnormal returns for the acquirers, even though there is a dissimilarity in results. During a financial crisis, the M&A returns are not known for certain, since it has not been widely researched in literature yet for all sectors. Nevertheless, looking at other articles, theory points towards more negative acquirer’s announcement returns during the economic crisis, because of the increased systematic risk in crisis times. This results in hypothesis 1:
H1: The acquirer’s announcement returns during a financial crisis are lower than these returns during non-crisis periods.
2.4 M&A deals and diversification during non-crisis periods
Over the last decades, the M&A landscape changed from simple same industry and same country takeovers, to global and industrial diversified ones (Grave et al., 2012). This can be partly explained by the immense evolution of IT, through which communicating with and getting information about different firms throughout the whole world became within reach.
Question remains what the effect of diversification in M&A deals is on the returns of these deals.
During non-crisis periods, it is assumed in literature that diversifying M&A deals destroy value.
This “diversification discount” is mainly caused by the agency problems between management and local departments and between management and shareholders. Agency problems are a conflict of interest between two groups within a firm, creating internal costs for the company.
Diversification increases these agency costs between management and departments, as it will
become more difficult to align interests with departments operating in different industries or
different countries. (Laeven and Levine, 2007). Furthermore, diversification also increases the
agency costs between management and shareholders. This is shown by Morck, Shleifer and
Vishny (1990), who find that the loss of the acquirer’s firm value in diversifying deals is caused
by personal objectives of management. Examples of these personal objectives are diversifying
the risk of the firm going bankrupt through buying a firm from a different industry or country,
increasing job security. Empire building, wanting to create a bigger firm to manage, is also a
personal objective which can be achieved by diversifying M&A deals. These personal
objectives lead to diversifying deals having more problems of overinvestment, investing in projects that do not have a positive value (Berger and Ofek, 1995; Denis et al., 2002).
Overinvestments decreases firm value, as these projects will only cost resources. As theory finds a value destruction caused by diversification, this diversification-discount during non- crisis periods is also expected in this research.
A value-destroying effect of diversification should normally imply a negative announcement return for the acquirer as well. This idea is supported by Moeller and Schlingemann (2005), who found that both global and industrial diversifying M&A deals have significant lower announcement returns than related M&A deals. These results are supported by Freund, Trahan and Vasudevan (2007), who also find that diversification has a negative effect on the announcement returns of M&A deals. All in all it is widely concluded in literature that, at least during non-crisis periods, both industrial and global diversification result in lower firm value.
Consequently, theory shows that this also results in lower acquirer’s abnormal returns around the announcement date of M&A deals.
2.5 The effect of diversification during a financial crisis
M&A returns during crisis periods are not widely researched in the literature yet. Likewise, this holds for the results of diversified M&A deals during a crisis. However there is reason to believe that this effect is different than the diversification discount during non-crisis periods.
The negative relationship between the crisis and the cumulative abnormal returns of M&A deals explained above is mainly based on the added uncertainty and risk that is caused by this financial crisis. Diversification lowers both actual and perceived risk as it creates a collection of assets and cash flows in different, normally not perfectly correlated, countries or industries (Shapiro, 1978; Denis et al., 2002). This could lead to an increase in value and consequently in returns of diversified deals in crisis times, as these deals should lower the risk that is systematically higher due to the economic circumstances. However, if every country and industry is hit by the financial crisis, the effect of diversification diminishes, even though one industry or country shall normally be hit harder than another.
This being said, theory still points towards a positive effect of diversification on firm value during the last financial crisis. As diversified firms have better access to credit markets and can use their internal markets more efficiently, they are able to lower their risk. This resulted in a significant increase in value of industrial diversified firms compared to single-industry firms.
Furthermore, these multi-segment firms are able to have a higher leverage, as the they are a
safer bet for banks, certainly during an economic crisis (Kuppuswamy and Villalonga, 2010).
The possibility of having a higher leverage is also confirmed in literature for a less recent crisis, which was the Asian crisis of 1997-1998 (Gerlach, Wilson and Zurbruegg, 2006). Not only industrial diversification seems to increase value during an economic crisis, as Chang, Kogut and Yang (2016) found this effect for cross-border diversification as well. They found a significant higher premium in excess value associated with cross-border diversified firms during the financial crisis. This was caused by an increased interest in flexibility during periods of high volatility, such as the financial crisis period. Taking into account the results of the effects of diversification during non-crisis periods, where the value-destructing effect of diversification also lead to lower abnormal returns, it can be expected that the abnormal returns of diversifying deals should be higher during crisis times, as the diversification increases firm value during the crisis-period. This implicates that during crises there would actually be a
“diversification premium”, a positive moderating effect of diversification on the acquirer’s abnormal returns. This leads to hypotheses 2 and 3:
H2: Cross-border diversification has a positive moderating effect on the abnormal M&A returns during crisis times.
H3: : Industrial diversification has a positive moderating effect on the abnormal M&A returns during crisis times.
3. Data and Methodology
This section describes the way the research is executed, what variables are included in the research, and how these variables are constructed. After an introduction into the sample and its data, both the event study methodology and OLS regression methodology are explained.
3.1. The sample
The data on the M&A deals are taken from the Zephyr database by Bureau van Dijk. Other data
needed on the acquirers in the deals is taken from the Thomson Reuters Datastream. In order to
analyse the impact of the crisis on M&A returns, the announcement date of the deals has to be
between 2006-2011, which is approximately 2 years before until 2 years after the financial crisis
of ‘08-’09. The crisis itself is determined by the American national bureau of economic research
to be starting in America in December 2007, and ending in June 2009 (NBER, n.d.). These
dates are also used in this research as the crisis period.
The sample consist of North American acquirers, meaning they are originating from the US and Canada. This part of the world is chosen as it has one of the biggest economic markets.
Europe also has a big economic market, but the European market had a second financial crisis around 2011, which would have biased the returns and corresponding results in the time period after the crisis.
As some of the data needed in the research is normally not disclosed in unlisted firms, only deals with a listed acquirer are used, whereas the target can be both listed and unlisted.
Furthermore, in order to increase the quality of the results, only deals with a substantial impact on the acquirer are wanted in the sample. That is why the M&A deals need a deal value of over 1 million euros and should be fully completed during the set time period between 2006 and 2011. Next to this, only complete takeovers, meaning that the acquirer gets a 100% stake in the target’s firm, are included. This discards deals in which just a small part of the target is bought and it discards mergers, as both have a smaller impact on the acquirer. In accordance with Denis et al. (2002), the utility, financial and government related industries are also discarded (SIC 4900-4999, SIC 6000-6999 and SIC 9111-9999). These industries have significantly different operations, which would normally result in different, potentially contradicting outcomes.
Lastly, as it is preferable to create as unbiased results as possible, all deals from acquirers that have another deal in the sample that takes place during the estimation period, are discarded.
These estimation periods are biased by the earlier executed M&A deal, resulting in biased parameters creating biased outcomes. After applying all these constraints, 898 deals end up in the sample.
3.2 Descriptive statistics
Table 1 shows the descriptive statistics of the data. According to the crisis period variable, the
sample has 201 deals during the financial crisis, enough for this sample to research the
hypotheses. Looking at the distance between acquirer and target, which is used for constructing
the cross-border diversification variable, it is important that the minimum and maximum are
very different from each other, in combination with a substantial standard deviation. More
diversity in the distance should help in testing for the cross-border diversification moderating
effect. Additionally, the cross-border dummy shows that 265 deals are classified as cross-border
deal, which is also sufficient for testing this moderating effect.
Table 1
Descriptive statistics of the data
Characteristics of the 898 deals in the sample. All the variables that will be implemented in the regression are included, except for the dependent variable from which the outcomes will be shown in the results section. From the dummy variables, the amount of deals classified with a value of 1, meaning the variable is present in that deal, are also stated, in order to create insights in the diversity of the data.
Mean Median Maximum Minimum Standard deviation
Amount classified
Crisis period 0.2238 0 1 0 0.417 201
Distance between acquirer and target (KM) 2525 1372 18175 0 1.1455 -
Cross-border diversification dummy 0.2951 0 1 0 0.4563 265
Industrial diversification 0.4143 0 1 0 0.4929 372
Acquirer's size (mln $) 13.2351 30.71 276 0.0211 315.934 -
Acquirer's leverage 0.7243 0.381 72.9182 -30.0465 3.7754 -
Tobin's Q 2.1652 1.7377 43.8569 0.4214 1.8812 -
Free cashflow (mln $) -0.2494 -0.1542 42.676 -29.96 3.74352 -
Target ownership status 0.0033 0 1 0 0.0577 3
Deal size (mln $) 791.7604 117.75 51000.4 1.36 3160 -
Cash deal 0.8441 1 1 0 0.363 758
Equity deal 0.255 0 1 0 0.4361 229
The industrial diversification dummy based on the two-digit SIC code shows there are 372 deals
classified as industrial diversifying, enough for testing the hypothesis on the industrial
diversification moderating effect. Comparing acquirer’s size with deal size, it is interesting to
find deal size having a higher mean as well as higher median. This indicates that, on average,
firms in this sample do pay more for their M&A deals than they own as assets in this sample,
which normally entails a high amount of risk. The highly negative minimum leverage could be
surprising. However, sometimes firms can have a slightly negative equity, for instance when
the liabilities are bigger than the assets. These negative values shall often be very small,
otherwise the firm will become bankrupt. As leverage is calculated by debt divided by equity,
this will end up in highly negative leverage values for the acquirer. Furthermore, only three of
the targets are listed, which is a rather low amount. Free cashflow turns out to have a negative
mean and median, which is noteworthy, as this means that on average firms in this sample
would have a negative financial performance. Normally these firms with a negative
performance are not the ones that would be expected to participate in M&A deals. This certainly
applies to the firm with the lowest free cashflow amount of -30 million dollars. A possibility
explaining this (highly) negative amount would be very high capital expenditures in the year of
the M&A deal, as these are extracted from the EBIT to calculate the free cashflow. Finally, it
is remarkable that the amount of deals paid (partly) in cash is very high, certainly compared
with the amount of deals paid (partly) in equity. A reason for this could be the negative effect
of using equity and the positive effect of using cash found in theory (Travlos, 1987; Walker, 2000). This will be covered more in depth later on at the explanation of both variables.
Table 2
Amount of deals per time period
Overview of the amount of deals in this sample that took place during the three different time periods important for this research. This data gives an insight in the distribution of the deals amongst the different time periods.
Time period Amount
Before crisis
(January 2006 – November 2007)
316 Crisis
(December 2007 – June 2009)
201 After crisis
(July 2009 – December 2011)
381
Total 898
Table 2 shows that the amount of deals per time period is not completely evenly matched. This result is not remarkable, as theory found that less M&A deals happened during the crisis period (Grave et al., 2012). Besides, the period after the crisis is somewhat longer, which consequently results in a higher amount of deals after the crisis. Nevertheless, the distribution is sufficient to test the hypotheses.
3.3 Event study and OLS regression methodology
To evaluate whether a merger is successful or not, the cumulative abnormal returns are used.
These are the sum of the returns on the stocks of the acquiring firm around the announcement date of the M&A deal. To research these cumulative abnormal returns, there are two frequently used methods, which are the event study and the OLS regression. As both methods have their advantages and disadvantages, which is why both will be used in this thesis.
First, an event study is constructed according to the model of MacKinlay (1997) to solely research the differences in abnormal returns of the M&A deals during the different time periods.
The model of MacKinley is chosen as this is widely used for constructing event studies researching announcement returns of M&A deals (Brunner, 2002; Corrado, 2011; Shah and Arora, 2014). The advantage of this model is that it purely researches the differences of the returns itself during changing events, in this case the financial crisis. The disadvantage is that one cannot add control variables, as it is a univariate model. This means that for instance deal values or other control variables could also explain part of the difference in the abnormal returns, if these control variables would be fundamentally different during the crisis period.
However, the event study would not be able to find this. That is why in a second step after this
event study, a regression analysis is constructed to control for the impact of this. Furthermore,
as the event study model of MacKinley is a univariate model, it is also unable to research the moderating effects of both cross-border and industrial diversification. Therefore, the regression analysis is also used to test the hypotheses of these two variables.
3.3.1 Event study methodology
As explained above, this research starts with the even study model created by MacKinlay (1997), which researches the effect of the crisis period on the abnormal returns. Two tests are executed, one in which the crisis and non-crisis periods are researched, and one in which the pre-crisis and post-crisis period are also split up, to research the three periods separately. This second test is executed in order to research whether a potential significant outcome in the first test might be caused by big differences between the crisis period and for instance the pre-crisis period. If this is the case, it would not be the effects of the crisis which causes the significant effect in the first test.
To be able to execute both tests, the independent variable, crisis period, will have two versions.
These are the dummy variable version and the nominal version. Both versions are coded according to the announcement date of the M&A deal, where the crisis period is taken from the American national bureau of economic research. The dummy variable is used in the first test.
In this version deals announced between 2006 and November 2007 and between July 2009 and 2011 will take value 0, meaning there was no crisis at the announcement date. The deals announced between December 2007 and June 2009 will have the value of 1, meaning there was a crisis at the announcement date. The nominal variable is used in the second test. In this variable the first time period between January 2006 and November 2007 will take value 0, the crisis period between December 2007 and June 2009 stays value 1, and the period after the crisis between July 2009 and December 2011 becomes value 2. This way it is possible to distinguish the pre-crisis and post-crisis M&A deals, and compare all three different time periods with each other.
The dependent variable, the cumulative abnormal return, is calculated using the event study method described by MacKinlay (1997). First, the abnormal return (𝐴𝑅
𝑖𝑡) is needed, which can be calculated by:
𝐴𝑅
𝑖𝑡= 𝑅
𝑖𝑡− 𝐸(𝑅
𝑖𝑡) (1)
where 𝑅
𝑖𝑡is the return of firm i at time t and 𝐸(𝑅
𝑖𝑡) is the expected return of this firm i at this
time t.
To calculate the expected return of the firms, a market model is used on the time period prior to the announcement date, which is called the estimation window. The different time periods in a market model method are stated in figure 1.
Figure 1
The different time periods in the abnormal returns calculation
Figure showing the two different time periods used in calculating the abnormal returns of the M&A deals. Time t at 0 is the announcement date of the M&A deal. Figure from MacKinley (1997).
The market model method for calculating abnormal returns compares the return of the market and the return of the firm during the estimation window. In this research, this estimation window is 150 days until 10 days before the announcement date. The amount of days used in this window is tested by a sensitivity analysis, which can be found in graph 1 in the appendix.
This sensitivity analysis is executed to make sure the results are not biased by the chosen length of the estimation window.
With the market model method, the Alpha and the Beta can be constructed for that particular firm, which are parameters showing the exact relationship between the return of the firm and the market. These parameters are used to calculate the expected return around the announcement date, which is called the event window. This is calculated through:
𝐸(𝑅
𝑖𝑡) = 𝛼̂ + 𝛽̂𝑅
𝑚𝑡(2)
where 𝑅
𝑚𝑡is the return of the market, and 𝛼 and 𝛽̂ are the parameters constructed. This research uses the S&P500 firms’ stock price return as market return, which is a widely used benchmark for US firms.
The event window in figure 1 is the amount of days around the announcement date used in the calculation. This research uses one main event window, which is the seven-day period [-3;+3].
As some deals can leak before the announcement date, the seven-day period is used to make sure that this leakage is still included in the research. Just as in the regression later on, the event study will also use the three-day period [-1+1] as a robustness test.
In order to get results in this univariate model, one needs the average and variance of the
abnormal returns. The average at one period in time can be calculated by equation (3):
AR ̅̅̅̅
𝑡= 1
𝑁 ∑ 𝐴𝑅
𝑖𝑡𝑁
𝑖=1
(3)
Where N is the amount of deals in the sample. However, the event study will not use the abnormal returns of separate days in the event window, but the returns of the event window as a whole. The abnormal returns of the whole event window are called the cumulative abnormal returns (CAR).
The average of the CAR for all firms in the population is calculated by adding up all the averages of the abnormal returns during the different days of the event window, which is shown in equation (4):
𝐶𝐴𝑅 ̅̅̅̅̅̅
(𝑇1,𝑇2)
= ∑ 𝐴𝑅
𝑇1̅̅̅̅
𝑡𝑇2
(4)
where T
1and T
2are the start and end date of the event window, also shown in figure 1.
Next to the average, the variance of the abnormal return is needed. MacKinlay (1997) uses the error term of the market model calculations in the estimation window to calculate this. The calculation of the variance of this error term is showed in equation (5):
𝜎
𝜀𝑖2= 1
𝐿
1− 2 ∑ (𝑅
𝑖𝜏− 𝛼̂
𝑖− 𝛽̂
𝑖𝑅
𝑚𝑡)
2𝑇1
𝑡 = 𝑇0+1
(5)
where 𝜎
𝜀𝑖2is the variance of the error term of a particular firm i, 𝐿
1is the length of the estimation window, T
0and T
1are the start and end of the estimation window, 𝑅
𝑖𝜏is the return of a particular firm i on a particular day t, 𝛼̂
𝑖and 𝛽̂
𝑖are the parameters of a particular firm i calculated in the estimation window and 𝑅
𝑚𝑡is the market return.
This variance of the error term 𝜀
𝑖𝑡is then used to calculate the variance of the average of the abnormal returns of all deals in the sample at a certain point in time using equation (6):
𝑣𝑎𝑟 (AR ̅̅̅̅
𝑡) = 1
𝑁
2∑ 𝜎
𝜀𝑖2𝑁
𝑖=1
(6)
where N is the amount of deals, and AR ̅̅̅̅
𝑡is the average abnormal return of all the deals at time t. Just as with the average of the abnormal returns, not the variance of the average abnormal returns of separate days in the event window is needed, but the variance of the average CAR. This is calculated in equation 7:
𝑣𝑎𝑟 (𝐶𝐴𝑅 ̅̅̅̅̅̅
(𝑇1,𝑇2)
) = ∑ 𝑣𝑎𝑟(𝐴𝑅 ̅̅̅̅
𝑡)
𝑇1
𝑇2
(7)
where T
1and T
2are the start and end date of the event window.
To draw conclusions on whether the CAR is different from zero, a t-statistic is needed, which is calculated by using equation (8):
Θ
1= 𝐶𝐴𝑅 ̅̅̅̅̅̅
(𝑇1,𝑇2)
√ 𝑣𝑎𝑟 (𝐶𝐴𝑅 ̅̅̅̅̅̅
(𝑇1,𝑇2)
)
(8)
where Θ
1is the t-statistic. However, a CAR significant from zero in different time periods does not mean that there is a significant difference between the different time periods. To research whether two groups are significantly different from each other, a Welch’s test is executed, shown in equation (9):
Θ
2= 𝐶𝐴𝑅 ̅̅̅̅̅̅
1− 𝐶𝐴𝑅 ̅̅̅̅̅̅
2√ 𝜎
12𝑁
1+ 𝜎
22𝑁
2(9)
where Θ
2is the t-statistic of the Welch’s test, 𝐶𝐴𝑅 ̅̅̅̅̅̅
1and 𝐶𝐴𝑅 ̅̅̅̅̅̅
2are the averages of the CAR, 𝜎
12and 𝜎
22the variances of the CAR and 𝑁
1and 𝑁
2the amount of deals of the two different time periods respectively. This t-statistic would then indicate whether there is a significant difference in CAR in the two different time periods compared.
3.4 Regression analysis methodology
After constructing this event study, an OLS regression will elaborate on the effect found in the event study, by controlling for other variables that could have also influenced the abnormal returns. These variables are added as control variables. Furthermore, it is possible to test the moderating effect of both cross-border and industrial diversification with the OLS regression method, something which is not possible with the event study method.
In the OLS regression model, a couple of regressions are run to test the research question and hypotheses stated earlier. The moderators are implemented one by one to make sure these do not influence each other or filter out effects. The final and most extensive model is:
𝐶𝐴𝑅
𝑖= 𝛽
0+ 𝛽
1𝐶𝑅𝐼𝑆𝐼𝑆 + 𝛽
2𝐶𝐵
𝐷𝐼𝑉+ 𝛽
3(𝐶𝑅𝐼𝑆𝐼𝑆𝑥𝐶𝐵
𝐷𝐼𝑉) + 𝛽
4𝐼𝑁𝐷
𝐷𝐼𝑉+ 𝛽
5(𝐶𝑅𝐼𝑆𝐼𝑆𝑥𝐼𝑁𝐷
𝐷𝐼𝑉) + 𝛽
6𝐴𝐶𝑄𝑆𝐼𝑍𝐸 + 𝛽
7𝐿𝐸𝑉 + 𝛽
8𝑇𝑂𝐵𝑄 + 𝛽
9𝐹𝐶𝐹 + 𝛽
10𝑂𝑊𝑁𝐸𝑅𝑆𝐻𝐼𝑃 + 𝛽
11𝐷𝐸𝐴𝐿𝑆𝐼𝑍𝐸 + 𝛽
12𝑃𝐴𝑌𝑀𝐸𝑇𝐻𝐶𝐴𝑆𝐻
+ 𝛽
13𝑃𝐴𝑌𝑀𝐸𝑇𝐻𝐸𝑄𝑈𝐼𝑇𝑌 + 𝛽
14𝐼𝑁𝐷_𝐶𝑂𝑁𝑇𝑅𝑂𝐿 + 𝜖
𝑖All these variables shall now be explained one by one. The dependent variable, the cumulative
abnormal return of the deals, is calculated by adding up all the abnormal returns of one firm
during the event window. The calculation of the abnormal returns (AR) is showed in equation (1), the calculation of the cumulative abnormal returns (CAR) used in the regression is present in equation (11):
𝐶𝐴𝑅
(𝑇1,𝑇2)= ∑ 𝐴𝑅
𝑖𝑡𝑇1
𝑇2
(11)
where T
1and T
2are the start and the end of the event window, which are also shown in figure 1. Just as in the event study, the seven-day period [-3+3] is used as main event window. Later on at the robustness analysis, the three-day period [-1+1] event window will be used.
In order to investigate the impact of the crisis on the abnormal returns of the M&A deals, the independent variable, crisis period, is regressed on the CAR. The crisis period variable is exactly the same as the dummy variable used in the first test of the event study, with a value of 0 for deals during non-crisis periods and a value of 1 for deals during the crisis period.
Since the diversification status of a deal can have impact on the abnormal returns and the effects of crisis periods on these, both cross-border and industrial diversification are included as moderating variables. The first moderator, cross-border diversification, is composed out of two different measures. This thesis uses a multiplication of the distance variable and the cross- border dummy. Firstly, the distance variable is used, which measures the distance between the acquirer’s and target’s home city. This variable is able to measure the amount of influence the cross-border deal has on the acquirer, as a bigger distance between the two home cities increases the strength the cross-border diversification effect has. However, this distance variable is not able to measure whether a deal is actually a cross-border deal or not, as it does not take the countries of both firms into consideration. This is why the distance is multiplicated with the cross-border dummy. This dummy variable is used to check whether the target firm is from a different country than the acquiring firm. The dummy has a value of 1 when the target is from another country than the acquirer, which indicates cross-border diversification, and a value of 0 if the target is from the same country. The multiplication of both variables, called the dummy*distance variable in this thesis, will show a value of 0 when the merger is domestic, as the dummy variable takes a value of 0. When the deal is cross-border diversifying, the dummy variable has a value of 1, and the dummy*distance variable will show the value of the distance.
This distance*dummy cross border variable is constructed this way, to be able to reflect both
whether the deal is cross-border diversifying, and what the influence of the diversification effect
is. In the robustness analysis, a test only using the cross-border diversification variable is
performed as robustness check.
The second moderator, industrial diversification, only uses a dummy variable. The variable is constructed by looking at the two-digit standard industrial classifications (SIC) codes of the acquirer and target, and comparing these. This results in a dummy variable with a 0 when both codes are the same, meaning no industrial diversification takes place, and a 1 when both codes differ, meaning industrial diversification does take place. As the SIC codes are rather specific, some firms will be classified as industrial diversification as they have a different two-digit SIC code, while actually these firms are rather similar. To overcome this drawback, a robustness test is performed using the industry groups of these SIC codes. The industrial diversification dummy will then be coded according to whether or not the industry groups, shown in table 11 in the appendix, differ from each other.
Besides the variables of interest, the regression includes a set of control variables. In articles researching M&A returns, the control variables used are frequently grouped by acquirer characteristics and deal characteristics (Moeller et al., 2004; Moeller et al., 2005; Masulis, et al., 2007). The acquirer characteristics controlled for in this research are firm size, leverage, Tobin’s Q and free cashflow. The deal characteristics controlled for are target ownership status, deal size and payment method. In this research an industry control is also added.
Firm size is the first variable that is controlled for. Research shows that bigger acquirers have a lower announcement return of almost two percentage points, which is explained by the higher premiums paid by bigger acquirers (Moeller, Schlingemann and Stulz, 2004). As this result shows that acquirer size could bias the returns, the logarithm of total assets of the year of the announcement date is included as control variable.
Leverage is also added as control variable, as leverage gives incentives to managers to increase firm performance. The possibility of a manager losing control is higher with higher leverage, as this increases the chances of bankruptcy. This gives the firm and especially management incentives to increase its performance in M&A deals (Masulis et al., 2007). The book value of debt divided by the market value of equity is therefore included as control variable.
Tobin’s Q, which is the acquirer’s market value of assets divided by the acquirer’s book value
of assets, is also used as control variable. Research shows that acquirers with a higher Tobin’s
Q value generally execute better acquisitions, which would lead to a higher announcement
return. This is explained by showing that Tobin’s Q can be interpreted as a measure of
managerial performance, and a better managerial performance normally leads to higher returns
(Servaes, 1991). To calculate the market value of assets divided by the book value of assets, the
market values and book values of equity and liabilities in the year of the announcement date are used, as the sum of these two is equal to the total assets.
Free cashflow (FCF) is the last acquirer characteristic controlled for. When having a lot of FCF, managers have the resources for empire building. This agency cost leads to overinvestment, investing in projects that do not have a significantly positive value or do not have a positive value at all. This overinvestment results in a negative effect of FCF on the announcement returns, as overinvestments generally leads to investments in low-benefit or even value- destroying M&A deals (Jensen, 1986). The FCF is calculated by taking the EBIT (earnings before interest and taxes), adding up the depreciation and amortisation, and then extracting the capital expenditures of the year of the announcement date. This number has to be scaled to the total assets of the firm to make it comparable.
The first deal characteristic that is controlled for is whether the target firm is a private or a public firm. Research shows that in M&A deals with a public target firm, more of the announcement returns go to the target firm than to the acquiring firm. This is the other way around when buying private target firms. This has to do with the illiquidity of private firms, which makes acquirers pay a lower price, resulting in a higher return for the acquirer (Fuller et al., 2002). Furthermore, as private firms have more concentrated ownership, taking over these private firms creates bigger blockholders, which increases the capacity to monitor the firm. This increases the value of the target firm, and therefore the value of the M&A deal (Chang, 1998).
As this could bias the results, target ownership status is included as control variable.
The second deal characteristic controlled for is deal size. Research shows that the abnormal return for the acquirer is positively related to the price paid for the target’s equity (Asquith, Bruner and Mullins, 1983). However, for a subsample of large acquirers, this effect turned out to be the other way around and became negative, because of the higher premiums that big firms have to pay (Moeller et al., 2004). Regardless of the way deal size influences abnormal returns, theory shows that there is a relationship which could bias the results. This is why the logarithm of the deal value is used as control variable.
The way the deal is paid by the acquirer can also have a significant influence on the
announcement returns. Research shows that equity financed deals have more negative
announcement returns than deals that are financed by cash. One of the explanations for this is
the signalling theory, which explains that the acquirer’s equity is overvalued at time of the deal,
otherwise the acquirer could better use cash to complete the deal. The shareholders could take
this as a sign to sell their stocks, resulting in a drop of stock price and consequently a lower
announcement return (Travlos, 1987). This also has to do with the adverse selection problem, where in this case the acquirer has more knowledge about its own stocks than the target firm has, even though the target will be paid in these stocks. Usage of cash leads to a higher return for the acquiring firm, again due to the adverse selection problem and the signalling theory.
Usage of cash might then indicate that the shares of the acquiring firm are undervalued, and therefore the acquirer rather wants to use cash instead of equity (Walker, 2000).
Table 3
Overview of all the variables included in the regression model
Explanation of all variables included in the OLS regression, shortly stating what they measure and how they are constructed.
CAR Cumulative abnormal returns of the acquirer, which is the abnormal return summed up over the seven-day period [-3+3] around the announcement date.
Crisis period Dummy variable whether or not the announcement date was during the economic crisis; 1
= crisis period, 0 = non-crisis period.
Cross-border diversification Multiplication of the distance variable, measuring the distance between acquirer’s and target’s home cities, and the cross-border dummy variable, measuring whether the acquirer is from a different country than the target. Variable gets value 0 when no cross- border diversification takes place, and shows the value of the distance variable when cross-border diversification does take place.
Industrial diversification Dummy variable whether the acquirer’s two-digit SIC code differs from the target’s code;
1 = industrial diversification, 0 = no industrial diversification.
Firm size The logarithm of the total assets of the acquiring firm.
Leverage The book value of debt divided by the market value of equity of the acquiring firm.
Tobin’s Q The market value of equity added with the market value of the liabilities of the acquirer, which is then divided by the book value of equity added up with the book value of liabilities of the acquirer.
Free cashflow The EBIT of the acquirer added up with the depreciation and amortisation, and extracted with the capital expenditures. This is scaled to the acquirer’s total assets to make it comparable with other deals.
Target ownership status Whether or not the target is a listed firm; 1 = listed target, 0 = unlisted target.
Deal size The logarithm of the amount of money paid for the target firm.
Payment method cash Whether or not the deal is paid with cash; 1 = cash used, 0 = no cash used.
Payment method equity Whether or not the deal is paid with equity; 1 = equity used, 0 = no equity used.
Industry control 48 dummy variables, 1 for every industry in the sample (except for one industry);
1 = acquirer’s SIC code same as dummy’s SIC, 0 = acquirer’s SIC code different from dummy’s SIC code.
To control for the way the deal is paid, the method of Masulis et al. (2007) is used, containing two dummy variables. The first dummy variable is for a cash deal, which becomes 1 for deals that are fully or partly paid in cash and 0 otherwise. The second dummy variable is for equity deals, which becomes 1 when the deal is fully or partly financed by equity, and 0 otherwise.
Lastly, industry controls are added. The industry controls consist of 48 industry dummies for each individual two-digit SIC codes but one, as including all SIC codes would create a perfect correlation. These control variables are used to control for industry specific changes, for instance a crisis in one of the industries of the sample. This could bias the results, while only one particular industry is actually hit. As the impact of a crisis can be significantly different between industries, it is necessary to control for these industry-specific effects.
An overview of all variables included in the regression is shown in table 3.
4. Results
This chapter contains the results of this thesis. First, the findings of both tests in the event study are shown. Next, a correlation researches whether there are problems of multicollinearity in the sample, before the OLS regressions can be presented.
4.1 Event study
The event study contains two parts. The first part researches the difference in average CAR between the crisis and non-crisis periods, while the second part looks at the difference between the pre-crisis, crisis and post-crisis periods.
Table 4
Difference in cumulative abnormal returns between crisis and non-crisis periods
Results of the first part of the event study, researching the differences between the crisis and non-crisis periods.
Panel A shows the t-test results whether or not the periods had a CAR significantly different from 0. Panel B shows the results of the Welch’s test, whether or not the CARs of the non-crisis and crisis periods are
significantly different from each other. P-values smaller than 0.01, 0.05 and 0.10 are indicated by ***, ** and *, respectively.
Panel A: crisis and non-crisis periods
Period N Event window CAR average t-statistic
Non-crisis 697 (-1+1) 0.219% 1.1370 *
697 (-3+3) 0.173% 0.7093
Crisis 201 (-1+1) -0.493% -1.3349 *
201 (-3+3) -0.898% -1.5915 *
Panel B: Difference between periods
N Event window CAR average t-value
Difference 898 (-1+1) 0.712% -1.7687 **
898 (-3+3) 1.072% -2.7423 **
Panel A of table 4 shows whether the average of the CAR of both the non-crisis and crisis periods is significantly different from zero. The non-crisis deals have a significantly positive CAR at a 10% level at the three-day event window, with an average of 0.219%. At the seven- day window, this CAR is insignificant from zero. The crisis deals have a significantly negative average CAR at a 10% level at the three-day window, with an average of -00493%. This average is even more negative at the seven-day period, with a value of -0.898%, also significant at a 10% level. These results seem to point towards a significant difference in average CAR between crisis and non-crisis times, which the Welch’s test has to confirm.
The results of this Welch’s test are listed in panel B of table 5. At the three-day window the test shows a lower average CAR of 0.712% for deals during crisis times, significant at a 5%
significance level. At the seven-day window, this difference in average CAR grows to 1.072%
which is also significant at a 5% level. All in all, the Welch’s test in table 4 finds that there are significant lower M&A abnormal returns during crisis times compared to non-crisis times.
It is important to know whether the results change when looking at the three time periods separately. The findings of the Welch’s test in table 4 could for instance be caused by a big difference between the CARs of only one of the non-crisis periods and the crisis period. If this is the case, it would not be the crisis that causes the significant differences in abnormal returns.
Table 5
Difference in cumulative abnormal returns before, during and after the crisis
Results of the second part of the event study, researching the differences between the pre-crisis, crisis and post- crisis periods. Panel A shows the t-test results whether or not these three periods had a CAR significantly different from 0. Panel B shows the results of the Welch’s test, whether or not the CARs of the periods are significantly different from each other. This Welch’s test is executed three times, in pairs of two time periods each. P-values smaller than 0.01, 0.05 and 0.10 are indicated by ***, ** and *, respectively.
Panel A: Before, during and after the crisis
Period N Event window CAR average t-statistic
Before 316 (-1+1) 0.140% 0.7991
(2006-2007) 316 (-3+3) 0.112% 0.4195
Crisis 201 (-1+1) -0.493% -1.3349 *
(2007-2009) 201 (-3+3) -0.898% -1.5915 *
After 381 (-1+1) 0.284% 1.1207
(2009-2011) 381 (-3+3) 0.224% 0.5771
Panel B: Difference between periods
N Event window CAR average t-value
Before vs. Crisis 517 (-1+1) 0.634% -1.5486 *
517 (-3+3) 1.011% -1.6177 *
Crisis vs. After 582 (-1+1) 0.777% -1.7347 **
582 (-3+3) 1.122% -1.6388 *
Before vs. After 697 (-1+1) 0.144% -0.4669
697 (-3+3) 0.111% -0.2359
