The effect of target age on the acquirers’ abnormal returns in the U.S. tech industry

UNIVERSITY OF AMSTERDAM AMSTERDAM SCHOOL OF ECONOMICS BSc Economics and Business Specialization Finance & Organization

THE EFFECT OF TARGET AGE ON THE ACQUIRERS’

ABNORMAL RETURNS IN THE U.S. TECH INDUSTRY

Author: T. van der Put Student number: 6096239 Thesis supervisor: J. J. G. Lemmen Finish date: June 2016

Statement of Originality This document is written by Timon van der Put 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.

Abstract

This thesis studies the bidding firms’ cumulative abnormal returns around the announcement date of mergers and acquisitions (M&A) in the technology sector, with the focus on target age. The technology sector can be regarded as one of the biggest sectors in the M&A market. Demand for external business expansion by acquiring unique products and technologies, is being met by the supply of various startup companies aiming for maximum profit, resulting in a takeover. However, startup acquisition doesn’t come without it’s risks. Insufficient information on young target companies can lead to uncertainty. On the other hand, unexploited synergies can result in profits. To get more insight on these speculations based on theory from previous research, the following hypotheses will be tested. The acquirers’ CARs around tech acquisitions are different from zero (H1), target age has a significant negative effect on the acquirers’ CARs (H2) and CAR variation is larger when acquiring younger targets and decreases when the target firm age increases (H3). An event study on 385 observations of acquisitions in the period between 1987 through 2015 on the 13 biggest U.S. technology companies, followed by a cross-sectional regression analysis, gives the subsequent results. CARs of the acquirer around acquisition announcements with an event window of three days (-1, 0, 1) are insignificant and not different from zero. Secondly, target age has a significant negative effect on the acquirers’ CARs, pointing towards profit opportunities in the form of potential synergy exploitation when acquiring young targets. The final conclusion shows a decreasing CAR variance with the increase of target age, which confirms that acquiring startups brings uncertainty and extra risk.

Table of Contents

ABSTRACT………...………3 TABLE OF CONTENTS………..4 LIST OF TABLES AND FIGURES………..5 CHAPTER 1 Introduction………6 CHAPTER 2 Related literature and hypotheses development………..………..8 2.1 Valuation theory involving younger targets………...8 2.2 Empirical evidence on CARs around M&A announcements……….8 2.3 Method of payment……….10 2.4 Target age and the valuation theory………10 CHAPTER 3 Data description and methodology………...11 3.1 Sample and data sources……….11 3.2 Dependent variable……….13 3.3 Target age and control variables………..………..14 3.4 Final data composition……….…16 CHAPTER 4 Results……….…16 4.1 Descriptive statistics and correlation matrix………..16 4.2 Cumulative abnormal returns………..18 4.3 Cross-sectional regression analysis………..19 4.4 Variance differences in cumulative abnormal returns………..21 CHAPTER 5 Conclusion……….22 REFERENCES………..24 APPENDIX 1……….…26 APPENDIX 2……….27 APPENDIX 3………...28 APPENDIX 4……….28

List of Figures and Tables

Table 1 Previous empirical results on abnormal returns 9 around M&A announcements Table 2 Description of the independent variables in the 15 regression model Table 3 CAAR per bidding firm for two different event 18 windows Table 4 Regression results for the three-day (1) and 20 five-day (2) event window Table 5 Variance testing per target age category 21

1.

Introduction

The high-tech industry is a fast changing, dynamic environment. It can be defined in the following four characteristics. First, products are subject to very rapid technological change, which requires quick innovation. Fast growth in market demand, creating high customer expectations is a second characteristic. Followed by a heavy dependence on scientific research and development to fend off competitors. Competition is intensive in the technology sector, therefore technological leadership is essential to maintain high sales and profits. Lastly, corporate restructuring, involving expansion in both smaller and big technology companies, makes a fourth important feature (Keeble, 1990). Acquisition of external technologies is a good solution to keep up with the high demands set by the technology industry. It can be an alternative for doing your own research and development (R&D), which is costly and time consuming. Rigby and Zook (2002) state that executives admit that the best innovation idea’s aren’t always coming out of their own R&D. By importing ideas from the outside, companies can collect more and better ideas from different kind of experts. These acquisitions conceive motivational incentives, both for buyers and sellers. Buyers gain strategically valuable resources in the form of new products and technologies, where the seller receives necessary (financial) resources without risking the uncertainty of an IPO. Second to that, the buyer enhances its market power and achieves strategic renewal, while the seller relieves personal pressures (Eisenhardt et al., 2010). This creates a market for mergers and acquisitions (M&A). In March 2016, the global technology M&A year to date (YTD) volume has already surpassed last years volume with $71.4 billion through 1,535 deals versus $46.6 billion in 2015 YTD. Accounting for 12% in global M&A volume and 26% of activity (Dealogic, 2016), the technology sector can be regarded as one of the largest sectors in the M&A market. The high demand of M&A in the technology sector, leads to the rise of small startup companies, as it is lucrative with the potential ahead of selling your startup for the millions of dollars that are being offered by big tech companies. Silicon Valley Bank reports in their 2016 Innovation Economy Outlook the future plans of startups in the U.S. The survey

points out that entrepeneurs rather choose being acquired (56%) as a long-term goal, instead of aiming for an IPO (17%) or staying private (19%). However, the startup acquisition strategy doesn’t come without its risks. According to Peemöller et al. (2001), company valuation tends to be difficult for young companies, given the lack of historical data and uncertainty about other development elements. How do executing managers value these startups, or is it a big gamble? Possibly the deal value is too high, or the manager makes a bad decision due to lack of information about the target. Should the acquiring party wait a few more years, until more information is available on the target company, or should they handle quickly before competition grabs away an unique key technology opportunity? To do research on this matter, acquisition perfomance can be measured by evaluating the stock returns, in the form of cumulative abnormal returns (CARs) of the acquiring company around the acquisition announcement date by using the methodology of an event study. Comparing these CARs with the targets’ age and other control variables by doing a cross-sectional regression analysis, will give a deeper insight in stock behaviour around acquisition announcements. Hence, the research question in this thesis. What is the effect of target age on the acquirers’ abnormal returns in the technology industry? For this study, data is collected from the largest U.S. technology companies, based on market capitalization. All known acquisitions by these companies will be evaluated individually by looking at available data, instead of picking random acquisitions made by random technology companies. This method has been chosen to get specific information and insights on individual acquisition strategies for each company, as well as overall results. The remainder of this thesis continues with a literature review, where hypotheses will be formed based on previous studies. Followed by a description of the sample data and methodology used for setting up the model. Finally, the results will be displayed and discussed with a brief conclusion summarizing the main thesis findings. Additional information and references can be found in the appendix and references section, disclosed at the end of the paper.

2.

Related Literature and Hypotheses Development

2.1 Valuation theory involving younger targets Previous research on stock return behaviour around M&A has been done earlier, both in theoretical and empirical studies. Before heading into detailed research on the different factors that potentially influence the bidding firm’s CARs, it is important to start with a more broader view. What effect do M&A have in general on the acquirers’ stock returns? With his theoretical study about returns to bidding firms in mergers and acquisitions, Barney (1988) focuses on the relatedness hypothesis. The relatedness hypothesis suggests that the economic value of the bidding firm can be increased by merging or acquiring strategically related firms (Salter and Weinhold, 1979). Barney (1988) states that positive abnormal returns are generated by synergetic cash flows between the bidding and target firm, luck and managerial decisions to keep information about valuable synergetic cash flows private. Opposed to that, mergers and acquisitions are more likely to result in negative abnormal returns for the acquirer, due to target firms are being overvalued by underestimating the costs of exploiting the target’s synergies. Combining these findings with the theory of Peemöller et al. (2001), where young target firms tend to be difficult to valuate due to the lack of historical data and development uncertainty, it is likely that the acquisition of young targets will lead to negative abnormal returns. However, when target age increases and more information becomes available, there is no guarantee for positive abnormal returns, considering Barney’s (1988) requirements. 2.2 Empirical evidence on CARs around M&A announcements Empirical evidence on the acquirers’ stock return around M&A can be found in several studies. Results tend to have different outcomes. Papers from Andrade et al. (2001), Smith and Ward (2007) and Wong and Cheung (2009) conclude no significant proof for abnormal returns around announcement dates. While Fuller et al. (2002), Loderer and Martin (1990), Moeller et al. (2004) and Ma et al. (2009) found significant positive CARs (Table 1). When comparing these empirical results with the theoretical studies from Barney (1988) and Salter and

Weinhold (1979), there seems to be some contradiction. Abnormal returns are expected to be negative because bidding firms typically overvalue target firms and underestimate synergy costs. However, the empirical results do not correspond with this theory, showing neutral to positive returns. Table 1 Previous empirical results on abnormal returns around M&A announcements

Study Period Region Sample Event Estimation Acquirer

Insignificant results size window window CAAR

Andrade et al. (2001) 1973-1998 U.S. 3.688 (-1,1) n.m. -0,70% Smit and Ward (2007) 2000-2002 S. Africa 27 (-1,1) n.m. -0,02% Wong and Cheung (2009) 1990 Asia 658 (-1,0) (-300, -51) -0,38% Significant results Fuller et al. (2002) 1990-2000 U.S. 3.135 (-2,2) n.m. +1,77%*** Loderer and Martin (1990) 1966-1984 U.S. 5.172 (-5,0) (-300, -100) +0,69%*** Ma et al. (2009) 2000-2005 Asia 1.477 (-1,1) (-125, -6) +1,28%*** Moeller et al. (2004) 1980-2001 U.S. 12.023 (-1,1) (-205, -6) +1,10%*** *** Significant at 1% n.m = not mentioned It is obvious that there are divided conclusions on CAR behaviour around M&A announcement dates. There is no single right answer on whether the outcome will be positive, negative or neutral. However, previous research has mostly evolved around overall sample data. It’s interesting to see whether abnormal returns behave different at a smaller scale, for example per individual firm. Does the unique acquisition strategy or acquisition experience and knowledge have any effect on different outcomes? CARs are difficult to predict and there are many dependent factors with a significant role in CAR prediction around M&A. Based on previous empirical studies and the contradiction with theoretical studies, we can make the following prediction: Hypothesis 1 (H1): Acquirers’ CARs around tech acquisitions are not different from zero. On firm level they might vary due to different acquisition strategies.

2.3 Method of payment One of the factors to explain CAR behaviour is the method of payment, which can be in cash, stock exchange or a combination of the two. Travlos (1987) distinguishes these two types of payment in his empirical research in the period 1972 through 1981. The sample contains 167 acquiring firms, composed out of sixty stock exchange offers and one hundred cash exchange offers. The results point out that stock exchange offers have significant (at 1% significance level) negative abnormal returns of -0.78% and -0.69% on (t=-1) and (t=0) respectively. Cash offers show insignificant abnormal returns of -0.05% and +0.29% on the same event days. These results are explained by the signalling theory, where the exchange using common stock hides the negative information that the bidding firm is overvalued (Leland and Pyle, 1977). Although it would be interesting to add the method of payment as control variable in the regression, there is only little information to be found on acquisition details, such as methods of payment in the tech industry M&A research field. Therefore it has not been included in this research. 2.4 Target age and the valuation theory The effect of target age on the acquirers’ return, has not been researched frequently in the past. There is one relevant study to be found on the relation between the acquirers’ abnormal returns and the targets’ age. Ransbotham and Mitra (2010) evaluated 249 acquisition announcements in their empirical research on the telecommunication industry through the years 1995 and 2001. Using market model returns with an event window of 1 day (day 0), they found a significant (1%) negative relation between the target age and the bidders’ CARs with a β-coefficient of -3.579. One of their main explanations is the unexplored growth potential in younger targets, which can be fully exploited in combination with the knowledge and resources that the acquirer can provide and weren’t at the targets’ disposal at first. A second statement mentions the valuation uncertainty of young targets, leading to lower bids and a selling price below the actual value of the target. This again corresponds to Peemöller et al. (2001) and Barney (1988), the theory of valuation difficulty for younger targets. We therefore make the following prediction:

Hypothesis 2 (H2): Target age has a significant negative effect on the acquirers’ CARs. One major distinction that has to be emphasized, is that Ransbotham and Mitra (2010) assume undervaluation by bidding firms, where Barney (1988) mentions overvaluation. This thesis can contribute to this contradiction, by doing additional research in the effect of the target age on the acquirers’ CARs around acquisition announcements. Misvaluation can be interpreted in two ways. This doesn’t automatically mean that all targets are being undervalued or the other way around. What would seem more natural, is that acquisitions can turn out in two ways, leading to eventually positive CARs in the case of undervaluation and negative CARs when overvalued. There might be no relation at all to be found between target age and CARs. However, the valuation uncertainty involving younger targets, should lead to a higher spread in CARs when target age decreases, due to the higher risk involved. When target age increases, the CAR variance should diminish according to theory, leading to the final prediction: Hypothesis 3 (H3): CAR variation is larger when acquiring younger targets and decreases when the target firm age increases. This can be tested by doing one-tailed F-tests for the equality of CAR variances in different target age intervals.

3.

Data Description and Methodology

3.1 Sample and data sources The sample covers all known acquisitions by the biggest U.S. technology companies, selected by market capitalization size and revenues. These companies are on alphabetical order, Amazon, Apple, Cisco, Facebook, Google, Hewlett Packard, IBM, Intel, Microsoft, Oracle, Salesforce, Twitter and Yahoo. The reason behind selecting individual firms, rather than random acquisitions by

random firms, is to get an insight in individual acquisition strategies. Do these individual companies select on target age and does this relate to their own specific abnormal returns? This also gives information on the availability of data. How many acquisitions have been made by each firm in their existence? Are the transaction details disclosed for each takeover, or has the majority of these deals been done under undisclosed details? By analysing takeovers individually, essential additional individual information becomes available. The final sample consists out of 385 observations of acquisitions in the period between 1987 through 2015 on the 13 biggest U.S. technology companies selected on market capitalization and revenues (Appendix 1). To obtain the final dataset, the following steps have been made. The first step was to conduct all known acquisitions made by the individual firms in the past. Both Crunchbase and Zephyr have been consulted. By using two different databases, missing observations can be complemented by the other database to rule out incompleteness. Before collecting more additional data for the abnormal returns and potential control variables, it was crucial to get the correct announcement dates for each observation. The first problem arose when Crunchbase and Zephyr maintained different announcement dates for some of the observations. This can be explained by information leakage prior to the event or inaccurateness of the databases. Mostly, the announcement dates differed from each other by only one or two days, which can make a big difference when doing an event study for the abnormal returns. To solve this, each observation has been checked manually to correspond with the correct announcement date by observing official statements from bidding firms. The same has been done for the accompanying acquisition values. These tend to differ due to different sources proclaiming different amounts. In this case, each transaction value has been observed manually according to the most reliable source selected by own arbitrage. Most of the times this could be confirmed afterwards by multiple reliable sources announcing the same values. Getting all the acquisitions with the correct announcement dates and values in order, formed a solid base which was required to get all additional information such as the dependent variable, being cumulated abnormal returns, the independent variable target age and other control variables such as acquirer,

target and deal characteristics. A description of the independent variables used for the regression-model, which will be discussed later, and the sources where they are collected from can be found in table 2. 3.2 Dependent variable To measure the abnormal returns for the bidding firms around the acquisition announcement, the methodology of an event study has been used (De Jong, 2007, MacKinlay, 1997). The first step in this process is to identify the event, which has been done in the previous part by gathering all corresponding acquisition announcements for each observation. The reason for choosing announcement dates over the actual takeover date is explained by the fact that the stock price has already got the opportunity to adapt to the potential value changes in the bidding firm. The announcement date is the first time the public gets informed, rumours excluded. The second step is identifying the abnormal returns by choosing a benchmark model. In this study the market model will be used, to prevent market biased results. Abnormal returns (AR) are defined by the following equation: 𝐴𝑅!" = 𝑅!"− 𝑁𝑅!" Where R is the bidder’s stock return and NR the normal return or benchmark return. The benchmark is an OLS estimation of the market model: 𝑅_!" =

α

! + 𝛽!𝑅!" + 𝜖!" à 𝑁𝑅!" = 𝛼 + 𝛽𝑅!" Where 𝑅!" is the return on a market index. For this study The Center for Research in Security Prices (CRSP) US total market index has been used for setting the benchmark. Unfortunately, the effect of an event is not always displayed in only one day, being the announcement day. In perfect conditions, the acquirer announces the event in the morning just before the market opens. This gives stock prices the opportunity to adapt valuation to the announcement, finished by market closure, resulting in an fully adjusted opening stock price the next day. In perfect

conditions only one day is needed to determine abnormal returns. However, rumours and data leakage require inclusion of the days prior to the announcement. If the announcement has been made after market closure, the next day is needed to include in your event window due to delayed information. Therefore an event window is constructed around the announcement date (t=0). Most event studies in research on M&A use an event window of three days, being (-1, 0, 1). To measure abnormal returns in an event window, covering multiple days, cumulative abnormal returns are used. Where: 𝐶𝐴𝑅_! = 𝐴𝑅_!" !!!! !!!! To obtain all CARs for each announcement date, the tool Eventus has been used from Wharton Research Data Services. The estimation window is set at 250 days, including a gap of 10 days prior to the announcement (-260, -10), based on earlier studies and an example set by MacKinlay (1997). The traditional three day event window of (-1, 0, +1) is set for the first series of CARs. A second event window of (-2, 0, +2) has been opted based on the CAR development graph over multiple days (Appendix 3) and an additional check for robustness. 3.3 Target age and control variables The dependent variable, explained in the previous part, will be regressed against the target age and control variables by doing a cross-sectional regression. The regression equation is as following: 𝐶𝐴𝑅_! = 𝛽_!+ 𝛽_!𝐴𝑔𝑒_! + 𝛽_!𝑆𝑖𝑧𝑒_! + 𝛽_!𝑅&𝐷_!+ 𝛽_!𝐸𝑥𝑝_! + 𝛽_!𝑉𝑎𝑙𝑢𝑒_! + 𝛽_!𝐶𝑟𝑜𝑠𝑠𝑏_! + 𝛽!𝐷𝑜𝑡𝑐𝑜𝑚! + 𝛽!𝐶𝑟𝑖𝑠𝑖𝑠! + 𝜀 Where 𝛽_! is a constant and 𝜀 is the error term. As mentioned earlier, the control variables and their description including data source can be found in table 2.

Table 2 Description of the independent variables in the regression model

Variable Definition Data source

Target characteristics:

Target age Acquisition year minus target Crunchbase, Bloomberg,

founding year news articles,

Bidder characteristics:

official announcements

Acquirer size (log assets) Log of book value of Compustat

total assets

R&D Intensity Quarterly R&D spending Compustat

divided by sales

Acquisition experience Number of acquisitions Crunchbase, Zephyr made previous year Deal characteristics: Acquisition value Deal value divided by the Crunchbase, Zephyr, Compustat, acquirers' market capitalization news articles, at the announcement date official announcements Cross-border dummy Dummy variable: 1 if transaction Crunchbase, Zephyr, Bloomberg, is cross-border, 0 if domestic news articles, official announcements Macro characteristics: Dotcom bubble dummy Dummy variable: 1 if acquisition Based on announcement date (1999-2000) is between years 1999-2000, 0 if not

Crisis years dummy Dummy variable: 1 if acquisition Based on announcement date (2007-2008) is between years 2007-2008, 0 if not. For the second hypothesis, which states that target age has a negative effect on the cumulative returns of the acquirer, the dependent variable target age has been admitted to the regression equation. Target age is defined as the year of announcement minus the target founding year. The target founding year is recorded in the Crunchbase database, but not for most observations. Missing values had to be added manually to the dataset, by searching in news articles, official announcements by the bidding firm and other sources like Bloomberg

and various financial information sources. This part took up most of the time when collecting all the sample data. 3.4 Final data composition The last step for finalizing the sample dataset is to rule out and remove observations. Observations with undisclosed vital information, like transaction value or target age, have been removed from the dataset. Overlapping acquisitions have also been removed. When two announcement days for the same bidding firm interfere with each other by crossing the 3 and 5 day event windows, the cumulative abnormal returns will become biased. Therefore all observations having less than 5 days in between them, by the same acquirer, have been removed, non-trading days exluded. Finally, extreme outliers and possible erroneous observations have been removed to prevent the regression parameters being influenced by them. Least squares estimates are very sensitive to outliers, especially in case of small samples. Appendix 1 shows the data coverage percentages, with an overall coverage of 27.4%. This means that out of all acquisitions made by the top 13 technology firms, only 27.4% of detail information is actually available. Surely, some removed observations are caused by overlapping acquisition data and extreme outliers, but this is only a small fraction. The most common reason for incomplete information turned out to be missing transaction values. Looking at individual coverage percentages shows high scores for Intel and Cisco, providing information on almost half of their acquisitions with 44.6% and 42.9%. Facebook, Twitter and Hewlett Packard show low percentages with 11.7%, 12.5% and 13.1% respectively.

4.

Results

4.1 Descriptive statistics and correlation matrix Appendix 2 provides a descriptive statistics overview expanded with a correlation matrix for every variable in the regression. In general, the average target age is 8.16 years, which is relatively young and meets the startup

definition with a maximum of 12 years old. When looking at individual average target age for each specific firm, we get different results. Appendix 1 includes a column average target age for every individual bidding firm. Facebook and Twitter have the lowest average target age, with 3.7 and 3.3 years. Suprisingly, these are the youngest two firms, both having their first acquisition respectively on April 25, 2013 and April 15, 2014 in the sample. Microsoft is the only firm who comes close to the overall average of 8.2 years, with an average of 7.8 years. For the other firms we can distinguish average target ages between 4 and 8 years, which is still relatively young. Except for Hewlett Packard, IBM, Intel and Oracle, whose average target age lies between 11 and 16 years until acquisition. The differences in average target age for each company, tend towards individual acquisition strategies amongst the sample firms, which has been suggested at the start of part 3.1. When looking at the target value, the overall average value is 0.86% of the bidder’s market capitalization. Again, when looking at individual results, a clear difference can be noticed between the companies, pointing out towards different acquisition strategies. Microsoft, Twitter and Google seem to play it safe, with transactions of only 0.28%, 0.27% and 0.17% of total market capitalization. Hewlett Packard (HP) has the highest average target value, with 3.31% of their market cap. Noticable is that HP also has a high average target age. This might suggest an increased average spending value on older targets. Oracle is a second firm with a high average target age combining a relatively high target value of 1.84%. The correlation coefficient of target age and acquisition value is 0.239, which indicates a weak positive linear relationship between the two, making older target firms seemingly to be a little more expensive than young targets. Acquirer size also has a slight positive correlation with target age, with a correlation coefficient of 0.274. Bigger acquirers seem to prefer older targets than their smaller competitors. Contradictionary, the correlation coefficient of acquirer size and acquisition value is slightly negative, being -0.174. While larger firms choose older targets on average, who seem to be more expensive than the younger ones, they still maintain a weak negative linear relationship with their overall acquisition value.

Furthermore, R&D expenditure is 11.12% on average of sales, having slightly negative correlation coefficients with target age and acquirer size (-0.141 and -0.126). The average acquisition experience is 8.6 takeovers in the previous year and a weak to moderate positive linear relationship with acquirer size. This seems logical as larger bidders make more acquisitions in a year than their smaller competitors. Out of all observations, 10.4% was during the dotcom bubble in the years 1999 through 2000 and 12.2% during the 2007 and 2008 financial crisis. One quarter of all transactions in the sample were international, where the remaining 74.8% concerns domestic takeovers. There are no high correlations between the predicting variables in the regression model, which precludes multicollinearity. 4.2 Cumulative abnormal returns Table 3 shows the cumulative average abnormal returns (CAARs) for each firm in the 3 day event window (-1, 0, +1) and the 5 day event window (-2, 0, +2). Table 3 CAAR per bidding firm for two different event windows Event window (-1, 0,+1) Event window (-2, 0,+2) Bidding firm CAAR Std. dev. t-stat

CAAR Std. dev. t-stat

Amazon -1.05% 0.0383 -1.026 -1.13% 0.0540 -0.780 Apple 0.80% 0.0418 0.953 1.87%** 0.0435 2.147 Cisco -0.32% 0.0260 -1.138 -0.36% 0.0388 -0.849 Facebook 0.39% 0.0563 0.183 -0.47% 0.0554 -0.224 Google 0.25% 0.0203 0.787 0.54% 0.0342 1.010 Hewlett Packard -0.39% 0.0148 -0.950 -0.60% 0.0162 -1.323 IBM 0.30% 0.0176 1.069 -0.26% 0.0278 -0.583 Intel -0.54% 0.0360 -0.866 -1.11% 0.0421 -1.528 Microsoft -0.35% 0.0230 -0.766 -0.81% 0.0340 -1.549 Oracle 0.72% 0.0241 1.662 0.82% 0.0344 1.335 Salesforce -0.12% 0.0430 -0.995 -1.98%* 0.0396 -1.804 Twitter 2.24% 0.0380 1.447 4.04% 0.0477 2.074 Yahoo -0.33% 0.0381 -0.539 0.24% 0.0422 0.344 Total -0.07% 0.0306 -0.458 -0.09% 0.0388 -0.474 * Significant at 10%; **significant at 5%; ***significant at 1%

These results show different outcomes per firm. In the 3 day event window, no significant abnormal returns can be found. Whereas the abnormal returns for the individual companies are both positive and negative, equally divided. Generally, for the whole sample, the CAARs are -0.07% and insignificant. The 5 day event window seems to be the better choice, as CAARs increase or decrease further from the values in the 3 day event window. Going from no significant CAARs to two significant values; +1.87% with 5% significance for Apple and -1.98% with 10% significance for Salesforce. These values are not major though, since the other 11 CAARs remain insignificant and divided. The overall CAAR of -0.09% in the 5 day event window is also insignificant. Because the results are divided and overall not significant, the first Hypothesis (H1) Acquirers’ CARs around tech acquisitions are not different from zero. On firm level they might vary due to different acquisition strategies, stated in part 2.2 will not be rejected. 4.3 Cross-sectional regression analysis Table 4 displays the cross-sectional multiple regression model results for the 3 day event window (-1, 0, +1) (1) and the 5 day event window (-2, 0, +2) (2). The regression is performed with robust standard errors to account for heteroskedasticity and acquisitions made by the same acquirer. The 3 day event window model shows no significant effect of target age on cumulative abnormal returns (𝛽_!= -0.0001). Look at model (2), for the five day event window, the target age coefficient is negative (𝛽_!= -0.0005) at a 5% significance level. Comparing the adjusted R squared values for model (1) and model (2), 0.0189 versus 0.0350, the five day event window model seems to be better at explaining the CARs. With a negative significant coefficient for target age, the predicion based on the valuation difficulty theory by Peemöller et al. (2001) and Barney (1988), leading to lower bids and lower selling prices for younger targets and vice versa for older targets, seems to be correct. Results also correspond with earlier research by Ransbotham and Mitra (2010), who also concluded a significant negative effect of target age on cumulative abnormal returns. Therefore hypothesis 2, Target age has a significant negative effect on the acquirers’ CARs, will not be rejected.

Table 4 Regression results for the three-day (1) and five-day (2) event window (1) (2) Target characteristics: 𝛽_!Target age -0.0001 -0.0005** (0.0002) (0.0003) Bidder characteristics: 𝛽_! Acquirer size (log assets) 0.0024 0.0020 (0.0021) (0.0024) 𝛽_! R&D Intensity 0.0506 0.0657 (0.0345) (0.0418) 𝛽!Acquisition experience -0.0004 -0.0003 (0.0002) (0.0003) Deal characteristics: 𝛽_! Acquisition value 0.0462 0.0355 (0.0827) (0.1098) 𝛽_! Cross-border dummy 0.0005 -0.0009 (0.0036) (0.0046) Macro characteristics: 𝛽_! Dotcom bubble dummy (1999-2000) -0.0127* -0.0178** (0.0073) (0.0083) 𝛽_! Crisis years dummy (2007-2008) 0.0066* 0.0150** (0.0039) (0.0060) 𝛽_! Constant -0.0601 -0.0503 (0.0532) (0.0608) Observations 385 385 Adjusted R-squared 0.0189 0.0350 Significant at 10%; **significant at 5%; ***significant at 1%. Robust standard errors between parantheses. Other significant coefficients are the dummy variables for the dotcom bubble (1999-2000) and the financial crisis (2007-2008). Acquisitions during the dotcom bubble lead to lower CARs (𝛽!= -0.0127) with 10% significance in model (1) and (𝛽_!= -0.0150) in model (2) with 5% significance. The financial crisis years show an opposite effect with higher CARs with (𝛽_!= 0.0066) for model (1) and (𝛽_!= 0.0150) for model (2) with 10% and 5% significance. Acquirer size, R&D intensity and acquisition value give positive insignificant coefficients. No definite conclusions can be made out of these

coefficients, but it could suggest that bigger acquirers have more capabilities for integrating target firms. The higher R&D expenditure would logically provide more knowledge for synergizing with the target’s unique characteristics. 4.4 Variance differences in cumulative abnormal returns Appendix 4 shows the scatterplot of the cumulative abnormal returns for the three day event window on target age. Noticable is the decreasing variance as target age increases. To test hypothesis 3, which states that CAR variation is larger when acquiring younger targets and decreases when the target firm age increases, the sample has been divided in four categories, sorted on target age. By performing F-tests for differences in variance, H3 can be accepted or rejected. Table 5 Variance testing per target age category

Target age df CAR std. dev F-stat 6-10 11-15 F-stat F-stat 16+

0-5 182 0.04191 1.279* 1.186 2.347*** 6-10 97 0.03706 - 0.927 1.835** 11-15 54 0.03848 - - 1.979*** 16+ 48 0.02735 - - - * Significant at 10%; **significant at 5%; ***significant at 1%. Table 5 shows the results per category. In the younger years, there seems to be no significant difference in variance. Except for variance in the years 0 to 5, which is significantly higher (10%) than the years 6 to 10. When comparing the first three categories (0-5, 6-10 and 11-15) with the last category of 16+ years, significant differences occur for all three categories. These results indicate that target firms older than 16 years have smaller variance in cumulative abnormal returns than the three younger categories. Therefore, hypothesis 3 will not be rejected. Simultaneously, heteroskedasticity can be assumed. Nevertheless it doesn’t have consequences for the regression performed in part 4.3, where robust standerd errors have been used.

5.

Conclusion

This final part contains the major findings in the study on the effect of target age on abnormal returns around M&A announcements under 13 big technology companies. This study has analysed 385 observations of acquisitions with announcement dates between the years 1987 through 2015. Firstly, abnormal returns are unpredictable. Previous studies lacked in providing a straightforward outcome. Instead, they showed multiple results between insignificant and significant positive abnormal returns for acquirers around M&A. The results of this study confirmed uncertainty in CAR behaviour, finding no different significant outcomes, stating that acquirers’ abnormal returns in this sample are not different from zero (H1). The importance of individual firm differences in stock return behaviour has to be emphasized, obviously suggesting the presence of individual acquisition strategies. The second finding is a negative significant effect of target age on the acquirers’ abnormal returns around acquisition announcements (H2). Mainly due to lack of information on young targets and the uncertainty on their future potential, causing underpricing. The last finding confirms the hypothesis of a decreasing CAR variance when target age increases (H3). This conclusion corresponds with, and strenghens the theory of uncertainty on younger targets, as stated earlier in the second hypothesis. The takeover of a young startup can evolve in both directions at extreme values compared to older targets, making them more risky. Noticable difference arises especially under targets in the last age category of 16 years and older. A concluding remark has to be made about the usefulness of multiple event windows. Especially in studies involving M&A events, where announcements are being made in multiple news sources and liable to information leakage. Future improvements can be made by investigating this subject in multiple industries, involving more individual companies. However, this is time consuming since data research depends heavily on manual assembly. Also, it has to be noted that data availabilty on targets is limited (private targets), which deteriorates the possibilities. More opportunities exist in further analysing the

unique acquisition strategies of the individual companies and their level of succes based on accounting performance.

References

Andrade, G., Mitchell, M. and Stafford, E. (2001) ‘New evidence and perspectives on mergers,’ Journal of Economic Perspectives, 15(2), 103-120. Barney, J. (1988) ‘Returns to Bidding Firms in Mergers and Acquisitions: Reconsidering the Relatedness Hypothesis,’ Strategic Management Journal, 9(1), 71-78. Dealogic – M&A Statshot (2016) ‘Insightful analysis of Merger & Acquisition Activity and trend around the world,’ http://www.dealogic.com/media/market-insights/ma-statshot/#MA050416 Eisenhardt, K.M., Graebner, M.E. and Roundy, P.T. (2010) ‘Succes and Failure in Technology Acquisitions: Lessons for Buyers and Sellers,’ Academy of Management Perspectives, 24(3), 73-92. Fuller, K., Netter, J. and Stegemoller, M. (2002) ‘What do returns to acquiring firms tell us? Evidence from firms that make many acquisitions,’ The Journal of Finance, 57(4), 1763-1793. Jong de, F. (2007) ‘Event Studies Methodology’ Lecture notes written for the course Empirical Finance and Investment Cases. Keeble, D. (1990) ‘High-Technology industry’ Geography, 75(4), 361-364. Leland, H.E. and Pyle, D.H. (1977) ‘Information Asymmetries, Financial Structure, and Financial Intermediation,’ Journal of Finance, 32, 371-387. Loderer, C. and Martin, K. (1990) ‘Corporate acquisitions by listed firms: the experience of a comprehensive sample,’ Financial Management, 19(4), 17-33. Ma, J., Pagán, J.A. and Chu, Y. (2009) ‘Abnormal Returns to Mergers and Acquisitions in Ten Asian Stock Markets,’ International Journal of Business, 14(3), 235-250. MacKinlay, A. C. (1997) ‘Event Studies in Economics and Finance,’ Journal of Economic Literature, 35(1), 13-39. Moeller, S.B., Schlingemann, F.P. and Stulz, R.M. (2004) ‘Firm size and the gains from acquisitions,’ Journal of Financial Economics, 72(1), 201-228.

Peemöller, V.H., Geiger, T. and Barchet, H. (2001) ‘Bewertung von Early-Stage Investment im Rahmen der Venture Capital Finanzierung,’ Finanz Betrieb, 5(1), 334-344. Ransbotham, S. and Mitra, S. (2010) ‘Target Age and the Acquisition of Innovation in High-Technology Industries,’ Management Science, 56(11), 2076-2093. Rigby, D. and Zook, C. (2002) ‘Open-market innovation,’ Harvard Business Review, 80(10), 129. Salter, M. and Weinhold, W. (1979) ‘Diversification Through Acquisitions: Strategies for Creating Economic Value’ Free press. Silicon Valley Bank U.S. Startup Outlook 2016 (2016) ‘New reality prompts U.S. startups to focus on fundamentals,’ http://www.svb.com/innovation- economy-outlook/startup-outlook/us/ Smit, C.J.B. and Ward, M.J.D. (2007) ‘The impact of large acquisitions on the share price and operating financial performance of acquiring companies listed on the JSE,’ Investment Analysts Journal, 65(1), 5-14. Travlos, N.G. (1987) ‘Corporate takeover bids, methods of payment, and bidding firm’s stock returns,’ The Journal of Finance, 42(4), 943-963. Wong, A. and Cheung, K.Y. (2009) ‘The effects of merger and acquisition announcements on the security prices of bidding firms and target firms in Asia,’ International Journal of Economics and Finance, 1(2), 274-283.

Appendix 3 CAR graph for estimating the event window Appendix 4 Scatterplot of CARs on target age