1 Market Power and Shareholder Returns from Mergers in Liberalized EU Markets
Name: M. van de Kraats
Student number: s1772066 Study Program: MSc IFM (DD)
Supervisor: Prof. dr. L.J.R. Scholtens Co-assessor: Dr. W. Westerman
ABSTRACT
This paper examines the relationship between market power increases through mergers and the shareholder returns upon the announcements of such mergers. I propose that this relation is positive and particularly strong in sectors which display characteristics that promote collusion and monopolistic markets. To empirically test this proposal, I perform an event study on a sample of mergers in the liberalized EU energy, telecommunications and air transportation sectors. The results indicate there is a significant, non-linear relationship between market power increases and the returns of shareholders of acquiring firms.
University of Groningen & Uppsala University, 19 June 2015
JEL: G34
2 1. INTRODUCTION
This study examines the relationship between market power increases and shareholder returns in mergers in the deregulated EU energy, telecommunications and air transportation sectors. The role of market power in mergers has been studied by a range of authors (e.g., Eckbo, 1983; Kim and Singal, 1993; Fee and Thomas, 2004), but empirical findings thus far have been inconclusive. To fill this research gap, I propose that previous findings have been incongruent because market power increases are only a significant factor in mergers in sectors that display certain industry characteristics.
In recent decades privatization and liberalization of formerly government-owned enterprises and markets has almost been universally adapted, including in the EU. In this study, privatization refers to the sale of state-owned companies, whereas liberalization or deregulation is defined as the opening up of monopoly markets to competition (Bauer, 2005). Sectors in the EU that witnessed particularly large developments in these areas include the energy (comprising electricity and gas), telecommunications and air transportation sectors. Following deregulation and privatization, as well as integration and internationalization, these markets experienced large sector consolidation through mergers and acquisitions (García and Trillas, 2013). The role of market power in these mergers is the main point of focus of this study. More specifically, this paper investigates whether in these three EU sectors, market power creation through mergers has resulted in positive abnormal shareholder returns upon merger announcements. To this purpose, I use a sample of mergers that were announced or rumored between 1997 and 2014. This study’s research subject is based on the conjecture that mergers both diminish the monitoring costs of oligopolies, which facilitates collusion, and reduce the number of active firms in a market (Perry and Porter, 1985; Stigler, 1964). These two factors are theorized to increase firm profitability, which, assuming capital markets are efficient, translates into higher shareholder returns. The terms mergers and acquisitions are used interchangeably in this text.
3 natural monopolies. Attributes such as these make it easier to exercise market power by exploiting oligopolistic market positions or by engaging in collusive behavior (Green, 2006).
Markets that are characterized by factors that promote monopolistic markets and collusion include the deregulated EU energy, telecommunications and air transportation sectors (Scharpenseel, 2001; Green, 2006; Torres and Bachiller, 2013). Factors that cultivate collusion and the creation of oligopolies in these markets include capital-intensive entry barriers, limited numbers of incumbent firms, limited (direct) substitutes and various types of capacity constraints and congestions (David and Wen, 2001; Morrell, 1998; Warf, 2003). These markets were traditionally characterized by natural monopolies and were commonly heavily regulated (Bilotkach, Clougherty, Mueller and Zhang, 2012). Moreover, state ownership has historically been very common in these sectors (Fink, 2011) and partial state ownership is still prevalent in most of these markets (Bauer, 2005; Jamasb and Pollitt, 2005). To examine the role of market power in mergers in these markets I pose the following research question: “What is the effect of the creation of market power through mergers on shareholder announcement returns in the deregulated EU energy, telecommunications and air transportation markets?”
To investigate the research question I develop a three-part methodology. First, I use stock price reactions of bidder and target firms to merger announcement and rumors to compute Cumulative Average Abnormal Returns (CAARs). Next, an OLS regression is employed to study to what extent changes in market power can explain these CAARs. Finally, stock price reactions of rival firms are examined to assess whether these competitors experienced wealth effects consistent with increases in market power. The results of this study indicate that market power increases through mergers are significantly related to shareholder returns upon merger announcements. More specifically, I find evidence of a non-linear relation between changes in market power and bidder returns. Additionally, I find that rival stock price returns are positively related to merger announcements. These combined results are consistent with the hypothesis that market power increases have a significant impact on shareholder announcement returns in deregulated EU energy, telecommunications and air transportation sectors.
4 The results of this study also have significant implications for managers in deregulated sectors. Firstly, the results suggest that market power increases are a relevant merger value creation channel in these sectors. For these managers, market power increases should be a valid consideration when deciding on potential merger attempts. Secondly, the results indicate that bidder shareholders in the energy sector generally experience negative returns. It may therefore be sensible for managers in this sector to refrain from pursuing an active acquisition strategy.
The remainder of this paper is structured as follows: first, I discuss the various value drivers in mergers in section 2. Here I also provide an overview of the deregulation and privatization processes of the three investigated sectors. Afterwards, I propose a number of hypotheses based on this discussion. Subsequently, to empirically test the hypotheses, various models and samples are constructed in section 3. Next, the results of the analyses are presented in section 4. Section 4 also presents the conclusions and implications of this study.
2. LITERATURE REVIEW
The effects of mergers on shareholder wealth have frequently been studied by measuring the impact of merger announcements on stock prices using event studies (e.g., Jensen and Ruback, 1983; Andrade et al., 2001; Moeller, Schlingemann and Stulz, 2004). Previous studies consistently find that shareholders of bidding firms earn normal or minimal abnormal returns. Andrade et al. (2001), who study a large sample of mergers and find returns similar to those of preceding studies, report three-day announcement returns of -0.7% for the shareholders of bidding firms. This can be explained by the so-called winner’s curse hypothesis, which proposes that winning bidding firms overestimate the benefits of a merger (Varaiya, 1988). On the other hand, shareholders of target firms generally experience substantial positive returns (Jensen and Ruback, 1983; Andrade et al., 2001); Andrade et al. (2001) find that target shareholders on average earn three-day announcement returns of 16%. The combined equity value of the bidder and target firms commonly increases by a moderate amount (Leggio and Lien, 2000; Devos, Kadapakkam and Krishnamurthy, 2009). Andrade et al. (2001) report that combined shareholders earn three-day announcement returns of 1.8%.
5 (Capron and Mitchell, 1997). Devos et al. (2009) also propose tax gains as a value creation channel. Another shareholder value source is the creation and exploitation of market power (Cox and Portes, 1998; Kim and Singal, 1993). The proposition that the creation of market power is a driver of shareholder wealth in mergers is commonly dubbed the market power hypothesis (Eckbo, 1983).
The market power hypothesis is based on the argument that a merger reduces the number of active firms in an industry. Industry models such as the Cournot model predict higher prices and higher firm profitability when the number of firms in an industry is reduced. This prediction is based on the conjecture that when a merger occurs, the new combined firm will typically reduce its production output below that of the two previously separate firms. Consequently, industry output decreases and industry prices increase, pushing industry profits upwards (Perry and Porter, 1985).
Additionally, a merger can diminish the monitoring costs of an oligopoly (Stigler, 1964). This facilitates (tacit) collusion, where individual firms coordinate production rates within an industry. Monitoring costs are decreased by mergers, as they reduce the number of firms in a sector. The fewer firms that are active, the more visible the actions of each individual firm are to its competitors. Consequently, if a cartel forms, this heightened visibility raises the probability of detecting cartel members that try to increase output or decrease prices. This increases the stability and profitability of a cartel (Eckbo, 1983). Effective collusion allows firms to reap monopoly (or monopsony) rents, which increase firm profitability. A positive relation between market power and profitability has empirically been established by examination of the effects of entry barriers and industry concentration on firm profits (Sullivan, 1974). Under the assumption of efficient markets, stock prices reflect the combined value of all the firm’s future cash flows. Thus, if an event increases profitability, this should translate into higher current stock prices.
6 rival firms do not benefit from efficiency gains and experience deteriorated relative efficiency instead. Similarly, antitrust filings should benefit rivals if mergers improve the efficiency of merging parties. Eckbo employs a market model to estimate the abnormal returns of rival firms, of which a sample is constructed using four-digit SIC codes. Eckbo finds that rival firms benefit from merger announcements and experience significant announcement day returns of 0.19%. However, Eckbo also finds that reactions to antitrust actions are not statistically significant. These findings do not clearly correspond with either of the two hypotheses. Eckbo suggests that the positive announcement reactions are caused by an additional effect; the information effect. The information effect proposes that an efficiency-driven merger can signal opportunities for competitors to improve their productivity through mergers. Eckbo concludes that as antitrust actions do not negatively affect rival stocks, the empirical evidence does not support the market power hypothesis.
Eckbo and Wier (1985), Fee and Thomas (2004) and Stillman (1983) use the same methodological approach as Eckbo (1983) and present comparable results. Additionally, Eckbo (1985) regresses the abnormal returns of rival firms against pre-merger market concentration and changes in market concentration, measured as the change in the Herfindal-Hirschman index (HHI). Eckbo finds that changes in market concentration are negatively related to rival stock price returns and rejects the market power hypothesis.
Subsequent papers, e.g. James and Wier (1987) and Devos et al. (2009), also find evidence that is inconsistent with the market power hypothesis. James and Wier (1987) study a sample of 60 US bank mergers to examine the role of various potential merger value drivers, including market power increases. To examine the market power hypothesis, the authors first use event study methodology to compute bidder announcement returns. Next, they use an OLS regression to assess whether bidder returns are positively related to increases in bidder deposit market shares. James and Wier find that the relevant coefficient is negative and propose that their empirical evidence is inconsistent with the market power hypothesis.
7 flows primarily through reductions in investment expenditures, as opposed to increased operating profitability. The authors conclude that anticompetitive effects are not a significant factor in mergers. Becker-Blease, Goldberg and Kaen (2008) do not explicitly discuss the market power hypothesis, but examine the stock price reactions to mergers of US electric power utility firms. The authors employ an OLS regression model to examine whether a range of deal, firm and industry variables can explain shareholder announcement returns. Becker-Blease et al. also include the change in the HHI as an explanatory variable, but find that the variable is not statistically significant, which seems inconsistent with the market power hypothesis.
Although the findings of the discussed studies are inconsistent with the market power hypothesis, they do not provide conclusive evidence to fully reject the role of market power in mergers. For example, the methodology based on competitor returns used by Eckbo (1983), Stillman (1983), Eckbo (1985), Eckbo and Wier (1985), and Fee and Thomas (2004) can be criticized on a number of grounds. Firstly, the efficiency and market power hypotheses are not mutually exclusive. As noted by Mullin, Mullin and Mullin (1995), a merger may increase efficiency, but may simultaneously still be anticompetitive. Thus, an insignificant rival stock price reaction to antitrust actions can be interpreted as evidence of the absence of both efficiency and market power effects, but also as evidence of an offsetting presence of both. Moreover, rival firms may be active in a large number of markets. If a merger impacts only a small proportion of the markets of a rival, the effect on the profits of a rival may be too small to generate significant stock price reactions (McAfee and Williams, 1988).
8 The discussed papers that do not (fully) rely on rival stock price reactions to reject the market power hypothesis also have methodological weaknesses that limit the strength of their claims. As noted by Cox and Portes (1998), cash flow predictions rely on many subjective and debatable assumptions. Consequently, the results and implications of Devos et al. (2009) should be interpreted with caution. Also, James and Wier (1987) and Becker-Blease et al. (2008) only examine whether there is a linear relation between market concentration changes and shareholder gains. According to Schumann (1993), the relation between market power increases and shareholder returns is likely to be non-linear. The omission to test for a non-linear relationship thus limits the strength of the findings of Becker-Blease et al. (2008) and James and Wier (1987).
Besides the discussed weaknesses of the studies that reject the market power hypothesis, a number of studies have found empirical evidence consistent with the hypothesis. Kim and Singal (1993) examine merger-induced market power changes in the US airline industry by comparing non-merged routes with merged routes. The authors test the role of market power using product price data. They argue that efficiency gains would decrease fare prices, while market power increases would raise prices. Kim and Singal pose that the airline industry provides a unique possibility for this type of analysis as each route can be seen as a separate market; the routes not impacted by a merger can act as a control group to account for industry-wide shocks. The authors study a total of 14 mergers and find that prices on merged routes increased in comparison with the control group. Moreover, the authors employ a range of regressions to examine explanatory factors of price increases. They find that changes in market concentration, measured as the change in HHI, are positively and significantly related to price increases. Kim and Singal conclude that market power increases have overshadowed efficiency gain effects in the airline sector and that the investigated mergers have resulted in a wealth transfer from consumers.
9 market power increases being the primary merger motive. Hankir et al. conclude that, according to capital markets, increased market power is an important motive in bank mergers.
Sapienza (2002) also examines bank mergers, but instead focuses on post-merger interest rate effects. She discovers that interest rates charged by the combined firm commonly decrease after a merger, which is consistent with efficiency gains. However, Sapienza also finds that if the target firm has a large market share, interest rates do not decline. She concludes this result is consistent with the exercise of market power.
Fraunhoffer, Freytag and Schiereck (2014) analyze 139 mergers in the US and German energy utility markets, focusing on the effects of market power increases. The authors use event study methodology to study rival stock price reactions upon merger announcements. Fraunhoffer et al. find that for US mergers, competitors generally experience negative returns. These returns are inconsistent with the market power hypothesis and support the efficiency hypothesis. However, for German mergers, it is found that competitors experience positive returns. These competitor returns are consistent with the market power hypothesis. Fraunhoffer et al. conclude that to increase market power is an important merger motive in the German energy utilities industry, but not in the US energy utilities sector. The authors suggest that the differences between US and German mergers are explained by regulatory differences. They argue that deregulation of energy wholesale markets has progressed much further in Germany, and that market power may not result in increased profitability if sectors are highly regulated.
García and Trillas (2013) study the effects of deregulation on mergers in the EU electricity sector. The authors examine target, bidder and rival shareholder returns for 11 merger announcements using an event study approach. García and Trillas find that target firms experience positive returns, while bidders earn negative or normal returns. Moreover, the authors find that the returns of competitors are much higher in countries with high market concentration. Consequently, the authors suggest that market power has a substantial impact on competitor stock price reactions to merger announcements.
The analysis of existing studies on market power and mergers highlights two major issues. Firstly, it indicates that it is highly difficult to assess the effects of market power by solely examining the stock price reactions of rival firms. In the current study I therefore examine the effects of changes in market power more directly. Nevertheless, I also examine rival stock price reactions to improve comparability with other studies and to reinforce the results of the primary analysis. These procedures are further elaborated on in the Data and Methodology section.
10 capacity constraints and a limited number of active companies (David and Wen, 2001; Green, 2006). Such factors increase the potential for oligopolistic behavior and successful collusion. Consequently, mergers in such industries may well be able to create value by increasing market power, as opposed to mergers in sectors where sustaining a non-competitive market is more difficult. Based on the proposal that market power motives are more important in certain specific industries, this study focuses on three industries that are characterized by the aforementioned factors that promote market power exploitation. Three sectors that are characterized by a strong presence of the aforementioned industry-specific factors are the deregulated EU energy, telecommunications and air transportation sectors (García and Trillas, 2013; Torres and Bachiller, 2013; Scharpenseel, 2001). These markets are also characterized by high degrees of market concentration (Green, 2006; Torres and Bachiller, 2013; Barros and Peypoch, 2009). Moreover, a primary reason why markets were previously regulated is that the potential for market power exploitation is high due to natural monopolies (Armstrong and Sappington, 2006). Deregulation and integration of these markets has largely taken place in the late 1980s and 1990s with the goals to create an internal market, to promote competition, to protect consumer interests and to increase economic efficiency (Becker-Blease et al., 2008; Belloc, Nicita and Parcu, 2013). These liberalization and deregulation trends were frequently preceded or accompanied by (partial) privatization (Jamasb and Pollitt, 2005).
In the EU energy, telecommunications and air transportation industries, deregulation and privatization have been followed by merger waves and increased ownership concentration (García and Trillas, 2013), even though regulatory institutions have commonly been established to monitor the liberalization process and promote competition (Ottow, 2012). Jamasb and Pollitt (2005) and Bauer (2005) state that these regulators are frequently ineffective in prohibiting increases in market concentration. With the general background of liberalization and deregulation in the EU established, I now turn to discussing the individual industries of interest and why market power increases through mergers may play a significant role.
The EU Energy Sector
11 state ownership in the energy sector is still prevalent in many member states.
One of the primary goals of the deregulation packages was to improve economic efficiencies and operating synergies by exposing previously inefficient firms to market pressures (Becker-Blease et al., 2008). However, previous studies have shown that deregulation has not led to effective competition in most energy markets and has generally resulted in increased ownership concentration instead (Green, 2006; García and Trillas, 2013). Moreover, Boudier and Lochard (2013) state that deregulation first allowed the entry of new operators, but subsequently led to a consolidation process and increased ownership concentration via asset distribution and in particular mergers. Other authors support the observation that the increased ownership concentration was primarily achieved via a large takeover wave (e.g., Datta, Kodwani and Viney, 2013; Verde, 2008). Industry specific factors that have prevented the development of effective competition since deregulation include high entry barriers in the form of capital-intensive investments (e.g., power generation stations) and the absence of direct substitutes. Other factors are the limited number of producers, the difficulty of storing energy, and transmission constraints and congestions (David and Wen, 2001; Green, 2006).
Based on these observations, I propose that market power increases are an important factor in mergers in the EU energy sector. This is supported by various empirical studies that suggest that market power plays an important role in energy markets. For example, García and Trillas (2013) find that the merger announcement returns of rival firms are strongly dependent on the degree of market concentration in EU electricity mergers. Moreover, David and Wen (2001) demonstrate that there is significant market power potential in the electricity sectors of Australia, the US and the UK. Finally, Datta et al. (2013) state that market power has increased substantially in the EU energy market due to acquisitions.
The EU Telecommunications Sector
12 end of the 1990s (Bauer, 2005). Due to the discussed industry characteristics, monopolies and market power are still present in the EU telecommunications sector (Torres and Bachiller, 2013). Based on these observations, I propose that market power increases are also an important factor in mergers in the EU telecommunications sector.
The EU Air Transportation Sector
Privatization and deregulation of the air transportation market occurred in phases in the 1980s and the 1990s. The first legislative package was introduced in 1988, the second one in 1990 and the most significant third one in 1993, which did not come into full effect until 1997 (Barros and Peypoch, 2009). Moreover, a large trend of privatization of airports occurred in the EU from 1987 onwards (Bilotkach et al., 2012).
Following these developments, the air transportation market witnessed rapid consolidation (Morell, 1998). Industry-specific factors that have prohibited the development of effective competition since deregulation include the dominance of partly state-owned carriers (Scharpenseel, 2001) and the importance of economies of scale (Barros and Peypoch, 2009). Issues with airport infrastructure, including congestions and slot allocation problems also play an important role (Scharpenseel, 2001). Finally, monopolies in input markets, which substantially increase the costs of potential new entrants, also prevent developments towards effective competition (Morell, 1998). Based on the prevalence of factors that promote market power exploitation, I propose that market power increases are also an important factor in mergers in the EU air transportation sector. The next section of this study presents a set of testable hypotheses based on the various propositions made in this text.
Hypotheses
13 9.55%, while bidders experience returns of -0.25%. Datta et al. (2013) find three-day returns of 7% for target firms and returns of -0.1% for bidders. For non-regulated sectors, Andrade et al. (2001) report returns of 16% and -0.7%. Cox and Portes (1998) argue that the low returns in (de)regulated markets are due to a higher probability that a deal is eventually prohibited by regulators. Based on the non-positive bidder returns commonly found by previous studies, I propose the following null hypothesis:
H10: Bidder shareholders experience positive abnormal returns upon the announcement or rumor of a
merger in the EU energy, telecommunications and air transportation sectors. The corresponding alternative hypothesis is:
H1A: Bidder shareholders experience normal or non-positive abnormal returns upon the announcement or
rumor of a merger in the EU energy, telecommunications and air transportation sectors.
If the null hypothesis H10 is rejected, this would indicate that bidder shareholders experience non-positive
or normal returns. This would imply that merger gains mostly accrue to the shareholders of target firms or that the merger destroys shareholder value. The former would be in line with a large collection of studies, including Andrade et al. (2001) and Jensen and Ruback (1983).
The second hypothesis examines the shareholder returns of the target firm. Based on the substantial positive target returns that previous studies report (e.g., Leggio and Lien, 2000), I propose the following second null hypothesis:
H20: Target shareholders experience normal returns or negative abnormal returns upon the announcement
or rumor of a merger in the EU energy, telecommunications and air transportation sectors. The corresponding alternative hypothesis is stated as follows:
H2A: Target shareholders experience positive abnormal returns upon the announcement or rumor of a
merger in the EU energy, telecommunications and air transportation sectors.
If the null hypothesis H20 is rejected, this would indicate that target shareholders experience positive
14 H30: The combined bidder and target shareholders experience normal returns or negative abnormal
returns upon the announcement or rumor of a merger in the EU energy, telecommunications and air transportation sectors.
The corresponding alternative hypothesis is therefore:
H3A: The combined bidder and target shareholders experience positive abnormal returns upon the
announcement or rumor of a merger in the EU energy, telecommunications and air transportation sectors. If null hypothesis H30 is rejected, this would imply that merger announcements increase total shareholder
wealth. Put differently, it would indicate that markets expect mergers to create value. As discussed, possible channels for mergers to create shareholder value include synergies and market power increases (Schumann, 1993). To examine whether market power increases play a significant role, I formulate a fourth null hypothesis. This hypothesis is primarily based on the sector overview, which proposes that the potential to exercise market power in the three investigated sectors is substantial. Arguably, mergers in these sectors can therefore create shareholder value by increasing market power. Higher market power allows firms to set prices above cost and increase profitability (David and Wen, 2001). According to the (semi-strong) efficient market hypothesis, where market prices reflect all publicly available information (Cox and Portes, 1998), higher expected profitability is directly translated into higher shareholder wealth, i.e. stock prices. Based on the conjecture that market power increases are a significant value driver in the discussed deregulated EU sectors, I propose the following fourth null hypothesis:
H40: Changes in market power through mergers are not related to abnormal shareholder returns.
The corresponding alternative hypothesis is:
H4A: Changes in market power through mergers are positively related to abnormal shareholder returns.
If the null hypothesis H40 is rejected, this would imply that changes in market power are (positively)
15 are expected to experience positive abnormal returns upon merger announcements and rumors. Thus, I pose the following fifth null hypothesis:
H50: Rival firms experience no positive abnormal returns upon the announcement or rumor of a merger in
the EU energy, telecommunications and air transportation sectors. The corresponding alternative hypothesis is:
H5A: Rival firms experience positive abnormal returns upon the announcement or rumor of a merger in
the EU energy, telecommunications and air transportation sectors.
If the null hypothesis H50 is rejected, the empirical results would be consistent with the market power
hypothesis. However, as argued by Eckbo (1983), positive rival stock price reactions can also be caused by information effects. It is not possible to distinguish between these effects based on announcement and rumor returns. However, the results of this analysis can be used to verify consistency with the results from the test of the fourth hypothesis. In the next section I construct a data sample to empirically examine the role of market power in mergers. Moreover, I develop three different models to empirically test the five formulated hypotheses.
3. DATA AND METHODOLOGY
This section describes how the hypotheses developed in the preceding section are tested empirically. First, I discuss how this study’s sample group is constructed. Next, the methodology that is used to compute shareholder returns is described. I also discuss how the shareholder returns are subsequently used to test hypotheses 1-3. Next, I develop a regression equation to examine the relation between merger event shareholder returns and changes in market power. Finally, I create a model to study the impact of merger announcements and rumors on the stock prices of rival firms.
16 The sample used for this study was obtained from the Zephyr database on mergers and acquisitions. The sample was created using a number of selection criteria. Firstly, I collected all mergers where the merging parties were active as an EU energy, telecommunications or air transportation company, based on their NACE code or Zephyr classification. Moreover, the bidder’s and target’s main activities were required to be within the same industry. The resulting sample consisted of 12,015 mergers. Secondly, I removed all mergers where one of the participants was not publicly listed at the time of the merger announcement. This reduced the sample size by 10,054 observations. Next, deals where the bidder did not aim to acquire at least a 50% stake with a maximum pre-announcement stake of 50% were removed (Akdoğu, 2009). This reduced the sample by 1,778 deals. Next, mergers that were announced or rumored outside of the period spanning 1 January 1997 to 31 December 2014 were removed. This reduced the sample size by 4, resulting in a sample of 179 observations. I use the sample period of 1997-2014 as the most significant deregulatory developments of the energy and air transportation did not come into full effect until 1997 (García and Trillas, 2013; Barros and Paypoch, 2009). Additionally, the Zephyr database does not cover mergers that occurred before 1997.
As recommended by Park (2004), I control the sample for confounding events, specifically CEO changes, profit warnings and stock splits. This was done by manually checking the company news headlines provided by the Zephyr database for the 250 days preceding merger rumors and announcements. In total, I identified seven such events. I corrected the sample by removing the affected firms from the sample on the dates of the confounding events. This procedure is recommended by Park (2004) for event studies with modest sample sizes, as it both corrects for confounding events and minimizes the loss of observations.
17 An overview of the bidder and target characteristics per sector is provided in table 1. For the purpose of this table, no distinction is made between rumored and announced deals. From the table it is clear that bidder firms are generally much larger than target firms. Moreover, a comparison of the mean and median values suggests that several particularly large firms dominate the sample. More detailed tables on the distribution of rumored and announced deals and the sample construction steps are found in appendix A. TABLE 1: Selection of Bidder and Target Firm Characteristics – Aggregate Sample and Per Sector
Aggregate sample Bidder Target
M ean Q1 M edian Q3 M ean Q1 M edian Q3
Total assets (in €m) 22,156 1,276 4,823 26,689 5,177 59 1,012 4,155 Sales (in €m) 4,260 23 426 3,062 1,299 14 186 734 M arket capitalization 15,098 2,311 8,244 24,254 12,032 2,632 5,362 24,669 Return on assets 0.0% -0.5% 0.4% 4.5% 0.1% -0.5% 1.0% 30.1% Return on sales -4.8% -17.5% 5.6% 22.8% -255.8% -2.5% 6.7% 30.1% Number of firms 107 135 Number of observations 179 179
Energy Bidder Target
M ean Q1 M edian Q3 M ean Q1 M edian Q3
Total assets (in €m) 30,568 1,881 15,125 51,118 7,984 190 2,261 6,643 Sales (in €m) 4,204 108 655 3,068 1,263 28 314 1,231 M arket capitalization 19,477 8,244 15,964 28,574 22,330 24,669 24,669 24,669 Return on assets 2.2% -0.5% 0.7% 5.0% 0.7% 0.0% 3.0% 8.2% Return on sales -3.2% -19.1% 5.6% 34.8% -146.9% 3.2% 10.3% 32.8% Number of firms 36 45 Number of observations 57 57
Telecom Bidder Target
M ean Q1 M edian Q3 M ean Q1 M edian Q3
Total assets (in €m) 28,533 1,249 6,007 40,501 5,231 399 1,198 7,867 Sales (in €m) 6,879 27 708 11,087 1,121 4 80 573 M arket capitalization 18,514 3,888 9,128 23,840 4,556 2,399 2,833 5,362 Return on assets -4.4% -0.4% 0.2% 4.3% -2.0% -0.3% 0.1% 3.2% Return on sales -13.0% -9.6% 0.8% 19.1% -537.4% -0.5% 9.8% 29.8% Number of firms 34 48 Number of observations 60 60
Transportation Bidder Target
M ean Q1 M edian Q3 M ean Q1 M edian Q3
18 To determine the impact of merger announcements or rumors on shareholder wealth, this study employs the same event study methodology as described by Armitage (1995) and Duso, Gugler and Yurtoglu (2010). Event studies are a common approach for measuring effects on stock prices (García and Trillas, 2013) and have empirically been shown to be a useful method for the (ex-ante) competitive analysis of mergers (Duso et al., 2010). Event studies have been frequently used for studies on deregulated industries, for example by Berry (2000) and Leggio and Lien (2000). Berry (2000), who examines a set of US electric utility mergers, finds that bidder target shareholders suffer losses, target shareholders benefit, and combined shareholders experience a net benefit. It is also found that the economic benefits are smaller than those earned in non-regulated mergers. Leggio and Lien (2000) also examine merger announcement returns in the deregulated US electric utility sector. They find that bidder firms earn significant negative returns, while target firms earn significant positive returns. Similar to Berry (2000), Leggio and Lien also note that these returns are smaller than those reported in non-regulated sector mergers. A possible explanation for the low returns is the higher probability that a deal is eventually prohibited by regulators. According to Cox and Portes (1998), as stock prices are determined by the interactions of multiple (presumably) rational actors, methods based on stock prices are generally less prone to individual bias and less reliant on debatable assumptions than alternative methods such as discounted cash flow analysis. Event study methodology thus seems an appropriate approach to examine merger wealth effects.
Event study methodology is based on the idea that stock prices represent the present value of the future profits of a firm (Eckbo, 1983). Therefore, the change in equity value (i.e. shareholder returns) upon the announcement of a merger can be interpreted as a measure of the additional discounted profits that are to be expected from this merger. The main issue is then to determine how the equity value would have evolved had the event not occurred, i.e. the counterfactual. The idea of event studies is to employ a model to predict this counterfactual (Duso et al., 2010). One of the most commonly used models to predict this counterfactual is the market model. Although alternative models exist, e.g. the capital asset pricing model, Armitage (1995) finds that the market model performs as well, if not better, in most circumstances. The market model is the following one factor OLS regression equation:
𝑅𝑖𝑡 = 𝛼𝑖+ 𝛽𝑖𝑅𝑚𝑡+ 𝑒𝑖𝑡 (1)
where the random error term eit is a normally distributed error term and the R variables are daily
continuously compounded returns. This model is used to obtain estimates of αi and βi, which can in turn
be used to predict the counterfactual, i.e. the “normal” return for firm i (Duso et al., 2010). Here, αi is the
intercept of the regression equation, while βi is the coefficient that represents the systematic risk of
19 (Modigliani and Modigliani, 1997). In this study local primary composite market indices are used to represent the market portfolio Rmt (e.g., the AEX, DAX and FTSE). This choice is based on a review of
event studies by Armitage (1995), who state that a composite market index is an appropriate proxy for the market portfolio. As mentioned, the market model uses daily continuously compounded returns. These returns are computed as:
𝑅𝑡 = ln ( 𝑅𝐼𝑡
𝑅𝐼𝑡−1) (2)
where ln is the natural logarithm and RI represents a stock’s return index. This study estimates the parameters (i.e. αi and βi) over a period of 250 trading days before the announcement or rumor date
(Leggio and Lien, 2000), using daily return index data from Thomson’s Datastream database. This is consistent with the estimation period range from 100 to 300 days for daily studies recommended by Armitage (1995). If the announcement or rumor date is on a non-trading date, the next trading day is taken as the relevant date. Subsequently, abnormal returns (ARs) are calculated by subtracting the counterfactual returns from the actual returns (Duso et al., 2010), i.e.:
𝐴𝑅𝑖𝑡 = 𝑅𝑖𝑡− (𝛼̂𝑖+ 𝛽̂𝑖𝑅𝑚𝑡) (3)
Next, the average abnormal return (AAR) for each date t is calculated as:
𝐴𝐴𝑅𝑡 =𝑁1∗ ∑𝑁𝑖=1𝐴𝑅𝑖𝑡 (4)
To incorporate information leakages around the event date, abnormal returns are calculated for the -1 to +2 period relative to the merger announcement (Duso et al., 2010; Leggio and Lien, 2000). Cumulative Abnormal Returns (CARs) are computed for the event period that spans from date m to date n by summing these daily AARs, i.e.:
𝐶𝐴𝑅𝑚,𝑛= ∑𝑡=𝑛 𝐴𝐴𝑅𝑡
𝑡=𝑚 (5)
Finally, cumulative average abnormal returns are calculated by taking the daily average of the CAR, i.e.: 𝐶𝐴𝐴𝑅𝑚,𝑛= 1
(𝑛−𝑚)𝐶𝐴𝑅𝑚,𝑛 (6)
20 (𝐶)𝐴𝐴𝑅𝑡𝑜𝑡𝑎𝑙= (𝐶)𝐴𝐴𝑅𝑡𝑎𝑟𝑔𝑒𝑡∗𝑀𝑉𝑡𝑎𝑟𝑔𝑒𝑡+(𝐶)𝐴𝐴𝑅𝑏𝑖𝑑𝑑𝑒𝑟∗𝑀𝑉𝑏𝑖𝑑𝑑𝑒𝑟
𝑀𝑉𝑡𝑎𝑟𝑔𝑒𝑡+𝑀𝑉𝑏𝑖𝑑𝑑𝑒𝑟 (7)
where MV is the market value of equity at the end of the month preceding the event window. Thus, returns are weighted by the market capitalizations of the two firms on the last day of the month before the announcement or rumor.
Both parametric and nonparametric tests are used to test the null hypotheses 1-4. For the parametric test, I use a test to determine whether mean abnormal returns are non-zero. According to Armitage (1995), a t-test is the generally favored parametric t-test method. Moreover, t-t-tests have been widely used in event studies to assess shareholder returns (e.g., Hankir et al., 2011). Additionally, I use a (nonparametric) Wilcoxon signed rank test to determine whether median abnormal returns are non-zero. The advantage of such a nonparametric test is that it does not assume a normally distributed sample. The Wilcoxon test is commonly used if a non-parametric test is performed in event studies (Armitage, 1995). For example, Hankir et al. (2011) use a Wilcoxon signed rank test to test for the significance of CARs in their study on banking mergers and market power. To examine bidder and combined CAARs (hypothesis 1 and 3), I use two-tailed tests. One-tailed tests are used to examine the target CAARs (hypothesis 2). This decision is based on the consistency and size of the shareholder returns reported by previous studies. For example, in an examination of 3,688 mergers from 1973 to 1998, Andrade et al. (2001) report highly significant average three-day abnormal returns of 16% for target firms. Andrade et al. also find positive returns for combined shareholder, but their average three-day returns are only 1.8% and much less frequently statistically significant. Finally, although bidders earn average three-day returns of -0.7%, Andrade et al. argue that these results are frequently unreliable due to low statistical significance levels. Thus, existing empirical evidence for positive target firm returns is quite strong, which supports a one-tailed test. As the (sign of the) empirical finding regarding bidder and combined returns are less reliable, I use two-tailed tests to examine these returns.
The analysis of the shareholder returns considers a number of time windows. As mentioned, the primary event window of interest spans the period from 1 day before to 2 days after the announcement or rumor. This is a common time window for shareholder return event studies and serves as a method to incorporate information leakages (Leggio et al., 2000). The other event windows that I examine are [-1;1], [0;1], [0;2] and [1;2]. Moreover, I estimate AARs for all the individual days within these event windows.
21 assumption of semi-strong efficient markets (García and Trillas, 2013). An extensive overview of this assumption and its critics is provided by Malkiel (2003). Malkiel proposes that the efficient market hypothesis has long been widely accepted by academics (Fama, 1970), but that its dominance has receded somewhat since the 21st century. Important points of critique on the efficient market hypothesis include the existence of behavioral effects, the January effect, size effects and the price-to-book effect. Malkiel reviews the existing evidence on these effects and finds that these effects are either economically insignificant, unsustainable over longer periods of time, or proxies for security risk that was not captured by Beta. Malkiel concludes that, although pricing irregularities may exist for short durations, markets are still exceptionally efficient in incorporating information. Thus, the assumption of (semi-strong) efficient markets seems appropriate for this study. The CAARs and their implications are discussed in the Results section of this paper.
Market Power Increases and Shareholder Returns
Next, to examine the relation between market power increases and merger announcement and rumor returns, I employ an OLS regression equation. The use of an OLS regression is based on Eckbo (1985) and James and Wier (1987), who use the same method to examine the relation between shareholder returns and changes in market power. Three important assumptions regarding OLS regression analysis are normality, homoscedasticity and serial independence of regression errors (Jarque and Bera, 1980). To test whether these assumptions are satisfied, I employ the Bera-Jarque test, White’s test, and the Durbin-Watson test, respectively (Brooks, 2008). The results of these tests are provided in appendix B. These tests suggest the regression may suffer from heteroskedasticity issues. To alleviate these concerns, I employ heteroskedasticity robust standard errors when estimating the regression (Books, 2008).
The CAAR-1,2 of a merger event serves as the dependent variable in the regression. I use the CAAR over
22 and are not possible to measure ex ante. An important assumption is that perceived market power increases are directly reflected in stock prices (James and Wier, 1987). However, the use of a computed combined concentration ratio is a common method for proxying future (potential) market power increases (e.g., James and Wier, 1987).
To measure market shares, I use market capitalizations and sales figures. Thus, I construct two alternative measures for ΔHHI, which are analyzed using two separate models. The former measure defines HHI by market capitalization on the day before the announcement or rumor. A set of listed EU firms is used to represent the market, with their market capitalization representing market shares. The peer groups for energy, telecommunications and air transportation consist of 202, 238 and 198 firms, respectively, and are constructed using the Zephyr database according to three steps. First, I selected all (previously) publicly listed EU companies, which resulted in a sample size of 21,506 firms. Next, all firms that did not have any of the four-digit SIC codes found in the merging firms samples were removed, based on Eckbo (1983). This reduced the sample size by 20,868. The remaining 638 firms were split up into the three sector groups according to their Bureau van Dijk industry classifications. I calculate ΔHHI by comparing the pre-announcement HHI with the post-announcement HHI, where the market capitalizations of the merging firms are combined. The Datastream database is employed to construct this variable. The second measure bases HHI on sales at the end of the fiscal year before the rumor or announcement, similar to Becker-Blease et al. (2008). This variable is constructed along the same method and uses the same peer groups and data sources as the ΔHHI variable based on market capitalizations. Schumann (1993) suggests that the relation between the change in HHI and abnormal returns is non-linear. I therefore also include a transformed variable that is computed as the square of the change in HHI in the models.
To control for deal- and firm-specific influences on merger returns, I also employ a number of commonly used control variables in the regressions. These control variables are primarily based on Becker-Blease et al. (2008), who examine the effects of deal and firm characteristics on merger returns in the US electric power industry. The model used to analyze the shareholder return effects of market power increases measured by market capitalizations is described by the following formula:
𝐶𝐴𝐴𝑅𝑖[−1;2]= 𝛽0+ 𝛽1𝛥𝐻𝐻𝐼(𝑀𝐶)𝑖+ 𝛽2𝛥𝐻𝐻𝐼(𝑀𝐶)𝑖2+ 𝛽
3𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦𝑖+ 𝛽4𝑆𝑖𝑧𝑒𝑖+
𝛽5𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑆𝑖𝑧𝑒𝑖+ 𝛽6𝐶𝑎𝑠ℎ𝑖+ 𝛽7𝐴𝑐𝑡𝑖𝑣𝑒𝑖+ 𝛽8𝑃𝑎𝑦𝑚𝑒𝑛𝑡𝑖+ 𝛽9𝐶𝑟𝑜𝑠𝑠𝑏𝑜𝑟𝑑𝑒𝑟𝑖+
𝛽10𝐴𝑛𝑛𝑜𝑢𝑛𝑐𝑒𝑚𝑒𝑛𝑡𝑖+ 𝛽11𝐸𝑛𝑒𝑟𝑔𝑦𝑖+ 𝛽12𝑇𝑒𝑙𝑒𝑐𝑜𝑚𝑖+ 𝛽13−20𝐶𝑜𝑢𝑛𝑡𝑟𝑦 𝑑𝑢𝑚𝑚𝑖𝑒𝑠 + 𝑒𝑖 (8)
23 𝐶𝐴𝐴𝑅𝑖[−1;2]= 𝛽0+ 𝛽1𝛥𝐻𝐻𝐼(𝑆𝑎𝑙𝑒𝑠)𝑖+ 𝛽2𝛥𝐻𝐻𝐼(𝑆𝑎𝑙𝑒𝑠)𝑖2+ 𝛽 3𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦𝑖+ 𝛽4𝑆𝑖𝑧𝑒𝑖+ 𝛽5𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝑆𝑖𝑧𝑒𝑖+ 𝛽6𝐶𝑎𝑠ℎ𝑖+ 𝛽7𝐴𝑐𝑡𝑖𝑣𝑒𝑖+ 𝛽8𝑃𝑎𝑦𝑚𝑒𝑛𝑡𝑖+ 𝛽9𝐶𝑟𝑜𝑠𝑠𝑏𝑜𝑟𝑑𝑒𝑟𝑖+ 𝛽10𝐴𝑛𝑛𝑜𝑢𝑛𝑐𝑒𝑚𝑒𝑛𝑡𝑖+ 𝛽11𝐸𝑛𝑒𝑟𝑔𝑦𝑖+ 𝛽12𝑇𝑒𝑙𝑒𝑐𝑜𝑚𝑖+ 𝛽13−20𝐶𝑜𝑢𝑛𝑡𝑟𝑦 𝑑𝑢𝑚𝑚𝑖𝑒𝑠 + 𝑒𝑖 (9)
In these models the CAAR[-1;2] represents the CAARs of the various shareholders as estimated by equation (6). C(A)ARs are frequently used as dependent variables to examine the effect of changes in market structures, e.g. Schumann (1993) and Becker-Blease et al. (2008) employ this methodology.
ΔHHI represents the change in Herfindal-Hirschman index following the merger announcement. As
mentioned, two separate measures based on sales and market capitalizations are used. The use of this variable is based on Schumann (1993), who finds a positive and statistically significant relation between ΔHHI and merger announcement returns. The ΔHHI based on sales is computed using the same methodology as Becker-Blease et al. (2008).
ΔHHI2 represents the squared version(s) of the Herfindal-Hirschman index. This variable is included to
examine whether there exists a non-linear relation between changes in market power and shareholder returns, as hypothesized by Schumann (1993). This variable is strongly correlated with its base variable (see table 3), raising concerns about multicollinearity. These are confirmed by an estimate of the Variance Inflation Factors, as displayed in appendix C. Mean centering of the predictor is used to diminish potential issues with multicollinearity, as proposed by Kreft, De Leeuw and Aiken (1995).
Volatility represents the riskiness of the target firm. It is measured as the standard deviation of the
monthly stock returns of the target firm for the 60 months preceding the announcement or rumor. This variable is based on Anderson and Reeb (2003) and is constructed using Datastream.
Size indicates the bidder’s size, and is measured as the natural logarithm of the bidder’s equity value. The
construction of this variable follows the methodology of Moeller et al. (2004), who find that smaller bidders experience higher abnormal returns. The variable is created using Datastream.
RelativeSize measures the deal size relative to the bidder’s size. This variable is computed as the deal
value divided by the bidder’s equity value. Construction of this variable again uses the methodology of Moeller et al. (2004). Asquith, Bruner and Mullins (1983) report that relative size is positively related to merger announcement returns. For (most) rumored mergers actual deal values were not available. In these cases, the variable was measured as the target’s equity value divided by the bidder’s equity value. The data required for constructing this variable was collected from Zephyr and Datastream.
Cash indicates the amount of cash that is held by the bidding firm. It is measured as the natural logarithm
24
Active indicates whether the bidding firm has announced acquisitions in the twelve months preceding the
rumor or announcement. This variable is based on Becker-Blease et al. (2008), who report lower announcement returns for active bidders. The variable is constructed using information from the Zephyr database.
Payment indicates whether the merger was paid for by cash or (a combination of cash and) shares. I use a
dummy variable that equals parity for all-cash deals, whereas it is zero otherwise, based on Becker-Blease et al. (2008). Andrade et al. (2001) report that all-cash deals are associated with higher premiums. The reason is that managers are more likely to pay with their own equity if they think that it is overpriced by the stock markets. Zephyr is used to construct this variable.
Crossborder indicates whether a merger is domestic or cross-border. A dummy variable that equals one
for cross-border deals is constructed using Zephyr. This variable is based on Goergen and Renneboog (2004), who find that domestic mergers generate larger wealth effects.
Announcement indicates whether an observation is a merger announcement or a merger rumor. This
dummy variable is modeled to be equal to one for announcements. As mentioned, there are, to the author’s knowledge, no studies on which a prediction regarding the sign of the coefficients can be based.
Energy and Telecom dummy variables are included to indicate industry classification. Thus, air
transportation is modeled as the default industry.
Finally, a set of country dummy variables is included to account for country-specific effects. Country dummy variables are frequently used when investigating merger effects across different countries (e.g., Lepetit, Patry and Rous, 2004). This study uses country combinations, which indicate the countries of origin of the bidder and the target. Dummy variables are included for the eight most frequent combinations. An overview of all combinations and their frequencies can be found in appendix D.
The estimated coefficients of the regression are used to assess the relation between market power and shareholder returns in the Results section. T-tests are employed to determine the statistical significance of the various variables (Becker-Blease et al., 2008). Descriptive statistics of the variables can be found in table 2. Moreover, a correlation matrix is displayed in table 3.
Market Power Increases and Returns of Rivals
25 TABLE 2: Descriptive Statistics of the Main Regression Variables
Note: DE|DE, DE|ES, etc. indicate the aforementioned country combinations. Here, DE=Germany, ES=Spain, GB=Great Britain, DK=Denmark, NL=Netherlands, FR=France
primarily increase efficiency, all non-merging market participants lose due to lower competitiveness vis-à-vis the merging firm. In this case a non-positive share price reaction of rivals is expected (Eckbo, 1983). To attempt to reject the fifth null hypothesis, which states that rivals experience no positive abnormal returns upon merger announcements or rumors, I employ a second model based on McAfee and Williams (1988). This model is used to examine the abnormal returns of rival firms and is captured by the following formula:
𝑟𝑝= 𝛼 + 𝛽𝑟𝑚+ 𝛾𝑑 + 𝜖𝑝 (10)
where rp represents the continuously compounded returns of an equally-weighted portfolio of
competitors, and rm indicates the daily continuously compounded returns of the equally-weighted relevant
market indices (e.g., the AEX, DAX and FTSE). As mentioned, Armitage (1995) states that a composite market index is an appropriate proxy for the market portfolio. Both returns are computed according to equation (2) using data from Datastream.
In the regression d is a dummy variable that takes a value of one if an observation falls in an event window (i.e., the period surrounding a merger announcement or rumor). The corresponding coefficient γ therefore represents the daily AAR of the portfolio of competitors. The random error term is a normally distributed error term.
M ean M edian M inimum M aximum # observations
26 TABLE 3: Correlation Matrix of the Main Regression Variables
ΔHHI(Sales) ΔHHI(MC) ΔHHI(Sales)2 ΔHHI(MC)2 Volatility Size Relative size Cash Active Payment Energy
ΔHHI(Sales) 1.00 ΔHHI(MC) 0.96 1.00 ΔHHI(Sales)2 0.71 0.70 1.00 ΔHHI(MC)2 0.50 0.53 0.91 1.00 Volatility -0.05 0.02 0.04 0.04 1.00 Size 0.45 0.40 0.50 0.32 0.21 1.00 Relative size -0.02 -0.01 0.04 0.06 0.06 -0.11 1.00 Cash 0.56 0.50 0.46 0.24 -0.08 0.81 -0.11 1.00 Active 0.16 0.12 0.18 0.14 0.08 0.37 0.06 0.21 1.00 Payment -0.16 -0.15 -0.04 -0.02 0.40 -0.28 -0.04 -0.30 -0.23 1.00 Energy 0.49 0.49 0.39 0.25 0.13 0.41 -0.06 0.40 0.12 0.02 1.00 Telecom -0.29 -0.29 -0.23 -0.20 -0.02 0.09 0.19 -0.04 0.16 -0.11 -0.42 Cross-border 0.07 0.07 0.02 0.01 0.23 -0.01 0.05 -0.01 -0.03 0.01 0.00 DE|DE -0.10 -0.08 -0.10 -0.06 -0.11 -0.08 -0.04 -0.20 -0.14 0.22 0.09 DE|ES 0.66 0.72 0.63 0.55 -0.30 0.40 -0.03 0.46 0.15 -0.28 0.49 DE|GB 0.00 -0.06 0.00 -0.05 0.08 0.12 -0.04 0.19 0.07 -0.12 0.21 DK|NL 0.06 0.05 0.02 -0.04 -0.06 0.04 -0.07 0.01 0.09 -0.17 -0.12 ES|ES -0.12 -0.14 -0.05 -0.10 0.21 0.14 -0.08 0.04 0.08 -0.15 0.08 FR|FR 0.17 0.20 0.01 -0.04 0.06 0.09 -0.02 0.18 0.02 -0.05 0.20 FR|GB -0.11 -0.10 -0.01 -0.06 0.10 0.16 -0.06 0.13 0.02 -0.05 -0.14 GB|GB -0.24 -0.21 -0.28 -0.17 -0.16 -0.40 -0.15 -0.47 -0.24 -0.09 -0.36 Announcement -0.04 -0.03 -0.02 -0.02 0.01 0.03 -0.01 -0.01 0.07 -0.01 0.00
Telecom Cross-border DE|DE DE|ES DE|GB DK|NL ES|ES FR|FR FR|GB GB|GB
27 An OLS regression is employed to estimate the parameters in the equation, which is the same method as used by McAfee and Williams (1988). Again, important assumptions regarding OLS regression analysis are normality, homoscedasticity and serial independence of regression errors (Jarque and Bera, 1980). I use the same tests as for the first OLS regression to determine whether these assumptions are satisfied. The results of these tests are provided in appendix E. These tests suggest that the regression may suffer from heteroskedasticity issues. To mitigate these concerns, I employ heteroskedasticity robust standard errors when estimating the regression (Brooks, 2008). The Bera-Jarque test suggests non-normality of the regression errors may also be an issue. However, Brooks (2008) proposes that for large samples, violation of normality assumptions is essentially inconsequential. Given the large number of observations in the affected regressions, i.e. 14,085 and 4,965, non-normality issues appear to be of minor concern.
I use a t-test to assess whether γ is statistically significantly different from zero, which is also used by McAfee and Williams (1988). The competitor portfolios are the same as those used earlier for constructing the ΔHHI variables. Thus, the energy, telecommunications and air transportation portfolios consist of 202, 238 and 198 firms, respectively. The estimation period for the regression is 250 days before the merger announcement or rumor and ten days after, based on the methodology of McAfee and Williams (1988) and Schumann (1993). In total, I perform four different regressions due to the use of two different event windows (day 0 and [-1;2]), and the use of separate regressions for rumor and announcement returns.
Overall, the sample sizes of this study compare favorably to previous studies. For reference, this study examines 179 mergers events, involving 172 different companies. Other studies that employ similar methodologies provide comparable figures. For example, Becker-Blease et al. (2008) examine 70 mergers, Kim and Singal (1993) review 27 mergers, and Eckbo (1983) uses a sample of 266 mergers. Given that the sample size of this study is restricted by its industry focus, the large sample size compares rather favorably. Probable causes for this are the use of a long time frame, a wide geographic focus and the choice to examine multiple, although comparable, industries. A more general but more exhaustive overview of event study sample sizes is provided by Park (2004).
Another methodological feature of this study is that it examines both merger announcements and rumors. Few previous studies have explicitly considered both merger rumors and announcements. Various studies, including Duso, Gugler, and Yurtoglu (2011), examine merger rumors, but these do not simultaneously examine the effects of official announcements on stock prices. However, rumors may significantly impact shareholder returns and may also influence announcement effects.
28 comparison of the two and may bear implications for the appropriate measure to use in future studies. Additionally, most studies on the market power hypothesis examine either the effect of changes in market concentration on merger returns, or the effect of merger events on rival stock prices. Combining these two methods may reinforce the strength of possible conclusions.
A limitation of the methodology of this paper, and in particular its sample choice, may lie in the fact that it only examines listed firms. This may limit the generalizability of any significant result. Moreover, the situation of private firms may be particularly interesting due to the fact that listed firms are commonly under much more public scrutiny than private firms. However, this limitation is shared by most of the existing literature and may be interpreted as a possible research venue for future studies. Another limitation is that the analysis does not discern between different rumor types, while in reality some rumors may be more substantial than others. The following section of this text discusses the results of the empirical analyses and how these relate to the previously posed hypotheses.
4. RESULTS AND CONCLUSION
In this section I first examine the returns experienced by the various types of shareholders and relate these findings to both previous literature and the first three hypotheses of this study. Next, I discuss the results of the main regression to examine the relation between changes in market power through mergers and shareholder returns. These findings are then used to test the fourth hypothesis. Subsequently, I examine the impact of mergers on rival firm stock prices to test the fifth and final hypothesis. Finally, I discuss the findings, contributions and limitation of this study, as well as future research opportunities.
29 TABLE 4: Merger Announcement Stock Returns for the Bidder, Target and Combined Entities
The table displays the shareholder returns of merger announcements for the various shareholders in the first column. AARs are provided for individual dates, e.g. day -1, while CAARs are provided for various time windows, e.g. [1;2]. P-values for the corresponding parametric t-tests are provided in the second column, while the third column displays p-values for the corresponding nonparametric Wilcoxon signed-rank tests. P-values in italics are for one-sided tests. Non-italic p-values are for two-sided tests.
Announcement returns - Aggregate sample
bidder target combined bidder target combined bidder target combined
-1 0.23% 1.81% 0.40% 0.45 0.00 0.15 0.80 0.00 0.22 0 0.43% 3.78% 0.76% 0.22 0.00 0.05 0.08 0.00 0.01 1 -0.11% 0.96% 0.11% 0.81 0.13 0.78 0.63 0.02 0.70 2 1.34% 0.33% 0.92% 0.01 0.19 0.00 0.00 0.46 0.01 [-1;2] 0.47% 1.72% 0.55% 0.02 0.00 0.00 0.02 0.00 0.00 [-1;1] 0.18% 2.18% 0.43% 0.37 0.00 0.06 0.41 0.00 0.02 [0;1] 0.16% 2.37% 0.44% 0.57 0.01 0.16 0.42 0.00 0.03 [0;2] 0.55% 1.69% 0.60% 0.03 0.00 0.01 0.02 0.00 0.00 [1;2] 0.61% 0.65% 0.52% 0.06 0.03 0.03 0.04 0.03 0.04
Announcement returns - Energy
bidder target combined bidder target combined bidder target combined
-1 0.24% 1.24% 0.25% 0.68 0.04 0.57 0.64 0.02 0.47 0 -0.01% 3.38% 0.21% 0.99 0.11 0.77 0.44 0.22 0.20 1 -0.21% 1.72% 0.07% 0.63 0.03 0.84 0.72 0.02 0.88 2 -0.20% -0.12% -0.04% 0.80 0.65 0.94 0.96 0.64 0.96 [-1;2] -0.04% 1.55% 0.12% 0.91 0.02 0.62 0.47 0.00 0.38 [-1;1] 0.01% 2.11% 0.18% 0.99 0.02 0.54 0.76 0.00 0.35 [0;1] -0.11% 2.55% 0.14% 0.80 0.04 0.70 0.76 0.04 0.38 [0;2] -0.14% 1.66% 0.08% 0.75 0.04 0.81 0.54 0.02 0.28 [1;2] -0.20% 0.80% 0.01% 0.63 0.05 0.96 0.96 0.04 0.68
Rumor returns - Telecommunications
bidder target combined bidder target combined bidder target combined
-1 0.38% 1.26% 0.39% 0.62 0.05 0.55 0.59 0.04 0.95 0 0.12% 3.08% 0.08% 0.79 0.08 0.90 0.56 0.08 0.56 1 -1.17% 2.20% -0.20% 0.19 0.04 0.81 0.33 0.12 0.65 2 2.99% -0.35% 1.53% 0.02 0.75 0.04 0.02 0.79 0.06 [-1;2] 0.58% 1.55% 0.45% 0.21 0.01 0.16 0.47 0.01 0.40 [-1;1] -0.22% 2.18% 0.09% 0.51 0.01 0.80 0.29 0.00 0.78 [0;1] -0.53% 2.64% -0.06% 0.23 0.02 0.91 0.45 0.03 0.71 [0;2] 0.64% 1.64% 0.47% 0.25 0.02 0.24 0.37 0.01 0.20 [1;2] 0.91% 0.93% 0.66% 0.21 0.08 0.18 0.22 0.24 0.20
Rumor returns - Air Transportation
bidder target combined bidder target combined bidder target combined
-1 0.11% 2.59% 0.50% 0.71 0.02 0.17 0.63 0.07 0.16 0 0.95% 4.56% 1.65% 0.07 0.04 0.01 0.11 0.00 0.01 1 0.74% -0.48% 0.38% 0.39 0.61 0.61 0.67 0.78 0.48 2 1.12% 1.14% 1.10% 0.02 0.08 0.01 0.02 0.14 0.01 [-1;2] 0.73% 1.95% 0.91% 0.01 0.02 0.01 0.02 0.00 0.00 [-1;1] 0.60% 2.22% 0.84% 0.06 0.06 0.06 0.03 0.00 0.01 [0;1] 0.85% 2.04% 1.02% 0.07 0.14 0.09 0.07 0.00 0.03 [0;2] 0.94% 1.74% 1.05% 0.01 0.05 0.01 0.01 0.00 0.01 [1;2] 0.93% 0.33% 0.74% 0.06 0.29 0.07 0.06 0.19 0.12
AAR/CAAR p-value (parametric) p-value (nonparametric)
AAR/CAAR p-value (parametric) p-value (nonparametric)
AAR/CAAR p-value (parametric) p-value (nonparametric)
30 TABLE 5: Merger Rumor Stock Returns for the Bidder, Target and Combined Entities
The table displays the shareholder returns of merger rumors for the various shareholders in the first column. AARs are provided for individual dates, e.g. day -1, while CAARs are provided for various time windows, e.g. [1;2]. P-values for the corresponding parametric t-tests are provided in the second column, while the third column displays p-values for the corresponding nonparametric Wilcoxon signed-rank tests. P-values in italics are for one-sided tests. Non-italic p-values are for two-sided tests.
Rumor returns - Aggregate sample
bidder target combined bidder target combined bidder target combined
-1 -0.26% 0.75% 0.00% 0.12 0.01 0.99 0.02 0.05 0.42 0 0.31% 3.75% 0.94% 0.17 0.00 0.00 0.23 0.00 0.00 1 0.43% 0.96% 0.43% 0.06 0.02 0.04 0.02 0.26 0.02 2 0.38% -0.21% 0.13% 0.18 0.79 0.53 0.16 0.90 0.42 [-1;2] 0.22% 1.31% 0.37% 0.07 0.00 0.00 0.10 0.00 0.00 [-1;1] 0.16% 1.82% 0.45% 0.15 0.00 0.00 0.32 0.00 0.01 [0;1] 0.37% 2.35% 0.68% 0.01 0.00 0.00 0.01 0.00 0.00 [0;2] 0.38% 1.50% 0.50% 0.01 0.00 0.00 0.00 0.00 0.00 [1;2] 0.41% 0.37% 0.28% 0.03 0.04 0.08 0.01 0.48 0.03
Rumor returns - Energy
bidder target combined bidder target combined bidder target combined
-1 -0.15% 0.22% -0.08% 0.45 0.22 0.59 0.60 0.22 0.72 0 -0.28% 1.01% -0.02% 0.49 0.03 0.95 0.83 0.02 0.51 1 0.33% 0.29% 0.19% 0.25 0.26 0.49 0.11 0.49 0.20 2 -0.29% -0.06% -0.27% 0.34 0.61 0.30 0.60 0.82 0.39 [-1;2] -0.10% 0.36% -0.05% 0.57 0.03 0.74 0.82 0.14 0.42 [-1;1] -0.03% 0.51% 0.03% 0.87 0.02 0.87 0.78 0.07 0.51 [0;1] 0.03% 0.65% 0.08% 0.92 0.03 0.73 0.45 0.12 0.37 [0;2] -0.08% 0.41% -0.04% 0.71 0.04 0.85 0.34 0.13 0.34 [1;2] 0.02% 0.11% -0.04% 0.91 0.33 0.82 0.35 0.46 0.48
Rumor returns - Telecommunications
bidder target combined bidder target combined bidder target combined
-1 -0.58% 1.04% -0.19% 0.11 0.07 0.54 0.00 0.33 0.11 0 0.58% 4.06% 1.27% 0.05 0.03 0.05 0.09 0.00 0.01 1 0.13% -0.08% 0.02% 0.70 0.56 0.95 0.36 0.63 0.45 2 0.23% -0.15% 0.13% 0.62 0.62 0.76 0.78 0.84 0.77 [-1;2] 0.09% 1.22% 0.31% 0.59 0.01 0.14 0.98 0.00 0.35 [-1;1] 0.04% 1.68% 0.37% 0.78 0.02 0.15 0.67 0.00 0.23 [0;1] 0.35% 1.99% 0.65% 0.10 0.04 0.08 0.14 0.01 0.01 [0;2] 0.31% 1.28% 0.47% 0.13 0.03 0.06 0.18 0.01 0.07 [1;2] 0.18% -0.11% 0.07% 0.55 0.62 0.78 0.42 0.93 0.56
Rumor returns - Air Transportation
bidder target combined bidder target combined bidder target combined
-1 -0.02% 0.99% 0.31% 0.96 0.03 0.32 0.77 0.05 0.41 0 0.65% 6.46% 1.63% 0.18 0.00 0.00 0.49 0.00 0.00 1 0.92% 2.94% 1.17% 0.11 0.01 0.03 0.11 0.07 0.04 2 1.31% -0.46% 0.59% 0.05 0.78 0.12 0.01 0.56 0.08 [-1;2] 0.71% 2.48% 0.93% 0.01 0.00 0.00 0.01 0.00 0.00 [-1;1] 0.52% 3.46% 1.04% 0.03 0.00 0.00 0.07 0.00 0.00 [0;1] 0.78% 4.70% 1.40% 0.01 0.00 0.00 0.02 0.00 0.00 [0;2] 0.96% 2.98% 1.13% 0.00 0.00 0.00 0.00 0.00 0.00 [1;2] 1.11% 1.24% 0.88% 0.02 0.01 0.01 0.00 0.00 0.00
AAR/CAAR p-value (parametric) p-value (nonparametric)
AAR/CAAR p-value (parametric) p-value (nonparametric)
AAR/CAAR p-value (parametric) p-value (nonparametric)
