The effects of acquisition size and target ownership on acquirer value in Great Britain

(1)

The effects of acquisition size and target ownership on

acquirer value in Great Britain

Master’s Thesis

MSC Economics and MSC Finance

Maarten W. Hoen University of Groningen

S2326884

Supervisor: Dr. J.H. Von Eije

Based on data on acquisitions by British firms over the period 2000 till 2017, this paper examines the influence of an acquisition’s size and its target’s ownership structure on the acquirer’s cumulative abnormal returns. Using an event study methodology, this paper concludes that acquisitions of listed firms perform worse than acquisitions of unlisted firms. Also, relative deal size is positively associated with cumulative abnormal returns. Yet, those cumulative abnormal returns turn strongly negative once it comes to the acquisition of relatively large, listed companies.

Keywords: Event study, Acquisitions, Great Britain, Relative deal size, Listed companies, Unlisted companies

(2)

2 1.0 introduction

Annually, billions are spent on the acquisition of corporate assets, and, whereas some may gain from those trades, Moeller, Schlingemann & Stulz (2005) claim that on average 12% of shareholder value is destroyed due to acquisition attempts. Those numbers make us wonder about the motives behind acquisitions and the drivers founding its potential value creation and/or destruction.

Based on a study involving roughly 4400 acquisitions in Western Eurpean countries, Faccio, McConnell & Stolin (2006) conclude that there is a significant difference between acquisitions of private firms and acquisitions of public firms. Where acquisitions of unlisted companies would provide the acquirer with an average abnormal return of 1.48% following its announcement, the average abnormal returns of listed targets would be negative (-0.36%). Concluding private firms to be sold at discounts, Koeplin, Sarin & Shapiro (2000) and Fuller, Netter & Stegemoller (2002) underscore this conclusion from an empirical as well as a theoretical perspective. Yet, Capron & Shen (2007), describe the relationship between ownership structure and abnormal returns to be less clear cut as there is a significant heterogeneity in results. As the relationship between a target’s ownership structure and acquirer’s takeover performance seems contingent on various determinants, this research will further investigate the influence from a target’s structure on acquisition performance.

Another major issue in economic research concerns the influence of acquisition size on shareholder value. Moeller, Schlingemann & Stulz (2004) claim that losses from acquisitions are mostly incurred by enterprises attempting to acquire relatively large counterparts. Companies purchasing small firms, on the other side, are claimed to achieve significantly better share price results. Switzer (1996), however, reaches completely opposite conclusions by noting that particularly the acquisitions of large companies would be value enhancing for acquirers. Also, Alexandridis, Fuller, Terhaar & Travlos (2013) conclude that potential for overpayment is smaller when targets are substantial. Some further research, like Sharma & Ho (2002), fails to find any relationship.

(3)

3 This paper particularly distinguishes itself from previous research by examining the interaction between size and ownership structure. Whereas previous research analyzed the influence of size and target’s ownership on an acquisition’s performance independently, this paper considers both aspects simultaneously by researching the interaction between size and ownership structure and its relation to acquisition performance. Further, compared to previous work, this paper is based on more recent data. Given the recent economic crisis and ever changing market conditions, investigation based on recent data is highly valuable. Further, contrary to most literature, this paper particularly addresses listed acquirers located in Great Britain. As such, this research expands and enhances today’s academic literature.

Moreover, given our major finding that acquirer’s shareholder returns are particularly hampered following the announcement of acquisitions on large listed companies, this research yields important implications for economic practitioners and managers.

The remainder of this paper is as follows. Section two describes the leading insights regarding shareholder responses on acquisition announcements and further elaborates on our research objectives. Section three describes our dataset and motivates the use of several variables. Section four summarizes our econometric analysis and its results, whereas section five provides a robustness analysis. The final section will conclude this research.

2.0 Theory and hypothesis

(4)

4 Companies aiming to overcome this adverse selection problem need to decrease the information asymmetry between the buyer and seller. Doing so requires the collection of additional information by buyers and/or provision of creditworthy signals by targets. As acknowledged by Becchetti & Trovato (2002), related costs to those search and due diligence processes increase the price of acquisitions of unlisted companies. This information problem is particularly an issue for companies which value highly relies on a proper evaluation of intangible assets (Capron & Shen, 2007).

An advantage for the acquirer of private firms being less visible to the public concerns the associated lower publicity during the acquisition negotiations. As mentioned by Conn, Cosh, Guest & Hughes (2005), (acquiring) managers usually face pressure during negations as their reputation is contingent on its outcome. As managers are inclined not to lose face following the termination of acquisition negotiations, managers engaged in easily observed negotiations have a tendency to overpay for a target rather than to terminate the negotiations. As acquisitions of private companies are less visible to the general public, it is easier to end acquisition negotiations regarding private companies without publicly damaging the acquiring manager’s reputation. As such, the limited exposure following private acquisition negotiations decreases the probability that managers will overpay, rendering acquisitions of private companies advantageous.

Further factors distinguishing public from private acquisitions involve the liquidity of transferable equity. As private companies are not tradable on exchanges, they are less liquid compared to public ones. Fuller et al. (2002) infer that this illiquidity of private companies causes them be sold at a discounts.

Also, Capron & Shen (2007) argue that there is lower bidder competition for private targets compared to public targets. Given supply, decreased demand generally puts downwards pressure on prices. Limited bidder competition for private targets, therefore, increases potential realization of supernormal returns following such acquisition.

Considering all arguments provided above, to analyze whether private acquisitions indeed have better performance compared to their public counterpart, the following hypothesis is established:

Hypothesis 1. Following an acquisitions announcement, cumulative abnormal stock market returns for British acquiring companies are lower when the target is listed compared to when the target is unlisted.

(5)

5 On the other side, when acquiring large companies, potential for synergies is generally substantial. Especially when an industry is characterized by a natural monopoly (that is, production exhibits subadditivity of costs) or by a degree of monopolistic and oligopolistic power, benefits from economics of scale are potentially large.

Further, when the acquisition of large companies leads to a substantial consolidated market share, the acquirer would be able to benefit from market power and the corresponding ability to earn supernormal profits.

It might prove difficult, however, to integrate and consolidate large companies in one’s existing business structure. Given the high costs of integrating and managing large companies and considering the potential of failure in doing so would hamper shareholder value (Clark & Ofek, 1994).

To analyze the relationship between deal size and acquisition performance empirically, our second hypothesis reads:

Hypothesis 2. Following an acquisition announcement, cumulative abnormal stock market returns for British companies relate negatively to relative deal size.

To review whether this relationship differs between acquisitions on listed and unlisted firms, the following hypothesis is investigated also:

Hypothesis 3. Following an acquisition announcement, the relationship between cumulative abnormal stock market returns and relative deal size is negatively influenced by acquisitions of listed companies.

Moeller et al. (2004) conclude that particularly acquisitions of the extremely large companies detriment average cumulative abnormal results following an acquisition. To analyze whether the relationship between size and cumulative abnormal returns is especially harmed by acquisitions of large, listed enterprises, our final hypothesis reads:

Hypothesis 4. Following an acquisition announcement, cumulative abnormal stock market returns for British listed acquiring companies relate negatively to relative deal size for the largest 5 percentiles of relative deal size data.

3.0 Data and methodology 3.1 Sample selection

(6)

6 internationally cross listed companies are dropped from the dataset. Further, this research only includes observations where the rumor date of an acquisition equals its announcement date. Companies without a traceable ISIN codes are discarded also, resulting in a final sample with 3290 observations of which 239 on listed and 3051 on unlisted targets.

3.2.1 Abnormal returns

Data on abnormal returns are obtained based on an event study methodology as inspired by MacKinlay (1997). This methodology, as outlined below, enables us to compare deviant returns on a security due to the announcement of an acquisition, to normal returns (returns for the security during this period without such event). Any difference between those normal and the observed returns (either positive or negative) is presumably caused by the acquisition announcement and is referred to as an abnormal return.

As normal returns are not observable in the period surrounding an announcement, the methodology entails examining what the returns would have been in absence of the acquisition disclosure. To do so, we estimate how sensitive a security is to market returns during the period before the announcement. Based on the estimated sensitivity coefficients (𝛼̂ and 𝛽𝑖 ̂ ) and the post announcement market returns, 𝑖 normal returns for the security can be extrapolated.

(7)

7 The sensitivity of normal returns to the market is estimated from an OLS regression of the market equation:

𝑅𝑖𝑡 = 𝛼𝑖+ 𝛽𝑖𝑅𝑚𝑡+ 𝜀𝑖𝑡 (1)

The variable 𝑅𝑖𝑡 represents the daily (normal) return of bidder 𝑖’s shares at day 𝑡, with 𝑡 being the 100 working days advancing the event window. These return data are retrieved from Thomson Reuters’ DataStream and are defined as the daily percentage change in the security’s index (reinvestment of dividends included). The explanatory variable 𝑅𝑚𝑡 proxies the corresponding market return at date 𝑡, calculated as the daily percentage change of the Financial Times Stock Exchange All-Shares index (FTSE ALL) including reinvested dividends. Coefficients 𝛼𝑖 and 𝛽𝑖 represent the intercept and the security’s systematic risk, respectively, whereas 𝜀𝑖𝑡 denotes the random disturbance term with expected value zero. Note that, based on equation 1, prevalence of stochastic trend processes, non-co-integration of return data and potential for serially correlated disturbance terms could hamper the reliability and precision of the coefficient estimates for some securities. A further elaboration on this matter and several alternatives are provided in section 5.

Presuming our estimates to be reasonably accurate for now, abnormal returns (ARit) are calculated as the difference between the returns during the event window (t=-2,-1,0,1,2) and the return as would be expected based on the (in equation 1) estimated 𝛼̂ and 𝛽𝑖 ̂ . That is: 𝑖

𝐴𝑅𝑖𝑡 = 𝑅𝑖𝑡− (𝛼̂ + 𝛽𝑖 ̂ 𝑅𝑖 𝑚𝑡) (2)

Consequently, cumulative abnormal returns (CARi) during the event window are given by:

𝐶𝐴𝑅𝑖 = ∑2𝑡=−2𝐴𝑅𝑖𝑡 ≡ ∑2𝑡=−2(𝑅𝑖𝑡− (𝛼̂ + 𝛽𝑖 ̂ 𝑅𝑖 𝑚𝑡)) (3)

3.2.2 Listed firms

Analyzing the potential difference between acquisitions of publicly listed and private (or unlisted) companies requires that all target companies can be categorized accordingly. Therefore, any observation involving a target company without information on their legal form is dropped. The remaining observations are assigned an indicator variable labeled 0 in case Bureau van Dijk classifies the company to be unlisted and 1 otherwise.

3.2.3 Size

(8)

8 (later also referred to as market values) of the acquiring company are collected from DataStream. Note that those market values are collected based on the trading day advancing the event period to ensure that rumors on potential takeovers are not incorporated in the security prices already. A third and final numerical variable concerning size is relative deal size, defined as an observation’s deal value divided by the acquirers pre-announcement market value multiplied by 100%.

One should note that data on relative deal size is highly dispersed. Whereas deal values for the vast majority of observations compromise merely one percent of the acquirer’s pre-announcement market value, relative deal values for observations in the largest percentile surge towards a striking 62484%. Part of those outliers are attributable to the structure or particular type of the acquisition. Reverse takeovers, for example, are structures in which a small company purposely acquires a larger enterprise in exchange for a significant part of its own shares. Such structures and particular deal subtype classifications relate more to legal and accounting maneuvers than to firms actually acquiring a significantly larger counterpart. As such, those observations hamper our estimations whilst being economically insignificant, requiring observations classified by Bureau van Dijk as reverse takeover to be discarded from our data. Following this adjustment, minimum and maximum relative deal values are 0.001% and 286%, respectively, averaging at 3.2% with a standard deviation of 40.7%.

As remaining deviant observations are a particular subject of interest for this research, no further alterations to size variables are performed.

Since Moeler et al. (2004) argue the negative relationship between relative deal size and cumulative abnormal returns to hold particularly for the largest percentiles of relative deal size data. To analyze the influence of the largest percentiles of relative deal size data, a dummy variable ‘Large’ is created reading a value of 1 in case the observation belongs to the largest 5 percentiles based on relative deal size and 0 otherwise.

3.2.4 Control variables

(9)

9 internationally could induce managers to pay ill-founded premiums, hampering shareholder value. The possibility to gain from international tax benefits (Servaes & Zenner, 1994), exchange rates (Froot & Stein, 1991) and a scope for cross-border synergies (Doukas, & Travlos, 1988) on the other side, potentially increase the value of international takeovers. To isolate our estimation results from any of the effects induced by international acquisitions, this paper includes a dummy variable ‘Cross-border’. This dummy obtains a value of 1 in case of a cross-border acquisition and 0 in case of a domestic (British) takeover.

Considering cross-border acquisitions only, Capron & Shen (2007) conclude that takeovers of listed companies appear more successful compared to takeovers of unlisted enterprises. Rosenkopf & Almeida (2003) substantiate this statement intuitively by concluding firms to have a lower ability to adequately value distant assets. As a corollary, they conclude the informational disadvantage from acquisitions on unlisted targets to be more pronounced when it comes to international transactions. To control for the conditional relationship between a target’s geographical location and ownership status, we include the interaction variable ‘Listed * Cross-border’, generated as the variable ‘Listed’ multiplied by the variable ‘Cross-border’. This variable yields value 1 in case a particular observation concerns an international takeover of a listed company and 0 otherwise. Based on data from Bureau van Dijk we conclude that 618 acquisitions were cross-border, of which 100 cross-border and listed. Beyond analyzing the effect of distance on an acquisition announcement, to analyze robustness of our estimation results, also a numerical measure of cultural difference between those countries is incorporated. This variable ‘Cultural distance’ is proxied by the absolute value of the difference between Hofstede’s cumulative culture code for Great Britain and the target country. In case of a domestic takeover this variable carries value 0.

(10)

10 observations (1615/3290) appears to be diversifying, whereas the remainder of acquisitions is conducted in the bidder’s main field of expertise.

Schwert (1989) suggests that stock market volatility changes over time due to economic activity, macroeconomic factors, leverage and stock trading activities. Also catastrophic events (Nikkinen & Vähämaa 2010) and changes in political regime (Santa-Clara & Valkanov, 2003) could highly influence the economic diligence. To compensate for these changes in economic vigor over time, year dummies are constructed to capture such time fixed effects. Further, as markets might respond particularly pronounced during recessions, recession dummies - indicating 1 in case the acquisition was announced during a recession and 0 otherwise - are constructed. Recession dates (defined as a period of two successive quarters of negative real GDP growth) are based on OECD reports as retrieved from the Federal Reserve of Saint Louis (FRED). Merging this OECD data with our sample concludes that roughly one-third of acquisitions were performed during a recession. Please refer to table 4.2 for more detailed descriptive statistics.

The fact that some companies engaged in acquisitions before could serve as a valuable signal to investors. Namely, if companies are known to be experienced in acquisitions, they will most likely be more successful in consecutive ones. Consequently, we construct an experience dummy indicating 1 in case the bidding company has participated in at least three acquisitions before (1017 observations) and 0 otherwise (2273 observations). Data on the number of acquisitions per company is collected from Bureau van Dijk. As their database only reports on acquisitions starting in 2000 and is not necessarily complete, one should be cautious inferring conclusions based on this variable.

Travlos (1987) argues that acquisitions financed using cash usually yield higher returns. To compensate for such effect, based on data from Bureau van Dijk, a dummy is constructed yielding the value of 1 in case the majority of the deal value is financed using cash and 0 otherwise.

Also the variable ‘Completed on announcement day’ is constructed, yielding a value of 1 in case the acquisition is completed on the same day as it was announced and 0 otherwise. Those acquisitions reveal dedication in the acquisition process and remove any uncertainty regarding possible termination of acquisition negotiations.

(11)

11 3.3 Descriptive statistics

Descriptive statistics for all numerical variables are provided in table 4.1 below. Table 4.2 displays numerical statistics for all indicator variables. As this research particularly focusses on listed and unlisted firms for the largest 5% of observations based on relative deal size, statistics are organized based on multiple categorizations.

Table 4.1: Descriptive statistics numerical variables

Complete Listed Unlisted Listed 5% Unlisted 5%

CAR (percentage) Observations 3290 239 3051 12 153 min -72.859% -31.928% -72.859% 16.262% 15.326% max 120.963% 120.963% 109.647% 120.964% 109.647% mean 1.706% 0.217% 1.792% 32.163% 27.217% SD 9.771% 12.374% 9.504% 29.926% 15.958%

Relative deal size (Percentage)

Observations 3290 239 3051 12 153

Min 0.001% 0.027% 0.001% 172.374% 89.088%

Max 286.696% 286.696% 271.627% 286.964% 270.627%

Mean 3.203% 44.474% 21,311% 206.167% 141.711%

SD 40.729% 55.480% 34,498% 37.348% 48.526%

Market value (millions)

Observations 3290 239 3051 12 153

Min 0.100 3.070 0.100 4.860 0.150

Max 201546.700 106988.100 201536.700 66193.310 587.580

Mean 3.203 7039.583 1120.353 7393.119 25.657

SD 40.729 18618.629 7551.881 18286.430 65.028

Deal value (millions)

Observations 3290 239 3051 12 153

Min 0.006 0.044 0.006 12.943 0.160

Max 189951.095 189951.095 8414.169 189951.095 612.699

Mean 149.895 1621.334 34.629 1891.966 25.388

SD 3502.226 12899.013 245.610 52145.869 65.028

Cultural distance (points)

Observations 3220 221 2999 11 152

Min 0 0 0 0 0

Max 18.766 18.766 18.766 2.538 5.207

Mean 0.852 1.435 0.810 0.231 0.206

SD 0.259 2.970 2.634 0.765 0.803

Note: The column ‘complete’ provides statistics for the overall data sample, whereas the columns ‘listed’ and ‘unlisted’

(12)

12 Table 4.2: Numerical tables indicator variables

Complete Listed Unlisted Listed 5% Unlisted 5%

Industry match Yes (1) 1615 140 1475 7 81 No (0) 1675 99 1576 5 72 Experience Yes (1) 1017 10 1007 0 37 No (0) 2273 229 2044 12 116 Cross-border Yes (1) 681 100 581 2 14 No (0) 2609 139 2470 10 139

Completed on announcement date

Yes (1) 1895 24 1870 0 43

No (0) 1395 215 1181 12 110

Recession

Yes (1) 1296 103 1193 5 55

No (0) 1994 136 1858 7 98

Majority paid in Cash

Yes (1) 1863 141 1722 5 76 No (0) 1427 98 1329 7 77 SIC Code 1 196 43 162 0 6 2 322 35 287 2 6 3 417 31 386 0 13 4 274 19 255 1 16 5 247 7 240 0 14 6 484 53 431 8 32 7 821 47 774 1 42 8 529 13 516 0 24 Year 2000 237 36 201 3 10 2001 240 15 225 0 13 2002 215 18 197 0 11 2003 200 15 185 1 12 2004 237 10 227 1 10 2005 287 16 271 1 16 2006 321 22 299 1 18 2007 312 16 296 1 16 2008 203 10 193 2 11 2009 126 14 112 0 4 2010 175 14 161 1 7 2011 129 8 121 0 6 2012 108 11 97 0 5 2013 127 10 117 1 6 2014 134 7 127 0 4 2015 128 11 117 0 3 2016 111 6 105 0 1

Note: The column ‘complete’ provides statistics for the overall data sample, whereas the columns ‘listed’ and ‘unlisted’

(13)

13 4.0 Model specification and results

4.1.1 Model specification

Our main regressions to test our four hypotheses are based on the specifications constructed in this section. Further specifications for all hypotheses and a robustness analysis are discussed in section 5. For our first two hypotheses, to estimate the influence of ownership status (private vs public) and relative deal size on the acquirer’s cumulative abnormal returns, the following regression is performed: 𝐶𝑎𝑟𝑖 = 𝛼 + 𝛽1𝐿𝑖𝑠𝑡𝑒𝑑𝑖+ 𝛽2𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑑𝑒𝑎𝑙 𝑠𝑖𝑧𝑒𝑖+ 𝛽3𝐸𝑥𝑝𝑒𝑟𝑖𝑒𝑛𝑐𝑒𝑖+ 𝛽4𝐶𝑟𝑜𝑠𝑠 𝑏𝑜𝑟𝑑𝑒𝑟𝑖+

𝛽5𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑚𝑎𝑡𝑐ℎ𝑖+ 𝛽6𝐷𝑒𝑎𝑙 𝑣𝑎𝑙𝑢𝑒𝑖+ 𝛽7𝑀𝑎𝑟𝑘𝑒𝑡 𝑣𝑎𝑙𝑢𝑒𝑖+ 𝛽8𝑅𝑒𝑐𝑒𝑠𝑠𝑖𝑜𝑛𝑖+ 𝛽7+𝑥𝑆𝐼𝐶𝑥,𝑖 (4) +𝛽15+𝑦𝑌𝑒𝑎𝑟(2000+𝑦),𝑖+ 𝜀𝑖

In this specification, the variables CARi, Listedi, Relative deal sizei, Experiencei, Cross-borderi, Industry matchi, Deal valuei, Market valuei, Recessioni, the SIC codes (with subscript x labeling the particular SIC code) and year dummies (identifiable by subscript y) refer to the variables as have been discussed in section 3. Subscript i denotes each individual observation, the coefficient 𝛼 is our intercept, 𝛽1 till 𝛽31 represent the estimation coefficients related to those variables, respectively and 𝜀𝑖 is a random disturbance term with expected value zero. Sic code 1 and the year 2000 are omitted to avoid collinearity.

To analyze whether a negative relationship between relative deal size particularly holds for firms acquiring listed targets (our third hypothesis), the interaction variable Listed * Relative deal size is included. Further, as outlined in section 3.2.4, since the information asymmetry concerning private and public targets appears more pronounced for international acquisitions, the control variable Listed * Cross border is incorporated also. The coefficients 𝛼, 𝛽𝑥 and 𝜀𝑖 have a similar interpretation to as what they were assigned in our previous estimation equation.

𝐶𝑎𝑟𝑖 = 𝛼 + 𝛽1𝐿𝑖𝑠𝑡𝑒𝑑𝑖+ 𝛽2𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑑𝑒𝑎𝑙 𝑠𝑖𝑧𝑒𝑖+ 𝛽3(𝐿𝑖𝑠𝑡𝑒𝑑𝑖∗ 𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑑𝑒𝑎𝑙 𝑠𝑖𝑧𝑒𝑖)

+𝛽4(𝐿𝑖𝑠𝑡𝑒𝑑𝑖∗ 𝐶𝑟𝑜𝑠𝑠 𝑏𝑜𝑟𝑑𝑒𝑟𝑖) + 𝛽5𝐸𝑥𝑝𝑒𝑟𝑖𝑒𝑛𝑐𝑒𝑖+ 𝛽6𝐶𝑟𝑜𝑠𝑠 𝑏𝑜𝑟𝑑𝑒𝑟𝑖 (5) +𝛽7𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑚𝑎𝑡𝑐ℎ𝑖+ 𝛽8𝐷𝑒𝑎𝑙 𝑣𝑎𝑙𝑢𝑒𝑖 +𝛽9𝑀𝑎𝑟𝑘𝑒𝑡 𝑣𝑎𝑙𝑢𝑒𝑖+ 𝛽10𝑅𝑒𝑐𝑒𝑠𝑠𝑖𝑜𝑛𝑖

+ 𝛽9+𝑥𝑆𝐼𝐶𝑥,𝑖+ 𝛽17+𝑦𝑌𝑒𝑎𝑟(2000+𝑦),𝑖+ 𝜀𝑖

(14)

14 𝐶𝑎𝑟𝑖 = 𝛼 + 𝛽1𝐿𝑖𝑠𝑡𝑒𝑑𝑖+ 𝛽2𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑑𝑒𝑎𝑙 𝑠𝑖𝑧𝑒𝑖+

𝛽3(𝐿𝑖𝑠𝑡𝑒𝑑𝑖∗ 𝐿𝑎𝑟𝑔𝑒𝑖∗ 𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑑𝑒𝑎𝑙 𝑠𝑖𝑧𝑒𝑖) + 𝛽4(𝐿𝑖𝑠𝑡𝑒𝑑𝑖∗ 𝐶𝑟𝑜𝑠𝑠 𝑏𝑜𝑟𝑑𝑒𝑟𝑖) (6) +𝛽5𝐸𝑥𝑝𝑒𝑟𝑖𝑒𝑛𝑐𝑒𝑖+ 𝛽6𝑐𝑟𝑜𝑠𝑠𝑏𝑜𝑟𝑑𝑒𝑟𝑖 +𝛽7𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑚𝑎𝑡𝑐ℎ𝑖+ 𝛽8𝐷𝑒𝑎𝑙 𝑣𝑎𝑙𝑢𝑒𝑖

+𝛽9𝑀𝑎𝑟𝑘𝑒𝑡 𝑣𝑎𝑙𝑢𝑒𝑖+ 𝛽10𝑅𝑒𝑐𝑒𝑠𝑠𝑖𝑜𝑛𝑖+ 𝛽9+𝑥𝑆𝐼𝐶𝑥,𝑖+ 𝛽17+𝑦𝑌𝑒𝑎𝑟(2000+𝑦),𝑖+ 𝜀𝑖

All equations are estimated based on an ordinary least square regression. 4.1.2 Specification tests

Ordinary least squares regression results from cross sectional data yield the risk of exhibiting heteroskedastic disturbance terms. To analyze whether our data indeed suffers from heteroscedasticity, a Breusch-Pagan heteroscedasticity test is performed on all of the above specifications (equation 4 till 6).

Regarding equation 4 (also referred to as model 1), The Breusch-Pagan test concludes with a P-value of 0.000 and a Chi2_{statistic of 258.91. As a consequence, we reject the null hypothesis that the variance} of our residuals is constant and conclude that our model most likely suffers from heteroskedastic error terms. Therefore, estimations based on model 1 commence by means of a robust regression.

Similarly, Breusch-pagan tests for the specifications based on model 2 (equation 5, P = 0.000, Chi2₌ 263.08) and model 3 (equation 6, P = 0.000, Chi2_{= 452.01) indicate that our explanatory variables are} most likely related to the disturbance term, causing us to use robust standard errors in all regressions. Further, multicollinearity of our explanatory variables could hamper our estimation results as its presence inflates our standard errors. Variance Inflation Factor (VIF) analyses for all specifications are displayed in appendix A. Given that none of the reported inflation factors exceeds (nor closely approaches) a value of 10, we consider it unlikely that our research suffers from multicollinearity. Observation of merely small correlation coefficients between our explanatory variables (refer to appendix A) substantiates this conclusion.

4.2 Results

Relevant estimation results are reported in table 4.4. Please refer to appendix A for the complete estimation output.

(15)

15 significant, casting doubt on the robustness of the coefficient and the validity of our conclusion for the overall dataset. Further, note that the R-squared for either of the models is extremely low (0.018 and 0.022). This implies that only a very small part of the variation in cumulative abnormal return data is explained by our specifications. This might prove problematic for a low R-square hampers the explanatory power of our specifications and signifies potential for a missing variable bias.

Contrary to what we expected based on previous research, our results indicate a positive relationship between an acquisition’s relative deal size and the associated cumulative abnormal returns. With a P-value of 0.035, we observe that this relationship is significant, and consider this as direct evidence against our second hypothesis. As a consequence, we conclude that it is unlikely that, following the announcement of an acquisition, relative deal size of an acquisition and cumulative abnormal returns for the acquirer are negatively associated for the complete dataset.

Table 4.4: Estimation results

Model 1 Model 2 Model 3

Dependent variable: CAR CAR CAR

Observations 3290 3290 3290

Listed -1.493 * (0.844) -0.751 (1.247) -1.397 (0.944)

Relative deal size 1.735 ** (0.825) 2.339 ** (0.926) 2.424 *** (0.509)

Listed * Relative deal size -3.824 ** (1.910)

Listed * Large * Relative deal size -4.439 ***(1.226)

Listed * Cross border 2.020 (2.074) 2.021 (1.394)

Market value (acquirer) -0.004 (0.009) -0.014 (0.010) -0.012 (0.020)

Deal value -0.066 (0.034) -0.014 (0.035) -0.007 (0.053) Industry match -0.230 (0.354) -0.224 (0.351) -0.237 (0.350) Cross border -0.721 (0.424) -0.976 ** (0.401) -0.966 ** (0.467) Experience -0.061* (0.334) -0.029 (0.334) -0.015 (0.367) Recession -0.840 (0.762) -0.899 (0.763) -0.916 (0.718) Constant 1.091 (1.174) 1.114 (1.162) 1.111 (1.161)

Refer to appendix A for coefficient estimates for the SIC codes Refer to appendix A for coefficient estimates for the year dummies

R-squared 0.018 0.022 0.024

Standard errors Robust Robust Robust

Note: Estimation based on an OLS regression with Robust standard errors. Year dummies and SIC codes are included in the

estimation but not reported in this table. Refer to appendix A for the overall estimation result. Standard errors are reported in parenthesis next to the coefficient estimate. *** indicates that the variable is significant at a 1% level, ** denotes significance at the 5% level and * represents significance at a 10% significance level. Dependent variable CAR is calculated as CARi= ∑2t=−2(Rit− (α̂ + βi ̂ Ri mt)), with α̂ and β being estimated from Ri it= αi+ βiRmt+ εit. The variable ‘Listed’ has

(16)

16 What we do conclude based on model 2, though, is that the interaction variable Listed * Relative deal size is significant and negative. Intuitively this suggests that the relationship between relative deal size and cumulative abnormal returns is negatively affected by acquisitions of listed companies. Given a P- value of 0.013 we accept our third hypothesis and conclude that it is likely that acquisitions on listed companies negatively affect the relationship between relative deal size and short term post announcement performance.

Narrowing the scope of this research to listed companies whose acquisitions belong to the largest 5 percentiles of relative deal size data (model 3), we observe a highly negative relationship between Listed * Large * Relative deal size and cumulative abnormal returns. Given a P-value of 0.000 and a t statistic of -3.62, we conclude, in line with our fourth hypothesis that cumulative abnormal returns are particularly harmed by acquisitions of listed companies with extremely large relative deal size. 5.0 Robustness and critiques

As mentioned in section 3.2.1, the estimates our cumulative abnormal returns are based on are potentially precarious. The event study methodology by MacKinlay (1997) requires normal returns during the event window to be estimated based on a security’s alpha and beta. Those latter coefficients are inferred from an OLS regression of the market equation (𝑅𝑖𝑡 = 𝛼𝑖+ 𝛽𝑖𝑅𝑚𝑡+ 𝜀𝑖𝑡). However, simply estimating those parameters based on an ordinary least square regression could be problematic for several reasons.

Firstly, for some time windows, data on market returns and security returns simultaneously exhibit stochastic trends processes. As those non-stationary series may not be co-integrated for some observations, coefficients 𝛼̂ and 𝛽𝑖 ̂ are possibly based on spurious relationships, casting doubt on their 𝑖 reliability.

Also, whereas not true for all observations, the random disturbance term in some of the estimations seems related to the variability of returns and/or to the value of lagged residuals. Those indications for heteroskedastic and serially correlated disturbance terms question whether 𝛽̂ are the best possible 𝑖 estimates for our securities’ sensitivities to the market. The latter causes the precision of our estimates to be low, rendering the calculation of some of the 𝐶𝐴𝑅𝑖𝑡 imprecise.

As standard errors for estimates 𝛼̂ and 𝛽𝑖 ̂ are generally large regardless, the precision of some of the 𝑖 in equation 2 calculated abnormal returns (𝐴𝑅𝑖𝑡 = 𝑅𝑖𝑡− (𝛼̂ + 𝛽𝑖 ̂ 𝑅𝑖 𝑚𝑡)), and as a corollary, the precision of some 𝐶𝐴𝑅𝑖𝑡, is limited.

(17)

17 Warner (1985), 𝛼̂ and 𝛽𝑖 ̂ are not estimated but set to 0 and 1, respectively for all observations. The 𝑖 assumption behind this methodology is that returns should on average fluctuate in line with the overall market. Cumulative abnormal returns calculated based on this methodology are referred to as 𝐶𝐴𝑅_𝑖𝑡𝐵𝑟𝑜𝑤𝑛.

Our second alternative (referred to as 𝐶𝐴𝑅_𝑖𝑡𝐴𝑣𝑒𝑟𝑎𝑔𝑒) entails calculating abnormal returns by subtracting the average security return from the returns observed during the event window. For this specification average returns are calculated based on the 100 trading days advancing the event window. That is: 𝐴𝑅𝑖𝑡 = 𝑅𝑖𝑡− ∑ (

𝑅_𝑖𝑡 100) 𝑡=−3

𝑡=−102 (7)

As can be observed from table 5.1, descriptive statistics for CAR data based on either of the three methodologies are relatively similar, signaling that application of those competing methodologies might not significantly alter our conclusions.

Table 5.1: Descriptive statistic CAR data based on various event study methodologies

CARMacKinlay _CARBrown _CARAverage

Min -72.859% -95.741% -74.470%

Max 120.963% 125.529% 120.120%

Mean 1.706% 1.609% 1.815%

SD 9.771% 10.077% 9.793%

Note: CARMacKinlay_{= ∑} _(R

it− (α̂ + βi ̂ Ri mt) 2

t=−2 ), with α̂ and β being estimated based on Ri it= αi+ βiRmt+ εit, CARBrown is

are generated as the sum of differences between security returns and market returns and CARavarage_{is calculated as the sum}

of differences between security returns during the event window minus its 100 working day average.

The results of our estimates per methodology for our first model (equation 4) are compared with each other in table 5.2 below. Alternative specifications for our second model (equation 5) are displayed in table 5.3 and results for the competing specifications based on equation 6 (model 3) are reported in table 5.4. To keep those tables organized, only our variables of interest are displayed. Please refer to appendix B for the complete regression results and corresponding specification tests.

Table 5.2: Model specification one based on various event study methodologies

Base model Alternative 1 Alternative 2

Dependent variable CARMacKinlay _CARBrown _CARAverage

Listed -1.493 * (0.844) -1.266 * (0.640) -1.476 * (0.893)

Relative deal size 1.735 ** (0.825) 1.330 * (0.807) 1.800 ** (0.824)

R-squared 0.018 0.019 0.019

Note: The variables ‘Recession’, ‘Experience’, ‘Cross border’, ‘Industry match’, ‘Deal value’, ‘Market value’, indicator variables

(18)

18 As can be noted from table 5.1 till 5.4, whilst coefficients and their standard errors experience minor changes in magnitude, the sign and significance of our major variables is hardly altered. This result substantiates our choice to calculate cumulative abnormal returns based on their estimated sensitivities to the market.

Table 5.3: Model specification two based on various event study methodologies

Listed -0.751 (1.247) -0.691 (1.232) -0.885 (1.348)

Relative deal size 2.339 ** (0.926) 1.927 ** (0.926) 2.351 ** (0.924)

Listed * Relative deal size -3.824 ** (1.910) -3.722** (1.937) -3.457 * (1.943)

R-squared 0.018 0.019 0.019

for all SIC codes and year dummies are included in the regressions as well, yet, are not reported on in this table. Please refer to appendix B for the complete estimation output. Robust standard errors are displayed in parenthesis to the right of the corresponding coefficient. *** indicates that the variable is significant at a 1% level, ** denotes significance at the 5% level and * represents significance at a 10% significance level.

Table 5.4: Model specification three based on various event study methodologies

Listed -1.397 (0.944) -1.322 (0.946) -1.391 (0.974)

Relative deal size 2.424 *** (0.509) 2.008 *** (0.510) 2.464 *** (0.525)

Listed * Large * Relative deal size -4.439 *** (1.226) -4.312 *** (1.228) -4.279 *** (1.265)

R-squared 0.023 0.024 0.023

for all SIC codes and year dummies are included in the regressions as well, yet, are not reported on in this table. Please refer to appendix B for the complete estimation output. Robust standard errors are displayed in parenthesis to the right of the corresponding coefficient. *** indicates that the variable is significant at a 1% level, ** denotes significance at the 5% level and * represents significance at a 10% significance level.

Inferring that our conclusions are not particularly sensitive to changes between the above described event study methodologies, using our initial methodology (inspired by MacKinlay’s, 1997), robustness of our results is further analyzed based on the inclusion of three additional variables.

(19)

19 analysis. Further, as described by Travlos (1987), acquisitions financed based on equity usually yield lower returns compared to acquisitions financed with cash. Consequently, a dummy depicting 1 in case the acquisition is financed mostly by cash in included. Our final additional variable measures whether the acquisition is completed on the same day as it was announced. Those acquisitions ensure dedication in the acquisition process and remove any uncertainty regarding termination of acquisition negotiations.

Table 5.5: Robustness analysis based on inclusion of further explanatory variables

Base 1 Alt 1 Base 2 Alt 2 Base 3 Alt 3

Dependent CAR CAR CAR CAR CAR CAR

Observations 3290 3220 3290 3220 3290 3220 Listed -1.493 * (0.844) -1.486 * (0.963) -0.751 (1.247) -1.034 (1.232) -1.397 (0.944) -1.626 * (0.969)

Relative deal size 1.735 **

(0.825) 1.734 ** (0.870) 2.339 ** (0.926) 2.299 ** (0.965) 2.424 *** (0.509) 2.382 *** (0.532) Listed * relative deal

size

-3.824 ** (1.910)

-3.503 * (1.887) Listed * large * relative

deal size

-4.439*** (1.226)

-4.027 *** (1.251)

Listed* cross border 2.020

(2.074) 3.002 (2.270) 2.021 (1.394) 2.928 ** (1.480) Market value (acquirer) -0.004 (0.009) -0.003 (0.010) -0.014 (0.010) -0.013 (0.011) -0.012 (0.020) -0.011 (0.022) Deal value -0.066 (0.034) -0.049 ** (0.020) -0.014 (0.035) -0.003 (0.026) -0.007 (0.053) 0.003 (0.053) Industry match -0.230 (0.354) -0.202 (0.355) -0.224 (0.351) -0.195 (0.353) -0.237 (0.350) -0.209 (0.355) Cross border -0.721 (0.424) -1.245 * (0.643) -0.976 ** (0.401) -1.620 ** (0.552) -0.966 ** (0.467) -1.620 * (0.630) Experience -0.061* (0.334) 0.010 (0.336) -0.029 (0.334) 0.046 (0.336) -0.015 (0.367) 0.055 (0.373) Recession -0.840 (0.762) -0.877 (0.776) -0.899 (0.763) -0.960 (0.777) -0.916 (0.718) -0.976 (0.726) Cash involved 0.290 (0.351) 0.266 (0.351) 0.287 (0.351) Cultural distance 0.121 * (0.070) 0.132 * (0.068) 0.135 (0.086) Completed on announcement day -0.283 (0.467) -0.228 (0.457) -0.203 (0.380) Refer to appendix C for coefficient estimates for the SIC codes Refer to appendix C for coefficient estimates for the year dummies

R-squared 0.018 0.019 0.022 0.022 0.024 0.023

Standard errors Robust Robust Robust Robust Robust Robust

(20)

20 Results for all the alterative specifications are listed in table 5.5 above. Standard errors are reported in parenthesis below the estimates. SIC codes and year dummies are included in the models, however, given to their limited significance, in order to keep our tables uncluttered, their results are left for appendix C.

Based on table 5.5, we again observe our results to be rather constant and our conclusions to remain valid given all alternative specifications.

Despite our results remaining relatively stable during this robustness analysis, one should note that all specification suffer from low R-squares. As it turns out, neither of our specifications seems able to explain a significant part of the variability of cumulative abnormal returns. The latter casts doubt on the reliability and explanatory power of our analysis. Also the corresponding potential for endogeneity make that our conclusions should be interpreted with caution.

6.0 Conclusion

Moeller et al. (2004) describe that failure to identify positive abnormal returns following acquisition disclosures is attributable to significant losses from acquisitions of relatively large listed companies. This paper examined the influence of ownership status (public vs private) and relative deal size on cumulative abnormal returns following the announcement of an acquisition.

Based on differences in liquidity, asymmetry in available information, a lower exposure related to private firms and less bidder competition for private companies, we expected acquisitions on listed companies to have lower performances compared to acquisitions on unlisted ones. Based on an event study methodology, evidence is detected that cumulative abnormal returns for British firms since 2000 are indeed negatively affected by takeovers of listed companies. However, this relationship is only marginally significant.

Further, this research concludes that relative deal size is overall positively associated with cumulative abnormal returns. However, in line with, Fuller et al. (2002), we observe the relationship between size and return to be negatively affected by acquisitions on listed firms. The more so, as also described by Moeller et al. (2005) for the largest percentiles of relative deal size data.

(21)

(22)

22 References

Akerlof, G. (1970). The Market for "Lemons": Quality Uncertainty and the Market Mechanism. The Quarterly Journal of Economics, 84(3), 488-500.

Alexandridis, G., Fuller, K., Terhaar, L., & Travlos, N. (2013). Deal size, Acquisition Premia and Shareholder Gains. Journal of Corporate Finance, 20, 1-13.

Becchetti, L., Trovato, G. (2002). The Determinants of Growth for Small and Medium Sized Firms. The Role of the Availability of External Finance. Small Business Economics, 19(4): 291-306.

Brown, J., Warner, J. (1985). Using Daily Stock Returns, The Case Of Event Studies. Journal of Financial Economics, 14, 3-31.

Bureau van Dijk (nd). Zephyr. Retrieved from https://zephyr-bvdinfo-com.proxy-ub.rug.nl/home.serv?product=zephyrneo&loginfromcontext=ipaddress

Capron, L., Shen, J. (2007). Acquisitions of Private vs. Public Firms: Private Information, Target Selection, and Acquirer Returns. Strategic Management Journal, 28(9), 891-911.

Clark, K., Ofek, E. (1994). Mergers as a Means of Restructuring Distressed Firms: An Empirical Investigation. The Journal of Financial and Quantitative Analysis, 29(4), 541-565.

Conn, R., Cosh, A., Guest, P., & Hughes, A. (2005). The Impact on UK acquirers of Domestic, Cross-border, Public and Private Acquisitions.Journal of Business Finance & Accounting, 32, 815–870. DataStream (nd). Thomson Reuters DataStream. Available at: Subscription Service (Accessed: June 2017)

Denis, D., Denis, D., & Yost, K. (2002). Global Diversification, Industrial Diversification, and Firm Value. The Journal of Finance, 57(5), 1951-1979.

Doukas, J., Travlos, N. (1988). The Effect of Corporate Multinationalism on Shareholders' Wealth: Evidence from International Acquisitions. The Journal of Finance, 43(5), 1161-1175.

Faccio, M., McConnell, J., & Stolin, D. (2006). Returns to Acquirers of Listed and Unlisted Targets. The Journal of Financial and Quantitative Analysis, 41(1), 197-220.

Fama, E. (1965). The Behavior of Stock-Market Prices. The Journal of Business, 38(1), 34-105.

FRED (nd). OECD based Recession Indicators for the United Kingdom from the Peak through the Trough. Retrieved from https://fred.stlouisfed.org/series/GBRRECDM

Froot, K., Stein, J. (1991). Exchange Rates and Foreign Direct Investment: An Imperfect Capital Markets Approach. The Quarterly Journal of Economics, 106(4), 1191-1217.

Fuller, K., Netter, J., & Stegemoller, M. (2002). What do Returns of Acquiring Firms tell us? Evidence from Firms that make Many Acquisitions. The Journal of Finance, 57(4), 1763–1793

(23)

23 MacKinlay, A. (1997). Event Studies in Economics and Finance. Journal of Economic Literature, 35(1), 13-39.

Moeller, S., Schlingemann, F., & Stulz, R. (2004). Firm Size and the Gains from Acquisitions. Journal of Financial Economics, 73(2), 201-228.

Moeller, S., Schlingemann, F., & Stulz, R. (2005). Wealth Destruction on a Massive Scale? A Study of Acquiring-Firm Returns in the Recent Merger Wave. The Journal of Finance, 60(2), 757-782.

Morck, R., Shleifer, A., & Vishny, R. (1990). Do Managerial Objectives Drive Bad Acquisitions? The Journal of Finance, 45(1), 31-48.

Nikkinen, J., Vähämaa, S. (2010). Terrorism and Stock Market Sentiment. Financial Review, 45(2), 263-275.

Reuer, J., Ragozzino, R. (2008). Adverse Selection in Acquisitions of New Ventures: The Roles of Alliances and IPOs in M&A Design. Journal of Economic Behavior & Organization, 66, 195-212.

Rosenkopf, L., Almeida, P. (2003). Overcoming Local Search Through Alliances and Mobility. Management Science, 49(6), 751-766.

Santa-Clara, P., Valkanov, R. (2003). The Presidential Puzzle: Political Cycles and the Stock Market. The Journal of Finance, 58(5), 1841-1872.

Schwert, G. (1989). Why Does Stock Market Volatility Change Over Time? The Journal of Finance, 44(5), 1115-1153.

Servaes, H., Zenner, M. (1994). Taxes and the Returns to Foreign Acquisitions in the United States. Financial Management, 23(4), 42-56.

Seth, A., Song, K., & Pettit, R. (2000). Synergy, Managerialism or Hubris? An Empirical Examination of Motives for Foreign Acquisitions of U.S. Firms. Journal of International Business Studies, 31(3), 387-405.

Shaked, I., Michel, A., & Mcclain, D. (1991). The Foreign Acquirer Bonanza: Myth Or Reality?. Journal of Business Finance & Accounting, 18, 431–447.

Sharma, D., Ho, J. (2002). The Impact of Acquisitions on Operating Performance: Some Australian Evidence. Journal of Business Finance & Accounting, 29, 155-200.

Switzer, J. (1996). Evidence on Real Gains in Corporate Acquisitions. Journal of Economics and Business, 48(5), 443-460.

Travlos, N. (1987). Corporate Takeover Bids, Method Of Payment, And Bidding Firm’s Stock Returns. Journal of Finance, 52, 943–963.

Von Eije, H., Wiegerinck, H. (2010). Shareholders' Reactions to Announcements of Acquisitions of Private Firms: Do target and Bidder Markets make a Difference?. International Business Review, 19(4), 360-377.

(24)

24 APPENDIX A

Table A.1: Variance inflation factor analysis.

VIF VIF VIF

Listed 1.11 2.95 2.11

Listed cross border 2.06 2.01

Large listed 2.32

Large listed size 1.56

Recession 4.31 4.32 4.32

Experience 1.06 1.06 1.06

Crossborder 1.14 1.25 1.26

Industry match 1.08 1.08 1.08

Deal value 1.09 1.21 1.19

Market value (acquirer) 1.15 1.17 1.16

Relative deal size 1.10 1.27 1.24

SIC2 2.42 2.42 2.42 SIC3 2.81 2.81 2.81 SIC4 2.25 2.25 2.25 SIC5 2.16 2.16 2.16 SIC6 3.04 3.04 3.04 SIC7 4.06 4.06 4.06 SIC8 3.26 3.26 3.26 Year 2001 1.91 1.91 1.91 Year 2002 1.94 1.95 1.95 Year 2003 2.30 2.30 2.30 Year 2004 1.88 1.88 1.88 Year 2006 3.11 3.11 3.11 Year 2006 3.34 3.34 3.34 Year 2007 3.28 3.29 3.29 Year 2008 1.79 1.79 1.79 Year 2009 1.62 1.62 1.62 Year 2010 2.34 2.34 2.34 Year 2011 2.02 2.02 2.02 Year 2012 1.86 1.86 1.86 Year 2013 1.99 2.00 2.00 Year 2014 1.89 1.89 1.89 Year 2015 1.51 1.52 1.52 Year 2016 1.45 1.45 1.45 Mean 2.14 2.21 2.16

Note: Variance Inflation Factor (VIF) analysis as based on the specifications described by

(25)

25 Table A.2: Correlation matrix

Lis te d 1 Lis te d Cr o ss b o rd er 0.6326 1 Larg e* Lis te d * siz e 0.3752 0.0690 1 R ec es sio n 0. 0212 0.0239 0.0179 1 Ex p eri en ce -0.0371 -0.0258 -0.0237 -0.0279 1 Lis te d * Siz e 0.6119 0.2103 0.9032 0.0296 _-0.0386 1 Cro ss b o rd er 0.1461 0.3466 _-0.0165 _-0.0326 _-0.0044 0.0145 1 In d u str y-matc h 0.0531 0.0599 0.0301 _-0.0139 _-0.0094 0. 0486 0.0761 1 De al valu e 0.1176 0.0474 0.3299 _-0.0037 0.0179 0.3284 0.0129 0.0324 1 Mar ke t valu e 0.1714 0.2104 0.0325 0.0169 0.0644 0.0400 0.1759 0.0341 0.2095 1 R elati ve d eal s iz e 0.1629 0.0127 0.4055 0.0036 _-0.0983 0.4195 _-0.0909 0.0430 0.1497 _-0.0625 1 Lis te d Lis te d * Cro ss b o rd ern b o r der _Larg e * lis te d * s iz e R ec es sio n Ex p eri en ce Lis te d * s iz e Cro ss b o rd er In d u str y matc h D eal valu e Mar ke t valu e R elati ve d eal s iz e

Note: Correlation matrix including our main variables as described in section 3. SIC codes

(26)

26 Table A.3: Complete estimation results

Dependent variable: CAR CAR CAR

Listed -1.493 * (0.844) -0.751 (1.247) -1.397 (0.944)

Relative deal size 1.735 ** (0.825) 2.339 ** (0.926) 2.424 *** (0.509)

Listed * Relative deal size -3.824 ** (1.910)

Listed * Large * Relative deal size -4.439 ***(1.226)

Listed * Cross border 2.020 (2.074) 2.021 (1.394)

Market value (acquirer) -0.004 (0.009) -0.014 (0.010) -0.012 (0.020)

Deal value -0.066 (0.034) -0.014 (0.035) -0.007 (0.053) Industry match -0.230 (0.354) -0.224 (0.351) -0.237 (0.350) Cross border -0.721 (0.424) -0.976 ** (0.401) -0.966 ** (0.467) Experience -0.061* (0.334) -0.029 (0.334) -0.015 (0.367) Recession -0.840 (0.762) -0.899 (0.763) -0.916 (0.718) SIC2 0.405 (0.723) 0.451 (0.721) 0.484 (0.883) SIC3 0.267 (0.736) 0.309 (0.737) 0.280 (0.850) SIC4 -0.902 (0.839) -0.890 (0.842) -0.873 (0.917) SIC5 0.851 (0.871) 0.901 (0.872) 0.905 (0.941) SIC6 -0.155 (0.774) -0.049 (0.776) -0.031 (0.831) SIC7 0.189 (0.667) 0.185 (0.667) 0.182 (0.786) SIC8 0.184 (0.716) 0.219 (0.715) 0.197 (0.830) Year 2001 -0.267 (0.779) -0.379 (0.778) -0.403 (0.897) Year 2002 -0.235 (0.996) -0.366 (0.996) -0.355 (0.953) Year 2003 1.061 (1.151) 0.938 (1.150) 0.928 (1.071) Year 2004 1.782 * (0.941) 1.725 * (0.944) 1.728 * (0.896) Year 2006 1.478 (1.085) 1.330 (1.084) 1.318 (1.055) Year 2006 0.530 (1.063) 0.385 (1.061) 0.379 (1.040) Year 2007 0.048 (1.090) -0.133 (1.089) -0.160 (1.045) Year 2008 0.206 (0.983) 0.163 (0.975) 0.191 (0.938) Year 2009 1.770 (1.379) 1.513 (1.372) 1.523 (1.120) Year 2010 0.414 (1.341) 0.304 (1.323) 0.247 (1.151) Year 2011 1.934 (1.479) 1.758 (1.477) 1.742 (1.236) Year 2012 -0.120 (1.363) -0.290 (1.368) -0.291 (1.291) Year 2013 1.880 (1.478) 1.729 (1.478) 1.744 (1.238) Year 2014 1.609 (1.180) 1.503 (1.178) 1.469 (1.174) Year 2015 2.659 ** (1.228) 2.622 ** (1.224) 2.532 ** (1.075) Year 2016 1.456 * (0.832) 1.358 (0.831) 1.373 (1.125) Constant 1.091 (1.174) 1.114 (1.162) 1.111 (1.161) R-squared 0.018 0.022 0.024

Note: Estimation based on an OLS regression with Robust standard errors. Standard errors are reported in parenthesis next

to the coefficient estimate. *** indicates that the variable is significant at a 1% level, ** denotes significance at the 5% level and * represents significance at a 10% significance level. Dependent variable CAR is calculated as CARi=

∑2t=−2(Rit− (α̂ + βi ̂ Ri mt)), with α̂ and β being estimated based on Ri it= αi+ βiRmt+ εit. The variable ‘Listed’ has value

(27)

27 APPENDIX B

Table B.1.1: Residual diagnostics based on dependent variable CARBrown

Null hypothesis: residuals have a constant variance

P-value 0.000 0.000 0.000

Chi2 _168.58 _226.52 _340.14

Note: Breusch-Pagan heteroscedasticity test basted on the regressions for equation 4, equation 5, equation 6, respectively.

The dependent variable cumulative abnormal returns is calculated as the sum of differences between security returns and market returns.

Table B.1.2: Estimation results based on dependent variable CARBrown

Dependent variable: CARBrown _CARBrown _CARBrown

Listed -1,266 * (0,640) -0,691 (1,232) -1,322 (0,946)

Relative deal size 1,330 * (0,807) 1,927 ** (0,926) 2,008 *** (0,510)

Listed * Relative deal size -3,722** (1,937)

Listed * Large * Relative deal size -4,312 *** (1,228)

Listed * Cross border 2,339 (2,062) 2,342 * (1,395)

Market value (acquirer) -0,004 (0,010) -0,014 (0,010) -0,012 (0,020)

(28)

28

R-squared 0.018 0.023 0.024

Note: Robust standard errors are displayed in parenthesis to the right of the corresponding coefficient. *** indicates that the

variable is significant at a 1% level, ** denotes significance at the 5% level and * represents significance at a 10% significance level. The dependent variable cumulative abnormal returns is calculated as the sum of differences between security returns and market returns.

Table B.2.1: Residual diagnostics based on dependent variable CARAverage

P-value 0.000 0.000 0.000

Chi2 _140.51 _192.86 _296.08

Note: Breusch-Pagan heteroscedasticity test basted on the regressions for equation 4, equation 5, equation 6, respectively.

Dependent variable cumulative abnormal returns is calculated as the sum of differences between security returns during the event window and its 100 working day average.

Table B.2.2 : Estimation results based on dependent variable CARAverage

Dependent variable: CARBrown _CARBrown _CARBrown

Listed -1,476 * (0,893) -0,885 (1,348) -1,391 (0,974)

Relative deal size 1,800 ** (0,824) 2,351 ** (0,924) 2,464 *** (0,525)

Listed * Relative deal size -3,457 * (1,943)

Listed * Large * Relative deal size -4,279 *** (1,265)

Listed * Cross border 2,029 (2,171) 1,968 (1,437)

Market value (acquirer) 0,001 (0,010) -0,008 (0,011) -0,006 (0,021)

(29)

29

Year 2015 3,903 ** (1,305) 3,871 ** (1,303) 3,781 ** (1,109)

Year 2016 2,554 ** (0,963) 2,462 ** (0,962) 2,474 ** (1,161)

Constant -0,167 (1,237) -0,139 (1,225) -0,147 (1,197)

R-squared 0.019 0.022 0.024

Note: Robust standard errors are displayed in parenthesis to the right of the corresponding coefficient. *** indicates that the

(30)

30 APPENDIX C

Table C.1.1: Complete results robustness analysis

Dependent variable: CARMacKinlay _CARMacKinlay _CARMacKinlay

Listed -1,486 * (0,963) -1,034 (1,232) -1,626 * (0,969)

Relative deal size 1,734 ** (0,870) 2,299 ** (0,965) 2,382 *** (0,532)

Listed * Relative deal size -3,503 * (1,887)

Listed * Large * Relative deal size -4,027 ***(1,251)

Listed * Cross border 3,002 (2,270) 2,928 ** (1,480)

Market value (acquirer) -0,003 (0,010) -0,013 (0,011) -0,011 (0,022)

Deal value -0,049 ** (0,020) -0,003 (0,026) 0,003 (0,053) Industry match -0,202 (0,355) -0,195 (0,353) -0,209 (0,355) Cross border -1,245 * (0,643) -1,620 ** (0,552) -1,620 * (0,630) Experience 0,010 (0,336) 0,046 (0,336) 0,055 (0,373) Recession -0,877 (0,776) -0,960 (0,777) -0,976 (0,726) Cash payment 0,290 (0,351) 0,266 (0,351) 0,287 (0,351) Cultural distance 0,121 * (0,070) 0,132 * (0,068) 0,135 (0,086)

Completed on announcement date -0,283 (0,467) -0,228 (0,457) -0,203 (0,380)

SIC2 0,643 (0,720) 0,716 (0,716) 0,740 (0,911) SIC3 0,578 (0,726) 0,644 (0,726) 0,611 (0,881) SIC4 -0,552 (0,837) -0,507 (0,840) -0,498 (0,944) SIC5 1,141 (0,858) 1,223 (0,860) 1,216 (0,967) SIC6 0,233 (0,780) 0,382 (0,789) 0,390 (0,861) SIC7 0,457 (0,661) 0,486 (0,662) 0,477 (0,814) SIC8 0,501 (0,713) 0,561 (0,714) 0,534 (0,858) Year 2001 -0,159 (0,784) -0,270 (0,784) -0,292 (0,904) Year 2002 -0,240 (1,008) -0,374 (1,008) -0,371 (0,962) Year 2003 1,004 (1,171) 0,883 (1,170) 0,867 (1,082) Year 2004 1,828 * (0,956) 1,784 * (0,958) 1,781 ** (0,904) Year 2006 1,545 (1,101) 1,417 (1,098) 1,396 (1,068) Year 2006 0,542 (1,076) 0,407 (1,075) 0,396 (1,050) Year 2007 0,011 (1,105) -0,163 (1,104) -0,189 (1,056) Year 2008 0,183 (0,999) 0,161 (0,990) 0,184 (0,948) Year 2009 1,849 (1,404) 1,593 (1,397) 1,600 (1,133) Year 2010 0,507 (1,360) 0,386 (1,348) 0,320 (1,168) Year 2011 2,092 (1,531) 1,939 (1,528) 1,915 (1,261) Year 2012 -0,562 (1,359) -0,733 (1,363) -0,736 (1,313) Year 2013 1,682 (1,493) 1,493 (1,494) 1,507 (1,253) Year 2014 1,659 (1,209) 1,551 (1,207) 1,517 (1,195) Year 2015 2,685 ** (1,280) 2,688 ** (1,275) 2,596 ** (1,101) Year 2016 1,511 * (0,847) 1,469 * (0,847) 1,474 (1,139) Constant 0,778 (1,276) 0,766 (1,256) 0,749 (1,219) R-squared 0.018 0.023 0.024

Note: Dependent variable CAR is calculated as CARi= ∑2t=−2(Rit− (α̂ + βi ̂ Ri mt)), with α̂ and β being estimated based on i

Rit= αi+ βiRmt+ εit. Robust standard errors are displayed in parenthesis to the right of the corresponding coefficient. ***

(31)

31 Table C.1.2: Residual diagnostics for robustness analysis

P-value 0.000 0.000 0.000

Chi2 _235.41 _323.42 _440.42