The influence of acquisition experience on acquisition performance : evidence from the European banking industry

(1)

The Influence of Acquisition Experience on

Acquisition Performance: Evidence from the

European Banking Industry

Master Thesis

Version: Final Version

Programme: Master in International Finance Thesis Supervisor: Dr. Florian Peters

Student: Xiaonan Wang Student Number: 11082046 Date: 31st_{August, 2016}

(2)

Abstract

This thesis is a detailed study of the wealth effects of horizontal acquisitions in European banking industry on shareholder value of acquiring firms and the value moderating effects of acquisition experience. A regression model is constructed, which delineates the complexity of identifying the specific kind of experience that is transferable (and thus possibly of value moderating effects) to the focal acquisition performance. Two samples are involved in this study. The main sample contains 141 horizontal acquisitions between 2006 and 2016 in the banking sector and the secondary sample consists of past acquisitions of acquirers in the main sample between 1997 and 2016. The study finds that first on average, horizontal acquisitions are of statistically insignificant wealth effects on

acquiring firms. Second, acquisition experience from past acquisitions similar to the focal one in terms of regulatory framework is of positive correlation to the performance of the focal acquisition.

(3)

I. Introduction

I.1. Background and research purpose

Acquisitions are an important corporate strategy by which companies respond to challenges. Many firms acquire others or end up being acquired at some point. Since the credit crisis, managerial risk aversion and shortage of financing has weighed upon mergers and acquisitions (M&A) market and led to a sharp decline of deal volume from 2007 to 2008. Last year, the market seems finally to see a full recovery. According to a report published in December, 2015 by McKinsey & Company (McKinsey & Company 2015), worldwide M&A activity across sectors in 2015 outruns an all-time record high last seen in 2007 with a combined announced value of $4.5 trillion as of December, 2015.

However, the academia does not seem to share the market’s enthusiasm about M&As. Research results such as meta-analyses of King, Dalton, Daily, and Covin (2004) show that M&A does not result in better performance of acquiring firms in the short term on average and even has a slightly negative effect in the long term. This discrepancy resulted in dozens of years of the academia being devoted to entangle the maze of M&A. Among the research questions, the value effect of acquisitions has long been at issue in both finance and management literature. Though M&A as a whole is often depicted as a game that one simply cannot win, the degree of success of an acquisition (acquisition performance) is usually measured by the creation of shareholder value (hence the term ‘value creation’).

There is extensive research on the extent of value creation of acquisitions and the key value drivers thereof. Past research investigated whether there is a causal relationship between acquisition activity and firm performance by examining short-term and/or long-term performance around the event window (e.g., Sudarsanam (2003), Betton, et al. (2008), Martynove and Renneboog (2008), and Haleblian et al. (2009) give a detailed review of past research). Scholars have been exploring the antecedent motivations as to why firms conduct M&As (the potential gains that firms can exploit from a deal, such as market power and tax benefits), factors that can moderate the acquisition-performance relationship (such as deal characteristics and firm characteristics), and possible strategies or outcomes after the transaction is completed (e.g., integration strategies and turnover).

(6)

In this thesis, the main concern is one of the value moderators of acquisition performance—the acquirer’s prior acquisition experience. In other words, the study’s first and foremost research question is whether acquirers with more experience can accrue more value in the event of acquisitions. To answer this question, the study first constructs a new system of measuring acquisition experience.

I.2. Research outline

Acquisition experience has long drawn scholars’ attention. However, there is mixed evidence of its value moderating effects as will be explained in detail in the literature review. This study contributes to the literature of M&A first by exploring acquisitions in the European banking industry to evaluate the short-term return consequences of acquisitions and second by constructing a regression model which delineates the complexity of identifying the specific kind of experience that is transferable (and thus possibly of value moderating effects) to the focal acquisition performance.

Three parameters are incorporated to better model acquisition experience, including the sum number of past acquisitions of the acquirer, similarity between past acquisitions and the focal acquisition, and temporal proximity of the past acquisitions. Both the standalone effects and the interactive effects of the three parameters are investigated. This model better identifies the part of the experience that can be transferred to the focal acquisition.

In respect of the recent macro-factor changes in the banking industry, the study also contributes by examining whether findings of the wealth effects of acquisitions born after the credit crisis still reconciles with those before. Acharya et al. (2011) contend that bank holdings in otherwise normal times can be explained by assets acquired in fire-sale due to the crisis and thus they conjecture that it is possible for acquirers to accrue value during the crisis, quite contrary to the results of the majority of past research (e.g., King, Dalton, Daily, and Covin (2004)).

In addition, the study provides comparability and validity check for previous studies. The majority of past research of M&A wealth effects are based on the U.S. data. Considering the size of the European M&A market, the empirical findings focusing on this market is rather scarce. This study, with a sample of European banking industry, not only provides industry-specific insight but also supplement the understanding of the European M&A market.

(7)

This thesis is organized as follows. In the literature review, a review of past studies of wealth effects of acquisitions, research methodologies, and acquisition experience will be presented. New evidence born out of behavioral finance, studies centering the credit crisis, and organization science requires researchers to re-examine the past evidence and incorporate the new evidence into the architecture of their studies.

The author, in the conceptual framework and hypothesis development, will break down acquisition experience with three parameters instead of singular parameters as prevalent in most of the past research and present the two groups of hypotheses of this study.

Then in the research methodology part, the author will analyse the sample used for this study and explain the construction of the variables used and the model built. The study involves two samples—a sample of transactions under study and a sample of past acquisitions of acquirers in the sample mentioned before. Therefore, all the independent variables are constructed in such a way of uncovering the connections of the two samples.

Results found by the research models and robustness tests are reported and analysed in the Empirical findings. Specifically, the study employs two methods to conduct robustness tests which not only yield more insights but also provides more validity to the study.

Last but not least, the conclusion remarks sum up conclusions answering the research questions that have been put forward. The conclusion remarks also present the limitations of the study and potential directions for futures studies.

(8)

II. Theoretical framework

The next part will first briefly review past research on wealth effects of acquisitions, the acquisition experience theory, the main research questions of this study, and a brief introduction of other value moderators including deal characteristics and firm characteristics.

II.1. Value creation of acquisitions

II.1.1. Literature review of the value creation effect of acquisition

There is mixed evidence about the value effects of acquisitions. Acquisitions, in a considerable number of research, are found to erode shareholder value of acquiring firms especially in short term. Most scholars provide evidence of negative abnormal returns for acquiring firms including Cybo-Ottone and Murgia (2000) with horizontal and vertical acquisitions (either the acquirer or the target being a bank) in Europe from 1988 to 1997, DeLong (2001) (only acquisitions focusing on both geographical and activity accounts can avoid destroying value) with a sample of U.S. mergers from 1988 to 1995 where at least one side of the deal is a banking firm, Fuller et al. (2002) with a sample of 3135 takeovers in U.S. from 1990 to 2000, Capron and Pistre (2002) with a sample of 101 horizontal acquisitions, Moeller et al. (2004), Moeller et al. (2005), Campa and Hernando (2006) with a sample of acquisitions in the EU financial industry in the period 1998-2002, and Bouwman et al. (2009) who investigate 2944 acquisitions happened between 1979 and 2002 in the U.S. market.

In a similar vein, Becher (2000) with a sample of U.S. 558 bank mergers from 1980 to 1997 and Masulis et al. (2007) with a sample of 3333 completed acquisitions from 1990 to 2003 in the U.S. market report that bidders can break even in a deal but nothing more.

Beltratti and Paladino (2013) find, with European banking data from 2007 to 2010, significantly positive abnormal returns for acquirers in the time window 10 days after the completion day. They argue that these banks boost investors’ confidence because they manage to complete transactions at a time of crisis. Moreover, with the consolidation, they achieve the ‘too big to fail’ status. These new findings shed light on how the crisis may have changed the landscape of M&A as a whole, awaiting confirmation by future research.

By contrast, not surprisingly, the targets seem to fare much better than acquirers considering that acquirers usually pay a premium to the target. Evidence shows that target firms’ abnormal stock

(9)

returns around the announcement date are usually significantly positive (e.g., Servaes (1991), Bargeron et al. (2008), and Kyriazopoulos and Drymbetas (2015)).

In summary, previous research show that acquisitions generally have a greater wealth effect on the target firm shareholders than those of the acquiring firm. However, considering the far reaching effects of the credit crisis particularly in the banking industry, the recent data of acquisitions may tell a different story of value creation effects.

II.1.2. Methodology issues

To assess the value moderating effects of a deal, scholars generally use the cumulative abnormal return (CAR) to measure the short-term wealth effect of acquisitions to acquiring and target firms in several event windows. The rationale underlying this method is that upon announcement of the deal, by assuming semi-strong market efficiency, the information of the prospect acquisition will be

factored respectively into stock prices of the acquirer and the target. Potential value from the deal will accrue to the acquiring firm shareholders and those of the target firm separately and instantly. This is the temporal window when researchers can measure the size of returns resulted from the particular event under study for each party.

Ex-post accounting measures such as operating income scaled by sales (e.g., Heron and Lie (2002)) and survival probability (e.g., Hebert, Very and Beamish (2005)) would not satisfy this research purpose because they are long-term measurement and thus, for one thing, involve too much ‘noise’ in the background and for another are more concerned with the post-acquisition outcomes. For

example, potential strategic changes of the firms may constitute a huge source of noise.

A large number of research is using short-term cumulative abnormal returns with the announcement date as the day 0 to measure acquisition performance. Though the announcement date seems to be a natural choice, Giglio and Shue (2014), from a behavioral perspective, argue that the passage of time between the announcement of a bid and the completion of a transaction is highly informative. More specifically, they establish that it is possible that the variation in hazard rates of completion after announcement could predict returns.

In a similar vein, Beltratti and Paladino (2013) claim that investors may delay the reward for an acquisition to completion date because more information could be revealed during the due diligence

(10)

process because of the accounting opacity and financial structure complexity which are common among financial services firms. These newly revealed information during due diligence process is likely to add new variables in the completion of a deal. Given this evidence, future research is recommended to examine cumulative abnormal returns centering both the announcement date and the completion date.

Another issue worth noticing is that past research is highly concentrated in U.S. market as about two-third research reviewed in this thesis is conducted in the U.S. market. Beitel, Schiereck, and

Wahrenburg (2004) report that the value moderators found in research with U.S. market data have quite strong explanatory power with European data. Nevertheless, it is highly necessary to isolate the evidence that is unique to European market.

In general, past research on the value creation effects of acquisitions provides as many challenges as answers not only theoretically but also in terms of methodology. It is crucial to examine whether acquisitions are of any wealth effect on the European banking industry, whether the credit crisis has changed the wealth effect of acquisitions on acquiring firms in the European industry, and whether the value determinants found in U.S. sample based research still have explanatory power in European banking sector sample.

II.2. The theory of acquisition experience

This section will first give an introduction of past studies of acquisition experience and then zoom in to three parameters of experience—accumulation, similarity, and degeneration.

II.2.1. Definition and past research

Levitt and March (1988) defines organizational learning as a process in which organizations engage in certain activities, draw inferences from the activities, and store the results for future reference and engagement. In this study, prior acquisition experience is thus defined as the inferences drawn from a firm’s previous acquisitions to be used for the focal acquisition.

The research community growingly realizes that experience from previous acquisitions could be crucial to an acquirer’s future acquisition performance. Bruton, Oviatt, and White (1994) find positive effects of acquisition experience (using the sum number of acquisitions that the acquirer made in the past four years as a proxy) if the target is in distress. Zollo and Singh (2004) find that acquisition

(11)

experience (measured by the sum number of all acquisitions that the acquirer has made since

founding) alone does not positively influence acquisition performance but is contingent on the degree of codification of experience. Ellis, Reus, Lamont, and Ranft (2011) find no significant relationship between long term acquirer performance (measured by return on asset) and acquisition experience (measured by the sum number of all past acquisitions and bidder-target similarities based on CEO survey). Within the acquirer and target nexus, Cuypers et al. (2016) claim that the party with more prior experience (measured by the difference between the sum numbers of acquirer and target past acquisitions) can accrue more value.

Organizational learning research has established that experience does change acquisition behaviours. For example, the ground-breaking work of Haleblian and Finkelstein (1999) conclude that less experienced acquirers misapply experience acquired from earlier acquisitions to later acquisitions which are not similar to the earlier ones (hence, experience in this case is not transferable), whereas more experienced acquirers are able to draw upon the right kind of experience. The authors further conclude that the relationship between acquisition experience (measured by the sum number of the acquirer’s all past acquisitions) and acquisition performance is U-shaped instead of positively linear. Namely, acquirers with no experience and those with a large amount of experience should accrue more value than acquirers with mediocre experience. Their conclusion is supported by a later study by Haleblian, King, and Rajagopalan (2006) who explore the U.S. commercial banking industry and find that acquirers are likely to make further acquisitions and adjust their acquisition behaviors based on previous acquisition outcomes.

The intuition behind the value moderating effect of the acquisition experience is very simple. In the short term, acquirers with more experience are more likely to pick up better targets, perform better due diligence, and negotiate a better bargain to avoid overpayment. In the long term, firms with more experience can better integrate the target to maximize value in the long term.

In fact, research show that the effect of acquisition experience is amplified by targets of lower integration difficulty. Zollo and Reuer (2006) show that there is a positive effect of experience on acquisition performance if the difficulty of integration and management replacement are low. For one, it shows that acquisition experience is of value effect at least in the long term. For another, it

(12)

indicates that firms may have a propensity to overgeneralize and misapply prior experience in terms that the effect of experience diminishes with the difficulty of integration. This observation can be reconciled with an earlier finding of Zollo and Singh (2004). The authors show that the number of codifications of experience (a conscious learning mechanism that can be used to avoid over-generalization) has a significant positive effect on acquirer performance though acquisition

experience itself (measured by the sum number of all past acquisitions) does not. This observation could be one of the reasons explaining why researchers find mixed evidence of value moderating effect of experience.

II.2.2. Experience accumulation

Following the logic that companies could draw inferences from previous acquisitions for future reference, the sum number of previous acquisitions is crucial in that companies which have made more acquisitions are more likely to draw more inferences and hence accumulate more experience. Therefore, naturally, the sum number of past acquisitions is necessarily the first measurement of experience. This straightforward approach has persisted in the research concerning acquisition experience. Every piece of research in acquisition experience employs the numerical count of past acquisitions as the only parameter or one of the parameters of measurement of experience. However, only using the sum number of the past experience may overlook some very important aspects of acquisition experience. Firstly, from the perspective of organizational learning, by using the sum number, researchers assume all the accumulated experience would be transferable to the focal acquisition. However, as explained in Haleblian and Finkelstein (1999), the heterogeneity of experience accumulation in the early stage determines that not all experience is transferable to the focal acquisition. Secondly, it could also overlook the degenerative properties of acquisition

experience. Using the sum number of past experience implies an ever increasing relationship between number of acquisitions made in the past and the amount of experience accumulated. This implied relationship is not consistent with the fact that experience accumulated in the past could become no longer referable at some future point.

(13)

II.2.3. Similarity

Since experience is defined as the inferences drawn from past activities for future use, experience can only possibly have a wealth effect if past inferences can be transferred to the focal acquisition

correctly. Therefore, it is necessary to first review past organizational learning research on experience transfer.

Evidence show that the transference of experience is contingent on the similarity of the focal and prior acquisitions. The importance of similarity is reflected in the review of Haleblian, Devos, McNamara, Carpenter, and Davison (2009). Specifically, Zollo and Winter (2002) argue that high heterogeneity in corporate decisions including acquisition decisions is likely to result in

misapplications of past experience because when past experience is of a high degree of heterogeneity, it is possible that the acquirer would overgeneralize the past experience which could explain why they hypothesize that experience codification systems can help acquirers enhance experience transfer. Hayward (2002) finds that acquisition experience boosts acquisition performance contingent on the degree of similarity (determined by business activity and motivation of the acquisition) between target of the focal acquisition and those of the previous ones.

This finding can help explain the previous observation that the effect of experience diminishes with the difficulty of integration and management replacement because high difficulty of integration and management replacement often implies excessive experience heterogeneity which complicates the identification of causal relationships.

Though the similarity between the previous deals and the focal acquisition can be framed in several parameters, past research show that the degree of similarity of previous acquisitions and the focal acquisition is heavily dependent on the degree of similarity of business activities of the targets. For example, Finkelstein and Haleblian (2002) in their study of the effect of the first acquisition on the second one find that if the targets are from different industries, the second will significantly

underperform the first. Hayward (2002) compares the industries that the previous targets and the focal target are from. Ellis, Reus, Lamont and Ranft (2011) employs the parameter of deal size, product offering, and geography to trace similarity. They find that acquiring firms are better off to start with ‘large’ acquisitions and dissimilarity of product offerings and geography reach is harmful to

(14)

experience transfer. Cuypers, Cuypers and Martin (2016) measure target similarity with target product offering scope. Therefore, business activity (including industry and product offering) is the essential determinant of target similarity.

The second determinant of similarity should be regulatory framework. Regulatory framework itself has very strong value moderating effects. One of the stylized facts about environmental factors is that regulatory events, particularly tax-related, can influence the attractiveness of acquiring. Regulatory changes are found to be harmful to acquirer performance (Malatesta and Thompson, 1993), although no recent studies are found directly examining whether and how recent regulatory changes ensuing the credit crisis influence acquisition performance. Considering the moderating effect of regulations, a change of regulations can potentially render experience generated in another regulatory framework completely not referable for the focal acquisition. In light of such possible outcomes, it is necessary to map similarity with regulatory framework.

Scholars have been trying to control the effect of regulatory framework changes by restricting the sample within a specific regulatory framework. Others set a time frame for the experience to control macro-economic changes, regulation changes, and the potential of degeneration of experience. For example, Haleblian and Finkelstein (1999) with a sample of acquisitions from 1980 to 1992, only take account of the sum of a certain acquirer’s acquisitions since 1948. Hayward (2002), Haleblian, Kim and Rajagopalan (2006), and Nadolska and Barkema (2007) take similar precautions. After the crisis, the most relevant regulation change to the banking industry is the introduction of Basel II. Therefore, this research will test whether this regulatory change has reduce the previous acquisition experience useless.

Third, the third parameter is deal size following Reus, Lamont and Ranft (2011). Intuitively, the experience drawn from large deals could be dramatically different from small deals. Hence, assuming large transaction experience could be transferable to small ones is over-generalizing and could

possibly cause negative transfer. Therefore, this parameter is included to explore the relationship to further map similarity.

Collectively, these findings show that the value moderating effect of acquisition experience depends not only on the sheer number of experience accumulated but also on the nature of the experience.

(15)

Specifically, whether the prior acquisition targets are similar to the focal one. Therefore, scholars have to isolate the portion of the experience that should be transferred to the focal acquisition before examining moderating relationship of experience and performance.

II.2.4. The degeneration of experience

The difficulty with mapping similarity may not be the only challenge. The degenerative quality of experience can also pose questions to measurement. In respect to experience degeneration, Barkema and Schijven (2008) raise the question as to whether experience should be discounted or depreciated. One may easily treat it as an intangible asset and assume that past experience is of upward

compression where it depreciates slower than linear initially and with accelerated rate over time. Nevertheless, there is no consensus yet of this question.

Scholars usually set a specific time frame to control the effect of degeneration. For example, Bruton, Oviatt, and White (1994) choose the sum number of acquisitions in the past four years while Reuer, Park and Zollo (2002), in the past 10 years.

However, there is no evidence directly recording the degeneration of experience. Particularly, in the aftermath of the crisis (some people may argue that we are still in the crisis), it is useful to explore whether and to what extent that the drastic changes in the financial market resulted from the crisis will reduce the past acquisition experience to obsolete. This research question has relevance to both the problem of degeneration and the regulatory changes.

In summary, the wealth effect of experience is easy to comprehend but difficult to prove. The fact that experience does not show significant wealth effect on acquisition performance may be a matter of theoretical discussion as much as of methodological because the measurement should take into account whether this specific experience transferable to the focal acquisition. However, the sum number of previous acquisition fails to capture the potential effect of negative experience transfer and overgeneralization. In other words, a measurement of sheer accumulation overlooks similarity and the degenerative effect.

II.3. Conceptual framework and hypothesis development

In this section, the hypotheses are presented, which are tested in this research. They are based on the theoretical discussions so far. The hypotheses consist of two groups of hypotheses. First, presented is

(16)

the hypothesis regarding the wealth effects of acquisitions in the European banking sector. The second group contains the hypotheses regarding the value moderating effects of acquisition experience on acquisition performance.

II.3.1. Acquisition performance in European banking sector–CAR

As mentioned in the previous discussions, assuming semi-strong efficiency in the market, an acquisition of positive net present value should result in an immediate rise in the acquirer’s stock price. The previous research on acquisition performance generally holds that acquirer could at best breakeven but recent research presents new evidence. Hence, there is no consensus yet of whether the average acquisitions create value or not. Therefore, the first hypothesis of the study is designed to test whether horizontal acquisitions have wealth effect on the shareholders of the acquiring firms. The first hypothesis is that acquisitions in the European banking sector has created value on average. Hypothesis 1: CAR is positive.

II.3.2. The value moderating effect of acquisition experience

From the literature review, it is evident that a positive relationship is expected between acquisition experience and acquisition performance. In this study, several hypotheses are formulated in order to test the expected relationship. As mentioned in the theoretical discussions before, acquisition

experience is a very complex concept and it is mapped with 3 parameters in this study, which are accumulation, similarity and degeneration. Among them, similarity is further mapped with three parameters, which are activity, regulatory framework, and deal size.

Hypothesis 2: CAR is positively related to acquirer experience, which is further specified with the following sub-hypothesis:

1) Sub-hypothesis 2.1: the acquirer’s sum number of prior acquisitions is positively related to CAR;

2) Sub-hypothesis 2.2: the temporal proximity of acquirer’s immediate prior acquisition is positively related to CAR;

3) Sub-hypothesis 2.3: the general similarity of acquirer’s prior acquisitions with the focal one is positively related to CAR;

(17)

4) Sub-hypothesis 2.4: the acquirer’s sum number of prior acquisitions with similar target activity is positively related to CAR;

5) Sub-hypothesis 2.5: the acquirer’s sum number of prior acquisitions within the same regulatory framework is positively related to CAR;

6) Sub-hypothesis 2.6: the acquirer’s sum number of prior acquisitions with similar deal size is positively related to CAR;

7) Sub-hypothesis 2.7: the temporal proximity of acquirer’s immediate prior acquisition with similar target activity is positively related to CAR;

8) Sub-hypothesis 2.8: the temporal proximity of acquirer’s immediate prior acquisition with similar deal size is positively related to CAR;

9) Sub-hypothesis 2.9: the temporal proximity of acquirer’s immediate prior acquisition within the same regulatory framework is positively related to CAR.

(18)

III. Research methodology

The M&A literature has witnessed the prevalence of the event study methodology. As mentioned before, the underlying logic is that, under the assumption that market is semi-efficient, the value effect of the event under study will immediately be factored in stock price. Following this

methodology, the event under study will be first defined. III.1. Event definition

The purpose with this study is to measure the value moderating effects of the acquirer’s previous acquisition experience on the performance of the focal acquisition. M&A research in general applies three methodologies: event studies (which is employed in this study), efficiency studies (Beitel and Schiereck, 2001), or long term performance studies (Owen and Yawson, 2010). As discussed in the theoretical framework, the only methodology, which directly allows evaluations of the value

implications of an acquisition, is event study methodology (Pilloff and Santomero, 1998; Kale, Dyer, and Singh, 2002). It is necessary to define the event day first. As explained in the theoretical

framework, the event day in this study will be two days—the announcement day of a bid and the completion day of the deal.

An event window has to be further defined, which are the days surrounding the event day when the stock price will reflect the value effect of the acquisition. Following the example of past research (e.g., Campa and Hernando (2006) and Zollo (2009)), relatively short time windows under the event study methodology are used. One of the reasons is that under the semi-efficiency assumption, the value effect will be captured immediately by the market. The bigger the window is, the more noise the result will potentially have because it is hard to separate the effects of a particularly event with those of other events. However, previous research show little consensus about how short the window should be. Hence, the choice of event windows has to be done with a theoretical argument and discretion of the researcher.

Following Sudarsanam and Mahate (2003) and Beltratti and Paladino (2013), a three day time window [-1,+1] is applied where the announcement day and the completion day are day 0

respectively. There is a trade-off between two issues when choosing an event window: on one hand, a shorter event window might not capture the effects of potential information leakage or analysts'

(19)

forecast to the event date or gradual diffusion of info or analysts inattention after the event. On the other hand, a longer event window increases the likelihood of contamination from any 'background noise' from other events. To control the first issue, a conservative event window is used that covers 3 trading days, the event day being the center. The results are further verified using multiple event windows to ensure that the results are not dependent on any particular assumptions about information dissemination in the market. Therefore, in total, this research explores CARs in eight different short event windows: 1,+1], 10,+10], 10,+5], 10,-5], 5,+10], 5,+5], 1,+5], and [-1,+10].

III.2. Sample description

The study defines a main sample and a secondary sample. Both samples are collected from the database Zephyr. The main sample selection criteria for this study is listed below:

 The transaction is announced between January 1, 2006 and January 1, 2016.  The transaction must be completed.

 The deal must be friendly

 The main sector for both the acquirer and the target must be banking.

 The acquirer must be a listed company with its headquarter in Western Europe

 The deal type must be acquisition (the acquirer control less than 50% of the target before the transaction and at least 50% after the transaction).

The time window of the sample covers the period of before-crisis, crisis, and after-crisis and thus it can yield informative results concerning the crisis. The deal attitude is restricted to friendly because deal attitude can potentially change the type of experience generated. The restriction made on the main sector is for two reasons. First, it could minimize inter-industry effects. Second, the effect of regulations (which are usually industry specific) could be more easily isolated if the sample is restricted to one sector. For similar reasons, the geographical scope of the sample is restricted to be Western Europe. In addition, a study of Western Europe will provide comparability with studies centering the U.S. market as mentioned before in the literature review.

It should be noted that originally there are 154 transactions collected from Zephyr. However, there are 40 transactions whose deal size is not publicly available. Since the deal size is a key parameter

(20)

that determines a number of variables, these transactions are removed from the sample. Removing 40 transactions could result in a loss of important information. However, the tradeoff has to be made between a loss of information and undue disruptions on the coefficient estimates.

The number of transactions and domestic ones in each year is plotted in Fig 3.1. Deal sizes are plotted versus completion dates in Fig 3.2. Fig 3.3 illustrates the geographical dispersion of acquirers and targets in the sample. Table 3.1. reports a comprehensive statistic summary of the main sample. From the inception of the crisis onwards, international flows of M&A activities in the banking sector have shrunk considerably with respect to levels reached in 2006. 2006 sees the largest volume of deals with 19 deals completed and the volume continually shrinks and drops to the bottom in 2010. The total volume begins to pick up since 2011 but has not touched the peak that it has once reached in 2006. This U-shaped trend is mainly due to the credit crisis that puts European banks under severe funding pressure and risk aversion which greatly reduces the number of participants in the M&A market.

Over half of all the deals are domestic. Notably, in 2010 and 2012, all deals are domestic and in fact over 76% of all deals after 2011 are domestic suggesting a trend of domestic consolidation in

European banking industry. It could also imply that after the crisis, the consolidation strategy in European banking industry becomes more conservative.

The average deal size in the sample is 1.55 billion. The average deal size decreases drastically after 2006 (from 3.17 billion to 1.68 billion). Most large deals (deals exceeding 1 billion) happen in 2006 and 2007. Lack of funding is the potential explanation for the decrease. Banks could be more cautious about paying premium after the crisis and it is also possible that there are a number of fire sales in the sample. The three largest deals are the acquisition of Gestitres, SA by Natexis Banques

Populaires, Finansbank AS by National Bank of Greece, SA, and HBOS PLC by Lloyds Banking Group PLC.

(21)

(22)

Figure 3.2 Sample summary: scatter plot of deal size versus completion date. The blue box labels the 2007-2010 global credit crisis period.

(23)

Figure 3.3 Sample summary: geographical distribution of acquirers and targets. [AL: Albania, AT: Austria, BA: Bosnia and Herzegovina, BE: Belgium, BG: Bulgaria, BR: Brazil, CH: Switzerland, CY: Cyprus, DE: Deutschland, DK: Denmark, EG: Egypt, ES: Spain, FI: Finland, FR: France, GB: Great Britain, GR: Greece, HR: Hungary, IT: Italy, LT: Lithuania, LU: Luxembourg, LV: Latvia, MC: Monaco, MD: Moldova, MK: Macedonia, MR: Mauritania, NL: Netherlands, NO:

(24)

Norway, PL: Poland, PY: Paraguay, RS: Serbia, RU: Russia, SE: Sweden, SK: Slovakia, TR: Turkey, UA: Ukraine, US: United States

Total 2006 2007~2010 2011~

# of deals 114 19 61 34

Percentage of domestic deals 50.0% 36.8% 39.3% 76.5% # of active acquirers 55

# of acquirers that have 1 deal 27 2 deals 15 3 deals 5 >=4 deals 8 # of acquirer country 15 # of target country 35

Percentage of public target 43.9% 57.9% 37.7% 47.1% # of days between announcement and completion:

mean 131.07 106.05 144.82 120.38

min 0 0 0 0

max 499 246 499 452

std. dev. 102.99 75.95 102.68 114.78 Percentage of cash payment 28.9% 26.3% 31.1% 26.5% Deal size (billion EUR)

mean 1.5549 3.1744 1.6842 0.4180

min 0.0004 0.0074 0.0004 0.0010

max 29.6097 29.6097 17.2664 3.9419 std. dev. 4.1542 7.5701 3.6567 0.8067 Deal size distribution

>=1 billion EUR 25 5 16 4

0.5~1 billion EUR 15 4 6 5

0.1~0.5 billion EUR 32 6 20 6

<0.1 billion EUR 42 4 19 19

Table 3.1 Statistical characteristics of the deals in the sample

The sample consists of 55 acquirers. As shown in Fig 3.3, acquiring firms from Italy, Spain, and France are the most active between 2006 and 2016. Italy, Denmark, and Ukraine are the countries where most targets originate. Most of the targets are from European Economic Area (EEA) countries. Italy sees the most domestic deals (accounting for more than 72% of all Italian acquirer related deals) followed by Denmark. Spain-based acquirers have the highest number of deals involving ‘Rest of World’, a label that refers to countries which are none of EEA, European micro-states, or European Neighboring Policy (ENP) and equivalent, highlighting deals involving U.S. targets.

The secondary sample contains the past acquisitions of the acquirers in the main sample. The selection criteria is listed below:

 The transaction is announced between January 1, 1997 and January 1, 2016.  The deal must be friendly

(25)

 The deal type must be acquisition (the acquirer controls less than 50% of the target before the transaction and at least 50% after the transaction).

Most of the criteria for the secondary sample is the same with those of the main sample. However, for the secondary sample, completion is not required because it is assumed that firms could also

accumulate experience from withdrawn deals. Following other research, it was intended that the sample should contain 10-year worth experience of acquirers (e.g., Reuer, Park and Zollo (2002)). However, Zephyr only houses data after January 1st_{, 1997. Therefore, the time span for the secondary}

sample is 1997 to 2016.

min max mean std. dev.

Time span 1997~2016

# total deals 301

Percentage of domestic deals 50.2%

# occurances for each acquirer 0 22 7.93 5.71

Deal size (billion EUR) 0.00 67.50 3.41 8.87

Deal size distribution

>=1 billion EUR 69 0.5~1 billion EUR 26 0.1~0.5 billion EUR 52 <0.1 billion EUR 79 unknown 75 Deals within Basel I framework 192 Deals within Basel II framework 109

Table 3.2 Statistical characteristics of the deals in the secondary sample

The secondary sample contains 301 deals, approximately half of which are domestic deals. As shown in Table 3.2, the average sum number of deals for all acquirers in the sample is 7.93. Societe Generale is the most active acquirer accumulating 22 deals in total. Average deal size is 3.41 billion and the largest deal is the withdrawn acquisition of ABN AMRO by Barclays. Notably, over two thirds of all deals happen within Basel I framework.

III.3. Variables III.3.1. Definitions

The independent variables are constructed to model the previous acquisition experience of the acquirer. As explained before, the following three experience attributes are of interest in the study: the sum number of previous acquisitions, the degree of similarity between previous transactions and

(26)

the focal one, and the temporal proximity from the immediate prior acquisition to the focal one. In particular, the degree of similarity is measured by three parameters: deal size, activity, and regulatory framework. It is based on the three parameters that the variables in the study are constructed. As mentioned before, the variables in this study is designed to reflect the connections and reactions between these variables. The structure and specifications of the experience-related attributes are illustrated in Fig 3.4.

The experience accumulation is measured by the sum number previous deals. The temporal

proximity is measured by natural days between the announcement date of two deals. Target activity similarity is measured by U.S. primary SIC code. The classification of deal size is drawn from Beltratti and Paladino (2013) where deal size is sorted into four categories as shown below. As explained in the theoretical framework, the regulatory framework in this study specifically refers to Basel I and Basel II and the deals are classified into ‘Basel I’ (before the enactment of Basel II in 2008) and ‘Basel II’ (after the enactment of Basel II in 2008).

Figure 3.4 Structure and specifications of the acquirer experience-related attributes in this study Acquisition experience Number of previous deals (experience accumulation) Temporal proximity of previous deals Similarity of previous deals with focal

Target activity similarity

Classification: 4-digit SIC code

Deal size similarity

Classification: >1 bn 500~999.9 m 100~499.9 m <100 m Bank regulatory framework similarity Classification: Basel I: before 2008 Basel II: after 2008

(27)

III.3.2. Independent variables

Figure 3.5 Structure of independent variables

Based on the definitions above, in total ten independent variables are constructed. The structure of the independent variables is illustrated in Fig 3.5.

First, three basic variables are constructed in correspondence to experience accumulation, temporal proximity, and similarity respectively.

Exp_N: the sum number of previous acquisitions

𝑬𝒙𝒑_𝑵 = 𝐥𝐨𝐠⁡(𝑵) where 𝑁 signifies the total number of previous acquisitions.

Exp_T: the temporal proximity of the immediate previous acquisition and the focal one 𝑬𝒙𝒑_𝑻 = 𝒆−𝐦𝐢𝐧⁡(𝑻𝒊)/𝟑𝟔𝟓

(28)

where 𝑇𝑖 signifies the temporal proximity (number of natural days) of each previous acquisitions to

the focal one.

Exp_Similar: whether the acquirer has previous acquisition experience that bears deal size, target activity, and regulatory framework similarities with the focal one

𝑬𝒙𝒑_𝑺 = 𝑫_𝑺𝒊𝒛𝒆 + 𝑫_𝑨𝒄𝒕 + 𝑫_𝑹𝒆𝒈 where 𝐷_𝑆𝑖𝑧𝑒 = {1, 𝑖𝑓⁡𝑡ℎ𝑒⁡𝑎𝑐𝑞𝑢𝑖𝑟𝑒𝑟⁡𝑒𝑥𝑝𝑒𝑟𝑖𝑒𝑛𝑐𝑒⁡ℎ𝑎𝑠⁡𝑑𝑒𝑎𝑙⁡𝑠𝑖𝑧𝑒⁡𝑠𝑖𝑚𝑖𝑙𝑎𝑟𝑖𝑡𝑦⁡𝑤𝑖𝑡ℎ⁡𝑡ℎ𝑒⁡𝑓𝑜𝑐𝑎𝑙⁡𝑜𝑛𝑒_{0, 𝑖𝑓⁡𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡} 𝐷_𝐴𝑐𝑡 = {1, 𝑖𝑓⁡𝑡ℎ𝑒⁡𝑎𝑐𝑞𝑢𝑖𝑟𝑒𝑟⁡𝑒𝑥𝑝𝑒𝑟𝑖𝑒𝑛𝑐𝑒⁡ℎ𝑎𝑠⁡𝑎𝑐𝑡𝑖𝑣𝑖𝑡𝑦⁡𝑠𝑖𝑚𝑖𝑙𝑎𝑟𝑖𝑡𝑦⁡𝑤𝑖𝑡ℎ⁡𝑡ℎ𝑒⁡𝑓𝑜𝑐𝑎𝑙⁡𝑜𝑛𝑒⁡ 0, 𝑖𝑓⁡𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡ 𝐷_𝑅𝑒𝑔 = {1, 𝑖𝑓⁡𝑡ℎ𝑒⁡𝑎𝑐𝑞𝑢𝑖𝑟𝑒𝑟⁡𝑒𝑥𝑝𝑒𝑟𝑖𝑒𝑛𝑐𝑒⁡ℎ𝑎𝑠⁡𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑜𝑟𝑦⁡𝑓𝑟𝑎𝑚𝑒𝑤𝑜𝑟𝑘⁡𝑠𝑖𝑚𝑖𝑙𝑎𝑟𝑖𝑡𝑦⁡𝑤𝑖𝑡ℎ⁡𝑡ℎ𝑒⁡𝑓𝑜𝑐𝑎𝑙⁡𝑜𝑛𝑒⁡_{0, 𝑖𝑓⁡𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡} On top of the above three basic variables, three groups of joint variables are constructed to model the interactions between the experience attributes.

1) the number of previous deals and the temporal proximity:

Exp_N_T: the total number of previous acquisitions weighted by the time degeneration effect: 𝑬𝒙𝒑_𝑰𝒏𝒅𝒆𝒙 = ∑ 𝒆−𝑻𝒊/𝟑𝟔𝟓

𝑵

𝒊=𝟏

2) the number of previous deals and similarity:

Exp_N_Act: the total number of previous acquisitions with the same target bank activity

𝑬𝒙𝒑_𝑵_𝑨𝒄𝒕 = 𝐥𝐨𝐠⁡(∑𝑲𝒊 𝑲𝟎

)⁡⁡

𝑵

𝒊=𝟏

where 𝐾0 is the total number of target bank SIC codes of the focal deal, and 𝐾𝑖⁡is the total number of

target SIC codes in each one of previous acquisitions that matches the focal one. Exp_N_Size: the total number of previous acquisitions with the similar deal sizes

𝑬𝒙𝒑_𝑵_𝑺𝒊𝒛𝒆 = 𝐥𝐨𝐠⁡(∑ 𝑫𝒊⁡ 𝑵

𝒊=𝟏

(29)

where 𝐷_𝑖= {1, 𝑖𝑓⁡𝑡ℎ𝑒⁡𝑖

𝑡ℎ_{𝑝𝑟𝑒𝑣𝑖𝑜𝑢𝑠⁡𝑑𝑒𝑎𝑙⁡ℎ𝑎𝑠⁡𝑡ℎ𝑒⁡𝑠𝑖𝑚𝑖𝑙𝑎𝑟⁡𝑑𝑒𝑎𝑙⁡𝑠𝑖𝑧𝑒⁡𝑤𝑖𝑡ℎ⁡𝑡ℎ𝑒⁡𝑓𝑜𝑐𝑎𝑙⁡𝑜𝑛𝑒}

0, 𝑖𝑓⁡𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡

Exp_N_Reg: the total number of previous acquisitions within the same regulatory framework (Basel I or II) 𝑬𝒙𝒑_𝑵_𝑹𝒆𝒈 = 𝐥𝐨𝐠⁡(∑ 𝑹𝒊⁡ 𝑵 𝒊=𝟏 ) where 𝑅𝑖 = {⁡⁡1,⁡⁡⁡⁡𝑖𝑓⁡𝑡ℎ𝑒⁡𝑖 𝑡ℎ_{⁡𝑑𝑒𝑎𝑙⁡𝑖𝑠⁡𝑤𝑖𝑡ℎ𝑖𝑛⁡𝑡ℎ𝑒⁡𝑠𝑎𝑚𝑒⁡𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑜𝑟𝑦⁡𝑓𝑟𝑎𝑚𝑤𝑜𝑟𝑘⁡𝑎𝑠⁡𝑡ℎ𝑒⁡𝑓𝑜𝑐𝑎𝑙⁡𝑜𝑛𝑒} 0,⁡⁡⁡⁡𝑖𝑓⁡𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡ 3) the temporal proximity and similarity

Exp_T_Act: the temporal proximity of the immediate previous acquisition in which the target bank SIC codes cover all of those in the focal one

𝑬𝒙𝒑_𝑻_𝑨𝒄𝒕 = 𝒆−𝐦𝐢𝐧⁡(𝒇𝟏(𝑻𝒊))/𝟑𝟔𝟓

where 𝑓1(𝑇𝑖) = {

𝑇𝑖,⁡⁡⁡𝑖𝑓⁡𝐾𝑖 = 𝐾0⁡⁡⁡⁡⁡⁡

𝐼𝑛𝑓, 𝑖𝑓⁡𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒⁡

Exp_T_Size: the temporal proximity of the immediate previous acquisition that has the similar deal size as the focal one

𝑬𝒙𝒑_𝑻_𝑺𝒊𝒛𝒆 = 𝒆−𝐦𝐢𝐧⁡(𝒇𝟐(𝑻𝒊))/𝟑𝟔𝟓

𝑇𝑖, 𝑖𝑓⁡𝐷𝑖= 1⁡⁡⁡⁡⁡⁡

𝐼𝑛𝑓, 𝑖𝑓⁡𝐷𝑖= 0⁡⁡⁡⁡⁡⁡⁡

Exp_T_Reg: the temporal proximity of the immediate previous acquisition that is within the same regulatory framework as the focal one

𝑬𝒙𝒑_𝑻_𝑹𝒆𝒈 = 𝒆−𝐦𝐢𝐧⁡(𝒇𝟑(𝑻𝒊))/𝟑𝟔𝟓

𝑇𝑖,⁡⁡⁡⁡⁡⁡⁡𝑖𝑓⁡𝑅𝑖 = 1⁡⁡⁡⁡

𝐼𝑛𝑓,⁡⁡⁡⁡𝑖𝑓⁡𝑅𝑖 = 0⁡⁡⁡⁡⁡⁡⁡

(30)

Independent

variable min max mean median std. dev. # obs.

exp_N -2.303 3.001 0.964 1.411 1.548 114 exp_T 0.000 0.978 0.342 0.249 0.339 114 exp_Similar 0.000 3.000 2.035 2.000 1.096 114 exp_N_T 0.000 4.773 0.685 0.334 0.954 114 exp_N_Act -2.303 2.839 0.684 1.131 1.638 114 exp_N_Size -2.303 1.808 -0.740 0.095 1.494 114 exp_N_Reg -2.303 3.001 0.117 0.742 1.790 114 exp_T_Act 0.000 0.973 0.091 0.000 0.214 114 exp_T_Size 0.000 0.978 0.155 0.000 0.285 114 exp_T_Reg 0.000 0.978 0.258 0.086 0.322 114

Table 3.3 Statistical summary of the independent variables

III.3.3. Control variables

Following Haleblian, Devos, McNamara, Carpenter, and Davison (2009), the control variables are divided into two categories: deal characteristics and firm characteristics.

III.3.3.1. Deal characteristics

In this research, two deal characteristics variables are controlled—payment medium and geographic diversification.

Stock-financing is perceived as a signal of overvaluation of the acquirer (King et al., 2004).

Malmendier et al. (2016), with a data sample containing unsuccessful bids between 1980 and 2008 in U.S., show that targets who receive cash offers are revalued on average by 15% after the deal is

withdrawn.

Announcement returns of cash financed deals are found to be higher than those of stock financed ones, both in the short run and in the long run and for both the acquirer and the target. With a sample of Canadian firms, Eckbo and Thorburn (2000) find domestic stock-financed acquisitions see higher returns on the acquirer’s side. Fuller, Netter, and Stegemoller (2002) find that returns are greater when the deal is financed with stock though they find insignificant abnormal returns on average.

Taking a somewhat different perspective, Beltratti and Paladino (2013), in the context of the credit crisis, find, with European banking data, that cash payment is bad news for acquirer abnormal returns because it is considered to be indiscrete to pay cash at a time of crisis.

(31)

Another deal characteristic is diversification, particularly in terms of activity and geography.

Research generally classifies acquisitions according to the degree of geographic and activity similarity between the bidder and the target. The fact that acquirers are making geographical and/or activity diversifying acquisitions suggests that the acquirer seeks to find a foothold in a market outside its original market and/or to a new line of business. This strategy seems to promise growth.

However, Schoar (2002) claims that diversification is not bad per se but it is a bad corporate strategy as she identifies two sources of value destruction in a diversifying deal—a decreased efficiency of the acquirer and an increased wage of the target. Devos, Kadapakkam, and Krishnamurthy (2009) find that operating synergies are higher with focusing deals.

Evidence on the detrimental value effect of diversifying deals is found for industrial firms in the U.S. (Morck, Sheleifer, and Vishny, 1990). DeLong (2001) find that only deals that are both geography and activity focusing do not destroy value at announcement with U.S. banking data. Goergen and

Renneboog (2004) find significant higher abnormal returns for both bidders and targets in intra-European deals. Moeller and Schlingemann (2005) find lower abnormal returns for bidders who acquire foreign targets in a sample of 4430 acquisitions between 1985 and 1995 and returns are negatively correlated with both geographical and activity diversification.

III.3.3.2. Firm characteristics

Firm characteristics here refer to those of the acquirer instead of the target. The firm characteristic discussed here is pre-acquisition operating performance of the acquirer.

Scholars have paid particular attention to the value moderating effects of historical operating performance in acquisitions. Recent research such as Heron and Lie (2002) find that the post-acquisition performance (measured by operating income) increase when acquirers with high market-to-book ratios is paired with low MTB ratio targets.

Another measure of pre-bid operating performance is more popular and that is the Tobin’s Q. Lang et al. (1989) find that high Tobin’s Q acquirers and low Tobin’s Q targets benefit more from M&As. Rau and Vermaelen (1998) show that companies with agency issues (indicated by low Tobin’s Q ratios) tend to destroy value with negative NPV acquisitions. This could be one of the reasons as to why low Tobin’s Q ratio could predict low returns.

(32)

In this study both the market to book ratio and the Tobin’s Q of the acquirer are controlled. III.3.3.3. List of control variables

The definitions and sources of the control variables in this study are summarized in Table 3.4, and the statistical summary is presented in Table 3.5.

Variable

name Description Source

Asset Log of the acquirer’s total assets in the year prior to the

acquisition. Calculated based on Datastream data

Concen Market concentration of the acquire country; represented

using the 3-bank asset concentration. World bank

CPM Medium of payment; 1 if paid in cash; 0 if otherwise. Zephyr DCR Domestic acquisition; 1 if the acquirer and the target

originate from the same country and 0 if otherwise. Zephyr

DS Deal size normalized by the acquirer market capitalization Calculated based on Zephyr and Datastream data

EAR Common equity to asset ratio of the acquirer Calculated based on Datastream data LVRG Leverage factor; ratio of total debt and the sum of total debt

and common equity of the acquirer Calculated based on Datastream data

MTB Market-to-book ratio of the acquirer Datastream

PUB 1 if the target if public, and 0 if otherwise Zephyr

TBQ Tobin's Q; the ratio of market capitalization to (total assets -

total liabilities) Calculated based on Datastream data

VOL Volatility; acquirer’s return idiosyncratic volatility from 250 days to 5 days before the announcement (Beltratti and Paladino 2013)

Calculated based on Datastream data Table 3.4 Control variables included in this study

(33)

Control

variable min max mean median std. dev. # obs.

Concen 0.217 1.000 0.567 0.460 0.265 114 DVOL 0.000 6.876 1.219 1.097 1.041 114 VOL 0.005 0.118 0.019 0.013 0.016 114 DCR 0.000 1.000 0.500 0.500 0.502 114 CPM 0.000 1.000 0.289 0.000 0.456 114 PUB 0.000 1.000 0.439 0.000 0.498 114 MTB -0.160 4.260 1.325 1.285 0.851 114 TBQ 0.004 0.474 0.075 0.054 0.067 114 DS 0.000 3.153 0.152 0.024 0.400 114 ASSETS 3.230 6.386 5.146 5.158 0.793 114 EAR -0.051 0.192 0.053 0.053 0.035 114 LVRG 0.000 1.097 0.834 0.867 0.167 114

Table 3.5 Statistical summary of the control variables III.4. Research models

III.4.1. Cumulative abnormal returns

The cumulative abnormal return (CAR) will be calculated following the event study methodology proposed by Brown and Warner (1985). The calculation is based on the market model:

𝑬(𝑹𝒊,𝒕) = 𝜶 + 𝜷 ∗ 𝑬(𝑹𝑴𝒕)

where 𝑹𝒊,𝒕⁡is the return of the stock 𝒊 on day 𝒕, and 𝑹𝑴𝒕 is the return of the market index. To estimate

the coefficient 𝜶 and 𝜷, OLS regression will be performed for the [-170, -20] estimation window using the Euro Stoxx 600 Bank Index as market index following Beltratti and Paladino (2013). This index tracks specifically the European banking sector. The abnormal return (AR) is defined as the

difference between the actual stock return and the expected return based on the market model. The ARs for each day in the event window will thus be calculated as

𝑨𝑹𝒊,𝒕= 𝑹𝒊,𝒕− (𝜶̂ + 𝜷̂ ∗ 𝑹𝑴𝒕)

and the CAR will be calculated by integrating the abnormal returns: 𝑪𝑨𝑹𝒊,[𝒔,𝒆]= ∑ 𝑨𝑹𝒊,𝒕

𝒆

(34)

Additionally, for statistical analysis on the abnormal returns during the Acquisition event, cross-sectional average abnormal returns (AAR) will be calculated from 10 days before the event to 10 days after the event:

𝑨𝑨𝑹𝒕= ∑ 𝑨𝑹𝒊,𝒕 𝑵

𝒊=𝟏

where 𝑵 is the sample size. Cumulative average abnormal returns will also be calculated for various windows:

𝑪𝑨𝑨𝑹[𝒔,𝒆]= ∑𝒆𝒕=𝒔𝑨𝑨𝑹𝒕

III.4.2. Regression model and robustness test The main OLS regression model is constructed as follows

𝑪𝑨𝑹 = 𝑪𝟎+ ⁡ 𝑪𝟏𝑬𝒙𝒑_𝑵 + 𝑪𝟐𝑬𝒙𝒑_𝑻 + 𝑪𝟑𝑬𝒙𝒑_𝑺𝒊𝒎𝒊𝒍𝒂𝒓⁡ + 𝑪𝟒𝑬𝒙𝒑_𝑵_𝑻 + 𝑪𝟓𝑬𝒙𝒑_𝑵_𝑺𝒊𝒛𝒆 + 𝑪𝟔𝑬𝒙𝒑_𝑵_𝑨𝒄𝒕⁡ + 𝑪𝟕𝑬𝒙𝒑_𝑵_𝑹𝒆𝒈 + 𝑪𝟖𝑬𝒙𝒑_𝑻_𝑺𝒊𝒛𝒆 + 𝑪𝟗𝑬𝒙𝒑_𝑻_𝑨𝒄𝒕⁡ + 𝑪𝟏𝟎𝑬𝒙𝒑_𝑻_𝑹𝒆𝒈⁡ + ∑ 𝜹𝒊 𝒏𝒖𝒎𝒃𝒆𝒓⁡𝒐𝒇⁡𝑪𝒐𝒏𝒕𝒐𝒍⁡𝒗𝒂𝒓𝒊𝒂𝒃𝒍𝒆𝒔 𝒊=𝟏 𝑪𝑽𝒊⁡⁡⁡⁡

where 𝑪𝟎 signifies the constant (intercept), 𝑪𝟏~𝟏𝟎 represent the coefficients of the independent

variables, 𝑪𝑽𝒊 represents each of the control variables, and 𝜹𝒊 represents the coefficient of each of the

control variables. The polarity and student’s t-statistic of the resultant coefficients reflect respectively the relationship and the significance of relevance of each variables with the CAR.

An auxiliary OLS regression will then be performed removing the independent variables as follows

𝑪𝑨𝑹 = 𝑪𝟎+ ⁡ ∑ 𝜹𝒊

𝒏𝒖𝒎𝒃𝒆𝒓⁡𝒐𝒇⁡𝑪𝒐𝒏𝒕𝒐𝒍⁡𝒗𝒂𝒓𝒊𝒂𝒃𝒍𝒆𝒔

𝒊=𝟏

𝑪𝑽𝒊⁡⁡⁡⁡

A comparison is made between adjusted-R squared scores of the main regression and that of the auxiliary regression to verify whether the addition of independent variables improves the fitting of the model.

The aforementioned procedures will be repeated twice, respectively for the announcement-day CAR and the completion-day CAR.

(35)

Robustness test I:

 The control variables that have non-significant coefficients in the main regression are selected as testing variables, while the independent variables and the control variables that have significant coefficient are selected as fixed variables;

 The OLS regression is repeated 2N_{times, where N equals to the number of testing variables. In}

each iteration, the regression is constructed with the fixed variables and one specific combination of the testing variables. Thus, all 2N_{possible combinations of testing variables}

are tested;

 Statistical analysis is performed on the repeated regression results and a comparison will be made with the main regression result.

Robustness test II:

 Select one of the independent variables and all of the control variables as fixed variables. Select the other nine independent variables as testing variables. The OLS regression is

repeated 29=512 times. In each iteration, the regression is constructed with the fixed variables and one specific combination of the testing variables;

 Repeat the above procedure ten times, in which the ten independent variables are selected as fixed variable one by one.

 After each of the ten iterations, a statistical analysis is performed on the independent variable which has been selected as fixed variable. Thus, the statistic descriptions of all the ten

independent variables are obtained after the ten iterations. The results will be then compared with results obtained in the main regression.

(36)

IV. Empirical findings

Empirical findings are summarized in this section with respect to the hypotheses listed in Section II.3. The robustness test results are presented in the end.

IV.1. Hypothesis I: polarity of CAR

The CARs for the [-1 1] days window around announcement dates and completion dates are plotted in Fig 4.1. It should be noted that variations of both CARs after 2012 increase significantly. It suggests that after the crisis the acquisition market in Europe is not quite stable in this period.

Figure 4.1 CAR[-1 1] around announcement dates (left) and completion dates (right) for the 114 deals in the sample. The positive CARs are coloured in red while the negative ones are in black.

The blue rectangle labels the 2007-2010 credit crisis period.

Table 4.1 summarized the statistics about the CARs around various event windows. Unlike the results obtained in Beltratti and Paladino (2013), there are no significantly positive CARs observed. This may be caused by the fact that their sample is more inclusive than sample in this study (it includes 139 deals happened within a four-year duration and does not exclude hostile deals). It is also noteworthy that the average days taken to complete a deal in their sample is much shorter in the sample here (86.80 days versus 131.70).

On the contrary, there is significantly negative CAR in the [-1 5] window around announcement dates, and in the [-10 5] and [-10 -5] windows around completion dates. This is consistent with the majority of past research which report zero or negative abnormal returns for acquiring firms.

(37)

Statistics summary of cumulative abnormal returns over different event

windows

Around announcement

Window

Mean

Min.

Max.

Std. dev.

T stat.

[-1 1]

-0.35%

-31.95%

18.01%

5.19%

-0.7105

[-10 10]

-2.14%

-114.90%

53.39%

14.86%

-1.5383

[-10 5]

-1.24%

-110.77%

47.99%

13.69%

-0.9659

[-10 -5]

-0.54%

-114.06%

19.00%

11.65%

-0.4909

[-5 10]

-1.96%

-38.00%

25.12%

9.44%

-2.2105**

[-5 5]

-1.05%

-33.88%

19.72%

7.33%

-1.535

[-1 5]

-0.72%

-26.40%

22.56%

6.21%

-1.239

[-1 10]

-1.62%

-35.78%

27.95%

9.39%

-1.8435*

Around completion

Window

Mean

Min.

Max.

Std. dev.

T stat.

[-1 1]

-0.97%

-64.43%

13.39%

7.71%

-1.3447

[-10 10]

-1.57%

-67.91%

23.02%

11.55%

-1.4531

[-10 5]

-2.01%

-60.41%

13.77%

10.19%

-2.1089**

[-10 -5]

-1.24%

-39.35%

11.50%

6.30%

-2.0947**

[-5 10]

-0.10%

-42.59%

20.69%

8.61%

-0.1245

[-5 5]

-0.54%

-43.65%

14.99%

7.68%

-0.7518

[-1 5]

-0.89%

-36.35%

13.60%

6.66%

-1.4204

[-1 10]

-0.45%

-36.22%

22.85%

8.06%

-0.5911

* 10% significance level

** 5% significance level

Table 4.1 Summary of announcement-day CAR and completion-day CAR within various windows

The AAR for each day in the [-10 10] window is plotted in Fig 4.2. The figure shows more days with negative AR than with positive AR around the announcement date; particularly, the AR is

consecutively negative in the period from 7 days to 4 days before announcement and in the period from 1 day to 3 days after announcement. In addition, both the immediate day before and after the

(38)

completion date experience negative AR. The distribution of CAARs for the 210 possible windows from -10 days to 10 days with respect to the event is plotted in Fig 4.3. The fitted normal distribution curves also shows negative means. These results imply that the acquisition event is more likely to induce negative CAR rather than positive.

Figure 4.2 AAR in the [-10 10] window around announcement (upper) and completion (lower) for the 114 deals in the sample.

(39)

Figure 4.3 Distributions of the CAARs among 210 possible event windows from the [-10 10] days period around announcement (upper) and completion (lower). Each bar in the figures represents the number of occurrence for each specific range of CAAR in the x-axis. The red lines

(40)

IV.2. Hypothesis II: relevance of acquisition experience with CAR

The regression results for the CAR around announcement and completion dates are presented in Table 4.2 and Table 4.3 respectively. In both tables, comparisons are made between the regression results with and without the independent variables (the variables modelling the acquirer experience). According to the results, the independent variables contribute to better fitting of the regression model. In both Table 4.2 and Table 4.3, the model with the inclusion of independent variables shows higher adjusted R-squared score. Particularly in the completion-day-CAR result (Table 4.3), the adjusted R-squared score is significantly decreased when the independent variables are removed. Besides, the F-test null hypothesis in the completion-day-CAR model cannot be rejected when the independent variables are removed (the F-test p-value is 033). Furthermore, there is one significant independent variable (exp_N_Reg) in the announcement-day CAR and three significant independent variables (exp_N_Reg, exp_T_Act, and exp_T_Reg) in the completion-day CAR. Therefore, there is a significant relationship between acquirer experience and CAR.

(41)

Coeff. Std. Err. T-stat. P-value Coeff. Std.Err. T-stat. P-value 0.2454 0.0969 2.5315 0.0131 0.2349 0.0840 2.7944 0.0062 -0.0048 0.0077 -0.6235 0.5345 -0.0257 0.0492 -0.5227 0.6024 -0.0248 0.0161 -1.5393 0.1272 -0.0018 0.0107 -0.1645 0.8697 0.0102 0.0070 1.4525 0.1498 0.0060 0.0073 0.8136 0.4180 0.0138 0.0060 2.312** 0.0230 -0.0412 0.0260 -1.5853 0.1163 -0.0200 0.0263 -0.7607 0.4488 0.0346 0.0505 0.6860 0.4944 Control Varaibles 0.028 0.023 1.200 0.233 0.017 0.021 0.836 0.405 0.082 0.418 0.197 0.845 -0.052 0.397 -0.131 0.896 -0.027 0.013 -2.056** 0.043 -0.031 0.012 -2.605** 0.011 0.003 0.011 0.251 0.803 0.002 0.011 0.175 0.862 -0.021 0.010 -2.098** 0.039 -0.018 0.010 -1.894* 0.061 -0.012 0.008 -1.423 0.158 -0.013 0.008 -1.668* 0.098 -0.064 0.122 -0.530 0.597 -0.054 0.121 -0.444 0.658 -0.013 0.013 -1.031 0.305 -0.010 0.012 -0.835 0.406 -0.019 0.010 -1.917* 0.058 -0.020 0.009 -2.294** 0.024 -0.215 0.326 -0.659 0.511 -0.392 0.308 -1.270 0.207 -0.063 0.049 -1.278 0.205 -0.089 0.048 -1.865* 0.065 * 10% signif. level ** 5% signif. level *** 1% signif. level 0.0154 0.0097

Regression with full variables Regression with control variables only

114 114 0.151 0.123 exp_T_Size exp_T_Reg (Intercept) Independent Variables exp_N exp_T exp_Similar exp_N_T exp_N_Act exp_N_Size exp_N_Reg exp_T_Act # Obs. Adj. R-squared F-stat. (P-value) Concen VOL DCR CPM PUB TBQ DS ASSETS EAR LVRG MTB

Table 4.2 Statistical summary of OLS regression results for announcement-day CAR (the announcement-day CAR is used as the dependent variable)

The influence of acquisition experience on acquisition performance : evidence from the European banking industry