Managerialism in cross-border Mergers and Acquisitions
An analysis of value creation and destruction related to the agency theory.
University of Amsterdam
Amsterdam Business School
MSc Business Economics Finance track Master Thesis
01-07-2017
Louren Matthanja van Garderen Supervisor: dhr. dr. V.N. Vladimirov
Statement of Originality
This document is written by Louren van Garderen 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 influences of agency problems in cross-border announcements of Mergers and Acquisitions (M&As). It is hypothesized that managers have incentives to destroy shareholder value by growing the firm beyond the optimal size or by lowering the variance of stock returns. Based on an event study of 21.500 completed transactions, of which 3.000 cross-border, executed by U.S. public acquirers between 1990 and 2010, it is shown that agency problems play an different role in cross-border M&As, influencing acquirer abnormal announcement returns negatively. It is then empirically shown that blockholders and managerial stock ownership positively influence announcement returns of cross-border M&As, mitigating these managerial incentives.
Table of contents
1 Introduction ... 5
2 Literature review ... 6
2.1 Wealth creation or destruction? ... 6
2.2 The Agency theory and managerialism ... 7
2.3 Hypothesis construction ...13
2.4 Factors influencing returns to acquirers ...14
3 Methodology and data ...16
3.1 Methodology ...16
3.2 Data ...18
3.3 Summary statistics ...21
4 Empirical results and discussion ...26
4.1 Acquirer CAR ...26
4.2 Regression model ...26
4.3 Robustness and limitations ...32
5 Conclusion ...34
6 References ...35
1 Introduction
This paper studies the relation between managerial objectives and announcement returns from cross-border Mergers and Acquisitions (M&As). The literature on managerialism and on cross-border M&As is discussed and integrated. Through an event-study of announcement returns, the effect of several acquirer, target and transaction characteristics on cross-border CARs are estimated.
The amount and volume of cross-border Mergers and Acquisitions (M&As) is ever growing because of deregulation, trade agreements and convergence of international cultures. 40% of transactions between 1999 and 2000 was a cross-border transaction (Hitt et al., 2001). In 2007 cross-border M&As are responsible for almost 45% of the total value in world-wide transactions (Erel, Lia & Weisbach, 2012).
M&As are amongst the largest investments firms make in their existence. Additionally, cross-border M&As have growing impact on national politics and labor markets. Although the potential for cross-border M&A acquirer synergies has been established (Mulherin & Boone, 2000; Bradley, Desai & Kim, 1988; Markides & Ittner, 1994; Morck & Yeung, 1992), the results are mixed. In 2000 for example, a few large loss deals offset any profits made on all other transactions (Moeller, Schlingemann & Stulz, 2005). Several characteristics, known from domestic acquisitions, are shown to be present in cross-border M&As as well, including type of payment and acquirer’s firm size. On the other hand, throughout the literature several idiosyncrasies have been established. Cultural differences, corporate governance structures and separated capital markets play a unique role in cross-border transactions (Erel et al., 2012). Shimizu, Hitt, Vaidyanath and Pisano (2004) conclude that the literature on cross-border M&As is spread thinly and identify the need for further research. Further analysis of this topic is important to substantiate the effects underlying returns, helping both academic researchers, executive managers and their boards in identifying value-creating strategies, thereby increasing the return on investment.
The aim of this paper is therefore to introduce managerial objectives and measure their influence on cross-border M&A acquirer returns. Both cross-border M&A synergies and managerialism have been studied before. It has been shown that managerialism and hubris have the potential to destroy potential for value creation in domestic transactions (Loderer & Martin, 1990; Roll 1986). But although several factors, including hubris and managerialism have been established in domestic M&As (Roll, 1986; Seth et al., 2002),
the relation to cross-border M&As has not been fully established. Seth. et al. (2002) do relate some of these factors to cross-border M&As, but don’t establish whether this effect is different between domestic and cross-border transactions. This analysis therefore adds to the literature by investigating the influence of managerial objectives on cross-border M&A transactions executed by U.S. public acquirers.
To analyze this effect, an event study is used to estimate Cumulative Abnormal Returns (CARs) to the Acquirer. As most attention has been focused on cross-border acquisition regarding non-U.S. acquirers (Collins et al., 2009) the focus on this paper is on U.S. bidders, acquiring both U.S. and non-U.S. targets.
This paper continues by discussing the current state of literature in section 2. In section 3 the methodology is presented, while also describing the dataset and summary statistics. In section 4 the results are presented, including robustness checks. Finally, section 5 concludes.
2 Literature review
The literature review consists of three parts. First, the wealth effects of cross-border M&A are discussed. Then known sources of synergy in M&As are discussed, in order to control for these effects. Finally, managerial objectives factors are introduced, in order to come to the hypothesis construction.
2.1 Wealth creation or destruction?
Whether M&As are associated with positive wealth effects has been an ongoing debate in the literature. Most authors find that M&As are able to create wealth trough generating synergies. For example, Mulherin and Boone (2000) show that shareholder wealth is created, and that a synergistic explanation is best suited. However, it is important to distinguish between target and acquirer. Generally, targets gain the majority of synergies. Bradley, Desai and Kim (1998) find that total returns average to 7.4%, but that acquirers at best break-even. Moeller et al. (2005) find that average acquirer return is -12% by using a dollar-weighted average. As this paper analyzes the influence of the acquirer’s managerial objectives, the focus of the literature review will be on wealth effects to acquirer’s shareholders.
Harris and Ravenscraft (1992) argue that several differences exists between domestic and cross-border M&As, including cultural, regulatory and capital market differences. They show that cross-cross-border M&As tend to be more R&D-oriented, and that cross-border M&As are more likely to occur after
domestic transactions. After controlling for all-cash bids and multiple-bid transactions, cross-border bidders pay 10 percentage points (50 percent) more than domestic acquirers. Kang (1993) also shows positive wealth effects to the Japanese acquirers, acquiring U.S. firms. Factors driving this wealth effect include dollar depreciation, a higher acquirer debt level, and stronger ties to financial institutions (Kang, 1993). Both higher debt and stronger ties to financiers indicate more manager’s scrutiny. According to the agency theory this leads managers to act more in the interest of shareholders. Markides and Ittner (1994) find that U.S. acquirers generate shareholder value through international acquisitions. Important factors include relatedness of activities, bidder industry characteristics, and bidder international experience. Morck and Yeung (1992) find a positive stock price reaction for cross-border acquiring firms that posses information-based assets. They argue that reverse internalization of the acquirer’s information-information-based assets to the target allows the acquirer to generate excess returns (Morck & Yeung, 1992). This effect was also shown in Harris and Ravenscraft (1991).
Cross-border M&As are, to some extent, comparable to domestic M&As. Both domestic and cross-border M&As have been extensively researched. However, returns from cross-cross-border M&A are still a topic of much debate, as opposed to returns from domestic M&As. It is especially interesting to research the relationship between transaction characteristics and returns to the acquirer. It is the aim of this paper to analyze some of these characteristics. The agency theory forms the basis for the characteristics discussed in this paper.
2.2 The Agency theory and managerialism
Managerialism is driven by conflicts of interest. This conflicts follow from the separation of ownership and control. This is what Coase (1937) and Jensen and Meckling (1976) describe as the principal-agent problem. The manager acts as the agent, hired by the shareholders (principal) to maximize their wealth. The agency problem is constituted by the misalignment of the utility functions of the principal and agent. Imperfect contracting leaves leeway for the manager to decide when unforeseen circumstances arise. However, this leavers room to act against shareholders’ interests (Schleifer and Vishny, 1997).
Throughout the corporate governance literature, many control mechanisms have been introduced that solve the principal-agent problem, including incentive plans, pay for performance and monitoring by the board of directors. But those measures are not always effective: the board is limited in the available
resources needed for monitoring and acts on limited information. Additionally, value-maximizing managers act to diminish the effectiveness of these mechanisms (Donaldson and Lorsch, 1983).
The agency theory states that managers optimize their own utility function, constraint on the company’s profits. This optimization of private utility constitutes a tradeoff between private and shareholders benefits when deciding on M&As. Conveniently assuming that managers earn private benefits from a merger, it follows that managers are willing to pay more for acquisitions than optimal from the perspective of shareholders. The assumption of private benefits trough M&As has been shown to hold throughout the literature. For example by improving the manager’s job security (Shleifer and Vishny, 1990; May 1995) or because of “empire building”:
A. Job security. Executive managers have several reasons to diversify the activities of the firm, sometimes at the expense of shareholders. Three reasons are discussed: diversification to improve job security and minimize employment risk, to pursue the independence and continuation of the firm, and to entrench themselves.
First, manager’s incomes are highly “undiversified”: as the manager’s professional reputation, career opportunities and generally the managers’ pay is dependent on the firms results, they face “employment risk”. While shareholders can diversify this risk, managers can’t. A risk-adverse manager therefore benefits from diversifying the firm’s activities. Diversification at the firm level is not beneficial for shareholders. Assuming perfect capital markets, they can build a portfolio that fits their diversification preferences themselves (Levi & Sarnat, 1970; Modigliani & Miller, 1958). Even worse, lowering the variance of the firm’s stock returns institutes a transfer of value from stockholders to bondholders due to risk shifting (Amihud & Lev, 1981). It follows that risk-averse managers have incentives to destroy shareholder value by diversifying the firms’ activities. Both domestic and cross-border M&As can serve this purpose.
Secondly, managers are inclined to pursue the independent operation of the firm, even when shareholder value is destroyed (Shleiefer and Vishny, 1990). Donaldson and Lorsch (1983) find from a case study of 12 firms that executives beliefs include the pursuit of independence and growth, regardless of shareholder value. This inclines managers to enter new lines of business in situations where shareholders
would be better off by liquidating the firm. Morck et al. (1990) empirically show that unrelated diversification leads to significantly lower announcement returns.
Thirdly, Schleifer and Vishny (1989) show that managers, especially when they face scrutiny, are inclined to attribute shareholder wealth to investments that are complementary to their own skills. This manager entrenchment gains the manager higher job security and more freedom. Especially when faced scrutiny, value-maximizing managers should buy and sell assets that raise their relative advantage over the next best manager in order to gain or regain competitive advantage. This often leads managers to expand excessively (Schleifer and Vishny, 1989). Additionally, managers pursuing diversification for job security purposes are willing to overpay, especially in situations of underperformance compared to the industry (Schleifer and Vishny, 1989).
Several papers have empirically shown a relation between agency problems and diversification. Denis, Denis and Sarin (1997) show that both managerial stock ownership and the presence of an outside blockholder are negatively related to diversification. Those factor are known to limit agency problems by respectively aligning incentives or introducing a monitoring function. May (1995) shows that managers with relatively more of their personal wealth vested in firms’ assets have a higher tendency to reduce risk trough diversification.
Concludingly, managers engage in diversification for private reasons, and have incentives to do so even tough diversification is not positively associated with shareholder value. Worse, diversification can destroy shareholder value by transferring wealth to bondholders due to lower return volatility or through overpriced acquisitions. Although corporate governance prescribes several solutions that alleviate the agency conflict, unforeseen situations, imperfect contracting and manager entrenchment still allow managers to pursue private benefits at the cost of shareholders. Diversification can be accomplished in several ways. For example by entering new, unrelated industries or by entering new geographic areas, for example trough unrelated M&As (targeting firms operating in other industries) or trough cross-border M&As, targeting firms in new geographic areas. If cross-border M&A is, at least partly, driven by job-related managerialism, it follows that value generated by cross-border M&A is not optimal.
B. Empire building.
Another example of managerial behavior is empire building. This theory states that managers prefer to run a big “empire” and therefore grow the firm beyond the size that maximizes shareholder value. This again boils down to the agency theory: if managers prefer to run big firms, than they have an incentive to do so at the cost of shareholders. There are several reasons why managers prefer to manage a big firm: First of all, it is suggested that growth is part of managers utility function, regardless of perquisites or compensation (Baumol, 1959). Second, running a big firm is associated with higher executive compensation. Managing a large firm is perceived as more challenging, justifying higher executive compensation. Additionally, a larger firm allows for more career opportunities as more key positions become available, diminishing competition and therefore increasing job security for the current top executives (Donaldson, 1984). Third, managers earn perquisites (for example corporate jets and chauffeur plans) from growing the firm. Although the resulting value destruction is not a priori shown (after all, the perquisites might allow the manager to be better at his job), the fact that the manager earns private rents from pursuing growth, specifically earning the corresponding perquisites, incentivizes the manager to grow the firm at the expense of the shareholders. Finally, managers might pursue growth in order to stay independent as a firm (Donaldson and Lorsch, 1983). Amihud and Lev (1981) argue that firms might pursue a merger after a hostile bid. If managers can negotiate a more certain position in the friendly merger, they are incentivized to destroy shareholder value by lowering the ask in merging with the friendly bidder.
Jensen (1986) provides empirical evidence for managers’ tendency to grow the firm beyond value-maximizing values. He shows that managers of oil firms chose to keep growing the firm, long after oil prices had collapsed in the 1980s, when shareholders would’ve been better off by ceasing growth. Jensen (1993) shows the same value destruction for large corporations.
Managers can pursue rapid growth in several ways, but M&As are the most obvious approach. Growing sales independently requires either marketing expenses, R&D investments or a combination of both. Furthermore, such projects are generally risky and have a long time horizon. M&As allow growth to be ‘bought’. Seth et al. (2002) empirically show that buying growth is one of the reasons managers pursue M&As, while also showing that buying growth is a value-destroying activity. Morck et al. (1990) find that
estimated bidder returns fall significantly when the difference in log sales over 5 years goes to 1, concluding that it is very expensive to buy growth.
In summary, managers have incentives to pursue rapid growth because it’s part of their utility function, because executive compensation is generally tied to firm size, in order to earn perquisites, and finally in order to pursue independence of the firm. M&As are an obvious way to pursue growth because of the speed and the relative certainty that growth will be realized.
Apart from the improvement of job security and empire building, another managerial reason has been identified in the literature. Either driven by overconfidence or plain error, hubris influences the returns from managerial decisions, and should be discussed in this context.
C. Hubris
Finally, Hubris could be the reason for overpaying acquirers. Roll’s Hubris hypothesis (1986) suggests that acquisitions occur because of mispricing of target firms. The takeover premium reflects the amount of mispricing. Under the hubris hypothesis, managers valuate potential target firms with some random error around the current market price. Therefore, some valuations will result below the market price, no public bid will be made, and the valuation error will never be observed. On the other hand, some evaluations will result in a valuation above the market price. Then an offer is made at the valuation amount. The public bid is announced and the valuation error is observed as the difference between the current market value and the transaction amount. Roll (1986) argues that the result is that the distribution of observed valuation errors is truncated at the current market price. The observed error will therefore be a positive takeover premium, which is absorbed by the target and paid for by the acquirer. The average overpayment is positively related to the average absolute valuation error.
In the original version of the hubris hypothesis, it is assumed that no synergies exist. Any observed takeover premia are then the observed tail of valuation errors. But Seth, Song and Pettit (2000) present empirical evidence for a more sophisticated version. They assume that synergies exist (the expected synergies are positive). Synergies are then a rational reason for managers to pursue M&As. However, valuation mistakes still result in a left-hand truncation of the distribution of valuation errors. Therefore, on average,
acquirers overpay for targets. This overpayment might or might not offset any synergies to the acquirer, but is a destruction of shareholder value nonetheless. Another implication of the hubris hypothesis is that managers of successful firms are more infected by hubris due to overconfidence. Due to arrogance, they overestimate synergies, resulting in a higher valuation error. As opposed to the general hubris hypothesis, where the distribution of errors is random, this results in a distribution of errors that is skewed to the right. The hubris hypothesis is mostly a conceptual model, empirical support is sparse. For example because the empirical challenge of measuring hubris is that it’s measured as the absence of a relation between the announcement returns and various sources of gains. Moreover, hubris is likely to be entangled with other deal motives, including synergy-seeking. Seth et al. (2002) find some support for the hubris hypothesis by disentangling deal motives and subsequently analyzing announcement returns for these motives. However, empirical evidence by Lang and Walkling (1989) shows the opposite: acquiring firms with a high Tobin’s q (a signal of managerial performance) earn greater announcement returns when taking over low q firms, then low q do when they take over low q firms, suggesting that the markets value a high q acquirer in takeover procedures. This opposes the adjusted hubris hypothesis that suggests that hubris occurs because of overconfidence.
Hubris can be dismissed as a random error that can’t be assessed or influenced. However, in practice, managers have influence on the level of hubris (Seth et al., 2002). Miscalculations can be caught by exercising costly efforts. From Lang and Walkling (1989) it follows that good managers are less obstructed by hubris. Apparently managers can influence, at least to some extent, the amount of hubris. Therefore, hubris is, in the context of this paper, considered to be influenced by agency problems as well.
There are several reasons why it is important to analyze the influence of hubris in the context of cross-border M&A. First, cross-border transactions add new layers of complexness to the valuation process. For example because the influence of cultural differences in the integration process or exchange rate influences. This results in larger valuation errors, and consequently in larger average overpayment. Second, as hubris is likely to be entangled in any transaction, it is important to consider the influence of hubris when interpreting results to other characteristics.
2.3 Hypothesis construction
As managers pursue job security, empire building or make erroneous decisions based on of overconfidence or miscalculations, shareholder value is destroyed. Therefore the following hypothesis is constructed:
H1: Agency problems are idiosyncratically related to cross-border M&A’s, leading to overpayment and negative abnormal announcement returns to the acquirer.
Based on the discussion of the literature, it is assumed that these agency problems lead managers to pursue diversification at the expense of shareholders. Diversification trough cross-border M&As is particularly fitting, since geographic diversification provides an efficient way to decrease stock return variance. Therefore, the following is hypothesized:
H2: Acquirers pursue unrelated diversification driven my managerialism, resulting in negative abnormal announcement returns in cross-border M&As.
Finally, from the literature discussion it follows that managers have incentive to engage in empire building, growing the firm beyond the level that is optimal to shareholders. Cross-border M&As fit in this strategy to pursue rapid growth. This leads us to the third and final hypothesis of this paper:
H3: Driven my managerialism and subsequent pursuit of excessive growth, acquirers overpay for rapidly growing targets in cross-border M&As.
Managerial behavior will be measured using factors indicating either the presence or absence of agency problems, including managerial stock ownership, blockholders and the level of financial constraints. The hypotheses specifically don’t state that negative acquirer returns occur. As discussed next, there are several factors influencing acquirer abnormal returns, allowing the possibility to generate shareholder value. Overpayment can occur in a transaction, regardless of whether wealth is created or destroyed in aggregate. Therefore, overpayment is measured as any negative announcement return following from managerial behavior. The return could then have been more positive (or less negative) if managers had decided in the interest of shareholders. This overpayment is destruction of shareholder value.
2.4 Factors influencing returns to acquirers
As discussed, cross-border M&As provide several unique challenges over domestic transactions. Differences in culture, regulation and economy trouble the integration process. On the other hand, a cross-border transaction provides direct access to a new market, and induces a transfer of knowledge to the acquirer. As literature on cross-border synergies is limited, it’s unclear whether the same effects apply (Shimizu, Hitt, Vaidyanath & Pisano, 2004). In this paper wealth effects stemming from managerialism in cross-border M&As will be discussed. In order to do so, the current state of literature on known wealth effects in M&As will be discussed. These effects will then be controlled for in the further analysis. The following factors are known to influence bidder announcement returns:
Payment method. Throughout the literature it is suggested that the announcement of a stock issue is surrounded with information asymmetry. Myers and Majluf (1984) argue that a one-sided information asymmetry between investors and management signals to investors that managers belief the firm is overvalued. Subsequently, Asquith, Bruner, and Mullins (1987) have argued that acquirer losses occur due to adverse information following from a stock financing of acquisitions. Moreover, cross-border transactions tend to get paid in cash more often, as targets shares mostly aren’t traded in the target’s country (Harris & Ravenscraft, 1990). These different sample characteristics need to be controlled for, to distinguish this effect from the takeover synergies. In line with Morck, Shleifer and Vishny (1990) a stock dummy will be used to control for this effect.
Tax considerations. Like domestic transactions, cross-border M&As can create shareholder value by leveraging a tax shield or deferred tax asset that couldn’t otherwise be put to use (Erel et al., 2012). Additionally, a cross-border transaction could be means to take advantage of more lenient tax systems. Erel et al. (2012) use a linear regression model on a sample of over 180,000 transactions (of which over 55,000 cross-border) to show that acquirers are more likely to be from countries with higher corporate taxes, while targets’ countries are more likely to have low tax rates (Erel et al., 2012). The influence of tax regimes on financial investments has also been established by Scholes and Wolfson (1990). They use the 1986 Tax Reform Act to show that a tax rule change can result in economically significant shifts in M&A activity, both domestically and internationally. Interestingly, Cakici, Hessel and Tandon (1996) find that the 1986
Tax Reform Acted generated no positive abnormal returns to acquirers. Concludingly, taxes influence cross-border M&As in several ways.
Accounting policies and corporate governance. Differences in accounting policies and corporate governance can provide incentives for cross-border M&As. Erel et al. (2012) show that purchasers tend to be from developed countries, while the inverse holds for targets. This is consistent with a governance argument, since development, accounting standards and corporate governance tend to be related (Erel et al., 2012). There is empirical evidence that accountancy policies influence cross-border M&A returns. For example, several papers found that the favorable accounting treatment of goodwill in targets’ countries generated possible returns to acquirers (Choi and Lee, 1992).
Relative size. Bidder announcement returns are higher when the relative size of the target firm is larger (Loderer & Martin, 1990). Loderer and Martin (1990) use a linear regression explaining target offer announcement effects. They show that acquiring firms pay too much for relatively large targets, resulting in a 1.7% acquirer abnormal announcement loss. These findings are consistent with the hubris hypothesis and the managerial objectives hypothesis.
R&D Intensity and information-based assets. Several authors have discussed that the possession of intangible assets can be a source of synergies trough asset sharing and reverse internalization in (cross-border) M&As (Morck and Yeung, 1992; Seth et al., 2002; Eun, Kolodny & Scheraga, 1996). The reverse internalization hypothesis states that a merger creates possible synergies through a transfer of knowledge, such as technical or managerial know-how from the target to the acquirer, whereas this transfer wouldn’t be possible otherwise. For example due to transaction costs or risk of misappropriation (Eun et al., 1996). Seth et al. (2002) distinguish between three M&A motivations (synergistic, managerial and hubris), and find that reverse internalization of intangible assets is an important and statistically significant factor in synergistic transactions. The same principle applies to asset sharing (or “forward internalization”), but in the other direction. By sharing the acquirer’s assets with the target, it benefits from economies of scale on the proprietary assets. Eun et al. (1996) find no empirical support for forward internalization, but they do find statistically significant support for reverse internalization. They argue that this could be a result of their selection. They selected U.S. targets and non-U.S. acquirers, and suggest that these non-U.S. firms buy U.S. firms because of the large market.
Economic considerations. Erel et al. (2012) find that the likelihood of M&As happening between a country pair is influenced by exchange rate returns. They find that mergers a priori likely to happen, become more likely after advantageous exchange rate movements. The conclusion is that other motivations, including synergies, drive M&As to be considered in the first place. However, the exchange rate movements influence the relative attractiveness of the target, increasing bidders propensity to acquire. Froot and Stein (1989) find that foreign buyers take advantage of the relatively cheaper proprietary information (such as technical knowledge), caused by exchange rate changes.
3 Methodology and data
3.1 Methodology
The aim of this paper is to estimate acquirer Cumulative Abnormal Returns, whereafter a linear regression model will be used to explain any abnormal announcement effects. The methodology is based on Bradley, Desai and Kim (1988): using the Capital Asset Pricing Model (CAPM), expected announcement returns are estimated, so that abnormal announcement returns can be calculated.
3.1.1 Cumulative Abnormal Returns and market model
Cumulative Abnormal Returns (CARs) are calculated using daily stock data, following the approach of Bradley, Desai and Kim (1988). To calculate daily stock returns, the corrected daily stock prices for the U.S. public acquirers are retrieved from Datastream, based on the DS code from Thomson One, and imported into Stata. A CAPM model is then estimated using 220 days of trading data, starting 250 trading days before the event (t=-250) and ending 30 days before the announcement (t=-30). Events where less than 100 trading days in the estimation period where available are eliminated. Additionally, events with 3 or more missing observations in the event window are eliminated. The risk-free and market returns are retrieved from Kenneth R. French’s website. The estimated CAPM beta is used to estimate event window predicted returns. The event window ranges from 10 days before to 5 days after the event (t=-10 to t=5), in order to capture any information leaking and overshooting (Eckbo, 2007). Additionally an event window ranging from two days before to two days after the event is estimated, with similar results. Appendix 1 shows that, although some information leaking seems to occur, most excess return occurs on the announcement day (t=0). Also,
there seems to be evidence of overshooting that is corrected at day 5. Using the predicted event returns, the abnormal residual to firm i on day t (ARi,t)is calculated using
𝐴𝑅, = 𝑅, − 𝛼 − 𝛽 𝑅 , (1)
where
𝑅, = realized return to firm i on day t, 𝛼 , 𝛽 = market model estimates for firm i, 𝑅 = return to the market portfolio.
The CAR is calculated as the sum of abnormal returns, given by AR, over the given event window. In line with Eckbo (2007) the event windows starts 10 days before the event, to capture all announcement effects, including pre-announcement information leaking. Consequently, the CAR is calculated as
𝐶𝐴𝑅 = ∑ 𝐴𝑅, . (2)
3.1.2 Linear regression model
CAR represents the announcement window cumulative abnormal return. A linear regression model will be used to explain the CAR in terms of target, acquirer and deal characteristics, while controlling for payment method, country effects, relative size and R&D intensity. The (simplified) regression equation is as follows:
𝐶𝐴𝑅 = 𝛼 + 𝛽 𝐴𝑔𝑒𝑛𝑐𝑦𝐹𝑎𝑐𝑡𝑜𝑟𝑠 + 𝛽 𝐷𝑖𝑣𝑒𝑟𝑠𝑖𝑓𝑖𝑐𝑎𝑡𝑖𝑜𝑛 + 𝛽 𝐴𝑙𝑙𝐶𝑎𝑠ℎ + 𝛽 𝑅𝑒𝑙𝑆𝑖𝑧𝑒 + 𝛽 𝑅&𝐷 𝑆𝑝𝑒𝑛𝑑 + 𝜖
(3)
Where AgencyFactors include several variables indicating the presence of absence of agency problems (as will be discussed in the next part), AllCash is a dummy indicating 100% of the deal price is paid in cash, RelSize controls for the effect of target size, R&D Spend is the amount the acquirer spend on Research and Development in the year prior to the announcement.
3.1.3 Discussion of methodology
There are several limitations inherent to the CAR approach. First, the announcement return might be anticipated by investors. This is partly corrected by including up to 10 days before the event, capturing
pre-announcement information leaking. However, the anticipation might happen long before, for example because of an announced acquisition program. Malatesta and Thompson (1985) show that stockholders of frequent acquirers partially anticipate future acquisition attempts. Announcement returns are then biased towards zero, since stockholders already anticipated the announcement (Malatesta and Thompson, 1985). A second limitation is that the timing of the event is not exogenous. The management of the acquiring firm chooses the announcement date, based on private information. The outcome of the CAR regression therefore gives CAR conditional on the management’s choice, and not the unconditional CAR. It is important to consider this in interpreting the results.
Another challenge is the possible introduction of heteroscedasticity by including accounting measures that relate to firm or deal size. Heteroscedastic standard errors will be used to control for heteroscedasticity.
Finally, as both accounting information and corporate governance are important indicators in the proposed regression model, the number of available observations drops to those firms where such measures are available. This introduces a selection bias, because the availability of these measures isn’t exogenous. For example, large firms are required to disclose more detailed information. Additionally, smaller firms, or firms trading over the counter (OTC) or on smaller exchanges are followed by less analysts and databases (Eckbo, 2007). Also, the selection bias might affect cross-border M&As differently. The effect of this bias is estimated in the data section of this paper, trough cross-sectional descriptive statistics. Also, control variables are used to capture this effet. Additionally, regressions are executed at cross-sections in terms of size deciles, in order to shed light on the distributional differences of our regression variables in different subsamples.
3.2 Data
The intent of this analysis is to study the influences of managerial objectives on the announcement returns to cross-border M&A acquirers. A sample of M&A deals is retrieved from Thomson One. The following initial criteria were applied: the deal is completed, a majority stake is acquired (51%), the acquirer is a U.S.-based public firm, and the target is either a U.S.-U.S.-based or a non-U.S.-U.S.-based public of private firm. Public acquirers are chosen because of the availability of stock data and firm characteristics, needed to reliably
a financial institution are excluded. Leveraged buyouts, spinoffs, recapitalizations, self-tenders, exchange offers, repurchases, acquisitions of remaining interest, minority Stake purchases and privatizations are eliminated as well.
The time 1990 to 2010 is chosen a comprehensive dataset of 32.282 transactions by 10.349 different acquirers. As discussed later, a large dataset is required because only a small subset of transactions is cross-border, and because limited data availability eliminates observations. A number of data items are retrieved from Thomson One, including announcement date, deal size, payment method and the cross-border mark. Additionally target and acquirer characteristics are retrieved where available, including nation, SIC codes and sales- and R&D measures. Datastream is used to retrieve acquirer characteristics in the year prior to the announcement, when this data was not available on Thomson One. Additionally, ExecuComp was accessed through CRSP to retrieve corporate governance indicators for our acquirers, including the presence of blockholders and the percentage of shares held by the firm’s executives. Sample statistics for all relevant variables are presented in part 3.3.
Finally, the following additional filters are applied. 1521 transactions from 820 acquirers are eliminated because they announce more than one transaction on a given date. For these transactions, the announcement effect can’t be attributed to a specific transaction. 4.581 transactions where the transaction value comprises less than 5% of acquirer’s assets are eliminated, because these transactions only add noise to the sample and have no economically significant impact, in line with Morck et al. (1990). 4.614 transactions where stock returns aren’t (reliably) available are eliminated. This results in a sample of 21.575 observations, of which 3.052 cross-border.
3.2.1 Variable construction
The focus of this paper is the effect of managerialism on the cross-border M&As. Based on the constructed hypotheses, the following variables are constructed. Unless stated otherwise, calculated and accounting variables are from the last 12 months prior to the announcement.
A. Dependent variable. The CAR is used as the dependent variable as discussed in part 3.1.1. Summary statistics for the CAR are presented in part 3.3. To correct for extreme values, the CAR is winsorized at 0.5%. Individual analysis shows that outliers tend to occur for OTC-traded or pink sheet firms. Due to a
low stock price (“penny stocks”) or low trading volume, the returns for these stocks tend to be more volatile, resulting in extreme values.
B. Independent variables. With regard to agency theory, the following variables are constructed. The amount of shares owned by the chief executive, including stock options is retrieved from ExecuComp. Additionally, an interaction between managerial stock ownership and a cross-border dummy is constructed. A blockholder dummy is constructed to equal 1 if a majority blockholder is present (50% or more). Again, an interaction dummy is generated. Finally, the ratio of net debt to acquirers assets is calculated, indicating the level of financial constraint. This ratio is winsorized at 0.02 to correct for outliers. Based on hypothesis 1, all three variables (including their interactions with the cross-border dummy) are expected to have a positive sign.
With regard to the diversification hypothesis, a RELATED dummy is calculated, equaling to 1 if the target and acquirer share any of the top three 4-digit SIC codes. A non-related acquisition is a sign of managerialist diversification. Additionally, managerial diversification is possibly present if an interaction between cross-border and agency cost indicators is shown. Based on hypothesis 2, it is expected that the RELATED dummy shows statistically significant positive effects. Additionally, it is expected that an interaction between CROSS, the cross-border dummy, explains additional variance. Another measure that will be included is the current level of diversification (SICCOUNT), measured by the count of SIC codes that the company currently operates in. Again, interaction with CROSS is an indication of idiosyncratic diversification related agency costs in cross-border M&As.
In order to measure the effect of empire building, the variables SALESGROWTHt,5 and
SALESGROWTHt,2 are generated, defined as the percentage change in sales over the last 5 respectively two
years. Five years is considered a better measure, since it requires more dedication to realize sales growth over five years then over two years. However, the availability of the sales variable six years before the announcement is limited, further limiting the amount of available targets1. Therefore both variables will be
estimated. Both variables are winsorized at 0.01. SALESGROWTH is both included as a standalone variable, and as an interaction with the cross-border dummy. The first captures announcement returns from
buying growth that are generic to both domestic and cross-border transactions, while the interaction term captures the cross-border specific effect. Based on the hypothesis it is expected the SALESGROWTH factor to have an statistically significant negative coefficient, and the interaction term of SALESGROWTH with CROSS to show an additional negative announcement effect.
C. Control variables. Several control variables are included: first, since many cross-border M&A aspects are country related (tax rate, accounting system, corporate governance), country dummies are generated to control for any systematical country differences. Because all the acquirers are U.S. firms it suffices to include a target-country dummy (as opposed to a country pair dummy). One country dummy is dropped to prevent perfect multicollinearity. Second, the natural logarithm of the assets of the acquirer firm is included to control for firm size effects (LN_ASSETSa). Third, an ALLCASH dummy is generated in line
with Morck et al. (1990). A positive coefficient is expected because all-cash offers are not associated with the announcement effects related to the information signaling effect from issuing equity. Forth, an HOSTILE dummy is generated to equal 1 if the deal was opposed by target’s management. Hostile takeovers are associated with lower bidder returns, so a negative coefficient is expected. Fifth, the amount of acquirer R&D spending in the year prior to the transaction (RDSPENDa) is included as an indication of the reverse
internalization hypothesis (Eun et al., 1996). Based on this hypothesis, a positive coefficient is expected, because reverse internalization is hypothesized to create synergies. Sixth, a CHALLENGED dummy is generated to equal 1 if there are more than one bidder, zero otherwise. A challenged bid is generally associated with a higher takeover premium and therefore lower acquirer returns.
3.3 Summary statistics
The summary statistics of variables relevant for testing the hypothesis are shown in Table 1. This table displays statistics for the full sample. Interestingly, a positive average CAR of 2.31% is found. Both the average and median CAR are positive across almost all cross-sections. The average deal size is $ 273 million, with a standard deviation of $ 2,054 million. The large variance in deal sizes can be explained by a few very large transactions. An example of this is the Time Warner takeover by America Online inc. (AOL) on January 10th. 2000 for $ 164,0 billion, a few months before the collapse of the Dot-com bubble. The abnormal stock return of America Online was -19.38% during the event period. The largest cross-border transaction in our sample is Reuters Group PLC, bought by Thomson Corp for $ 17,6 billion on May 7th,
2007. The abnormal announcement return of Thomson was -4.82%. These large transaction skew the distribution, but elimination is not appropriate because that would introduce selection bias. Summary statistics across size deciles are displayed in Table 3.
In Table 2 the differences between domestic and cross-border M&As are analyzed. The differences in CAR, transaction value and acquirer’s assets have the expected sign. However, the difference between CARs in domestic and cross-border transactions is neither statistically nor economically significant.
The average cross-border transaction size is $ 87 million smaller. This difference is statistically significant at the 5% level. However, the size distribution is highly skewed. Therefore differences can be caused by a small number of very large transactions. The average cross-border transaction has a slightly lower CAR, although this difference is not statistically significant. Interestingly, average cross-border acquirers are larger than domestic acquirers, while the average transaction value is $ 87 million less than an average domestic transaction.
It is important to note the difference between equally- and value-weighting the CARs. Moeller et al. (2005) show that a few very large transactions offset any dollar-gains generated by the majority of value-creating transactions. But while Moeller et al. (2005) use a value-weighted average (by calculating dollar returns), this paper uses an equally-weighted average, by focusing on relative returns to the acquirer. These CARs are therefore not directly comparable.
Table 1 Summary statistics for regression variables
VARIABLES Mean Median SD P25 P75 N
Dealsize ($ mil) 272.923 28.790 2,054.489 7.780 112.500 21,575 Assets acquirer ($ mil) 2,227.377 244.762 12,742.791 52.152 1,013.085 19,990
CAR acquirer 0.0213 0.0048 0.1552 -0.0470 0.0707 21,575
Net debt/Assets acquirer 0.1436 0.0788 0.2600 0.0046 0.2681 18,476
Ebit/sales acquirer -0.0107 0.0081 0.2405 -0.0064 0.0356 4,803
Executive stock ownership (%) acquirer 4.7772 1.9000 7.5825 0.6820 5.2000 3,837
Blockholder (50% or more) acquirer 0.0788 0.0000 0.2694 0.0000 0.0000 4,253
Price-to-book acquirer 2.9501 2.1400 2.6765 1.3700 3.4800 16,164
Sales growth 5 year (%) target 20.7147 11.5390 35.2514 2.2850 27.8465 2,004
Sales growth 2 year (%) target 46.7071 12.6108 142.0989 -2.1657 41.8609 4,337
ln(Salest-3/salest-1) target 0.2087 0.1188 0.5009 -0.0219 0.3497 4,337
Cross-border dummy 0.1415 0.0000 0.3485 0.0000 0.0000 21,575 Relatedness 0.5568 1.0000 0.4968 0.0000 1.0000 21,575 All Cash 0.2361 0.0000 0.4247 0.0000 0.0000 21,575 Hostile 0.0021 0.0000 0.0456 0.0000 0.0000 21,575 Challenged 0.0087 0.0000 0.0927 0.0000 0.0000 21,575
This balance table shows the mean, median, standard deviation and first and third quartiles of key parameters for the sample of M&A transactions. The sample consists of M&A deals retrieved from Thomson ONE for the period 1990-2010, requiring a completed deal where a majority stake is acquired (51%), where the acquirer is a U.S.-based public firm, and the target is either a U.S.-based or a non-U.S.-based public of private firm. Accounting information is retrieved from Thomson One. Execucomp is used to retrieve blockholders and executive share ownership information. CAR is the cumulative abnormal return, defined as the percentage change of the acquirer’s stock price during the event window. Blockholder is a dummy equaling 1 if a blockholder exists that holds more than 50% of the firm’s stock. Exec. Stock ownership gives the percentage of stock (including options) that is held by the company’s CEO. Sales growth is calculated as the percentage change in sales over resp. 5 and 2 years. The cross-border dummy equals 1 if the taget is a non-U.S. firm. Relatedness equals 1 if the target and aquirer share at least one of their top 3 SIC industry codes. All cash equals 1 if the acquirer used cash to pay for the transaction. Hostile equals 1 if the target’s management has at some point opposed the deal. Challenged equals 1 if there are more bidders.
Table 2 Sample differences between cross-border and domestic M&As
Difference in mean t-statistic
CAR -0.000696 (-0.23)
Transaction value ($ mil.) 86.78** (2.16)
Assets acquirer -1,377.2*** (-5.32)
N 21,575
t-statistics in parentheses * p<.10, ** p<.05, *** p<.01
This table shows a t-test for differences in averages between domestic and cross-border transactions. The sign of the difference in CAR suggests that domestic transactions have a lower average announcement return, as expected. But in contrast to the expectation, this difference isn’t statistically different. The average transaction value in a cross-border transaction is almost $ 87 million lower. This difference is statistically significant at the 5% level. Finally, the cross-border acquirer has almost $ 1,400 million more assets, significant at the 1% level.
Table 3 presents a cross-sectional analysis of the deal size and CAR over size deciles. Panel 1 uses deal size quintiles. Consequently, median and mean deal size are significantly different. Interestingly, the average and median acquirer CAR are lowest in the largest deal size quintiles. Besides the 5th quintile, the
median CAR is around 0.6% while the mean CAR ranges from 1.7% to 3.8%, suggesting that relatively high CARs occur across all deal sizes. The standard deviation across the quintiles suggests the same. Panel 2 displays characteristics across acquirer size quintiles. This is calculated using the acquirer’s book value of net assets in the year prior. The same pattern emerges, now showing a negative median and mean CAR in the 5th size decile. Interestingly, the 1st quantile shows a relatively high mean CAR while the median is only
slightly higher, suggesting that high CARs especially occur in smaller transactions. Not surprisingly, the standard deviations vary across deal size. The logic behind this is that, ceteris paribus, a larger deal will have a larger impact on the acquirer’s stock price. Variance is based on the absolute size of values, therefore the largest deal values result in higher variation. In the linear regression model, the effect of size on the estimates and residuals is further analyzed.
Not all variables are available for all observations. Including variables with missing values in a regression model would introduce a selection bias. Therefore, three samples are reported in Appendix 3. Firstly, the deal characteristics are summarized for the whole sample, as in Table 3. Then the samples where respectively R&D data of the acquirer and Sales growth data of the target is available are presented.
R&D data is available for 4,756 firms. Median and average transaction size show large differences, corresponding to a high standard deviation. The average CAR differs significantly at the 1% level compared to the whole sample (t=2.90).
Sales data is available for 2,004 targets. These targets tend to be very big. The median deal size ($) is over 9 times larger while the average is almost 6 times larger than the median and average deal size in the full sample. The CAR for these deals is negative and statistically different from the full sample at the 1% level with t=9.53.
These effects will be considered in the discussion of the results. Also, size variables will be included as introduced in part 3, to control for this effect.
Appendix 4 summary statistics are calculated across target country. Apart from the fact that 87.1% of transactions is domestic (the target country is U.S.), the three most frequent target countries are United Kingdom, Canada and Grenada. CARs differ between countries, but standard deviations are high so that the differences aren’t statistically significant.
Table 3 Cross-sectional deal and CAR summary statistics
Panel 1 - Characteristics per deal size quintile Median
dealsize
($ mil) Mean dealsize ($ mil) Median CAR Mean CAR SD CAR N
1 (smallest) 2.45 2.57 0.004 0.038 0.218 4319 2 10.30 10.71 0.006 0.022 0.157 4314 3 29.00 29.96 0.008 0.025 0.146 4344 4 82.00 88.88 0.006 0.017 0.119 4284 5 (largest) 394.23 1,233.21 0.000 0.004 0.110 4314 Total 28.79 272.92 0.005 0.021 0.155 21575
Panel 2 - Characteristics per acquirer size quintile Median
dealsize
($ mil) Mean dealsize ($ mil) Median CAR Mean CAR SD CAR N
1 (smallest) 4.50 19.53 0.019 0.062 0.241 3998 2 13.30 34.40 0.010 0.019 0.149 4000 3 33.00 95.63 0.006 0.012 0.119 3996 4 67.75 173.30 0.003 0.006 0.094 3998 5 (largest) 260.00 1,138.13 -0.001 -0.001 0.079 3998 Total 32.55 292.19 0.005 0.020 0.149 19990
Panel 3 - Characteristics for domestic and cross-border deals Median
dealsize
($ mil) Mean dealsize ($ mil) Median CAR Mean CAR SD CAR N
Domestic 28.00 285.20 0.005 0.021 0.155 18523
Cross-border 33.64 198.42 0.005 0.022 0.156 3052
Total 28.79 272.92 0.005 0.021 0.155 21575
This table shows deal characteristics over cross-sections on deal size, acquirer size (as measured by assets) and on domestic and on cross-border classification. For the deal size, the median and mean are given. CAR is the cumulative abnormal return, defined as the percentage change of the acquirer’s stock price during the event window. The CAR is described using median, mean and standard deviation. Finally, the number of observations per cross-section is given. Interesting to see is that the CAR is lower in the higher deal and acquirer size quintiles, consistent with existing literature.
4 Empirical results and discussion
This section contains the regression results for the three hypotheses. Additionally, robustness checks are presented in interpreting the results. Before discussing the hypothesis results, the cumulative abnormal returns found using the CAPM model are discussed.
4.1 Acquirer CAR
The first analysis based on the methodology is the calculation of CARs. Table 4 displays the result of a test on the statistical significance of the CARs, showing an economically and statistically significant announcement effect of 2.19%. In Appendix 2 a distributional graph of the CAR is presented.
Table 4 Test on statistical significance of CAR's
(1) (2) (3)
VARIABLES CAR CAR cross-border CAR domestic
Constant 0.0219*** 0.0231*** 0.0218***
(0.0012) (0.0031) (0.0012)
Observations 21,575 3,052 18,523
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
This table displays the result of a t-test on the CAR, with h0: CAR is equal to zero. With statistical significance of 1% it is shown that the CAR is different from zero in the sample and both cross-sections. This effect is economically significant as an announcement effect of 2.18% to 2.31% of acquirer’s stock price is shown.
4.2 Regression model
Table 5 displays the results for the regression models. Models 1 and 2 display the effect of blockholders and executive ownership on the CAR. Model 3 and 4 analyze the effect of diversification on announcement returns. Model 5 and 6 combine these effects. Model 7, 8 and 9 focus on the buying growth hypothesis. The results will be discussed, first for the control variables and then per hypothesis.
4.2.1 Control variables
The control variables are mostly in line with expectations based on current literature. A challenged transaction, meaning more than 1 bidder, results in a statistically significant lower CAR. The coefficient indicates a 2.69%-point decrease in acquirer CAR. However, in challenged takeover, the bidding process
to effectuate. The event window in this paper isn’t specified to capture this effect correctly. The coefficient should be considered in the light of this context, and cannot be taken on its own. The HOSTILE indicator is never statistically significant. This could be because only 45 firms in our sample fulfill this criterium. This could be because entrenched target managers protect their position, while acquirers stockholders see synergy potential, for example trough the target’s adaptation of corporate governance, at the target. As expected, the ALLCASH dummy had a significantly positive effect. The interpretation should not be that an all-cash bid is more profitable in itself, the dummy merely captures the absence of the negative announcement effect of an equity issue in these all-cash deals. Further, a statistically negative effect is found for the log of acquirer’s assets. An increase in acquirer’s assets of 10% will, ceteris paribus, result in a 0.076%-point lower CAR. The direction of this effect is in line with the agency theory, that firms with access cash will destroy shareholder value by pursuing negative-NPV projects. Unexpectedly, the cross-border dummy has no significant impact. It seems to be, that a cross-border acquisition on its own doesn’t create or destroy shareholder value, when controlling for other deal characteristics. Furthermore, the acquirer’s R&D over sales is included as a measure of synergy trough reverse internalization. The data don’t support this hypothesis: a statistically significant negative effect is shown to exist. However, this could be a selection bias because our sample, opposed to that of Eun et al. (1996) includes U.S. acquirers, where Eun et al. (1996) include U.S. targets. This selection bias is suggested and discussed by Eun et al. (1996, p.1578). It should be noted that the inclusion of the R&D/Sales variable highly limits the number of observations as discussed in the summary statistics (part 3.3). Finally, all models excluding model 3 and 6 include country fixed effects, whereas models 3 and 6 include year fixed effects. Although the constant is reported, is has no economic interpretation, since it simultaneously captures the effect of being a transaction corresponding to the multicollinearity-driven dropped country and year dummies, and any unexplained CAR.
4.2.2 Agency cost hypothesis
With regard to the agency cost hypothesis the following results have been found. The output of the regression models is displayed in Table 5. First, although the executive ownership variable is not statistically significant, the interaction between executive ownership and cross-border is at the 1% or 10% level, depending on the model used. According to this, an increase in executive ownership of 1%-point results in a 0.10 to 0.36 %-points higher CAR in cross-border M&As. This suggests that unique agency costs surround
cross-border M&As, that can be mitigated by the presence of blockholders or by increasing executive stock ownership.
The blockholder variable is less clear: in model 1, 5 and 6 the interaction with the cross-border dummy is statistically significant, while in model 2 only the blockholder dummy is significant. In all cases, the significance is limited to the 10% level. This could be to the relatively restrictive definition that a blockholder is only present if the blockholder owns 50% or more. The coefficient, albeit mostly insignificant, suggests that, as firms gain blockholder status, their CAR increases with 0.79 to 1.10 %-point in respectively model 1 and 2. In a cross-border merger this effect is even stronger: on top of this effect, an extra announcement effect of 4.95%-point is suggested by model 1 and 4.90% by model 5. However, although this effect is economically significant, the fact that less than 100 firms constitute the sample of cross-border acquirers with a blockholder indicates that this coefficient, although conclusive, can’t be extrapolated. Additionally, the fact that models 2 and 9 fail to find a significant effect, and the fact statistical significance is limited to the 10% level, are somewhat equivocal.
Concludingly, there is evidence that executive ownership mitigates specific cross-border related agency costs. An even stronger effect seems to apply to the presence of blockholders, although this effect is only significant at the 10% level. The data seems to support the first hypothesis.
4.2.3 Diversification hypothesis
Based on the literature review the expectation is that unrelated diversification has a negative effect on announcement returns. Some support for this diversification hypothesis is found. Diversification effects are shown in model 3 and 4. In contrast to Eun et al. (1996), the Related dummy shows no effect. The SIC count however, a proxy for current level of diversification, does have a statistically significant effect at the 1% level in models 5 and 6. In line with the diversification hypothesis, it seems that managers of diversified firms engage less in value-decreasing M&As. It could be that managers prefer cross-border M&As as a means of differentiating above differentiation across SIC codes. However, a statistically significant effect in cross-border M&As, as measured by the SIC Count interaction with Cross, is not supported by the data. Additionally, although firm size is controlled for, a higher SIC count could also be related to other acquirer characteristics, including prior takeover experience or a longer existence (age), factors that this model
4.2.4 Buying growth hypothesis
Sales growth is the measure for the buying growth hypothesis. Before the results can be interpreted, it’s good to mention that both the 5- and 2-year sales growth metrics severely limit the number of observations. As discussed in part 3.3 this introduces a selection bias, especially in model 7. Regardless, in model 7, the control variables behave as expected. By including control variables for firm size, deal characteristics and year fixed effects, the selection bias is at least partly mitigated.
There is some evidence for the buying growth hypothesis: while the 2 year sales growth measure shows no effect, the 5 year measure shows a small effect, significant at the 5% level. The difference between the 2- and 5-year measure could be because managers consider 2 years too short to conclude that further growth will occur, while this future growth is important to accomplish the managerial objectives. The negative coefficient supports the hypothesis that managers overpay for growth through M&As, but the effect is small: an increase in sales growth over the last 5 years of 1% results in a 0.03%-point decrease of the CAR. The median firm grew 12% in sales over 5 years, while the firm on the 3th quartile has grown 28%. This difference of 16%-points would result in a 0.5% lower CAR.
The hypothesis suggests that negative result follows from manager maximizing their private value, thereby destroying shareholder value. This effect is supported by the data, but a cross-border effect is not established. Concludingly, cross-border M&As are not differently affected by the buying growth hypothesis.
4.2.5 Empirical discussion
In model 5 and 6 the agency variables and the diversification variables are combined. Model 5 uses country fixed effects, while model 6 uses year fixed effects. No significant changes arise between model 5 and 6. The interpretation of the agency variables is unchanged between models 1-2 and 5-6, both in terms of significance and coefficients. However, the diversification effects lose the slight statistical significance visible in models 3 and 4.
The buying growth hypothesis (model 7) is supported by the data, but this effect is accompanied by a selection bias, complicating empirical conclusions. While managers don’t overpay for short term growing firms, the purchase of a long-term growing firm is accompanied by slight overpayment resulting in a negative CAR.
The first hypothesis, that additional cross-border agency aspects negatively influence bidder returns, is supported by the data. The aim of the diversification hypothesis is to relate this agency cost to diversification trough cross-border M&As. However, the data doesn’t support the diversification hypothesis, neither in domestic nor in cross-border transactions. The data does support the buying growth hypothesis, but there is no unique cross-border effect. The buying growth hypothesis therefore finds no empirical support.
Exec. Ownership -0.0003 -0.0001 -0.0002 -0.0002 -0.0001 (0.0005) (0.0002) (0.0005) (0.0005) (0.0007) Exec. Ownership*Cross 0.0035*** 0.0010* 0.0035*** 0.0036*** 0.0032** (0.0011) (0.0006) (0.0011) (0.0012) (0.0013) Blockholder 0.0079 0.0109* 0.0083 0.0087 0.0031 (0.0118) (0.0060) (0.0118) (0.0117) (0.0179) Blockholder*Cross 0.0495* 0.0317 0.0490* 0.0469* 0.0395 (0.0286) (0.0205) (0.0284) (0.0276) (0.0380)
Net debt/assets acq. 0.0226** 0.0125** 0.0228** 0.0231** 0.0191
(0.0103) (0.0052) (0.0103) (0.0103) (0.0139) SIC Count 0.0008 0.0014*** 0.0009 0.0009 0.0015* (0.0006) (0.0003) (0.0006) (0.0006) (0.0008) SIC Count*Cross 0.0002 0.0003 -0.0004 -0.0006 -0.0006 (0.0015) (0.0008) (0.0016) (0.0015) (0.0024) Related -0.0009 (0.0046) Related*Cross 0.0012 (0.0105)
Sales growth 2yr t (%) 0.0001 -0.0000 -0.0000
(0.0001) (0.0000) (0.0000)
Sales growth 5yr t (%) -0.0003**
(0.0001)
Sales grw 5yr t (%)*Cross 0.0002
(0.0002) R&D/Sales a -0.0029*** -0.0028*** -0.0028*** -0.0029*** -0.0017 (0.0009) (0.0009) (0.0009) (0.0009) (0.0012) Ln(Assets a) -0.0076*** -0.0099*** -0.0081*** -0.0101*** -0.0081*** -0.0085*** -0.0028* -0.0097*** -0.0094*** (0.0012) (0.0007) (0.0013) (0.0007) (0.0013) (0.0013) (0.0016) (0.0012) (0.0017) Cross -0.0199 -0.0022 -0.0152 -0.0021 -0.0183 0.0025 -0.0001 -0.0144 -0.0339 (0.0214) (0.0106) (0.0221) (0.0103) (0.0224) (0.0098) (0.0308) (0.0193) (0.0313) AllCash 0.0137*** 0.0035* 0.0130*** 0.0025 0.0135*** 0.0135*** 0.0302*** 0.0191*** 0.0260*** (0.0041) (0.0020) (0.0040) (0.0020) (0.0040) (0.0041) (0.0053) (0.0044) (0.0056) Hostile 0.0019 0.0028 0.0042 0.0020 0.0010 0.0032 0.0118 0.0004 0.0053 (0.0109) (0.0102) (0.0101) (0.0102) (0.0110) (0.0115) (0.0097) (0.0099) (0.0114) Challenged -0.0268*** -0.0213** -0.0266*** -0.0218** -0.0271*** -0.0279*** -0.0157* -0.0251*** -0.0247** (0.0099) (0.0093) (0.0098) (0.0092) (0.0099) (0.0095) (0.0094) (0.0085) (0.0107) Constant 0.0387 0.0676** 0.1919*** 0.0668** 0.1259*** 0.0675*** 0.1054*** 0.2084*** 0.1889*** (0.0243) (0.0287) (0.0250) (0.0292) (0.0252) (0.0142) (0.0327) (0.0249) (0.0355) Observations 4,489 18,476 4,590 19,277 4,489 4,489 1,919 4,064 2,550 R-squared 0.0349 0.0367 0.0313 0.0348 0.0351 0.0344 0.0281 0.0491 0.0461 Year FE NO NO NO NO NO YES NO NO NO
Country FE YES YES YES YES YES NO YES YES YES
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
This table summarizes the regression output. Models 1 and 2 analyze agency effects on the acquirers return. It is found that agency effects play a different role in cross-border M&As, concluding from the coefficients of the interaction between Executive ownership and Blockholders with Cross. The cross-cross-border dummy has no significant effect, suggesting that any differences between domestic and cross-border stem from characteristic differences. Model 3 and 4 introduce the differentation
variables. No cross-border effect is found. Model 5 and 6 summarize models 1-4. Model 5 includes Country fixed effects, while model 6 includes Year fixed effects. Models 7 and 8 discusse the effect of buying growth. The Sales growth (5 year) variable limits the number of observation, but is statistically significant. Model 8 and 9
4.3 Robustness and limitations
Several limitations and setbacks complicated this analysis. With Eun et al. (1996) we conclude that it is hard to explain acquirer wealth gains in M&As. Limitations will be discusses as either methodologic, or data-related.
Some methodologic limitations have been discussed in the methodology section of this paper, including the fact that the event date is not exogenous but based on private information. Apart from the endogenous event, this could disclose private information.
Another is that the CAPM model is used to predict returns. Any test on the returns is a joint test on the correctness of the CAPM model and subsequent excess returns. As presented in in part 4.1 the abnormal returns are statistically significant at the 1% level. However, this is assuming the CAPM is a correct prediction model. To test this assumption, another event window is predicted on a random sample of 3.245 firms. The chosen event window still ranges 15 days, but now starts 150 days before the announcement. The expectation is that during this arbitrary event window no consistent abnormal returns occur. A t-test is performed on the hypothesis that the average CAR is equal to zero. This hypothesis can’t be rejected with t(3244)=-0.83, p = 0.407. The CAPM therefore is considered is a suitable model for this analysis.
Additionally, abnormal-returns are expected to vary across cross-sections (Eckbo, 2007). For example because large firms tend to have more analysts or because large firms more often announce acquisition plans. Therefore, cross-sectional regressions are executed across firm size deciles, measured by accounting assets. The results are based on model 5. The variables of interest, including Executive Ownership*Cross and Blockholder*Cross generally are interpreted alike. The exception is Blockholder*Cross in model 1. As this coefficient is based on 11 observations2 the result is spurious, based
on a non-representative sample.
Although a rich set of data is available on M&As, the data collection introduced several challenges. Most importantly, the selection bias introduced by the absence of accounting variables for cross-sections of firms complicates generic conclusions to be drawn from this paper. Furthermore, accounting variables, where present, include outliers and require normalization. Scrupulous normalization trough winsorizing
