1
Date and version:
29.01.2014, final version
Name and student number:
Nadine van Meel, 10263314
Supervisor:
dr. P.J.P.M. Versijp
BSc Finance and Organisation, University of Amsterdam
University of Amsterdam
Time-varying of
systematic risk
An event study of how the announcement of mergers and acquisitions
affects systematic risk
University of Amsterdam
Time-varying of
systematic risk
An event study of how the announcement of mergers and acquisitions
affects systematic risk
2
Abstract
Economic literature still has not reached consensus about the time-varying properties of beta – the company’s systematic risk. This empirical study investigates whether an announcement of a merger or acquisition affects beta of AEX-listed companies in the period 2006-2010. Other events, such as the impact of the recent financial crisis and the effect of multiple announcements in a short period of time by the same company, are also taken into account. The company beta is estimated before the announcement of a takeover deal, and after completed date of the deal. The betas of both event windows are compared to each other by conducting a paired t-test. The research shows that on average, the announcement of a takeover deal does not influence the company’s systematic risk, thus implying that beta is constant over time with respect to the announcement of mergers and acquisitions.
3
Table of Contents
1 Introduction ... 4
1.1 Discussion about beta’s properties ... 4
1.2 Broader perspective ... 5
1.3 Structure of the empirical research ... 5
2 Theoretical background ... 6
2.1 CAPM and its underlying assumptions ... 6
2.2 The market portfolio ... 7
2.3 Beta in the spotlight ... 8
2.4 Hypotheses ... 11
3 Methodology ... 13
3.1 Collecting data for M&A’s ... 13
3.2 Event windows and erratically behaving company returns ... 13
3.3 Assigning dummy values ... 14
3.4 Selection criteria ... 14
3.5 Collecting data about stock- and market prices ... 15
3.6 Running regressions and conducting a t-test ... 15
4 Analysis ... 19
4.1 Outcome paired t-test ... 19
4.2 Some remarks ... 19 4.3 Validation of hypotheses ... 23 4.4 Explanation ... 24 5 Conclusion ... 25 Reverences ... 26 Appendix ... 29
4
1 Introduction
1.1 Discussion about beta’s properties
Systematic risk, i.e. the sensitivity of company returns to market fluctuations – as measured by beta – is a very well-known concept in finance. However, within the existing literature there is still no consensus about the assumed characteristic of beta to be time-varying. This lack of consensus is already mentioned by Choudhry (2002). In light of the recent economic and financial crises several experiments are conducted to investigate whether the beta of a company changes over time due to the effect of severe economic market shocks. For instance, Choudhry (2002) found that the betas of twenty Taiwanese and Malaysian companies were affected by the Asian financial crisis in the years 1997-1998. This result is consistent with the opinion of Fabozzi and Francis (1978) that betas can fluctuate in response to a changing market environment. Others however, are convinced that betas remain constant over time (Bos & Newbold, 1984, p. 35) – irrespective of the economic situation.
The same discussion in respect of beta to be time-varying applies to mergers and acquisitions. Mandelker (1974) was the first to conduct an event study to assess whether company betas change as a consequence of a merger or acquisition. He observed significant changes in company betas. The main drawback of his study though, is that he used the merging date to define the two event
windows (pre- and post-event) of the data set instead of the date of the takeover announcement. This makes his results biased because the period between announced date and completion date of a takeover deal is often characterized by erratic behavior (Garabato, Gerpe & Maiztegui, 2008, p. 53; Brealey, Cooper & Kaplanis, 2010, p. 1724), and should therefore be excluded from the investigation. For his research however, Mandelker takes into account this period to base his calculations on. From Hackbarth and Morellec (2008) it follows that the beta of the bidding company increases/decreases prior to the takeover when the acquiring company has a higher/lower pre-announcement beta than its target. For their research they used daily stock returns and consequently their focus was on the immediate market reaction of announced mergers and acquisitions. Furthermore, they predicted that when the takeover becomes more/less likely, the value of the option to merge
increases/decreases as a percentage of the total firm value. That is why the priced risk of the acquiring company increases/decreases and likewise does its beta. In contrast to the findings of Hackbarth and Morellec (2008), Amihud, DeLong and Saunders (2002) could not find evidence of a significant change in the systematic risk of companies engaged in takeovers when comparing the beta prior to the announcement of a merger or acquisition to the beta afterwards. This study as well is based on daily stock return data, and therefore provides a short-term view on the dynamics of beta. Madura and Wiant (1994) found the same results in an earlier study already, but their focus
5 was on the long-term effects by making use of monthly returns for their research. Hence, their findings are not readily comparable to the results of the study of Hackbarth and Morellec (2008) and to those of Amihud, DeLong and Saunders (2002).
1.2 Broader perspective
Having knowledge of the behaviour of betas is relevant for investors and other stakeholders when making rational investment decisions. Therefore, this study will investigate how beta of AEX-listed companies in the period 2006-2010 is affected by the announcement of a merger or acquisition. Where the studies of Amihud, DeLong and Saunders (2002) and of Madura and Wiant (1994) only investigated the effects on systematic risk of takeovers in the banking industry, this study will take this broader by not imposing a restriction on sectors. In various industries and business segments, the possible effects on systematic risk will be investigated for AEX-listed companies. Therefore, the results of this study will be comparable to those of Hackbarth and Morellec (2008), who conducted a similar study for a number of U.S. publicly traded companies. On the other hand, monthly stock and market data is used to investigate the long-term effect of mergers and acquisitions on beta. In this respect, research findings correspond as well to the conclusion of the study of Madura and Wiant (1994), despite the fact that their sample set only consists of takeover deals in the banking sector.
1.3 Structure of the empirical research
The research is of empirical nature. For a number of takeover deals the company betas will be estimated in two different periods – before the announced date and after the completion date of the merger or acquisition – with the help of financial company data and the Capital Asset Pricing Model (hereinafter: CAPM). The focus is solely on the beta of the acquiring company, as the target in many cases will not be listed. The estimation will be done using the linear regression technique. Finally, the estimated betas will be compared by means of a paired t-test in order to assess whether on average the beta of AEX-listed companies in the period 2006-2010 changed due to the announcement of a takeover deal.
Paragraph 2 will elaborate on the CAPM and an hypothesis is formulated. Then paragraph 3 raises the matter of the methodology, describing into more detail the characteristics of the data used and the process of selection for investigation. Lastly, paragraph 4 provides an analysis of the research results and the conclusion will follow in paragraph 5.
6
2 Theoretical background
2.1 CAPM and its underlying assumptions
The objective of this empirical study is to investigate any possible change in company betas. The estimation will be fundamentally based on the CAPM, which is the most common method used in practice for estimating betas (Berk & DeMarzo, 2011, p. 322; Fama & French, 2004, p. 25). The CAPM is firstly introduced by Sharpe (1964), building his study on the earlier work of Markowitz (1952), who formalized the role of diversification in forming an optimal stock market portfolio. Before evaluating whether an announcement of a merger or acquisition had any influence on beta, getting a better understanding of the CAPM is essential.
As mentioned above, one of the applications of the CAPM is estimating company betas. According to Berk and DeMarzo (2011) and Fama and French (2004) there are several assumptions underlying the model. Firstly, investors should be able to trade their stocks at competitive market prices, and money can be lent and borrowed at the risk-free interest rate. Furthermore it is assumed that investors are holding efficient stock portfolios (which thus will yield the maximum expected return given a certain volatility). The third underlying assumption is that investors have homogeneous expectations with regard to volatility, correlation and expected return of stock. Since investors have the same expectations, they will all invest in the market portfolio. As a consequence, all idiosyncratic risk is diversified and the portfolio only contains systematic risk, i.e. market risk. The risk level related to the performance of the market portfolio represents therefore the company’s systematic risk, captured by beta.
According to Lee, Reed and Robinson (2008), the main determinant of the magnitude of the
systematic risk of a company is the company's financial condition. Olsen and Patel (1984) argued that both short-term financial leverage and business risk are directly related to the level of systematic risk. The more recent study of Allen, Madura and Springer (2000) as well as the results of Delcoure and Dickens (2004) identified a positive and statistically significant relation between long-term debt and systematic risk. A change in capital structure may therefore cause a change in the company’s beta. There is also work that sees a company's equity as a call option on the value of the company, with the level of debt being the exercise price, e.g. the research study of Grass (2010). In that sense, the leverage would influence the beta. A second determinant of systematic risk that is cited in literature is company size. Firstly Gyuorko and Nelling (1996), but also the later work of Conover, Friday and Howton (1998), as well as that of Tan (2004) found evidence for a positive and statistically significant relation between size and systematic risk. However, consensus about this topic is not
7 reached yet; the studies of Litt, Mei and Webber (1999), Ambrose and Linneman (2001) and Byrne and Lee (2003) produced contrasting results. Overall, the above research results confirmed that company characteristics may be of influence on the systematic risk the company bears.
The natural way to estimate the CAPM is by linear regression. This technique identifies the best-fitting line through a set of points. The linear regression expresses excess return as the sum of three components, so that the CAPM-equation will look as follows (Berk & DeMarzo, 2011, p. 386):
(Ri− rf) = αi+ βi× (Rmarket− rf) + εi
The first element represents the risk premium on the investment concerned. The corresponding risk premium on the market is given by (Rmarket− rf), and εi is the residual term. This term represents the deviation from the estimated best-fitting line and is assumed to be zero on average. The
reasoning underlying this assumption is that εi corresponds to the diversifiable risk of the investment, which is unrelated to the market. This way, positive deviations will be cancelled out against negative deviations, yielding zero on average. The constant alpha in the above equation quantifies the historical performance of the investment relative to the expected return; it measures the variance of the investment’s return (higher/lower) compared to the security market line. Therefore, alpha can be interpreted as a risk-adjusted measure of the investment’s historical performance. Once again, this term of the equation should not be significantly different from zero, according to the presumptions underlying the CAPM (Berk & DeMarzo, 2011, p. 386). The beta eventually – the component of interest – measures the expected percentage change in the return of the investment given a 1% change in the return on market portfolio (Van Ophem, 2012, p. 18). The empirical study will focus on this measure of sensitivity.
2.2 The market portfolio
As becomes clear in the above paragraph, applying the CAPM requires the returns of a market portfolio. Such a portfolio contains only systematic risk. This implies that risk cannot be reduced any further without lowering the expected return. The Standard & Poor’s Europe 350 Index (hereinafter: S&P 350) is chosen as benchmark for the market. Similar to the AEX-companies, the companies included in this index operate on European markets. As well as in the study of Alam and Rajjaque (2010), and relying on the expertise of Standard and Poor’s (S&P Europe 350, 2014), the S&P 350 qualifies as an appropriate approximation for the market portfolio. The index represents circa 70% of the market capitalization of the region and therefore complies with the criterion of being large enough to be considered essentially fully diversified. While this portfolio is not a global market portfolio, many investors have considerable home bias. Lau, Lilian and Zhang (2010) found
8 evidence for this phenomenon and cite several reasons such as information asymmetries, behavioral biases, and accounting standards. Moreover, a genuine market portfolio containing all assets - the only setup that can be guaranteed not to have any further diversification benefits - is impossible. For the above reasoning the S&P 350 qualifies as market portfolio. Logically, the beta of the market portfolio equals one. Linear regression namely measures beta relative to a percentage change of the market return, thus yielding an identity.
2.3 Beta in the spotlight
Since the focus of this study is on company betas, an understanding of what beta exactly embodies is fundamental. In short, it represents systematic risk. The beta of a company measures the sensitivity of company returns to fluctuations in the market. Unlike idiosyncratic risk (that has little or no correlation to the market), systematic risk cannot be eliminated by diversification (Berk & DeMarzo, 2011, p. 311). A distinction can be made between betas with a value above one, and those with a value between zero and one. The values of beta distinguish between cyclical and non-cyclical companies, referring respectively to betas with a value above one and those with a value between zero and one. Berk and DeMarzo (2011) refer to cyclical companies as being more sensitive to risk, which is characterized by more strongly fluctuating revenues. This kind of companies therefore has a beta that exceeds the beta of the market portfolio. Non-cyclical companies are to a much lesser extent subject to fluctuations and hence show a lower beta than the market portfolio. Lastly, there is the special case of companies having negative betas; this indicates that the returns of the investment move oppositely relative to the market portfolio.The companies that will be investigated in the following empirical study operate in several different sectors, so it may well be the case to find both cyclical and non-cyclical betas.
As indicated in the introduction, not only the announcement of a merger or acquisition may exert influence on the company’s beta. A change may be caused as well by other market shocks that effect the economy as a whole. Since the financial crisis started in the specified time period subject to investigation (2006-2010), there has to be controlled for the effect of this market shock on beta. For the purpose of this research the financial crisis is set to last from 22 September 2008 to 21 December 2009. The crisis period starts just after the fall of Lehman Brothers, where the graphs show a
significant decrease of stock prices (S&P Dow Jones Indices, 2014). The moment ING made the first partial repayment of received state aid (21 December 2009) (Stuiveling & Van Schoten, 2010, p. 26) is used to mark the end of the period. Among the Dutch companies that received state aid, ING was the very first in making a (partial) repayment of the debt. It is assumed that this first repayment indicated
9 a recovery of the Dutch economy and it was a first step in restoring economic confidence. Therefore the end of the crisis period is set on December 21st of 2009.
Figure 1, graphical representation of the performance of the S&P 350 and S&P 500 Indices1
Brealey, Cooper and Kaplanis (2010) state that it has become conventional in literature on excess comovement to call the measured coefficients that result from the multiple regression rather than from the standard simple regression betas. From a friction-based view, excess comovement arises because of investor-trading patterns that are not explained either by cash flows or by a frictionless stochastic discount factor. Similar to Vijh (1994), Barberis, Shleifer and Wurgler (2005) provided evidence in support of this friction-based view against the traditional theory – the fundamental approach – that assumes that comovement between share prices is caused by common fundamental factors in a frictionless market. From the friction-based view, an AEX-listed share will tend to fall – despite global cash flows – when the Dutch market falls, just because the share is mainly held by Dutch investors. Selling by those investors causes a further decrease of the market. As a
consequence, all betas will tend to move towards 1 - the beta of the market portfolio. This way, a change in beta can be explained by excess comovement.
Likewise, multiple takeover announcements of a company in a short period of time could also have an impact on the company’s beta, and there has to be controlled for as well. Variables for both factors of potential influence should be included in the regression model to prevent from omitted variables bias. The variables concerned will be simple dummy variables2. Dummy
1 and dummy2 represent respectively the values indicating the occurrence of multiple announcements in the period of investigation, or that the event window of the deal includes the financial crisis period.
1Derived from S&P Dow Jones Indices
0 200 400 600 800 1000 1200 1400 1600 1800 2-1-2007 2-1-2008 2-1-2009 2-1-2010 2-1-2011 2-1-2012 S&P 350 S&P 500
10 Those are multiplied by the risk premium as well as by the beta to isolate the effect of other takeover announcements and/or the financial crisis. Not including these dummy variables would introduce omitted variables bias into the coefficient of interest. The occurrence of multiple announcements or an overlap of the event window with the financial crisis will yield a different company beta compared to the situation of a single takeover announcement and no overlap with the financial crisis. The beta-coefficient in front of the two dummy variables will capture that part of the company beta that is caused by other announcements and/or the financial crisis. Since also these dummy variables have to be taken into account when estimating beta, the basic equation of the CAPM will be extended to:
(Ri,t− rf) = αi,t+ β1× (Rmarket,t− rf) + β2× dummy1× (Rmarket,t− rf) + β3× dummy2 × (Rmarket,t− rf) + εi,t
Again, the first element represents the computed stock returns adjusted for the risk free rate; the alpha is the historical performance relative to the return and is assumed to be zero; and εi,t is the residual term. In the above equation index i refers to the respective company, and index t refers to the time period. β2 and β3 are the respective beta coefficients of the multiple announcement effect and the financial crisis effect, so that β1 – the one of interest – captures only the change of beta pertaining to the announcement concerned among changes caused by other effects. A comparable study with a slightly different research objective used a very similar regression model to estimate company betas (Brealey, Cooper & Kaplanis, 2010, p. 1720).
For the purpose of this research, the amount of leverage that a company has taken on does not require a variable. There may be other effects on beta than a change in leverage, but since this study investigates any possible total change in beta, including a variable for the effect of leverage would be irrelevant and would introduce bias into the regression coefficients. This follows from the reasoning that the intention of the takeover deal may be to bring the company's leverage to the desired level. The larger deals in particular should not be seen separately from the acquiring company's capital structure. In this way, beta could be affected directly as a consequence of the takeover, and then indirectly by the leverage.
2Year fixed effect dummy variables are generally considered more refined and accurate (Lazear, 2000, p. 1352). However, for the
purpose of this study they would provide biased results, since the financial crisis does not adhere to the calendar year. An example for clarification: when using a year fixed effect dummy variable, a crisis effect will be added for deals completed in 2008 before initiation of the financial crisis whereas this is not required.
11
2.4 Hypotheses
Finally, to assess whether the resulting company betas changed due to the announcement of a merger or acquisition, a statistical test needs to be performed. In line with the technique proposed by Brealey, Cooper and Kaplanis (2010), average pre- and post-deal betas will be compared using a paired t-test, firstly introduced by William Sealy Gosset (1925). The test statistic for this test can be calculated as follows:
t = ∑𝑑
√𝑛(∑𝑑2) − (∑𝑑)2 𝑛 − 1
In the above equation, the difference between all pairs of betas before announcement and betas after completion must be calculated. The element ∑𝑑 represents the sum of all differences between the pairs of variables in the two sample groups – estimated beta before announced date and after completion date. The bottom of the formula respectively uses also the sum of all squared differences (∑𝑑2) and the squared sum of all differences ((∑𝑑)2).
The number of observations is given by 𝑛, and the degrees of freedom used for this test is found by (𝑛 − 1). This test will be two-sided, to allow for changes of negative as well as positive nature. The hypotheses for this test are as follows:
H0: βbefore= βafter H1: βbefore≠ βafter
To put it into words, the null hypothesis supposes the betas before the announced date and those after completion date to be equal, whereas the alternative hypothesis formulates that both categories of betas differ significantly from each other. The two-sided t-test will provide a 95% confidence interval and if the test statistic does not fall in this range, the null hypothesis must be rejected. This would imply that the announcement of a takeover deal does actually influence the acquirer’s beta.
The most straightforward expectation in respect of the outcome of this study is to assume that the betas of the acquiring companies are not affected by the announcement of a merger or acquisition. The amount of total risk could possibly be changed due to the announcement, but this change concerns idiosyncratic risk and can thus be diversified. This does clearly not affect systematic risk and leaves beta therefore unchanged. This hypothesis is in line with the findings of Amihud, DeLong and Saunders (2002), and also with the earlier work of Madura and Wiant (1994).
12 Another possible hypothesis is that beta in fact does change; depending on the value of the beta of the target company, a decrease or increase of the acquiring firm’s beta could occur. Please note that this will only be the case if the target company is of considerable importance compared to the acquiring company, otherwise the target company’s beta will be of negligible influence. So after completion of the merger or acquisition the new beta value falls in between the two former values. Since one of the reasons behind undertaking mergers and acquisitions is decreasing the acquirer’s beta (Garabato, Gerpe & Maiztegui, 2008, p. 52), the above expectation may as well be a plausible one. This hypothesis finds support in the study of Hackbarth and Morellec (2008).
13
3 Methodology
3.1 Collecting data for M&A’s
In order to answer the above defined research question, data in respect of the mergers and acquisitions of interest is collected. Those mergers and acquisitions took place in the period 2006-2010 and the acquiring company must be listed on the AEX-index. The information is obtained by use of the database Zephyr. The search strategy will be described below.
‘Zephyr Advanced: For advanced queries and analysis’ is needed to gather the required information. Firstly, a time period is selected. The selected period runs from January 1st 2006 up to and including December 31st 2010. The selection has to be restricted to contain only completed-confirmed deals, in conformity with the assumption that as soon it becomes public a takeover deal is cancelled, any possible change in beta reverses. The second step was to specify the deal types of interest, which are mergers and acquisitions. Since the research will be done on mergers and acquisitions undertaken by AEX-listed companies, the AEX INDEX is selected under ‘stock data, indices’. In the field ‘Match’ only the box pertaining to ‘Acquirer’ needs to be ticked. Moreover, several options are added for
additional information. Under ‘add, deal, deal structure & dates’, the announced and completed date of the deals are added. Some deals have several announced dates. Only the first is retained, since this is considered to be the event that has potential impact on the firm’s systematic risk. Also some stock data about both the acquirer and target companies are considered relevant and are added to the deals for the purpose of further analysis. Under ‘add, company, stock data, acquirer stock data’ the boxes ‘acquirer listed’ and ‘acquirer ISIN number’ are ticked. This action is repeated for ‘target listed’ and ‘target ISIN number’. The above search strategy provides a list of 201 merger and acquisition deals.
3.2 Event windows and erratically behaving company returns
For all takeover deals an event window is defined. For practical reasons, the actual announcement dates of the mergers and acquisitions are set to have been announced at the first day of the concerning month and completion dates are set to have taken place at the first day of the next month. The first event window comprises a 30 months period prior to the announcement date for each takeover deal. Likewise, the second event window comprises the 30 months period after the completion date. In the period in between these dates, returns are often characterized by erratic behaviour. Following Brealey, Cooper and Kaplanis (2010), this transition period will thus be omitted from the analysis. An event window of 30 months using monthly returns yields 30 observations. According to the central limit theorem (Keller, 2012, p. 306)this is sufficient to perform a regression,
14 because this number of observations will yield an approximately normal distribution. For that reason it is possible to draw meaningful conclusions regarding the results of the regression.
3.3 Assigning dummy values
As mentioned before, dummy variables have to be generated to correct for any possible effect on beta of multiple announcements and/or the financial crisis. So, the individual mergers and acquisitions are analysed into more detail. To determine the value of the dummy variables, the following has to be considered. In case one or multiple other announcements by the same acquirer occurred in the event window of a certain deal, the following applies: for all later announced deals in the defined 30 months period the dummy variable will be set to 1 for the following months in the event window indicating that company stock prices are possibly already influenced by the other takeover announcements. The other months all receive value zero. A similar method is used for the values of the dummy variable that controls for any potential effect of the financial crisis. The individual months of the event window showing an overlap with the financial crisis received value 1, indicating any possible effect of the financial crisis on company betas during this period. Again, to the other months a value of zero is assigned. As specified above, the financial crisis is defined as the period between September 22nd of 2008 and December 21st of 2009.
3.4 Selection criteria
Now, a further selection is required. To be included in the sample, the deal value for the mergers and acquisitions as provided by Zephyr must be available. However, for a significant number of takeover deals this is not the case, and these are excluded from investigation. Furthermore, all acquiring companies must be listed not only for the complete duration of the specified period, but also during the event windows. Otherwise essential data will be missing and no regression can be performed for these deals. Lastly, when looking at the event windows of the deals of the individual companies, some of them show collinearity in their dummy variables. These 15 deals obviously had to be
excluded from the data set. This resulted in 12 out of 20 deals of the company Fugro being removed. For the remaining deals of Fugro, the elimination of these deals would have a significant impact in respect of assigning a value of 1 for the multiple announcement dummy variable to the individual months within the event windows. For this company, eliminating only the deals showing collinearity would therefore imply the selection of deals not being random anymore, and thus possibly
introducing bias into the test statistics. The above supports exclusion of all 20 takeover deals of the company Fugro. Several deals of other companies also have the same announced or completion dates, and thus identical event windows, but do not create collinearity. Therefore it can be
15 coefficients. The only effect of maintaining these deals could be that the standard errors of the estimated coefficients turn out higher. Applying the three selection criteria above results in a list of 59 deals qualifying for investigation. Table A in the appendix provides an overview of these deals and relevant information.
3.5 Collecting data about stock- and market prices
After the selection of deals qualifying for investigation, historical stock prices of the acquiring, AEX-listed companies are collected. These historical prices – adjusted for dividend payments and stock splits – are retrieved on a monthly basis via Datastream. Besides the companies’ historical stock prices, also the historical market prices of the S&P 350 – chosen representative of the market portfolio – are retrieved in the same manner. Based on the stock and market prices, the monthly returns are computed for the specified event windows. These returns are subsequently required to perform the ordinary least squares regressions in order to estimate the company betas before the announcement of the deal and after its completion.
Furthermore, an average yearly risk free rate for the period 2006-2010 is determined from the data of the Agentschap van de Generale Thesaurie (Agentschap van de Generale Thesaurie – Ministerie van Financiën, 2014). This is done by calculating the average of the interest rate on government bonds with a maturity of 10 years issued on the first and last day of the period. Berk and DeMarzo (2011) justify the use of the interest rate on government bonds, on the basis that a majority of companies and financial analysts makes use of it as well. Subsequently, the determined yearly risk free rate is converted to a monthly rate in order to fit the regression. This yields a monthly risk free rate of 0,270%.
3.6 Running regressions and conducting a t-test
Having collected and structured the data, the ordinary least squares regressions are performed to estimate the company betas for each deal in both periods. The software package STATA is the statistical tool used to run the regressions. These are performed with robust standard errors to avoid that the standard errors may be biased. If not robust, standard errors may be misleading if there is non-zero correlation or heteroskedasticity in residual terms. The estimated company betas over the two event windows before announced date and after completed date – together with their standard errors, t-value and p-value – are summarized in Table I and II below. The numbering of the deals in both tables corresponds to the sequence of deals as listed in Table A in the appendix.
16 Table I: Estimated company betas before announced date
Takeover Takeover Takeover
deal coefficient std. err. t-value p-value deal coefficient std. err. t-value p-value deal coefficient std. err. t-value p-value 1 1,181229 0,4827788 2,45 0,021 21 0,6513201 0,4316784 1,51 0,143 41 0,5504608 0,254465 2,16 0,04 2 1,280784 0,3843571 3,33 0,002 22 2,039088 0,3228809 6,32 0 42 1,561319 0,4517119 3,46 0,002 3 1,087196 0,4378175 2,48 0,02 23 2,015662 0,2474428 8,15 0 43 1,980871 0,1568299 12,63 0 4 0,8502469 0,323919 2,62 0,014 24 1,811904 0,3587081 5,05 0 44 1,279339 0,6071591 2,11 0,045 5 1,807714 0,3338903 5,41 0 25 2,059714 0,318495 6,47 0 45 0,8205062 0,4184072 1,96 0,06 6 1,968019 0,5431023 3,62 0,001 26 1,339673 0,4483853 2,99 0,006 46 1,554184 1,010071 1,54 0,136 7 1,557616 0,4209998 3,7 0,001 27 1,505608 0,8321765 1,81 0,082 47 1,241511 0,6114329 2,03 0,052 8 1,636991 0,304628 5,37 0 28 1,663348 0,6960861 2,39 0,024 48 1,149283 0,5865507 1,96 0,06 9 0,4883439 0,4731468 1,03 0,312 29 2,037091 1,05892 1,92 0,065 49 2,352671 0,609662 3,86 0,001 10 0,6030654 0,3660764 1,65 0,111 30 0,7270543 0,830776 0,88 0,39 50 1,778933 0,3343893 5,32 0 11 0,0376762 0,3053394 0,12 0,903 31 0,0284446 1,037134 0,03 0,978 51 1,011444 0,608481 1,66 0,108 12 0,5483632 0,3736185 1,47 0,154 32 2,105992 0,9837894 2,14 0,041 52 1,015803 0,3392854 2,99 0,006 13 1,757238 0,4151602 4,23 0 33 0,726886 0,8304636 0,88 0,389 53 2,047293 0,5048326 4,06 0 14 1,876793 0,4002665 4,69 0 34 1,662167 1,093372 1,52 0,141 54 1,256118 0,7293954 1,72 0,096 15 2,257887 0,5350528 4,22 0 35 -0,0089552 1,091276 -0,01 0,994 55 1,498141 0,6050951 2,48 0,02 16 1,397858 0,7063095 1,98 0,058 36 1,798556 0,4640266 3,88 0,001 56 2,174513 0,584309 3,72 0,001 17 1,825856 0,4338288 4,21 0 37 0,8484445 0,3836738 2,21 0,036 57 1,448831 0,2358238 6,14 0 18 1,270457 0,2156122 5,89 0 38 0,8342164 0,3656176 2,28 0,031 58 1,728672 0,2927129 5,91 0 19 0,2053026 0,603337 0,34 0,736 39 0,8446384 0,3579662 2,36 0,026 59 0,9318547 0,1609141 5,79 0 20 0,6561814 0,4223158 1,55 0,131 40 0,7654882 0,406031 1,89 0,07
17 Table II: Estimated company betas after completed date
Takeover Takeover Takeover
deal coefficient std. err. t-value p-value deal coefficient std. err. t-value p-value deal coefficient std. err. t-value p-value 1 1,537025 0,2681948 5,73 0 21 0,9511234 0,5123785 1,86 0,075 41 -0,0672339 0,2651012 -0,25 0,802 2 1,543604 0,3287493 4,7 0 22 -0,0582191 0,3914943 -0,15 0,883 42 0,746381 0,4844658 1,54 0,135 3 0,8209167 0,2160172 3,8 0,001 23 1,410132 0,5370197 2,63 0,014 43 0,6777428 0,3368902 2,01 0,054 4 0,2283201 0,1669119 1,37 0,183 24 0,8958411 0,3035407 2,95 0,006 44 2,229731 0,471903 4,72 0 5 2,169012 0,4418277 4,91 0 25 0,9091079 0,2522515 3,6 0,001 45 1,140625 0,3598475 3,17 0,004 6 2,031291 0,4732006 4,29 0 26 0,8877551 0,2702584 3,28 0,003 46 1,787777 0,4216875 4,24 0 7 2,141608 0,465631 4,6 0 27 0,3302701 0,2044231 1,62 0,118 47 2,229731 0,471903 4,72 0 8 0,4139527 0,6860073 0,6 0,551 28 0,3549756 1,410849 0,25 0,803 48 2,45211 0,4822926 5,08 0 9 1,142169 0,2489158 4,59 0 29 0,3549756 1,410849 0,25 0,803 49 1,678541 0,30237 5,55 0 10 1,9332 0,4409819 4,38 0 30 -0,4191423 1,417043 -0,3 0,77 50 1,775392 0,4719119 3,76 0,001 11 0,8456925 0,5103909 1,66 0,11 31 1,271174 0,7109446 1,79 0,085 51 1,04692 0,2773225 3,78 0,001 12 0,6028755 0,3660344 1,65 0,111 32 0,6076884 0,3206877 1,89 0,069 52 0,9556414 0,3127099 3,06 0,005 13 1,76711 0,281968 6,27 0 33 -3,829829 1,498492 -2,56 0,017 53 1,223989 0,2538953 4,82 0 14 1,558767 0,2902766 5,37 0 34 2,413667 0,2705316 8,92 0 54 1,444681 0,8546361 1,69 0,103 15 1,801512 0,5560923 3,24 0,003 35 1,066146 0,7585564 1,41 0,172 55 1,397159 0,786005 1,78 0,087 16 1,636341 0,2955787 5,54 0 36 1,373637 0,1765882 7,78 0 56 1,909699 0,4147265 4,6 0 17 1,671337 0,2431501 6,87 0 37 0,5901782 0,2878738 2,05 0,051 57 1,928481 0,9376056 2,06 0,05 18 0,2724285 0,2513208 1,08 0,288 38 0,3414631 0,29066 1,17 0,251 58 0,9724068 0,225152 4,32 0 19 0,9068956 0,2280536 3,98 0 39 0,6354136 0,2360093 2,69 0,012 59 0,7840779 0,153564 5,11 0 20 1,206502 0,5601005 2,15 0,041 40 0,6354136 0,2360093 2,69 0,012
18 Finally, a paired t-test as described in theoretical section 2.4 is conducted. It can be assumed that the variances of the pairs are equal, for they are derived from the same group of data. Each company is used as its own control, so random between-company variation has been eliminated. The results of the test can be found in Table III. The table shows the test statistics and the p-value 0,0670 of the two-sided test.
Table III: Paired t-test on betas; complete sample set
Variable Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] BetasBefore 59 1,307337 0,0790705 0,6073518 1,149061 1,465614 BetasAfter 59 1,072308 0,121483 0,9331287 0,8291335 1,315483 Difference 59 0,2350292 0,125986 0,9670254 -0,016979 0,4870374
t = 1,8669 degrees of freedom = 58
H0: βbefore= βafter H1: βbefore≠ βafter
19
4 Analysis
4.1 Outcome paired t-test
As can be seen from Table III, the betas of the companies investigated did not significantly change as a result of the announced mergers and acquisitions. The outcome of the conducted paired t-test shows that the betas estimated in the event window before announced date do not differ
significantly from those estimated after completed date. The p-value 0,0670of the two-sided test is above the significance level of 0,05, and thus on the basis of this test the null hypothesis cannot be rejected in favour of the alternative hypothesis. The betas of both event windows of all deals in the data set show that there was no significant change. Another indication that the null hypothesis cannot be rejected is given by the 95% confidence interval. The difference between the two tested variables lies with 95% certainty within the interval: [-0,016979 ; 0,4870374].
4.2 Some remarks
The outcome of the paired t-test implies that company betas remain unchanged after the
announcement of a takeover deal. However, it should be noted that for a majority of the cases in this study the deal value represents only a very small percentage of the value of the acquirer. In Table A (to be found in the appendix) the deal value is compared to the acquirer’s market capitalization at the announced date. The market value – i.e. the present value of all future economic benefits – would be a more accurate measure for this purpose, but since this value is not available on the exact announced dates, market capitalization is used instead. For only 13 out of 59 takeovers the deal value represents 5% or more of the acquirer’s company value. Consideration should be given that the other deals could be too small to be of any possible and significant influence on the acquirer’s beta. This follows from the assumption that undertaking a relatively small deal with respect to the acquirer’s market capitalization does not change the company’s capital structure. As indicated in section 2.1, a company’s systematic risk is to a large extent determined by the capital structure. To verify whether the large number of small deals influences the test statistics too much, another paired t-test is performed. Table IV below contains the outcome of a test on the betas of only the deals with a deal value that make up for 5% or more of the acquirer’s market capitalization.
20 Table IV: Paired t-test on betas; deals representing 5% or more of acquirer’s value
Variable Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] BetasBefore 13 1,151225 0,1914181 0,6901679 0,7341606 1,568289 BetasAfter 13 1,342793 0,1391816 0,5018263 1,039542 1,646043 Difference 13 -0,1915679 0,2203223 0,7943834 -0,671609 0,2884732
t = −0,8695 degrees of freedom = 12
H0: βbefore= βafter H1: βbefore≠ βafter
Pr([T] > [t]) = 0,4016
For this test as well, the p-value 0,4016 is higher than the significance level of 0,05. Consequently, neither in this case the null hypothesis can be rejected, and beta appears to be unchanged on average. In addition, the standard error is quite high, probably as a consequence of the low number of observations. However, when looking at the estimated betas for the deals that make up for 5% or more of the acquirer’s value, the change of beta is not unambiguous; beta equally often tends to increase as to decrease. Table V lists the estimated coefficients for beta of these deals in both event windows. Note that the numbering refers to the respective takeover deals as listed in Table A in the appendix.
Table V: Estimated betas before announced date and after completed date; deals representing 5% or more of acquirer’s value
To test if the paired t-tests show a non-significant outcome because possibly the change in increasing betas offsets a change in decreasing betas, a split sample test is conducted. For this purpose, the original data set is split in two subsets: one contains the deals in which the beta increased after the takeover and a second subset containing all deals in which the beta shows a decrease after the
Takeover beta-coefficient 1st beta-coefficient 2nd
deal event window event window
1 1,181229 1,537025 3 1,087196 0,8209167 9 0,4883439 1,142169 10 0,6030654 1,9332 13 1,757238 1,76711 19 0,2053026 0,9068956 31 0,0284446 1,271174 36 1,798556 1,373637 42 1,561319 0,746381 43 1,980871 0,6777428 44 1,279339 2,229731 45 0,8205062 1,140625 56 2,174513 1,909699
21 takeover. Then for each deal in both subsets is verified whether the estimated beta of the second event window falls in the confidence interval of the beta coefficient of the first event window, in which case no significant change of beta is observable. If not however, with a confidence level of 95% it can be assumed that beta after completion date does significantly differ from beta before
announced date. From Table VI on the next page it can be seen that for a majority of the deals (46 out of 59) the estimated beta of the second event window has a value that falls in the 95%
confidence level of the beta coefficient of the first event window. This confirms the outcome of the both paired t-tests that were conducted, implying no increasing betas offset decreasing betas in the statistical tests.
22 Table VI: Split sample on betas
Takeover Confidence interval of beta Beta coeffiecient Takeover Confidence interval of beta Beta coeffiecient Takeover Confidence interval of beta Beta coeffiecient deal coefficient 1st event window 2nd event window deal coefficient 1st event window 2nd event window deal coefficient 1st event window 2nd event window
1 0,1906484 2,171809 1,537025 21 -0,2344109 1,537051 0,9511234 41 0,0283416 1,07258 -0,0672339 2 0,4934643 2,068104 1,543604 22 1,376591 2,701584 -0,0582191 42 0,6328118 2,489826 0,746381 3 0,1872491 1,987143 0,8209167 23 1,508799 2,522526 1,410132 43 1,658502 2,303239 0,6777428 4 0,1867289 1,513765 0,2283201 24 1,074569 2,549239 0,8958411 44 0,0335516 2,525127 2,229731 5 1,122628 2,4928 2,169012 25 1,406216 2,713211 0,9091079 45 -0,0365621 1,677574 1,140625 6 0,853665 3,082373 2,031291 26 0,4180034 2,261342 0,8877551 46 -0,518311 3,626678 1,787777 7 0,6937957 2,421436 2,141608 27 -0,2018767 3,213093 0,3302701 47 -0,0130462 2,496067 2,229731 8 1,012989 2,260993 0,4139527 28 0,2350972 3,091599 0,3549756 48 -0,0542198 2,352785 2,45211 9 -0,4842232 1,460911 1,142169 29 -0,1356329 4,209816 0,3549756 49 1,103835 3,601507 1,678541 10 -0,1480613 1,354192 1,9332 30 -0,9806302 2,434739 -0,4191423 50 1,093967 2,463898 1,775392 11 -0,5877832 0,6631355 0,8456925 31 -2,096028 2,152917 1,271174 51 -0,2370554 2,259944 1,04692 12 -0,2182388 1,314965 0,6028755 32 0,0874234 4,124562 0,6076884 52 0,3208085 1,710798 0,9556414 13 0,9054001 2,609077 1,76711 33 -0,9801564 2,433928 -3,829829 53 1,013191 3,081396 1,223989 14 1,054034 2,699553 1,558767 34 -0,5852921 3,909625 2,413667 54 -0,2404774 2,752714 1,444681 15 1,161881 3,353893 1,801512 35 -2,248068 2,230158 1,066146 55 0,2586599 2,737622 1,397159 16 -0,0539821 2,849698 1,636341 36 0,8480403 2,749071 1,373637 56 0,9756102 3,373416 1,909699 17 0,9357132 2,716 1,671337 37 0,0612109 1,635678 0,5901782 57 0,965768 1,931894 1,928481 18 0,8280572 1,712856 0,2724285 38 0,0840311 1,584402 0,3414631 58 1,128075 2,329269 0,9724068 19 -1,032643 1,443248 0,9068956 39 0,1113779 1,577899 0,6354136 59 0,602237 1,261472 0,7840779 20 -0,2088933 1,521256 1,206502 40 -0,0676185 1,598595 0,6354136
23 Furthermore, this study illustrates some other points. In almost all performed regressions α is
estimated to be not significantly different from zero. This implies that the calculated risk free rate for the period is appropriately chosen. Moreover, it implies there are no abnormal returns to be earned. Unless several estimation deviations of opposite effects cancel each other out, but no indication is found, which makes this scenario highly improbable. Table C and D in the appendix summarize the estimated alphas together with their standard error and p-value for both event windows of each takeover deal. In addition, the estimated coefficient for the multiple announcement effect appears to be present in only six out of 68 performed regressions where this dummy variable had to be included – before announced date as well as after completed date. And for a total of 68 regressions where the dummy variable for any change in beta caused by the financial crisis had to be taken into account, 16 estimated coefficients were significantly different from zero. Apparently, for the majority of the deals neither multiple announcements nor the financial crisis has been of any significant influence on the company beta. The low score of significant coefficients for the multiple announcement variable and the financial crisis variable may indicate that not even a severe market shock such as the recent financial crisis has an effect on a company’s systematic risk, and thus beta not being time-varying at all, but rather unchanging irrespective of economic circumstances. All estimated values for the dummy variables are recorded in Table E and F – standard errors and p-values are included as well.
4.3 Validation of hypotheses
The results are consistent with the formulated hypothesis of market shocks having no influence at all on the company’s systematic risk. The outcome of the empirical research conducted is that beta remains fixed despite an announcement of a takeover. This study thus validates the earlier work of Madura and Wiant (1994), and likewise that of Amihud, DeLong and Saunders (2002). The findings of this study correspond most to those of the former mentioned duo, since they investigated the long-term effect of an takeover announcement by using monthly data, whereas Amihud, DeLong and Saunders (2002) made use of daily data and hence focused on the immediate market reaction.
On the other hand, like the last column of Table 1 shows, most takeover deals are relatively small in comparison to the market value of the acquiring company. A vast majority represents less than 5% of the acquirer’s market capitalization. Therefore, on the basis of the firstly conducted paired t-test that takes into account all deals of the sample set, the alternative hypothesis cannot be ruled out. This hypothesis specifies that beta will take a value in between the former two betas of both the target and acquiring company in case the target company is of considerable importance compared to the acquirer. To see whether this is true, a second paired t-test on only the betas of the relatively big takeover deals was conducted. The test statistics proved that beta remained the same on average,
24 and therefore imply beta to be stable over time with respect to the announcement of mergers and acquisitions. However, a number of only 13 out of 59 deals representing 5% or more of the acquirer’s market capitalization is too low to generalize the results of this test. Furthermore, beta does not seem to behave unambiguous. Some companies show a lower beta after completion of the deal than before announcement, others on the contrary show a higher beta afterwards than before. Since a paired t-test provides an average result, it cannot be claimed that none of the changes in beta is significant. A final test is performed on the betas per deal to verify whether the outcome of no significant change in beta is the result of increasing and decreasing betas canceling out. The low number of takeover deals for which the value of beta did significantly change according to this test, confirms the outcome of the both paired t-tests.
4.4 Explanation
The paired t-test proved that beta did not change on average due to announcements of mergers and acquisitions – as well for the complete sample set of takeover deals as for the relatively big deals only. Mergers and acquisitions with a deal value below 5% of the acquirer’s value are assumed to have a negligible effect on the acquirer’s systematic risk, for the reason that they do not alter the company’s capital structure. Therefore, the outcome of the test on the complete data set that beta did not significantly change is not astonishing. However, the outcome of the secondly performed paired t-test is unexpected, because, as Table V shows, the beta estimated in the first event window differs from the one in the second event window. Lastly, a split sample test is performed to
investigate any possible change in beta per individual takeover deal instead of an average
investigation on the entire data set. This test as well shows no significant change of beta for a vast majority of the deals investigated. Beta does not always increase or decrease after the
announcement of a takeover deal; its behaviour is clearly ambiguous. A possible explanation is the opinion of investors about the takeover. They may feel positive about the by the company adopted strategy or disagree with the merger or acquisition, causing respectively a decrease or increase of the company’s beta.
25
5 Conclusion
Some economist assume a company’s beta can vary over time as a result of severe market shocks and other events that affect the company as a whole, whereas others are convinced betas cannot fluctuate at all, regardless changes in the market environment. This study tried to answer the question how beta of AEX-listed companies in the period 2006-2010 is affected by the announcement of a merger or acquisition.
By running regressions and then conducting a paired t-test, research showed that on average the announcement of a takeover did not have any effect on a company’s systematic risk. The split sample test provided the same result for the majority of the takeover deals included in the data set. This is in line with the formulated hypothesis that beta is not time-varying with respect to the announcement of mergers and acquisitions. However, one should bear in mind that for most takeovers investigated the deal value is small in comparison to the acquirer’s value, and in that case it is unlikely that beta will change as a result of the announcement, since the target company is of negligible importance compared to the acquirer. The low number of deals of considerable importance – and thus of possible influence on the acquirer’s systematic risk – places a limitation on this study. It is not possible to generalize the results obtained. The limitation can be eliminated by repeating this investigation for a higher number of takeovers for which the deal value represents at least 5% of the acquirer’s value. Only then it will be justifiable to draw general conclusions about the time-varying of beta as a result of the announcements of mergers and acquisitions.
26
References
Agentschap van de Generale Thesaurie (2014). Interest rate series of Dutch obligation bonds. Source: http://dsta.nl/Actueel/Statistieken/Rentereeksen_Nederlandse_staatsobligaties.
Alam, N. & Rajjaque, M.S. (2010). Shariah-compliant equities: empirical evaluation of performance in the European market during credit crunch. Journal of Financial Services Marketing, 15, 3, 228-240.
Allen, M.T., Madura, J. & Springer, T.M. (2000). REIT Characteristics and the Sensitivity of REIT Returns. Journal of Real Estate Finance and Economics, 21, 2, 141-152.
Ambrose, B.W. & Linneman, P.D. (2001). REIT Organizational Structure and Operating Characteristics.
Journal of Real Estate Research, 21, 3, 141-161.
Amihud, Y., DeLong, G.L. & Saunders, A. (2002). The effects of cross-border bank mergers on bank risk and value. Journal of International Money and Finance, 21, 857-877.
Barberis, N., Shleifer, A. & Wurgler, J. (2005). Comovement. Journal of Financial Economics, 75, 2, 283-317.
Berk, J. & DeMarzo, P. (2011). Corporate Finance (2nd ed.). Edinburgh Gate: Pearson Education Limited.
Brealey, R.A., Cooper, I.A. & Kaplanis, E. (2010). Excess Comovement in International Equity Markets: Evidence from Cross-border Mergers. The Review of Financial Studies, 23, 4, 1718-1740. Bos, T. & Newbold, P. (1984). An Empirical Investigation of the Possibility of Stochastic Systematic
Risk in the Market Model. The Journal of Business, 57, 35-41.
Choudhry, T. (2002). Time-varying beta and the Asian financial crisis: Evidence from Malaysian and Taiwanese firms. Pacific-Basin Finance Journal, 13, 1, 93-118.
Conover, C.M., Friday, H.S. & Howton, S. (1998). The Relationship between Size and Return for Foreign Real Estate Investments. Journal of Real Estate Portfolio Management, 4, 2, 107-112. Delcoure, N. & Dickens, R. (2004). REIT and REOC Systematic Risk Sensitivity. Journal of Real Estate
Research, 26, 3, 237-254.
Fabozzi, F. & Francis, J. (1978). Beta as a random coefficient. Journal of Financial and Quantitative
Analysis, 13, 101-116.
Fama, E.F. & French, K.R. (2004). The Capital Asset Pricing Model: Theory and Evidence. Journal of
Economic perspectives, 18, 3, 25-46.
Fischer, R.A. (1925). Applications of ‘Student’s’ Distribution. Metron, 5, 90-104, Collected papers, 2,
no. 43.
Garabato, N., Gerpe, J. & Maiztegui, A. (2008). Can mergers and acquisitions help companies manage market risk? Watson Wyatt’s Horizons, 3, 52-58.
27 Grass, G. (2010). The Impact of Conglomeration on the Option Value of Equity. Journal of Banking
and Finance, 34, 12, 3010-3024.
Gyuorko, J. & Nelling, E. (1996). Systematic Risk and Diversification in the Equity REIT Market. Real
Estate Economics, 24, 4, 493-515.
Hackbarth, D. & Morellec, E. (2008). Stock Returns in Mergers and Acquisitions. The Journal of
Finance, 63, 3, 1213-1252.
Keller, G. (2012). Managerial Statistics (9th ed.). Cengage Learning.
Lau, S.T., Lilian, N. & Zhang, B. (2010). The World Price of Home Bias. Journal of Financial Economics,
97, 2, 191-217.
Lazear, E.P. (2000). Performance pay and productivity. American Economic Review, 90, 5, 1346-1361. Lee, C.L., Reed, R. & Robinson, J. 2008. Listed Property Trusts and Downside Systematic Risk
Sensitivity. Journal of Property Investment & Finance, 26, 4, 304-328.
Litt, J., Mei, J. & Webber, P. (1999). A Risk Adjustment Model for REIT Evaluation. Real Estate
Finance, 16, 1, 9-19.
Madura, J. & Wiant, K.J. (1994). Long-term valuation effects of bank acquisitions. Journal of Banking
& Finance, 18, 1135-1154.
Mandelker, G. (1974). Risk and Return: The Case of the Merging Firm. Journal of Financial Economics,
1, 4, 303-335.
Markowitz, H. (1952). Portfolio Selection. The Journal of Finance, 7, 1, 77-91.
Olsen, R.A. & Patel, R.C. (1984). Financial Determinants of Systematic Risk in Real Estate Investment Trusts. Journal of Business Research, 12, 4, 481-491.
Van Ophem, H. (2012). Syllabus Research Project Economics and Business (ed. unknown). Amsterdam: Faculty Economics and Business, University of Amsterdam.
Sharpe, W.F. (1964). Capital Asset Prices: A Theory of Market Equilibrium Under Conditions of Risk.
The Journal of Finance, 19, 3, 425-442.
S&P Dow Jones Indices (2014). Index performances. Source: http://us.spindices.com/indices/equity/sp-europe-350.
S&P Europe 350 Constituent List, Standard and Poor’s (2014). Financial market indicators. Source: http://www.standardandpoors.com/en_EU/web/guest/home.
Stuiveling, S.J. & Van Schoten, E.M.A. (2010). Krediet crisis: interventies en vervolg. Derde rapportage:
vierde kwartaal 2009 en eerste kwartaal 2010. Algemene Rekenkamer, Tweede Kamer der
Staten-Generaal.
Tan, Y.K. (2004). The Effect on Management Structure on the Performance of Listed Property Trusts.
28 Vijh, A.M. (1994). S&P 500 Trading Strategies and Stock Betas. Review of Financial Studies, 7, 1,
29
Appendix
Table A provides an overview of the takeover deals included in the sample set and relevant
information that is used for this research study. Table B and C summarize the estimated alphas per deal – together with their standard error and p-value – for respectively the event window before announced date and after completed date. Lastly, Table D and E display the estimated coefficients for the dummy variables that have to control for any multiple announcement and/or financial crisis effect. For Tables B-E holds that all values are rounded to four decimal places. P-values are included in the tables, and the standard errors are put in parentheses in Tables B and C. The numbering of the deals in all tables – in the working paper as well as in the appendix – corresponds to the sequence of deals as listed in Table A.
30 Table A: Selected deals
Deal Number Acquiror name Target name Announced Completed Deal value Hist. mkt capitalization Deal value/mkt
date date EUR (th) acquirer EUR (mln) capitalization %
1 432400 Akzo Nobel N.V. Imperial Chemical Industries PLC 13.08.2007 02.01.2008 10.888.399,12 16.610,84 66%
2 434816 Akzo Nobel N.V. Sico Inc. 05.04.2006 10.07.2006 201.707,97 12.533,24 2%
3 1601119158 Heineken N.V. FEMSA Cerveza S.A. de C.V. 11.01.2010 30.04.2010 4.434.177,00 16.659,13 27%
4 643929 Heineken N.V. SP Rechitsapivo OAO 22.05.2008 01.08.2008 6.370,65 18.084,96 0%
5 491805 ING Groep N.V. Oyak Bank A.S. 19.06.2007 08.07.2008 1.702.968,30 73.809,88 2%
6 639112 ING Groep N.V. CitiStreet LLC 02.05.2008 01.07.2008 570.780,00 56.414,41 1%
7 650971 ING Groep N.V. Oyak Emeklilik A.S. 17.06.2008 04.12.2008 110.000,00 51.745,95 0%
8 593985 ING Groep N.V. Chongqing Longhu Real Estate Development Co., Ltd 31.01.2007 31.01.2007 20.356,20 73.978,38 0%
9 1601145561 Corio N.V. Multi Corporation BV'S Certain Portfolio Assets in Germany 25.03.2010 25.03.2010 1.300.000,00 3.507,35 37%
10 1601090850 Corio N.V. Príncipe Pío Gestion S.A. 25.06.2009 25.06.2009 183.000,00 2.550,91 7%
11 429898 Corio N.V. Outlet Marques Avenue's Cholet 23.03.2006 23.03.2006 26.000,00 3.634,97 1%
12 508539 Corio N.V. Mei Lulin Ood 16.11.2006 16.11.2006 15.500,00 3.819,92 0%
13 523187 AEGON N.V. OPTAS N.V. 15.03.2007 28.06.2007 1.300.000,00 23.662,27 5%
14 1601020368 AEGON N.V. Mongeral Seguros e Previdência S.A. 01.10.2008 29.05.2009 55.000,00 9.439,53 1%
15 397157 AEGON N.V. Seguros Navarra S.A. 22.11.2005 10.05.2006 42.000,00 21.154,46 0%
16 1601041098 AEGON N.V. BT AEGON Fond De Pensii S.A. 07.01.2009 23.06.2009 11.000,00 8.206,78 0%
17 522214 AEGON N.V. Clarke Inc. 20.02.2007 13.03.2007 231.857,60 25.171,59 1%
18 1601059997 Unilever N.V. Sara Lee Household and Bodycare International B.V. 25.09.2009 06.12.2010 1.200.000,00 30.212,64 4%
19 1601157496 Unibail-Rodamco SE Simon Ivanhoe 05.02.2010 15.07.2010 715.000,00 14.260,18 5%
20 449822 Unibail-Rodamco SE Société de Tayninh S.A. 12.06.2006 08.08.2006 2.342,00 5.832,70 0%
21 449837 Unibail-Rodamco SE Société de Tayninh S.A. 18.07.2006 04.09.2006 1.458,00 6.235,88 0%
22 445215 Koninklijke Philips Electronics N.V. AVENT Holdings Ltd 23.05.2006 04.09.2006 682.902,86 32.625,36 2%
23 308475 Koninklijke Philips Electronics N.V. Partners in Lighting International N.V. 13.11.2006 06.02.2007 590.000,00 34.862,89 2%
24 1601095791 Koninklijke Philips Electronics N.V. InnerCool Therapies Inc.'s assets 15.07.2009 24.07.2009 7.949,97 14.197,21 0%
25 547837 Koninklijke Philips Electronics N.V. Color Kinetics Inc. 19.06.2007 27.08.2007 578.616,50 35.907,59 2%
26 1601064899 Koninklijke Philips Electronics N.V. Saeco International Group S.p.A. 25.05.2009 27.07.2009 200.000,00 13.501,93 1%
27 615235 ArcelorMittal S.A. Shaktar Pervomaiskaya OAO 01.02.2008 10.04.2008 414.280,80 68.283,13 1%
28 595995 ArcelorMittal S.A. Kalagadi Manganese (Pty) Ltd 20.11.2007 03.09.2008 298.944,00 70.470,81 0%
29 1601007388 ArcelorMittal S.A. Koppers Holdings Inc.'s Monessen Metallurgical Coke Plant 04.08.2008 01.10.2008 113.568,00 77.352,75 0%
30 1601076834 ArcelorMittal S.A. Noble European Holdings B.V. 08.05.2009 21.07.2009 78.295,71 30.903,43 0%
31 409928 Arcelor S.A. Dofasco Inc. 14.01.2006 09.03.2006 4.009.969,60 5.935,53 68%
32 629739 ArcelorMittal S.A. China Oriental Group Co., Ltd 14.01.2008 04.02.2008 327.875,05 67.095,06 0%
33 1601075949 ArcelorMittal S.A. ArcelorMittal Ostrava a.s. 07.05.2009 28.07.2009 275.054,99 29.425,65 1%
34 1601128392 ArcelorMittal S.A. ArcelorMittal Ostrava a.s. 12.11.2009 29.01.2010 261.867,67 38.616,99 1%
31 Table A: Selected deals, continued
Deal Number Acquiror name Target name Announced Completed Deal value Hist. mkt capitalization Deal value/mkt
date date EUR (th) acquirer EUR (mln) capitalization %
36 372895 Wolters Kluwer N.V. NDCHealth Corporation's Pharmaceutical Information Management Business 29.08.2005 06.01.2006 315.684,80 4.678,64 7%
37 476950 Koninklijke KPN N.V. Tiscali B.V. 15.09.2006 19.06.2007 236.000,00 20.850,86 1%
38 447468 Koninklijke KPN N.V. TDINL B.V. 02.06.2006 15.06.2006 69.000,00 19.362,24 0%
39 445286 Koninklijke KPN N.V. Enertel N.V. 24.05.2006 31.10.2006 10.000,00 18.802,88 0%
40 481212 Koninklijke KPN N.V. Siemens Nederland B.V.'s Enterprise Networks Division 04.10.2006 04.10.2006 5.000,00 21.101,82 0%
41 1601137117 Koninklijke KPN N.V. IBASIS INC. 23.11.2009 21.12.2009 65.212,90 19.750,08 0%
42 1601028452 Imtech N.V. NVS Installation AB 04.11.2008 01.12.2008 235.000,00 1.161,50 20%
43 1601198946 Imtech N.V. NEA Gruppen AB 23.06.2010 30.07.2010 103.260,70 1.940,12 5%
44 529073 Imtech N.V. Peek Traffic Holdings Ltd 05.04.2007 30.04.2007 80.000,00 1.465,58 5%
45 428249 Imtech N.V. Radio Holland Group B.V. 16.03.2006 11.05.2006 47.000,00 1.040,24 5%
46 623478 Imtech N.V. Pertec Pty Ltf 26.02.2008 26.02.2008 45.000,00 1.293,78 3%
47 424086 Imtech N.V. The Aqua Group Ltd 01.05.2007 01.05.2007 24.000,00 1.587,65 2%
48 468824 Imtech N.V. Seacoast Electronics Inc. 05.03.2007 05.03.2007 4.500,00 1.333,03 0%
49 504725 ASML Holding N.V. Brion Technologies Inc. 19.12.2006 08.03.2007 205.551,00 9.092,52 2%
50 548032 Air France-KLM S.A. VLM Airlines NV 24.12.2007 17.03.2008 180.000,00 7.235,27 2%
51 505502 TNT N.V. Expresso Mercúrio S.A. 10.01.2007 10.01.2007 151.000,00 14.981,63 1%
52 400624 TNT N.V. Hoau Logistics Group 06.12.2005 15.03.2007 102.262,50 11.484,38 1%
53 446690 Koninklijke Ahold N.V. Konmar B.V.'s 29 Superstores 31.05.2006 14.11.2006 101.000,00 9.921,70 1%
54 574400 Randstad Holding N.V. Team BS Betriebs-Service GmbH Gesellschaft für Zeitarbeit 14.09.2007 30.09.2007 71.000,00 4.279,06 2%
55 423716 Randstad Holding N.V. PinkRoccade N.V.'s Human Resources Services 17.08.2006 02.10.2006 65.000,00 4.815,32 1%
56 415278 Randstad Holding N.V. Vedior N.V. 01.04.2008 01.07.2008 2.853.002,00 3.383,93 84%
57 480290 Koninklijke DSM N.V. Lipid Technologies Provider AB 29.09.2006 29.09.2006 18.000,00 6.985,55 0%
58 523196 Koninklijke DSM N.V. Pamako AG 15.03.2007 15.03.2007 15.000,00 6.539,02 0%
32 Table B: Estimated alphas before announced date
Takeover Takeover Takeover
deal coefficient p-value deal coefficient p-value deal coefficient p-value 1 0,0052 (0,0118)a 0,663 21 0,0120 (0,0136) 0,387 41 0,0042 (0,0119) 0,73 2 -0,0040 (0,0106) 0,706 22 -0,0229 (0,0077) 0,006 42 0,0183 (0,0145) 0,217 3 0,0042 (0,0104) 0,691 23 -0,0211 (0,0063) 0,002 43 0,0281 (0,0146) 0,066 4 0,0070 (0,0089) 0,437 24 -0,0105 (0,0143) 0,47 44 0,0108 (0,0099) 0,286 5 -0,0103 (0,0077) 0,193 25 -0,0128 (0,0076) 0,104 45 0,0117 (0,0126) 0,361 6 -0,0096 (0,0100) 0,342 26 -0,0035 (0,0152) 0,818 46 0,0063 (0,0119) 0,603 7 -0,0101 (0,0095) 0,299 27 0,0184 (0,0181) 0,32 47 0,0126 (0,0103) 0,234 8 -0,0046(0,0076) 0,549 28 0,0163 (0,0203) 0,43 48 0,0100 (0,0101) 0,334 9 -0,0041 (0,0083) 0,621 29 0,0216 (0,0166) 0,204 49 -0,0193 (0,0149) 0,207 10 -0,0051 (0,0103) 0,625 30 0,0113 (0,0180) 0,535 50 0,0031 (0,0124) 0,801 11 0,0210 (0,0086) 0,022 31 0,0741 (0,0403) 0,077 51 0,0057 (0,0069) 0,415 12 0,0091 (0,0085) 0,296 32 0,0202 (0,0186) 0,287 52 0,0001 (0,0057) 0,985 13 -0,0067 (0,0089) 0,457 33 0,0113 (0,0180) 0,535 53 -0,0270 (0,0110) 0,021 14 -0,0162 (0,0072) 0,033 34 0,0074 (0,0207) 0,725 54 -0,0089 (0,0150) 0,558 15 -0,0177 (0,0109) 0,116 35 0,0688 (0,0407) 0,103 55 0,0057 (0,0123) 0,645 16 0,0002 (0,0185) 0,993 36 -0,0174 (0,0127) 0,182 56 -0,0057 (0,0146) 0,7 17 -0,0098 (0,0092) 0,298 37 0,0028 (0,0078) 0,724 57 -0,0002 (0,0081) 0,985 18 0,0108(0,0108) 0,326 38 0,0001 (0,0090) 0,991 58 -0,0090 (0,0105) 0,397 19 0,0008 (0,0089) 0,926 39 -0,0009 (0,0085) 0,919 59 -0,0079 (0,0046) 0,096 20 0,0110(0,0130) 0,404 40 0,0055 (0,0081) 0,502 aStandard error
33 Table C: Estimated alphas after completed date
Takeover Takeover Takeover
deal coefficient p-value deal coefficient p-value deal coefficient p-value 1 0,0072 (0,0132)a 0,592 21 -0,0013 (0,0126) 0,916 41 -0,0163 (0,0108) 0,145 2 0,0126 (0,0160) 0,436 22 -0,0042 (0,0143) 0,77 42 0,0145 (0,0111) 0,202 3 0,0065 (0,0064) 0,319 23 -0,0055 (0,0168) 0,745 43 -0,0068 (0,0122) 0,581 4 0,0083 (0,0096) 0,398 24 -0,0059 (0,0106) 0,58 44 0,0227 (0,0152) 0,147 5 -0,0007 (0,0190) 0,969 25 0,0032 (0,0167) 0,849 45 0,0222 (0,0150) 0,15 6 -0,0010 (0,0187) 0,959 26 -0,0108 (0,0113) 0,347 46 0,0271 (0,0129) 0,046 7 -0,0085 (0,0154) 0,583 27 -0,0177 (0,0192) 0,366 47 0,0227 (0,0152) 0,147 8 0,0111 (0,0191) 0,568 28 -0,0230 (0,0120) 0,066 48 0,0232 (0,0158) 0,155 9 -0,0135 (0,0112) 0,239 29 -0,0230 (0,0120) 0,066 49 0,0216 (0,0142) 0,141 10 -0,0076 (0,0116) 0,52 30 -0,0194 (0,0116) 0,107 50 -0,0042 (0,0206) 0,84 11 0,0001 (0,0108) 0,995 31 0,0209 (0,0178) 0,252 51 -0,0071 (0,0134) 0,601 12 -0,0051 (0,0103) 0,625 32 -0,0114 (0,0190) 0,552 52 -0,0070 (0,0133) 0,604 13 -0,0006 (0,0245) 0,98 33 -0,0269 (0,0103) 0,015 53 0,0113 (0,0112) 0,323 14 -0,0161 (0,0127) 0,216 34 -0,0278 (0,0095) 0,007 54 0,0244 (0,0193) 0,218 15 -0,0153 (0,0083) 0,078 35 0,0255 (0,0192) 0,195 55 0,0056 (0,0209) 0,79 16 -0,0201 (0,0125) 0,118 36 -0,0025 (0,0071) 0,734 56 0,0292 (0,0165) 0,088 17 0,0053 (0,0214) 0,804 37 0,0055 (0,0131) 0,681 57 0,0083 (0,0126) 0,519 18 0,0056 (0,0071) 0,442 38 0,0096 (0,0092) 0,305 58 0,0127 (0,0130) 0,337 19 0,0039 (0,0091) 0,669 39 -0,0017 (0,0109) 0,874 59 -0,0069 (0,0098) 0,487 20 -0,0034 (0,0134) 0,799 40 -0,0017 (0,0109) 0,874 aStandard error
