Cross border versus domestic mergers and acquisitions:
performance measurement in the beverage market
Amsterdam Business School
Name Christopher Robert Zeitlin Student number 10371893
Program Economics & Business Specialization Finance & Organization Number of ECTS 12
Supervisor J.J.G. Lemmen Target completion 28 / 06 / 216
Abstract
In this study, I investigate if there is a difference in performance of cross border and domestic M&A in the beverage market worldwide. I investigate if cross border abnormal returns are significantly lower than domestic abnormal returns. With 121 observations in a 3 day (-1,1) and a 5 day (-2,2) event window I found negative abnormal returns for both cross border and domestic M&A. Results show that domestic abnormal returns are significantly higher than cross border abnormal returns. Although both show negative returns, domestic acquisitions perform less poorly. This implies that it is not profitable to invest in the beverage industry, but if investment takes place, the best way to invest would be a domestic investment.
Table of contents
1. Introduction ...4 2. Literature review ...6 2.1 Cross border versus domestic acquisition
2.2 Beverage market specifications 2.3 Existing literature
3. Hypothesis, methodology and data ...10 3.1. Data and descriptive statistics
3.2. Hypothesis and methodology
4. Empirical result ...17 4.1. Empirical results
4.2 Robustness check
5. Conclusion and discussion ...22 References ...24 Appendix………..25
Statement of Originality
This document is written by Student [fill out your Given name and your Surname] who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table number Name
Table 1 Full regression statistics
Table 2 Descriptive statistics domestic CAR (-1,1) Table 3 Descriptive statistics domestic CAR (-2,2) Table 4 Descriptive statistics cross border CAR (-1,1) Table 5 Descriptive statistics cross border CAR (-2,2)
Table 6 Matching criteria
Table 7 t-test values
Table 8 Descriptive statistics full regression CAR (-1,1)
Table 9 Model summary CAR (-1,1)
Table 10 Pearson correlations event window (-1,1) Table 11 Coefficients full regression (-1,1)
Table 12 Descriptive statistics full regression CAR (-2,2)
Table 13 Model summary CAR (-2,2)
Table 14 Pearson correlations event window (-2,2) Table 15 Coefficients full regression (-2,2)
Table 16 Table 17
Levene’s test for equality of variances Literature time window summary
CHAPTER 1 Introduction
The world has never been as global as it is today. Due to globalization, a lot of investment opportunities for firms and corporations arise not only in their own country, but also abroad. Besides an increase of globalization, also liberalization and deregulation makes cross border mergers and acquisitions more accessible and profitable for investors. The creation of Anheuser-Busch InBev is an example of a typical corporation that expanded cross border to gain market share and profitability. In 2004 the Belgium brewery Interbrew merged with the Brazilian AmBev, which together became InBev. 4 years later, in 2008, they acquired the American firm Anheuser-Busch and in 2012 the Mexican Grupo Modelo has been added to the firm, which is now known as Anheuser-Busch InBev. Anheuser-Busch InBev or AB InBEv is the current market leader in the brewery industry with a market share of approximately 21%.
In this paper there is a focus on cross border versus domestic acquisitions explained by abnormal returns. In other words, I am researching if a cross border effect exists in the
beverage industry. A cross border effect can be explained as a significant difference between returns of cross border compared to domestic acquisitions (Moeller and Schlingemann, 2004). This thesis’ research question is: do cross border abnormal returns significantly differ from domestic abnormal returns in the beverage industry in the time period from 2000 until 2016? I want to answer this research question with a rejection of either the null hypothesis or the alternative hypothesis. The null hypothesis is that domestic abnormal returns are positive and cross border abnormal returns are negative. My two alternative hypothesis are (1) cross border acquisitions gain positive abnormal returns after a M&A and (2) cross border acquisitions have negative abnormal returns and do not significantly differ from domestic acquisitions.
Previous literature already researched whether a cross border effect is significant or not. Most of them are country specific and only little literature is focussed on cross border versus domestic M&A performance measurement in a specific industry. This paper shows if a cross border effect can also exist in the beverage market, a market which is divided between a couple of big firms and a lot of small firms. The focus on a specific market distinguishes this research from previous researches.
In chapter 2 existing literature and important differences between cross border and domestic M&A are summarized. After that, in chapter 3, I state the hypothesis, explain how the data is gathered and show the methodology of the research. In chapter 4 all empirical
results are explained and in chapter 5 I discuss them. Finally, I give a suggestion for a next research with the knowledge of this one.
CHAPTER 2 Literature review
2.1 Cross border versus domestic acquisition
In this research, cross border M&A are compared with domestic. Before we take a look at the data and methodology, it is important to distinguish these two acquisitions.
First of all, going global means that a company has to adapt to regulations and corporate laws that exist in the country of the target firm. King and Segain (2007) showed that adapting goes paired with a lot of costs for lawyers, training employees, consultants and many more. Returns of the acquisition have to offset this costs to be profitable for the bidder. The bigger and more complex the target firm, the higher the costs of acquisition.
As mentioned above, lawyers are needed for negotiations. The negotiation costs of lawyers for cross border acquisitions are higher than those for domestic ones (King and Segain, 2007), mainly because of their education. Lawyers are often solely taught their own law system, therefore a firm needs specialized lawyers to negotiate with the cultural
difference of the target firm.
Generally, extra costs and time investigated in a cross border M&A result in lower abnormal returns than domestic M&A. This is the main theory of the existing literature focussed on cross border versus domestic M&A. The hypothesis for this research will be explained in chapter 3.
2.2 Beverage market specifications
The beverage industry can be divided in roughly 4 different sections: beer and wine, spirits, carbonated beverage and coffee and other drinks department. As shown in Figure 1, the industry for brewery companies looks like it is divided exponentially. Anheuser-Busch InBev NV is the market leader and has got more market share than number 2, South African
Brewery Miller Plc e.g. SABMiller Plc, and number 3, Heineken NV, combined. In the carbonated beverage industry (Figure 2), the difference between the three biggest firms is not as big as in the brewery industry, but in this industry the three biggest firms together possess over 90% of the total market shares compared to just over 40% in the brewery industry.
Figure 3 shows the exponentially nature of the total beverage industry expressed in net revenue in 2014, where SABMiller was smaller than Heineken NV compared to 2015.
0 5 10 15 20 25 Anhseuser-Busch InBev NV
SABMiller Plc Heineken NV Carlsberg A/S China Resources Enterprises Ltd Tsingtao Brewery Co Ltd Moison Coors Brewing Co Beijing Yanjing Brewery Co Ltd Kirin Holdings Co Ltd Asahi Group Holdings
Market Share in %
Figure 1
Global Market Share Brewery Industry 2015
0 10 20 30 40 50
The Coca Cola Company PepsiCo, Inc. Dr Pepper Snapple Group Private label Other
Market Share in %
Figure 2
US Carbonated Beverage Industry 2015
0 10000 20000 30000 40000 50000 Anhseuser-Busch InBev NV
The Coca Cola Company PepsiCo, Inc. Nestlé SA Heineken NV SABMiller Plc Diageo Starbucks Corp. Pernod Ricard
Market Share in mil. US$
Figure 3
2.3 Existing literature
Research for a difference in domestic and cross border M&A is done by different people for different subjects like emerging markets, shareholders wealth, global diversification etc.
Moeller and Schlingemann (2004) published an article in Elsevier where they compared the difference in global diversification and bidder gains between domestic and cross border acquisitions in a period from 1985 until 1995. This research has been done because Moeller and Schlingemann (2004) saw a rise in new markets, an integration of capital and product market and they believed that globalization is a good strategy to improve
performance of a corporation. Other than this research, Moeller and Schlingemann (2004) don’t focus on an industry, but on the United States of America as a country. Their research shows significantly lower cross border abnormal returns compared to the domestic. Moeller and Schlingemann (2004) call the difference between the cross border and the domestic, a cross border effect which is only significant in the second part of their time period.
Another article that researches the difference between domestic and cross border performance after a M&A is published in the Journal of Business Finance & Accounting in 2012. Danbolt and Maciver (2012) researched if there exists a significant difference in the impact on shareholder wealth in a domestic versus a cross border acquisition. Just like
Moeller and Schlingeman (2004) they did not focus on a market, but on a country. In this case the focus lies on the United Kingdom. Their regression for the calculation of the CAR for an acquirer uses 12 different variables including LnCompanySize and CrossBorder Dummy that are also used in my research. Contrary to the results of Moeller and Schlingemann (2004), the research of Danbolt and Maciver (2012) showed that cross border cumulative abnormal returns are significantly higher than domestic cumulative abnormal returns. Bidder corporations have significant negative abnormal returns while the results show that cross border abnormal returns are not significant. Both are negative, but the domestic abnormal returns are more negative than the cross border abnormal returns.
Ma et al. (2009) did a research about abnormal returns in ten Asian stock markets focussing on M&A. Other than Moeller and Schlingemann (2004) and Danbolt and Maciver (2012) , Ma et al. (2009) used 3 different time windows. A 2 day time window (0, 1), a 3 day time window (-1,1) and a 5 day time window (-2,2) as event window for the expected
abnormal returns for the time period of 2000-2005. Most of the results from the research Ma et al. (2004) did not match expectations of previous literature.
Rai et al. (2015) did a study to find any short run effects in shareholders’ wealth after a M&A and found different strategies for investors. First of all, their research shows that
returns after a cross border M&A announcement are significantly higher than domestic returns. Furthermore, they state that a share purchase of the acquiring firm 2 days prior to the announcement day and to sell them 2 days after the announcement day gives significant returns. Also, M&A completed with cash and no shares perform better with substantial returns flowing from these acquisitions. Finally, cross border CAR is also significant in the long run, not only in the short run.
The conclusion of the summary of the researches above show a possibility for the rejection of the null hypothesis. Due to different findings in more or less the same researches, it is hard to predict the outcome of this research. My research is related to the previous ones in several ways. First I also use CAR to find if a cross border effect exists. I also use the 3 day (-1,1) window but just like Ma et al., I added a 5 day window (-2,2) for robustness and to compare both event windows.
CHAPTER 3 Data & methodology
3.1 Data and descriptive statistics
Four databases, Orbis, ZEPHYR, Thomson Reuters and Datastream, were used to collect the appropriate data for the M&A in the beverage industries. Collecting the data was a three stage process. The first part was to find all listed bidder firms in the beverage industry through Orbis. It is necessary that the bidder firms in that industry are listed, because the International Securities Identification Number e.g. ISIN is needed to find the stock price, index price and beta in Datastream. Orbis found 250 listed firms active in the beverage market. Finally, missing data was added manually.
ZEPHYR was used to filter all listed M&A. The following filters were added in ZEPHYR: 1) Time period: 01/01/2000 – 01/01/2016; 2) The M&A completion date; 3) NACE Rev.2: 11- Manufacture of beverages, 110 – Manufacture of beverage (Acquirer or Target); 4) Methods of payment: Cash, Shares; 5) All 250 ISIN numbers 6a) Domestic deals: Global; 6b) Cross border deals: Global; 7) Percentage of stake: Percentage of final stake (min: 50%). The outcome of the search was 79 cross border and 234 domestic deals. The data for the variables deal value, total assets of the acquirer, total assets of the target, completion date, method of payment and if the M&A was cross border or domestic were needed for the
regression. After a manual second filter in ZEPHYR, deleting all deals without the
information of all needed variables, 35 cross border and 86 domestic deals were left for the last step.
Finally, Datastreams’ EventStudyMatchingTool was used to find the price of the stock, price of the index and beta for the acquirer firms. These betas were necessary for the calculation of the expected return calculated with the CAPM model. See formula (2) below.
Every country-specific index was matched to the company. For example: the AEX index was used for Heineken, S&P500 for PEPSICO. etc. However, no index values could be found for the Chinese, Malaysian, Philippines (until 2014) and Serbian firm’s databases. For China, Malaysia and the Philippines I used the Morgan Stanley Capital International e.g. MSCI Emerging Markets Index and for Serbia the MSCI World Index. The difference can be explained that firms from 23 countries including China, Malaysia and the Philippines were used for the calculation of the MSCI Emerging Market Index. Serbia was not used for the emerging market index, that’s why I used the MSCI World Index for Serbia. Datastream found most company specific beta, which are needed for the calculation of the expected return
with the CAPM model, but not for firms settled in China and a couple of other countries with emerging markets. These betas were manually found on the THOMSON Reuters database.
Literature showed different estimation windows to calculate the expected price in the CAPM model. As shown in table 16 in the Appendix, often windows of 200+ trading days were used. The choice for (-123,-3) e.g. 120 trading days was made, because firms like Heineken, SABMiller and The Coca Cola Company made several acquisitions in short time periods. The (-123,-3) window therefore excludes overlapping in estimation and event windows of one specific firm. Below the statistics of the full regression are shown.
Table 1
Full regression statistics CAR
(2,2)
CAR (-1,1)
DealAss LogAqAss DummyGEO DummyMeth LOGDealVal
N 121 121 121 121 121 121 121 Mean -0.036 -0.023 -1,63 6.522 0.710 0.830 4.858 Median -0.013 -0.010 -1,618 6.522 1.000 1.000 4.746 Std. deviation 0.129 0.083 0.842 0.903 .455 .373 1.112 Skewness -3.773 -2.915 -0,146 -0.353 -0.941 -1,825 0.215 Std. error of skewness 0.220 0.220 0.220 0.220 0.220 0.220 0.220 Minimum -0.963 -0.559 -4,000 4.711 0 0 2.248 Maximum 0.309 0.177 0,419 8.209 1 1 7.615 Percentile 25 -0.059 9.510 -2,162 5.872 0.000 1.000 -0.041 Percentile 50 -0.012 50.000 -1,618 6.522 1.000 1.000 -0.010 Percentile 75 0.015 100.000 -1,088 7.287 1.000 1.000 0.012 Notes:
The results are based on the regression model calculated with the CAR from the 3 day event window (-1,1) and from the 5 day event window (-2,2):
Studying the results in the statistics, the CAR values are important for this research. It is expected that cross border CAR have negative results and domestic positive. As shown in table 4 an table 5 below, cross border CAR have a negative mean. However, table 2 and table 3 below show that domestic CAR are also negative, which is against expectation. Table 1 summarizes the descriptive statistics of the full regression. Also in the full regression the CAR´s are negative for both time windows. The negative mean of the LOGDealAss makes sense, because LOGDealAss e.g. Deal Value/Total Assets Acquirer is a fraction. Often is the value of the fraction between 0 and 1. Every number calculated with a logarithm bigger than 0 and smaller than 1 has a negative outcome. The mean 0.965 in table 1 and table 2 and 0.800 in table 3 and table 4 of the DummyMeth makes sense, because the values of DummyMeth are
either 0 if the deal is financed with shares or 1 with cash. This implies that most acquisitions have been financed with cash.
Table 2
Descriptive statistics domestic CAR (-1,1)
Mean Std. Deviation N CAR -0.017 0.057 86 DummyMeth 0.965 1.152 86 LOGDealVal 4.554 1.039 86 LOGDealAss -1.687 0.850 86 LOGAssAq 6.249 0.848 86 Table 3
Descriptive statistics domestic CAR (-2,2)
Mean Std. Deviation N CAR -0.025 0.075 86 DummyMeth 0.965 1.152 86 LOGDealVal 4.554 1.039 86 LOGDealAss -1.687 0.851 86 LOGAssAq 6.249 0.848 86 Table 4
Descriptive statistics cross border CAR (-1,1)
Mean Std. Deviation N CAR -0.038 0.126 35 DummyMeth 0.800 0.406 35 LOGDealVal 5.603 0.928 35 LOGDealAss -1.590 0.812 35 LOGAssAq 7.193 0.656 35 Table 5
Descriptive statistics cross border CAR (-2,2)
Mean Std. Deviation N CAR -0.066 0.209 35 DummyMeth 0.800 0.406 35 LOGDealVal 5.603 0.928 35 LOGDealAss -1.590 0.812 35 LOGAssAq 7.193 0.656 35 Notes:
Results from all table 1 until table 4 calculated with the regression:
Table 6 Matching criteria
Cross Border Domestic
Panel A: Country top 6
n % n %
Belgium 7 18.42 China 20 24.39
Great Brittain 7 18.42 South-Korea 14 17.07
Netherlands 7 18.42 Great-Brittain 8 9.76
United States 4 10.53 United States 8 9.76
Japan 3 7.89 Malaysia 4 4.88 Denmark 3 7.89 Netherlands 4 4.88 Total 31 81.57 58 70.74 Panel B: Year 2000 0 2000 1 2001 4 2001 6 2002 5 2002 1 2003 2 2003 3 2004 4 2004 7 2005 6 2005 8 2006 3 2006 7 2007 0 2007 4 2008 1 2008 3 2009 2 2009 8 2010 2 2010 11 2011 0 2011 5 2012 1 2012 3 2013 2 2013 7 2014 2 2014 8 2015 1 2015 3 35 85
3.2 Hypothesis and methodology
This research contains a null hypothesis and two alternative hypothesis. Existing literature show that on average, abnormal returns of domestic acquisitions are positive and cross border acquisitions have negative abnormal returns. This is the null hypothesis that has been used in this research. The alternative hypothesis are (1) cross border acquisitions gain positive abnormal returns after a M&A and (2) cross border acquisitions have negative abnormal returns and do not significantly differ from domestic acquisitions. The second alternative hypothesis can be translated that both cross border and domestic M&A have significant negative abnormal returns and do not significantly differ. I test the hypothesis in two different ways. First, for the calculation with CAR, I’m doing a Levene’s test of Equal variances with hypotheses:
where H1 is a 2-tailed test. The Levene’s test does a one way ANOVA test of the absolute difference between the value X and the mean for each cross border and domestic group. If the value is significant, the null hypothesis has to be rejected and the two groups, cross border and domestic, have different variances. The other way to test the hypothesis is to focus on the DummyGEO. If the DummyGEO has a significant value in either of the full regressions, there is a significant difference between cross border and domestic M&A with CAR as dependent variable.
Just like Danbolt and Maciver (2012) I used abnormal returns to discover whether there is a difference in performance comparing domestic versus cross border acquirer firms after a M&A. First of all, it is important to know what regression is used. The regression that I want to use for this research is the following:
where GEO is a dummy variable with value 0 if the M&A is cross border or value 1 if the M&A is domestic and is shown as DummyGEO in the regression, METHPAY is a
dummy whether the M&A has been financed with Cash (value 0) or with Shares (value 1) and is shown as DummyMETH in the regression, DEALVALUE is a logarithm of the value of the deal e.g. log(DealValue), SIZE is the logarithm of the difference of the acquirers total assets minus the targets total assets e.g. log(total assets acquirer – total assets target), DEALAS is the logarithm of the fraction of the deal value compared with the target e.g. log(Deal
Value/Total Assets Acquirer) and is the standard error term. The variables GEO and SIZE were also used by Danbolt and Maciver (2012). They didn’t have a variable linked to the deal value, but I think it is important for the regression.
My research differs from previous studies, because I use the SIZE and DEALAS variable to measure the relative size between the acquirer and the target. However, these variables can only be used if there is no multicollinearity. Between the variables deal value and total assets acquirer there was no multicollinearity, but between the variables total assets acquirer and total assets target there was. Multicollinearity exists if | | . The Pearson correlation between total assets acquirer and total assets target was 0.720, which concludes there is multicollinearity between the two variables. Because of multicollinearity, I am not allowed to have total assets acquirer and total assets target together in my regression. I have to choose the variable that gives the highest Rsquare e.g. R² in my full regression. The
variable that gave the highest R² was total assets acquirer with a value of 0.048 compared to total assets target with an R² of 0.042 and the SIZE e.g. (total assets acquirer – total assets target) with an R² of 0.042. The regression I used for my research is:
where the variables are the same as explained above and AssAq is a logarithm (total assets acquirer). The logarithm was necessary to improve the B in the unstandardized coefficients. Without the logarithm the value was 1.835E-009, which is about the same as 0 and thus negligible. Big differences in total assets of the acquirer cause the low B. This is a characteristic of the beverage industry as explained above.
It was important to combine the deal value in a logarithm, because of the huge difference between the biggest deals of 42 billion and the smallest deals of 100.000. I also added the method of payment, but I left out the gain of the original owners stake for 2 reasons. First I couldn’t find enough information and had to filter the firms again and wouldn’t have enough M&A left for my regression. Second I searched for a final stake of at least 50% so I more or less filtered M&A with this restriction. The reason why I chose a final stake of 50% is, because I want to focus on a M&A where the acquiring firm will eventually have a majority of the target firm, which mostly is combined full control of the target firm.
For the calculation of the cumulative abnormal returns, which are the actual return minus the expected returns e.g. returns the firm would have if the M&A did not take place, the following formula has been used:
(1)
where i = 1, 2, 3, …, N, t is the time index, are the actual returns and are the expected returns if the M&A did not take place. Both returns and are returns during period t on their securities i. Equation (1) is calculated for two different time windows. The actual returns are calculated in a 5 day time window, 2 days before until 2 days after the acquisition, and in a 3 day window, 1 day before until 1 day after the acquisition. Following Ma et al. (2009), the expected returns can be calculated in several ways: the single-index model, the market model and the CAPM model. In this research the Capital Asset Pricing
Model is used, because it focuses on the volatility of the market return which is a good prediction of the flow of the market return:
(2)
where is the expected return, is the risk free rate, β is the beta of the company and is the market return of the index. To calculate the risk free rate, the average of 4 months, which is nearly the same as the estimation window of 118 days, monthly returns of US Treasury Bills e.g. 3 months T-Bills were used. The is the calculated average return of the index price for the estimation window from 121 days until 3 days prior of the event. Both equation (1) and (2) are needed for the calculation of Average Abnormal Returns e.g. AAR and the Cumulative Abnormal Return e.g. CAR, which is needed to measure performance of the acquirer:
∑ (3)
where AAR is the same as the calculated mean in the descriptive statistic part of the regression. The formula for the CAR is:
∑ (4)
where T1-T2 are the two different event windows of (-2,2) and (-1,1). In the next chapter these two windows will be compared.
To calculate the significance of the cumulative abnormal returns I used a student t-test. Just like equations (1), (2), (3) and (4) the t-test was done manually in excel with the
following formula:
(5)
where t is the t-value, N is the number of days during the event, CAR the value of the cumulative abnormal return and the standard deviation of the abnormal returns.
CHAPTER 4 Empirical results
4.1 Empirical results
In this chapter, all important results are summarized and in chapter 5 they will be discussed. Because of the 2 different alternative hypothesis, (1) cross border M&A have positive CAR and (2) cross border M&A CAR are negative but do not significantly differ from domestic M&A, it is important to distinguish both domestic and cross border CAR.
The results in table 7 are calculated with a student t-test, which is formulated in equation (5). Except for 2 days prior to the announcement day, which is 5% significant, all other domestic CAR are significant on a 1% significance level. Cross Border results on the other hand are not all significant. Both extreme days, 2 days prior and after the announcement day, and 1 day prior the announcement day are significant on a 1% significant level, but both CAR on the announcement day and 1 day after the announcement day are not significant. To translate significant values to economic understanding, all significant values provide
information about the abnormal returns. The significance of the abnormal returns says that there is indeed a difference between expected returns and actual returns, no matter if they are positive or negative.
Table 7 t-test values
Domestic Domestic Cross Border Cross Border
(-2,2) (-1,1) (-2,2) (-1,1) CAR -2 -2.0623** 4.4649*** CAR -1 -6.4186*** -8.2864*** 2.8234*** 3.6450*** CAR 0 -12.2573*** -15.8241*** -0.4581 -0.5914 CAR 1 -14.3614*** -18.5405*** 0.6914 0.8926 CAR 2 -7.3044*** -2.4206*** ***, **, * significant in 1%, 5%, 10% respectively
Despite some insignificant values of the cross border CAR, summary 1 and summary 2 below show the results of the full regression for the 3 and 5 day event window. The
GEOGRAPHIC Dummy is added into the full regression. Significance of this dummy would show a difference between cross border and domestic CAR. The results show that only the GEOGRAPHIC Dummy is significant for both the 5 day event window the Geographic dummy variable has got a significant outcome of 0.034 or 3.4% (table 15) and the 3 day event
window has a significant outcome of 0.048 or 4.8% (table 11). Summary 1
Full regression CAR (-1,1) Table 8
Descriptive statistics full regression CAR (-1,1)
Mean Std. Deviation N CAR -0.024 0.083 121 DummyGEO 0.710 0.455 121 DummyMeth 0.830 0.373 121 LOGDealVal 4.858 1.112 121 LOGDealAss -1.659 0.837 121 LOGAssAq 6.522 0.903 121 Table 9 CAR (-1,1)
Model R R square Adjusted R
square Std. error of the estimate Durbin-Watson 1 0.042 0.000 0.083 2.046
a. Predictors: (Constant), LOGDealAss, DummyGEO,DummyMeth, LogAqAss, LOGDealVal
b. Dependent Variable: CAR Table 10
Pearson correlations event window (-1,1)
CAR DummyGEO DummyMeth LOGDealVal LOGDealAss LOGAssAq
CAR 1.000 0.118 -0.002 0.029 -0.061 0.093 DummyGEO 0.118 1.000 0.060 -0.429 -0.053 0.476 DummyMeth -0.002 0.060 1.000 -0.283 -0.216 -0.257 LOGDealVal 0.029 -0.429 -0,283 1.000 0.605 0.664 LOGDealAss -0.061 -0.053 -0.216 0.605 1.000 -0.192 LOGAssAq 0.093 0.476 -0.257 0.664 -0.192 1.000 Table 11 full regression (-1,1)
Unstandardized coefficients Standardized coefficients
B Std. error Beta T Sig.
(Constant) -0.168 0.075 -2.222 0.028 DummyGEO 0.039 0.020 0.214 2,000 0.048 DummyMeth 0.002 0.021 0.008 0.079 0.938 LOGDealVal 0.012 0.013 0.157 0.912 0.363 LOGDealAss -0.015 0.012 -0.151 -1.217 0.226 LOGAssAq 0.006 0.012 0.076 -0.547 0.585
a. Dependent variable: CAR Notes:
The results are based on the regression model calculated with the CAR from the 3 day event window (-1,1):
Summary 2:
Full regression CAR (-2,2) Table 12
Descriptive statistics full regression CAR (-2,2)
Mean Std. Deviation N CAR -0.036 0.129 121 DummyGEO 0.710 0.455 121 DummyMeth 0.830 0.373 121 LOGDealVal 4.858 1.112 121 LOGDealAss -1.659 0.837 121 LOGAssAq 6.522 0.903 121 Table 13 CAR (-2,2)
Model R R square Adjusted R
square Std. error of the estimate Durbin-Watson 1 0.048 0.007 0.129 2.043
a. Predictors: (Constant), DealAss, DummyGEO,DummyMeth, LogAqTar, LOGDealVal b. Dependent Variable: CAR
Table 14
Pearson correlations (-2,2)
CAR DummyGEO DummyMeth LOGDealVal LOGDealAss LOGAssAq
CAR 1.000 0.145 -0,028 -0.002 -0.080 0.072 DummyGEO 0.145 1.000 0.060 -0.429 -0.053 0.476 DummyMeth -0.028 0.060 1.000 -0.283 -0.095 -0.257 LOGDealVal -0.002 -0.429 -0,283 1.000 0.605 0.664 LOGDealAss -0.080 -0.053 -0.095 0.605 1.000 -0.192 LOGAssAq 0.072 0.476 -0.257 0.664 -0.192 1.000 Table 15 full regression (-2,2)
Unstandardized coefficients Standardized coefficients
B Std. error Beta T Sig.
(Constant) -0.253 0.110 -2.299 0.036 DummyGEO 0.064 0.030 0.227 2.149 0.034 DummyMeth -0.013 0.033 -0.038 -0.395 0.693 LOGDealVal 0.035 0.232 0.301 0.151 0.880 LOGDealAss -0.041 0.234 -0.264 -0.174 0.862 LOGAssAq -0.010 0.231 -0.071 -0.044 0.965
a. Dependent variable: CAR Notes:
The results are based on the regression model calculated with the CAR from the 5 day event window (-2,2):
The other way to test the hypothesis is to compute a Levene’s test of equal variances. I did an independent samples t-test in SPSS with CAR as test variable and DummyGEO as grouping variable with cross border M&A as group 1 with value 0 and domestic M&A as group 2 with value 1.
Table 16
Levene’s test for equality of variances
F Sig
Equal variances assumed CAR(-1,1) 6.126 0.015** Equal variances assumed CAR(-2,2) 10.709 0.001*** Notes:
This test is calculated with an independent t-test. *** and ** are significant values for alpha of 1% and 5% respectively.
The hypothesis for this test are:
H0: = and H1: ≠
Table 16 shows, that for both event windows, the values of the Levene’s test for equality of variances are both significant of 1.5% for the 3 day event window and 0.1% for the 5 day event window. Due to insignificant values of the test, the null hypothesis has to be rejected, which concludes that the variances of the cross border CAR and the domestic CAR are significant different.
4.2 Robustness
To compare the outcome of this research I used two different time windows of 3 days, 1 prior until 1 after the announcement day, and 5 days, 2 prior until 2 after the announcement day. This is important to compare the result to show that there is a difference in time windows. Both time windows show a significant value in the GEOGRAPHIC dummy which can be translated to a significant difference between a domestic or cross border M&A. With 4.8% for the 3 day event window and 3.4% for the 5 day event window, both DummyGEO are
significant with an alpha of 5%.
Multicollinearity, heteroskedasticity and autocorrelation have impact on the use of the regression. If two variables are correlated too much, in this research the variables total assets acquirer and total assets target, they can affect the results. Correlation tables 10 and 14 show no multicollinearity between the six variables. When the absolute correlation between 2 variables is the same or higher than 0.7 there is multicollinearity, but neither regression has a value higher or the same as 0.7. For a valid regression, the variance of the error variable is required to be constant. Regression plots did not show a pattern, so there was no
close to 0 implies positive autocorrelation e.g. a correlation of a residual in time t with a residual in time t-1. A value close to 4 implies negative autocorrelation. As shown in table 9 and table 13, both Durbin-Watson values are close to 2, which implies no evidence of autocorrelation.
CHAPTER 5 Conclusion and discussion
In this research, I was looking for a difference in abnormal returns between cross border and domestic merges and acquisitions in the beverage industry worldwide. With the databases Orbis, ZEPHYR, Datastream and Thomson Reuters I eventually found 35 cross border and 86 domestic M&A within the beverage industry for the time period from 01/01/2000 until
01/01/2016. For the 3 day event window (-1,1) and the 5 day event window (-2,2) I found negative cumulative abnormal returns for cross border and domestic M&A. The student t-test showed that most CAR are significant. Finally, the full regression calculated a significant difference of the Geographic dummy of 4.8% in the 3 day event window and 3.4% in the 5 day event window, which concludes that there is a significant difference of abnormal return whether the M&A was domestic or cross border. The Levene’s test for equality of variances also showed a difference between cross border and domestic M&A. With the values of 1.5% for the 3 day event window and 0.1% for the 5 day event window, the null hypothesis has been rejected which concludes that the CAR for cross border and domestic M&A are differ significantly. Given that both domestic and cross border CAR are negative and the domestic CAR differs significantly from the cross border CAR, the null hypothesis, that cross border CAR are negative and significantly different from domestic CAR, is not rejected. Acquiring firms that invest in a domestic M&A perform better, or in other words they perform less poorly, than acquirers of a cross border M&A.
The rejection of both alternative hypotheses are generally in line with most existing literature that domestic performs better than cross border M&A. The results of Rei et al. (2015) would reject the null hypothesis and accept the alternative hypothesis that both
domestic and cross border CAR are positive. Danbolt and Maciver (2012) found significantly higher cross border CAR which is the opposite of my results. Moeller and Schlingemann (2004) found the same as this research i.e. significant lower cross border CAR. Ma et al. (2009) didn’t find a significant result. The four researches mentioned above and in chapter 2 show that different results were found for the rejection of the null and both alternative
hypothesis. This states that there it is not expected that domestic CAR are always higher than cross border CAR.
During the project, a couple of limitations and road blocks did arise. One limitation was the fact that only firms that were listed could be used as acquiring firm because crucial information like stock prices are not available for most of the non-listed firms. Furthermore, I had to delete more than half of the data, because information like total assets of the target was
not available and couldn’t be found manually. Also a couple of firms were not added to the research due to lack of information of the share price during the event window. The major limitation in this research was that I couldn’t use the SIZE variable, because of
multicollinearity between the variables total assets acquirer and total assets target. I wanted to research whether this relative size variable had a significant effect on the CAR. I had to replace this with a logarithm function of the total assets of acquirer.
As seen in table 7, a couple of CAR from cross border M&A are not significant. Luckily the Geographic dummy is significant in both event windows to answer the research question whether there is a difference in performance between cross border and domestic M&A but besides the constant variable, all other variables were not sufficient and do not matter for determining the value of the CAR. Due to the focus on the beverage industry, only 121 acquisitions could be used for the regression which is a relative small amount. A higher amount of observations could cause significance for insignificant variables or vice versa.
Finally, it would be interesting to investigate whether CAR are also negative in more equally divided industries, compared with industries such as the beverage market. Also it would be interesting what the influence of the kind of the acquisition on the CAR i.e. if a hostile acquisitions causes negative CAR or not?
References
Danbolt & Maciver, 2012, Cross-Border versus Domestic Acquisitions and the Impact on Shareholder Wealth, Journal of Business Finance & Accounting 39(7) & (8), 1028-1067. Georgopoulos, 2008, Cross-Border Mergers and Acquisitions: Does the Exchange Rate Matter?, The Canadian Journal of Economics / Revue Canadienne d’Economique 41(2), 450-474.
King, & Segain, 2007, Cross Border Negotiated Deals: Why Culture Matters?, Economic
Council on Foreign Relations, 126-166.
Ma, Pagán & Chu, 2009, Abnormal Returns to Mergers and Acquisitions in Ten Asian Stock Markets, International Journal of Business, 14(3), 235-250.
Moeller & Schlingemann, 2004, Global diversification and bidder gains: A comparison between cross-border and domestic acquisitions, Journal of Banking & Finance, 29, 533-564. Rani, Yadav & Jain, 2015, Impact of Mergers and Acquisitions on Shareholders’ Wealth in the Short Run: An Event Study Approach, The Journal for Decision Makers, 40(3), 293-312.
Appendix
Table 17
Literature time window summary
Author Estimation period Event window(s)
Moeller and Schlingemann (2005)
(-205, -6)
200 trading days
(-10,10) and (-1,1)
Danbolt and Maciver (2012) (-260, -41) 220 trading days (-5,5) and (-1,1) Ma et al. (2009) (-125, -6) 120 trading days (-2,2), (-1,1) and (0,1) Rani et al. (2015) (-280, -26) 255 trading days (-20, 20)
