The long run performance of Mergers and Acquisitions:
Evidence from the S&P500 from 1996-2004 with Calendar Time models
Thesis: MSc. BA, Corporate Financial Management Student: L.R. Swart (0772941)
Professor: Prof. Dr. B.W. Lensink
Period: Feb – Aug 2005
Abstract
In this report we would like to answer the question whether mergers and acquisitions create value for the acquirers. To see which models are used best for answering this question, we review existing literature. We observe that long run calendar time models for performance measuring of stock returns of acquisitive firms are recently most used. We empirically test how stock returns of acquisitive S&P500 firms from 1996 until 2004 are performing compared to a Fama-French three-factor model and compared to a reference portfolio. We conclude to find no evidence of underperformance of acquisitive firms in the three years after the announcements of M&As and conclude that value is at least conserved in M&A activities. In some models we find outperformance of small acquisitive firms and non-acquisitive firms. In a first approach the outperformance of non-acquisitive firms is higher than of acquisitive firms. In a more thorough approach we find however no evidence of any underperformance of acquisitive firms. The exact cause of the outperformance is recommended for further research.
Index
1. INTRODUCTION ... 4
2. MERGERS AND ACQUISITIONS THEORY... 7
2.1INTRODUCTION... 7
2.2WHAT IS VALUE CREATION?... 7
2.2.1 Introduction ... 7
2.2.2 Value creation... 8
2.2.3 Value conserved... 8
2.2.4 Value destruction ... 8
2.3EMPIRICAL STUDIES OF SHORT-TERM PERFORMANCE... 9
2.3.1 Event studies ... 9
2.3.2 Accounting studies ... 9
2.3.3 Surveys with executives... 9
2.3.4 Clinical studies ... 10
2.4CONCLUSIONS OF EMPIRICAL RESULTS... 10
3. REVIEW OF MODELS FOR LONG-TERM PERFORMANCE ANALYSES... 12
3.1INTRODUCTION... 12
3.2EVENT STUDIES... 12
3.3SIZE CONTROLLED EVENT STUDIES... 12
3.4BOOK-TO-MARKET AND SIZE CONTROLLED EVENT STUDIES;FAMA AND FRENCH... 13
3.5BUY AND HOLD METHODOLOGY AND NEW STATISTICAL METHODS... 15
3.6CALENDAR TIME STUDIES... 16
3.6.1 Calendar time Fama-French three-factor model ... 17
3.6.2 Mean monthly calendar time abnormal return model ... 19
3.7CONCLUSIONS... 21
4. SAMPLE SELECTION AND PORTFOLIO CONSTRUCTIONS ... 22
4.1INTRODUCTION... 22
4.2DATA COLLECTION... 23
4.3FAMA-FRENCH THREE-FACTOR PORTFOLIOS CONSTRUCTION... 25
4.3.1 Acquisitive Portfolio construction ... 25
4.3.2 Mimicking Portfolios constructed by Fama and French ... 26
4.4MEAN MONTHLY CALENDAR TIME ABNORMAL RETURN REFERENCE PORTFOLIO CONSTRUCTION ... 27
5. RESULTS OF THE CALENDAR TIME FAMA-FRENCH THREE-FACTOR MODEL... 29
5.1INTRODUCTION... 29
5.2EQUALLY WEIGHTED PORTFOLIO WITH WEIGHTED LEASED SQUARES REGRESSION... 30
5.3VALUE-WEIGHTED PORTFOLIO WITH WEIGHTED LEASED SQUARES REGRESSION... 31
5.4EQUALLY WEIGHTED PORTFOLIO WITH OLS REGRESSION... 32
5.5VALUE-WEIGHTED PORTFOLIO WITH OLS REGRESSION... 33
5.6CONCLUSIONS... 34
6. REGRESSION RESULTS OF NON-ACQUISITIVE AND DIFFERENCE PORTFOLIOS.. 35
6.1INTRODUCTION... 35
6.2RESULTS OF REGRESSIONS OF THE NON-ACQUISITIVE PORTFOLIO... 35
6.3RESULTS OF REGRESSIONS OF THE DIFFERENCE PORTFOLIO... 37
7. RESULTS OF THE MEAN MONTHLY CALENDAR TIME ABNORMAL RETURNS
MODEL ... 39
7.1INTRODUCTION... 39
7.2RESULTS AND CONCLUSIONS... 39
8. CONCLUSIONS AND RECOMMENDATIONS ... 41
8.1CONCLUSIONS... 41
8.2RECOMMENDATIONS... 42
REFERENCES ... 43
APPENDIX A... 46
APPENDIX B... 47
1. Introduction
During the last decade, Merger and Acquisition (M&A) activity has grown enormously. The falling stock prices in the year 2000 of especially firms with a history of many takeovers such as Enron, Nortel and Worldcom fuelled the debate on whether these M&A activities are value creating or value destroying activities for acquirers.
Research on value creation of M&As for the acquirer is in general done by the examination of stock returns to shareholders, where positive returns are associated with value creation and negative returns with value destruction. This stock return research has for a long time been concentrated on short-term announcement studies (Agrawal and Jaffe, 2000). These announcement or event studies examine the behavior of stock returns in a predefined period around the announcements of a merger or acquisition of firms involved. In short-term studies this period is in the order of days and in long-term studies in the order of months or years. In these studies not the actual returns but abnormal returns are examined.
These are defined as the difference between the actual returns and expected returns. These expected returns can be either from an asset-pricing model or from a benchmark of firms with the same characteristics as the examined firms but without M&A activities. The average of all abnormal stock returns of firms with M&As before, during and after the announcement of the M&A activities in a sample are examined. In this way general information is obtained about how stock prices of acquisitive firms behave around the announcement of M&As. Due to the fact that the moment of announcement in these studies is the central position around which a period is defined, these studies are called event time studies. This is in contrast with calendar time studies where calendar months in a sample are the starting point. Each calendar month in a sample is scanned on firms that have announced an M&A activity within a last predefined period prior to that calendar month. These firms will be defined as acquisitive firms later. To obtain general information about the performance of these firms the abnormal returns of the acquisitive firms are averaged per calendar month. Thus, while in the extreme case, where all M&As occur in the same calendar month, results will not differ from another case where all M&As are randomly distributed in time in an event study approach. Results of the two cases will differ in calendar time studies, however. Short-term event studies are based on the assumption that stock prices of a firm are the result of all discounted cash flows of the firm in the future. In these models it is therefore assumed that all effects of future performance of the M&A are therefore incorporated in the stock returns today (Bruner, 2001).
However, since the beginning of the 1990s questions have arisen concerning whether all information about future cash flows of the M&A is really incorporated in a short time period surrounding the announcement. Therefore also research in the field of long-term performance of M&As came up.
These studies examine the behavior of stock returns typically 12, 36 or 60 months around the
announcements of M&As (Bruner, 2001), and thus are long-run event studies. Results from these long-run event studies surprisingly show underperformance of the acquirer in contrast with short-term studies where around zero returns are observed (for example: Agrawal, 1992 and Gregory, 1997 versus Bruner, 2001). This apparent underperformance deviates from the famous efficient market hypothesis that explains that all information is almost immediately incorporated in asset prices. It implies that there are no arbitrary possibilities, and thus there should not be any underperformance one can make advantage of. The possible underperformance of acquirers in the long run is also surprising because one would expect a decline in M&A activity when this would be the case. The opposite is observed however; there is a continuing amount of M&A activities.
Due to the development of new theories and methodologies, some researchers question whether the outcomes of the former long-run event studies are statistically controlled and unbiased (Barber and Lyon, 1997a and Kothari and Warner, 1997) or whether the control asset-pricing model is used within these models. Fama and French, 1992, 1993) show, for example, that in models for calculating expected returns book-to-market ratio, firm size and firm beta should be taken into account, where former studies did not. Another paper that criticizes long-term event studies is from Andrade, Mitchell and Stafford (2001). They show that M&A activity is clustered in time and in industry and thus returns from M&As are not randomly distributed in time. Because of the fact that long run event studies assume that returns from M&As are random, results from these studies will be biased and unreliable.
This will be explained in more detail in section three.
In this report we would like to answer the central question whether mergers and acquisitions are value- destructing activities in the long run for the acquirers. We will do this by examining the performance of stock returns of the acquirer. For this purpose we first have to examine which methodology is used best. We will review articles and studies on this subject through time and see that there is a constant interaction between new theoretical models and the results of empirical studies. We will follow the articles by Fama (1998) and Mitchell and Stafford (2000) that recommend using calendar time models for examining long-term stock performance.
We will empirically test whether M&As show any underperformance in a sample of S&P500 firms from January 1996 until December 2004. In general, the performance of M&As can be tested by comparing the returns of firms with M&A activity with the returns one would expect from an asset- pricing model or with the returns of a portfolio of firms with comparable characteristics but without M&A activities, a reference portfolio. The first method used in literature is the calendar time Fama- French three-factor model where expected returns are calculated with a Fama-French three-factor asset-pricing model. This asset-pricing model will be explained in detail in section 3.4. Another method is the mean monthly calendar time abnormal return method where the returns of firms with
M&A activities are compared with the returns of the reference portfolio. Both methods will be explained in section 3.6. We will use both calendar time methods with different techniques on our sample of S&P500 firms, which we believe is representable for all American acquisitions. Because of the fact that American companies have transactions with the highest values in the world and the time frame examined contains the most recent M&A wave, which consists of transactions with the highest values ever (Weston, Mitchell and Mulherin, 2003), we believe our sample is ideal in drawing conclusions about value creation of M&As.
This research is unique because this most recent M&A wave in the United States has not been investigated before with calendar time studies. Next to that, there have only been few calendar time studies until now, of which only one examined American stocks (see table 3.1). Further, we are the first that examine the M&A activities of a sample of S&P500 firms. Finally, there has not been a report that used different calendar time methods with diverse techniques within these models so extensively, such as the calendar time Fama-French three-factor model and the mean monthly calendar time abnormal returns model with value and equally weighted portfolios and ordinary and weighted leased squares regressions.
This report is constructed as follows. In section two we will review the literature on mergers and acquisitions in general. We will look at how value creation is defined and what the results are from research in the field of short-term event studies. In the third section we review the research of long- term performance of M&As. The development of new theories and therefore new methodologies will be discussed next to the results of empirical research in this field. We will also discuss where the debate on long-term stock performance of M&As is at this moment. Possible explanations why any underperformance may exist will be described hereafter. In the fourth section we will explain how the sample is created and data retrieved. After that, the methodology of portfolio constructions for the different models is described. In section five we explain the results of a calendar time Fama-French three-factor model on our sample. The articles from Mitchell and Stafford (2000) and André, Kooli, L'Her (2004) are followed for the construction of this model. A value-weighted and an equally weighted portfolio are modeled with ordinary and weighted leased squares regressions. The results are compared with other research and will be discussed. In the sixth section we will compare the returns from acquisitive firms directly to non-acquisitive firms in our own model as a first approach. The difference portfolio of returns from acquisitive firms and non-acquisitive will be regressed over a Fama-French three factor model. Another used variant of calendar time studies in literature will be used as well: the mean monthly calendar time abnormal returns model. We will follow articles from André, Kooli and L’Her (2004) and Lyon, Barber and Tsai (1999) for the construction of this model, and results will be explained in section seven. In section eight final conclusions are drawn and recommendations are made for further research on long-term performance of M&As.
2. Mergers and Acquisitions theory
2.1 Introduction
In the last decade interest in the theory of mergers and acquisitions has grown increasingly. This can be explained by the fact that since the middle of the nineties of the last century, there has been an enormous increase in merger and acquisition activity throughout the world. In the years 1999 and 2000 in the U.S. alone, there have been more than 9000 M&A transactions, with a total value of more than
$1 trillion (Andrade, Kooli and L’Her, 2004). This huge amount of money involved in mergers and acquisition transactions can have an enormous influence on possible value creation or value destruction for firms. Next to that, due to faster computers since the beginning of the 1990s, more and bigger empirical studies could be made. Although there has been a lot of research on the subject of value creation of mergers and acquisitions, there still remain important questions about whether M&A activities are value creating or value destructing for acquirers. For example, the studies on long-term performance of buyers seem to show underperformance (Gregory, 2004 and Agrawal and Jaffe, 2000), where studies on the short-term show a neutral performance (Bruner, 2001).
2.2 What is Value creation?
2.2.1 Introduction
A first question that can be asked is how performance and thus value creation is actually measured.
Theories on value creation are almost always based on returns to shareholders. This is the so-called shareholder approach. It is believed that when one focuses primarily on returns to shareholders, this will also be beneficial to other stakeholders. For example, Copeland, Koller and Murrin (2000) state that in countries where firms have a shareholder approach, economies are more prosperous and have better capital markets. Other stakeholders benefit from this. For example, employees in these countries do benefit from better job perspectives and higher loans. Salomon and Salomon (2004) argue that, although firms can have a broader stakeholder approach in general and thus also responsible for other stakeholders for the firm, a firm should in the first place realize positive returns to shareholders. In literature a distinction between the returns to shareholders to the acquirer, the target or the combination of the two is made. Roll (1986) stated that for answering whether M&As do create value one should look at the effect of the combination of the two.
There are basically three possible simple different answers to the question whether firms do create value with Merger and Acquisition transactions. The first category of theories claims that there is value creation for the combined firms due to synergy and economies of scale. The second category claims that value is conserved; there is no net gain or loss. The value created by the target is offset by the value destructed for the bidder in this category. Finally, a last group of theories describes that
M&As destroy value due to overconfident managers and managers acting only for their own benefit at a cost of shareholders of the acquirer (which is not compensated by any value creation for the target).
2.2.2 Value creation
Weston, Mitchell and Mulherin (2003) explain that value can be created due to economies of scale.
The total of fixed costs of a firm can be divided over a larger amount after an M&A transaction, decreasing the cost per unit. Manne (1965) and Alchian and Demsetz (1972) describe theories based on the disciplinary motives effects of M&As. In these theories takeovers can be used to replace bad functioning management of the target. This will improve the performance of the company, creating value. Bradley, Desai and Kim (1983) explain that M&As can create value due to synergies between companies involved in the transactions. For example, due to bigger internal markets more efficient financing can take place for the combined firm.
2.2.3 Value conserved
The second group describes the conservation of value. Roll (1986) describes that in a bidding process the bidder that is willing to pay the most is the one that will get the deal. This means that on average the winner will always pay the highest amount. The reason why the manager is willing to pay this amount is that he is overconfident of getting positive results. The possibility that this is a higher amount than the “actual” price is the largest of all offers, thus the possibility of overpayment is highest for the winner. This is what Roll (1986) calls the winner’s curse. Wealth will be transferred from the bidder to the target in that case, having a zero net effect when the combined firm is considered. This is known as Roll’s hubris theory.
2.2.4 Value destruction
Examples of theories on value destruction are the following. Jensen (1986) describes in his article that managers can have incentives to overinvest. That means that there are circumstances under which managers can finance projects with very low or even negative positive net present value. These circumstances are cases in which the firm has a high positive free cash flow, thus high internal financial means, and where firms have limited liability. Managers then have incentives to invest in risky projects, even in projects with negative expected cash flows. In case of positive outcomes managers profit, in case of negative outcomes they do not loose. But investments in negative present value projects will reduce the value of the firm in the end. In another model, Shleifer and Vishny (2003) claim that managers are likely to enhance value for themselves at a cost of shareholders, thus rather overpaying M&As to keep their positions than creating shareholder value. The combined returns in M&As are negative in these theories (Weston, Mitchell and Mulherin, 2003).
2.3 Empirical Studies of short-term performance
2.3.1 Event studies
According to Bruner (2001), most empirical studies on M&A can be divided into different groups. In short-term event studies, abnormal returns to shareholders in a time period (also called ‘window’) around the announcement day of the M&As are examined. The announcement here is the central time reference and set as t=0. The stock returns in a time window before (pre) or after (post) this date are examined. The window is for example defined as [-3, +3], where returns are examined three days before the announcement until three days after the announcement, independent of the calendar month of occurrence. Due to the small time windows, these studies are called short-term event (time) studies.
Due to the fact that windows in short-term studies are in general three days (Mitchell and Stafford, 2000) outcomes do not rely much on the model used, and differences between models will be small (Weston, Mitchell and Mulherin, 2003). Fama (1998) explains; “because daily expected returns are close to zero, the model for expected returns does not have a big effect on inferences about abnormal returns”. In long-term event studies a much larger window is used to investigate the long-term stock performance.
With event studies the direct stock reactions of announcements of takeovers can be examined. These studies are considered to be forward-looking on the assumption that share prices are simply the present value of expected future cash flows to shareholders (Bruner, 2001). This is based on the efficient markets hypothesis that claims that any new information tied to an event such as an acquisition announcement will be incorporated into market prices quickly and accurately. Therefore, the abnormal change of stock prices at the announcement days are claimed to be a direct measure of a firm’s value creation or destruction. To get a clear picture of the movements of these abnormal returns they are cumulated over the event period (or event windows) to form cumulative abnormal returns (CARs).
With statistical measures, normally a t-statistic, one can infer whether the CARs differ significantly from zero (Weston, Mitchell and Mulherin, 2003).
2.3.2 Accounting studies
Another category does not focus on shareholder returns but on accounting returns. These studies incorporate the research question whether acquirers have better accounting figures after an M&A than a comparable peer-group. These accounting figures are in most cases net income, return on asset or equity and earnings per share (Bruner, 2001).
2.3.3 Surveys with executives
Surveys with executives are a next survey group. This group of research is based on surveys that executives can fill out. The many different opinions of executives about whether they think M&A activities created value are examined (Bruner, 2001).
2.3.4 Clinical studies
Clinical studies are the last category of Bruner (2001). These studies examine one or only a few transactions in depth. Usually this is done by interviews with executives and other insiders of the deal.
Due to the fact that deep knowledge of the transaction is obtained, new insights about details of M&As can be retrieved. As these transactions are chosen such that they are representable of all M&As, general conclusions about value creation might be drawn (Bruner, 2001).
2.4 Conclusions of empirical results
Most research has been focusing on returns to shareholders of M&As in the last decade, and thus most empirical outcome has been generated in this group until this moment, and we will focus on that group as well. Bruner (2001) summarizes the findings of former studies and concludes that the average gain for targets is about 20-30 percent. The gains to buyers in the studies show a more mixed picture. On average however, buyers show abnormal returns around zero. This means that buyers earn the expected returns, and thus takeovers can be seen as zero net present investments for the bidder (Bruner, 2001). Roll (1986) was the first who claimed one should be careful interpreting empirical outcomes form mergers and acquisitions. One should take the combined effect of both into account when observing performance M&As. The sizes of the firms should be taken into account as well. Due to the fact that bidders are larger than targets in most cases, even small negative abnormal returns after a takeover for a bidder can result in a net zero gain for the combined firm despite the large gains for the (small) target. If these size differences are taken into account, the most general conclusion for combined firms made in research is that there is a net positive return. Bruner (2001) concludes that this suggests that M&A does pay the investors in the combined buyer and target firm, according to short-term announcement effects studies.
In empirical research the complete dataset of M&As is further analyzed. For example, Bruner (2001) shows that transactions paid with stock results seem to perform worse than when a transaction is paid with cash. A few studies conclude that stock-based deals are associated with significantly negative returns at deal announcements, whereas cash deals are zero or slightly positive. Examples of these studies are Loughran and Vijh (1997), Franks, Harris and Titman (1991) and Gregory (1997).
Loughran and Vijh (1997) created a theory of M&As based on the theory of Meyers and Majluf (1984). These latter authors state that when managers need new financing, new stock will be issued in cases where stock prices are overvalued. Loughran and Vijh (1997) apply this theory on M&As and describe that in cases where firms offer stock for a transaction, stock prices tend to be overvalued.
Returns after the announcement will therefore underperform, because the actual returns will be lower than the overvalued expected returns.
Rau and Vermaelen (1998) define a firm with a low book-to-market ratio (also defined as ‘growth stock’) as a glamorous firm. They claim that a firm that was a glamorous acquirer in the past is overvalued at the time of a new announcement of acquisition. This is what they call the performance extrapolation hypothesis. Shareholders expect a good deal and a high profitability from a glamorous acquirer. In the period after the acquisition new information about the true quality of the acquirer reaches the shareholder, now realizing that the expected profitability was too high. This will cause an underperformance in the post-acquisition period.
Bruner (2001) adds other findings, of which we mention only a few. There seems to be a difference in performance of M&As when one looks at the method of acquisition. Bruner (2001) states that tender offers seem to be more successful than other offers, such as proxy contests. In a proxy contest the acquirer tries to persuade to join forces and gather enough shareholder proxies to win a corporate vote for the M&A. The observation that tender offers seem to perform better might be explained by the fact that they are in most cases unfriendly and bypasses management of the target firm, in contrast with proxy contests. Due to the fact that target firms in tender offers seem to be underperformers with relatively low share prices and bad management, they will be a bargain for acquirers.
Bruner (2001) concludes that the success of an M&A activity is also dependent on how well the post- merger is dealt with. Strategy and skills of the management formed in the new company can be of major importance of the profitability after the M&A activity.
3. Review of models for long-term performance analyses
3.1 Introduction
In the last section we have seen that most research of the performance of mergers and acquisitions have concentrated on short-term event studies. These take into account only few days after an announcement of an M&A. Recently there has been more interest in long-term studies in the performance of takeovers. Typically periods of 12, 36, or 60 months after the M&A are taken into account (Agrawal and Jaffe, 2000). These studies question whether all effects of possible value creation are incorporated during announcements which short-term event studies imply (see section 2.3.1). In this section we will review the different models developed for calculating long run performance through time and summarize empirical results obtained from these models.
3.2 Event studies
Agrawal and Jaffe (2000) give a useful basic review on long-term performance of mergers and acquisitions until 2000. They argue that for a long period of time the study of the performance of M&As has been concentrated on short-term event studies. It has however already been noticed in the eighties of the last century that there seemed to be negative abnormal returns in the years after mergers and acquisitions. For example, in 1983 Jensen and Rubach (1983) already made a comment in their article that observed negative abnormal returns in the long run seem to violate the efficient market hypothesis. When there is any underperformance of M&As in the long run, one can say that the asset price returns at the announcement of the merger and acquisitions are overestimated and not efficient.
These possible deviations from the efficient market theory are also called anomalies.
Agrawal and Jaffe (2000) state that research on the long-run performance can be divided into two periods: a period before and the period after the article of Franks, Harris and Titman (1991). These authors did alter the way of measuring long-term performance. Next to that, the articles until then only paid attention to the long-term performance as a part of a bigger research. The articles that appeared before the article of Franks, Harris and Titman (1991) are all event studies using the CAPM (or a variant of this model) to calculate expected returns and thus can be seen as extended (long-term) announcement studies. In general, Agrawal and Jaffe (2000) conclude that there seems no strong evidence of significant negative abnormal returns from these papers. The results only imply the possibility of an anomaly. Due to the application of new methodologies Agrawal and Jaffe (2000) conclude that the methods from these studies are outdated.
3.3 Size controlled event studies
The paper of Franks, Harris and Titman (1991) focuses completely on the long-term effects, and can be seen as the first article that does so. Although they also use an event study, they use new
methodologies for calculating expected returns (see section 2.3.1). In their article, they use the returns of firms with the same characteristics, but without these events as benchmark to calculate expected returns. This latter portfolio is also defined as a reference portfolio. They construct four different benchmarks, but claim that an eight-portfolio is most important. This reference portfolio is made up of four portfolios based on firm size, three based on dividend yield and one based on past returns. This article is thus the first that controls for firm size in expected returns, next to the firm’s beta. The authors do not find any significant abnormal returns, and thus conclude that the efficient market hypothesis should not be rejected and that there are thus no mispricings at the announcement dates of mergers and acquisitions. Further, they conclude that any found abnormal returns from earlier research must be the result of errors in the benchmarks used.
Another paper that takes firm sizes into account in expected returns is the paper by Agrawal, Jaffe and Mandelker (1992). They calculate abnormal returns in the five years after acquisitions and tender offers, where they control for firm size and beta (see section 3.4). In contrast with the study of Franks, Harris and Titman (1991), they report to find a significant underperformance of –10% up to five years after the investigated mergers. They find that the difference observed must be due to the specific time period that Franks, Harris and Titman used. The authors claim to have found a significant anomaly that post-mergers underperform in the long-term.
Other studies that controlled for size are those of Loderer and Martin (1992) and Kennedy and Limmack (1996). Loderer and Martin (1992) find that abnormal returns are significantly negative in the period 1966-1969 but in later periods are insignificantly different from zero. Kennedy and Limmack (1996) find insignificant underperformance of mergers in the period 1980-1989 in the UK up to twelve months after the events.
3.4 Book-to-market and size controlled event studies; Fama and French
The papers that control for firm size are only partly in line with how expected returns should be calculated, according to the research by Fama and French (1992, 1993). According to them, expected returns should be controlled for book-to-market ratios, size, and firm beta, and not only firm size.
In 1990 Sharpe achieved the Nobel Price for economics for his work (together with Lintner) in the 1960s on a revolutionary asset-pricing model, the Capital Asset Pricing Model, or CAPM. Expected excess returns from this model are simply calculated by taking the excess market returns and multiplying them with a factor of the analyzed portfolio. This factor is called the beta factor. The beta of the portfolio is a measure of the risk of the portfolio and is measured as the covariance of returns of that portfolio with the returns of the market portfolio divided by the variance of the stock returns of the market portfolio. Thus risk is directly coupled to the volatility of the returns according to CAPM.
Fama and French (1992, 1993) studied the returns of a very long history of American stocks. They had to conclude from this research that the CAPM is hardly capable of explaining stock returns.
Surprisingly, they showed that a model with other factors in combination with beta is much better in explaining the stock returns. Especially the factors book-to-market ratio and market capitalization (or size) together with the beta factor explain stock returns much better; they show much higher R2- statistics. The theoretical backgrounds of why just these factors are so important in stock returns are still under debate. In theoretical perspective, a higher return should imply a higher risk. The risk factor of a very high volatile stock simply implies the large possibility of getting a very high or very low return, thus the uncertainty in stock return is very high.
Of the other factors size and book-to-market ratios, it is harder to explain why these are related to risk premiums. Fama and French (1998) believe that financial distress and the possibility to borrow money are important factors. The size of a firm can be important as a risk factor because small firms do not have the same benefits as large companies. That is because small firms have to pay more to borrow from capital markets than big firms. Therefore an investor may want to get a higher risk premium for these stocks, implying higher expected returns. Fama and French define firms with high book-to- market ratios as value stocks. This is the opposite of stocks of firms with low book-to-market ratios, which are defined as growth stocks. Value stocks have smaller market capitalizations with the same book values as growth stocks due to worse prospective. These are in general financially distressed firms that have bad financial performance, irregular earnings and/or poor management. Therefore these firms also have to pay more for capital and one can state that investors do want to have higher returns for these stocks to compensate for the extra risk bared. Due to the higher risk of high book-to- market and small firms, higher returns are expected for these firms. Because these risk factors are not directly related to the volatility of stock returns they are not incorporated in CAPM.
Anderson and Mandelker (1993) and Gregory (1997) do take book-to-market ratios into account to calculate abnormal returns compared to benchmark portfolio returns in their event studies. They form book-to-market and size deciles of control portfolios in the same way as Fama and French (1993) describe in their article. The returns of the appropriate size and book-to-market control portfolios are subtracted from the returns of the acquiring firms up to five years after the acquisitions. Anderson and Mandelker (1993) find significant negative abnormal returns over five years. This is in line with the abnormal returns found in Agrawal, Jaffe and Mandelker (1992), but now with taking book-to-market into account as well.
Gregory (1997) researches the stock returns in the UK market. He calculates abnormal returns with six different models in event time, including a CAPM and a Fama-French three-factor model. Gregory (1997) reports a significant underperformance of –12% to -18% of five-year post-merger returns with
the Fama-French three-factor benchmark. Research with these size and book-to-market controlled event studies thus seems to have evidence of post-merger underperformance.
There has been growing critique however about the conclusions from these event study models.
Barber and Lyon (1997a) and Kothari and Warner (1997) state that measurement methods in the long run abnormal return event studies are biased and not reliable. They explain that there are three major reasons for this. First, the reference portfolios used as a benchmark contain only firms that will survive during the period analyzed, therefore these reference portfolios are biased. They call this the problem of survivor bias. Second, in event studies the portfolio analyzed is not rebalanced, where the reference portfolio, in most cases a market index, is monthly rebalanced. This is what they call a rebalancing bias. Finally, they show that the distribution of long-run abnormal stock returns is positively skewed.
A t-test, which assumes a normal distribution of the stock returns, will therefore give a misspecification of the statistics. Lyon, Barber and Tsai (1999) conclude that the survivor bias in the sample will cause in general a positive bias, where the rebalancing and skewness bias give a negative bias. To overcome these bias problems Lyon, Barber and Tsai (1999) recommend to use either an approach based on the traditional event study, but with carefully constructed reference portfolios together with taking other t-tests into account or to use an approach in calendar time instead of event time.
3.5 Buy and Hold methodology and new statistical methods
The first method Lyon, Barber and Tsai (1999) recommend is an event study where buy-and-hold abnormal returns are calculated. In the buy-and-hold methodology the gain from a virtual investment in a portfolio of firms with M&As are compared to the gains from an investment of a reference portfolio. This difference is in this model defined as abnormal returns. The gains are calculated by a virtual investment in the portfolio at the starting date, keeping them for a predefined period and selling them at the end date. As there is no acquisitive trading considered this is also called a passive buy and hold strategy. The Buy and Hold methodology is also an event time study because the time frames considered are defined starting from the announcement date, and are thus independent of calendar month. According to Lyon, Barber and Tsai (1999) an advantage of this methodology is that actual investment opportunities for investors are analyzed. The returns on an investment in a portfolio under test can be directly compared with the returns of an investment in a clean (without events) reference portfolio. In this way investor experience is directly examined, according to the authors. In the new approach the reference portfolios are carefully cleaned from survivor and rebalancing bias. This is done by keeping an unadjusted (thus not refreshed) reference portfolio through the whole time period and by excluding newly listed firms in the reference portfolio. To get rid of the skewness bias, two different methods can be used in random samples where distributions are positively skewed; a bootstrapped version of a skewness-adjusted t-test and empirical p-values calculated from a simulated
distribution of mean abnormal returns of pseudo-portfolios (for details see Lyon, Barber and Tsai, 1999). The first method is well established in literature on t-tests of positively skewed distributions.
The second method is described by Ikenberry, Lakonishok and Vermaelen (1995). Here the statistical significance of the mean of the skewed distribution is compared to an empirically generated distribution. These methods seem to work very well in random distributions (Kothari and Warner 1997). Two important examples of empirical studies that use these methods for research of long-term performance of M&As are the following.
Loughran and Vijh (1997) calculate the five-year post-performance of acquisitions. They calculate the difference between the returns from a portfolio of stocks bought at the announcement of the M&A and sold after a certain period, and a control (or reference) portfolio of firms without these activities but of similar size and book-to-market values. They also argue that this buy-and-hold method is recommendable as investor experience is directly visible. Results from their study are a significant underperformance for acquisitions and a significant positive abnormal return for tender offers after five years.
Rau and Vermaelen (1998) calculated the performance of M&As from 1980-1991 in an event time study. They calculated reference portfolios that are controlled for size and book-to-market ratios values into account in event time. The method is performed under a bootstrap version of an adjusted t- test, as explained by Lyon, Barber and Tsai (1997a). They report significant positive abnormal returns up to three years after tender offers and significant negative abnormal returns following mergers.
3.6 Calendar time studies
Lyon, Barber and Tsai (1999) state that the before mentioned bias problems with event studies are solved with the buy-and-hold method. However, according to Mitchell and Stafford (2000) two other problems still remain. The first is that when the sample is cross-sectionally dependent and therefore not random, the buy-and-hold method will give biased results. Brav and Gompers (1997) explain this as follows. The pervasiveness of underperformance may be misleading in event studies because the returns may be correlated in calendar time. If a shock to the economy in for example calendar year 1983 decreased all share prices of firms with M&As, then in an event study all years from 1979 to 1983 will show underperformance, although the underperformance only takes place in one single year.
Next to that investor sentiment is likely to be market wide and not only firm specific, causing the returns of events to be correlated. Andrade, Mitchell and Stafford (2001) show that M&A activity is clustered in time and also in industry and thus returns are cross-sectionally dependent.
Another problem is the problem of a poorly specified asset pricing method, explained by Fama (1998).
He explains that there does not exist a 100% reliable asset pricing theory that explains all the stock
behavior. This is what he calls a bad-model problem. Any found abnormal returns are therefore the result of a combination of the underlying model to calculate expected returns and the abnormal returns from an event. This is known as the joint-hypothesis problem. Mitchell and Stafford (2000) show for example that in the Fama-French model in three out of the twenty-five constructed portfolios by Fama and French (1993) significant abnormal returns can be seen, where in case of a perfect model zero returns are expected. These abnormal returns are in the order of 0.3% in monthly abnormal returns (Fama and French, 1992).
Brav and Gompers (1997) recommend using calendar time studies instead of event studies, as these will not have the problem of any cross-sectional dependence. Fama (1998, p 291) shows with a technical analysis that the bad model problem is less severe in calendar time studies than in event time studies and therefore recommends using calendar time models as well. Loughran and Ritter (2000) criticize the calendar time portfolio method because it averages over months with a lot of activity and months of little activity, making it hard for abnormal returns to be detected. Mitchell and Stafford (2000) find that this low power of explaining abnormal returns of the calendar time portfolio method finds no support in their empirical tests. On the contrary, they find more reliable outcomes for the calendar time approach where the dependency problem is taken care of.
3.6.1 Calendar time Fama-French three-factor model
In general, the most common method used to calculate performance in the calendar time is to regress excess returns per calendar month from a portfolio of firms that had an M&A activity in a preceding period on a Fama-French three factor model. Any significant values of the intercept of this calendar time Fama-French three-factor model are defined as abnormal returns. The dependent portfolio in the regression is a portfolio consisting of firms that had an M&A activity in a preceding defined period.
We will use a post-acquisition period of three years, because this is used in other recent research as well (André, Kooli, L’Her, 2004 and Gregory, Matatko, 2004). We will call this dependent portfolio the acquisitive portfolio, to distinguish from a portfolio with firms without an activity in the last three years, which is defined as non-acquisitive firms. To form the acquisitive portfolio, for each month in our time period under investigation (Jan 1996 until Dec 2004) the average of returns of the firms that announced an M&A activity with up to three years prior to this month is calculated. If a firm had an M&A more than three years ago it will not be in the acquisitive portfolio anymore. Next to that, new firms might have just announced a new M&A and need to be included in the portfolio. Therefore the portfolio needs to be refreshed every month. For example, in January 2000 the returns from the firm Halliburton will be included as it had an acquisition in January 1997 where American Express is not in the acquisitive portfolio, as it did not acquire other firms in the three years prior to January 2000. In February 2000 Halliburton is not in the acquisitive portfolio anymore as it is longer than three years before February 2000 that the acquisition took place, but now American Express is in the acquisitive
portfolio as it acquired another firm (with a transaction value of above $10m) in February 2000. The acquisitive portfolio can be made by equally weighting (EW) the returns of the firms in the acquisitive portfolio but also by value weighting the returns with respect to firm size (VW).
The calendar time Fama-French three-factor model is given by the following expression:
(1) Rp,t-RF,t = α + β0 HMLt + β1SMBt + β2 (Rm,t-RF,t) + error-term
Here, the dependent variable Rp,t – RF,t is the monthly excess return of the acquisitive portfolio over the risk free rate RF for a given calendar month t. The risk free rate is taken as the monthly rate of a 1- month American Treasury Bill, which is taken for Ibbotson and Associates Inc. The betas, β0, β1 and β2 are loadings of the portfolio on each risk factor. The intercept α is a measure of the average monthly abnormal return of the M&A sample.
The independent variables are the excess market return (Rm,t – RF,t) and two zero-investment portfolios HML and SMB. These are constructed to mimic the risk factors common to all securities, and are thus called mimicking portfolios. They are zero-investment portfolios as they can be seen as a long investment in one portfolio but at the same time a short position in another portfolio. HML is the difference in returns from a (size) value-weighted portfolio of high market-to-book factors and low market-to-book factors, for a given month. SMB is the difference in returns from a (size) value- weighted portfolio of small firms and a portfolio of big firms, for a given month. The construction of the SMB and HML portfolios of Fama and French is discussed in section 4.4 in detail.
In table 3.1 all results of known calendar time Fama-French three-factor models to calculate long run performance of M&As can be seen.
Mitchell and Stafford (2000) find small underperformance of M&As with their calendar time study in the United States for stocks from 1961 until 1993. With equally weighted portfolios they find a significant negative abnormal return. In a value-weighted portfolio they find insignificant underperformance. They conclude, with taking the bad model problem into account, that they find no evidence of underperformance of the acquirers.
Gregory and Matatko (2004) re-examined their earlier research of the long-run performance of takeovers in the UK with the new methodologies. They use a bootstrapped t-test for event studies but also a calendar time Fama-French three-factor model. With the latter method they report a significant negative average monthly abnormal return of –0.30% in five years of an equally weighted portfolio of acquiring firms.
André, Kooli and L’Her (2004) find for an equally weighted portfolio of Canadian acquirers a significant monthly average underperformance of –0.8% in the three years after the acquisitions.
When a value-weighted portfolio is used, they also find a significant negative abnormal return of – 0.8% and conclude that in general M&As in Canada are significantly underperforming in the long run.
Table 3.1 Results from calendar time Fama-French three-factor models for M&A long-term performance studies
3.6.2 Mean monthly calendar time abnormal return model
Another variation of the calendar time approach is the mean monthly calendar time abnormal returns model. We will use the same methodology as André, Kooli and L’Her (2004) and Lyon, Barber and Tsai (1999). In this method the performance of acquisitive firms is not compared to expected returns of an asset-pricing model but compared to reference portfolios. These reference portfolios consist of firms that did not have an M&A announced within the last three years prior to the calendar month considered, but of the same comparable size and book-to-market ratio.
For each calendar month the average of abnormal returns is calculated, which is the difference between the returns for each stock of an acquisitive firm and the return from a reference portfolio. The monthly calendar time abnormal return (CTAR) of a security i in calendar month t is given by:
(2) CTARi,t = Ri,t – Rcpi,t
Here Ri,t is the return of the acquisitive stock and Rcpi,t is the return of the reference portfolio. In each calendar month the average of all monthly CTARs is calculated, CTARt. This can be done with or without any weights:
Value- weighted Equally weighted
portfolio portfolio Stock regression
post- acquisition Intercept (%) Intercept(%) Market Time Period method period
(t-statistic) R2 (t-statistic) R2
Mitchell and Stafford (2000) -0.20 0.97 -0.03 0.95 U.S. Jul 1961- Dec 1993 OLS 60months
(-3.7) (-0.48)
André, Kooli, L'Her (2004) -0.75 0.62 -0.84 0.70 Canada Jan 1984-Dec 2000 WLS 36 months
(-2.77) (-3.34)
Gregory, Matatko (2004) -0.30 0.90 U.K. Jan 1977-Dec 1983 OLS 36 months
(-2.69)
(3)
∑
=
= N
i
i t i
t w CTAR
CTAR
1 ,
Where Nt is the number of firms in the calendar month t and wi,t = 1/Nt in case of equally weighted
abnormal returns and wi,t = sizei,t-1 /
∑
= −
N
i t
sizei 1
1
, where the sum is taken over all firm sizes in that calendar month for (size) value-weighted abnormal returns. Size values at the end of year t-1 are taken for this.
Now, a grand mean monthly calendar time abnormal return (MCTAR) can be defined as the average of the monthly average abnormal returns:
(4) MCTAR = (1/T)
∑
= T
t
CTARt 1
Here T is the amount of months taken into consideration and the sum is taken over all calendar months. To test whether the distribution of the grand mean abnormal return is significantly different from zero the variance (σ) of the mean monthly abnormal returns is used in the following t-test:
(5) t(MCTAR) = MCTAR / ( σ (CTARt) / T )
There are only two papers known with reported results where long run performance of M&As are examined with this model. The first is by Mitchell and Stafford (2000) that also researched the U.S.
stocks with a calendar time Fama-French three-factor model. They find insignificant underperformance of acquirers in a 5 years period after the M&A activity. Conn, Cosh, Guest and Hughes (2005) examined the performance of M&As in the U.K. and also find insignificant underperformance (with a significance level of 5%, two sided) with this model, which can be seen in table 3.2.
Equally weighted portfolio
Value-weighted
portfolio Stock post-acquisition
Intercept(%) Intercept(%) Market Time Period period (t-statistic) (t-statistic)
Mitchell and Stafford (2000) -0.04 -0.03 U.S. Jul 1961- Dec 1993 60 months (0.78) (-0.58)
Conn, Cosh, Guest and Hughes (2005) -0.21 U.K. Jan 1984-Dec 1998 36 months (-1.58)
Table 3.2 Results of research on the mean monthly calendar time abnormal return method
3.7 Conclusions
New theoretical frameworks and empirical studies have alternated since the last decade and triggered each other. Therefore comparisons between empirical results from different studies should be kept in the right perspective and time frame. Although not all researchers agree which model is best to use, the calendar time studies are nowadays always included in the research. Positive arguments for the calendar time method are that there is no cross-sectional dependence and the bad model problem is less severe in calendar time studies than in event time studies.
Results from calendar time studies show mixed results. With the calendar time Fama-French three- factor model Gregory and Matatko (2004) and André, Kooli, L’Her (2004) find significant underperformance in both equally and value-weighted acquisitive portfolios in the UK and Canadian stock markets. Mitchell and Stafford (2000) conclude to find no evidence in the US stock markets of underperformance when the bad model problem is taken into account. With the mean monthly calendar time abnormal return model, Conn, Cosh, Guest and Hughes (2005) and Mitchell and Stafford (2000) find insignificant monthly abnormal returns.
The debate on which model to use and conclusions from empirical results is still ongoing. We would like to contribute to this discussion with an empirical study of the most recent M&As in the United States, which has not been done before, and will examine M&As in the S&P500 with different calendar time studies.
4. Sample selection and portfolio constructions
4.1 Introduction
We choose to research the long-term performance of M&As of S&P500 firms for several reasons.
The country where the most M&A activity recently occurred, representing the highest value in the world is the United States. Any value creation due to M&As is therefore likely to be best reflected in this stock market. The amount of the worldwide M&A activity in the last years is shown in table 4.1.
Worldwide M&A activity, 1998-2001 ($ billion) 1998 1999 2000 2001
World 2092 2373 3565 2063
U.S. 1343 1395 1786 1143
Rest of world 749 979 1779 920
Table 4.1. Worldwide M&A activity, 1998-2001
Source: Mergers and Acquisitions Magazine and Weston, Mitchell, Mulherin (2003)
The S&P500 index is representative for all American stock; any observed behavior in firms within the S&P500 is therefore representative for the behavior of all American stocks. The shares of these firms are liquidly traded; therefore stock prices of these firms are believed to represent the fundamental values (Weston, Mitchell and Mulherin, 2003). Due to the fact that the S&P500 firms contain firms in the highest market capitalization (size) categories and bigger firms have more acquisitions than small firms (Weston, Mitchell and Mulherin, 2003) one can expect to find highest value acquisitions in this index.
We choose for the time window of 1996-2004. The reason is that the amount of M&As in the beginning of the nineties was very low for S&P500 firms. The amount of M&As increased starting from 1996 and peaked in 2000. The activity level decreased after 2001 but stayed high. To observe the behavior of stocks in the bull market at the end of the nineties through the bull market since 2001 might give more reliable results than when only activities in a bull or a bear market are examined, we average out any dependency on the market in this case. Most importantly, there have not been any reports yet that document long run performance of M&As until 2004 with a calendar time model.
0 50 100 150 200 250 300
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 year
Number of M&A transactions
Graph 1. Number of M&A transactions per year of all American acquirers with deal size >$10m. Source: Zephyr database
4.2 Data collection
The Zephyr database is able to provide the useful data on mergers and acquisitions. Data of mergers and acquisitions can be found for all different kind of countries, deal sizes, deal type (merger or acquisition), method of payment (cash, stock or a combination), etc. Because we are interested in companies from the S&P500 we make a search for American companies that completed a merger or acquisition with a deal size larger than $10m to prohibit the sample to be biased with too many small valued activities. This number is arbitrary but used in comparable research such as Gregory, Matatko (2004) and André, Kooli, L’Her (2004).
For the Fama-French three-factor model we need to obtain data about the following financial and accounting variables of the S&P500 firms: monthly stock returns, book-to-market ratios, market capitalization and the monthly risk-free rate. The latter is obtained via Kenneth French’s homepage of a 1-month Treasury Bill, which is from Ibbotson and Associates, Inc. (http://mba.tuck.dartmouth.
edu/pages/faculty/ken.french/). The stock returns and financial data are obtained via Thompson’s DataStream database.
As a first step we downloaded the end of the year book-to-market and market capitalization accounting data and end of the month stock return financial data from DataStream from the first of January 1996 until the end of December 2004. One needs to be careful to download the right data from DataStream. Dividend payouts or stock repurchases will change stock prices “artificially” and we need to control for this. Because we only want information on capital gains without these effects DataStream has the option to download returns of a data type that is already controlled for these distortions, which is called “Total Returns” in DataStream.
The obtained return data consist of five hundred Excel files of end of the month return data and another five hundred Excel files with all kinds of accounting data of the S&P500 firms for our time period. Because DataStream does not give book-to-market and market capitalization directly we calculate these ourselves. The book-to-market ratio is calculated by taking the inverse of the given market-to-book ratio of DataStream. The market capitalization is taken as the difference between the total enterprise value and the total debt value available from DataStream. In the program language Visual Basic for Applications a macro is written to handle the copying of the useful information of the five hundred Excel files into just one file. This macro can be seen in Appendix A. For each calendar year we construct a ranking in size from small to large and a ranking in book-to-market ratio from low to high for later portfolio creation. To get an idea about these figures, the market capitalizations are ranged from $970m for Delta Airlines to the very large General Electric conglomerate, which had a value of $385 billion in 2004.
The downloaded file from Zephyr contains all American acquirers that have completed mergers and acquisitions with a deal size of $10m or greater, from January 1996 until December 2004. This Zephyr data-file contains 7321 mergers and acquisitions. After investigating the file (making sorts on date, deal size and firm names respectively and creating a Macro that identifies doubles) we noticed some double counting. When we controlled the file for this, there are 7119 M&As left. We are only interested in the S&P500 firms, so from this file we need to pull out the S&P500 firms. Unfortunately, Zephyr does not give the short name for the S&P500 firms. Therefore we need to identify the S&P500 in this list ourselves one by one due to the different naming of the firms in the different databases. The amounts of M&A activities per month for the S&P500 firms in our time frame 1996-2004 can be seen in the figure below. The firms in this time frame contain a total of 1097 activities.
0 5 10 15 20 25 30 35
jan-96 jan-97
jan-98 jan-99
jan-00 jan-01
jan-0 2
jan-0 3
jan-0 4
number of transactions
Graph 2. Number of M&A activities per month for S&P500 firms with deal sizes >
$10m. Source: Zephyr database
This graph just looks the same as the total amount of all American acquirers in the Zephyr database we saw before. We conclude that the M&A activity levels in the S&P500 behave in the same manner as all American stocks. From the downloaded data portfolios for the different models can be constructed.
4.3 Fama-French three-factor portfolios construction
4.3.1 Acquisitive Portfolio construction
Now we know for each S&P500 firm on which date they announced an M&A activity of larger than
$10m and thus are able to make the acquisitive portfolio from this. For each month in the time frame between January 1996 and December 2004 we create an overview of which firms had an M&A activity within the last three years. For example, Alcoa Company announced an M&A activity in August 1999, and will therefore be in the acquisitive portfolio in August 1999 until July 2001. When a firm had an activity more than 36 months after the announcement it will be left out. Therefore acquisitive calendar time portfolios are refreshed monthly.
1999 2000
5 6 7 8 9 10 11 12 1 2 3 4 5 6 7
3M
ABBOTT LABS
ACE
ADC TELECOM. 1 1 1 1 1 1 1 1 1 1 1 1 1 1 ADOBE SYSTEMS ADVD.MICRO DEVC.
AES
AETNA 1 1 1 1 1 1 1 AFFILIATED CMP.SVS.'A' 1 1 1 1 1 1
AFLAC
AGILENT TECHS.
AIR PRDS.& CHEMS. 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 ALBERTO CULVER
ALBERTSONS
ALCOA 1 1 1 1 1 1 1 1 1 1 1 1 ALLEGHENY EN.
ALLEGHENY TECHS. 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
ALLERGAN
ALLIED WASTE INDS.
ALLSTATE
ALLTEL 1
Graph 3. Zoomed in part of overview of all calendar months between January 1996 until December 2004 that contain firms that had an M&A activity within the last three years.
To calculate the excess returns of the acquisitive portfolio we need to take the returns from the acquisitive firms and equally or value weight these and subtract the risk free rate. Because we have downloaded the monthly returns from all the S&P500 firms already we can now combine the acquisitive firms with their returns to construct the acquisitive portfolio. In the equally weighted portfolio returns of the acquisitive firms are equally weighted to form one monthly average return of the acquisitive portfolio. In the value-weighted portfolio the returns are value-weighted by the size of the firms. This is a bit more complex due to the fact that the sizes of the firms should be taken into
account. The sizes of the firms at the end of the former year (t-1) will be used as weighting factors of the returns in year t. Thus for the calendar month August 1999, firms sizes at the end of 1998 will be used as weighting factors. To construct excess returns of the portfolio, the risk free return is simply subtracted from the average monthly return of the equally or weighted acquisitive portfolio. In Appendix B the number of firms in the acquisitive portfolio in our time frame of 1996-2004 can be seen. Clear to see is that activities are concentrated in time as stated by Mitchell and Stafford (2000).
In the same manner excess returns of a non-acquisitive portfolio can be constructed, which contains the excess returns of all firms in our S&P500 sample that did not have any M&A activity within the last three years.
4.3.2 Mimicking Portfolios constructed by Fama and French
The mimicking portfolios of the Fama-French three-factor model need to contain as many American stocks as possible to get the best risk premiums of size, book-to-market and market excess returns.
Fortunately, we can make use of the data library of Kenneth French (http://mba.tuck.dartmouth.
edu/pages/faculty/ken.french) where the mimicking portfolios are already given. The data library contains the universe on stock return data for the United States of America, from 1927 until today. It contains all stocks from the NYSE, AMEX and NASDAQ.
Fama and French (1992, 1993) made the SMB and HML portfolios as follows. In June of each year t from 1963 to 1991, all NYSE stocks are ranked on size. The median size is used to split NYSE, Amex and Nasdaq stocks into two groups. They call these small (S) and big (B) groups. The NYSE, Amex and Nasdaq stocks are also broken up in three book-to-market groups based on the breakpoints for the bottom 30% (Low), middle 40%, (Medium) and top 30% (High) of the ranked book-to-market ratios for NYSE stocks. The book-to-market ratio is taken as the value at the end of December of calendar year t-1. Now six portfolios are constructed from the intersections of the two size and three book-to- market portfolios. For example, the B/H portfolio contains the big firms that are also in the highest book-to-market group. Monthly value-weighted returns of the six portfolios are calculated from July of year t to June of year t+1. The portfolios are reformed every June of year t+1. This is done to assure that the book-to-market ratios are known from the year before. The mimicking portfolio SMB is the difference per month between the average of the three portfolios with the small size group (thus SH, SM, SL) and the average of the three portfolios with the big group (BH, BM, BL). The mimicking portfolio HML is the difference of the average of returns between the two portfolios with high book- to-market ratios (SH, BH) and the portfolio with the low book-to-market ratios (SL, BL).
Kenneth French’s database also contains return data for a 1-month Treasury Bill that French obtained from Ibbotson and Associates, Inc. This is simply the end-of-month return data of a one-month Treasury Bill, and can be used as the risk free return RF. The excess market returns are also taken from
