Mergers and Acquisitions in European Banking

(1)

Mergers and Acquisitions in

European Banking

An Analysis of Shareholder Wealth Effects

1

Mergers and Acquisitions in

European Banking

An Analysis of Shareholder Wealth Effects

Rense Matthias Keij

12

th

_{July, 2011}

Mergers and Acquisitions in

European Banking

(2)

Mergers and Acquisitions in European Banking

Mergers and Acquisitions in

European Banking

An Analysis of Shareholde

This thesis adds to the literature on M&A in the European banking sector by exploring the effects of both activity and geographical relatedness in a standard event

including the concept of institutional distance shareholders and both forms of relatedness. Furthermore, unrelatedness on activity

deals results in the most positive shareholder wealth be associated with the most negative results.

important moderator in cross-border deal performance does not turn out to be significant. Contrary, low market-to-book ratio and cash as method of payment are associated with shareholder value creation.

M&A, European banking, Shareholder value, Relatedness,

Author Supervisor Co

Date

University of Groningen

MSc. International Business & Management Faculty of Economics and Business

2

Master’s Thesis

gers and Acquisitions in

European Banking

An Analysis of Shareholder Wealth Effects

Abstract

to the literature on M&A in the European banking sector by exploring the effects of both activity and geographical relatedness in a standard event-study. It contributes to the literature by institutional distance_{on the relationship between abnormal returns to} shareholders and both forms of relatedness. The data shows that focused deals are value destroying.

relatedness on activity but relatedness on geography in the form of positive shareholder wealth effects, whereas focused cross

negative results. Furthermore, institutional distance although considered an border deal performance does not turn out to be significant. Contrary, low book ratio and cash as method of payment are associated with shareholder value creation.

Key words:

M&A, European banking, Shareholder value, Relatedness, Institutional d

Author Rense Matthias Keij Supervisor Dr. B. Qin

Co-assessor Dr. W. Westerman Date 12th_{July, 2011}

University of Groningen

MSc. International Business & Management Faculty of Economics and Business

Uppsala University MSc. Economics & Business Faculty of Economics & Business

gers and Acquisitions in

European Banking

r Wealth Effects

to the literature on M&A in the European banking sector by exploring the effects of both It contributes to the literature by on the relationship between abnormal returns to The data shows that focused deals are value destroying. of diversified domestic cross-border deals tend to Furthermore, institutional distance although considered an border deal performance does not turn out to be significant. Contrary, low book ratio and cash as method of payment are associated with shareholder value creation.

utional distance

(3)

3

1. Introduction

The number of banks in the European Union fell over the period 1985 to 2003 from approximately 12,670 to slightly more than 7,400, a reduction of 42% (ECB, 2000; 2004). This trend is not remarkable realizing that the number of banks per 1,000 inhabitants in Europe is almost double (0.49) the number in the United States (0.27), suggesting that there is room for consolidation potential through Mergers and Acquisitions (M&A)1_{in the European Union}

(Altunbas & Ibáñez, 2004; European Central Bank, 2004). The introduction of the euro and the creation of a single market for financial services by the European Union strongly stimulate this process. More recently the European Central Bank (ECB) has indicated that a further integration through cross-border M&As in the banking sector is one of the main objectives pursued by the ECB (Trichet, 2007). The main rationale for this consolidation process lies in the efficiency improvements that can be achieved. More specifically, banks try to improve earnings and reduce expenses by achieving increased market power, reduced earnings volatility, and scale and scope economies by e.g. cutting costs through elimination of overlapping operations and consolidating backroom operations. The ECB therefore supports M&A in order to achieve a more efficient banking sector.

Despite the promised positive effects and increasing level of M&A transactions in the European banking sector, research does not uniformly support managers’ apparent enthusiasm for the practice, with the impact of acquisitions on acquiring firm performance remaining ‘inconclusive’ (e.g. Haspeslagh & Jemison, 1991; Roll, 1988; Sirower, 1997). A consensus of estimates, in fact, places the M&A failure rate2_{somewhere in the range of 60 to 80%, a figure}

that Moeller et al. (2005) translate into annual losses in the range of $60 billion. Empirical evidence from the US banking industry on M&A operations provide mixed results and most of the results show no improvement in shareholder returns due to M&A (Amel et al., 2004; Berger et al., 1999; DeLong & DeYoung, 2007). Evidence from European banking sector so far also seem to conclude that positive performance improvements are, at best, very limited if not statistically insignificantly proven (Altunbas & Ibáñez, 2004; Vander Vennet, 1996).

Since it is generally claimed that the primary goal of publically listed organizations is to create value for the shareholder, then based on aforementioned, one might ask whether we can speak of value creation in European M&A practices. As it is also mentioned in the works of Pilloff & Santomero (1997) and Oladepo (2010), “has the industry followed a path of massive restructuring on a misguided belief of value gains? Is management in this sector just incompetent? Or, are they merely lying to shareholders about the effect of their activity on

1_{The terms ‘merger’, ‘acquisition’, ‘takeover’ and ‘M&A’ in this paper are used synonymously.}

(6)

6

shareholder value?” The broad dispersion of findings around the zero return to buyers suggests that executives should approach the activity of M&A with caution (Bruner, 2002).

The process and actual performance outcome of M&A is an important topic and better insight into the determinants of either successful or failed post-merger results can help managers as well as shareholders to make better decisions concerning M&A. This paper aims to identify characteristics of the deal or the transaction parties that have explanatory power for bidder bank returns in EU between 1996 and 2009.It contributes to the research on European banking sector M&A activity because there are to date very few articles that try to identify such variables for bank transactions. Research on antecedents of successful M&A transactions will therefore add to the academic field and hopefully in the future a significant and generalizable theory will result. This, moreover will help managers to make better decisions when it comes to the selection of targets for M&As in the European banking sector, thereby improving the shareholder value creation.

Besides the overall performance of M&A deals for acquiring shareholders in the European banking sector, this paper will further address aspects of relatedness. Specifically, activity relatedness (diversification vs. focus) and geographical relatedness in a transaction (domestic vs. cross-border deals) and their effect on the variation in value creation. Furthermore, and up until now not previously done in this type of M&A research and therefore a contribution to the literature is the introduction of a moderator effect of institutional distance on the relationship between M&A performance and relatedness in case of cross-border deals. This institutional distance between the countries of the acquiring and target bank is hypothesized to explain variation in the M&A performance. Consequently, the research question of this thesis is stated as follows:

What is the overall performance of M&A transactions in the European banking sector, what is the effect of activity and geographical relatedness on this performance, and does institutional distance moderate the performance pattern?

(7)

7

These findings show that focused deals are value destroying for bidder firm shareholders. Diversification has a weak significant positive association with shareholder value, and strong significance when the deal takes place in a domestic setting. Focused cross-border deals turn out to be the most value destroying for bidder firm shareholders among the merger types examined. Institutional distance was expected to be related to weaker performance the larger the distance between the bidder and target firms’ countries. This study does not confirm this relationship due to insignificant results. Lastly, cash as method of payment and low market-to-book ratios are associated positively related to shareholder value creation.

(8)

8

2. Mergers, Acquisitions and Shareholder Value Creation

Merger and acquisition activity, as the literature of industrial organizations and strategic management suggest, is motivated by improving the performance of the acquiring firm or the new combined entity. In this respect, the most common theoretical rationale for M&A activity is the search for synergy where shareholders benefit when the value of the consolidated post-merger entity is larger than the sum of the two separate pre-post-merger companies alone. In economic theory three broad areas are often mentioned as to where this value creation may come from, namely improved efficiency, increased market power, or larger diversification.

The key synergy potential, especially for the banking sector, lies in the improvement of efficiency by exploitation of scale and scope economies. Scale economies arise when the unit cost savings increase the more the scale of an activity is expanded. These can be attained through asset divesture, such as eliminating overlapping operations, personnel, or management practices or consolidating back-office activities thereby reducing costs. Economies of scale in the production process are often mentioned, but they may also be achieved in other functional areas of a business (e.g., R&D, distribution, sales or administrative activities) through the spreading of fixed costs over a higher total volume (Shepherd, 1979). Economies of scope, on the other hand have to do with cost savings or revenue improvements due to an increase in variety of activities of the company. A firm, through a M&A transaction may acquire complimentary resources and capabilities which can be leveraged with those of the capabilities of the acquirer and so enhance the competitive advantage of the post-merger firm. Here, the potential for greater revenue due to cross-selling various products of each of the two firms to customers of the other party is expected to improve profit efficiency.

Value due to a M&A transaction may also stem from increased market power. Besides the increase in market share of targeted customers due to the earlier mentioned larger variety of products and cross-selling, market share also increases since taking over a bank in the same geographical region will reduce competition for the surviving organization. It should be noted that the European Union has a Competition Law3_{in effect that will prohibit mergers with}

substantial anti-competitive effects or monopoly formation. However, despite these policies there is still considerable room for value creation in the European banking sector due to increased market power through external growth.

Lastly, value creation may come from a rise in the level of bank diversification. This may be due to increased geographical reach, either more domestic regions or otherwise expanding the

(9)

9

market through cross-border acquisitions. In line with the efficiency gains due to economies of scope, an increase in the breadth of services offered or the addition of newly acquired assets facilitates diversification. The advantage of diversification is that it may stabilize returns, and lower volatility may reduce bankruptcy costs. Taxes paid can be lower in case the firm faces a convex tax shield4_{, and some risky, yet profitable, activities such as lending may be expanded}

without additional capital needed. All the above aspects are the motives for M&As and are supposed to lead to value creation for the shareholders, the ultimate goal of listed organizations.5

To date, most of the empirical studies that have been conducted on the post-acquisition performance, however, do not find significant positive value effects for the shareholders. Whereas for the target firm shareholders there often seems to be some value gains, for the acquiring firm shareholders mergers do not provide them with a return greater than it would receive from other investment-production activities with similar levels of risk. As Lubatkin appealingly stated, “If no […value creation for shareholders is created], as empirical studies completed mostly in the field of finance conclude, then why do firms continue to merge? If mergers do provide real benefits, as the literature of industrial organizations and strategic management suggest, then why have the empirical studies not found any evidence of real benefits?” (1983, p. 218).

Although his work dates back about three decades, the more recent findings are still inconclusive concerning the positive value effects of mergers and acquisitions. The empirical studies can broadly be categorized in three ways, depending on their approach in measuring performance. One way is to measure value creation on the basis of accounting measures of profitability, such as cost (Altunbas & Ibáñez, 2004; Fiordelisi & Molyneux, 2007; Vander Vennet, 1996; 2002), return on assets or return on equity (Altunbas & Ibáñez, 2004; Beccalli & Frantz, 2009; Vander Vennet, 1996), profit X-efficiency (Beccalli & Frantz, 2009; Fiordelisi & Molyneux, 2007; Huizinga et al., 2001; Vander Vennet, 1996) amongst others. In these studies, it is customary to determine whether the acquirers outperform their peers, where peers are the non-acquiring firms in the same industry and similar in size and profitability prior to the deal. Secondly, and also mostly done by testing performance in relation to non-acquiring peers is the approach of using long-term event-study methodology using stock performance of the acquiring firm for up to five years after the announcement. Lastly, the most widely used method in studies is the short-run event-study (Beitel & Schiereck, 2001a; Beitel et al., 2004; Cybo-Ottone &

4_{A tax schedule where the relative amount of taxes payable to total taxable income increases the higher} this taxable income is. A firm has a convex tax function if it has tax preference items (e.g. tax loss carry forwards or tax credits) or if it has income in the progressive region of the statutory tax schedule. 5_{It appears that the most popular assumption in the academic field of finance is that companies should act}

(10)

10

(11)

11

3. Literature Review and Hypotheses Development

This section presents a short overview of event studies using abnormal returns on the basis of daily equity returns to shareholders. Since the focus in this paper is on the short-term abnormal returns, results on the basis of monthly returns will not be discussed. Similarly, the results about target firm returns in studies are also excluded, since the focus in this paper rests exclusively on acquiring banks’ shareholder value creation.

The findings of event studies dealing with M&A in the banking sector are mixed (see Appendix A for an overview of event-studies done on European banking sector M&As). While some results are significantly positive in terms of value creation to shareholders (e.g. Cybo-Ottone & Murgia, 2000), the majority of event studies find negative or insignificant results, especially for the cumulative abnormal returns to acquirer bank shareholders. Due to deregulation via the Second Banking Directive (1989)6_{, the creation of a single market for}

financial services through the Financial Services Action Plan (1999-2004) and the introduction of the Euro, especially the cross-border or international aspect of M&A is becoming more interesting for research. Still, to date relatively little research has been conducted about EU banking M&A value creation so far. Papers by Tourani-Rad & Van Beek (1999), Cybo-Ottone & Murgia (2000), Beitel & Schiereck (2001b), Beitel et al. (2004) and Ismail & Davidson (2005) apply event study methodology in order to analyze announcement effects of European bank mergers and acquisitions. All of them focus on the question whether bank M&As in Europe have created or destroyed shareholder value. This paper will also be interested in the value creation of all the selected M&A deals in the European banking sector. Despite the “overwhelming majority” of studies find that “M&A activity does not positively contribute to the acquiring firm’s performance” (King et al., 2004) and mixed results for the European banking sector as explained afore, the first hypothesis will be based on the theoretical rationale as described in section 2 on the value drivers of M&A. For that reason, the M&A related synergy potential in the European banking sector through improved efficiency, increased market power and diversification in combination with the assumption that managers have as their primary goal the creation of shareholder value will lead to H1 and is thus:

H1: M&A transactions in the European banking sector have a positive effect on the value creation of acquiring firm shareholders.

(12)

12

Besides the fact that the outcomes of event studies are mixed, there is also a large amount of variation in hypotheses used to explain differences in M&A outcomes. Unfortunately, the studies on post-acquisition performance have not (yet) identified antecedents that can be used to predict post-acquisition performance in a consistent manner (Hitt et al., 1998; Hoskisson et al., 1994; Sirower, 1997). And according to the work of Beitel et al. (2004) none of the EU studies done up till that time have provided an in-depth analysis of such antecedent. Such an extended analysis can, however, address the question whether the market is able to differentiate among the M&A transactions that ultimately create value and those that fail to do so.

The extended analysis in this paper is based on the concept of relatedness.7_{This is}

underpinned by the results of McCarthy & Weitzel (2011)8_{, who state that ‘product}

diversification’ [activity unrelated] and ‘geographical expansion’ [geographically unrelated] are not only the “two most commonly announced merger motives”, but they are even most likely to be the only announced reasons for a takeover by firms. Shareholders, however, do not seem to receive these merger motives with so much enthusiasm, and deals including these motives are each associated with significantly lower announcement returns9_{. Hence, in this paper these}

aspects of relatedness will be incorporated into such an analysis, thereby trying to explain variation in the outcomes of sample abnormal return variation. As the preceding indicates this idea of relatedness is twofold. First, it has to do with the activity of the two parties that merge and to what extent these are the same (focused deal) or different (diversified deal). Secondly, there can be relatedness in terms of geographical choice of the target firm. Consequently, the differences between domestic deals on the one hand and cross-border deals on the other hand will be included in the analysis.

Next, the interactive effect of combining these two forms of relatedness in relation to the M&A post-acquisition performance is tested. Lastly, a moderator effect will be investigated where the level of similarity (institutional distance) between two countries in case of a cross-border deal is expected to have an impact on the value creation due to M&A. These variables will be explained in the subsequent section, followed by the hypotheses used in this paper.

3.1. Activity Relatedness: Focus versus Diversification

In terms of activity relatedness, a M&A deal that leads to an expansion into new services or products is described as a form of diversification, whereas focus encompasses a deal where a target firm is acquired that is similar in terms of resources, products or markets. Diversification

7_{Relatedness has to be understood in terms of the degree of similarity between (two) characteristics.} 8_{By McCarthy & Weitzel. A version of this paper was presented at the 2009 SOM PhD Conference,}

Groningen, (March 2009)

(13)

13

has often been seen by academics as a value generating strategy (Martynova & Renneboog, 2008), since it has the potential of reducing volatility in earnings as well as the risk of bankruptcy (Shleifer & Vishny, 1992), both expected to positively influence shareholder wealth.

However, empirical evidence on the diversification effect on post-acquisition performance shows that diversification is not costless (Datta et al., 1991). The cost of control and coordination in the newly formed entity may rise and production inefficiencies may result (Lancaster, 1990). These costs rise with the level of diversification (Hitt et al., 1994) and thus have a negative effect on shareholder wealth to the extent that at a certain point in time the diversification benefits are outweighed by these costs, and therefore result in a destruction of shareholder value in terms of stock price decline. Research indicates that some acquirers benefit from a diversified deal but that, on average, most firms do not (Loughran & Vijh, 1997). Maquieira, Megginson & Mail (1998) found negative, but insignificant, returns to buyers in diversified deals. The preponderance of M&A research shows that buying unrelated firms leads to value destruction or no value creation at best (Datta, 1991; Lubatkin, 1983; Salter & Weinhold, 1979). European Bank M&A research by Beitel et al. (2004) and Ismael & Davidson (2005) show that bidding banks are more successful in focused transactions and thus significantly determines the value-creation in European bank M&As. Therefore, based on the market power hypothesis and the synergy hypothesis acquiring bank firms can succeed in reducing price competition in the European market by acquiring some of its European competitors, and thus improve bank performance. Benefits can also come from cutting operating expenses via economies of scale or consolidating overlapping business and operations, such as back office operations. Improved bank performance is expected to result in positive shareholder reaction to an announcement of a focused deal. This leads to the following hypothesis in this research:

H2: Activity diversification in the European banking sector M&A is negatively related to shareholder value, compared to focused deals which have a (more) positive impact on shareholder value.

3.2. Geographical Relatedness: Domestic versus Cross-border Deals

(14)

14

initiatives at the EU designed to stimulate financial integration through creation of a single market for financial services and the introduction of the euro suggests that there is considerable consolidation potential left (Goddard et al., 2007), and thus shareholder value creation. The event-study on EU banking M&A by Ismael & Davidson (2005) finds positive, although insignificant shareholder value effects with cross-border deals.

However, as with the previous section about relatedness of activity, most studies document poor performance for cross-border mergers. Child et al. (2001) suggests that cross-border mergers are more likely to encounter ‘unforeseen and insurmountable challenges’. Chatterjee, in a study with Aw (2004) reports that domestic takeovers appear to be more successful taking into consideration the better management of a common language, politics and accounting; which is probably a serious obstacle in the case of cross-border takeovers, where many differences might be observed, resulting in increased agency problems (Denis et al., 2002). Softer factors such as the cultural fit between two companies and different management styles present more difficulties in cross-border transactions, which increases the uncertainty associated with the transaction. Campa & Hernando (2004) find that domestic deals show higher returns than cross-border mergers in the financial sector (consistent across a number of different event windows). Other works that find negative results for cross-border acquisitions are for example Moeller & Slingemann (2005), Martynova & Renneboog (2006) and in European studies by Cybo-Ottone & Murgia (2000), Beitel et al. (2004) and Beccalli & Frantz (2009).

In line with the above mentioned empirical evidence, in this paper the stance is taken that cross-border deals are negatively related to acquirer returns since considerable differences exist between European countries in terms of management style, national and organizational culture and institutional characteristics. Although room for consolidation and shareholder value seems to exist and is stimulated by the European Union, these country differences are hypothesized to result in negative shareholder value and a geographical focus on the domestic market has higher value potential. This leads to the following hypothesis in this research:

H3: Cross-border deals in European banking sector M&A are negatively related to shareholder value, compared to domestic deals which have a (more) positive impact on shareholder value.

3.3. Activity and Geographic Relatedness Combined

(15)

15

acquisitions) lead to a strong diversification discount of about 24%. Lins & Servaes (1999) conclude that the benefits of diversification are not enough to offset the substantial agency costs impaired in case of a cross-border acquisition. Furthermore, DeLong (2001) finds that mergers that focus both activity and geography enhance buyer’s share value by 2% to 3% more than other types of mergers.

Based on these findings and by taking into account H2 and H3, sound logic results in the fourth hypothesis in this research:

H4: Activity as well as geographically unrelated (cross-border diversified) deals in European banking sector M&A are most negatively related to shareholder value, whereas activity as well as geographically related (domestic focused) deals have the highest positive impact on shareholder value.

3.4. Country Distances

Finally, this study will include the concept of institutional distance into the analysis. When comparing cross-border deals with domestic deals in terms of shareholder value creation as will be done with H3, there is no way of detecting any variation between these dichotomous options. Emilia di Patri (2009) in her discussion paper on the work of Beccalli & Frantz (2009)10_suggests

that measures of institutional proximity matter and should be included in the analysis. Furthermore, Datta & Puia (1995) suggest that significant national differences may be perceived by investors as a factor increasing the postacquisition administrative and consolidation costs. Inadequate understanding of the foreign market and institutional characteristics may cause acquirers to overpay for the target. This combination of potentially large post-acquisition costs and pre-acquisition premiums may have an adverse effect on the acquirer firms’ market value and thus shareholder value. Therefore, a moderator effect11_{will be included in case of a}

cross-border deal to investigate whether countries that are more similar in terms of institutional architecture have better performance results than deals between countries that are more dissimilar. Since the geographically unrelated (cross-border) deals are expected to have a negative impact on shareholder value as stated in H3, the following and final hypothesis of this research including the moderation effect of institutional distance will be as follows:

10_{The authors found a slight deterioration in ROE, cash flow return and profit efficiency and a marked} improvement in cost efficiency due to M&A in European banking. Furthermore, a particular negative trend was found for cross-border deals, so geographical relatedness matters.

(16)

H5: High institutional distance between cross

negative impact on shareholder value, whereas low institutional distance has a less impact on shareholder value.

3.5. Research Design

To summarize the above section where the hypotheses in this research are developed 1 gives a graphical overview of the proposed res

Hypotheses 1 will be tested u

tested using mean univariate analysis, mean regression analysis, whereas hypothesis 5 w signs of the relationships ment

of the boxes.

Figure 1

Now that the research in this article has been presented, supported by a literature review of empirical evidence and the final research de

and methodology applied in order to test the hypotheses.

16

High institutional distance between cross-border European banking M&A part impact on shareholder value, whereas low institutional distance has a less

To summarize the above section where the hypotheses in this research are developed gives a graphical overview of the proposed research design presented in this paper

sing mean univariate analysis only, hypotheses

mean univariate analysis, mean comparative statistics as well as regression analysis, whereas hypothesis 5 will only be answered using the latter.

tioned in the hypotheses are mentioned in the

Graphical presentation of the hypotheses

Now that the research in this article has been presented, supported by a literature review of empirical evidence and the final research design, the next section will deal with the data selected and methodology applied in order to test the hypotheses.

ing M&A parties has a impact on shareholder value, whereas low institutional distance has a less negative

To summarize the above section where the hypotheses in this research are developed, Figure earch design presented in this paper. ypotheses 2 until 4 will be comparative statistics as well as multiple ill only be answered using the latter. The expected e upper right corner

Graphical presentation of the hypotheses

(17)

17

4. Data & Methodology

4.1. Data Sources

The databases used to extract data on deals are Zephyr, Thomson Datastream Advance (from now on called Datastream) and Eurostat. Zephyr, a database of Bureau van Dijk electronic publishing, is used to acquire a sample of M&A deals. This database provides e.g. the announcement date, country of origin as well as industry and subsector of the parties involved. Daily stock returns as well as daily market index data are directly extracted from Datastream, which is considered as the world’s largest and most respected financial statistical database. Lastly, real GDP growth rate data is collected through Eurostat, the statistics organization of the European Union and the largest source of statistical information about European member states.

4.2. Sample

An important step in the research is to select the sample of bank acquisitions for the study. The empirical analysis of this study is based on a final sample of 260 M&A announcements over the period 1996-2009. 1996 was chosen since it is the first year Zephyr has data on mergers and acquisitions. The following criteria have been taken into account:

1. They have been announced between January 1, 1996 and December 31, 2009 and are completed.

2. The acquiring firm is a publically listed European bank (using NAICS 2007 industry code 52211 which stands for commercial banks).

3. The target is a bank, financial service provider (FSP) or insurance company (using NAICS 2007 first two digit 52 code).

4. The deal value exceeds 1 million euro.

5. The acquisition results in a controlling majority for the acquiring firm of at least 51% in the target firm.

6. The bidding firm is listed for the total time of the event, so 260 days before and 20 days after the announcement date in order to retrieve data necessary for statistical analysis.

(18)

their strong M&A activity in the sample

dataset of 260 M&A deals is the basis for this paper.

Table 1 Descriptive statisti

Table 1 reports this M&A sample by geographical origin of th

by merger completion year (right column). Notable is that Italy has made most acquisitions in the period with over 25% of the total sample, followed by

(10%). Czech Republic, Finland and

M&A announcement. Furthermore, as shown in

geographical relatedness concepts, we see that about 60% of the deals are activity focused and 40% diversifying deals. Cross

56% domestic deals. Combining the two forms of relatedness results in the four options (displayed in the boxes) being rather equally divided, except for di

having a 16% proportion compared to about 30% for the other three deal types. Mergers and Acquisitions in European Banking

18

their strong M&A activity in the sample period. Finally, based on the six criteria a final sample dataset of 260 M&A deals is the basis for this paper.

Descriptive statistics for sample of 260 M&A deals

reports this M&A sample by geographical origin of the bidding bank (left column) and by merger completion year (right column). Notable is that Italy has made most acquisitions in

of the total sample, followed by the United Kingdom (10

(10%). Czech Republic, Finland and Luxembourg are strongly underrepresented with just one M&A announcement. Furthermore, as shown in Graph 1, when looking at the activity and geographical relatedness concepts, we see that about 60% of the deals are activity focused

eals. Cross-border deals amount to 44% of the total sample compared to 56% domestic deals. Combining the two forms of relatedness results in the four options (displayed in the boxes) being rather equally divided, except for diversifying cross

having a 16% proportion compared to about 30% for the other three deal types.

criteria a final sample

cs for sample of 260 M&A deals

(19)

19

Graph 1 Categories of bank mergers based on activity and geographical relatedness

In Figure 2 the amount of deals per year are shown, as well as a partitioning into domestic and cross-border deals. The trend line is included to give a visual indication of the variation in number of deals over the years, where the years 2000 and 2006 show peaks of total annual M&A deals.

Figure 2 Graphical representation of the dataset including real GDP growth

(20)

20

Lastly, real GDP growth data retrieved from Eurostat12_{has been included with percentages}

indicated on the right-hand side of the graph, giving evidence of the fact that the trend line total M&As and the real GDP growth clearly move in the same way. In other words, an increase in M&A activity goes hand in hand with improved economic development.

4.3. Variables

Here a short description is given for the variables that are used in the research. Besides the dependent variable, cumulative abnormal returns, the independent variables explained in the literature review will be discussed. Furthermore, a large number of antecedents for explaining variation in shareholder wealth results due to M&A announcements have been dealt with in other research by academics. Several of these variables will be included in the multiple regression analysis in order to control for these effects on the hypotheses in this paper.

4.3.1. Dependent Variable

The dependent variable in this research is a metric indicating the financial performance, or otherwise stated the shareholder value effect for the selected M&A transactions. Here stock-market performance is assessed by looking at either the average cumulative abnormal returns (average CARs) in the simple mean comparison tests of the hypotheses or the cumulative abnormal return (CAR) for each transaction in the multiple regression analysis. These metrics are based on daily Return Index (RI)13_{data retrieved from Datastream. The exact steps towards}

calculating the CAARs and CARs by the use of RI data will be discussed in methodology section.

4.3.2. Independent Variables

The independent variables used in this paper are the activity and geographical relatedness aspects. Zephyr provides information for each and every transaction about the industry classification of the acquirer and the target. In case the target firm has the same NAICS 2007 industry code (in this case 522110, “commercial banking”), a binary dummy variable FOCUS takes the value of 1 indicating a focused deal, and 0 when these are not the same and thus is classified as a diversified deal in my research design.

The same way of reasoning applies to the geographical relatedness, where according to the acquirer and target country code specified by Zephyr, a transaction is either cross-border or domestic and so a corresponding binary dummy variable CROSS receives the value of 1 or 0 respectively.

12 _{http://epp.eurostat.ec.europa.eu/tgm/table.do?tab=table&init=1&plugin=1&language=en& pcode=} tsieb020

(21)

21

Figure 3 Variable Definitions

Lastly, in a separate multiple regression the above mentioned variables will be replaced by four new variables (Var 2.1, 2.2, 3.1 and 3.2), which represent every combination of the two forms of relatedness. These will be used to test the interactive effect represented in hypothesis 4. Again, the value of 1 is given in case a transaction applies to both forms of relatedness proposed, and a 0 otherwise. A detailed overview of variables is given in Figure 3.

4.3.3. Moderator Variable

Institutional distance is used as the moderator variable in this study. This distance in this paper is based on the Worldwide Governance Indicators (WGI) which is a long-standing research project to develop cross-country indicators of governance as developed by Kaufmann et al. (2006; 2010).14_{It consists of six composite indicators of broad dimensions of governance:}

Voice and accountability (VAC), political instability and violence (PIV), government effectiveness (GEF), regulatory quality (REG), rule of law (LAW), and control of corruption (COR). These indicators all measure another aspect of governance of a country. The aggregate governance indicators cover 215 countries and territories since 1996. For some years, data is missing since in the starting years until 2002 this study was done only once every two years . For these missing years values are proxied by interpolating data.

(22)

22

Using these WGI as the basis for the institutional distance metric in this paper is underpinned by the following arguments. First of all, to date, these indicators cover the broadest array of institutional aspects and years of data. Moreover, the indicators are “based on several hundred variables obtained from 31 separate data sources, capturing governance perceptions as reported by survey respondents, non-governmental organizations, commercial business information providers, and public sector organizations worldwide” (Kaufmann et al., 2010 ,p. 2). This means that these aggregate estimates are informative about changes over time in the relative institutional positions of individual countries (Dikova & Van Witteloostuijn, 2007).

The WGI presents data for every year and country for each of the six indicators separately. In order to come to a single metric to be used in this paper as the moderator variable, a mathematical calculation will be used to combine the six data sources and so end up with one value. In line with Lensink et al. (2008) institutional distance is measured by the Euclidian distance between the institutions of the home and the host country by applying equation 1:

= ∑௜ୀଵ ( ௜− ௜)ଶ

…଺ (Eq. 1)

For each cross-border deal the appropriate institutional distance value is selected from the compiled database15_{and included in the multiple regression analysis in order to answer}

hypothesis 5 of this research. The taken stance in this paper is that higher values for governance or institutional distance will lead to lower post-acquisition performance.

4.3.4. Control Variables

King et al. (2004) states in the end of his work that future M&A researchers are advised to use variables from existing M&A research as a basis for building new models of post-acquisition performance. Since variables of demonstrated significance are often excluded from M&A studies, many research models may form biased conclusions (see Griffiths et al., 1993). Therefore, several control variables will be included in the multiple regression analysis in the hope to receive more reliable results.

Relative size (SIZE): There is substantial evidence showing that the potential to create value from an acquisition depends upon the relative size of the merging firms. Post-acquisition performance may be lower for larger deals resulting from an increased complexity in these deals (Akhavein et al., 1997). Beitel et al. (2004) argues that for bidding banks, the relative size does not significantly drive M&A-success although transactions with smaller targets seem to be more value creating. Still, if targets are relatively small in relation to the bidding firm, attempts to restructure the target may be unnecessary or will have a limited impact to the shareholder value

(23)

23

creation. In this paper SIZE, as a measure of the significance of a transaction, is the ratio of the transaction value (V) to the market value of the acquiring firm (MV). In this paper the stance is taken that the larger the relative size of the acquirer, the larger the shareholder value creation. For larger deals the potential to be more efficient by cutting overlapping business necessary to support the servicing of financial products to clients as well as taking over large competitors will improve performance.

Method of payment (PAY): This variable has turned out to be a significant determinant of M&A success. The use of stock as payment for an M&A deal is believed to signal to investors that the bidding firm managers consider their equity to be overvalued (Loughran & Vijh, 1997; Shleifer & Vishny, 2003). Consequently, equity deals tend to be related with weakly performing M&A (Kiymaz, 2004; Lubatkin, 1983), whereas cash payment is considered to positively influence shareholders. The PAY variable in this research is a binary variable where the value of 1 stands for 100% cash financed, and 0 for a mix of cash and equity and PAY is expected to have a positive sign.

Market-to-book ratio (MB): Rau & Vermaelen (1998) argue that glamour firms (high MB-ratio) perform less well than value firms (low MB-MB-ratio) since good past performance, as hypothesized by the authors, is overextrapolated by the stock market when reacting to a M&A transaction and moreover, management overestimates its own liabilities. The MB-ratio four weeks prior to the M&A announcement is used and is retrieved from Datastream calculated four weeks prior to the announcement. Based on the concepts of glamour and value firms, a negative sign is expected where the higher the MB ratio, the lower the M&A performance.

Leverage (LEV): Jensen (1986) shows that debt reduces agency costs, in a way that it reduces the cost of free cash flow by limiting the cash flow available for spending at the discretion of managers. Hereby, managers are bonding their promise to pay out future cash flows in a way that cannot be accomplished by simple dividend increases. By increasing the debt, the managers are compelled to work efficient and work towards performance in order to make the interest payments and pay back the loans. Other empirical work by Hitt et al (1998) also shows that leverage affects the outcome of acquisitions, with higher leverage resulting in better post-acquisition performance compared to lower leverage. LEV is calculated by dividing acquirer total debt by acquirer total assets, both extracted from Datastream and four weeks prior to the announcement. It is expected to negatively relate to abnormal returns.

(24)

24

premiums for targets, which in the end make it more difficult to fulfill the desired and promised rates of return to shareholders. Therefore, higher economic state announcement are expected to negatively influence the abnormal return pattern. This paper includes the real GDP growth in Europe for the appropriate year of each transaction in order to test for the economic state.16

Governance system (CIVIL): According to Martynova & Renneboog (2008) part of the value creation in cross-border M&A deals comes from a synergy effect which arises due to bidders from countries with stronger shareholder orientation improving the governance system of the targets. La Porta et al. (2000) finds that acquisitions are more efficient if they are located in a common law country compared to any of the civil law systems in their study. Investor protection for the target firms may be improved, which will benefit the shareholders of the bidding firm. The other way around, civil law country targets are expected to result in lower shareholder value creation compared to common law countries, and so a negative sign for CIVIL is expected to arise in relation to abnormal returns.

Announcement and rumor date (ANN = RUM): In many instances there is a risk of an information leakage prior to the announcement date of a M&A deal. Sometimes this might be considered illegal, and insider trading may be the result. In these circumstances the pure announcement returns will be heavily influenced. The adjustment of the price to new information to the market or to some participants of the market (insiders) will have adapted the price already accordingly. Therefore, it is important to control for this effect when investigating abnormal returns around the announcement date of M&A deals. Based on the information that the Zephyr database provides concerning announcement dates and rumor dates for each of the deals in the sample, this effect is included in this study. Neither a negative, nor a positive sign is expected since they depend on the sentiment of those shareholders that dealt on rumor information, but a significant relationship between this variable and the abnormal returns is sought.

(25)

25

5. Methodology

Mean difference tests are used in order to find significant differences with respect to average cumulative abnormal returns (CARs). I also run multiple cross-sectional regressions to test for several variables simultaneously. The dependent variable in this analysis is the list of CARs for bidder firm shareholders of each of the transactions. First the event study methodology is explained which determines the way the dependent variable CARs is computed.

5.1. The Event Study Methodology

The methodology used in this thesis is an event study in order to calculate the abnormal returns to shareholders of the bidding firm for each transaction. An event study is an analysis of whether there is a reaction in financial markets to occurrences of a given type of event (merger or acquisition in this thesis) that is hypothesized to affect the public firms’ market values (Grinblatt & Titman, 2002). The usefulness of an event study lies in the assumption that the effect of an event will be reflected immediately in asset prices in the marketplace (Campbell et al., 1997). The assumption of market efficiency is important in this respect. On the rumor or announcement date of a merger or acquisition, the prices will immediately change to reflect the new information.

The event day is either the public announcement day, or the next trading day if this announcement date is not an actual trading day such as the weekend or holidays. This event day will be termed day 0 and the event window will be based on this t = 0.

First, I calculate the daily return for all the bidding firms in the sample for the period t=-260 until t=+20 trading days. Here, return data are obtained from Datastream and based on dividends reinvested. The following equation (Eq. 2) represents this first step:

ܴ

=

_೔೟శభ _೔೟

_೔೟

(Eq. 2)

Where: Rjt = Equity return for bank i on day t

Pjt = Closing price (dividends reinvested) for bank i on day t

Pjt+1 = Closing price (dividends reinvested) for bank i on day t+1

(26)

26

their market returns. The CAPM model17_{is used to get an estimate of these normal returns of the}

security, with an estimation window of 240 historical trading days (starting 260 days until 20 days before the announcement day) to calculate the alpha and beta coefficients of each company relative to the market.18_{I use the European banking sector index and the associated daily}

returns as the benchmark market returns. 19_{The CAPM market model equation (Eq. 3) is given}

by:

௜௧ = ௜+ ௜௠௧+ ε௜௧ (Eq. 3)

Where: E(Rit) = Expected equity return for bank i on day t

αi = Intercept coefficient estimated by the least squares regression

βi = Slope coefficients capturing the systematic risk of the bank’s equity

Rmt = Return on the benchmark European banking market index on day t

εit = The error term of the least squares regression

Next, and as indicated before, these estimated parameters of the regression analysis i and i

are used to predict the normal returns for each bank in the event window. When we subtract these predicted normal returns from the actual returns that the company has achieved we end up with the abnormal return for a bank for each day in the event window (ARit). This standard

event study methodology (see Brown & Warner, 1985)with the market model is represented in the next equation (Eq. 4):

A௜௧ =௜௧− (௜+௜௠௧) (Eq. 4)

Now that the daily abnormal returns are calculated for every bank for each t in the sample period (starting at t=-20 and ending t=+20 trading days), I am interested in finding the cumulative abnormal returns (CARiT) of each bank for the pre-defined event windows. By

accumulating the abnormal returns for all the days defined in the event window (T), we find the cumulative abnormal return for event i as follows (Eq. 5):

17 _{In finance, the capital asset pricing model (CAPM) is used to determine the theoretically appropriate} required rate of return of an asset (security, debt, etc).

18_{In this regression analysis using time series data, autocorrelation of the errors can be a problem. It} would possibly give unreliable alpha and beta values, which in turn might influence subsequent statistical analyses in this paper. Autocorrelation is tested for by using the Durbin-Watson test for each equity and market index in the sample. Fortunately, none of the results have values below 1 or above 3, and most are close to 2, indicating that autocorrelation is not present.

(27)

27 ௜்=∑ ௜௧

்

௜௧ (Eq. 5)

These CARs are the basis for the statistical tests executed in this paper. For each event window I will test if the average (mean) of all the CARs in the sample differs significantly from 0, indicating the wealth effect to shareholders surrounding the announcement of an M&A deal. This will answer my first hypothesis (H1). Furthermore, subsamples according to type of deal are tested in a similar way, and lastly the differences between average CARs of these subsamples are tested for their difference from zero. In the second form of testing the hypotheses, I will use multiple regression analysis and then all the CARs will be taken together as the dependent variable. More specifics about these two tests will follow in the subsequent sections 5.2 and 5.3.

Finally, in Figure 4 it is shown in what way time is used in this paper, where an indication is given about the beta calculation in the period (-260;-20) and the actual period (-20;+20) wherein the performance around the announcement of a transaction is tested.

Figure 4 Graphical Representation of the Beta and CAR Calculation

5.2. Test 1: Sample Mean Tests

(28)

28

chance – the higher the t-value, the lower the probability of a chance occurrence (strictly speaking, one never proves the hypothesis, one only disproves the “null hypothesis” that the phenomenon is due to chance).20_{The significance levels used in this paper are the 1%, 5% and}

10% levels.

The tests executed are two-tailed, since positive as well as negative results are of importance. Furthermore, an important requirement with using the parametric t-tests is that the underlying separate CARs are normally distributed around the mean. In section 4.5.2 details will be provided about how this paper deals with normality of the CARs. However, in case of deal type comparison the full sample will be split into subsamples which will likely change the normal distribution of these selected CARs. For this reason the analysis will be supplemented with non-parametric mean difference testing using Wilcoxon rank-sum and Mann-Whitney U tests.21

These test results and significance levels will be a robustness check of the results found with the parametric t-test for independent samples.

5.3. Test 2: Multivariate Cross-Section Analysis

I also test the hypotheses within the setting of the following multiple cross-section regression analysis. “Model 1” in equation 6 is as following:

CAR௜ = αଵ+ αଶFOCUS + αଷCROSS + αସSIZE + αହPAY + α଺MB + α଻LEV + α଼GDP +

α9CIVIL+ α10ANN=RUM+ ε݅ (Eq. 6)

CARi is the cumulative abnormal return for the bidder in acquisition announcement i, where

FOCUS is a dummy variable that equals 1 if the acquisition is a focused deal and 0 in case of a diversifying deal. CROSS is a dummy variable that equals 1 or 0 for cross-border deals and domestic deals respectively. Furthermore, I include the control variables SIZE (ratio of deal value V to market value of acquirer MV), PAY (method of payment where 1 represents full cash financed deals and 0 otherwise), MB (market to book value of acquirer), LEV (standing for the leverage ratio of the bidder firm), GDP (numerical variable for real GDP growth rate in Europe), CIVIL (governance system where 1 represents deals where targets are located in civil law countries) and ANN =RUM (value of 1 when announcement date is same as the rumor date and value of 0 otherwise) . Finally, the alpha terms 1 until 10 are the regression coefficients and εi is

the error term of the equation.

20_{Tests of significance also depend on sample-size. The t-values discussed here implicitly assume} relatively large samples of observations, such as more than 100.

(29)

29

To test the interactive effect (H4), I replace the CROSS and FOCUS variables in equation 6 by four new variables that stand for every combination to be made with these two relatedness variables. Therefore “model 2” in equation 7 becomes:

CAR௜ = αଵ+ αଶ

.ଵFOC/CROSS + αଶ.ଶFOC/DOM + αଷ.ଵDIV/CROSS + αଷ.ଶDIV/DOM +

αସSIZE + αହPAY + α଺MB + α଻LEV + α଼GDP + αଽCIVIL + αଵ଴ANN = RUM + ε௜

(Eq. 7)

FOC/CROSS stands for focused cross-border deals, FOC/DOM for focused domestic deals, DIV/CROSS for diversified cross-border deals and DIV/DOM for diversified domestic deals. The other variables are the same as specified for model 1 in equation 6. Now the alpha terms 2 and 3 have been replaced by 2.1, 2.2, 3.1 and 3.2 in order to keep the remaining alpha terms consistent with equation 6.

Finally, equation 8 will test for the moderator effect of institutional distance on the relationship between cross-border deals and the abnormal returns (H5). Compared to equation 6, “CARi (CROSS)” now stands for the subsample of geographically unrelated deals and the

variable DISTANCE is added with alpha term 11. Again, the variables keep the same alpha terms in order to keep them consistent in the equations, despite removing α3CROSS.

CAR௜(CROSS) = αଵ+ αଶFOCUS + αସSIZE + αହPAY + α଺MB + α଻LEV + α଼GDP +

α9CIVIL +α10ANN=RUM+ α11DISTANCE +ε݅ (Eq. 8)

5.4. The Event Windows Used For Testing

Correct timing is likely to be deal-specific and requires careful analysis in each and every case, but this weakens a broad-based cross-sectional analysis, since when time periods varies across transactions, the standardization is lost. In the abnormal return model developed by the author22_{of this paper a large variation in event windows has been included.}23_{Some of the}

graphs that will be presented in the results section of the paper will be based on this larger variation of windows. However, a few windows that are considered important in most of the empirical works done to date have been included in the actual hypotheses testing in this paper. Event-window (-10/0) is chosen to detect rumoring before the announcement date. Often it is

22_{The interactive and user-friendly abnormal return model (excel 2007) including sample means} comparison tests, significance levels, results tables and graphs is available upon request from the author.

(30)

30

seen that before the announcement of a merger or acquisition there are signs in that direction, which might lead to abnormal returns earned before the announcement date. Furthermore, (-1/1) and (-3/3) are included for the pure announcement return and (-1/20) for a wider view on shareholder value creation in terms of time elapsed after the announcement.

5.5. Statistical Requirements of the Data

5.5.1. Missing Data

Missing data for some Datastream variables reduces the total sample from 265 to 260 observations.24_{Furthermore, the distance variable based on WGI studies provides data since}

1996, but until 2002 this study has only been done once every two years. Data for the missing years have been calculated by interpolating the data.

5.5.2. Normality and Extreme Outliers

The assumption of normal distribution of the dependent variable is an important requirement for the reliability of test results done on a dependent variable. In order to test whether the dependent variable is normally distributed, the Jarque-Bera test is executed. This goodness-to-fit measure is based on skewness and kurtosis.25_{Appendix B}_{shows the test results,}

showing that when taking normal (discrete) returns for the banks and the market in the estimation of beta as well as for calculating abnormal returns, the Jarque-Bera results have high values with significance levels of p<0.01. This provides clear evidence that the alternative hypothesis of non-normality can be accepted.

Since it is often mentioned in the literature that logarithmic returns are more likely to have a normal distribution26_{, I calculated the CARs accordingly to test if this will improve the statistical}

properties of the dependent variables. However, still the results indicate non-normality.

Lastly, winsorization is conducted in order to improve the distribution to normality. Hereby, all values beyond two standard deviations from the mean are replaced by the value of two standard deviations from the mean, either below or above the mean. Now, 2,5% on both tails have been replaced and so the problem of (extreme) outliers has been mitigated. The Jarque-Bera test results for both discrete as well as logarithmic returns now show that the dependent

24_{Missing data values included the variables (and Datastream codes) for market-to-book value (MTBV)} and the two variables used for calculating leverage, namely total assets (WC02999) and total debt (WC03255).

25_{The null hypothesis assumes a normal distribution of the skewness being 0 and the excess kurtosis} being 0 (which is the same as a kurtosis of 3). The alternative hypothesis assumes a non-normal distribution.

(31)

31

variables have acceptable levels of distribution to normality. In the statistical tests in this paper I have chosen to work only with the winsorized dependent variable based on normal returns, since logarithmic returns do not improve the normality assumption.

5.5.3. Multicollinearity

To test for Multicollinearity, the correlations between the dependent, independent and control variables are estimated (see Appendix C). Based on the results the highest correlation detectable is only 0.40 between the variables FOC/DOM and FOC/CROSS, which is below the cut-off value of 0.5 used in this thesis.27_{Therefore, none of the proposed variables will be removed}

from the analyses. Besides, I test for multicollinearity by implementing the variance inflation factor (VIF) in each model. I use a cut-off value of five (Stundenmund, 2001) indicating that a figure larger than five is a sign that multicollinearity is a severe problem.

(32)

32

6. Empirical Results

As explained in the methodology the hypotheses in this study will be tested through sample mean tests (H1 until H4) and subsequently through multiple regression analysis. The later including the regressions provide an alternative way to test H2 until H4 by including control variables and furthermore will be used to test H5 with the institutional distance assumption. The robustness check that can be made by testing the H2 until H4 in two ways adds to the reliability of the results found.

In order not to have overlapping information concerning the hypotheses in the two separate results sections of the mean tests (6.1) and the multiple regression tests (6.2), the economic significance of the variables important for the hypotheses will be given in the discussion, presented in the next chapter, where implications of the results for the European banking sector will be outlined. Results of this study will now follow.

6.1. Sample Mean Tests

The mean tests are twofold. First I will discuss the results of univariate analysis, looking at single variables (merger types) and test whether the average of the related CARs is statistically different from zero. Thereafter, the results for comparing the means of two subsamples for their dependency (means are significantly different from each other) will be presented.

6.1.1. Univariate Analysis

Table 2 reports univariate results for the cumulative average abnormal returns (CARs) in event windows 10 days before the announcement until the announcement day (-10/0) as well as (-1/1), (-3/3) and (-1/20).

Panel A testing H1 and H2 indicates that for the total sample of 260 deals none of the results shows any significant abnormal returns different from zero. Still, by looking at the CARs for the four windows, all of them are negative varying from -0.138% with (-1/20) to -0.322% with (‒3/3). These are only very small percentages, but it does give an indication that M&A activity in the European banking sector tends to be slightly value destroying, which is opposed to the sign reflected in H1. Nevertheless, and mainly due to the insignificance, H1 has to be rejected. The value creation or destruction of shareholder value due to M&A transactions in the European banking sector based on these results remains undetermined.

Mergers and Acquisitions in European Banking

Mergers and Acquisitions in

European Banking

An Analysis of Shareholder Wealth Effects

Mergers and Acquisitions in

European Banking

An Analysis of Shareholder Wealth Effects

Rense Matthias Keij

12

July, 2011

Mergers and Acquisitions in

European Banking

Mergers and Acquisitions in

European Banking

An Analysis of Shareholde

Master’s Thesis

gers and Acquisitions in

European Banking

An Analysis of Shareholder Wealth Effects

gers and Acquisitions in

European Banking

r Wealth Effects

Table of Contents

1.

Introduction

2.

Mergers, Acquisitions and Shareholder Value Creation

3.

Literature Review and Hypotheses Development

3.1.

Activity Relatedness: Focus versus Diversification

3.2.

Geographical Relatedness: Domestic versus Cross-border Deals

3.3.

Activity and Geographic Relatedness Combined

3.4.

Country Distances

3.5.

Research Design

4.

Data & Methodology

4.1.

Data Sources

4.2.

Sample

4.3.

Variables

5.

Methodology

5.1.

The Event Study Methodology

ܴ

=

5.2.

Test 1: Sample Mean Tests

5.3.

Test 2: Multivariate Cross-Section Analysis

5.4.

The Event Windows Used For Testing

5.5.

Statistical Requirements of the Data

6.

Empirical Results

6.1.

Sample Mean Tests

_{July, 2011}