Value performance of European acquisitions; rumors or announcements?

(1)

Value performance of European acquisitions;

rumors or announcements?

N. Kokhuis

S1257528

March, 2007 Rijksuniversiteit Groningen Faculty of Business Administration

Department of Finance

Under the Supervision of: Dr. Ing. N. Brunia

(2)

Value performance of European acquisitions;

rumors or announcements?

Abstract

In this research the effect of an acquisition on return is examined with a sample including 397 European bidders. I look at the effects rumors and announcements have on the short term abnormal return following a rumor and an announcement. In addition, I look at the difference between these abnormal returns and the possible relationship. I find that in the short term, European acquirers earn small but statistically significant positive abnormal returns around the time of the deal rumor and the announcement. The average abnormal return on the announcement date is significant larger when it is preceded by a rumor. Looking at the timing of the rumor, there is a significant difference between the abnormal returns on the announcement day. After conducting an OLS regression between the number of days between a rumor and an announcement on one hand, and the abnormal returns following an announcement date on the other hand, I find a small but significant relationship. The closer the rumor is to the announcement, the larger the market reaction on the announcement date. Finally, although the rumor and announcement of an

(3)

Contents

1. Introduction ...4

2. Literature review...6

3. Data sample and selection events. ...10

Data sample ...11

Selection events ...12

4. Methodology ...13

Defining the time frame ...14

Choosing a normal specification model ...14

Market and risk adjusted model...15

Aggregation of the Abnormal Return (AR), Average Abnormal Return (AAR) and Cumulative Abnormal Return (CAR ...16

Significance ...17 Non-parametric test...18 OLS-regression...19 Chi-square test ...20 Normality ...20 5. Results...21

6. Conclusions and recommendations for further research ...26

References ...29

Appendix 1: Summary statistics ...30

Appendix 2: Distribution...31

Appendix 3: Abnormal Returns (AR’s) and Cumulative Abnormal Returns (CAR’s)...33

(4)

1. Introduction

-M&A volume in the first half of 2006 jumped to US$1.8 trillion worldwide, a 44% increase from the US$1.3 trillion recorded during the same period last year.- Thompson Financial, 2006.

Not only do these figures stress the magnitude of M&A all over the world, they also show us how interesting it is to know what kind of effects these M&As have for shareholders. Whether or not M&As actually achieve the expected performance gains is an important question. If consolidation causes an increase in the share price, then shareholder wealth is increased. Moreover, it is very interesting for stakeholders, speculators and business analysts to know when the expected performance gains will take place. Do they occur when a rumor is on the streets or when there is an announcement of a merger or acquisition? Another possibility is that there is a relationship between both dates; for example, will a rumor effect cancel out an announcement effect? These kinds of questions are becoming more and more interesting in an increasingly integrating economy like ours.

A very large merger and acquisition (M&A) wave occurred in the European Union in the early 1990s. Besides deregulation, technological and financial innovations and the introduction of the euro, others factors were important in this process, such as value creation, market power and efficiency. Cost and revenue synergies are playing an important role in gaining efficiency (Ayadi and Pujals, 2005).

Another surge in M&A activity occurred in 1996, but there was a break in 2001 and 2002 due to the economic downturn. Currently the economy is catching up, and this is also reflected in M&A activity, which has been increasing these last few years in the European Union. Besides the number of M&As, the size of the acquisitions have become much larger (Goergen and Renneboog, 2004). Another typical phenomenon is that more acquisitions in the Euorpean Union tend to be cross-border. Cross-border M&As have increased relative to the period 1993-1998, both in absolute and relative terms. They account for 30% of the number of M&As and for 24% of the value of all deals in the more recent period (José Manuel González-Páramo, Member of the Executive Board of the ECB Hong Kong, 24 February 2006).

(5)

Union for the bidder firms. Moreover, this paper aims at the question when the value creation is realized.

This leads to the following main research question: What are the short term effects of rumors and

announcements with respect to European acquisitions, and is there a relationship between these effects? To find an answer to this question using empirical research and statistical analyses, the

following testable hypotheses are developed:

Concerning the short term effects on stock return, I expect, following similar prior research positive abnormal returns. Therefore, hypothesis 1 becomes:

Hypothesis 1. The announcement or rumor of a European acquisition leads to a positive stock return effect for the bidder.

The second question of this thesis is if there are significant differences between the abnormal returns following an announcement when looking at the timing of the rumor. In other words, what is the difference, in terms of abnormal returns on the announcement day, between rumors that are very close to the announcement date and rumors that are heard on the street a couple years before the announcement? Therefore, hypothesis two becomes:

Hypothesis 2. There is no difference between the abnormal returns following an announcement when looking at the timing of a rumor.

The third and most interesting question for investors is if we can deduce the sign of the market reaction on the announcement day from the market reaction on the rumor date. Therefore the last hypothesis becomes:

Hypothesis 3. There is no relationship between the sign (negative or positive) of the market reaction following a rumor and the sign of the market reaction following an announcement regarding a European acquisition.

(6)

2. Literature review

Halpern (1983) emphasizes in his research that when the market is efficient, the abnormal returns through acquisitions will follow mainly on the rumor date and to some extent on the announcement date. To have an effect on the market, news has to be unexpected, because, in an efficient market, all information is already reflected in the price. But what if taking over another company is totally in line with the strategic vision of a company? A new market, raw material, production efficiency, knowledge and political safety, all are strategic motives to take over another company (Eiteman, Stonehill and Moffet, 2004). If these motives are part of the strategic vision of a company it is well conceivable that the company will take over another company in the future. Probably it is unknown which company exactly will be taken over at the moment of forming the strategic vision but the announcement of an acquisition has been expected. If the particular target is expected by the market the effect on the stock price of the bidder company is probably not so large because the information is not really news. However, if the target company is not in line with the market expectations because the target company is better/worse than expected by the market, there can be a price appreciation/ depreciation for the bidder firm. For the targets stock price such a rumor/ announcement also can have a big impact on the stock price. In figure 1 the accompanying time frame is presented.

Figure 1. Time frame of an acquisition.

_______________________________ ____________________ _____________

Strategy forming Rumor of an acquisition Official Announcement

As Clarkson, Joyce and Tutticci (2006) say: “In financial markets any information not capable of

objective verification can be classified as rumor. Although many rumors are spread by ‘word of mouth’, there also are formalized networks for rumor spreading.” Knapp (1944) describes rumor

as ‘a proposition for belief of a topical reference disseminated without official verification’. More recently, in analyzing the term rumor within the context of financial markets, Kosfeld (1998) defines it as ‘unofficial public information with an unknown quantity of truth and untruth’ and argues that financial markets are ‘the perfect breeding ground for rumors because highly

competitive industry participants value every piece of information in vying for a comparative advantage’.

(7)

refers to ‘information leakage’ (Aktas, de Bodt and Declerck, 2005). In the stock price of a firm, there is a so-called pre-announcement run-up. In this situation there are abnormal returns only a couple days before the announcement, capturing one-third of the ultimate takeover premium of the target before any formal public news of the bid (Jarrel and Poulsen (1989). If takeover rumors and pre-bid run-ups in prices reflect misuse of private information, we face a significant legal problem. If it is only due to the existence of so-called shark watchers speculating on an upcoming takeover, we do not face a legal problem. However, the economic effects remain the same because in both situations there is a price run-up before the announcement. The second description refers to a rumor that has been presented in newspapers, magazines, television or some other form of media. One recognized outlet is the Wall Street Journal through its Heard on

the Street and Abreast of the Market editorials (Pound and Zeckhauser, 1990; Zivney et al.,

1996)). These rumors can occur a long time (a couple of years) before the announcement of the acquisition has been made.

In this paper the primary purpose is to look at the abnormal returns of the rumor as defined in Zephyr: “A deal status indicating that there is an unconfirmed report, or an announced deal but

the identity of one of the parties is not known e.g. Company A is to buy a German engineering firm for GBP 5 million”. Moreover, this paper will look whether there is a price run-up in the

days before the rumor as defined in Zephyr.

It is remarkable that in prior literature about mergers and acquisitions the timing of stock price appreciation is subordinate to the question whether there is stock price appreciation in the first place. Of course it is very interesting to know in advance whether the stock will go up with respect to a merger or acquisition, but it is also very interesting to know in advance when this value creation takes place. Will this effect take place at the rumor date or at the announcement date? On the stock exchange it is crucial to know, in order to make a profit, when you have to buy or sell a stock or other kind of security. Especially in an efficient market, timing is crucial (Merton, 1981).

As presented in table 1a and 1b, several papers have been written about abnormal returns in terms of stock prices during a merger or acquisition for bidding firms. However, these papers do not reach the same conclusion (Campa and Hernando, 2004) and all of them use the announcement date to estimate abnormal returns. For the U.S., the evidence is equally distributed between

(8)

Table 1a. Comparison of Findings with similar empirical work; studies reporting negative returns to acquirers.

Study CAR Sample

size Sample period window Event (days)

Industry

coverage coverage Country Notes

Beitel et al. (2002) -0.14% 98 1985-2000 0 Financial,

Insurance Developed and Developing Countries

Targets worldwide being acquired by European banks

-0.01% (-1, 1)

-0.20% (-20, 20)

Doukas-Holmen-Travlos

(2002) -2.37% 101 1980-1995 (-5, 5) Diverse Sweden Diversifying acquisitions display negative returns

-1.12% (-5, 1)

-0.52% (-1, 1)

0.62% ( -1, 0)

-0.91% (0, 1)

Unless otherwise noted, event date is announcement date of merger/bid

Table 1b. Comparison of findings with similar empirical work; studies reporting positive returns to acquirers.

Study CAR Sample

size Sample period window Event (days)

Industry

coverage coverage Country Notes

Raj and Forsyth (2002) 1.60% 340 1994-1998 (-15, 15) Diverse U.K. Related sample

0.75% Unrelated sample

Cybo-Ottone et al. 0.99% 54 (-1, 1) Banks E.U

1.40% (-2, 2)

Floreani and Rigamonti

(2001) 3.65% 56 1996-2000 (-20, 2) Insurance U.S., Europe, Australia Beitel et al. (2002) 0.42% 98 1985-2000 (-20, 0) Financial,

Insurance Developed and Developing Countries

Targets worldwide being acquired by European banks 0.38% (-5, 0) 0.06% (-1, 0) 0.18% (-2, 2) 0.24% (-10, 10) Doukas-Holmen-Travlos

(2002) 2.74% 101 1980-1995 (-5, 5) Diverse Sweden Focused acquisitions display opsitive returns

1.38% (-5, 1)

1.19% (-1, 1)

0.83% ( -1, 0)

0.95% (0,1)

Georgen/Renneboog

(2004) 0.70% 139 1993-2000 (-1,0) Diverse E.U Deal value over 100 million

1.18% (-2,2)

0.39% (-30.30)

0.41% (-90,90)

This paper: deals with a

rumor 0.50% 187 (-2, 2) Diverse E.U. Event date is rumor date

0.54% (0, 1)

0.33% (-2, 2)

(9)

This paper: deals without

a rumor 0,17% 210 (-2, 2)

0,15% (0, 1)

Unless otherwise noted, event date is announcement date of merger/bid

Most event studies use the announcement day to measure whether there is value creation or destruction with respect to the announced merger or acquisition in the short run. The explanation is clear and simple; at this time, it is almost certain that the merger or acquisition will actually take place because this is the first time official news about the acquisition is presented. However, to have an effect on the market, the news has to be unexpected.

Pound and Zeckhauser (1990) are focusing their study on the rumors of an acquisition. Using the Market and Risk Adjusted Model (see methodology) they find an average abnormal return of seven per cent. They included 42 events in their sample for the period from January 1983 until December 1985. They started their analysis with public takeover rumors from the “Heard on the street “column that appears daily in the Wall Street Journal. However, their paper is not quite comparable with this paper for three reasons. The most important difference is that they are focusing on the target of an acquisition while this paper is focusing on the bidder of an acquisition. Secondly, Pound en Zeckhauser (1990) are looking at the period before the event, while this paper focuses on the days around the day of the event. The third reason is that their research is conducted for the U.S while this paper focuses on the European Union.

A problem for the choice of a rumor is the fact that it is difficult to determine what the normal returns are preceding the rumor (see methodology). However, the choice of an announcement of an acquisition brings about the same problem and is no solution for this problem. Besides, rumors about an expected merger prior to the announcement could weaken the effect on the announcement because rumors about an acquisition are already on the street and reflected in the price. The question remains how reliable the rumor is and how certain it is that the acquisition actually will take place.

Studies that analyze long-term (up to sixty months) returns to shareholders of acquiring firms tend to find negative significant cumulative abnormal returns for acquirers. Gregory and McCorriston (2002), Faccio et al. (2002) and Raj and Forsyth (2002) report significant long-term negative cumulative abnormal returns to acquirers.

(10)

(11)

3. Data sample and selection events.

Data sample

The primary data employed in this paper are the daily stock returns of European companies involved in M&As during the sample period from January 2001 through December 2005. The daily stock returns are obtained from Datastream International, and the list of M&As rumors and announcements from the Zephyr database. In this section will be showed how the test sample is constructed.

I start out with a list of mergers and acquisition deals in the period between January 2001 and December 2005 in the EU countries. I require that both the acquirer and the target are situated in the EU because I want to look at the effects of acquisitions within the European Union. The acquirer is required to be listed on a stock exchange and the acquired stake is hundred percent for all deals. This makes the comparison between deals with and without a rumor date clearer. Given these requirements, there are 5,481 deals left. Moreover, the top 500 deals are examined out of this extensive list for two reasons. First, it is important that the acquisition of the target is large enough for the bidder to have an impact on the value of the bidder and its stock price. Secondly, time is limited and there is no real surplus value to do research for such a large list of deals. This sample is large enough to find significant results if there are effects that are inclined to this direction.

This selection procedure yields a total of 500 merger deals which I extract from Zephyr with their accompanying ISIN codes. However, there are only 415 bidders accompanied with an ISIN code. The most important reason is that these companies have been de-listed. For the acquirers in this sample, I extract adjusted daily prices from at least 12 months before the announcement date or rumor date, if present (taking into account the estimation period). According to the definition of Datastream, adjusted prices are closing prices adjusted for capital gains (dividends for example). This definition is the most economically relevant for the calculation of returns.

(12)

As pointed out in table 2 and 3, these events are divided over 24 different industry classifications in 15 different EU countries. The consolidation process appears to be the most intense in the banking and IT sector. The United Kingdom is the leading country in this sample.

Table 2. Events in sample by industry classification Table 3. Events in sample by country

Industry Classification

Agriculture, Horticulture & Livestock 1 Banking, Insurance & Financial Services 63 Biotechnology, Pharmaceuticals and Life Sciences 12 Chemicals, Petroleum, Rubber & Plastic 8

Communications 22

Computer, IT and Internet services 41

Construction 11

Food & Tobacco Manufacturing 33

Hotels and Restaurants 16

Industrial, Electric & Electronic Machinery 29 Leather, Stone, Clay & Glass products 10

Metals & Metal Products 6

Mining & Extraction 2

Personal, Leisure & Business Services 28

Printing & Publishing 4

Property Services 15

Public Administration, Education, Health Social Securities 5

Retailing 11

Textiles & Clothing Manufacturing 3

Transport Manufacturing 8

Transport, Freight, Storage & Travel Services 14

Utilities 31

Wholesaling 15

Wood, Furniture & Paper Manufacturing 9

Total 397

Selection events

Out of these 397 events there are 187 events involving a rumor date and 210 that only involve an announcement date. I split these 187 events with rumor date into three groups, which gives a total of four groups (see appendix 3, table 2, 3, 4 and 5). In this paper I discriminate between the three rumor groups because there could be a difference in the effect of a rumor that is close to the announcement date and a rumor that has been known for a couple of years. I use the time frame

(13)

of an event study to make a distinction between the three groups. The first group concerns the events in which the rumor date lies before the estimation window. In other words, the rumor is heard more than 272 days before the announcement date. This group counts 25 events. The second group consists of events in which the rumor date is present in the estimation window. This comes down to the fact that the rumor day occurs between 272 and 20 days before the announcement date. In this group 118 events are included. In the third group, the rumor date lies in the event window. In other words, the rumor date lies between the 20th and 1st day before the announcement day. This group consists of 44 events.

(14)

4. Methodology

Defining the time frame

In this research the event study methodology is used to measure the abnormal returns around the rumor date and announcement date of the acquirers. The event study methodology assumes a certain time frame. This time frame consists of an estimation window and an event window. For an event study concerning daily returns, according to Mackinley (1997), 120 days could be used in the window to determine the normal return. However, in many studies an estimation window of 252 days is used (e.g. Beitel, Schiereck and Wahrenburg (2004)). This is equal to the amount of trading days in one year. In this paper there will also be used an estimation window of 252 days. The day of the rumor and the day of the announcement are both considered t = 0, as is the same in prior research. Although a rumor and an announcement of a merger or acquisition are determined by Zephyr on one day, it is usual to enlarge the event window. This is due to the occurrence of information leakage (see introduction), which may occur a number of days before the rumor or announcement day (James and Houston (2001)). The event window includes several days after the announcement/rumor because it is possible that the market is not efficient and in that case it will take longer than one day to react to the news. In this paper there will be looked at different event windows, starting with a rather wide one (-20, 20) and then slowly narrowing it. This corresponds with the paper of Cybo-Ottone and Murgia (2000) and MacKinley (1997) in order to make the results in line with the literature. Assuming an efficient market, the impact on the day of the announcement/ rumor will be the largest because in an efficient market news is immediately reflected in the stock price. In this paper abnormal returns will be observed twenty days before the rumor or announcement. After the rumor or announcement day the market reaction in a period of twenty trading days will be observed. This also is in line with Mackinley (1997). In this paper an event window of 41 days is used.

Figure 2. Time frame of an event study.

t = -272 t = -20 t = 0 t = 20

________________________ _________ _________ ________________________

Estimation window Event window

Choosing a normal specification model

(15)

and risk adjusted model. Although the first two simpler models are as useful as the more complex models according to Brown and Warner, this research will make use of the market and risk adjusted model.

Another model is the Fama-French three-factor model. Fama and French (1992) started with the observation that two classes of stocks tend to do better than the market as a whole, namely (i) small caps and (ii) stocks with a high book-value-to-price ratio (regularly called "value" stocks; their opposites are called "growth" stocks). They then added two factors to CAPM to reflect a portfolio's exposure to these two classes. However, the broad market index weighs stocks according to their market capitalization, making it size-biased and valuation blind. In general, the gains from employing multifactor models for event studies are limited. As MacKinley (1997) says “The reason for the limited gains is the empirical fact that the marginal explanatory power

of additional factors the market factor is small, and hence, there is little reduction in the variance of the abnormal return. The variance reduction will typically be greatest in cases where the sample firms have a common characteristic, for example they are all firms concentrated in one market capitalization group.” Because in this paper 397 events are divided over 24 different

industry classifications in 15 different countries (see data sample) and thus are not concentrated in one market capitalization group, the use of a multifactor model, like the Fama-French three-factor model, warrants consideration. For that reason this paper will not make use of the Fama-French three-factor model.

Market and risk adjusted model

In this paper the market and risk adjusted model is used in order to make the results in line with the literature. To define the abnormal returns it is important to choose a model that defines normal returns. The return of a stock is calculated by using the following formula:

(

)

1 1 , − − − = t t t t i I I I R (1)

Where Ri,t is equal to the total return of the stock of company i on time t. It en It-1 are equal to the

total return index of the stock, on respectively time t and t-1. The abnormal returns are defined as:

) ( it it

it R E R

AR = − (2)

Where ARit , Rit, and E(R_it) are respectively the abnormal return, actual and normal returns for

time period t.

(16)

it mt i i it R R =

α

+

β

( )+

ε

t=−272,...,−21 (3)

Where

α

_i and

β

_iare estimated values of the true parameters through Ordinary Least Squares (OLS) Regression. ) 0 ( it = E

ε

it it) 2 var(

ε

=

σ

Where R_mt is the return on the market portfolio and

ε

_it is the zero mean disturbance term. The market return in this paper is the return on the FTSE Europe.

(4)

Aggregation of the Abnormal Return (AR), Average Abnormal Return (AR) and Cumulative Abnormal Return (CAR )

After running the market and risk adjusted model regression for each company in the sample for the estimation period, I use the values of the model parameters (

α

_i,

β

_i) to forecast the expected value of returnE(Ri,t). The observed abnormal return on stock i on day t is calculated as the

difference between the actual returns on this stock R_i_,_t and the forecast (expected) return during the period: ) ( , , ,t it it i R E R AR = − where(E)R_i,_t =(E)

α

_i +(E)

β

_iR_m,_t. (5)

The abnormal returns of the stocks are added by using AR_i,_t for every separate time in the event

window. The event window takes from t₁ until t₂. The Average Abnormal Return is calculated by using the following equation:

it N i t AR N AR

Σ

= = 1 1 ₍₆₎

The ARt shows the average abnormal return on a specific day. If the ARt differs significantly

from zero, the market reacted on the news of the event on that specific time. The Cumulative Abnormal Return is defined as:

t t t t AR t t CAR

Σ

= = 2 1 ) (1, 2 (7)

(17)

while the CAR_{( t}_t₁_,₂₎shows a possible extended reaction on an interval. The CAR's₍t₁_,t₂₎ will be

tested for different time frames to get a better insight in the timing of the value creation/destruction and thus the market efficiency.

Significance

Calculating the size and the sign of the average abnormal returns around the event day is not sufficient to conclude that firm value has been created or destroyed. The observed average abnormal return must be statistically different from zero. To test whether the results are statistically different from zero, there has to be examined if the distribution is normal. The normality assumption is important for the exact finite sample to hold. Without assuming normality, all results would be asymptotic and the presence of non-normally distributed variables can decrease the parametric tests’ validity (Brooks, 2002). However, this is in general not a problem for event studies because for the test statistics, convergence to the asymptotic distributions is rather quick (MacKinley, 1997). In addition, removing outliers artificially increases the validity of the tests but will not make it more realistic. Still, to test the robustness of the results there will be conducted a non-parametric rank test for abnormal performance in event studies (Corrado, 1989), see further.

A one sided test is necessary to show that the observed average abnormal returns are statistically different from zero. A one sided t-test (because a positive abnormal return is expected (Moore and McCabe (1993)) with the null hypothesis that average abnormal returns and cumulative mean abnormal return for a given event window is equal to zero.

The degrees of freedom and the critical values under different levels of significance can be found in table 4.

Table 4: Critical values one sided t-test with df=40.

Level of Significance Critical value ( )

10% 1,303

5% 1.684

1% 2.423

The variance of the ARt can be calculated in several ways. The time series method and the

(18)

paper the time series method will be used to determine the variance because there is only one source, namely an acquisition, which is responsible for the possible abnormal return. The cross-sectional method can be helpful when multiple hypotheses exist for the cause of the abnormal return.

The formal specifications of the null hypothesis for both the rumor and the announcement day for the cumulative abnormal return (McKinlay 1997) are:

0 ) ( ) ( : ) 2 , 1 ( ) 2 , 1 ( 0 > t t t t CAR CAR H

σ

µ

(8a) 0 ) ( ) ( : ) 2 , 1 ( ) 2 , 1 ( ≤ t t t t CAR CAR Ha

σ

µ

(8b) Where ) var( )) , ( var( 2 1 2 1 = = t t t t AR t t CAR . (9)

If the null hypothesis is rejected, I will conclude that the acquisition rumor or announcement did carry information content and that based on market perceptions, the acquisition either created or destroyed firm value.

To determine if there is a significant difference between abnormal returns during an announcement following rumors and abnormal returns during an announcement without preceding rumors, I will conduct a t-test with the following formula:

2 1 2 2 2 2 1 1 2 1 n n s n s n AR AR t + + − =

Where n is the number of events and s is the standard deviation of the average abnormal return.

Non-parametric test

(19)

test and the signed rank test, as emphasized by Brown and Warner (1980, 1985), is the requirement that the distributions of the abnormal returns have to be symmetrical to do these tests correctly. Corrado (1989) notices that the Corrado rank test has the advantage of doing the test correctly despite the skewness of the distribution of the abnormal returns. Moreover, it is noticed that the results of the rank test are less influenced by an increase of the event day abnormal return variation in comparison with the results of the parametric tests. When there are abnormal returns, the rank test is also more powerful than the parametric counterparts. This is the result of the strong non-normal distribution, which is common for data of daily returns. Because of these reasons there is chosen for the use of the Corrado rank test. The event window consist of L abnormal returns for each of the N companies. To adopt the rank test the abnormal returns within the event window have to be ranged from one until L. Kit is equal to the rank of the abnormal

return of company i for the event window time t.

Kit = rank (ARit), where t = -20, …., +20.

The test value under the null hypothesis of no abnormal returns is equal to:

) ( / ) 2 1 ( 1 1 0 K s L K N t N i i + − = = (10) Where 2 1 2 )) 2 1 ( 1 ( 1 ) ( 1 + − = = = L K N L K s N i it t t t

By using the rank method there is made a distribution of the abnormal returns transformed in uniform distribution of the rank values independent of the possible asymmetry in the original distribution. Corrado (1989) mentions that the distribution of the above test value is approximated by a standard normal distribution. The null hypothesis of no abnormal returns on a particular day is tested by a one sided t-test. The degrees of freedom and the critical values under different levels of significance can also be found in table 4.

OLS-regression

To test whether the timing of a rumor is of importance for the size (or even the sign) of the abnormal return on the announcement date there will be conducted an ordinary least squares (OLS) regression analysis. The regressions take the following form:

(20)

Where AR_k is the abnormal return,DAYS_k is the number of days between rumor and announcement and

ε

is the error term.

Chi-square test

Chi square is a non-parametric test of statistical significance for bivariate tabular analysis. Any appropriately performed test of statistical significance shows the degree of confidence you can have in accepting or rejecting an hypothesis (Connor-Linton, 2003). Typically, the hypothesis tested with chi square is whether or not two different samples (abnormal returns following rumors and abnormal returns following announcements of an acquisition) are different enough in some characteristic or aspect of their behavior that we can generalize from our samples that the populations from which our samples are drawn are also different in the behavior or characteristic. To test whether there is a relationship between the sign of the abnormal return on the rumor date and the sign of the abnormal return on the announcement date, I will conduct a chi-square test. The accompanying formula is

= − = K i i i i E E O 1 2 2 ( )

χ

(12)

where (K-1) represents the degrees of freedom. In this case K = 4.

i

O is the observed frequency and E_i is the expected frequency. In the case of abnormal returns, the expectations for each cell are equal. In other words, there will be expected that after the abnormal return following a rumor, irrespective of the sign (positive for example), the chance is 50% that it will have the same sign again (positive) and 50% chance that the sign will be different (negative). Moreover the expected frequency for the sign of the abnormal returns on the rumor date is also 50% positive and 50% negative. Concluding there will be expected that each cell consists of 25% of the total frequency.

Normality

(21)

hypothesis of this test assumes normality; therefore, rejection of the null corresponds with a non normal distribution.

(22)

5. Results

First, the results will be presented on whether there is value creation or destruction. In addition, the difference between an acquisition involving a rumor and an acquisition without a rumor in terms of abnormal returns will be presented. Second, the results of the three different groups with a rumor involved will be discussed for both the rumors and announcements. After that, the relationship between the outcomes of the abnormal returns following a rumor and the outcomes of the abnormal returns following an announcement will be discussed.

Table 5 presents the average abnormal return (AR) for the bidders of M&A transactions in the European Union between 2001 and 2005. A difference has been made between deals with (n = 187) and without a rumor (n = 210). The abnormal return of deals with a rumor on the rumor date is 0,42 per cent and is significant on a five per cent confidence interval. The average abnormal return on the announcement is also significant positive on a one per cent level with an average of 0,73 per cent. The abnormal return of deals without a rumor on the announcement date is 0,31 per cent and is significant on a five per cent confidence interval. These results correspond to a large extent to the values and positive signs reported by Renneboog, Simons and Wright (2004). They also found a small but significant positive value on a short interval.

Table 5: Average abnormal return of deals with and without rumors. Deals with rumor ( 187

events ) Deals without rumor ( 210 events )

rumor date announcement date announcement date

AR on t = 0 t-value AR on t = 0 t-value AR on t = 0 t-value

0,0042 2,1676** 0,0073 3,5022* 0,0031 2,0953**

*** = significant at the 10 % level, ** = significant at the 5 % level, * = significant at the 1 % level. To measure the robustness of the results in table 5, outliers will be excluded to see if the results are still significant. The results on the announcement date are still significant when the outliers are excluded. However, the results are not significant for the results on the rumor date after excluding one very large outlier (47 per cent).

(23)

study methodology. Consequently there will be attached more importance to the results on the announcement date.

Table 6: Results of the Corrado rank test Deals with rumor (

187 events ) Deals without rumor ( 210 events )

rumor date announcement

date announcement date

AR on t = 0 Corrado AR on t = 0 Corrado AR on t = 0 Corrado

0,0042 0.7895 0,0073 5.7594* 0,0031 3.8004*

*** = significant at the 10 % level, ** = significant at the 5 % level, * = significant at the 1 % level. There are two remarkable findings of the results in table 5 to emphasize. Firstly, because the abnormal return on the announcement date with a rumor involved is significantly higher on a five per cent level, the combined abnormal returns of 0,42 per cent and 0,73 per cent make the distinction even larger. This implies that the combined reaction of the rumor and the announcement is larger than the reaction on the announcement date when there is no rumor involved. The second finding is in line with the first. The deals without a rumor date have a lower average abnormal return of 0,31 per cent on the announcement date in comparison with the average abnormal return of deals with a rumor on the announcement date of 0,73 per cent. The difference of the abnormal returns on the announcement date is significant on a five per cent level, see table 7. Because the abnormal return on the announcement date with a rumor involved is significantly higher than the abnormal return without a rumor, I assume that a rumor date strengthensthe reaction on the stock market on the announcement date.

Table 7: Difference between abnormal return on announcement date between deals with and without a rumor involved.

difference in AR t-value

0.0042 2.3934

** = significant at the 5 % level

Another finding is that there is no signal referring to any form of information leakage, as presented in appendix 3. Before the event day there are no significant positively cumulative abnormal returns (CAR’s) while on the event day itself all abnormal returns are positive. In other words, a stock price run-up before the rumor or announcement date is not found.

(24)

before. Although the CAR’s are calculated for many time frames, only the results of the two most important intervals, namely (0, 1) and 0, will be shown. For results of the larger event windows I refer to the appendix because these results are not significant.

In the first group the rumor date lies before the estimation window. As shown in table 8, a significant t-value for either the rumor or the announcement can not be found. This may be due to the fact that there are only 25 samples included in this sample. However, it is remarkable that the average abnormal return on the rumor date is 0,53 per cent with an accompanying tvalue of -1,1291, which indicates a negative market reaction. The average abnormal return of the announcement date is close to zero and logically not significant.

Table 8: Cumulative abnormal return: rumor before estimation window (n=25).

Interval Average CAR Rumor t-value Average CAR Announcement t-value

0,1 -0.0055 -0.8298 0.0002 0.0445

0 -0.0053 -1.1291 0.0001 0.0304

*** = significant at the 10 % level, ** = significant at the 5 % level, * = significant at the 1 % level. The second group is the group in which the rumor date lies in the estimation window. As presented in table 9 the market reaction on both the rumor date and the announcement date is significant. For the rumor date this is a positive reaction of 0,74 per cent while for the announcement date this is 0,63 per cent. However, the reaction on the interval (0, 1) of the announcement date (0,9 per cent) is larger than on the rumor date (0,84 per cent).

Table 9: Cumulative abnormal return: rumor in estimation window (n=118).

0,1 0.0084 2.8467* 0.0090 3.0805*

0 0.0074 3.5310* 0.0063 3.0443*

*** = significant at the 10 % level, ** = significant at the 5 % level, * = significant at the 1 % level. The last group consists of deals where the rumor date is situated in the event window. Table 10 presents the results of this group. The CAR’s are not significant on the rumor date, however on the announcement date they are positively significant at the one per cent level.

Table 10: Cumulative abnormal return: rumor in event window (n=44).

0,1 0.0037 0.6845 0.0172 2.7885*

0 0.0012 0.3176 0.0141 3.2428*

(25)

When comparing these three groups there is a remarkable difference. Looking at the abnormal returns on the announcement day in the different groups, there is a striking phenomenon. For the first group, which has the largest distance between the rumor date and the announcement date, the abnormal return is only 0,01 per cent and not significant. The second group, where the rumor date is closer to the announcement date, namely in the estimation period, a significantly positive value of 0,63 per cent is found. For the last group, where the rumor date is very close to the announcement date, namely in the event window, an even larger significantly positive value of 1,41 per cent is found. These abnormal returns differ significantly on a five per cent confidence level (appendix 4, table 1 and 2). In other words, the smaller the time between the rumor and the announcement of a merger or acquisition, the larger is the market reaction on the announcement date for these three groups. Moreover, after conducting an OLS regression between the number of days between rumor and announcement on one hand, and the abnormal returns on the announcement date on the other hand, small but significant results are found (appendix 4, table 3). Although the relationship is significant on the five per cent confidence interval, the R–squared is only 3,3 per cent which is very low (adjusted R-squared is even negative). Summarizing, the smaller the time between the rumors and the announcement of an acquisition, the larger is the market reaction on the announcement date.

(26)

(27)

6. Conclusions and recommendations for further research

This paper provides the first analysis of value performance on the relationship between rumors and announcements on mergers and acquisitions. This research is conducted for mergers and acquisitions in the European Union for the period from January 2001 until December 2005. Using a sample of 397 publicly traded companies divided over 24 different industry classifications in 15 different EU countries, I examine gains to acquirers on both the rumor and announcement date. Based on merger and acquisition theory and previous empirical findings I form three hypotheses. The first hypothesis puts forward that acquisition in the European Union during 2001-2005 created positive wealth gains for bidding firms on as well the rumor as the announcement date. The second hypothesis states that there is no difference between the abnormal returns following an announcement looking at the timing of a rumor. The third hypothesis proposes that the market reaction on the announcement day can be deduced from the market reaction on the rumor date. The first two hypotheses are tested using an event study method, namely the market and risk adjusted model. The third hypothesis is tested using a chi-squared test.

On the basis of the market and risk adjusted model, it is concluded that European acquirers earn positive and statistically significant abnormal returns around the time of the deal rumor and announcement. These results support the first hypothesis and correspond to a large extent to the values and positive signs reported by Renneboog, Simons and Wright (2004). The results on the rumor date are not as consistent and significant as the results on the announcement date. Consequently, more importance will be attached to the results on the announcement date. This is not in line with Halpern (1983) because his research emphasizes that when the market is efficient, the abnormal returns through acquisitions will follow mainly on the rumor date and to some extent on the announcement date.

Moreover, for the European Union, most studies report positive cumulative abnormal returns. Because the abnormal return on the announcement date with a rumor involved is significantly higher than the abnormal return without a rumor, I expect that a rumor date strengthens the reaction on the stock market on the announcement date.

(28)

conducting an OLS regression between the number of days between rumor and announcement on one hand, and the abnormal returns on the announcement date on the other hand, small but significant results are found which lead to a rejection of the second hypothesis.

Although the rumor and announcement of an acquisition brings about an averagely statistically significant positive market reaction the sign of the market reaction on the announcement day can not be deduced from the market reaction on the rumor date. This finding leads to the rejection of the third hypothesis.

This paper provides a novel insight in the separation of value gains following rumors and announcements. Prior event studies concerning mergers and acquisitions had their main focus on one out of two event dates while this paper shows a statically positive value gain on both dates. I would recommend taking into account both dates for further research when interested in the market reaction as a whole. When taking one date in account, I would recommend the announcement date because this research puts forward that the average abnormal returns on this date are more significant and consistent. Secondly, I think it is interesting to have a better insight in the relationship between the signs of abnormal returns on announcement dates following rumors. In this case, the market reaction following an announcement could be predicted looking at the market reaction of the preceding rumor. Although a significant relationship between the signs of the market reaction following rumors and announcements is not found, there could be a relationship when looking at a larger sample size.

An interesting extension (and limitation of this research) is to conduct a research on the relationship between the strategic vision of a company and the effect on the stock price following the announcement of an acquisition. An acquisition that is in line with the strategic motives of a company is less surprising. In an efficient market, prices reflect all information so the strategic vision of a company could influence the abnormal returns on the announcement/ rumor date of an acquisition.

(29)

References

- Aktas, N., de Bodt, E., DeClerck, F., 2005, Is there information leakage around business combinations on the French market?

- Amihud Y, De Long G. L., Saunders A., 2002, The effects of cross-border bank mergers on bank risk and value.

- Ayadi R., Pujals G., 2005, Banking mergers and acquisitions in the EU: overview, assessment and prospects.

-Beitel, P., Schiereck D., Wahrenburg M., 2004, Explaining M&A Succes in European Banks. - Brooks, C., 2002, Intoductory Econometrics for Finance, Cambridge University.

- Brown, S.J en J.B. Warner, 1980, 1985, Using daily stock returns: The case of event studies, Journal of Financial Economics 14, pp. 3-31.

- Campa, J. M.l, Hernando, I., 2004, Shareholder Value Creation in European M & As.

- Clarkson, P. M., Joyce, D., Tutticci, I., 2006, Market reaction to takeover rumour in Internet Discussion Sites, Accounting and Finance 46, 31-52.

- Connor-Linton, J, Department of Linguistics Georgetown University, available online: -

http://www.georgetown.edu/faculty/ballc/webtools/web_chi_tut.html#overview

- Corrado, C.J., 1989, A nonparametric test for abnormal security-price performance in event studies, Journal of Financial Economics 23, pp. 385-395, North-Holland Publishing Company. - Cybo-Ottone, A. and Murgia, M., ‘Mergers and shareholder wealth in European banking’, Journal of Banking and Finance, Vol. 24, 2000, pp. 831–59.

- De Long, G.L., 2003, ‘The Announcement Effects of US versus Non-US Bank Mergers: Do they Differ?’ The Journal of Financial Research 26 (4), 487-500.

- Eiteman, D.K., Stonehill, A.I., Moffett, M.H., 2004, Multinational Business Finance, 10th

edition.

- Goergen M., Renneboog L., 2004, Shareholder wealth effects of European domestic and cross-border takeover bids.

- González-Páramo, J. M., Member of the Executive Board of the ECB Hong Kong, 24 February 2006. Available online: http://www.ecb.int/home/html/index.en.html

- Halpern, P., 1983, Corporate Acquisitions: A Theory of Special Cases, A Review of Event Studies Applied to Acquisitions, Journal of Finance, Vol. 38, No. 2, Maart, 297-317.

(30)

- Knapp, R. H., 1944, A psychology of rumour, Public Opinion Quarterly 8, 22-27.

- Kosfeld, M., 2005, Rumours and Markets, Journal of Mathematical Economics 41, 646-64. - MacKinlay, A.,1997, Event studies in economics and finance, Journal of Economic Literature, 35(1): 13-39.

-Merton R. C., 1981, On market timing and investment performance. I. An equilibrium theory of value for market forecasts.

- Moore, D.S., McCabe, G.P., 2001, Statistiek in de Praktijk, 3e herziene druk.

- Pound J., Zeckhauser R. J., 1990, Clearly Heard on the street: The Effect of Takeover rumors on Stock Prices.

- Thompson Financial, Press Release, 2006. Available online:

http://www.thomson.com/solutions/financial/

(31)

Appendix 1: Summary statistics

Table 1. Average beta and alpha of publicly traded companies in sample

beta alpha

0.5736 0.0006

Table 2. Number of days between the rumor date and the announcement date

Figure 1: M&A deals included in the sample by year

Number of M&A deals

(32)

Appendix 2: Distribution

Figure 1. Distribution AAR’s: deals where a rumor is involved on the rumor date (187 deals).

Figure 2. Distribution AAR’s: deals where a rumor is involved on the announcement date (187 deals).

(33)

Appendix 3: Abnormal Returns (AR’s) and Cumulative Abnormal Returns (CAR’s)

Table 1. Abnormal Returns; deals with and without a rumor ( 397 deals in total). deals with

rumor deals without rumor

rumor date announcement

date announcement date

day abnormal

return t-value abnormal return t-value abnormal return t-value

-20 -0.0019 -0.9513 -0.0003 -0.1687 -0.0020 -1.2993 -19 0.0005 0.2467 0.0007 0.3595 0.0006 0.3648 -18 -0.0005 -0.2512 -0.0015 -0.7453 -0.0013 -0.8664 -17 -0.0003 -0.1663 -0.0021 -1.0511 -0.0006 -0.3976 -16 0.0008 0.4108 -0.0002 -0.0759 0.0022 1.4419 -15 0.0012 0.6156 -0.0015 -0.7642 -0.0003 -0.2278 -14 0.0000 -0.0230 -0.0029 -1.4733 -0.0016 -1.0515 -13 -0.0028 -1.4054 -0.0006 -0.3080 0.0000 -0.0005 -12 0.0009 0.4653 -0.0010 -0.5065 -0.0012 -0.8095 -11 -0.0005 -0.2497 0.0003 0.1662 -0.0016 -1.0501 -10 0.0015 0.7625 0.0001 0.0448 -0.0023 -1.5267 -9 0.0005 0.2633 0.0029 1.4547 -0.0004 -0.2416 -8 0.0035 1.7356*** 0.0004 0.2182 0.0001 0.0348 -7 0.0007 0.3277 0.0033 1.6636*** -0.0003 -0.1993 -6 -0.0002 -0.0883 0.0023 1.1380 0.0007 0.4478 -5 -0.0026 -1.2984 -0.0009 -0.4707 -0.0033 -2.1691** -4 -0.0025 -1.2361 0.0015 0.7586 0.0016 1.0773 -3 0.0034 1.7153*** -0.0014 -0.6906 0.0002 0.1617 -2 0.0015 0.7492 -0.0007 -0.3501 -0.0028 -1.8698 -1 -0.0001 -0.0633 -0.0028 -1.4076 0.0013 0.8939 0 0.0042 2.1253** 0.0073 3.6636* 0.0031 2.0549** 1 0.0024 1.2153 0.0049 2.4497** 0.0023 1.4918 2 0.0006 0.2809 -0.0007 -0.3468 0.0008 0.5073 3 -0.0037 -1.8655*** 0.0001 0.0660 0.0000 0.0036 4 0.0013 0.6486 0.0006 0.2783 -0.0009 -0.6265 5 0.0024 1.1967 0.0003 0.1739 -0.0014 -0.9453 6 0.0039 1.9355** 0.0010 0.4919 0.0011 0.7461 7 0.0030 1.5057 0.0005 0.2665 0.0001 0.0484 8 0.0012 0.6263 0.0005 0.2752 -0.0004 -0.2686 9 0.0014 0.6823 0.0005 0.2451 -0.0023 -1.5496 10 0.0029 1.4789 -0.0008 -0.4200 0.0000 -0.0208 11 -0.0036 -1.8022*** -0.0011 -0.5630 -0.0002 -0.1157 12 -0.0003 -0.1586 -0.0032 -1.6225 -0.0029 -1.9413*** 13 0.0006 0.3068 -0.0011 -0.5362 -0.0025 -1.6384 14 0.0001 0.0356 -0.0013 -0.6647 0.0014 0.9171 15 -0.0013 -0.6438 0.0011 0.5659 0.0006 0.4155 16 -0.0010 -0.5016 0.0001 0.0611 -0.0010 -0.6549 17 -0.0010 -0.5066 -0.0012 -0.5825 0.0007 0.4707 18 -0.0013 -0.6414 -0.0005 -0.2266 -0.0005 -0.3120 19 -0.0004 -0.2219 0.0005 0.2527 -0.0003 -0.1873 20 -0.0005 -0.2671 -0.0042 -2.1153*** 0.0009 0.5758

(34)

Figure 1a: CAR; deals where a rumor is involved on the rumor date (187 deals).

CAR: deals with rumor on rumor date ( 187 de als )

-0.005 0 0.005 0.01 0.015 0.02 0.025 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20 day re tu rn

Figure 1b: CAR; deals with rumor on announcement date ( 187 deals )

CAR: deals with rumor on announcement date ( 187 deals )

-0.015 -0.01 -0.005 0 0.005 0.01 0.015 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20 day re tu rn

Figure 2: CAR; deals without rumor on announcement date ( 187 deals )

CAR deals without rumor ( 210 deals )

(35)

Table 2: Cumulative Abnormal Returns without a rumor ( 210 events )

Interval Average CAR Rumor t-value

-20, 20 -0,0079 -0,8313

-5, 5 -0,0001 -0,0196

-2, 2 0,0017 0,5152

0,1 0,0016 0,7408

0 0,0031 2,0953**

*** = significant at the 10 % level, ** = significant at the 5 % level, * = significant at the 1 % level Table 3: Cumulative Abnormal Returns for rumor before estimation window ( 25 events )

rumor Average CAR Announcement announcement t-value

-20, 20 0,0118 0,5517 0,0009 0,0526

-5, 5 -0,0077 -0,5007 -0,0096 -0,7917

-2, 2 -0,0099 -0,9495 -0,0016 -0,1946

0,1 -0,0055 -0,8298 0,0002 0,0445

0 -0,0053 -1,1291 0,0001 0,0304

*** = significant at the 10 % level, ** = significant at the 5 % level, * = significant at the 1 % level Table 4: Cumulative Abnormal Returns for rumor in estimation window ( 118 events )

rumor Average CAR Announcement t-value announcement

-20, 20 -0,0008 -0,0882 -0,0066 -1,3659

-5, 5 0,0008 0,1107 0,0035 0,5087

-2, 2 0,0067 1,4313 0,0048 1,0504

0,1 0,0084 2,8467* 0,0090 3,0805*

0 0,0074 3,5310* 0,0063 3,0443*

*** = significant at the 10 % level, ** = significant at the 5 % level, * = significant at the 1 % level Table 5: Cumulative Abnormal Returns for rumor in event window ( 44 events )

rumor Average CAR Announcement announcement t-value

-20, 20 0,0281 1,1464 0,0226 0,8083

-5, 5 0,0110 0,8618 0,0094 0,6518

-2, 2 0,0090 1,0509 0,0017 0,1792

0,1 0,0037 0,6845 0,0172 2,7885*

0 0,0012 0,3176 0,0141 3,2428*

(36)

Appendix 4: Difference and relationship between Abnormal Returns.

Table 1: Difference between abnormal return on announcement date of deals where rumor lays before estimation period compared with deals where rumor lays in estimation period.

difference in AR t-value

0.0062 2.0847

Table 2: Difference between abnormal return on announcement date of deals where rumor lays in estimation period compared with deals where rumor lays in event window.

difference in AR t-value

0.0079 2.3081

Table 3: OLS between the abnormal returns on the announcement date and the number of days from rumor until announcement.

(37)

Chi-squared tests

Table 3: Rumor date before estimation period Announcement

Rumor AR: Min AR : Plus

AR: Min 6 7 13

AR:Plus 7 5 12

13 12 25

² = 0.44

Table 4: Rumor date in estimation period Announcement

AR: Min 27 24 51

AR:Plus 30 37 67

57 61 118

² = 3.15

Table 5: Rumor date in event window Announcement

AR: Min 9 7 16

AR:Plus 12 16 28

21 23 44

² = 4.18

Table 6: Total of deals with rumor included Announcement

AR: Min 42 38 80

AR:Plus 49 58 107

91 96 187