University of Groningen
MSc Business Administration – Finance
Master Thesis
“Does Size Matter?”
A study on the European Retail Takeover Sector
Author: Joost van Droffelaar
Mail: jhd.vandroffelaar@gmail.com
Phone: +31655861111
Student number: s1531433
Place and Date: Groningen, January 8, 2013
Supervisor: Dr. B.A. Boonstra
Contents
Contents 2 Figures 3 Tables 3 Abstract 4 1 Introduction 5 1.1 Introduction 5 1.2 Objective of research 7 1.3 Relevance of research 7 1.4 Structure 8 2 Literature review 92.1 Takeover and the Takeover Market 9
2.1.1 Takeovers 9
2.1.2 Efficient Market Hypothesis 11
2.1.3 Modern Portfolio Theory 12
2.1.4 Profitability & takeovers 13
2.2 The Retail Sector & Perspectives on Size 14
2.2.1 The Retail Sector 14
2.2.2 Bidder Size 18
2.2.3 Deal Size 22
3 Methodology and Data 25
3.1 Event Study Methodology 25
3.1.1 Abnormal return 25
3.1.2 Cumulative Average Abnormal returns 27
3.1.3 Test statistic under Null Hypothesis 27
3.2 Variables & Data 28
4 Results 32
4.1 Bidder size 32
4.2 Deal size 34
5 Robustness Analysis 37
5.1 Nonparametric tests 37
5.2 Sample size analysis 39
5.3 Time window analysis 40
5.4 Market return analysis 41
5.5 Relative deal size 42
Figures
Figure 1: EU M&A activity 10
Figure 2: MTP Investment Process (Gupta, Fabozzi & Markowitz, 2002) 13
Figure 3: Development of discount format in EU-27 16
Figure 4: Announced Mergers & Acquisitions: Retail, 1985-2011 17
Figure 5: Acquirer return versus industry index, by size of the deal 23
Figure 6: Time line for event study 25
Figure 7a: 𝜃!-values – Bidder size 33
Figure 7b: 𝜃!-values – Bidder size (market-risk adjusted) 34
Figure 8a: 𝜃!-values – Deal size 35
Figure 8b: 𝜃!-values – Deal size (market-risk adjusted) 36
Figure 9: CAAR relative deal size 43
Tables
Table 1: Merger waves explained 9
Table 2: Summary Statistics 31
Table 3: Summary of Sample Size for different studies 39
Abstract
1
Introduction
1.1 Introduction
In pursuing growth, the retail sector has been one of the most compelling sectors in realising immense expansion and economies of scale through mergers and acquisitions. The big question is however: have all these takeovers have been successful? More specifically, does the size of the firm and the deal really matter? This study explores these questions.
In corporate finance mergers and acquisitions are important events, for both individual firms and the economy as a whole. During the recent financial crisis the world has observed an enormous decline in takeovers, defined as acquisitions of over 50% of a target’s share capital. Takeovers are one of the mayor sources of growth, or at least proven to be a necessity for firms to survive. If a firm is not active in takeovers, the value of the firm’s share will drop, making itself a cheap takeover candidate (Fama & Jensen, 1983).
There are widespread views on the existence of takeover profitability. For instance, there is a popular view of M&A activity destroying value; “The sobering reality is that only about 20% of all mergers really succeed. Most mergers typically erode shareholder wealth… the cold, hard reality that most mergers fail to achieve any real financial returns… very high rate of merger failure… rampant merger failure (Grubb & Lamb, 2000). Executives, consultants and journalists often bring up the popular view of M&A activity destroying value. However, there is also evidence from several studies that disagree with these statements. Bruner (2002) concludes that M&A does pay summarizing evidence from 130 studies from 1971 to 2001. This study investigates whether size, to be specific bidder firm size and deal size, matters in a takeover. Classical perspective, such as i.e. Fama (1970) and Markowitz (1952, 1959) argue that the information about size has no influence. For example, Fama (1970) explains this in his influential article “Efficient Capital Market”, that securities markets reflect information on individual stocks in full. When information arises, it spreads very quickly and is incorporated into the prices of securities.
valuations are not warranted by fundamentals. Moeller, et al. (2004) shows that large firms have higher Tobin’s Q (market value of firm’s assets divided by the book value) and lower BM (book-to-market) ratios than small firms, which according to Dong, et al. (2002) proxy for overvaluation. Second, Roll’s (1986) hubris theory states that managers of bidding firms may suffer from hubris and hence they may overpay in a deal. A third theory concerns synergy gains (Shleifer & Vishny, 2003). Large firms are consistent with negative synergies and small firms exhibit positive synergies (Moeller, 2002). Fourth, the free cash flow theory (Jensen, 1986) states that empire-building managers would rather make acquisitions than increase pay-outs to shareholders.
A sector that has shown strong evidence on creating positive return and creates value through takeover is the Retail sector. This industry consists of individuals and firms engaged in selling of finished products to ‘end user’ consumer. The definition of ‘end user’ is that the person uses the product that has been manufactures and marketed. In the last decennia the retail landscape has changed dramatically. The big retailers have shown an aggressive global expansion strategy, not only by store openings but also through mergers and acquisitions, enabling diversification into various store formats. Examples of firms, which were leading in this expansion strategy, are the UK based Tesco, the French based Carrefour and Auchan, the German based Metro and Aldi, the Dutch based Ahold, and the Hong Kong based AS Watson.
1.2 Objective of the research
The size of the bidder firm, the size of the deal, the announcement of a takeover and the stock prices of both the market and the individual firm brings forth an interesting subject for research: Does size have an impact on the returns for bidder firm’s shareholders in a takeover situation? And does this effect show in the European Retail Market? Or put it in its most straightforward form:
Does size matter in the European Retail Takeover Market?
The main objective in this paper is to find out whether size, namely bidder size and deal size, has an impact on takeover returns. In order to answer the research question the following sub-research questions are investigated:
1 What is bidder size and how can it be measured?
2 What is deal size and how can it be measured?
3 Does bidder size matter for bidder firms in a takeover?
4 Does deal size matter for bidder firms in a takeover?
Questions 1 and 2 will be answered on the basis of a literature review. Question 3 and 4 are answered by means of an event study. This combined will give an answer to the main research question. Moreover, a thorough robustness analysis of the outcomes is presented.
1.3 Relevance of research
First of all, this research may be valuable to investors and shareholders who need to make takeover decisions, by explaining the impact of bidder size and deal size around the announcement of a takeover. Furthermore, the outcomes and robustness analysis may add to the academic knowledge on mergers and acquisitions.
depends completely on the event study methodology depicted by Brown & Warner (1980, 1985) and displays all outcomes and complemented calculations according to this study. Finally, this paper can be helpful for students and other researchers. The literature review provides an overview of research and knowledge on the takeover market, retail sector, and the impact of size in takeovers. Moreover, the methodology is written comprehensively for students and the extensive calculations will explain how to perform a study on cumulative average abnormal returns. The section on the data is also written in such extend that data sources, types of data and characteristics can be found fast and qualitatively.
1.4 Structure of the paper
2
Literature review
2.1 Takeovers and the Takeover market
2.1.1 Takeovers
There are three ways in which a takeover can take place: through a merger, a tender offer or proxy contest. With regard to both mergers and tender offers, the bidding firm will make an offer to buy the common stock of the target at the price that is higher than the market value at that time. The difference is that with a merger the offer is first negotiated with the target management and board, while in a tender offer the bidding firm will make an offer directly to the target firm’s shareholders (hostile takeover). A proxy contest is a rebel group (e.g. dissatisfied former managers) who attempts to get controlling seats on the target’s board (Jensen & Ruback, 1983).
Simply buying a share in a firm does not imply that a takeover has occurred. Shares of firms are being traded every (trading) day. Fama and Jensen (1983) argue that a takeover will be acknowledged when there has been a change in corporate control. They define this as “the rights to hire, fire and set the compensation of top-level managers”. In effect controlling majority of the company’s board and therefore the entire firm. This can be achieved by controlling a share of over 50% of share capital.
It is known for a fact that the world has seen several takeover waves. These waves are presented in Table 1 below. This table presents the time period, the name of the wave, and the specific characteristic (facet) of the takeovers.
Period Name Facet
1897 - 1904 First wave Horizontal mergers
1916 – 1929 Second wave Vertical mergers
1965 – 1969 Third wave Diversified conglomerate mergers
1981 – 1989 Fourth wave Congeneric mergers; hostile takeovers; corporate raiding
1992 – 2000 Fifth wave Cross-border mergers
2003 – 2008 Sixth wave Shareholder Activism; Private Equity; LBO
The characteristics like horizontal mergers, vertical mergers, hostile takeovers, corporate raiding, cross-border mergers, shareholder activism, private equity, and LBO are quite straightforward. However, conglomerate mergers and congeneric mergers need some explanation. The first one is a merger between firms that are involved in totally unrelated business activity. The second is a type of merger where two companies are in the same or related business but not offer the same products.
Rhodes-Kropf and Viswanathan (2004) state that in a sequence of two or more time periods the probability of a merger occurring is above the unconditional expected probability for a merger to occur, it is defined as a merger wave. They indicate this expected probability as the benchmark for a wave. Rhodes-Kropf and Viswanathan (2004) find that patterns of takeover activity and their profitability vary significantly across takeover waves. Despite such diversity, all waves still have some common factors: they are preceded by technological or industrial shocks occur in a positive economic and political environment, amidst rapid credit expansion and stock market booms (Rhodes-Kropf & Viswanathan, 2004). Figure 1 summarizes the EU M&A activity for the past several years.
Figure 1 EU M&A Activity (Source: European Commission calculations from Thomson Financial Service Data)
blue line represents their value from 1992 till 2006. This figure identifies several merger waves. At the end of the 1990s the European Union has seen a big wave of takeovers, with most of these deals being equity funded and often with bidder firms being in the same industry as targets. Several industries experienced this, including commercial banking, telecommunications, investment banking, hotels and casinos, oil and gas (Gorton, Kahl, & Rosen, 2005).
Several perspectives are offered by theory on mergers and acquisitions. Shleifer & Vishny (2003) elaborate on the Neoclassical Economic Theory that works with Efficient Market Hypothesis, stating that all players have only one goal, which is to maximize their own returns. Neoclassical theory sees mergers as an efficiency-improving response to various industry shocks, such as antitrust policy or deregulations (Mitchell & Mulherin, 1996). Jovanovic and Rousseau (2002) present the Q-theory. This perspective explains that companies only invest if the rate of return on the invested capital is higher than the firm’s weighted average cost of capital (WACC). If managers see these opportunities in the market they will have to invest in order to create maximum shareholder value.
2.1.2 Efficient Market Hypothesis
A few decades ago, the efficient market hypothesis (EMH) (see Fama’s (1970) article, “Efficient Capital Markets”) was generally accepted by academic financial economists. According to the EMH securities markets are efficient in reflecting information about individual stocks and about the stock market as a whole. The EMH states that when information arises, it spreads very quickly and is incorporated into the prices of securities. Thus neither technical analysis (which is the study of past stock prices in an attempt to predict future prices) nor even fundamental analysis (which is the analysis of financial information such as company earnings and asset values to help investors select “undervalued” stocks) would enable an investor to achieve returns greater than those that could be obtained by holding a randomly selected portfolio of individual stocks, at least not with comparable risk (Fama, 1970; Malkiel, 2003).
adjusted returns. Malkiel (2003) defines efficient capital markets as markets that do not allow investors to earn above-average returns without accepting above-average risks. Malkiel
(2003) illustrates it by citing a well-known story1.
Still, there is evidence that that the efficiency issues is never entirely resolved (Fama, 1991). Although prices on average adjust quickly to firm-specific information, a common finding in event studies is that the dispersion of returns (measured across firms, in event time) increases around information events.
2.1.3 Modern Portfolio Theory
Modern portfolio theory (MPT) attempts to maximize expected portfolio returns for a given amount of risk by carefully choosing combinations of assets (Markowitz, 1952, 1959). For example, Fabozzi, Gupta, & Markowitz (2002) present this particular MTP investment process in their paper, which is also displayed in Figure 2. They state that the theory dictates that given estimates of the returns, volatilities, and correlations of a set of investments and constraints on investment choices, it is possible to perform an optimization that results in the risk/return or mean-variance efficient frontier. Figure 2 shows these inputs (expected return model, volatility & correlation estimates, and constraints on portfolio choice) together are used in the portfolio optimization. Subsequently, the portfolio optimization results in the risk-return efficient frontier. This frontier is a line on which portfolios lie that makes up the set of efficient portfolios (Fabozzi, Gupta & Markowtz, 2002). Finally, depending on the particular objectives the investor has a choice for a particular portfolio of assets.
1 It is a story that tells of a finance professor and a student who come across a $100 bill lying on the ground. As the student stops to pick it up, the professor says, “Don’t bother- if it were really a $100 bill, it wouldn’t be there.” The story well illustrates what financial economists usually mean when they say markets are efficient. Markets can be efficient if, they sometimes make errors in valuation, or if many market participants are quite irrational, or if stock prices exhibit greater volatility than can apparently by explained by fundamentals. In short, these efficient capital market academics believe that $100 bills are not lying around for the taking, either by the professional or the amateur investors.
Figure 2 MTP Investment Process (Source: Fabozzi, Gupta, & Markowitz, 2002)
Portfolio theory is based on the premise of passive management, that cash flows can be combined but not altered, corporate diversification theory assumes that managers can actively intervene to lower corporate risk in a manner not available to shareholders (Lubatkin & Chatterjee, 1994). Proponents of corporate diversification theory make opposite predictions proposing that unsystematic and systematic risk are both best minimized by “putting all of one’s eggs in similar baskets” – by bringing together synergistically interrelated business units so that each business influences the other (Bettis & Hall, 1982; Lubatkin & O’Neill, 1987; Chatterjee & Lubatkin, 1990). They recognize that it was originally intended for securities managers’ use when assessing the risk characteristics of a portfolio of stocks, not for corporate managers. When corporations diversify, management’s actions can influence the underlying risk profiles of the combining businesses, and thus the expected variance of the combined returns need not be linear extension of historical variances (Lubatkin & Chatterjee, 1994).
2.1.4 Profitability & takeovers
2002). Researchers have been unable to successfully explain much of this variation, partially because the announcement of a takeover reveals information about numerous things.
A number of studies address various issues arising in M&A transactions in general. Earlier analysis in the UK (Franks & Harris, 1989; Firth, 1979) focused on profitability of the mergers and acquisitions irrespective of sector affiliation. These studies found that on average there have been no gains associated with takeovers; on the contrary, a very small loss of the shareholders of the acquiring companies was observed (224 UK companies were examined). Furthermore, this “no gain – small loss” position had been maintained 24 months after the announcement of the bid. The results were in conflict with the findings of some other similar studies in the USA (Halpern, 1983), which found significant gains arising from the consolidating transactions (the study sample included 120 North American companies). On the profitability of takeovers Jensen and Ruback (1983) conclude “corporate takeovers generate positive returns, that target firm’s shareholders benefit, and that bidding firm’s shareholder do not lose”. Target firms exhibit significant abnormal stock price gain of 20% in mergers and 30% in tender offers. Bidding firms have a statistically significant gain of 4% in tender offers and zero in mergers (Jensen & Ruback, 1983).
Moreover, recent work has been done on the topic of profitability around takeovers. In the article of Bruner (2002) the returns around 130 studies from 1971 to 2001 are investigated. Bruner (2002) finds that target firms earn sizable positive returns, bidders earn zero returns and bidder and targets combined earn positive adjusted market-returns, where positive returns exceed the required rate of return of shareholders. Bruner (2002) concludes that takeovers do pay, but presents a note of caution w.r.t. broad dispersion of findings around a zero return to bidder firm’s shareholders.
2.2 The Retail Sector & Perspectives on Size 2.2.1. The Retail Sector
and non-food. Food implies groceries and perishables, whereas non-food products are for instance appliances, electronics, clothing, etc.
In the last decennia the retail landscape has changed dramatically. The big retailers have shown an aggressive global expansion strategy, not only by store openings but also through mergers and acquisitions, enabling diversification into various store formats. A generally accepted taxonomy for store formats is: Hypermarkets, Supermarkets, Discount stores, Cash & Carries, C-Stores and Drugstores.
Examples of firms, which were leading in this expansion strategy, are the UK based Tesco, the French based Carrefour and Auchan, the German based Metro and Aldi, the Dutch based Ahold, and the Hong Kong based AS Watson. The key drivers for this expansion strategy were extrinsic factors as global economic growth and consumer spending, opening of new markets (Central and Eastern Europe, Asia), the introduction of the Euro and the rapid growth of discounters like Aldi and Lidl. Discounters have now a market share growing towards 18% in the near future. Their competitive advantages are: focused assortment, standardized outlets, efficient replenishment, a consistent no-frills approach and smart innovations and
investments.2
This has all led to concentration: the top-5 Retailers (Carrefour, Metro, Ahold, Tesco, Rewe) counted for 55% market share in Europe in 2008, whilst globally the top 15 retailers had a
market share of more than 30%3. Also intrinsic factors played an important role in driving this
transformation. Due to economies of scale, retailers discovered their bargaining power towards suppliers, which fueled their business models. They started materializing their purchasing power by demanding higher discounts, over-riders and extending payment terms, which created working capital surplus, as means to finance further takeovers and store openings. Besides, economies of scale created opportunities for cost synergies, supply chain efficiencies and growth of private label, which has grown over 30% of grocery sales in
Europe1. By internally harmonizing their IT systems retailers and suppliers started the process
of Global Data Synchronization1, which also enabled to reduce mutual inefficiencies by
collaboration at operational level. Figure 3 presents this change in the retail sector by showing
2 “What traditional retailers can learn from the discounters”, A.T. Kearney Online Paper, July 2010. 3 M&M PlanetRetail Ltd Presentation.
the increase in actual and forecasted discount sales (bars) in EUR millions, the increase in sales space (red line) in thousand m2 and the discount share (small blocks) in percentages over several years.
Figure 3 Development of discount format in EU-27 (Source: Planet Retail, A.T. Kearney analysis)
Strengthening the evidence on the necessity of takeovers, it is seen that firms in the retail sector are also trying to obtain positive return and create value through takeovers. Acquisition activity is one well-established growth mechanism in the retail sector (Burt & Limmack, 2001) next to investing into store–openings and growth of market share by extensive marketing. Previous research and literature depict three main perspectives on the reasoning of retail firms for doing a takeover.
First, follow the shopper. Takeovers have played an important role in the restructuring of the retail sector over the past two decades and appear likely to do so in the future. The debate about this impact of restructuring has particularly focused on the impact on competition, customers, and employees (Burt & Limmack, 2001). Retailers notice that there are no average families, that the customer has more choices but less time, and new technologies emerge every day. Retail firms create a stronger focus on diversity, or the so-called ‘cross-shopping’ formats. An example is Tesco, which has focused on obtaining small drug stores or variety stores like SuperDrug and Wilkinsons.
the relationship between concentration and price, with some studies from the US claiming a significant positive relationship (Lamm, 1981), and others find a negative relationship (Newmark, 1990).
Thirdly, go where the growth is. Internationalization is a large motive for a retailer. Largely unnoticed in much of the literature on economic globalization, a small group of elite transnational retailers have rapidly expended their overseas store operations beyond the core markets of North America and Western Europe through sustained merger and acquisitions activity, enabling them to assume dominant market positions in many countries across East Asia, Eastern Europe, and Latin America (Coe & Hess, 2005).
Figure 4 Announced Mergers & Acquisitions: Retail, 1985-2011 (www.imaa-institute.org/statistics-mergers-acquisitions)
Much has been written about the reasons for and nature of the growth; consolidation transactions have also been extensively documented. The world largest retailer, Wal-Mart has become a sizable player in two European markets via acquisitions in the UK (ASDA) and Germany (withdrawal July 2006). The French national champion Carrefour responded by first acquiring Comptiors Modernes in 1998 and merging with Promodés in 1999. Although a flurry of consolidation activity has been anticipated ever since, the actual M&A trends were more subdued. With the exception of the Casino-Monoprix merger in October 2000, the leading European retailers have preferred to pause for a while, in the meantime expanding into emerging markets of Latin America and Asia. However, there remains expectation of much further consolidation to be done (Dragun & Howard, 2003).
2.2.2 Bidder Size
One particular case of firm-specific information that displays an increase of dispersion of return is size. The strongest effects investigators have found are the tendency over long periods of time for smaller-capitalization stocks to generate larger returns than those of large-company stocks. Since 1926, small-capitalization stocks in the United States have produced annual rates of return over 1 percentage point larger than large-capitalization stocks (Keim, 1983). Fama and French (1993) examined data from 1963 to 1990 and ranked all stocks into deciles according to size measured by total capitalization. Decile one contained the smallest 10 percent of all firms, while decile ten contained the largest firms. The result clearly showed that the deciles composed of portfolios of smaller stocks generate higher average monthly returns than deciles composed of larger stocks.
In Moeller et al. (2002) it is shown that small firms fare significantly better than large firms when making an acquisition announcement. Overall, the abnormal return associated with the acquisition announcement for small firms exceeds the abnormal returns associated with acquisition announcements for large firms by 2.24 percentage points. Moeller et al. (2002) gives several theories on the impact of size when making an acquisitions announcement, which will be explained below.
for takeovers by documenting the empirical relations between the market valuations of firms and a comprehensive set of takeover characteristics. By testing two alternative theories of takeovers, one based upon the stock market overvaluation, and the other based upon extensions of the Q-theory of investment, Dong et al. (2002) show that firms with higher valuations have worse announcement returns. This could be that highly valued acquirers communicate to the market that these high valuations are not warranted by fundamentals, perhaps because they are undertaking efforts to acquire less overvalued assets with more
overvalued equity (the overvaluation theory)4. Since large firms have higher Tobin’s Q and
lower BM ratio than small firms it could be that when they announce the acquisition, they might signal something about their true value to the market, especially when they use equity (Moeller, et al., 2002).
Second, Roll’s (1986) “hubris” theory is presented. The possibility that mergers may be driven by biases of the acquiring managers has long been investigated. Roll (1986) first introduced the “hubris theory” of corporate takeovers in the finance literature. He predicts that managers are overconfident and overpay. Malmendier and Tate (2003) take it a step further and analyse the impact of “hubris” on mergers and acquisitions. They construct a simple model of the merger decision for CEO’s who are overconfident w.r.t. their abilities. This model makes three clear predictions. First, overconfident managers are more likely to conduct mergers when they have access to sufficient internal financial sources. Second, overconfident managers are more likely to conduct “bad” mergers. Thirdly, the announcement effect will be lower for overconfident managers, on average, since they are more likely to make value-destroying bids. In Malmendier and Tate (2003) overconfidence is measured by the options a manager has left unexercised. In a standard event study methodology used in order to explore the market’s reaction to merger announcement, they show that outside investors react more negatively to the announcement of a bid by an overconfident CEO than by a rational CEO. Moeller et al. (2002) relates the hubris theory to firm size and state that there are good reasons that managers of large firms can be more prone to overconfidence. Such managers have had a lot of success in the past that resulted in a large and successful firm or they might have to overcome more obstacles to become CEOs than managers of small firms. In order to confirm the statement above Moeller et al. (2002) look at the premium paid by the bidder (Officer, 2003) and the probability of success (Schwert, 2000). Officer (2003) and Schwert (2000)
provide evidence that firm and deal characteristics affect the premium paid by the bidder. It could be that large firms pay more because they acquire targets or enter a deal that requires a large premium. If large firms pay more, presumably they should have more success in making acquisition (Moeller et al., 2002), which can relate to the hubris theory. The results show evidence that premiums and probability of success are consistent with the hubris being more of a factor for the managers of large firms.
The third hypothesis is about synergy gains (synergy theory). An acquisition that has synergy gain increases the combined value of the acquired and acquiring firms. Bradley, et al. (1988) show that on average, acquisitions have synergy gain for their sample of tender offer acquisitions in the 1970s and 1980s. In Shleifer and Vishny (2003) the following formula is displayed. If the two firms (1 and 2) merge, S denotes the short-run valuation of the combined
equity per unit of capital, so the market value of the two firms together is 𝑉 = 𝑆(𝐾!+ 𝐾!).
They indicate that 𝐾 are the capital stocks for firm 1 and 2 and call 𝑆 the “perceived synergy” of the merger. In effect 𝑆 is the variable that the market consensus holds about the benefits of the merger. For example, it could be high if the market favours diversification and the two firms come from different industries or 𝑆 can be high when well performing firm acquirers a poor firm (Lang et al., 1989), or it could be that 𝑆, meaning synergy gain, differs between small firms and large firms from acquisitions. Following the method of Bradley, et al. (1988), Moeller et al. (2002) measures the impact of the acquisitions announcement on the combined value of the acquiring and acquired. They show that the acquisition announcements of large firms are consistent with the existence of negative synergies. In other words, the firms resulting from the mergers are worth less than the firm individually. In contrast, announcements by small firms exhibit positive synergies when using the percentage abnormal return measure.
stockpile free cash flow rather than spending it immediately. Harford (1999) presents evidence using stock returns that shows that acquisitions by cash-rich firms are value decreasing. Consistent with the stock return evidence, mergers in which the bidder is cash-rich are followed by abnormal declines in operating performance. Overall, the evidence supports the free cash flow theory and states that empire-building managers and poor investment opportunities prefer to invest the firm’s excess cash flow rather than pay it to shareholders (Harford, 1999).
When making a relation with the size of the firm to excess cash, a puzzle emerges. One can easily assume that empire-building managers are of a large firm, as they have a higher incentive to expand their firms beyond the size that maximizes shareholder wealth. Growth increases managers’ power by increasing the resources under their control. In additions, changes in management compensation are positively related to growth. The tendency of firms to reward middle managers through promotion rather than year-to-year bonuses also creates an organizational bias toward growth to supply the new positions that such promotion-based reward systems require (Jensen, 1986). However, empirical evidence does not give a clear conclusion on this matter. Chudson (1945), Vogel and Maddala (1967) give a positive relation between size and cash ratios, and Opler, Pinkowitz, Stulz, & Williamson (1998) display a negative relation. Moreover, the results of Harford (1999) show an insignificant relation between size (natural log of the market value of assets) and the cash ratio. Table 3 (Appendix A) of the paper of Moeller, et al. (2002) small firms have a higher mean and median ratio of cash to total assets. Furthermore, the opposite of free cash flow is leverage. Subsequently, Maloney, McCormick and Mitchell (1993) show that firms with higher leverage make better acquisitions and small firms have higher leverage than large firms.
Backed by the above literature on efficient markets and the diverse set of theories considering bidder size, the following “null” hypothesis is stated:
𝑯𝟎𝟏= Bidder size has no impact on the bidder abnormal returns in the European Retail
2.2.3 Deal Size
In a takeover the bidder and target firm will have different ideas about the value of the company that is being sold. Deal-makers (M&A facilitating companies) will employ a variety of methods and tools to assess the value of the target company. Some of these methods and tools are: comparative ratios (P/E ratio), replacement cost, and discounted cash flow (DCF). Furthermore, bidder firms nearly always pay a substantial premium on the stock market value (target value) for the firms they buy. This premium is the difference between the actual costs of acquiring a target firm against the estimate made of its value before the acquisition.
The size of this premium has several causes who largely coincide with the theories stated in the previous section. Bargeron, et al. (2008) show that unlisted acquirers offer significantly lower premiums than their publicly-listed counterparts. Moreover, takeover premiums tend to increase with the degree of competition in the market for corporate control (Walkling & Edminister, 1985) and the level of takeover hostility (Schwert, 2000). Another reason for the size for this premium comes from the notion of synergy (synergy theory); a merger benefits shareholders when a bidder firms’ post merger share price increases by the value of the potential synergy. This means that a bidder firm will need to pay this premium if they want to takeover the target firm, regardless of what the pre-merger valuation (target value) results into. Savor and Lu (2009) show that premiums in cash-financed acquisitions are larger than those paid in share-for-share transactions, as target shareholders are to be compensated for the immediate tax implications of cash offers. Finally, Roll (1986) posits that “hubris-infected” acquirers pay higher premiums (hubris theory). Similarly, Malmendier and Tate (2008) argue that managerial over-optimism leads to overpayment and thus results in value-destroying acquisitions.
Existing literature attributes to the fact that large acquisitions destroy more value for shareholders of acquiring companies. BusinessWeek (2002) reports that 61% of merger deals
worth at least $500 million end up costing shareholders.5 Similarly, research by Boston
Consulting Group (2007) shows that acquisitions priced more than $1 billion destroy nearly
twice as much value relative to smaller transactions.6 In 2012 Accenture7 came with an article
“who says M&A doesn’t create value?” on their online journal of high-performance business
5 “Mergers: Why most big deals don’t pay off”, BusinessWeek, 14 October 2002.
6 “A brave new world of M&A: How to create value from mergers and acquisitions”, The Boston Consulting
Group, July 2007.
“Outlook”. In their study Herd & McManus (2012) highlight the case of influence of deal size in a takeover. They state that in fact the opposite of the myth “the bigger the deal, the bigger the return” is true. Smaller deals do better than larger ones. Herd & McManus (2012) present their results and findings in Figure 5. The horizontal axes represents the range of deal sizes in billions USD and the vertical axes represent the acquirer total return to shareholders versus industry index. A possible explanation is that smaller deals minimize a variety of risks, such as building executive and board alignment regarding the deal, or maintaining the secrecy that keeps other potential suitors in the dark. Smaller deals also typically make it easier to conduct effective due diligence and mobilize the resources required for governing complex merger integration efforts. Loderer and Martin (1990) argue that “megadeals” tend to reduce shareholders wealth more than acquisitions involving smaller targets because acquiring firms tend to pay too much for large firms. Moreover, Bayazitova, et al. (2010) presents evidence that the well-documented value loss for acquirers is confined to a subset of excessively large deals.
Figure 5 Acquirer return versus industry index, by size of the deal (Source: Accenture Outlook)
and Boone (2000) study the acquisition and divestiture activity from 1990 through 1999 of 1305 Value Line firms (sample of firms covered by the Value Line Investment Survey).There it is found that combined bidder and target returns are significantly related to the relative value of the acquisitions (target value/bidder value). The wealth effects are directly related to the size of the acquisitions and are consistent with a synergistic explanation for the transactions. Officer, et al. (2008) concludes from an OLS regression of relative size to a 3-day (one 3-day previous to the announcement date till one 3-day after) cumulative average abnormal return, that the coefficient of relative size is positive and significant. Furthermore, Masulis, et al. (2007) concludes that this coefficient is positive, but insignificant. However, Boone & Mulherin (2007) conclude that the coefficient is negative and significant. Likewise, the coefficient is negative in Travlos (1987).
Fuller, et al. (2002) gives some further interesting results. Bidder firms experience significant negative abnormal returns when buying public firms and significant positive abnormal returns when targets are private companies or subsidiaries. Additionally, the returns to acquisitions of subsidiaries are greater than the returns to acquiring private firms, because subsidiary targets are larger than private targets.
Backed by the above literature on efficient markets, the characteristics of the deal size and previous research on deal size, the following “null” hypothesis is stated. This hypothesis will be tested using the methodology, formulas, variables and data described in this paper.
𝑯𝟎𝟐= Deal size has no impact on bidder abnormal returns in the European Retail Market
3
Methodology and Data
This section provides an overview of the data that are used in this research, including the sources, definitions and uses. To thoroughly test the hypotheses stated the methodology is divided in two comprehensive steps. First, a description is made of the measurement of the returns to shareholders through the calculation of abnormal returns. Secondly, using the abnormal returns, these results will be split into subsets in order to investigate the hypothesis stated.
3.1 Event Study Methodology 3.1.1 Abnormal Return
The methods implemented in this research are proposed in Brown & Warner (1980, 1985) and MacKinlay (1997). Brown & Warner (1980, 1985) explain the properties of daily stock returns and how these particular characteristic of data affect event study methodology. MacKinlay (1997) is used for their reasoning and variables to perform an event study in Economics. Finally, considering the reasoning for the time window Dodd & Ruback (1978) and Jensen & Ruback (1983) is used.
This event study will use a maximum of 250 daily return observations for the period around its respective event, starting at day -239 and ending at day +10. Figure 6 below displays the indexation of the returns using 𝑡. It defines T as the event date, which in this paper is the
announcement date (T), 𝑡! to 𝑡! (days -10 through +10) represents the event window, and 𝑡!
to 𝑡! constitutes the estimation window.
(Definition) Estimation Window Event Window
(Days) -239 -10 0 +10
𝑡! 𝑡! T 𝑡!
Figure 6 Time line for the event study (Source: MacKinlay, 1997)
In this event study two different abnormal returns are begin calculated by subtracting the daily observed return during the time window from: EUROSTOXX 50 daily returns / market-risk adjusted return. Consequently the formula used to calculate the return per individual (𝑖) bidder firm during the time window is:
𝑅!,!= !!,!!!!,!!!
!!,!!! (1),
where the stock price of an individual bidder (𝑃!,!) is subtracted from and divided by the
stock price of an individual bidder the day before (𝑃!,!). The return on the EURO STOXX 50
index during the time window (𝑅!,!) is calculated accordingly:
𝑅!,!=!!,!!!!,!!!
!!,!!! (2),
and the market-risk adjusted return per individual bidder firm (𝑅!,!) during the time window
is as follows:
𝑅!,!= 𝑅!− 𝛽! (𝑅!,!− 𝑅!) (3),
where 𝑅! is the risk free rate and 𝛽! is the beta of an individual company that describes its
movement with the market return and in this event study is estimated according to the following formula: 𝛽! = !!!(!!!! !!)(!!!!!!) !!!! !! ! ! !! (4),
where 𝑅! is the mean individual return and 𝑅! is the mean market return.
Finally, the formula used to calculate the abnormal returns is:
𝐴𝑅!,!= 𝑅!,!− 𝑅!,! (5),
which is the market model and is calculated by subtracting the daily observed return (𝑅!,!)
from the EUROSTOXX 50 daily returns (𝑅!,!) during the time window, and:
which is the market-risk adjusted model and is calculated by subtracting the daily observed
return (𝑅!,!) from the market-risk adjusted EURO STOXX 50 daily returns (𝑅!,!) during the
time window.
3.1.2 Cumulative Average Abnormal Returns
Tests with one event observation are not useful so the abnormal return observations are aggregated for the event window and across observations of the event (MacKinlay, 1997). Consequently the formula used to calculate average abnormal returns (AAR) will be:
𝐴𝐴𝑅! = !! !!!!𝐴𝑅!,! (7),
where 𝐴𝑅!,! is the individual abnormal return during the time window (𝑡), summing over
events (𝑁).
The average abnormal returns can then be aggregated over the ‘event window’ using the same approach as was used to calculate the cumulative abnormal return for each security (MacKinlay, 1997). Consequently the formula to calculate the cumulative average abnormal returns will be:
𝐶𝐴𝐴𝑅 𝑡!, 𝑡! = !!!!! !𝐴𝐴𝑅! (8),
where 𝐴𝐴𝑅! is the average abnormal return during the ‘event’ window, which is 𝑡! to 𝑡! (days
-10 through +10).
3.1.3 Test statistic under the Null Hypothesis
The Null Hypothesis tested entails that either bidder or deal size has no impact on the abnormal return. The test statistic is used to either confirm or reject the Null hypothesis
statistically. It is the ratio of cumulative average abnormal return 𝐶𝐴𝐴𝑅! and its estimated
standard deviation (𝛿𝐶𝐴𝐴𝑅!). The formula for any given event day 𝑡 is:
𝜃!! = !""!!
with
𝛿 𝐶𝐴𝐴𝑅! = 𝛿 𝐴𝐴𝑅! ∗ 21 (10), (total days in the event window is 21) and
𝛿 𝐴𝐴𝑅! = !! !!! 𝐴𝐴𝑅!− 𝐴𝐴𝑅 !
!! !!"# /228 (11),
with
𝐴𝐴𝑅 = !!"! !! !! 𝐴𝐴𝑅!
!! !!"# . (12).
3.2 Variables & Data
In this section will provide an overview of the variables & data used in this research. It will explain the data and were the variables come from, how they are calculated and used.
Sample Characteristics
In order to present an industry specific result this sample contains only deals in the retail sector. Between 1988 and 2010, worldwide 40.788 mergers and acquisitions with a total value of $2.225 billion have been announced (Figure 4). The particular region of the deals in this paper is limited to Europe. The reason is that it will be able to benchmark the returns on equity against the Index (EURO STOXX 50) and their respective risk-free rates. Furthermore, the sample is specified to bidder firms who are publicly held and conduct a takeover of more than 50% of the target firm. The total number of deals in the retail sector after specification occurred between January 1998 and January 2012, is 158.
Announcement Date
announcement date occurs at random times prior to the effective date, using the latter as the event date makes it difficult to identify changes in security prices that are due to the takeover event itself (Jensen & Ruback, 1983).
Daily Share Prices
Daily share prices for the bidding firms in this sample have been derived from DataStream for the past 14 years, from 01-01-1998 till 01-01-2012. The share prices are EUR denominated and have been retrieved in an indexed form and corrected for dividend payments. For each security in this research there is a maximum of 250 daily return observations in the period around its respective event starting at day -239 and ending at day +10 relative to the event. For a security to be included in the sample it must have at least 30 daily returns in the entire 250 day period and no missing return data in the last 20 days (Brown and Warner, 1985).
Daily Share Return
During the time window starting at day -239 and ending at day +10 the individual company return on their shares is calculated. This return is simply calculated by taking the daily change in share value over the time window. Because of the selection procedure of the share price date mentioned earlier dividend payments have been taken out of account.
EURO STOXX 50 Return
The EUR values for the EURO STOXX 50 have been retrieved from DataStream for the past 14 years. It is generally accepted for benchmarking a stock’s performance. The EURO STOXX 50 will be used to calculate the ARs for the individual European firms.
Company Beta
In order to calculate market-risk adjusted return, the company beta is needed. For every individual company the beta is calculated from historical stock and market returns. By taking the slope of these two daily returns these individual betas are obtained.
Risk Free Rate
Abnormal Returns
The abnormal return is calculated as the stock price return during the period surrounding the announcement date less a benchmark of what investors required during that period. That benchmark can be derived from the market model or from economic models. In the application of the market model a broad based stock index is used for the market portfolio (e.g. EURO STOXX 50). Economic models can be cast as restrictions on the statistical models to provide more constrained normal return models (MacKinlay, 1997). Two common economic models that provide restrictions are the Capital Asset Pricing Model (CAPM) and
the Arbitrage Pricing Theory (APT)8.
The ‘event window’ used in many previous event studies is a period starting at the effective date of merger or acquisition (the date of final approval by target shareholders). Investors will value the deal at or shortly after the announcement and not before (Mandelker, 1974; Ellert, 1976, Langetieg, 1978). However, at present the accepted method to consider the event window is to take the period around the announcement date, because the expected price effects will take place on or before the first public announcement, taking early price movements (information leaked to the market or the anticipations of the market that a merger or acquisition will take place) into account (Dodd & Ruback, 1978).
Bidder Size
The particular size of the bidder firm will be presented by the market capitalization. This is the total value of shares outstanding. The data concerning market capitalization is retrieved from the WRDS Compustat database, daily, per firm. According to market capitalization, the sample is split up into three bidder size subsets: “small cap (< €2 billion), “midcap” (€2 billion < €10 billion) and “large cap” (> €10 billion).
Deal Size
The deal size as well as the announcement data, is retrieved from Merger Market, Thomson Reuters, Zephyr, company presentations, Annual Reports, Press releases and private information. The sample contains a wide variety of deals with the smallest deal size being EUR 5.92m and the largest being 16,027.48m (Carrefour’s acquisition of Promodes). Deal size is the amount paid for the acquired (target) firm at the announcement date. According to
deal size, the sample is split up into three deal size subsets: “small deal” (< €200 million), “medium deal” (€200 million < €1000 million) and “large deal” (> €1000 million).
Summary statistics Daily Share
Return EURO STOXX 50 Return Company Beta Risk Free Rate Bidder Size Deal Size
Mean 0.11% 0.01% 0.50 2.78% 7781.03 797.07 Standard Deviation 0.26% 1.58% 0.64 1.31% 17890.88 1958.82 Skewness 2.50 0.13 0.14 -‐0.33 6.72 5.21 Kurtosis 11.63 4.32 16.37 -‐0.90 62.22 32.68 Number 250 3652 158 3652 158 158
Table 2 Summary statistics
4
Results
In this section the results of the event study are presented. These results will tell whether the Null hypothesis is accepted or rejected. Both the “market” model and the “market-risk adjusted” model will be displayed accordingly.
The Cumulative Average Abnormal Returns (“CAAR”) graphs are displayed in Appendix B.
Moreover, the graphs displaying the test statistics (𝜃!) are displayed in this section and are
the key indicators for addressing the hypotheses stated.
4.1 Bidder size
The graphs of the 𝜃!-values are displayed in Figure 7a and 7b together with the 10%
significance. In both figures the horizontal axes indicates the days in the event period and the
vertical axes gives the range of 𝜃!-values.
Focussing on the announcement day (𝑇) the CAAR using the market model for the “small
cap” firm is 2.51%. Given the value of 𝜃! is 1.40 for the market model, the first ‘null’
hypothesis (𝐻0!) that “bidder size has no impact on the behaviour of abnormal returns in the
European Retail Market” is rejected with a 10% significance level. Figure 7a representing the
𝜃!!-values for the different bidder sizes and the 10% significance level confirms the rejection
of the null hypothesis. The situation is different for the “midcap” and “large cap” firms. The announcement day (𝑇) CAAR for both “midcap” and “large cap” in the market model is respectively 0.12% and 0.05%. However, for midcap and large cap bidder firms the
corresponding 𝜃!-values are 0.07 and 0.03, which implies that the abnormal returns are not
Figure 7a Plot of 𝜃!-values vertically of cumulative average abnormal return (CAAR) for different bidder sizes from event day -10 to event day 10. Horizontally, the fat line respresents the 10% significance level for the 𝜃!
-value.
The findings using the abnormal returns from the market-risk adjusted model are consistent with those from the market model. However, there can be some loss of precision using the market-risk adjusted model, as the CAAR for all the three different subsets the CAAR is slightly higher. For the “small cap” firms the CAAR increases from 2.51% to 3.00%, the “midcap” firms from 0.12% to 1.12% and with the “large cap” firms the CAAR increases from 0.05% to 0.41%. These increases are to be expected when considering the market-risk adjusted return as a benchmark, because these stocks tend to have an important market component whose variability is eliminated using the market model (McKinlay, 1997). Consistent with the market model above only the “small cap” firms CAAR show a 10%
significance of rejecting 𝐻0! with 𝜃! being 1.76. Again, the 𝜃!-value plots for all three
categories can be found in figure 7b with the 10% significance level included. -‐1.00 -‐0.50 0.00 0.50 1.00 1.50 2.00 -‐10 -‐8 -‐6 -‐4 -‐2 0 2 4 6 8 10
Market
Figure 7b Plot of 𝜃!-values vertically of cumulative abnormal return (CAAR) for different bidder sizes from
event day -10 to event day 10. Horizontally, the fat line respresents the 10% significance level for the 𝜃!-value.
The CAAR plots (Appendix B) show that to some extent the market gradually learns about the announcement that is coming. The CAAR of the “small cap” firms gradually shifts up in days -20 to -1. Both “midcap” and “large cap” firms show a less clear gradually learning curve towards the announcement date. In the days after the announcement the CAAR for “small cap” is relatively stable. However, the CAAR of “midcap” firms shows a slight increase in days zero through one and days four through to five, but these increases are statistically insignificant. In the case of CAAR of “large cap” firms it shows a slight increase from the days six through ten, but again this is statistically insignificant.
4.2 Deal Size
Focussing on the announcement day the CAAR using the market model is respectively 2.33%
for small deals. Given that the value of 𝜃! is 1.42 for the market model, the second ‘null’
hypothesis (𝐻0!) that “deal size has no impact on the behaviour of the abnormal returns in the
European Retail Market” is rejected with a 10% significance level. Figure 8a representing the
𝜃!-values for the different sizes and the 10% significance confirms the rejection of the null
hypothesis. The picture is different for the “medium deal” and “large deal” bidder firms. The announcement day (𝑇) CAAR for both “medium deal” and “large deal” in the market model is respectively 0.29% and 1.06%. However, for medium deal and large deal the corresponding
𝜃!-values are 0.13 and 0.62, which gives the statistical evidence that the abnormal returns on
the announcement date are not large enough in order to reject 𝐻0!.
-‐1.00 -‐0.50 0.00 0.50 1.00 1.50 2.00 2.50 -‐10 -‐8 -‐6 -‐4 -‐2 0 2 4 6 8 10
Market-‐risk Adjusted
Figure 8a Plot of 𝜃!-values vertically of cumulative average abnormal return (CAAR) for different deal sizes from event day -10 to event day 10. Horizontally, the fat line respresents the 10% significance level for the 𝜃!
-value.
The findings using the abnormal returns market-risk adjusted model are again consistent with those from the market model. For the “small deal” firms the CAAR increases from 2.33% to 2.71%, the “medium deal” firms from 0.29% to 0.98% and with the “large deal” firms the CAAR increases from 1.06% to 1.76%. Consistent with the market model above only the
CAAR of bidder firms who have a small deal show 10% significance of rejecting 𝐻0! with 𝜃!
being 1.75. Again, the 𝜃!-value plots for all three deal sizes can be found in figure 8b with the
10% significance included. -‐1.00 -‐0.50 0.00 0.50 1.00 1.50 2.00 -‐10 -‐8 -‐6 -‐4 -‐2 0 2 4 6 8 10
Market
Figure 8b Plots of 𝜃!-values vertically of cumulative average abnormal return (CAAR) for different deal sizes
from event day -10 to event day 10. Horizontally, the fat line respresents the 10% significance level for the 𝜃! -value.
However, in the days after the announcement the CAAR plots of deal size shows some interesting differences with the CAAR plots of bidder size (Appendix B). Again the “small deal” CAAR is relatively stable as identical to the case of small bidder size. Conversely, the “large deal” CAAR displays an increase from days one through two and remains relatively stable from days 2 through 10, all being significant at 10%. Subsequently, this shows that a
few days after the announcement date (𝑇) large deal also rejects 𝐻0!. Finally, the CAAR of
medium deal firms does not show any changes up or down towards the end of the event (𝑡!).
-‐1.00 -‐0.50 0.00 0.50 1.00 1.50 2.00 2.50 -‐10 -‐8 -‐6 -‐4 -‐2 0 2 4 6 8 10
Market-‐risk Adjusted
5
Robustness Analysis
5.1 Nonparametric tests
Clearly, the method, using the test statistic (9) under the Null hypothesis, is parametric. There are however alternative nonparametric approaches available. For example, tests such as the sign test and the rank sign test. These particular tests are discussed and displayed next.
The sign test, which is based on the sign of the abnormal return, requires that the abnormal returns are independent across securities and that the expected proportion of positive abnormal returns under the null hypothesis is 0.5 (MacKinlay, 1997). The basis of this particular test is that, under the Null hypothesis, it is equally probable that the abnormal return will be positive or negative. So, in the sign test, the Null hypothesis states that there is a
positive abnormal return associated with a given event. This is displayed as 𝐻!: 𝜌 ≤ 0.5 and
the alternative hypothesis is 𝐻!: 𝜌 > 0.5 . 𝜌 is the probability of the abnormal return
exceeding zero return (𝜌 = 𝑝𝑟 𝐴𝑅! ≥ 0.0 ).
In order to calculate the sign test statistic, the number of cases where the AR is positive, 𝑁!,
must be evaluated. Furthermore, the total number of cases, 𝑁, is needed, which in this research is the total amount of takeover event in the sample. Finally this results in the following formula for calculating the test statistic under normal distribution 𝑁(0,1):
𝜃! = !!!− 0.5 !.!! ~ 𝑁 0,1 (13).
The Figures in Appendix C display the results of the sign test. The results show a 5% or 10% significant result for bidder size and deal size, which resembles the outcomes displayed in Section 4. Thus, according to the sign test the outcomes are considered to be robust.
In this test 𝐾!,! denotes the rank of the abnormal return 𝐴𝑅!,! in bidder firm 𝑖’s time window of 250 abnormal returns:
𝐾!,!= 𝑟𝑎𝑛𝑘 𝐴𝑅!,!
𝑡 = −239, … , +10 (14) where 𝐴𝑅!,! ≥ 𝐴𝑅!,! implies 𝐾!,! ≥ 𝐾!,! and 250 ≥ 𝐾!,! ≥ 1.
The rank statistic substitutes 𝐾!,!− 125.5 for the abnormal return 𝐴𝑅!,! resulting in an
particular ‘event’ day test statistic:
𝜃! = !! !!!! 𝐾!,!− 125.5 𝛿 𝐾 (15),
where the average rank is one-half plus half the number of observed returns. This means that
with the 250 observed return in this research the average rank is 125.5 (!"#! + 0.5 = 125.5).
The standard deviation 𝛿 𝐾 is calculated using the entire 250-day sample period:
𝛿 𝐾 = !"#! !! ! 𝐾!,!− 125.5
!!!
! !!"
!! !!"# (16).
The ranking procedure transforms the distribution of bidder firm’s abnormal returns into a uniform distribution across the possible rank values regardless of any asymmetry in the original distribution. Under the Null hypothesis, an ‘event’ day abnormal return rank is a drawing from a uniform distribution (Corrado, 1989).
5.2 Sample Size Analysis
The sample size is an important feature of any empirical study in which the goal is to make inferences about a population from a sample. In practice, the sample size used in a study is based on the expense of data collection, and the need to have sufficient statistical power. Generally, large sample sizes lead to increased precision when estimating unknown parameters. However, it can also be stated that it is more likely that in smaller samples the variation is bigger.
Study Sample Size Study Sample size
Bradley, Desai & Kim (1988)
236 Jarrell & Poulsen (1989) 526
Franks, Harris & Titman (1991)
399 Kaplan & Weisbach (1992) 209
Schwert (1996) 666 Mulherin & Boone (2000) 376
Mulherin (2000) 202 DeLong (2001) 280
Dodd (1980) 126 Asquith, Bruner & Mullins
(1987)
343
Morck, Shleifer & Vishny (1990)
326 Healy, Palepu & Ruback (1992) 50
Mitchell & Stafford (2000) 366 Dodd & Ruback (1977) 172
Bradley (1980) 134 Asquith (1983) 285
Lang, Stulz & Walking (1989)
87 Firth (1980) 434
Table 3 Summary of sample size (no. of events) for different studies.
Table 3 above displays the different sample sizes used in different studies. Roughly most of these studies are cited in the literature review. The sample size ranges from 50 events (Healy, Palepu & Ruback, 1992) towards 666 events (Schwert, 1996). It can be stated that a large sample is needed to generate significant results in the case of an event study.
In this section the robustness of the sample is investigated. The analysis will be of 10 randomly selected samples of 50 takeover events. From every individual sample the
Next, the average of the 10 samples’ CAAR and 𝜃!values is taken. In Appendix E the Figures
display the plots of the 𝜃!-values of both the total sample and the random sample.
The results show that the 𝜃!-values of the total sample above the 10% significance level,
whereas the 𝜃!-values of the random sample are below the 10% significance level. Hence, it
can be concluded that the sample size of 158 events is sufficient for a precise estimation and to have sufficient statistical power. Remarkably, this robustness check also shows the “size effect”. A smaller sample contains far less companies of different sizes, and these sizes have a much lower effect on the behaviour of the abnormal returns. The mere presence of a population of small firms has a significant positive cumulative average abnormal return influence on the total sample. Previous research also shows hints of this possibility. Healy, et al. (1992) present a non-significant negative cumulative abnormal return of -2.3% when investigating 50 largest U.S. mergers during the period from 1979 till 1984, where Schwert (1996) shows a significant positive cumulative abnormal return of 26,3% investigating 666 mergers during the period 1975-1991.
5.3 Time Window analysis
The test statistic (9) is the ratio of the particular day calculated cumulative abnormal return to its estimated standard deviation. Curiously, this standard deviation is estimated from the ‘estimation’ window of abnormal returns. In this study this ‘estimation’ window is day -239 through -11. Table 4 below shows previous studies accompanied with their event windows. There is a large variation in event windows, which comparably also indicates a large variation in estimation windows.
Study Event
Window
Study Event
Window
Bradley, Desai & Kim (1988) (-5, 5) Jarrell & Poulsen (1989) (-20, 10)
Franks, Harris & Titman (1991) (-5, 5) Kaplan & Weisbach (1992) (-5, 5)
Schwert (1996) (-42, 126) Mulherin & Boone (2000) (-1, +1)
Mulherin (2000) (-1, 0) DeLong (2001) (-10, 1)
Dodd (1980) (-1, 0) Asquith, Bruner & Mullins
(1987)
(-1, 0)
