Long-term post-M&A performance in US high-tech industries

(1)

1

Long-Term Post-M&A Performance in US

High-Tech Industries

Abstract

High-tech firms have great growth potential. Potential which can be obtained by M&A. But with this high growth potential comes high risk. Are acquirers of high-tech firms able to deploy this potential or does their performance lack on the long-term? High-tech acquiring firms are better able to manage their risk and hence have a greater ability at realizing potential synergies. To study this a long-term event study is conducted using the BHAR and calendar-time portfolio approach. Results show that M&A do not earn negative abnormal results as previous literature suggests. Further analysis shows that high-tech M&A significantly performs better than M&A in other industries during a three-year post-merger time horizon.

July 1st₂₀₁₇ Master Thesis MSc Finance

Duisenberg Honours Programme in Corporate Finance and Banking Student: Dion Langedijk

Student number: 10657843 Supervisor: Vladimir Vladimirov

(2)

2

Statement of Originality

This document is written by Dion Langedijk who declares to take full responsibility for the

contents of this document.

I declare that the text and the work presented in this document is original and that no

sources other than those mentioned in the text and its references have been used in

creating it.

The Faculty of Economics and Business is responsible solely for the supervision of

completion of the work, not for the contents.

(3)

3

1. Introduction

Do mergers and acquisitions create or destroy shareholder value? A question that has drawn a lot of attention of the corporate and academic world over the past decades. Managers should always act in the interest of their shareholders to create wealth. Most analyses focus on the short-term reactions of investors to the share price just before and after the merger announcement. The common result of studies on the short-term effect on wealth creation of M&A are positive. Acquiring shareholders tend to earn no or negative abnormal returns, while target shareholders earn significantly positive abnormal results leading to positive wealth creation overall (Loughran and Vijh, 1997). With the ever growing high-technology sector that currently compose more than half of the total GDP in wealthy economies, takeover activity in these industries is increasing (Kohers and Kohers, 2000). Since mid-1994, technology indexes have climbed twice as fast as the S&P 500. Targets from high-tech sectors may be able to provide greater shareholder wealth benefits to acquiring companies in comparison with lower growth target companies. Therefore these high-tech industries distinguish themselves from other industries. According to the study of Kohers and Kohers (2000) acquirers of high-tech targets earn significant positive abnormal results at the time of the merger announcement. The question is if these are unbiased forecasts for the future. Limited literature on long-term

performance of acquiring firms post-M&A generally present negative abnormal returns over a period of three to five years after the merger announcement date (Gregory, 1997; Anderson and

Mandelker, 1993; Asquith, 1983). These negative abnormal results are inconsistent with market efficiency and suggest that the changes in stock prices overestimate the future efficiency gains from mergers (Jensen and Ruback, 1983). However, Fama (1998) states that consistent with the market efficiency hypothesis that the anomalies are chance results, apparent overreaction of stock prices to information is about as common as under-reaction. And post-event continuation of pre-event abnormal returns is about as frequent as post-event reversal.

Bena and Li (2014) and Sevilir and Tian (2012) show that a firms innovative performance improves in the three years post-merger. Bloom and van Reenen (2002) and Nicholas (2008) provide evidence that that innovation has an economically and statistically significant effect on firm-level productivity and market value. Because for firms in high-tech industries innovation is a big part of their value creation, we should expect a positive relation between M&As in these industries and their post-merger firm performance. But with high-growth potential comes greater risk. Sorescu, Chandy and Prabhu (2007) state that high-tech firms are better at selecting targets with innovation potential and deploying its potential to gain competitive advantage. High-tech firms are better at evaluating the innovation of target firms. Due to their prior related knowledge, they are better at evaluating outside knowledge and handling the risk associated with high-tech targets resulting in abnormal

(5)

5

performances (Sorescu et al., 2007). The superiority in selection and deployment of external assets turns in greater ability in realizing potential synergies and will be transposed in long-term financial rewards to the acquiring firm. So on the long-term high-tech M&A should outperform M&A in other industries and would be interesting to research.

In this study a comprehensive sample of 24,332 US M&A events during the period of 1 January 1981 – 31 December 2006 is used. A sample size which has not reported in an M&A study before. In this paper the buy-and-hold abnormal returns approach is used which has been used before by Loughran and Vijh (1997) and Mitchell and Stafford (2000). The equal- and value-weighted mean BHARs will be computed for all acquiring firms and subsamples. Since previous studies have presented different results using different benchmarks, for robustness six different benchmarks will be used in this study. The BHARs are then used in regressions with control variables to check whether high-tech M&A outperforms other acquiring firms. Since methodological problems seem to be the main issue in long-term M&A studies, another approach is also conducted; the calendar-time portfolio approach (Fama, 1998; Mitchell and Stafford, 2000). Using this approach the monthly excess returns for the whole sample and subsamples will be regressed on the Fama and French three-factor model and four-factor model including the Carhart (1997) momentum factor. According to Fama (1998) this is the preferred methodology, due to its robustness to statistical errors. A long-short portfolio will be formed consisting of a long portfolio of high-tech acquiring firms and a long-short portfolio of non-high-tech acquiring firms to inspect the difference in performances.

The remained of the paper is organized as follows. In section two the literature relevant to long-term M&A and high-tech M&A is discussed. In the third section the hypotheses are developed. The research methodology is explained in section four. Section five describes the data being used. In section six the results are presented. And at last, section seven will present the conclusion and discussion.

2. Literature

This section provides a comprehensive description of the literature related to long term M&A performance. Thereafter this subject will be discussed regarding high-tech industries.

2.1 M&A Performances

According to economic theory managers should always act in the interest of their shareholders. In fact, managers have a fiduciary duty that requires them to act in the best interest on the behalf of their shareholders, which is most likely to make profit. It would be unethical to break this duty. So based on this theory corporate actions as mergers and acquisitions should be focused on maximizing

(6)

6

shareholder wealth. Benefits of mergers and acquisitions can come from different sources. According to efficiency theory M&A deals result in benefits acquired from synergies and economies of scale. Theories of market power suggest that M&A delivers advantages from enhanced market power, creating oligopolies and wiping out competition. Then there are agency theories which argue that M&A is used to solve agency problems by removing ineffective managers. M&A could also provide benefits by diversification, reducing its risk exposure when they have too much of their business invested in one particular industry. By investing in business in other industries their risk exposure may be reduced. Also tax considerations could be motives for M&A.

The amount of possible benefits of M&A would let one believe that most M&A deals would lead to positive wealth creation for shareholders. These theories however might be a great

misconception of the outcomes in reality. There have been many studies published about the short term effect of mergers and acquisitions. These studies suggest that overall M&A produces wealth for shareholders. Using time periods surrounding the announcement date target shareholders

immediately earn significantly positive abnormal results from all acquisitions. Meanwhile acquiring shareholders earn no abnormal returns and negative abnormal returns from tender offers and mergers respectively (Loughran and Vijh, 1997). The abnormal returns from acquirers and targets combined are positive overall and thus suggest that mergers and acquisitions create value and that those deals do increase efficiency. A fewer amount of studies have been done on the performance on the longer term after the effective merger or acquisition date questioning the assumption of market efficiency. A great amount of these studies show negative abnormal returns over a period of three to five years after the announcement date. These negative abnormal results are inconsistent with market efficiency and suggest that the changes in stock prices overestimate the future efficiency gains from mergers (Jensen and Ruback, 1983). Thus short term returns may be a biased forecast for the future.

Concerning long-term event studies on the performance of mergers and acquisitions Mandelker (1974) was the first to study this topic using a modern methodology, namely the Fama-MacBeth two-factor model. His results show negative abnormal returns, but does not report the statistical significance of them. Concluding that the short-term stock price reaction to mergers exhaust all valuable information in mergers and that capital markets are efficient. Langetieg (1978) finds large negative significant cumulative abnormal returns of more than 20% for a time period of 70 months after the merger using CAPM, industry indices and the Black two-factor model. However when using a control firm approach the merging firms performances are still worse than their

counterparts, but not statistically significant. Considering the latter he concludes the efficient market hypotheses still holds. Asquith (1983) introduces a method of using beta control portfolios as a benchmark to calculate excess returns. Using daily data to calculate the abnormal returns he finds

(7)

7

significant negative abnormal returns for a 240 day time period, where most of the decline in the equity value is delayed by more than 60 days. He is the first to question the efficiency of capital markets regarding merged firms. Malatesta (1983) does not agree with the questioning of market efficiency, despite finding negative significant negative abnormal results over a 12 month time period, using the market model, himself. He states that changes in risk parameters around the merger event lead to statistical issues resulting in losses after the mergers. He argues it is nearly impossible that investor systematically misinterpret the widely disseminated information about the characteristics of the mergers. Thus rejecting market inefficiency. Also Bradley and Jarrel (1988) find insignificant negative abnormal returns using the same method as Asquith (1983), but over a 3 year time period. Using the market model, Magenheim and Muller (1988) are the first to make a

distinction for post-acquisition returns between mergers and tender offers. However, they do not report the statistical significance of their results, only Z-scores. But they do find an economical significant difference of around 30 percent between merger and tender offers, where tender offers receive positive returns. Bradley and Jarrel (1988) also criticize the method of Magenheim and Muller (1988), arguing that market model parameter estimates based on monthly data are inefficient and nonstationary. Measurement of abnormal performance in forecast periods will be biased when abnormal performance is non-zero during the estimation period (Agrawal and Jaffe, 2000). Franks, Harris and Mayer (1988) use four different methods to calculate the cumulative abnormal returns in the US and UK for two years after the merger. They also make a distinction between all-cash

acquisitions and equity takeovers. Firstly, for equity takeovers three out of the four measures lead to significant negative abnormal returns. The method resulting in the insignificant result, the market model is used as a benchmark with the estimation period 25 until 60 months after the merger. if the abnormal performance during that estimation period is negative, this will lead to biased estimates. Secondly for all-cash takeovers the results are insignificantly different from zero for all four methods. For the UK two methods are used which both lead to different results. For equity takeovers leading to significant negative results for one method and insignificantly different from zero for the other. For cash takeovers the results are positively significant for one method and insignificant for the other. These results suggest that the benchmark used for long-term studies is important and that returns following all-cash acquisitions are higher than equity funded acquisitions. Since mergers are likely to be financed with equity and tender offers with cash, these results are in line with

Magenheim and Mueller (1988). Limmack (1991) inspects the performance of acquirers for both successful and unsuccessful bidders during the two years after the merger. He finds significant equally weighted negative abnormal returns for are significant for both successful and unsuccessful bidders, with the performance of unsuccessful bidders being much worse. For value weighted abnormal returns the significance is lost for the successful bids but remains for the unsuccessful

(8)

8

bidders. Frank, Harris and Titman (1991) argue that previous findings of negative performances after M&A are likely due to benchmarks errors instead of mispricing at the time of the takeover. They used four different benchmarks. Namely the CRSP value and equally weighted indexes, a ten-factor model provided by Lehman and Modest (1987) and an eight-portfolio model from Grinblatt and Titman (1989). They find different results for their benchmarks with some significant nonzero performances. However, they find no significant abnormal returns when using their preferred eight-portfolio method as a benchmark. This made them to believe that the underperformance of mergers and acquisitions on the long-term are likely due to benchmark errors. Loderer and Martin (1992) studied the performance of four different type of acquisitions over a five year time period after the

acquisition. They divided the acquisitions in the categories mergers, tender offers, other acquisitions and acquisitions of assets. They find insignificant abnormal performances for all categories for a five year period. They do however find some negative results for some sub-periods, but they seem to disappear in later years. Hence concluding that in the later years the performances after acquisitions do not provide strong evidence against market efficiency. Whereas Agrawal, Jaffe and Mandelker (1992) do find significant negative abnormal returns for the same time five year time period for mergers and insignificant positive results for tender offers. The studies both use size and beta adjustments for abnormal returns yet find different results. The factor in these differences here may be the time periods. As Loderer and Martin (1992) mention that the negative abnormal results seem to disappear in later years. Anderson and Mandelker (1993) extend the research of Agrawal, Jaffe and Mandelker (1992) and find the same results for mergers after controlling for the book-to-market affect under the size and book-to-market adjustments for abnormal returns. Gregory (1997) finds significant negative abnormal returns for six different methods that he uses. His sample includes only mergers and acquisitions from the UK. Using the Dimson and Marsh (1986) model he finds even worse returns for the UK than Agrawal, Jaffe and Mandelker (1992) find for the US. However, the time periods studied differ a lot between the two studies. So the results may me more sample-specific than country-sample-specific. In 1997 Loughran and Vijh (1997) introduce the buy and hold abnormal return (BHAR), where the BHAR is the buy and hold return of the acquiring firm minus the buy and hold return of a control firm based on the market-to-book ratio and size for a five year period. They find significant negative abnormal returns following mergers and non-significant differences from zero for tender offers. This is in line with previous research subject to the difference of abnormal returns between tender offers and mergers. Mitchell and Stafford (2000) also use the BHAR method and include the calendar-time portfolio regression method. They make a distinction between equal and value weighted averages of returns and portfolios of acquirers. For the BHAR they find significant negative returns and for the calendar-time portfolio regressions insignificant negative excess returns. However the economic magnitude of the BHAR are very small and hence a hard conclusion is still

(9)

9

hard to make. Moeller, Schlingemann and Stulz (2003) also find large significant negative abnormal returns using the BHAR method over the same three year time period, but of much larger magnitude. When they use the calendar-time portfolio method the returns are again insignificant. Andre, Kooli, and L’Her (2004) find the same results for the calendar-time portfolio with Fama-French regression. Dutta and Jog (2009) do again not find clear evidence for negative abnormal returns for Canadian acquiring firms for a three year time period using different methodologies. Hence the puzzle about long-term underperformance of mergers might all be due to benchmark errors. Fama (1998) believes that these anomalies are chance results. He concludes that post-event continuation of pre-event abnormal returns is about as frequent as post-event reversal. He found that the anomaly about negative abnormal returns disappears when incorporating changes in estimation methods and he recommends the calendar-time portfolio approach. This is in line with Mitchell and Stafford (2000). They do not find any statistical evidence of underperformance when correcting for methodological discrepancies like cross-correlation. Roll (1986) links anomalies in post-merger performance with underreaction to poor investment decisions by overconfident CEO’s, while Mitchell and Stafford (2000) mention it could also be due to the overreaction to strong performances of the acquiring firm beforehand the mergers.

2.2 M&A Performances in High-Tech Environment

Bena and Li (2014) show that an acquiring firm’s innovative output improves after the deal when there is prior technological linkage to their target firms. They conclude that synergies obtained from combining innovation capabilities are important drivers of acquisitions. Also Sevilir and Tian (2012) find an increase in innovation outcomes for the acquirers after M&A, particularly if the target firms are innovative before the merger. This effect is even higher when the innovation efficiency of the acquirer firm is lower than that of the target firm. Thus suggesting that less innovative firms acquire innovative firms to enhance their innovation output. Rhodes-Kropf and Robinson (2008) state that M&A enhances innovation by bringing together firms with complementary assets which will generate more innovative products and technologies together than alone. Firms that are not very good at innovating themselves can acquire this by acquiring firms which are more efficient at innovation (Aghion and Tirole, 1994). This may be more efficient than to invest internally in innovation within the firm. Cloodt, Hagedoorn and van Kranenburg (2006) show that in a high-tech setting the

acquisition of a large absolute knowledge base only contributes to improved innovative performance during the first couple of post-M&A years. Targets from high-tech sectors may be able to provide greater shareholder wealth benefits to acquiring companies in comparison with lower growth target companies. Therefore these high-tech industries distinguish themselves from other industries (Kohers & Kohers, 2000). They find that acquirers of high-tech targets experience significant positive

(10)

10

abnormal returns at the time of the merger announcement. They state that further research of the long term effect would be interesting to see if the initial positive reaction is an unbiased forecast for the future. Bloom and van Reenen (2002) and Nicholas (2008) find that innovation indeed is a key driver for firm value. Therefore it is interesting to see if the long-term firm performance is positive for high-tech firms after M&A, or if they at least perform better than other industries.

Takeovers that focus in the same industry enhance shareholder wealth, whereas diversifying takeovers decrease shareholder wealth (Kohers and Kohers, 2000). So when both the acquirer and target firm are classified as high-tech, this should result in positive wealth creation. Kohers and Kohers (2000) also state that due to their high growth potential, high-tech targets might add more value than other sectors. However this also brings more risk caused by the uncertainty of these companies whose values rely on the unproven future outcomes of developments. Sorescu, Chandy and Prabhu (2007) show that firms with high product capital select targets with greater innovation potential and deploy this innovation potential better than firms with low product capital. This

superior selection and deployment are lead to superior long-term financial performance. They define product capital as the product development and product support assets that a firm’s current and prior investments create. Where product support assets are assets that are devoted to the promotion of consumer adoption of new products and product development assets are those devoted to the creation, development and improvement of new products. In their study they used the pharmaceutical industry across seven different countries. However high technology firms are classified as high product capital firms according to their definition. High-tech firms are better at evaluating the innovation of target firms. Due to their prior related knowledge, they are better at evaluating outside knowledge and handling the risk associated with high-tech targets resulting in abnormal performances (Sorescu et al., 2007). Firms that conduct their own research and

development are better able to use externally available scientific and product-related information. High-product-capital firms have a greater ability to assimilate new assets (Cohen and Levinthal, 1990) and realize potential synergies with the target (Makadok, 2001) by transforming promising ideas into productive outputs. The acquiring firm must be able to understand and use the new assets. By conducting their own R&D firms with high-product-development assets are more likely to possess the knowledge and use externally available scientific and product related information that is crucial to understanding new technologies (Sorescu et al., 2007). Also high-tech firms are more disciplined in choosing right targets, because they stand to lose more by diluting a more valuable asset base by choosing badly. Hence, such companies have a greater incentive for monitoring a target. Ranft and Lord (2000) and many other studies mention that in high-tech acquisitions key is to effectively integrate and hold the newly acquired personnel. Those are the people that will make the innovative process a success. They can however leave at any time. Due to the expectations of greater

(11)

11

professional success of high-tech firms the key employees will have greater incentives to stay post-acquisition because they will perceive the long-term benefits when remaining with the merged firm (Sorescu et al., 2007).

3. Hypotheses

According to the literature on long-term performance of mergers and acquisitions there is a believe that short term returns do overestimate the future potential benefits of such a deal. However Fama (1998) and Mitchell and Stafford (2000) concluded that these anomalies of market efficiency are due to benchmark errors. This leads to the first hypothesis: The long-term abnormal returns of post-M&A

performance are insignificantly different from zero. Mergers and acquisitions in a high-tech setting

perform really well on the short-term (Kohers and Kohers, 2000). Other studies described in the literature section provided evidence that innovative performance of high-tech companies increases in the long-run after an M&A deal. Furthermore, studies have shown that innovation is a big part of value creation (Bloom and van Reenen, 2002; Nicholas, 2008). High-tech firms are better at selecting, deploying and performing better in an M&A process (Sorescu et al., 2007). Thus high-tech M&A deals should overall be performing better than other industries. Hence leading to the second hypothesis:

The long-term abnormal returns of post-M&A performance in high-tech industries are insignificantly different or significantly higher than zero. And the third hypothesis: High-tech M&A outperforms M&A in other industries.

4. Methodology

The research methodology section contains an illustration of the buy-and-hold abnormal returns approach and calendar-time portfolio approach.

4.1 Buy-and-Hold Abnormal Returns

To investigate the long-term performances of mergers and acquisitions firstly buy-and-hold abnormal returns (BHARs) of the acquiring firms will be calculated for a time period of three years after the merger announcement date to measure the long-term performance. The BHAR is calculated as the difference between the buy-and-hold returns of the acquirer’s stock and a benchmark. The BHAR is defined as: 𝐵𝐻𝐴𝑅𝑖𝑇= ∏(1 + 𝑅𝑖𝑡) − ∏(1 + 𝐸(𝑅𝑖𝑡)) 𝑇 𝑡=1 𝑇 𝑡=1

BHARiT is the buy-and-hold return of the acquirer at t = 36, with t = 0 being the merger

announcement date and proceeding in calendar months. Rit is the monthly return of acquirer i at

(12)

12

In this study several benchmarks will be used. Since previous studies have shown that different benchmarks yield different results, for robustness checks more benchmarks will be used to compare. The benchmarks that are included are:

- The CRSP equally-weighted index - The CRSP value-weighted index - The S&P 500 index return

- Market capitalization decile portfolio based upon the CRSP Portfolio Assignment database - Beta decile portfolio based upon the CRSP Portfolio Assignment database

- Standard deviation decile portfolio based upon the CRSP Portfolio Assignment database At the announcement date of the deal the acquirer gets assigned with three different matching decile portfolios based upon market capitalization, beta and standard deviation. The matching portfolios are not rebalanced and will be the same for the three years after the deal. In that case the differences in buy-and-hold returns are true BHARs.

After the BHARs are generated, the mean buy-and-hold return will be computed. An equally-weighted and value-equally-weighted mean will be calculated. Where the value-equally-weighted mean is given by:

𝐵𝐻𝐴𝑅

̅̅̅̅̅̅̅̅ = ∑ 𝑤𝑖𝐵𝐻𝐴𝑅𝑖 𝑁

𝑖=1

The value weight is based on the acquirer’s market capitalization on the announcement date.

Previous studies have shown that there are differences between equal-weighted and value-weighted returns and that the abnormal returns are generally smaller when value-weighted. This might be due to value-weighted returns more accurately capture the total wealth effect experienced by investors (Fama, 1998).

To test the if the null hypotheses if the mean BHAR differs from zero the following t-statistic is used:

𝑡 = 𝐵𝐻𝐴𝑅̅̅̅̅̅̅̅̅/(𝜎𝐵𝐻𝐴𝑅 √𝑁 )

Where σBHAR is the standard deviation of the buy-and-hold abnormal returns and N is the number of

sample events. From this statistic the p-value is derived and checked if there is a significant difference from zero abnormal returns.

The BHARs will be used for further analysis in a regression analysis. To test if the mergers and acquisitions in high-tech industries outperform other industries a regression with control variables will be run. There are several factors that are proven to have an effect on the outcome of the post-acquisition performances of the acquirers.

Bidders that use stock might imply that the board believes that their shares are overvalued and would make the target shareholders decline the bid (Myers and Majluf, 1984). Therefore the

(13)

13

market will perceive cash offers as more positive than stock offers. On the other hand, since targets will only accept cash offers greater than its current value, the bidder might not want to pay with cash when it is uncertain about the target’s value as it will be overpaying (Fuller, Netter and Stegemoller, 2002). A model developed by Eckbo and Thorburn (2000) that the expected overpayment cost of cash is greater than the expected overpayment cost with stock. Stock offers might therefore be preferred. Loughran and Vijh (1997) and Gregory (1997) have shown that for bidding firms that finance their M&A deals with cash have higher long-term abnormal returns than those that are financed with stock and outperform their matched firms.

Previous studies also showed that there is a difference in long-run abnormal returns between merger and tender offers (Rau and Vermaelen, 1998; Loughran and Vijh, 1997; Agrawal, Jaffe and Mandelker, 1992). Tender offers tend to be hostile and are usually paid with cash. Since cash payments have higher abnormal returns this in line with previous research. Also, these hostile takeovers could result in the current management of the target company to be replaced to more efficient managers with wealth gains as a result (Loughran and Vijh, 1997).

Rau and Vermaelen (1998) prove that highly-valued acquirers, which are mentioned as “glamour-firms”, underperform their characteristic-matched firm portfolio. These glamour-firms are probably affected by overconfident CEO’s that overestimate their own abilities and pay too much for their targets (Roll, 1986). Roll calls this phenomenon the “hubris”. effect The shareholders and of these firms are also more likely to trust the management’s and approve of its decisions whereas in low value firms the opposite effect holds where the survival of the company has to be considered (Rau and Vermaelen ,1998).

Further does the public status of the target seem to influence the post-M&A performance. Prior research shows that acquiring a public target generally leads to worse post-merger

performance than when acquiring a private target. Chang (1998) argues this by saying that large blockholders may be created by making a stock acquisition due to the small amount of shareholders in target firms. If these blockholders are better at monitoring the bidder’s management, its

performance could improve.

At last Moeller, Schlingemann and Stulz (2004) show that bigger sized acquiring firms tend to worse after an merger-event.

Controlling for these factors the following regression is run:

𝐵𝐻𝐴𝑅𝑖𝑇 = 𝛽1+ 𝛽2𝐻𝐼𝐺𝐻𝑇𝐸𝐶𝐻𝑖+ 𝛽3𝐹𝐼𝑁𝐴𝑁𝐶𝐼𝑁𝐺𝑖+ 𝛽4𝑂𝐹𝐹𝐸𝑅𝑖+ 𝛽5𝑄𝑖𝑡+ 𝛽6𝑃𝑈𝐵𝐿𝐼𝐶𝑖

(14)

14

Where BHARiT is the 36 month buy-and-hold return for M&A event i. HIGHTECHi is a dummy variable

that is equal to one when both the acquirer and target company are considered high-tech and equal zero otherwise. Information regarding the classification of high-tech industries is described in the data section. The payment method is expressed in the dummy variable FINANCINGi.This dummy

variable is equal to one if the medium of exchange is 100% cash and equal to zero otherwise. The dummy variable OFFERi is equal to one in the case of a tender offer and equal to zero otherwise. Qit is

the variable Tobin’s Q, measured by the market-to-book ratio at the announcement date. If the target is a public company the dummy PUBLICi is equal to one and zero otherwise. SIZEit is the size of

the firm measured by the natural logarithm of the assets at the time of the announcement date. To test the second hypothesis if post-acquisition performance of high-tech M&A outperforms other industries the dummy for indicating a high-tech merger should be significantly positive.

Regarding the previously described information on the control variables first of all FINANCINGI should

be positive, as cash-acquisitions seem to outperform stock transactions in previous literature. OFFERi

should also be positive as tender offers are 1) for a great deal paid with cash and 2) higher chance of replacement current target’s management into more efficient management. The variable for Tobin’s Q is expected to be negative, because high-value bidders are likely to be infected by hubris. As M&A of private target firms seems to be more successful than acquiring or merging a public firm, PUBLICi is

expected to be negative.

4.2 Calendar-Time Portfolio Approach

The second approach that is used is the calendar-time approach. The BHAR approach is potentially subject to statistical issues resulting in biased estimations. Kothari and Warner (2005) mention that buy-and-hold abnormal returns exhibit cross correlations because: (i) the abnormal returns of event firms are likely to share common calendar months due to the long measurement period; (ii)

corporate events like mergers occur in waves; (iii) mergers tend to be clustered within particular industries. This cross-dependence is ignored in the BHAR approach and will lead to overstated test statistics. The BHAR methodology is subject to the “bad model” problem mentioned by Fama (1998). The assumption in the BHAR approach is that the benchmark portfolios used completely describe the expected returns. However all models are incomplete descriptions of the systematic patterns in average returns during any sample period (Fama, 1998). This problem grows with the time horizon. Due to compounding the buy-and-hold abnormal performance measure will increase in holding period. Furthermore the BHARs tend to be skewed, due to being unbounded above but bounded below -100%.

(15)

15

Fama (1998) suggest using the calendar-time approach. In this approach each month a portfolio of firms that completed the event is formed. This makes it robust to the cross-sectional correlations and are each month automatically accounted for in the portfolio variance.

As just mentioned, each month a portfolio is formed of firms that have acquired a target firm in the last three years. So from each month from January 1981 until December 2009 there will be a portfolio of acquirers. These portfolios will be formed based on equal-weights and value-weights, where the value-weight is based on the market capitalization. Each months these portfolios are being rebalanced by dropping the acquirers that past their three year time period and adding firms that just executed a transaction. This will be done for high-tech only firms, the other firms in the sample and the whole sample of firms. The time-series of these returns will be regressed on the three Fama and French (1993) factors. Also the Carhart (1997) momentum factor will be included in the

regressions. Hence the regression will look as follows:

𝑅𝑖,𝑡− 𝑟𝑓,𝑡 = 𝛼𝑖+ 𝛽𝑖,𝑀(𝑅𝑚,𝑡− 𝑟𝑓,𝑡) + 𝛽𝑖,𝑆𝑀𝐵𝑆𝑀𝐵𝑡+ 𝛽𝑖,𝐻𝑀𝐿𝐻𝑀𝐿𝑡+ 𝛽𝑖,𝑀𝑂𝑀𝑀𝑂𝑀𝑡+ 𝜀𝑖,𝑡

Where Ri,t is the portfolio return at time t and rf,t is the one-month treasury bill rate. Hence Ri,t – rf,,t

measures the excess return of the portfolio at time t. Rm,t is the market return at time t so Rm,t – rf,t

measures the excess market return. SMBt is the difference in returns of a portfolio of small minus big

stocks and HMLt is the difference between a portfolio of high to-market stocks and low

book-to-market stocks. MOMt is the monthly momentum factor calculated by subtracting the equal

weighted average of the lowest performing firms from the equal weighed average of the highest performing firms, lagged one month (Carhart, 1997). In this case, the intercept αi measure the

average monthly abnormal return on the portfolio. For the first hypothesis that the abnormal return for all firms should be zero the intercept should be statistically insignificant different from zero. For the second hypothesis the intercept has to be significantly higher than zero.

To test the third hypothesis using the calendar-time portfolio approach, a long-short

portfolio will be formed. In this case the long position will be in high-tech acquirers that participated in a M&A event in the past three years and short in the non-high-tech acquirers that been in an event in the past three years. These portfolios will also be equally and value weighted. The regression will then look like:

𝑅𝐻𝑇,𝑡− 𝑅𝑁𝐻𝑇,𝑡= 𝛼𝑖+ 𝛽𝑖,𝑀(𝑅𝑚,𝑡− 𝑟𝑓,𝑡) + 𝛽𝑖,𝑆𝑀𝐵𝑆𝑀𝐵𝑡+ 𝛽𝑖,𝐻𝑀𝐿𝐻𝑀𝐿𝑡+ 𝛽𝑖,𝑀𝑂𝑀𝑀𝑂𝑀𝑡+ 𝜀𝑖,𝑡

Where RHT,t is the return on the high-tech portfolio at time t and RNHT,t the return on the portfolio of

(16)

16

t. Therefore the intercept αi measures the average monthly excess return of the long-short strategy.

This intercept is expected to be positive according to the third hypothesis.

5. Data and descriptive statistics

In this section first the construction of the sample is clarified after which descriptive statistics are presented.

5.1 Construction of sample

The data regarding the merger transactions is accessed through the Thomson One Mergers and Acquisitions database. The deals included are in the time period from 1 January 1981 until December 2006. Search criteria is then filtered by only including only deals where the acquirer is a public firm. Further, only deals that took place in U.S. are included. To eliminate small and insignificant deals only deals with a value of more than one million dollars will be included. Also firms in the financial sector are excluded (SIC 6000-6999). This leads to a sample of 27,799 M&A events. The data obtained from Thomson One includes: The announcement date of the deal, acquirer name and cusip, target name, deal number, acquirer and target macro industry, acquirer and target SIC code, value of the

transaction, target’s public status, method of payment and a flag indicating a tender offer. The acquirer 6-digit CUSIPs provided by the Thomson One database are then used to gain their PERMNOs. This is done by using the CRSP Translate Tool. The PERMNOs are then used in the CRSP Monthly Stock File database to gain the stock data from January 1981 to December 2009. The acquired data from this database includes the PERMNO, NCUSIP, company name, the monthly holding period and delisting returns, CRSP value- and equally-weighted as well as the S&P 500 holding period returns. This data is then merged with the Thomson One data by using the first six digits of the CRSP’s NCUSIP, that corresponds to the CUSIP provided by Thomson One. From this data the three year buy-and-hold returns after the merger event are calculated. The CRSP holding period return is already adjusted for stock splits, exchanges, and cash distributions. Some acquirers

disappear from CRSP in the three year period after the event. Following a delisting the delisting return is used to compute the stock buy-and-hold return. This buy-and-hold return is then carried forward to the last observation. In this case the hold return reflects the return of a buy-and-hold investor the same way as investing in a listed firm for three years.

To obtain the portfolio assignments the CRSP Stock / Portfolio Assignments database is used. The PERMNOs acquired CRSP translation tool are used as company identifiers for this database. This database assigns the acquiring firm to one of ten deciles based on market capitalization, beta or standard deviation if data is available. Then each firm is matched with the portfolio that it has been assigned in the year of the merger announcement date. The indexes and returns for each of those

(17)

17

ten decile portfolios for the three variables are obtained from the CRSP Index/Stock File Indexes database.

From the CRSP/Compustat Merged database the total assets, common shares outstanding and fiscal annual close price of the acquirers is obtained. Then the market capitalization is computed by multiplying the common shares outstanding with the share price. The market capitalization is divided by the total assets to derive Tobin’s Q. The data of most recent preceding fiscal year end is matched with the acquirer to its announcement date. Ultimately the sample consists of 24332 M&A events with available three year buy-and-hold returns.

And at last the Fama-French factor returns and the momentum factor return are obtained from the Fama-French data library.

Whether the M&A deals are classified as high-tech is based on the SIC code

recommendations by the study of Kile and Phillips (2009). Also the industries based on the OECD (2011) report on classification of high-tech industries is taken into the analysis, which classifies industries as high-tech based on their R&D intensity. A list of the industries is found in the appendix. When the industry of both the acquiring and target company are classified as high-tech the M&A deal is considered to be high-tech.

5.2 Descriptive statistics

Table I provides a summary of the descriptive statistics of acquirer and deal characteristics. The characteristics from the acquiring firm are computed from data at the end of the fiscal year

preceding the event. First of all the table shows that the differences in means between high-tech and non-high-tech M&A deals is significant for all characteristics. The deal value seems to be bigger for high-tech M&A deals than for non-high-tech deals, with 295.07 and 235.10 respectively. Although the median for deal value for high-tech mergers is lower a few big deals seem to pull this average up, as seen in the larger standard error. The next three rows in the table concerning market

capitalization, Tobin’s Q and assets indicate that the sample of acquirers in high-tech deals contains a lot of “glamour-firms” in comparison with acquirers from the other deals. The market capitalization of the average high-tech acquiring firm is more than three times bigger than non-high-tech acquirers, while book value of assets is differ a lot less with 700 million difference in means. Therefore it’s no surprise that the Tobin’s Q is also a lot bigger for high-tech acquirers. But, again, this difference is likely the effect of a few really high valued firms. This appears from the high standard errors for the high-tech sample, which is more than double the value for Tobin’s Q and nearly two and a half times bigger for the market capitalization. However, the differences are statistically significant. Non-high-tech M&A deals seem to be more often financed with cash, 34.39%, than high-Non-high-tech deals, 31.30%. Because high-valued firms might rather want to pay with stock, this seems plausible.

(18)

18

Table I

Descriptive Statistics

This table shows the key variables of acquiring firm characteristics and deal characteristics for a sample of 25,226 M&A events during the time period of 1 January 1981 until 31 December 2006. Deal value is the dollar value of the transaction in millions. Market cap is the market capitalization (common shares outstanding times fiscal-year closing price) in millions of dollars at the announcement date of the event. Tobin's Q is the ratio of market capitalization to total assets. Total assets is the book value of assets in millions of dollars. Cash is a dummy indicating a cash only transaction. Tender is a dummy variable indicating a tender offer. Public is a dummy variable indicating a public target. The p-value of difference refers to the difference in means between

high-tech and non-high-tech industries.

Panel A

High-tech Non-high-tech

p value of difference

Mean Median SE N Mean Median SE N

Deal value 295.07 21.08 2346.24 7588 235.10 25.00 1968.80 16744 0.04 Market cap 10683.81 482.09 41793.41 7314 3272.94 401.68 16885.36 16144 0.00 Tobin's Q 2.81 1.68 4.66 7312 1.29 0.89 1.72 16143 0.00 Total assets 3970.23 266.89 14402.60 7326 3271.58 428.00 15745.08 16167 0.01 Cash 31.30% 0.46 7588 34.39% 0.48 16744 0.00 Tender 4.14% 0.20 7588 6.13% 0.24 16744 0.00 Public 26.87% 0.44 7588 25.81% 0.44 16744 0.08

Tender offers do not often appear in the sample. Only 4.14% of the high-tech deals are tender offers with 6.13% for the other deals. This seem quite low in comparison with previous research. There are however no missing values in the data for the tender flag. The reason might be the time period, in which case the relative amount of tender offer seems to have dropped in comparison with earlier time periods from other studies. The higher percentage is in line with the assumption that tender offers are more often paid with cash. The difference in public targets, 1.06 percentage points, is again economically very small but statistically still significant at the 10% level.

6. Results

This section describes first the results of the mean buy-and-hold abnormal returns using various benchmarks. Next the regression results using control variables are presented. Thereafter the calendar-time portfolio approach is described and at last the long-short portfolio returns.

6.1 Mean Buy-and-Hold Abnormal Returns

Table II present the main results of the three year mean buy-and-hold abnormal returns for acquiring firms between 1 January 1981 and 31 December 2009. In panel A the results regarding the whole sample of acquirers are presented. The second column of panel A shows the equally-weighted mean BHARs for each benchmark, ranging from -4.48% to very large percentage of -32.78%. All BHARs are significantly different from zero, which indicates that post-acquisition acquirers underperform in

(19)

19

comparison with their benchmark portfolios. For the value-weighted mean BHARs the results are still all negative except when using the S&P 500 returns as a benchmark. However, the significance of the results vanish when using value-weighted and size adjusted benchmarks (CRSP VW and market capitalization portfolios). This is in line with Fama (1998) and Mitchell and Stafford (2000) who show that significance and magnitude of abnormal returns generally vanish using value-weighted returns and size adjusted benchmarks. Value-weighted returns more accurately capture the total wealth effects experienced by investors. Quite alarming is the large underperformance of acquirers to their risk-adjusted portfolios of -25.50% and -28.83% for beta and -32.78% and 20.12% for standard deviation. Acquiring firms tend to have a good performance before a merger announcement and earn substantial positive abnormal results. Investors tend to give too much weight to these past performances ignoring the mean-reverting effect. This could lead to serious misinterpretations of these risk coefficients leading to wrong portfolio assignments. In this analysis the beta and standard deviation portfolios are assigned at the announcement date. Merging with another firm most likely changes the firm’s pre-merger risk coefficients. Not allowing to vary these risk parameters over time may cause these high and statistically significant negative abnormal returns. Hence, as mentioned by Fama (1998) most of the anomalies regarding long horizon abnormal returns may be the cause of benchmark errors. However, in undisclosed results, the abnormal returns using these portfolios do not significantly change when they are allowed to change over time. The results in panel A show how important it is to choose a good benchmark, since all benchmarks yield very different results. Using value-weighted BHARs and size adjusted benchmarks should resolve some of the bad-model problems. Mispricing tends to occur more often in small stocks due to behavioral financial reasons (Fama, 1998). Using value-weighted returns and size adjusted benchmarks, small stocks are given less weight in the analysis and measured against an appropriate benchmark, therefore gives better results. As seen in panel A this seems to solve the anomaly of negative abnormal returns for long term mergers, given the small economic magnitude and statistical insignificance of these mean BHARs. The acquirers also significantly outperform the market with a mean BHAR of 5.20% over the S&P 500 market index. The value-weighted BHAR methodology is preferred interpreting these results as equal-weight is suggestive of serious mispricing.

In panel B the equal-weighted mean BHARs for the high-tech subsample and the rest of the acquirers are shown. The value-weighted mean BHARs for the subsamples are presented in panel C. Again can be seen that there is a big difference between benchmarks and value- and equal-weighted mean BHARs. For instance for high-tech acquirers the mean BHAR is nearly twice as negative using the CRSP equal-weighted portfolio benchmark, -15.16% and -28.87% for equal- and value-weighted BHARs respectively.

(20)

20

Table II

Three-Year Mean Buy-and-Hold Abnormal Returns (BHARs) for acquirers (January

1981-December 2006)

The BHARs are calculated as the difference between the equal- and value-weight average three year buy-and-hold period return for the acquirers and benchmark portfolios. The benchmark portfolios are reported in the first column. CRSP EW and CRSP VW are the CRSP equally- and value-weighted indexes. S&P 500 is the S&P 500 index. Mkt Cap, Beta and Std dev are the benchmark portfolios based on market capitalization, beta and standard deviation deciles respectively. The BHAR is the mean buy-and-hold abnormal returns. SE indicates the standard error and N the sample size. The p-value refers to the mean BHAR differing from zero. In panel A the equal- and value-weighted mean BHARs for all acquirers in the sample are presented. In panel B and C the equal- and value-weighted mean BHARs of the two subsamples are reported respectively. The p-value of difference in panel B and C refers to the difference in means between high-tech and non-high-tech acquirers.

Panel A: Whole Sample

Equal-Weight Value-Weight

Benchmark BHAR Median SE N _{p-value BHAR Median} SE N p-value CRSP EW -.1465 -.3603 .0089 24332 .000 -.2359 -.4308 .0222 23457 .000 CRSP VW -.1102 -.3127 .0089 24332 .000 -.0175 -.1528 .0157 23457 .265 S&P 500 -.0438 -.2342 .0090 24332 .000 .0520 -.0788 .0156 23457 .001 Mkt Cap -.1149 -.3106 .0090 23784 .000 -.0116 -.1278 .0153 23063 .448 Beta -.2550 -.3744 .0106 10366 .000 -.2883 -.4038 .0223 10137 .000 Std dev -.3278 -.2498 .0253 10168 .000 -.2012 -.3072 .0227 9942 .000 Panel B: Sub Samples Equal Weight

High-tech Non-high-tech p-value

of diff. Benchmark BHAR Median SE N _{p-value BHAR Median} SE N p-value

CRSP EW -.1516 -.4674 .0186 7588 .000 -.1442 -.3204 .0098 16744 .000 .701 CRSP VW -.0558 -.3404 .0183 7588 .002 -.1348 -.2987 .0100 16744 .000 .000 S&P 500 .0123 -.2618 .0184 7588 .503 -.0692 -.2221 .0101 16744 .000 .000 Mkt Cap -.0895 -.3769 .0184 7488 .000 -.1266 -.2860 .0100 16296 .000 .055 Beta -.2399 -.3693 .0226 1927 .000 -.2585 -.3766 .0120 8439 .000 .496 Std dev -.4156 -.2863 .0623 1897 .000 -.3076 -.2436 .0277 8271 .000 .097 Panel C: Sub Samples Value Weight

High-tech Non-high-tech p-value

of diff. Benchmark BHAR Median SE N _{p-value BHAR Median} SE N p-value

CRSP EW -.2887 -.5388 .0325 7314 .000 -.1580 -.2838 .0239 16143 .000 .001 CRSP VW -.0297 -.2067 .0235 7314 .207 .0006 -.0338 .0166 16143 .973 .294 S&P 500 .0386 -.1231 .0233 7314 .098 .0718 .0302 .0166 16143 .000 .245 Mkt Cap -.0199 -.1818 .0228 7253 .384 0.008 -.0240 .0167 15810 .961 .464 Beta -.3272 -.5060 .0359 1891 .000 -.2590 -.3511 .0278 8246 .000 .132 Std dev -.2934 -.4223 .0361 1860 .000 -.1312 -.2339 .0280 8082 .000 .000

While using the CRSP value-weighted portfolio as benchmark the value-weighted mean BHAR halves in comparison to the equal-weighted equivalent from -5.58% to -2.97%. Anyway, the equal-weighted mean BHARs in panel B are all negative and significant. The only exception is using the S&P 500 benchmark, where high-tech acquirers outperform the market with 1.23% post-M&A. The results is of small economic magnitude and statistically insignificant though. The magnitude of the negative results are rather big, ranging from -5.58% to -41.56% for high-tech acquirers and -6.92% to -30.76% for the rest of the sample. The high-tech acquirers outperform the other acquirers from the sample for four of the six benchmarks with a difference of 7.9% and 8.15% at the 1% level using the CRSP

(21)

21

value-weighted portfolio and S&P 500 respectively. Using the size adjusted portfolio the

outperformance is 3.71% but only significant at the 10% level. However, the medians in the high-tech sample are lower, suggesting there might be some firms with a really good performances that lift up the mean BHARs.

Panel C present the value-weighted mean BHARs for the two subsamples. Again using the CRSP equal-weighted, beta adjusted and standard deviation adjusted portfolios the acquirers have large negative mean buy-and-hold abnormal returns. For the high-tech sample all significant, ranging from -28.87% to -32.72%. For the non-high-tech acquirers the values using the CRSP equal-weighted standard deviation adjusted portfolios halved to -15.18% and -13.12% respectively. These differences are both significant at the 1% level, meaning the non-high-tech acquirers outperformed the high-tech acquirers in the three years post-merger. However, these benchmarks are likely to produce biased estimates. For both samples using the value-weighted and size adjusted portfolios the abnormal returns are insignificant. Further, the table shows that both samples significantly outperformed the S&P 500 market index at the 10% and 1% level for the high-tech and non-high-tech sample

respectively. The high-tech acquirers did with 3.86% while the other subsample did with 7.18%. However, the difference is insignificant just as for the value-weighted and size-adjusted benchmark where the high-tech sample is outperformed.

6.2 Buy-and-Hold Abnormal Return Regressions

The equal-weighted mean BHARs in Table II suggest that high-tech M&A outperforms non-high-tech M&A on the long-term. However the preferred value-weighted method results in insignificant differences for value- and size-adjusted as well as the market portfolio. To check whether these effects hold or change when controlling for factors that in previous literature have been proven to influence abnormal returns post-merger, regressions are run with the equal- and value-weighted BHARs. The results of these regressions are presented in Error! Reference source not found..

In the table can be seen that post-M&A performance is only significantly better for the equal-weighted BHARs using the value-equal-weighted CRSP and size adjusted portfolios and S&P 500 as

benchmark. This is in line with the results in Table II. The magnitudes of the differences are for these values are also nearly the same with high-tech BHARs increasing with 7.95%, 8.53% and 3.48%. Using the standard deviation adjusted portfolios the effect becomes significantly negative for both equal- and value-weighted mean BHARs. This also corresponds to the results in Table II.

The financing variables indicates M&A transactions that are 100% funded with cash. The coefficient is highly significant in almost every regression except (11) and (12) where its only significant at the 10% level. With certainty and in line with previous research (Frank, Harris and Mayer, 1988; Gregory, 1997) can be concluded that cash offers indeed perform better post-merger.

(22)

22

Table III

Buy-and-Hold Abnormal Returns Regressions

The regression in all columns is: 𝐵𝐻𝐴𝑅𝑖𝑇= 𝛽1+ 𝛽2𝐻𝐼𝐺𝐻𝑇𝐸𝐶𝐻𝑖+ 𝛽3𝐹𝐼𝑁𝐴𝑁𝐶𝐼𝑁𝐺𝑖+ 𝛽4𝑂𝐹𝐹𝐸𝑅𝑖+ 𝛽5𝑄𝑖𝑡+ 𝛽6𝑃𝑈𝐵𝐿𝐼𝐶𝑖+ +𝛽7𝑆𝐼𝑍𝐸𝑖𝑡 + 𝜀𝑖𝑡.The dependent variable is the equal-weighted BHAR of

acquiring firms in columns (1) through (6). In columns (7) through (12) the dependent variable is the value-weighted BHAR of acquiring firms. Each column represents the BHAR against a different benchmark portfolio indicated in the second row. Market cap, beta and standard are the benchmark portfolios based on market capitalization, beta and standard deviation deciles respectively. High-tech is a dummy variable indicating a deal with a high-tech acquirer and target. Financing is a dummy variable indicating whether the deal is financed with cash only. Size is the natural logarithm of the acquirer’s total assets. Tobin Q measure the market capitalization over book value of assets, where market capitalization is defined as common shares outstanding times fiscal year closing price. Offer is a dummy variable indicating if the M&A deal is a tender offer. Public is a dummy variable indicating whether the target is a public company. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

Dependent

Variable: Equal-Weighted BHAR Value-Weighted BHAR

Benchmark Portfolio: Value-weighted CRSP Equally-weighted

CRSP S&P 500 Market cap Beta

Standard deviation Value-weighted CRSP Equally-weighted

CRSP S&P 500 Market cap Beta

Standard deviation (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) High-tech 0.0795*** -0.0027 0.0853*** 0.0348* 0.0067 -0.213*** 0.0387 0.0102 0.0397 0.0415 -0.0348 -0.106** (0.0208) (0.0209) (0.0208) (0.0207) (0.0260) (0.0751) (0.0302) (0.0400) (0.0295) (0.0298) (0.0473) (0.0463) Financing 0.141*** 0.130*** 0.152*** 0.127*** 0.0530** 0.119** 0.0108*** 0.174*** 0.125*** 0.101*** 0.0814* 0.0746* (0.0201) (0.0200) (0.0202) (0.0201) (0.0213) (0.0493) (0.0336) (0.0427) (0.0328) (0.0334) (0.0445) (0.0435) Size 0.0499*** 0.0278*** 0.0516*** 0.0474*** 0.0288*** 0.210*** -0.0412*** -0.0932*** -0.0375*** -0.0325*** -0.0520** -0.0487*** (0.0043) (0.0043) (0.0043) (0.0043) (0.0064) (0.0096) (0.0096) (0.0122) (0.0094) (0.0096) (0.0172) (0.0169) Tobin Q 0.0065* 0.0001 0.0050 0.0067* -0.0004 0.0502*** -0.0113*** -0.0235*** -0.0121*** -0.0102*** -0.0210*** -0.0385*** (0.0035) (0.0035) (0.0034) (0.0036) (0.0061) (0.0185) (0.0014) (0.0027) (0.0014) (0.0013) (0.0055) (0.0053) Offer 0.0178 0.0178 0.0205 0.0168 -0.0309 -0.0522 -0.121** -0.155** -0.132** -0.121** -0.122* -0.235*** (0.0389) (0.0390) (0.0393) (0.0395) (0.0416) (0.101) (0.0537) (0.0703) (0.0520) (0.0527) (0.0725) (0.0792) Public 0.0533** 0.104*** 0.0538*** 0.0626*** 0.0210 0.0124 0.110*** 0.126*** 0.101*** 0.106*** 0.083* 0.111** (0.0212) (0.0213) (0.0214) (0.0213) (0.0226) (0.0557) (0.0319) (0.0420) (0.0313) (0.0315) (0.0459) (0.0443) Constant -0.505*** -0.380*** -0.452*** -0.477*** -0.485*** -1.902*** 0.337*** 0.679*** 0.371*** 0.254*** 0.227 0.354** (0.0306) (0.0306) (0.0307) (0.0306) (0.0502) (0.0202) (0.0830) (0.104) (0.0813) (0.0827) (0.152) (0.148) Obersvations 23455 23455 23455 23061 10136 9940 23454 23454 23454 23060 10136 9940 R-squared 0.010 0.006 0.011 0.009 0.003 0.024 0.019 0.054 0.021 0.016 0.020 0.043

(23)

23

The size of the acquirer does have a significant positive effect on the BHARs when equal-weighted but when using the value-equal-weighted BHARs this effect completely reverses. The all highly significant negative effects range from -3.25% to -9.32% for a 1% increase in acquirer size for the value-weighted BHARs. Smaller stocks are more likely subject to mispricing and with equal weights these stocks are weighted to much leading to biased estimates. Assuming the value-weight is the better approach, it can be concluded that bigger acquirers earn worse abnormal returns in the long-term. This results is in accordance with of Moeller, Schlingemann and Stulz (2004), who show that large acquirers perform worse than small acquirers.

According to the value-weighted BHARs regressions high valued firms earn significantly lower abnormal returns for the three years post-merger. This matches (Rau and Vermaelen, 1998) that “glamour firms” tend to make poor acquisitions. However the magnitude is really small ranging from 1.02% to 3.85% decrease in BHAR for one unit increase in Tobin’s Q.

Against the results of existing literature tender offers tend to have a significant negative effect ranging from -12.1% to -23.5% on the value-weighted BHARs. When examining the data, it seems that a lot of the bad performing tender offers are actually acquired by large sized firms. This gives the negative outcomes more weight, resulting in negative coefficients. For equal-weights this effect disappears as seen in regressions (1)-(6).

When acquiring a publicly owned firm the BHARs significantly increase in ten out of the twelve regressions with magnitudes of around 10% on the value-weighted and around 6% on the equal-weighted BHARs.

6.3 Calendar-Time Portfolio Regressions

The BHAR approach is subject to bad-model problems which lead to biased results. They can exhibit cross-correlations explained in the methodology section. Using the calendar-time portfolio approach the time-series variation of the monthly abnormal returns on a portfolio accurately capture the effects of cross-correlation and makes it robust to these concerns. The results of the calendar-portfolio regressions are reported in Table IV. Every equal- and value-weighted calendar-portfolio of acquirers is regressed on the Fama and French three-factor model and the four-factor model including the Carhart (1997) momentum factor.

Regression (1) to (6) show the results of the equal-weighted portfolios. For the total sample, regression (1) and (2), the average monthly excess return is not significantly different from zero. Hence according to this method the efficient market hypotheses cannot be rejected for M&A in general. Regressing the equal-weighted high-tech portfolio the alpha is insignificant for the three-factor model, but significant at the 1% level when including the momentum three-factor. With a monthly excess return of 0.36%. This might seem small in economic significance, but does really add up on the

(24)

24

Table IV

Calendar-Time Portfolio Regressions

This table shows three- and four- factor calendar-time portfolio regressions. Each column shows a regression on a different portfolio, where the portfolio is indicated in the second row. A high-tech acquirer portfolio consists of high-tech acquirers that acquire a high-tech target. Each portfolio is regressed on a three- and four- factor model. In each calendar month from January 1981 until December 2009 a portfolio is formed consisting of firms that acquired a target in the last three years. Portfolios are rebalanced each month. For the regressions (1) to (6) the portfolios are formed based on an equal weight for each stock. For the regressions (7) to (12) these portfolios are value-weighted based on their market capitalization. The dependent variable is the monthly excess return of a portfolio on the one month treasury bill rate. The explanatory variables are the excess market return (MktMRF), the returns from Fama and French (1993) mimicking portfolios (SMB and HML) and, in columns (2), (4), (6), (8), (10) and (12) the Carhart (1997) momentum factor (MOM). Alpha measures the monthly excess return of the portfolio. Standard errors are reported in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

Equal-Weighted Portfolio Value-Weighted Portfolio

Portfolio: All acquirers All acquirers High-tech acquirers High-tech acquirers Non-high-tech acquirers Non-high-tech acquirers All acquirers All acquirers High-tech acquirers High-tech acquirers Non-high-tech acquirers Non-high-tech acquirers (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) MktMRF 1.024*** 0.956*** 1.107*** 1.033*** 0.985*** 0.928*** 1.013*** 0.957*** 1.002*** 0.938*** 0.978*** 0.942*** (0.0214) (0.0172) (0.0318) (0.0293) (0.0229) (0.0205) (0.0276) (0.0261) (0.0462) (0.0461) (0.0183) (0.0174) SMB 0.614*** 0.624*** 0.771*** 0.784*** 0.545*** 0.552*** -0.0316 -0.0240 -0.0177 -0.0064 -0.0454* -0.0404* (0.0304) (0.0236) (0.0454) (0.0404) (0.0325) (0.0281) (0.0392) (0.0359) (0.0660) (0.0636) (0.0261) (0.0239) HML 0.0455 -0.0504* -0.410*** -0.510*** 0.255*** 0.173*** -0.282*** -0.362*** -0.484*** -0.571*** 0.0514* -0.0010 (0.0329) (0.0263) (0.0491) (0.0449) (0.0352) (0.0313) (0.0424) (0.0399) (0.0715) (0.0707) (0.0282) (0.0266) MOM -0.237*** -0.257*** -0.203*** -0.198*** -0.223*** -0.130*** (0.0158) (0.0269) (0.0188) (0.0239) (0.0423) (0.0159) Constant -0.0014 0.0008 0.0012 0.0036*** -0.0024** -0.0006 0.0027** 0.0045*** 0.0045** 0.0066*** 0.0007 0.0010** (0.0009) (0.0007) (0.0014) (0.0012) (0.0010) (0.0009) (0.0012) (0.0011) (0.0020) (0.0020) (0.0008) (0.0008) Observations 348 348 345 345 348 348 348 348 345 345 348 348 R-squared 0.908 0.944 0.874 0.901 0.876 0.908 0.836 0.864 0.672 0.696 0.901 0.917

(25)

25

long term. Non-high-tech acquirers have a significant negative abnormal performance of -0.24%, as seen in regression (3). The value is insignificant using the four-factor model though. Hence the there is a hint the high-tech portfolio outperforms the non-high-tech portfolio.

Using value-weighted portfolios the post-performance of M&A looks really good. The portfolio of all event firms have a significant and positive monthly excess return of 0.27% and 0.45% for the three-factor and four-factor model respectively. This suggest that M&A overall does better than expected on the long-term. For high-tech portfolios the excess returns are even higher with values of 0.45% and 0.66%, both significant. This argues that high-tech acquirers outperform the average acquiring firm post-M&A, which becomes apparent in regression (11) and (12). They do not earn significant excess returns in (11) but do in (12), but only 0.10%.

Both methodologies of equal- and value-weighted portfolio show that overall M&A does not have negative abnormal returns on a long-term horizon of three years. It even does better than expected using value-weighted portfolios. It looks like high-tech M&A outperforms the average M&A post-merger. In the next section this effect is investigated more profoundly.

6.4 Long-Short Calendar-Time Portfolio Regressions

To test the third hypotheses a long-short portfolio is formed consisting of a long portfolio with high-tech firms that acquired a high-high-tech target in the past three years and a short portfolio of acquirers in non-high-tech M&A deals. The portfolios are again equal- and value-weighted. The returns of these portfolios are then regressed on the Fama and French three-factor model in regression (1) and (3) reported in Table V. Regression (2) and (4) also include the Carhart (1997) momentum factor.

The monthly excess returns on the equal-weighted long-short portfolio are 3.51% and 4.04%, significant at the five and one percent level respectively. For the value-weighted long-short portfolio the value is 3.82%, significant at the 10% level, when regressed on the three-factor model. When regressing the portfolio on the four-factor model the result is 4.71% and significant at the 5% level.

These results provide statistical evidence that high-tech M&A does outperform M&A in other industries on the long-term horizon of three years. The monthly excess returns in TABLE 4 suggest that the abnormal performance from the long-short portfolio is mainly driven by the high abnormal returns for high-tech M&A. Mergers in other industries do not expressly have negative abnormal returns but are on average statistically insignificant. This means that high-tech M&A actually does better than expected in the first place and the long-term M&A performance anomaly is largely due to statistical errors and methodological issues.

(26)

26

Table V

Long-Short Calendar-Time Portfolio Regressions

This table shows three- and four- factor calendar-time portfolio regressions. In each calendar month from January 1981 until December 2009 a portfolio is created based on a long portfolio in stocks of all high-tech acquirers and a short portfolio containing stocks of all non-high-tech acquirers that have completed an M&A deal in the previous three years. The stocks in the portfolio are equally-weighted in regression (1) and (2), and value-weighted in regression (3) and (4). Portfolios are rebalanced each month. The dependent variable is the monthly return of the long-short portfolio. The explanatory variables are the excess market return (MktMRF), the returns from Fama and French (1993) mimicking portfolios (SMB and HML) in columns (1) and (3) and, in columns (2) and (4) the Carhart (1997) momentum factor (MOM). Alpha measures the monthly excess return of the long-short strategy. Standard errors are reported in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

Portfolio: Equally weighted Equally weighted Value weighted Value weighted

(1) (2) (3) (4) MktMRF 0.116*** 0.0994*** 0.0185 -0.00950 (0.0332) (0.0342) (0.0498) (0.0514) SMB 0.226*** 0.228*** 0.0342 0.0381 (0.0472) (0.0470) (0.0709) (0.0706) HML -0.659*** -0.683*** -0.523*** -0.563*** (0.0510) (0.0524) (0.0766) (0.0786) MOM -0.0581* -0.0992** (0.0313) (0.0471) Alpha 0.00351** 0.00404*** 0.00382* 0.00471** (0.00145) (0.00148) (0.00219) (0.00222) Observations 348 348 348 348 R-squared 0.495 0.500 0.154 0.165

7. Conclusion

This study empirically analyzes the long-run abnormal returns of US acquiring firms. A distinction is made between high-tech M&A and non-high-tech M&A to check whether high-tech acquiring firms perform better. A comprehensive sample of 24,332 M&A events are used over the period of 1 January 1981 until 31 December 2006. A sample size that has not been studied before. For the analysis two different approaches are used. First the buy-and-hold abnormal returns approach is applied where the equal- and value-weighted mean BHARs are calculated with six different benchmarks. The outperformance of high-tech M&A is checked with a regression and control variables. Secondly, the calendar-time portfolio approach is conducted, by forming a monthly time series of acquiring firms portfolio returns. A long-short portfolio is created to check whether high-tech M&A performs better.

The BHAR approach gave a lot of different results. It is apparent from these results that choosing the right benchmark is very important. Also equal- and value-weighted mean BHARs yield very different results. Size- and value-adjusted benchmarks are the most common used benchmarks in previous research they also seem to yield the economic most reliable results. Equal-weighted mean BHARs are approximately -11% for a three year time period for the whole sample which is somewhat of the same magnitude as the results of Anderson and Mandelker (1993) and Gregory (1997). For the high-tech subsample these BHARs are significantly smaller than the other acquiring firms. High-tech acquiring firms

Long-term post-M&A performance in US high-tech industries