Description
The International Finance Corporation (IFC) is an international financial institution which offers investment, advisory, and asset management services to encourage private sector development in developing countries.
ABSTRACT
Title of dissertation:
ESSAYS IN INTERNATIONAL FINANCE TANAKORN MAKAEW, Doctor of Philosophy, 2010
Dissertation directed by:
Professor VOJISLAV MAKSIMOVIC Department of Finance
The dissertation consists of three essays on international capital ?ows.
In the ?rst essay, titled “Do small ?rms bene?t more from foreign portfolio investment? Evidence from a Natural Experiment,” I test whether an increase in the supply of foreign portfolio capital bene?ts small ?rms by using the Thai government’s unique restriction on capital in?ows as a natural experiment. The Thai government imposed a very stringent capital control on December 19, 2006, and then quickly abandoned it one day later. Although many other studies have been plagued with the di?culty of separating the impact of foreign capital from the impact of other concurrent events, this experiment helps me solve the time-series identi?cation problem. My results suggest that foreign portfolio investment helps large ?rms the most, contrary to existing evidence, which ?nds a bene?t in foreign portfolio investment for small ?rms. I also investigate the importance of other ?rm characteristics correlated with size, which includes a ?rm’s exchange rate exposure, foreign ownership, and political connection.
The next two essays are on the dynamic patterns of international mergers and acquisitions.
In the second essay, I uncover key facts about international M&As by estimating a variety of reduced form models. I ?nd that: (1) Cross-border mergers come in waves that are highly correlated with business cycles. (2) Most mergers occur when both the acquirer and the target economies are booming. (3) Merger booms have both an industry-level component (productivity shocks) and a country-level component (?nancial shocks). (4) Across over one million observations, acquirers tend to be more productive and targets tend to be less productive, compared to their industry peers. These facts are consistent with the neoclassical theory of mergers in which productive ?rms expand overseas to seize new investment opportunities, but not with the widely held views that most crossborder mergers occur when the target economies are in a recession or face a ?nancial crisis.
In the third essay, I construct a dynamic structural model of cross-border mergers and integrate the important facts above into the model. This dynamic structural approach allows me to quantify the e?ects of productivity and ?nancial shocks on M&A decisions. In addition, this approach provides a proper analytical framework for conducting policy experiments. As an example of such analyses, I investigate the impact of President Obama’s proposal on multinational corporation taxation. My simulation results suggest that the foreign operation tax has economically signi?cant e?ects on productive ?rms and can be very distortionary for cross-border mergers.
ESSAYS IN INTERNATIONAL FINANCE by TANAKORN MAKAEW
Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park in partial ful?llment of the requirements for the degree of Doctor of Philosophy 2010
Advisory Committee: Professor Vojislav Maksimovic, Chair Professor Albert ‘Pete’ Kyle Professor Gordon Phillips Professor Lemma Senbet Professor Peter Murrell
ACKNOWLEDGEMENTS
I owe the deepest gratitude to my advisor, Vojislav Maksimovic. This dissertation would not have been possible without him. I am also deeply indebted to my committee members, Albert “Pete’ Kyle, Gordon Phillips, and Lemma Senbet, for their help and guidance throughout my doctoral studies.
I thank Jerry Hoberg and Mark Loewenstein for introducing me to the ?nance faculties. It was a pleasure working for Gurdip Bakshi, Steve Heston, Alex Triantis, and Haluk Unal. My thanks also go to my coauthors, Minwen Li and Juan Contreras.
I thank Peter Murrell for serving on my dissertation committee and Charle Lahaie for data supports. I appreciate valuable job market advices from Ethan Cohen-Cole, Shawn Cole, Michael Faulkender, Dalida Kadyrzhanova, Anna Obizhaeva, N.R. Prabhala, Georgios Skoulakis, and Russ Wermers. I thank Elinda Kiss for her teaching advices.
I thank Mara Faccio, Nandini Gupta, Anton Korinek, Andrew Karolyi, Carmen Reinhart, Antoinette Schoar, and Shangjin Wei, as well as seminar participants at FMA, EFA, and SWFA for comments and suggestions on Chapter 1.
I thank Rui Albuquerque, Sudipto Dasgupta, Antonio Felato, Amar Gande, Jarrad Harford, Pab Jotikastira, Sandy Klasa, Anton Korinek, Yrjo Koskinen, Michael Lemmon, Peter MacKay, Robert Marquez, Darius Miller, Kanda Naknoi, Mark Seascholes, Karin Thorburn, Missaka Warusawitharana, and Shangjin Wei, as well as seminar participants at Boston University, Federal Reserve Board of Governors, HKUST, Norwegian School of Economics, Southern Methodist University, University of Maryland, University of South Carolina, and University of Utah for many helpful suggestions on Chapters 2 and 3.
ii
Contents
1 Do small ?rms bene?t more from foreign portfolio investment? Evidence from a Natural Experiment 1.1 1.2 1.3 1.4 1.5 1.6 1.7 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Related Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Natural Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1 4 8
Data and Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Robustness Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Conclusion and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2 Waves of International Mergers and Acquisitions 2.1 2.2 2.3 2.4 2.5
40
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 Empirical Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 Merger Activities and Macroeconomic Conditions . . . . . . . . . . . . . . . 48 Firm Characteristics and Industry Merger Waves . . . . . . . . . . . . . . . 54
iii
2.6 2.7
Additional Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 Conclusion and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3 A Dynamic Model of International Mergers and Acquisitions 3.1 3.2 3.3 3.4 3.5 3.6
93
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 Conceptual Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 The Model and its Basic Properties . . . . . . . . . . . . . . . . . . . . . . . 96 Solution Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 Simulations and Policy Experiments . . . . . . . . . . . . . . . . . . . . . . 104 Conclusion and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
iv
Chapter 1
Do small ?rms bene?t more from foreign portfolio investment? Evidence from a Natural Experiment
1.1 Introduction
Small ?rms play an important role in emerging market economies since they are often associated with employment generation, economic diversity, balanced income distribution as well as being a source of entrepreneurship, innovation, and economic growth. While it is apparent that foreign portfolio investment has a signi?cant impact on ?rms in emerging markets, it is less clear whether ?rms of di?erent sizes are a?ected by foreign portfolio investment symmetrically. In this paper, I test whether an increase in the supply of foreign portfolio capital bene?ts small ?rms by using the Thai government’s unique restriction on capital in?ows as a natural experiment. I ?nd that foreign portfolio investment helps large ?rms the most, contrary to existing evidence, which ?nds a bene?t in foreign portfolio investment for small ?rms.
1
Existing papers tend to argue that small ?rms bene?t from foreign portfolio investment more than large ?rms (for example, Gelos and Werner 1999, Knill 2005, and Patro and Wald 2005). A number of authors document a positive correlation between foreign portfolio investment and small ?rms’ growth, both in terms of capital accumulation and the ability to access external capital markets. Others study the impacts of ?nancial liberalization, the event that leads to a large increase in foreign portfolio investment. They ?nd that, after ?nancial liberalization, small ?rms have lower investment-cash ?ow sensitivities, face lower cost of capital, and invest more. Another important ?nding is that, during the time of liberalizations, small ?rms, on average, experience higher stock returns compared to large ?rms. However, from the existing literature, it is ambiguous whether small ?rms bene?t from foreign portfolio investment or from other factors correlated with the surge in foreign capital. Foreign portfolio investment is potentially correlated with a number of macroeconomic variables. Stock market and capital account liberalizations are usually concurrent with other major changes such as trade liberalizations, reforms in stock market regulations, and reforms in banking supervisions. I believe that this time-series identi?cation problem is severe since it is virtually impossible to list all the events that a?ect ?rm value. Even if I can identify all the relevant factors, it is still hard to identify the exact time these changes took place (in order to control for them in a panel data study) or the exact time the market learned about them (in order to control for them in an event-study). I analyze the stock market impacts of Thailand’s unique restriction on portfolio capital in?ow. The Thai government imposed a very stringent capital control on December 19, 2006 and then quickly abandoned it on December 20, 2006. The fact that the control only lasted for one day provides an excellent framework for a natural experiment study. It is di?cult to come up with another factor that is unrelated to capital control, has as dominant of an e?ect on ?rms compared to capital control, and changes back and forth overnight like capital control. For example, one might argue the stock return on the capital control day may re?ect both changes in foreign capital and changes in investor’s perception about the Thai government’s ability to run the economy e?ectively. However, it is not likely that this perception was largely reversed in a day at the same time that
2
the government reversed its decision about the capital control. Natural experiments have recently become popular in social sciences, especially in economics. A recent search using the term “natural experiment” on Google Scholar yields more than one million results. The 2002 Nobel Laureate in Economics, Vernon L. Smith, also stated in the Journal of Economic Perspective, “Natural experiments occur all the time and it would be desirable to develop a professional readiness to seize upon these occasions (p.155).” This paper joins a growing literature in ?nance that uses a natural experiment as a solution to the identi?cation problems. I show that large ?rms experienced more negative abnormal returns on the capital control day (December 19) and more positive abnormal returns on the liberalization day (December 20), suggesting that foreign portfolio investment by itself bene?ts large ?rms the most. Compared to small ?rms, large ?rms have higher fractions of foreign ownership and are more likely to have political and business connections. In order to examine how much of the size e?ects are due to the di?erence in other ?rm characteristics correlated with size, I control for (1) ?rm ?nancial characteristics (pro?tability, investment opportunities, leverage, accounting liquidity, and industry dummy), (2) ?rm international involvements (exchange rate exposure, foreign ownership, and foreign control), and (3) ?rm connections (both political and business connections). I ?nd that size still has a large and signi?cant explanatory power after including these variables in our regressions. In the full speci?cation, ?rms that are one standard deviation larger earn 83.35 basis points less on the capital control day and 84.78 basis points more on the liberalization day. I further show that size is correlated with visibility to foreign investors and past capital market activities - large ?rms tend to be included in key stock market indices, to be rated by rating agencies, and to have issued securities in international markets. In addition to my ?ndings about size, I ?nd that ?rms with higher pro?tability, exportoriented ?rms, and ?rms with foreign directors are less a?ected by the capital control. Most interestingly, I ?nd that the stock prices of ?rms connected to the former Prime Minister Thaksin Shinawatra, the major opponent of the incumbent coup government, reacted more strongly to the capital control and the subsequent liberalization. Consistent with Johnson and Mitton (2003) which views Malaysian capital control as a way to support
3
?rms connected to the incumbent government, here investors view Thai capital control as a way to punish ?rms connected to the opponent of the government in power. One might be concerned about the ?ndings that large ?rms are a?ected by the capital control more are due to the market microstructure e?ects, not changes in ?rms’ fundamental value. In other words, since small ?rms are less liquid and small ?rms’ stocks are more closely held, stock prices of large ?rms might be more sensitive to any negative news that has an impact on the macroeconomy. To address this concern, I use the day the market learned about the September 2006 coup as a placebo test. I regress abnormal returns on the coup date on ?rm size together with other ?rm characteristics and ?nd that while the market return was negative on that day, small ?rm returns were signi?cantly more negative, ruling out the claim that large ?rms are more sensitive to any bad news. My results are robust to various econometric speci?cations and variable de?nitions. To correct the potential problems from the non-normality in error terms such as heteroscedasticity and cross-sectional correlations, I (1) use Huber/White/Sandwich standard errors, (2) cluster standard errors at ?rm-level, and (3) cluster standard errors at industry-level. I also (4) compute the empirical standard errors from bootstrapping and (5) compute the empirical standard errors from historical data. Finally, I use alternative de?nitions of size, industry classi?cation, pro?tability, and liquidity and use raw returns instead of abnormal returns. All of the results I ?nd are qualitatively the same. The paper proceeds as follows. Section 2 summarizes related literature. Section 3 describes the natural experiment. Section 4 outlines the empirical strategy and provides the data sources. Section 5 estimates the e?ects that ?rm size and other control variables have on the bene?ts from foreign portfolio investment and analyzes the results. Section 6 performs robustness checks. Section 7 concludes.
1.2
Related Literature
Foreign portfolio investment has become an important part of international capital ?ow. According to Bosworth et al.(1999), the composition of capital ?ow has shifted away from
4
foreign direct investment and bank loans to portfolio investment; the fraction of foreign portfolio investment in emerging markets has increased from 9% in 1978-1981 to 44% in the 1990’s. Consequently, costs and bene?ts of foreign portfolio investment are usually at the heart of any ?erce debates on ?nancial globalization. Bene?ts of Foreign Portfolio Investment Foreign portfolio investment is believed by many to have large potential bene?ts. The in?ow of foreign fund increases the supply of capital in a domestic economy. With more capital, ?rms can expand their existing capacities and undertake more projects. From the ?nancial markets perspective, foreign portfolio investment increases market liquidity and hence improves asset-pricing e?ciency (Levine and Zervos 1998). Li et al. (2006) provide evidence that capital account liberalization lowers the co-movement and raises idiosyncratic variation in stock prices, suggesting that stock prices contain more ?rmspeci?c information and the stock market becomes more e?cient. Some policy experts (for example, Evan 2003) additionally argue that foreign portfolio investors have superior technologies to value ?rms compared to domestic investors. Therefore, they create informational externalities that help domestic investors identify the best place to invest. The Importance of Small Firms Small ?rms have long been a center of attention in academia and policy circles. In the Federal Reserve’s Economic Quarterly, Weinberg (1994) stated, “It seems that a necessary part of the debate over any proposed public policy action, from healthcare to tax policy, is the question of how it will a?ect small ?rms (p.1).” Internationally, the World Bank has approved more than $10 billion support to the small and medium business enterprises during 1998-2002 (Beck et al. 2005). The attention that small ?rms received comes as no surprise since the growth of small ?rms and the growth of large ?rms are perceived to have very di?erent impacts on an economy. Small ?rms are often associated with employment generation, economic diversity, balanced income distribution, and being a source of entrepreneurship, innovation, and economic growth. While small ?rms are looked fondly upon, large ?rms are often associated with entrenchment and economic ine?ciencies. For example, Fogel, Morck, and Yeung (2005) ?nd that in countries whose large ?rms are doing well, overall economic growth, productivity growth and capital accumulation is lower.
5
They interpret this evidence as a support of Schumpeter’s theory of creative destruction, in which growth comes from small creative ?rms destroying large old ?rms. Small ?rms in emerging markets are generally considered the ones that su?er more from informational problems (Kang and Stulz 1997; Dahlquist and Robertsson 2001; and others) since large ?rms tend to be better known, older and have a longer track record of past performance. Analysts and the media also tend to cover large ?rms more frequently, making it harder for executives of large ?rms to hide mistakes or overstate pro?ts. Given that small ?rms su?er more from informational problems, I could easily deduce from the classical theories of corporate ?nance that small ?rms will be more ?nancially constrained. For example, small ?rms will face more credit rationing according to Stiglitz and Weiss (1981) and will face a higher cost of equity according to Jensen (1976) and Jensen and Meckling (1986). Small Firms and the Bene?ts from Foreign Portfolio Investment While foreign portfolio investment mechanically increases the aggregate supply of capital and hence should bene?t all ?rms in the domestic economy, it remains an empirical question whether or not small ?rms bene?t more than large ?rms. Theoretically, if foreign portfolio investment does help alleviate asymmetric information and agency problems for all ?rms, then small ?rms should bene?t more since they are the ones who su?er more from these problems and starve for capital in the ?rst place (see Section 1 in Evan (2003) and Section 4 in Forbes (2005) for the detailed arguments how foreign capital might solve informational problems). Even though it is well-documented empirically that foreign portfolio investors prefer to invest in larger ?rms (Kang and Stulz 1997; Dahlquist and Robertsson 2001; and others), many researchers in corporate ?nance note that small ?rms do not have to be the direct recipients of foreign portfolio investment in order to bene?t more. In one example, Knill (2005) suggests the monetary transmission mechanism through bank loans (Bernanke and Blinder 1988; Kashyap and Stein 1995; Kashyap, Rajan and Stein 2002); small ?rms and large ?rms are competing for the same pool of bank loans. When additional supply of capital ?ows to large ?rms, small ?rms receive more bank loans since large ?rms have less demand for loans. In another example, Gallego and Hernandez (2003) suggest the
6
trade credit channel; small ?rms and large ?rms are competing for the same pool of trade credits. When foreign portfolio investment ?ows to large ?rms, small ?rm receive more trade credit since large ?rms demand less trade credit. Gallego and Hernandez give the 1998 ?nancial market turmoil in Chile as an anecdotal example: “When interest rates in Chile (and in other emerging market economies) reached extremely high levels. During this period a group of large ?rms arbitrarily extended the payment period to suppliers from 90 to 180 days, forcing smaller ?rms to assume the increase in the cost of funds (p.17).” Existing empirical evidence strongly supports the hypothesis that small ?rms, not large ?rms, bene?t more. These works come in a variety of forms, including panel data studies, event studies, single-country studies, and cross-country studies. Examples that represent each genre of work are summarized in the appendix. The ?rst example is a cross-country panel data study by Knill (2005). She studies a panel of ?rms from 53 countries during 1996 to 2005 and ?nds that foreign portfolio investment is associated with an increase in the ability to issue securities for small ?rms. Additionally, she ?nds that foreign portfolio investment increases the maturity of bank loans, leading her to conclude that small ?rms bene?t more because they rely more on bank loans than large ?rms. The next group of papers (Harris, Schiantarelli, and Siregar 1994; Jaramillo, Schiantarelli, and Weiss 1996; Gelos and Werner 2002; Laeven 2003; Koo and Shin 2004; Contreras and Makaew 2007) analyzes ?rm behaviors before and after ?nancial liberalization, the event that leads to a large change in foreign investment. Most of these papers ?nd that ?nancial liberalization has di?erent impacts on small and large ?rms: compared to large ?rms, it a?ects small ?rms by further relaxing ?nancial constraints, lowering investment-cash ?ow sensitivity, increasing investment, and lowering purchasing price of capital. The last example, which is closest to my work, is Patro and Wald (2005). They study the impacts of stock market liberalization on small and large ?rms by extending the event study framework of Henry (2000) to a cross-sectional event study. Using the stock market data from 18 developing countries, they ?nd that small ?rms earn signi?cantly higher abnormal returns when stock markets are liberalized. 7
Even though the amount of the existing evidence supporting the “small-?rms-bene?tmore” hypothesis is overwhelming, it is not clear whether small ?rms bene?t from foreign portfolio investment or from other factors correlated with the surge in foreign capital. Stock market and capital account liberalizations are often concurrent with (1) banking deregulations (reduction in reserve requirements and credit controls; privatizations of state banks; allowing foreign bank entries), (2) reforms in stock market regulations and banking supervisions, and (3) trade liberalization. For instance, the Korean stock market liberalization was concurrent with interest rate deregulations and a strengthening of prudential regulations. The Colombian capital account liberalization was also concurrent with constitutional reforms and banking deregulations. I believe that identi?cation of the bene?ts from foreign portfolio investment is di?cult, given that countries’ economic prospects are changing rapidly along with their capital account policies. It is virtually impossible to list all the events that a?ect the value of small ?rms. Even if I could identify all the relevant factors, the liberalization process is still complicated and dynamic by nature. It is di?cult to identify the exact date these changes took place (in order to control for them in a panel data study) or identify the exact time the market learned about them (in order to control for them in an event study). In this paper, I propose a cross-sectional analysis of a unique event in Thailand. The Thai government imposed a draconian capital control on December 19, 2006 and then quickly abandoned it on December 20, 2006. This experiment-like event helps me separate the impact of foreign capital from the impact of other concurrent events.
1.3
Natural Experiment
In this section, I describe the capital control and liberalization which is the event of interest but I will ?rst discuss the political and economic situations in Thailand that lead up to the event. Thaksin Shinawatra and the September 2006 Coup From January 2001 to September 2006, Thailand was under the administration of Prime
8
Minister Thaksin Shinawatra who was also a successful businessperson and one of the richest people in the country. He and his family were major shareholders of many listed ?rms in the Stock Exchange of Thailand including Advanced Info Service - the largest mobile phone operators in Thailand, Shin Satellite - the only operator of Thailand’s commercial satellites, and ITV - a television station. Even though Prime Minister Thaksin swept the elections in 2001 and 2005, his popularity started to decline in late 2005 when he was accused of fraud, human rights o?enses and lese-majeste. On September 19, 2006, the Thai Military staged a coup against Prime Minister Thaksin and overthrew his government while he was attending the United Nation Assembly in New York. The new Prime Minister, as well as the new cabinet, and the new governor of the Thai central bank were appointed in October 2006 and November 2006, respectively. Prime Minister Thaksin is currently in exile. The One-Day Capital Control On December 19, 2006, the Thai central bank had decided to implement a reserve requirement on short-term capital in?ows. Under this new regime, foreigners bringing portfolio capital into Thailand had to deposit 30% of the funds into an account at the central bank which would earn no interest. This meant that only 70% of the funds would be available for investment in the Thai market. Moreover, if foreign investors wished to withdraw their money within one year, they would be ?ned 1/3 of the amount. The control was targeted straight at future portfolio capital in?ows. The central bank stated clearly that foreign direct investment was not subjected to this reserve requirement.1 Any foreign exchange transactions which had been traded before the announcement were also exempted. As anyone would expect, the Thai stock market reacted to this surprising news immediately; the Stock Exchange of Thailand (SET) Index dropped from 730.55 to 622.14 (i.e. 14.84% reduction in one day). On December 20, 2006, the central bank announced that in?ows for the investment in the Stock Exchange of Thailand, the Thai Market for Alternative Investment, the Thai Futures Exchange, and the Agricultural Futures Exchange of Thailand, which are basically most of
1
Initially, foreign direct investment was also required to place 30% of the in?ow as a reserve requirement,
but after submitting the relevant documents to support the claim of legitimacy, the central bank would refund the reserve amount.
9
the portfolio ?ows into Thailand, were no longer subjected to the 30% reserve requirement. Again, this announcement took the market by surprise and the SET Index bounced back from 622.14 to 691.55 (or 11.16% increase in one day).
[INSERT FIGURE 1.1 HERE]
The fact that the capital control restriction only lasted for one day provides a proper framework for a natural experiment study. It is di?cult to come up with another factor that is unrelated to capital control, has as dominant of an e?ect on ?rms compared to capital control, and changes back and forth overnight like capital control. For example, one might argue validly that the stock return on the capital control day may re?ect both changes in foreign capital and changes in investor’s perception about the Thai government’s ability to run the economy e?ectively. However, it is not likely that this perception was largely reversed back to normal when the government reversed its decision about the capital control the next day.
1.4
Data and Methodology
In this section, I describe the sample ?rms, the variables used and how they are constructed. The dataset consists of all Thai ?rms listed in the Stock Exchange of Thailand. All of the trading data are from the Reuters Database. All of the ?nancial statement data are from Reuters and the Stock Exchange of Thailand Market Analysis and Reporting Tool (SETSMART) Database. Daily stock prices are the last reported trade prices. Other ?rm characteristics are measured at the end of 2005 since most ?rms in Thailand report their ?nancial status at the end of December and ?rm characteristics measured at the end of 2006 might be contaminated by the e?ects of the experiment already. The details how each ?rm characteristic is constructed are in the appendix. The summary statistics are provided in Table 1.1A. I compute abnormal returns using the market model as the benchmark: Ri = ?i + ?i RM + 10
where RM is the percentage change in the MSCI Emerging Markets Asia Index. The market model is estimated by the daily returns from September 29, 2005 to August 31, 2006 (a 261-trading-day period). Abnormal returns, ARi , are de?ned as: ˆi RM ) ARi = Ri ? (? ˆi + ? ˆi are stock i’s estimated market model coe?cients. I also exclude all the where ? ˆi and ? ?rms that are not traded on the capital control day or the liberalization day. My empirical strategy is to link the abnormal returns on the capital control day and on the liberalization day with ?rm sizes and other control variables. Firms that experienced larger reduction in value on the capital control day and ?rms that experienced larger gain on the liberalization day should be the ones that bene?t more from foreign portfolio investment. [INSERT TABLE 1.1 HERE]
1.5
Analysis
Table 1.1B presents a univariate comparison of the abnormal returns and ?rm characteristics across four size quartiles. On average, ?rms in the largest quartile earn 2.94% less on the capital control day and 3.27% more on the liberalization day, compared to ?rms in the smallest quartile. The di?erences are statistically signi?cant at a 95% con?dence level. This suggests that large ?rms are a?ected by foreign portfolio investment more than small ?rms. Table 1.1B also suggests that there are systematic di?erences in the characteristics of small and large ?rms. Therefore, multiple regression analysis will be performed in the next section, but for now I use the propensity score method as a preliminary analysis to get a better feel of the data. The propensity score method matches treated ?rms with control ?rms that have the nearest propensity scores. I assign the largest size quartile as the treatment group and the smallest quartile as the control group. Propensity scores are computed from the probit model predicting the probability of being in the treatment
11
group using four covariates: pro?tability, exchange rate exposure, foreign director, and Thaksin connection. (These four variables are chosen because in the next section, I ?nd that they are indeed the most relevant variables.) Table 1.1C compares the abnormal returns of ?rms in the largest quartile to the abnormal returns of propensity-score matched ?rms from the smallest quartile (average treatment e?ect on the treated). After matching, the di?erences in abnormal returns between small and large ?rms change slightly; the magnitude of the t-statistic drops from 2.78 to 2.18 on the capital control day and from 3.25 to 2.72 on the liberalization day. The E?ects of Firm Size and Other Financial Characteristics In this section, I analyze the e?ects of ?rm size on the bene?t from foreign portfolio investment by regressing the abnormal return on size and other ?rm characteristics. I use the least square method with robust standard errors. Firm Size is measured by log of total asset. Besides ?rm size, other ?nancial characteristics might also determine how ?rms will be a?ected when the supply of capital decreases so I have to add these variables to the regressions. The ?rst set of control variables I include are the 9 Industry Dummies classi?ed by the Stock Exchange of Thailand. These industry dummies capture any industry-level changes in goods and capital market conditions. They also ?lter out any e?ects of the cross-sectional co-movement in stock returns that are driven by industry-level factors. Next, I include accounting pro?tability, market-to-book ratio and cash ?ow growth to capture ?rm investment opportunities. Pro?tability is measured as pre-tax pro?t or loss scaled by lagged total asset. Market-to-Book is measured as market value of equity divided by book value of equity. Cash Flow Growth is measured as lagged annual growth in operating cash ?ow. Theoretically, ?rms with better investment opportunities should have higher demand for fund and hence should be a?ected by the capital control more. Unlike market-to-book that captures investment opportunities in the future, accounting pro?tability and cash ?ow growth measured in the previous year are also proxies for ?rm ability to generate internal fund in the short-run. Firms that can generate more internal fund, and hence rely on external ?nancing less, should be a?ected by the capital control less. Finally, I include accounting liquidity and leverage. Liquidity is measured as net working capital (current asset minus current liability) scaled by lagged
12
total asset. Leverage is measured as total debt scaled by lagged total asset. When the supply of capital decreases, ?rms that have less current asset compared to current debt are more likely to face a liquidity problem. At the same time, ?rms that have higher leverage might have trouble paying interests and are more likely to be bankrupt. Therefore, ?rms with lower liquidity and higher leverage should be a?ected by the capital control more. [INSERT TABLE 1.2 HERE] Results Table 1.2A provides correlations between ?rm size and other characteristics. From the size column, I can see that size is not strongly correlated with any other variables. Large ?rms tend to be slightly less liquid and have higher leverage. No variables are strongly correlated with one another; the highest correlation of 24% is between liquidity and pro?tability. Therefore, multi-colinearity should not be a problem in my analysis. The regression results are reported in Table 1.2B. In Panel A, the dependent variable is the abnormal return on the capital control day. In Panel B, the dependent variable is the abnormal return on the liberalization day. I ?nd that the coe?cients on size are statistically signi?cant at a 95% or 99% con?dence levels in all speci?cations. The coe?cients on size are uniformly negative on the capital control day and uniformly positive on the liberalization day. This means that large ?rms lose more when a capital control is imposed and bene?t more when a capital control is lifted. In other words, large ?rms bene?t from foreign portfolio investment more. The estimated size coe?cients indicate that the size e?ect is economically large. From the full speci?cation (Model 7), one standard deviation increase in size leads to 0.97% reduction in ?rm value on the capital control day and 1.13% increase on the liberalization day. Other results suggest that ?rms with higher pro?tability bene?t less from foreign portfolio investment. This is consistent with the hypothesis that ?rm with higher pro?tability are less sensitive to changes in external capital markets. The economic signi?cance of pro?tability is comparable to size’s. From the full speci?cation, one standard deviation increase in pro?tability leads to 1.27% increase in ?rm value on the capital control day and 1.05% decrease in ?rm value on the liberalization day. In panel A, market-to-book also 13
has statistically negative coe?cients, suggesting that ?rms that have better investment opportunities are a?ected more from the capital control. The coe?cients on cash ?ow growth, liquidity and leverage are not statistically signi?cant and these variables do not increase the statistical ?t of my models. Firm Size and International Involvement In this section, I examine the e?ects of ?rm’s international involvement. In particular, I analyze the relationship between ?rm size, exchange rate exposure, foreign ownership, and foreign control. I then include these variables in the regressions in order to examine how much of the size e?ects found in the previous section are due to the di?erent degrees of international involvement between small and large ?rms. A ?rm’s Exchange Rate Exposure is proxied by exchange rate beta calculated from a factor model (see details in the appendix). Firms that have a positive exchange rate beta are likely to be ?rms that earn income in US dollars and have expenditure in the local currency (Thai Bahts) such as export-oriented ?rms and ?rms that own income-generating assets abroad. The high-beta ?rms should su?er less or even pro?t from the capital control. Since a control on capital in?ow automatically reduces the demand for Thai Bahts relative to US dollars, these ?rms’ cash ?ow in Thai Bahts will increase as a result of exchange rate depreciation. The next variable is Foreign Ownership which is measured as the fraction of ?rm’s equity owned by non-Thai citizens. Firms with a higher foreign ownership fraction tend to rely on foreign capital more and hence should be a?ected by the capital control more. The last control variable is the Foreign Director Dummy. This dummy takes the value of one if a ?rm has at least one non-Thai citizen as a director and zero otherwise. I have to include Foreign Director dummy because foreign control and foreign ownership sometimes do not go hand in hand as foreign ownership might be di?used. Theoretically, ?rms that have a foreign director should be a?ected by the capital control less since (1) foreign directors might provide better corporate governance. Firms with better corporate governance are able to attract more external capital when needed and hence are a?ected by the capital control less. (2) The existence of a foreign director might re?ect the fact that foreign investment in that ?rm is a non-di?used direct investment rather than portfolio investment and foreign direct investment is exempt from the December 19 capital control
14
in the ?rst place. [INSERT TABLE 1.3 HERE] Results Table 1.3A reports correlations between ?rm size and international involvement variables. Size is strongly and positively correlated with foreign ownership with the correlation coe?cient of 40.5%. This is consistent with the ?ndings of Kang and Stulz (1997) as well as Dahlquist and Robertson (2001) that foreign institutional investors tend to invest in larger ?rms. Size is also negatively correlated with exchange rate exposure which is partially due to the industry e?ects; ?rms in export-oriented industries (such as textile and food processing) tend to be smaller than ?rms domestic-oriented industries (such as real estates and telecommunication). As expected, ?rms that have higher foreign ownership fractions are more likely to have foreign directors. Therefore, foreign director dummy is strongly correlated with foreign ownership (51.24%). However, the correlation between size and foreign director is much weaker, only 9.64%. The regression results are reported in Table 3B. In Panel A, the dependent variable is the abnormal return on the capital control day. In Panel B, the dependent variable is the abnormal return on the liberalization day. I ?nd that the coe?cients on exchange rate exposure are signi?cant at a 95% or 99% con?dence level. As anticipated, these coe?cients are positive on the capital control day and negative on the liberalization day. It is likely that this ?nding re?ects the fact that ?rms with income in foreign currencies should be a?ected by the capital control less. The economic signi?cance of exchange rate exposure is large; from Model 2, one standard deviation increase in exchange rate beta leads to 1.45% increase in ?rm value on the capital control day and 1.16% decline on the liberalization day. From Model 3, I ?nd that the coe?cients on foreign ownership fraction are not statistically signi?cant. One of the plausible explanations is that ?rm size is a better proxy for the bene?t from future foreign investment, compared to foreign ownership fraction which re?ects past investment. The insigni?cance of foreign ownership also rules out another alternative explanation for the size e?ect: one might claim that the size e?ect found the 15
previous section is simply a result of foreign investors getting panic and liquidating their positions (which are mostly large ?rms) on the capital control day. If this explanation were valid, the foreign ownership variable would have driven out the signi?cance of ?rm size. From Model 4, I ?nd that ?rms that have a foreign director are a?ected less by the capital control. On average, ?rms with foreign directors earn 1.16% more on the day of the capital control and 2.19% less on the liberalization day. The e?ect of ?rm size is still large and signi?cant even after controlling for a ?rm’s international involvement. Comparing before and after including the control variables, on the capital control day, the magnitude of the size coe?cients drops slightly from -0.74 in the baseline model (Model 1) to -0.60 in the full speci?cation (Model 4) but remains statistically signi?cant at a 95% level. Similarly, on the liberalization day, the magnitude drops from 0.70 to 0.55 but remains statistically signi?cant at a 95% level. The ?ndings that size and foreign ownership is positively correlated and that size might be better as a proxy for the bene?t from future foreign investment are interesting. Therefore, I further investigate the relationship between ?rm size and other capital market activities. Table 3C reports the ?rms’ activities classi?ed into four size quartiles. From the ?rst three columns, I ?nd that large ?rms are more visible to foreign investors in the sense that most ?rms that are constituents of key national indices and have credit ratings are from the largest quartile. I also ?nd that large ?rms are more likely to engage in international capital market activities: In the secondary market, 9 out of 12 ?rms in the sample that are cross-listed or have over-the-counter ADRs are from the largest quartile. In the primary market, 75% of ?rms that have issued equities or debts in international capital market (from 1990 to 2006) are from the largest quartile. In short, I con?rm that size is highly correlated with international capital market activities in general. Firm Size and Political / Family Connections In this section, I analyze the relationship between ?rm size and its political and family business group connections. A number of studies have documented that (1) large ?rms are more likely to have political connections (Faccio 2006), (2) drastic government policies tend to a?ect connected ?rms and unconnected ?rms di?erently (Johnson and Mitton 16
2003; Faccio, McConnell and Masulis 2006; and others), and (3) a ?rm’s international capital market activities are in?uenced by its political connections (Leuz and OberholzerGee 2006). The event of interest, the capital control, was imposed during the tenure of the coup government, shortly after throwing out Prime Minister Thaksin. Therefore, it is important that I include these political and family business group connections in the regressions in order to examine how much of the size e?ect is due to the di?erent degrees of connectedness. The ?rst variable captures the direct political connection to the former Prime Minister Thaksin who is the rival of the coup government. Thaksin Connection is a dummy variable that takes the value of one if the ?rm’s major shareholder is, or is blood-related to, a member of Thaksin’s Cabinet and zero otherwise. Firms that are politically connected to Prime Minister Thaksin should su?er from the capital control more since these ?rms are more likely to have di?culties obtaining funds from domestic ?nancial institutions when the coup government is in power. Next, I include the family business group variable. In Thailand, like in many other East Asian countries, business groups consist of ?rms whose major shareholders are relatives. Family Business Group dummy takes the value of one if the ?rm’s major shareholder is from the 50 largest family business groups in Thailand and zero otherwise. Theoretically, ?rms in large business group can internalize many capital allocation functions and hence rely less on external capital markets. Therefore, ?rms that belong to a business group should su?er less from the capital control. Finally, I also allow for a more general de?nition of political connections. Faccio (2006) de?nes politically connected ?rms broadly as ?rms that (1) have a major shareholder or a top executive who is a parliament member, minister or head of the state, or (2) have a major shareholder or a top executive who is related to a top politician or a political party. Political Connection is a dummy variable that takes the value of one if the ?rm is politically connected according to Faccio (2006) and zero otherwise. I note that Political Connection captures political connections to anyone in Thai politics and Thaksin Connection should be a subset of Political Connection. [INSERT TABLE 1.4 HERE]
17
Results Table 1.4A reports correlations between ?rm size and the connection variables. I con?rm that ?rms that are connected to Thaksin, ?rms that are connected to a top family business group and ?rms with any political connections tend to be larger. The correlation between ?rm size and the political connection dummy is 30.19% suggesting that large ?rms tend to have some sorts of connections, either with Thaksin or with other people in high o?ces. The correlation between Thaksin connection and family business group connection is very strong (42.31%) suggesting that many of the top family ?rms in Thailand have someone representing them in the Thaksin administration. The regression results are reported in Table 4B. As usual, the dependent variable in Panel A is the abnormal return on the capital control day. The dependent variable in Panel B is the abnormal return on the liberalization day. From Panel B Model 1, I ?nd that the coe?cient on Thaksin Connection is signi?cant while all other connection coe?cients are not. (Since only 5% of ?rms in the sample are directly connected to Thaksin and the connected ?rms are concentrated in a few industries, it is natural that the level of signi?cance is not high.) This supports the belief that political connection does a?ect ?rm value. In fact, my ?nding here is closely related to the ?ndings of Johnson and Mitton (2003). They view the 1998 capital control in Malaysia as a way to support ?rms connected to the incumbent government. Here the market views Thai capital control as a way to punish ?rms that are connected to the opponent of the people in power. After taking a closer look at the ?rms connected with Thaksin, I ?nd that these ?rms are also the ones that heavily engage in international capital market activities. For example, 7 out of 12 ?rms in the sample that have ADRs are connected to Thaksin and approximately half of ?rms that are connected to Thaksin have issued equities in international capital markets. I ?nd that the e?ect of ?rm size is still large and signi?cant after controlling for the connection e?ects. On the capital control day, the magnitude of the size coe?cients drops from -0.74 in the baseline model to -0.53 in the model with Thaksin Connection dummy but remains statistically signi?cant at a 95% level. Similarly, on the liberalization day, the coe?cient drops from 0.70 to 0.54 but remains statistically signi?cant at a 95% level.
18
1.6
Robustness Tests
Are large ?rms more susceptible to any bad news? One might argue large ?rms are more susceptible to any negative national news compared to small ?rms, possibly because small ?rms are less liquid and small ?rms’ stocks are more closely held. My ?rst response to this argument is that I have excluded all the stocks that are not traded on the capital control day or the liberalization day. Conditioned on being traded, simple microstructure models predict that less-liquid ?rms should su?er more on a bad day since the prices of illiquid stocks have to decline more in order to induce trade. My second response is to use the day the market learn about the September 2006 coup in Thailand as a placebo test. The coup is a good candidate for a placebo as it is a bad news that a?ects the entire economy within a short period of time before the capital control. If the hypothesis that large ?rms are more susceptible to any bad macro news is true, I should ?nd that, after controlling for other factors, large ?rms were a?ected more. [INSERT TABLE 1.5 HERE] Results Table 1.5 reports the placebo regression results. The dependent variable is the abnormal return on September 23, 2006 which is the ?rst trading day that the market learned about the coup. In all speci?cations, I ?nd that the coe?cient on size is positive and signi?cant at a 99% con?dence level. In other words, large ?rms are a?ected by the coup less. This is consistent with the ?ight to quality hypothesis: when bad things happen, large ?rms become more attractive relative to small ?rms. More importantly, this ?nding rules out the alternative explanation that large ?rms are more sensitive to any bad news. I also note that the coe?cient of Thaksin Connection is negative and signi?cant. From Model 6, ?rms that are connected to Thaksin earn 3.23% less than other ?rms when the Thaksin government lost power. I take this as a con?rmation that my measure of connections with Prime Minister Thaksin is legitimate. Econometric Issues I use a number of techniques to address econometric concerns regarding the non-normality 19
of the error terms. First, I note that in all of The previous regressions the t-statistics are computed from Huber/White/sandwich standard errors so they are already robust to the heteroscedasticity problem. Second, a more serious problem here is the cross-sectional correlation problem because the event dates are the same across all ?rms in my analysis. My solution is to cluster standard errors both at the ?rm-level and the industry-level to eliminate the biases from stock return co-movements. I ?nd that the signi?cance of the size coe?cient barely changes when clustered at ?rm-level and even improves when clustered at the industry-level. Bootstrapping Third, I address any non-normality problems in the error terms by using empirical standard errors. I use two methods to generate an empirical distribution. In the ?rst one, I create 1,000 synthetic samples from the original dataset and estimate the full speci?cation using these samples. I then compute the z-scores of the size coe?cient using the empirical standard errors from these 1,000 synthetic betas. Once again, I obtain a similar result the size coe?cient is still statistically signi?cant at a 95% con?dence level. In the second method, I use the historical data to generate the empirical distribution. I perform the cross-sectional regression everyday from January to September 2006. I then compute the z-score of the size coe?cient using the standard errors computed from these daily beta estimates. This time, the empirical distribution computed from historical data yields a very high signi?cance level because size is consistently a poor predictor of daily abnormal returns. The mean of the size coe?cient on a regular day is -0.0005 (compared to -0.53 and 0.54 on the event days). The empirical standard deviation is 0.0227. Therefore, the z-value of the size coe?cient becomes 23.28 which is statistically signi?cant at a 99.99% con?dence level. Alternative De?nitions I con?rm that my results are robust to alternative variable de?nitions. I use log of market capitalization and log of market capitalization adjusted for free-?oat as alternative proxies for size, GIC industry classi?cation for industry dummies, sales for pro?tability, and cash holdings for liquidity. I also use raw returns instead of abnormal returns. All the results are qualitatively similar to what I found earlier. 20
1.7
Conclusion and Discussion
In this paper, I examine the e?ect foreign portfolio investment has on ?rms of di?erent sizes using Thailand’s unique restriction on short-term capital in?ow as a natural experiment. The fact that this restriction was a surprise and only lasted one day makes it an appropriate set-up for an event-study. In contrast to the majority of existing literature, my evidence suggests that large ?rms bene?t from foreign portfolio investment more; I ?nd that large ?rms stock market valuations were hurt by the capital control and helped by the subsequent liberalization. Compared to small ?rms, large ?rms have a higher fraction of foreign ownership and are more likely to have political and business connections. After controlling for ?rm ?nancial characteristics, international involvements and connections, size still has a large and signi?cant explanatory power. Other results suggest that ?rms with higher accounting pro?tability, export-oriented ?rms, and ?rms with foreign directors are less a?ected by the capital control while ?rms with higher market-to-book and ?rms connected to Prime Minister Thaksin, the major opponent of the incumbent coup government, were a?ected more. My results are robust to various econometric speci?cations and variable de?nitions. This paper contributes to the international ?nance literature in several ways. First, I provide a clean framework to test the size e?ects and, from this, I ?nd that the results from the existing literature are reversed. Second, by including a wide range of control variables, I identify that certain types of ?rms are a?ected more when the supply of foreign fund declines. Third, I con?rm that, in emerging markets, political connection does a?ect ?rm value. Firms connected to Prime Minister Thaksin su?ered more when the market learned about the coup and when the capital control was announced. The main implication of my study is not that foreign portfolio investment does not bene?t small ?rms. Rather what I ?nd is that, after separating other events or factors that are typically correlated with foreign portfolio investment, foreign portfolio investment by itself does not favor small ?rms. Therefore, my results, taken together with the existing literature, imply that the “other factors” concurrent with stock market and capital account liberalizations are very important. Whatever governments were doing at the time of
21
liberalizations - deregulating the banking systems, improving capital market supervisions, liberalizing international trade etc. - is probably good for small ?rms. In sum, my study calls for more research on how domestic reforms could channel funds to small ?rms who need it the most, and how domestic market conditions interact with foreign portfolio investment.
22
Table 1.1A: Summary Statistics Quartile Variable Firm Size Pro?tability Market-to-Book Cash Flow Growth Liquidity Leverage Exchange Rate Exposure Foreign Ownership Foreign Director Thaksin Connection Family Business Group Political Connection Mean 22.2 2.07 2.59 -58.03 19.33 51.23 -0.63 19.32 0.34 0.05 0.10 0.08 SD 1.57 3.66 21.89 576.8 29.38 28.38 0.66 21.22 0.47 0.22 0.29 0.26 Min 19.46 -15.45 -5.03 -3567.23 -51.15 0 -2.81 0 0 0 0 0 0.25 21.1 0.42 0.65 -107.74 1 29.93 -1.07 1.45 0 0 0 0 Median 21.88 1.85 1.02 -21.77 14.97 50.67 -0.58 12.19 0 0 0 0 0.75 22.97 3.43 1.53 41.98 36.51 70.68 -0.13 30.6 1 0 0 0 Max 27.97 17.01 409.18 5304.19 161.81 150.39 0.75 95.56 1 1 1 1
Firm size is measured as the log of total asset; Pro?tability is measured as pre-tax pro?t or loss scaled by lagged total asset; Market-to-Book is measured as market value of equity divided by book value of equity; Cash Flow Growth is measured as lagged annual growth in operating cash ?ow; Liquidity is measured as net working capital (current asset - current liability) scaled by lagged total asset; Leverage is measured as total debt scaled by lagged total asset; Exchange Rate Exposure is measured as the exchange rate beta estimated from the multi-factor model; Foreign ownership is a fraction of the ?rm that is owned by non-Thai citizens; Foreign Director is a dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen as a director and zero otherwise; Thaksin Connection is a dummy variable that takes the value of one if the ?rm’s major shareholder is, or is related to (has the same last name as), a member of Thaksin’s Cabinet and zero otherwise; Family Business Group is a dummy variable that takes the value of one if the ?rm’s major shareholder is from the 50 largest family business groups in Thailand and zero otherwise; Political Connection is a dummy variable that takes the value of one if the ?rm is politically connected according to Faccio (2006). The details how each variable is constructed are in the appendix.
23
Table 1.1B: Univariate Analysis Size Quartile Smallest Capital Control Day Abnormal Return Liberalization Day Abnormal Return Firm Size Pro?tability Market-to-Book Cash Flow Growth Liquidity Leverage Exchange Rate Exposure Foreign Ownership Foreign Director Thaksin Connection Family Business Group Political Connection -10.38 7.29 20.56 1.18 6.31 -65.62 19.09 47.91 -0.43 7.61 0.22 0.02 0.01 0 Second -11.05 8.15 21.43 2.09 1.04 35.05 22.87 49.48 -0.57 13.32 0.27 0.02 0.05 0.02 Third -11.16 8.25 22.38 1.9 1.33 -83.31 17.15 58.32 -0.63 22.96 0.43 0.05 0.07 0.1 Largest -13.28 10.67 24.34 1.73 1.94 -117.68 12.13 54.62 -0.86 33.37 0.4 0.12 0.24 0.17
The table reports summary statistics of abnormal returns and ?rm characteristics classi?ed into four size quartiles. The capital control day is December 19, 2006 and the liberalization day is December 20, 2006. Abnormal returns are from market model estimated by the daily returns from September 29, 2005 to August 31, 2006 (a 261-trading-day period). All ?rms that were not traded on the capital control day and the liberalization day are excluded. Firm size is measured as the log of total asset; Pro?tability is measured as pre-tax pro?t or loss scaled by lagged total asset; Market-to-Book is measured as market value of equity divided by book value of equity; Cash Flow Growth is measured as lagged annual growth in operating cash ?ow; Liquidity is measured as net working capital (current asset - current liability) scaled by lagged total asset; Leverage is measured as total debt scaled by lagged total asset; Exchange Rate Exposure is measured as the exchange rate beta estimated from the multi-factor model; Foreign ownership is a fraction of the ?rm that is owned by non-Thai citizens; Foreign Director is a dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen as a director and zero otherwise; Thaksin Connection is a dummy variable that takes the value of one if the ?rm’s major shareholder is, or is related to (has the same last name as), a member of Thaksin’s Cabinet and zero otherwise; Family Business Group is a dummy variable that takes the value of one if the ?rm’s major shareholder is from the 50 largest family business groups in Thailand and zero otherwise; Political Connection is a dummy variable that takes the value of one if the ?rm is politically connected according to Faccio (2006).
24
Table 1.1C: Propensity Score Matching Treated (Largest) Capital Control Day Abnormal Return Unmatched -13.3067 -10.3577 -2.949 [-2.78] Average Treatment E?ect on the Treated -13.3067 -10.2044 -3.1023 [-2.18] Liberalization Day Abnormal Return Unmatched 10.6555 7.3836 3.2719 [3.25] Average Treatment E?ect on the Treated 10.6555 7.115 3.5405 [2.72] Controls (Smallest) Di?erence (Largest-Smallest)
The table compares the abnormal returns from the treatment group and the control group. The treatment group is the largest size quartile and the control group is the smallest quartile. The capital control day is December 19, 2006 and the liberalization day is December 20, 2006. Abnormal returns are from market model estimated by the daily returns from September 29, 2005 to August 31, 2006 (a 261-trading-day period). Both unmatched e?ects and propensity-score-matched e?ects (average treatment e?ects on the treated) are reported. Propensity scores are computed from the Probit model using four covariates: profitability, exchange rate exposure, foreign director dummy, and Thaksin connection. Firm size is measured as the log of total asset; Pro?tability is measured as pre-tax pro?t or loss scaled by lagged total asset; Foreign Director is a dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen as a director and zero otherwise; Thaksin Connection is a dummy variable that takes the value of one if the ?rm’s major shareholder is, or is related to (has the same last name as), a member of Thaksin’s Cabinet and zero otherwise. Numbers in the brackets are t-statistics.
25
Table 1.2A: Correlation Matrix: Size and Basic Firm Characteristics Firm Size Firm Size Pro?tability Market-to-Book Cash Flow Growth Liquidity Leverage 1 -0.0538 -0.0911 -0.0015 -0.1926 0.148 1 -0.197 -0.0107 0.2405 0.0536 1 0.0128 -0.0761 0.0413 1 -0.0079 0.0427 1 -0.1998 1 Pro?tability Market-to-Book Cash Flow Growth Liquidity Leverage
Firm size is measured as the log of total asset; Pro?tability is measured as pre-tax pro?t or loss scaled by lagged total asset; Market-to-Book is measured as market value of equity divided by book value of equity; Cash Flow Growth is measured as lagged annual growth in operating cash ?ow; Liquidity is measured as net working capital (current asset - current liability) scaled by lagged total asset; Leverage is measured as total debt scaled by lagged total asset.
26
Table 1.2B: Basic Firm Characteristics and the E?ects of Foreign Portfolio Investment Panel A Dependent Variable Firm Size Model 1 -0.8392*** [-4.24] Pro?tability Model 2 -0.7564*** [-3.3] Model 3 -0.7457*** [-3.31] 0.3619*** [3.65] Market-to-Book Model 4 -0.742*** [-3.08] 0.3042*** [3.01] -0.0274*** [-5.28] Cash Flow Growth Model 5 -0.7682*** [-3.14] 0.3038*** [3.1] -0.0282*** [-5.39] -0.0009 [-0.81] Liquidity Model 6 -0.6195** [-2.09] 0.346*** [3.18] -0.0259*** [-4.93] -0.0009 [-0.65] -0.0026 [-0.17] Leverage Model 7 -0.6148** [-2.08] 0.3481*** [3.18] -0.0257*** [-4.88] -0.0008 [-0.64] -0.0032 [-0.21] -0.0024 [-0.16] Constant Industry Dummy R-squared Observations Yes No 0.0367 312 Yes Yes 0.0927 309 Yes Yes 0.1232 306 Yes Yes 0.1326 294 Yes Yes 0.1374 293 Yes Yes 0.1479 265 Yes Yes 0.148 265
27
Table 1.2B (Continued): Basic Firm Characteristics and the E?ects of Foreign Portfolio Investment Panel B Dependent Variable Firm Size Model 1 0.7397*** [3.92] Pro?tability Model 2 0.7267*** [3.37] Model 3 0.7092*** [3.27] -0.2642*** [-2.64] Market-to-Book Model 4 0.7385*** [3.18] -0.2419** [-2.24] 0.003 [0.6] Cash Flow Growth Model 5 0.7804*** [3.38] -0.2398** [-2.32] 0.0042 [0.92] 0.0014 [1.42] Liquidity Model 6 0.753*** [2.63] -0.2731** [-2.47] 0.0029 [0.6] 0.0014 [1.1] -0.0011 [-0.07] Leverage Model 7 0.7204** [2.49] -0.2878*** [-2.61] 0.0015 [0.31] 0.0013 [1.05] 0.0031 [0.19] 0.0164 [1.12] Constant Industry Dummy R-squared Observations Yes No 0.0324 312 Yes Yes 0.0779 309 Yes Yes 0.0969 306 Yes Yes 0.1007 294 Yes Yes 0.1151 293 Yes Yes 0.127 265 Yes Yes 0.1313 265
The table reports the coe?cient estimates from regressions of abnormal returns on ?rm characteristics. The dependent variable in Panel A is the abnormal return on December 19, 2006 (the capital control day) and the dependent variable in Panel B is the abnormal return on December 20, (the liberalization day). Firm size is measured as the log of total asset; Pro?tability is measured as pre-tax pro?t or loss scaled by lagged total asset; Market-to-Book is measured as market value of equity divided by book value of equity; Cash Flow Growth is measured as lagged annual growth in operating cash ?ow; Liquidity is measured as net working capital (current asset - current liability) scaled by lagged total asset; Leverage is measured as total debt scaled by lagged total asset. All variables are from December 2004 and 2005 ?nancial statements. Also estimated but not reported are a constant term and 9-industry dummy variables. Numbers in the brackets are heteroscedasticity-robust t-statistics. *, **, and *** indicate statistical signi?cant at 10, 5, and 1 percent levels, respectively. R-squared and the number of observations are reported in the last two rows.
28
Table 1.3A: Correlation Matrix: Size and Firm’s International Involvement Firm Size 1 -0.0418 -0.2724 0.405 0.0964 0.0207 -0.0726 0.5124 1 0.0495 -0.1343 1 0.1234 1 1 Pro?tability Exchange Rate Exposure Foreign Ownership Foreign Director
Firm Size
Pro?tability
Exchange Rate Exposure
Foreign Ownership
Foreign Director
Exchange Rate Exposure is measured as the exchange rate beta estimated from the multi-factor model; Foreign ownership is a fraction of the ?rm
that is owned by non-Thai citizens; Foreign Director is a dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen as
a director and zero otherwise.
29
Table 1.3B: Firms International Involvement and the E?ects of Foreign Portfolio Investment Panel A Model 1 -0.7457*** [-3.31] 0.3619*** [3.65] 2.1986*** [3.18] 0.0172 [0.95] 1.1604 [1.18] Yes Yes 0.1232 306 306 302 302 0.159 0.1598 0.1642 Yes Yes Yes Yes 0.0969 306 Yes Yes Yes Yes Yes Yes 0.1229 306 Yes Yes 0.1216 302 [0.11] 0.0025 [3.13] [3.24] [-2.69] 2.2274*** 2.2872*** -1.7571*** [3.31] [3.24] [3.26] [-2.64] [-2.3] [-2.27] -1.7922*** [-2.68] -0.0129 [-0.6] 0.3266*** 0.316*** 0.3169*** -0.2642*** -0.236** -0.2291** [-2.35] [-2.62] [-2.42] [3.27] [2.5] [2.64] [2.3] -0.2308** [-2.3] -1.9052*** [-2.88] 0.0148 [0.54] -2.1936** [-2.12] Yes Yes 0.1394 302 -0.5426** -0.6438*** -0.6026** 0.7092*** 0.5469** 0.6273*** 0.5494** Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4 Panel B
Dependent Variable
Firm Size
Pro?tability
Exchange Rate Exposure
Foreign Ownership
Foreign Director
Constant
30
Industry Dummy
R-squared
Observations
The table reports the coe?cient estimates from regressions of abnormal returns on ?rm’s international involvement. The dependent variable in Panel
A is the abnormal return on December 19, 2006 (the capital control day) and the dependent variable in Panel B is the abnormal return on December
20, (the liberalization day). Firm size is measured as the log of total asset; Pro?tability is measured as pre-tax pro?t or loss scaled by lagged total
asset; Exchange Rate Exposure is measured as the exchange rate beta estimated from the multi-factor model; Foreign ownership is a fraction of the
?rm that is owned by non-Thai citizens; Foreign Director is a dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen
as a director and zero otherwise. Also estimated but not reported are a constant term and 9-industry dummy variables. Numbers in the brackets
are heteroscedasticity-robust t-statistics. *, **, and *** indicate statistical signi?cant at 10, 5, and 1 percent levels, respectively. R-squared and the
number of observations are reported in the last two rows.
Table 1.3C: Firm Size and Activities in International Capital Markets Number of Firms Are SET 100 Index Constituent 0 1 20 69 90 42 12 82 33 9 56 6 0 16 6 40 46 3 2 4 0 0 1 6 0 Ratings uity Abroad Abroad Have TRIS Credit Have ADR Have Issued EqHave Issued Bond
Size Quartile
Are SET 50 Index
Constituent
Smallest
0
Second
0
Third
2
Largest
44
All
46
Stock Exchange of Thailand selects the constituents of SET 50 and SET 100 indices by choosing the top 50 and top 100 listed ?rms in terms of
market capitalization, liquidity and compliance with requirements regarding the distribution of shares to minor shareholders. TRIS (Thailand Rating
Information Services Co.) is Thailand’s ?rst and most prominent rating agency. The list of Thai ?rms with ADRs (American Depository Receipts)
31
is from JP Morgan’s and Bank of New York’s Databases. Data on the issuances of equity and debt securities by Thai ?rms outside of Thailand (from
1990 to 2006) are from Thomson’s SDC Platinum Database.
Table 1.4A: Correlation Matrix: Size and Firms Political/Family Connections Firm Size Firm Size Thaksin Connection Family Business Group Political Connection 1 0.1327 0.3290 0.3019 1 0.4231 0.1248 1 0.1004 1 Thaksin Connection Family Business Group Political Connection
Thaksin Connection is a dummy variable that takes the value of one if the ?rm’s major shareholder is, or is related to (has the same last name as), a member of Thaksin’s Cabinet and zero otherwise. Family Business Group is a dummy variable that takes the value of one if the ?rm’s major shareholder is from the 50 largest family business groups in Thailand and zero otherwise; Political Connection is a dummy variable that takes the value of one if the ?rm is politically connected according to Faccio (2006): A ?rm is politically connected if (1) its major shareholder or top executive are parliament member, minister or head of the state or (2) its major shareholder or top executive are closely related to a top politician or a political party.
32
Table 1.4B: Political/Family Connections and the E?ects of Foreign Portfolio Investment Panel A Dependent Variable Firm Size Model 1 -0.5309** [-2.26] Pro?tability 0.3165*** [3.3] Exchange Rate Exposure 2.244*** [3.22] Foreign Director 1.1498 [1.49] Thaksin Connection -2.2718 [-1.24] Family Business Group -1.2385 [-0.95] Political Connection 0.5489 [0.31] Constant Industry Dummy R-squared Observations Yes Yes 0.1702 306 Yes Yes 0.1677 306 Yes Yes 0.1657 306 Yes Yes 0.1511 306 Yes Yes 0.1434 306 Model 2 -0.4991** [-2.04] 0.3069*** [3.11] 2.2457*** [3.24] 1.1881 [1.54] Model 3 -0.6117** [-2.41] 0.3185*** [3.27] 2.287*** [3.34] 1.1939 [1.55] Model 1 0.54** [2.41] -0.2213** [-2.23] -1.8366*** [-2.76] -1.8121** [-2.46] 3.1052* [1.91] 1.2429 [1.16] -1.7614 [-1.13] Yes Yes 0.1448 306 Panel B Model 2 0.5289** [2.23] -0.2138** [-2.18] -1.8530*** [-2.85] -1.8656** [-2.53] Model 3 0.6941*** [2.95] -0.2167** [-2.17] -1.9001*** [-2.93] -1.878** [-2.55]
The table reports the coe?cient estimates from regressions of abnormal returns on ?rms political/ family connections. The dependent variable in Panel A is the abnormal return on December 19, 2006 (the capital control day) and the dependent variable in Panel B is the abnormal return on December 20, (the liberalization day). Firm size is measured as the log of total asset; Pro?tability is measured as the pre-tax pro?t or loss scaled by lagged total asset; Exchange Rate Exposure is measured as the exchange rate beta estimated from the multi-factor model; Foreign Director is a dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen as a director and zero otherwise; Thaksin Connection is a dummy variable that takes the value of one if the ?rms major shareholder is, or is related to (has the same last name as), a member of Thaksins Cabinet and zero otherwise. Family Business Group is a dummy variable that takes the value of one if the ?rms major shareholder is from the 50 largest family business groups in Thailand and zero otherwise; Political Connection is a dummy variable that takes the value of one if the ?rm is politically connected according to Faccio (2006) and zero otherwise. Also estimated but not reported are a constant term and 9-industry dummy variables. Numbers in the brackets are heteroscedasticity-robust t-statistics. *,**, and *** indicate statistical signi?cant at 10, 5, and 1 percent levels, respectively. R-squared and the number of observations are reported in the last two rows.
33
Table 1.5: Placebo Test: Are large ?rms more susceptible to bad news? Dependent Variable Firm Size Model 1 0.4772*** [2.95] Pro?tability Model 2 0.4438*** [2.67] Model 3 0.4481*** [2.74] 0.3057*** [4.14] Exchange Rate Exposure Model 4 0.5496*** [3.25] 0.2881*** [4.19] 1.0988* [1.79] Foreign Director Model 5 0.5311*** [3.2] 0.2864*** [4.22] 1.1337* [1.83] 0.484 [1.14] Thaksin Connection Model 6 0.6125*** [3.64] 0.278*** [4.12] 1.0762* [1.8] 0.4255 [1.01] -3.232* [-1.87] Constant Industry Dummy R-squared Observations Yes No 0.0257 312 Yes Yes 0.0746 309 Yes Yes 0.1231 306 Yes Yes 0.1424 306 Yes Yes 0.1447 306 Yes Yes 0.1661 306
These placebo regressions examine the e?ects of ?rm size and other characteristics on abnormal returns when there is bad news. The dependent variable in is the abnormal return on September 22, 2006 (the ?rst day the market learned about the coup). Firm size is measured as the log of total asset; Pro?tability is measured as the pre-tax pro?t or loss scaled by lagged total asset; Exchange Rate Exposure is measured as the exchange rate beta estimated from the multi-factor model; Foreign Director is a dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen as a director and zero otherwise; Thaksin Connection is a dummy variable that takes the value of one if the ?rms major shareholder is, or is related to, a member of Thaksins Cabinet and zero otherwise. Also estimated but not reported are a constant term and 9-industry dummy variables. Numbers in the brackets are heteroscedasticity-robust t-statistics. *, **, and *** indicate statistical signi?cant at 10, 5, and 1 percent levels, respectively. R-squared and the number of observations are reported in the last two rows.
34
Figure 1.1: Stock Market Response to the Imposition and the Reversal of Capital Control
The Stock Exchange of Thailand (SET) Index is a composite calculated from the prices of all common stocks on the main board of the Stock Exchange of Thailand. The SET Index dropped from 730.55 to 622.14 on the capital control day (December 19, 2006) and bounced back from 622.14 to 691.55 on the liberalization day (December 20, 2006).
35
Appendix 1.1: Summary of Related Empirical Work Country Time Period Methodology Finding about Firm Size
Paper
Cross-Country Study 53 countries 1996-2005 Panel Data Regression Foreign portfolio investment is associated with an increase in the ability to issue securities for small ?rms; Foreign portfolio investment increases the maturity of bank loans. 13 countries 1988-1998 Euler Equation Method Financial liberalization relaxes ?nancial constraints for small ?rms.
Knill (2005)
Laeven (2003)
Single-Country Study Indonesia large ?rms. Ecuador 1983-1988 Euler Equation Method The 1986 liberalization did not relax ?nancial constraints, even for small ?rms. Mexico Korea 1980-2000 Tobin Q Mothod 1984-1994 Euler Equation Method The 1989 liberalization relaxed ?nancial constraints for small ?rms. The 1995 liberalization relaxed ?nancial constraints for small ?rms; small ?rms get much wider access to outside credit after the liberalization. Colombia 1989-1993 Structural Estimation The 1991 liberalization increased investment rate and lowered buying price of capital for small ?rms. 1981-1988 Euler Equation Method The 1983 liberalization relaxed ?nancial constraints for both small and
Harris et al. (1994)
36
18 countries rolling window Event Study
Jaramillo et al. (1996)
Gelos and Werner (2002)
Koo and Shin (2004)
Contreras
and
Makaew
(2007)
Event Study Small ?rms earn signi?cantly higher abnormal returns when stock markets are liberalized.
Patro and Wald (2005)
Appendix 1.2: De?nitions and Data Sources Description and Source The last reported trade price each day, Source: Reuters Percentage changes in daily prices, Source: Authors Calculation Everything that the company or group owns at the end of the ?nancial year, expressed in local currency. Is equal to total current assets plus total ?xed assets, Source: Reuters and SETSMART Log of total assets, Source: Authors Calculation Pre-tax pro?t or loss, expressed in local currency, Source: Reuters and SETSMART Dummy variables representing 9 industries: Agro and Food, Consumer Products, Financials, Industrials, Property and Construction, Resources, Services, Technology, and Others, Source: SETSMART Pro?t scaled by lagged total assets, Source: Authors Calculation Number of shares outstanding times market price, Source: Reuters Number of shares outstanding times book value per share, Source: Reuters Market equity scaled by book equity, Source: Authors Calculation The net cash in?ow from operating activity, expressed in local currency. Is equal to the sum of various operating activities., Source: Reuters and SETSMART Lagged annual growth in operating cash ?ows, Source: Authors Calculation
Variable Name
Daily Stock Price
Daily Stock Return
Total Assets
Firm Size
Pro?t
Industry Dummy
37
Pro?tability
Market Equity
Book Equity
Market to Book
Operating Cash Flows
Cash Flow Growth
Variable Name
Description and Source The cash and other assets that the company or group expects to turn into cash. Is usually equal to cash and equivalents plus receivables plus inventories plus any of the following items: short term investments; marketable securities; prepayments, Source: Reuters and SETSMART All of the liabilities that the company or group expects to have to meet within 12 months, Source: Reuters and SETSMART Current asset minus Current liability, Source: Authors Calculation Working capital scaled by lagged total assets, Source: Authors Calculation Current Liabilities plus Long-term liabilities, Source: Reuters and SETSMART Total debts scaled by lagged total assets, Source: Authors Calculation Local Currency (Thai Baht) / US Dollar Exchange Rate, Source: WRDS The beta coe?cient estimated from a multi-factor model regressing stock returns on percentage changes in the exchange rate. The factor model is estimated by the daily returns from September 29, 2005 to August 31, 2006 (a 261-trading-day period), Source: Authors Calculation A fraction of the ?rm that is owned by non-Thai citizens, Source: SETSMART A dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen as a director and zero otherwise, Source: SETSMART A dummy variable that takes the value of one if the ?rms major shareholder is, or is related to, a member of Thaksins Cabinet and zero otherwise, Source:Ownership information is from SETSMART. The list of Thaksins Cabinet members is from the Thai government website: www.cabinet.thaigov.go.th/ A dummy variable that takes the value of one if the ?rms major shareholder is from the 50 largest family business groups in Thailand and zero otherwise, Source: Ownership information is from SETSMART. The list of family business groups is from a Brooker Group Publication: Thai Business Group: A Unique Guide to Who Owns What
Current Assets
Current Liabilities
Working Capital
Liquidity
Total Debts
Leverage
Exchange Rate
Exchange Rate Exposure
38
Foreign Ownership
Foreign Director
Thaksin Connection
Family Business Group
Variable Name
Description and Source A dummy variable that takes the value of one if the ?rm is politically connected according to Faccio (2006) and zero otherwise. Faccio (2006) de?nes politically connected ?rms broadly as ?rms that (1) have a major shareholder or a top executive who is a parliament member, minister or head of the state or (2) have a major shareholder or a top executive who is related to a top politician or a political party, Source: Faccio (2006)’s dataset is available on American Economic Review website: www.aeaweb.org/ Number of shares outstanding times market price, Source: Reuters The cash and cash equivalents value as shown in the company’s balance sheet. Cash and equivalents can include cash at bank and in hand, short-term deposits or other liquid assets, Source: Reuters and SETSMART The amount derived from the provision of goods and services falling within the company or group’s ordinary activities, otherwise known as turnover. The data is entered as reported in the pro?t and loss account of the annual report, Source: Reuters and SETSMART Stock Exchange of Thailand selects the constituents of SET 50 and SET 100 indices by choosing the top 50 and top 100 listed ?rms in terms of market capitalization, liquidity and compliance with requirements regarding the distribution of shares to minor shareholders, Source: Stock Exchange of Thailand website: www.set.or.th/ TRIS (Thailand Rating Information Services Co.), Thailands ?rst and most prominent rating agency, provides credit ratings of Thai ?rms, Source: TRIS website: www.trisrating.com/ American Depository Receipts of Thai ?rms, Source: JP Morgan website: www.adr.com/ and the Bank of New York website: www.adrbny.com/ Issuances of equity securities by Thai ?rms outside of Thailand from 1990 to 2006, Source: SDC Platinum Issuances of debt securities by Thai ?rms outside of Thailand from 1990 to 2006, Source: SDC Platinum
Political Connection
Market Capitalization
Cash Holdings
Sales
39
SET 50/SET 100 Index
TRIS Ratings
ADR
Equity Issuance Abroad
Debt Issuance Abroad
Chapter 2
Waves of International Mergers and Acquisitions
2.1 Introduction
In the past two decades, 26% of worldwide M&A activities involve acquirers and targets from di?erent countries. The aggregate volume of cross-border mergers from 1989 to 2008 totals more than 8 trillion dollars. In spite of such a large volume, most of the M&A literature focuses on domestic mergers. Moreover, the amount of cross-border mergers varies greatly from year to year. For example, the volume of worldwide M&A deals dropped by 62% from 2000 to 2003 but bounced back by 158% in 2006.1 Despite such a large year-to-year ?uctuation, most papers on cross-border M&As study the e?ects of long-run determinants like corporate governance and capital market development. These gaps in the literature motivate the research questions that are at the core of this paper: what are the dynamic patterns of cross-border mergers, and what are the factors that drive them? Using the data from 50 countries over the period of 1989-2008, I document the following facts about international M&As:
1
See table 2.1 and ?gure 2.1 for details
40
(1) International mergers come in waves that are highly correlated with business cycles. Merger booms coincide with booms in the real sector and in the ?nancial market. While the literature on merger waves shows that domestic mergers are pro-cyclical, I ?nd that cross-border mergers are even more pro-cyclical than domestic mergers. (2) Mergers are more likely to occur when both the acquirer and the target economies are booming. This is true even when I eliminate the e?ects of global booms. My ?nding refutes the widespread belief that most cross-border mergers occur when the target economies are in a recession or face a ?nancial crisis, and that acquirers are vulture investors taking advantage of liquidity-constrained targets (Krugman, 1998; Aguiar and Gopinath, 2005; Desai, Foley, and Forbes, 2007; Acharya, Shin, and Yorulmazer, 2009). Although such “?re sale” mergers can happen under speci?c circumstances, most mergers do not follow this pattern. (3) Merger booms have both an industry-level and a country-level component. Given that productivity shocks are better measured at the industry level and that ?nancial shocks such as a change in monetary policy are more of a country-wide phenomenon, this ?nding is consistent with the notion that M&As are driven by productivity shocks and facilitated by macro liquidity shocks (Harford, 2005; Eisfeldt and Rampini, 2006). (4) Across over one million ?rm-year observations, acquirers tend to be more productive than their industry peers and targets tend to be less productive than their industry peers. This ?nding supports the neoclassical theory of mergers in that high productivity ?rms acquire low productivity ?rms in order to redeploy their assets toward more pro?table uses (Maksimovic and Phillips, 2001). This ?nding is at odds with the “like-buys-like” theory in which high productivity acquirers seek high productivity targets to realize gains from asset complementarity (Rhodes-Kropf and Robinson, 2008). This paper makes contributions to several strands of merger literature. Most papers on cross-border M&As put their emphasis on long-run determinants of mergers. Starting in the 1980s, Errunza and Senbet (1984) develop a theory of why corporations diversify internationally based on capital and goods market imperfections. Rossi and Volpin (2004) ?nd that acquirers are more likely to be from countries with stronger investor protection than targets. Bris and Cabolis (2008) ?nd that the merger premium is higher when the acquirer 41
country’s investor protection is stronger than the target country’s. Ferreira, Massa, and Matos (2009) ?nd that cross-border mergers increase with foreign institutional ownership. Di Giovanni (2005) ?nds that mergers are more likely to originate from countries with developed ?nancial markets. Although corporate governance and capital market development are undeniably important for mergers, it is hard to imagine that the large year-to-year ?uctuations (i.e., waves) in M&A activities are driven by these long-run determinants. Variables such as investor protections are extremely persistent and can be traced back to colonial origins. The central contribution of this paper is to incorporate the dynamic dimensions from business cycle theories and help explain the cyclical ?uctuations of international mergers. My paper is also related to the neoclassical theory of mergers. This literature typically argues that merger waves are driven by productivity shocks (Maksimovic and Phillips, 2001; Ditmar and Ditmar, 2008; Yang, 2008). Some authors further argue that, for waves to be formed, productivity shocks must be accompanied by liquidity shocks (Eisfeldt and Rampini, 2003; Harford, 2005; Maksimovic, Phillips, and Yang, 2009). Although my international paper draws insights from relatively well-established domestic merger literature, I also address important issues unique to cross-border mergers, such as the role of exchange rates and government policies on multinational corporations. In relation to empirical work on domestic mergers, cross-border M&As are an excellent setting for out-of-sample tests and provide more comprehensive data. For instance, Dittmar and Dittmar (2008) suggest that U.S. domestic merger activities ?uctuate in response to GDP shocks. However, it is di?cult to identify the mechanisms by which GDP e?ects mergers using data from a single country. All acquirers, targets, and non-merging ?rms face the same macroeconomic shocks, and most macroeconomic shocks are highly correlated. In my international context, acquirers and targets are from di?erent countries so I can better identify where the shocks originate. Moreover, using the data from 50 di?erent countries, or 50x50 = 2,500 country pairs, will substantially increase the degrees of freedom of the analysis.
42
2.2
Empirical Framework
The goal of this paper is to document the important facts about cross-border mergers. More precisely, I ask the following four questions. (1) How do cross-border mergers behave over a business cycle? I ?rst present exploratory evidence on the cyclicality of international mergers. Without making any structural assumptions, I compute the correlations between merger activities and a number of macroeconomic indicators measuring the real sector, the ?nancial sector, and the external sector. The lead-lag correlations are computed within a seven-year window around mergers to o?er a clear picture of what happens before, during, and after merger waves. (2) Where do shocks that e?ect cross-border mergers come from? The results from the correlation analysis can be driven by the co-movements of the acquirer economy and the target economy or by global economic booms. Following Harford (2005) and Dittmar and Dittmar (2008), I assume that lagged macroeconomic indicators are proxies for exogenous shocks and that M&As ?uctuate in response to these shocks. Then, I regress mergers on acquirer country shocks and target country shocks, controlling for the year ?xed-e?ects to eliminate the e?ects of global booms. This exercise will help me identify where the cyclical nature of cross-border mergers originates. (3) What type of shocks (real or ?nancial) e?ect cross-border mergers? Mergers come in waves either because productivity shocks occur in waves or because ?nancial shocks occur in waves. Using industry-level data, I form indices of industry-level productivity and valuation shocks. Then, I regress mergers on the industry indices, controlling for country-level shocks and year ?xed-e?ects. Given that the macro ?nancial shocks (such as changes in lending rates, monetary policy, and security issuance cost) are likely to be highly correlated across di?erent industries within the same country, these regressions will allow me to distinguish the e?ects of productivity shocks from the e?ects of ?nancial shocks. (4) What types of ?rms engage in cross-border mergers? To complete the analysis, I examine the characteristics of the merging ?rms. Using data from WorldScope, I compute a number of productivity and valuation measures. Then, I
43
compare the characteristics of the merging ?rms with non-merging ?rms and compare the characteristics of acquirers with targets. Studying what types of ?rms engage in merger activities will shed some light on the motives behind M&A decisions.
2.3
Data
Mergers and Acquisitions Data The source of M&A data is Thomson’s Securities Data Corporation (SDC) database. My sample covers all deals announced and completed between 1988 and 2008. To ensure that my results represent a wide range of countries but are not driven by countries that rarely have mergers, I require that acquirers and targets must be from 25 developed countries with the most M&A deals and 25 developing countries with the most M&A deals. These 50 countries are listed in Table 2.2. There are 412,810 deals in my sample. The aggregate value of these deals is approximately 40 trillion dollars. Eight trillion dollars are from cross-border deals. [INSERT TABLE 2.1 AND FIGURE 2.1 HERE] Table 2.1 reports the aggregate volume and aggregate frequency of M&A activities. In Figure 2.1, aggregate volume exhibits both growing trends and large cyclical ?uctuations. The volume of all M&A deals grows from around 500 million dollars in the early 1990s to more than 3 trillion in 2006-2007. The cyclical component is very large, especially in recent years. For example, the volume of all M&A deals dropped by 62% from 2000 to 2003 but bounced back by 158% three years later. Cross-border M&A deals are more volatile than domestic deals. The standard deviation scaled by mean is 70% for all deals but 84% for cross-border deals. [INSERT TABLE 2.2 HERE] Table 2.2 shows the breakdown by country. Most M&As are between high-income countries. The countries that have a large number of acquirers also have a large number of targets. The developed countries that have the highest number of acquirers and targets 44
are the G-7 countries (except Italy): the U.S., UK, Germany, Canada, and Japan. The top developing countries are the BRIC countries (Brazil-Russia-India-China) plus Malaysia. SDC provides detailed information on deal characteristics. Aside from basic information such as country, industry, and year, the SDC data include deal size, percent acquired, method of payment, and acquirer/target public status. The SDC collects data from a number of sources including the SEC and international stock exchange ?lings, news wires, trade publications, as well as surveys of banks and advisory ?rms. It might be a concern that some deals do not have the size attached to them because ?rms are not required to report the transaction values to SDC. Di Giovanni (2005) ?nds no pattern in which industries, countries, or years have more missing values than others. As a precautionary measure, I also compare data from the SDC to the country-level FDI data from UNCTAD and the transaction-level data from Capital IQ. At the aggregate level, I observe similar cyclical patterns from these three sources. Macroeconomic Data Most of the macroeconomic data are from the World Bank’s World Development Indicator (WDI) database. I use the variables that capture the states of an economy in terms of the real sector, the ?nancial sector, and the external economy sector. For the real sector, I use the data on GDP, gross value added, gross capital formation, and total population. For the ?nancial sector, I use the data on domestic credit and stock market capitalization. For the external sector, I use current account balance and nominal exchange rate. I augment the WDI data with the foreign portfolio investment data (net foreign portfolio investment) from the IMF’s Coordinated Portfolio Investment Survey (CPIS) and stock price data (Average M/B and Average P/E ratios) from Kenneth French’s Website. The average M/B and average P/E are equal-weighted. The data from WDI cover all 50 countries over the 20-year sample period, but the CPIS and French’s data have less coverage. The CPIS data are available from 2001 and French’s data cover the entire period but are only available for 21 countries. The country-pair variables, geographical distance, and common language dummy are from Di Giovanni (2005). Micro Data I use the WorldScope database, which covers over 95% of world market capitalization. It 45
provides the ?nancial statement information and market price of ?rms around the world. From the 50 countries in Table 2.2, WorldScope provides full coverage of the listed ?rms in 31 countries, 10 of which are developing countries. WorldScope also provides targeted coverage (all listed ?rms with a market capitalization higher than 100 million dollars) for 16 countries. The missing countries are Slovakia, Lithuania, and Ukraine. The list of these countries is available in the appendix. I construct a 1988-2008 annual data set of all the public ?rms available. There are 1,104,516 observations in this data set.2 [INSERT TABLE 2.3 HERE] To ensure that my results are not speci?c to a particular variable de?nition, I compute six di?erent measures of productivity: return on assets, pro?t margin, labor productivity, sales growth, employment growth, and payout ratio. Return on assets is pro?t (EBITDA) divided by total assets; pro?t margin is pro?t divided by sales; labor productivity is pro?t per worker; sales growth is the percentage change in annual sales; employment growth is the percentage change in number of workers; and payout ratio is dividends divided by total assets. I compute three valuation measures: M/B, past one-year return, and past three-year returns. The M/B is market capitalization divided by total assets less total debts; past return is the percentage change in market prices. I also collect four other variables that might e?ect M&As: size as measured by log of total assets (book value), age calculated from incorporation date, age calculated from listing date, and leverage. Leverage is total debts divided by total assets. To ensure that my results are not driven by outliers or any mistakes in the original data set, I winsorize the data at 0.025. I report the descriptive statistics of WorldScope variables in Table 2.3.
2
I believe that WorldScope is an appropriate database for this research. Compared to practitioner-
oriented products like Reuters, WorldScope retains inactive ?rms but Reuters does not. M&As can result in the de-listing of target ?rms; therefore, information for many targets are not available in Reuters. Compared to other popular research-oriented products, such as S&P’s Research Insight or Compustat Global, which also cover inactive ?rms, WorldScope provides better coverage. For example, there are only 27,805 ?rms represented in Research Insight [but there are 52,596 ?rms represented in WorldScope]. Finally, many premium databases, such as Dun and Bradstreet’s WorldBase or CapitalIQ, cover larger cross-sections of ?rms. They provide limited or no historical data, which I deem necessary for a time-series study.
46
Filtering Procedure There are challenges associated with using raw data directly. First, most of the variables of interests such as merger activities, stock market capitalizations, and GDPs, are increasing over time. If I use the raw data to compute correlations or run regressions, then the results are likely to be spurious. Second, the focus of this paper is the cyclical properties of M&As rather than their crosscountry variations. If I run a panel-data regression using raw data as seen in Di Giovanni (2005), then the estimated coe?cients will combine the time-series and cross-sectional e?ects together. For example, Di Giovanni (2005) ?nds that that larger stock market capitalization in year t-1 leads to more acquisitions in year t. This ?nding could be driven by cross-country di?erences. For example, countries like the U.S. have larger stock markets than Sub-Saharan African countries do; therefore, his results might re?ect the fact that there are more American acquirers than Nigerian acquirers. My research question is di?erent: I ask whether there are more U.S. acquirers and targets when there is an economic boom in the U.S. Mendoza and Terrones (2008) develop an algorithm to identify and analyze credit booms. I use a minor variation of their algorithm to transform my data: (1) I de?ate all the nominal variables with GDP de?ators and scale them by total population. Because I try to measure various shocks to the economy, scaling by total population is more appropriate for my application than scaling by GDP. GDP itself is also a?ected by the shocks; thus, scaling by GDP will confound the e?ect of the shocks in the original variables. (2) I ?lter out trends in all variables by using the Hodrick-Prescott (HP) ?lter. Hodrick and Prescott (1997) propose a de-trending method, which is now commonly used in the business cycle literature. The HP ?lter decomposes the raw variable Xt into the trend component, trendt , and the cyclical component, shockt . Given the smoothing parameter ? , the ?lter will choose the trend component that minimizes the objective function:
T t=1 (Xt
? trendt )2 + ?
T ?1 t=2 ((trendt+1
? trendt ) ? (trendt ? trendt?1 ))2
47
The ?rst term in the objective function penalizes the deviations from the trend, while the second term penalizes the ?uctuation in the growth rate of the trend components, i.e., the non-smoothness of the trend. Following Mendoza and Terrones (2008), I apply the HP ?lter to the full sample period 1988-2008 and set the smoothing parameter equal to 100, which is commonly used with annual data.3 (3) I compute the standard deviations of shocks in each country and then scale the shocks with their standard deviations. Mendoza and Terrones (2008) de?ne “boom” as a situation in which the deviation from trend is unusually large relative to the country’s typical cycle. Scaling by the country’s standard deviation is necessary because some economies are more volatile than others. Moreover, such scaling will eliminate the cross-country di?erences in size and allow me to run a panel data regression in which all countries are treated equally. [INSERT FIGURE 2.2 HERE] Figure 2.2 shows an example of the raw and the de-trended series of U.S. ?rm acquisitions of assets in other countries. Consistent with Mendoza and Terrones (2008), I identify a “wave” as a situation in which the deviation from trend is unusually large. Instead of providing a speci?c cuto? and discretizing the wave variable, I use the HP-de-trended and standard deviation-normalized variables directly to preserve their information content. From the ?ltered variables, I observe many “merger waves” in the data. Out of 1,000 country-year observations, there are 165 observations in which total M&As activities exceed one standard deviation and 109 observations in which M&As exceed two standard deviations.
2.4
Merger Activities and Macroeconomic Conditions
Correlation Analysis Business cycle theories predict that there are systematic relations among macroeconomic
3
I examine the stationarity of the detrended data using the Levin-Lin-Chu test (a pooled Dickey-Fuller
test). The test con?rms that the panel is indeed stable.
48
variables. These variables might e?ect one another endogenously or be driven jointly by some unobservable factors. As a ?rst step, I do not make any causality or structural assumptions. To see which variables coincide or have lead-lag relations with M&As activities, I compute sample correlations at di?erent time periods in a seven-year window around mergers. This exercise o?ers a clearer picture of what happens before, during, and after merger waves. To examine how M&As ?uctuate over a business cycle, I compute the following correlations: Correlation(M ergerc,t , Xc,t+j ) j ? {?3, ?2, ?1, 0, 1, 2, 3}, where M ergerc is the aggregate volume of mergers in country c. The Xc s are the macroeconomic indicators capturing: (1) the real sector of the economy (gross value added, GDP, and capital formation); (2) the ?nancial sector of the economy (market capitalization, M/B, P/E, and domestic credit); as well as (3) the external sector of the economy (current account, exchange rate, and foreign portfolio investment). Both M ergerc and Xc are de-trended using the Mendoza and Terrones’ ?ltering procedure described earlier. [INSERT TABLE 2.4 HERE] The results are reported in Table 2.4. Because cross-border deals are associated with two countries, there are three sets of correlations: (1) between cross-border deals and the acquirer country’s characteristics, reported in Table 2.4A; (2) between cross-border deals and the target country’s characteristics, reported in Table 2.4B; and (3) between domestic deals and the corresponding country characteristics, reported in Table 2.4C. It is apparent from these tables that M&As are pro-cyclical and that merger waves coincide with macroeconomic booms. The correlations between mergers and real/?nancial indicators show a similar pattern across the board. The correlations between mergers at time t and the indicators at time t-3 are negative; then these correlations increase, become positive at time t-2 or t-1, and peak at time t; they then remain positive for a period of time and then turn negative at t+3. According to this cyclical pattern, the lagged values of the indicators will be able to predict mergers. 49
The ?uctuations of the real indicators should be highly correlated with aggregate productivity shocks facing an economy.4 The contemporaneous correlations between M&As and gross value added are positive and statistically signi?cant in all speci?cations. A one standard deviation shock in the acquirer’s value added is associated with a 0.26 standard deviation change in cross-border mergers; a one standard deviation shock in target’s value added is associated with a 0.18 standard deviation change in cross-border mergers; a one standard deviation shock in domestic value added is associated with a 0.07 standard deviation change in domestic mergers. I ?nd that the correlations between M&As and the ?nancial indicators are higher than the correlations between M&As and the real indicators. For example, stock market capitalization in the acquirer country has a 42% correlation with cross border M&As. A one-year lagged stock market capitalization has a 31% correlation with cross-border M&As, but gross value added only has correlations of 26% and 14%, respectively. This is not surprising given that the ?nancial indicators are forward-looking, but the real indicators are accounting numbers measuring past performance. As Harford (2005) points out, the ?uctuations in valuation measures can come from any source, including productivity shocks, liquidity shocks, and misvaluations. Comparing Tables 2.4A and 2.4B, real and ?nancial conditions of the acquirer country have a higher impact on M&As than the conditions in the target country. This ?nding can be explained by a variety of reasons. One example could be that acquirers have to raise funds for the acquisitions, and that the cost of ?nancing is lower when the acquirer country is booming. Another might be that the acquirers must take control of the targets, so it is more important that the acquirers receive high productivity shocks. Comparing Tables 2.4A and 2.4B with Table 2.4C, I ?nd that cross-border mergers are much more correlated with real and ?nancial conditions than domestic mergers. Harford (2005), Eisfeldt and Rampini (2006), Dittmar and Dittmar (2008), and Yang (2008) document that domestic M&As are pro-cyclical. The comparison between Table 2.4A-2.4B and Table 2.4C shows that cross-border mergers are even more pro-cyclical than domestic mergers.
4
Wurgler (2000) and Maksimovic and Phillips (2001) use value added as a proxy for productivity shocks.
50
Turning to the measures of the external sector, the dynamic pattern of correlations between cross-border mergers and the exchange rate is interesting. Merger waves do not occur when the domestic currency is strongest. The contemporaneous correlation between M&As and exchange rates is statistically zero. The domestic currencies are strong two to three years before the peak of the merger waves and become weak two to three years after the merger’s peak. This ?nding suggests that M&As do not react directly to exchange rate appreciations. Instead, both mergers and exchange rate movements are more likely to be part of a larger business cycle model in which symptoms such as the appreciation of local currencies and the run-up in real estate prices are typical during economic expansions. The correlations of other external indicators are less signi?cant. Cross-border mergers do not have a signi?cant relation with the acquirer’s current account. There is weak evidence that target countries run current account de?cits during the merger waves. This is expected because M&As are a part of the capital in?ow that might worsen the current account balance (current account de?cit = capital in?ow + change in foreign reserve). I detect a small correlation between mergers and foreign portfolio investments. This is probably due to the fact that the foreign portfolio investment data from CPIS have much less coverage than the domestic variables, which come from WDI. In sum, the correlation analysis reveals that M&As exhibit a strong cyclical pattern. Real and ?nancial indicators coincide with and predict mergers. Fact 1: Cross-border mergers are highly correlated with business cycles. Regression Analysis Because the correlation analysis is univariate in nature, it is possible that the results in Table 2.4 are driven by global economic booms or by the co-movements of the acquirer economy and the target economy. To answer this question, I move to a multivariate framework. Following Harford (2005) and Dittmar and Dittmar (2008), I assume that lagged macroeconomic indicators are proxies for exogenous shocks and that M&As ?uctuate in response to these shocks. Speci?cally, I regress mergers between country c1 and c2 on the lagged conditions of c1 and the lagged conditions of c2 , controlling for the year ?xed-e?ects. By
51
putting the acquirer conditions and the target conditions side by side, I can identify how much of the M&As are driven by the acquirer conditions and how much are driven by target conditions. Additionally, by including the year ?xed-e?ect, I can determine whether the variations beyond the global averages still have an e?ect on M&As. My speci?cation is: M ergerc1 ,c2 ,t = ?0 + ?1 Exc1 ,c2 ,t?1 + ?2 Xc1 ,t?1 + ?3 Xc2 ,t?1 +
c1 ,c2 ,t
where M ergerc1 ,c2 is the volume of deals with the acquirer in c1 and the target in c2 . Exc1 ,c2 is the exchange rate (acquirer currency per one unit of target currency). The Xc s are the real and ?nancial indicators I used earlier. All the variables are again de-trended by the Mendoza and Terrones’ ?ltering procedure. Putting all seven real and ?nancial indicators in a regression at the same time will result in a multi-collinearity problem. I solve this problem in two ways: (1) by picking a representative variable and (2) by using all the indicators to form economic shock indices. First, I pick the representative variable based on data availability. If data are equally available, then I try all of the indicators in a regression and select the horse-race winner. Second, I adopt Harford’s (2005) approach by forming indices using the ?rst principal component of all the indicators. (The real economy indicator is constructed from the gross value added, GDP, and capital formation; the ?nancial market indicator is constructed from market capitalization, M/B, P/E, and domestic credit.) [INSERT TABLE 2.5 HERE] The results are reported in Table 2.5. In column 1, the coe?cient of exchange rate is signi?cant and negative, which suggests that the acquirer currency is strong relative to the target currency prior to merger booms. In column 2, I use gross value added as a real economy indicator. The coe?cient of the acquirer country is estimated at 3%, and the coe?cient of the target country is estimated at 2%. These numbers are positive and statistically signi?cant even after controlling for the year ?xed-e?ect. These estimates suggest that mergers react to shocks both from acquirer countries and from the target countries. Because the year ?xed-e?ect has removed the global averages from all the 52
variables year by year, my result does not depended upon the worldwide merger booms in a particular time period. In column 3, I use market capitalization as a ?nancial market indicator. The coe?cient of the acquirer country is estimated at 6% and the coe?cient of the target country is estimated at 4%. Again, both numbers are positive and statistically signi?cant. Mergers are more likely to take place when stock markets in both the acquirer and the target country are booming. Consistent with the correlation analysis, ?nancial indicators are a better predictor of mergers than the real indicators. In addition, the conditions of the acquirer country are more signi?cant than the conditions of the target country. The speci?cations in columns 4 and 5 are similar to those in columns 2 and 3, except that I use ?rst principal component indicators. The results in columns 4 and 5 are similar to the ones in columns 2 and 3. In sum, the regression analysis con?rms that M&As react to shocks both in the acquirer country and in the target country. In other words, there are more M&As when both the acquirer and the target economies are booming. My results are also robust to the inclusion of the year ?xed-e?ects. Fact 2: There are more M&As when both the acquirer and the target economies are booming. Mergers and Global Economic Conditions From Table 2.5A, the year ?xed e?ect explains approximately 2% of the variations in cross-border merger activities. This ?nding suggests that there must be global factors driving mergers across di?erent countries. As an example of such factors, I replace the year dummies with three candidate measures of global economic conditions. Following Albuquerque, Loayza and Serven (2005), I use (1) World Equity Market, measured by the return on Morgan Stanley World Capital Index, (2) World Interest Rate, measured by the average of American, Japanese, and German three-month treasury rates, and (3) Credit Spread which is Moody’s AAA bond rate minus Moody’s BAA bond rate. I collect raw monthly data from Bloomberg, convert them into annual series, and detrend the series using the Mendoza and Terrones’ procedure.
53
The results are reported in Table 2.5B. The sign and the magnitude of the countrylevel indicators are similar to the ones in Table 2.5A. Even though the coe?cients of the acquirer and the target country indicators are still statistically signi?cant after controlling for the worldwide economic conditions, the e?ects of the global variables are relatively large - in some cases, larger than the e?ects of the acquirer and the target countries. For instance, the coe?cient of acquirer’s market capitalization is 0.07 and the coe?cient of target country is 0.03. The coe?cient of World Equity Market is estimated at 0.07.
2.5
Firm Characteristics and Industry Merger Waves
In this section, I examine the characteristics of merging ?rms and show how these characteristics change along with the merger waves. Characteristics of the Merging Firms In the previous section, I examined merger waves at the country level. In this section, I look inside each country and identify which ?rms engage in M&As activities. Analyzing the characteristics of the merging ?rms will shed some light on the main motives behind M&As in my sample. From the WorldScope data, I construct 13 measures of ?rm characteristics: 6 productivity measures, 3 valuation measures, and 4 other measures that might e?ect mergers. The productivity measures consist of return on assets (ROA), pro?t margin, labor productivity, sales growth, employment growth, and payout ratio.5 Although these measures are positively correlated, each represents di?erent concepts of productivity and has its own strength. For example, labor productivity captures technological shocks; pro?t margin captures demand conditions; the level measures, such as return on assets, capture productivity more directly; growth measures, such as sales growth, are less a?ected by ?rm-speci?c reporting practices or earning management. I examine all six to ensure that my results are not speci?c to a particular measure. For the valuation measures, I compute the market-to-book ratio, past one-year return, and past three-year returns, which
5
The term “productivity” is used loosely here- some of these should be labeled “pro?tability” measures
instead.
54
are similar to the measures used in Harford (2005). The other potential determinants of mergers are size, age based on incorporation date, age based on the listing date, and leverage. Because my sample consists of ?rms from di?erent countries, industries, and time periods, it is di?cult to interpret any di?erences in the unadjusted ?rm characteristics. For example, it is unclear whether a 2% ROA of a food factory in Thailand means the same thing as a 2% ROA of a car company in the U.S. To address this issue, I normalize each characteristic, labeled i, by: (1) Grouping all observations by country-industry-year and, for each group, computing the means and the standard deviations of i I use Fama-French’s 16 industries instead of the four-digit SIC code provided by WorldScope. The four-digit SIC industry is rather small, leaving some industries in small countries empty or sparsely populated. The de?nitions of Fama-French’s 16 industries and the
mapping from SIC code can be found on Kenneth French’s website:http://mba.tuck.dartmouth.edu/pages/ (2) Using the means and standard deviations to compute i’s Z-score (Z-score = (i-mean)/ standard deviation) for each observation. In other words, using the distribution of i in each country-industry-year, I convert the raw value of i into its position in the distribution. Mechanically, these Z-scores will have the mean of zero and the standard deviation of one. This adjustment will eliminate all cross-country and cross-industry di?erences. To investigate which ?rms are more likely to engage in M&As, I ?rst match the WorldScope data with the SDC data. For mergers that take place at time t, I compare the characteristic at time t-1 to avoid the reverse causality problem. Of all 173,357 deals that involve public acquirers, 122,118 deals can be matched with WorldScope ?rms. Of all 52,524 deals that involve public targets, 46,607 deals are matched. I then compare the Z-scores between the population of acquirers, targets, cross-border acquirers, cross-border targets, and other non-merging ?rms. [INSERT TABLE 2.6 HERE] The results are reported in Table 2.6. The ?rst four columns compare acquirers with 55
targets. Column 1 reports the di?erence between the average Z-scores of all acquirers and the average Z-score of all targets. Column 2 reports the di?erence between the crossborder acquirers and the cross-border targets. The numbers in the parenthesis are the t-statistics from the t-tests. Across all the productivity and valuation measures, I ?nd that the acquirers’ Z-scores are consistently higher than the targets’. The magnitudes of the di?erence are around 0.2-0.3 standard deviations. The economic signi?cance of these numbers is large but varies from country to country and from industry to industry. For example, in 2008, one standard deviation of ROA of the food industry in Thailand is 13% and one standard deviation of ROA of the American auto industry is 35%. The t-test only compares the averages of acquirers and targets. In columns 3 and 4, I compare the whole distributions using the Kolmogorov-Smirnov’s D-statistics. Column 3 reports the distributional distances between all acquirers and all targets. Column 4 reports the distributional distances between the cross-border acquirers and the cross-border targets. I ?nd that the di?erences are statistically signi?cant. This signi?cance con?rms that not only do the acquirers and the targets have di?erent means but they also come from di?erent distributions. In the last four columns, I compare the Z-score of the merging ?rms with the non-merging ?rms. Column 5 compares all of the acquirers with the non-acquirers, and column 6 compares cross-border acquirers with the non-acquirers. I ?nd that acquirers are more productive and have higher valuations than their industry peers. Column 7 compares all of the targets with the non-targets, and column 8 compares cross-border targets with non-targets. I ?nd that targets are less productive and have lower valuations than their industry peers. These results are virtually uniform across all measures. This is consistent with the neoclassical theory of mergers in which more productive ?rms purchase less productive ?rms to realize e?ciency gain. Turning to other characteristics, ?rms that participate in mergers are likely to be older and larger than non-merging ?rms. Among the merging ?rms, ?rms participating in cross-border mergers are older and larger than ?rms participating in domestic mergers. The acquirers are, generally, larger and older than the targets. For leverage, I ?nd some evidence that the targets tend to have higher leverage than the acquirers.
56
Because acquirers are larger and are more likely to be public than targets, acquirers ?nd more matches in WorldScope than targets. There is concern that the comparisons between acquirers and targets in columns 1-4 are potentially biased. However, Maksimovic, Phillips, and Yang (2009) report that non-listed ?rms are smaller and less productive than listed ?rms. This ?nding will bias my results toward zero. The fact that I still ?nd signi?cant di?erences between the acquirers and the targets implies that the unbiased di?erences must be very large and that my results in columns 1-4 can be thought of as a lower bound of the true di?erences. Fact 3: Acquirers are more productive ?rms. Targets are less productive ?rms. Comparisons between Acquirers and Targets In Table 2.6A, for cross-border mergers, the di?erence between the acquirers’ and the targets’ Z-scores might not be aligned with the di?erence in the original variable, i. In other words, the acquirers and targets are benchmarked by their industry peers in their own countries. It is unclear if the cross-border acquirers are more or less productive than their targets since the Z-scores come from two di?erent distributions. As a robustness check, I combine all countries together and assign a new Z-score based on industry-year grouping instead of country-industry-year grouping. The results are reported in Table 2.6B. I ?nd that, on average, the acquirers are still more productive and have higher valuation than the targets, but the signi?cance is not as strong as those of the original grouping. This is consistent with the country-level cross-sectional results in Section 2.5 that acquirers seek less productive ?rms in high income/productive countries, and not the least productive ?rms anywhere in the world. Characteristics of Merging Firms during Booms and Busts Motives behind mergers during booms and busts can be di?erent. Therefore, comparing the characteristics of the merging ?rms at di?erent points along the business cycle might shed some light on the factors driving merger activities. From the country- and the industry- level regressions, there are more mergers during booms. The standard neoclassical explanation is that productivity shocks are di?erent during booms and busts. Alternatively, more mergers during booms can be a result of the increase in participation by less productive ?rms. Less productive ?rms might engage in
57
MAs during booms due to higher capital liquidity, more free cash ?ows, or more intense product market competitions. In Table 2.6C, I compare the merging ?rms during booms and busts by computing another set of Z-scores based on country-industry grouping. I de?ne two types of booms (busts): (1) real booms (busts) as the periods in which HP detrended Gross Value Added is above (below) one standard deviation (2) ?nancial booms (busts) as the periods in which HP detrended Stock Market Capitalization is above (below) one standard deviation. I ?nd that more ?rms participate in MAs during booms and that the average acquirers during booms are smaller, younger, and have less leverage. However, I ?nd no evidence that economic booms lower the productivity threshold for mergers. The acquirers during booms are more productive than the acquirers during busts and the targets during booms are more productive than the targets during busts. Post-Merger Operating Performance To examine post-merger operating performance, I compute the percentage change in return on assets (Pro?t/Total Assets) over four windows: [t-1, t], [t-1, t+1], [t-1, t+2], and [t-1, t+3] where t denotes the year of the acquisition. The results are reported in Table 2.6D. Numbers in the table are the average performance of the treatment group (merging ?rms) minus the performance of the control group. From the ?rst four columns, without controlling for the selection issues, I ?nd that the acquirers and the targets underperform after mergers. Performance of the acquirers is worse than performance of the targets. Performance of the ?rms involving cross-border deals is worse than performance of the ?rms involving domestic deals. In Section 2.4, I ?nd that merging ?rms and non-merging ?rms have di?erent characteristics. Therefore, the unmatched results might re?ect di?erent initial characteristics rather than merger outcomes. To solve the problem, I use the propensity scores matching method. My probit selection model uses basic ?rm characteristics as covariates: lagged pro?tability, sales, total assets, and ?rm age. Columns 5-8 reports propensity-score-matched e?ects (average treatment e?ects on the treated). I ?nd that the most underperformance in the ?rst four columns disappears after controlling for basic ?rm characteristics. A potential explanation is that ?rms receive large productivity shocks prior to engaging in MA and the shocks revert in subsequent years.
58
Industry Shocks Next, I examine how the industry-level M&As are a?ected by the year-to-year ?uctuations in productivity and valuation measures. From section 2.4, it is not obvious whether the country-level indicators represent productivity shocks or other macroeconomic shocks because these shocks are highly correlated at the country level. Performing industry-level regressions will help me identify what types of shocks are driving the results. Although productivity shocks are probably best described at the industry level, ?nancial shocks (such as changes in lending rates, monetary policy, and security issuance cost) are more of an economy-wide phenomenon. If the industry-level regressors are signi?cant after controlling for the country-level regressors, then the productivity shocks are likely to be the factor driving M&As. At the same time, if the country-level regressors are signi?cant after controlling for the industry-level shocks, then the true shocks e?ecting M&As must have an economy-wide component. Using the six productivity measures and three valuation measures from the previous section, I construct the industry shock indices. I average the ?rm characteristics for each country-industry-year and normalize each series using the Mendoza and Terrones’ procedure. Similar to section 2.4, I either choose one variable (ROA) to represent productivity shocks and choose another variable (market-to-book) to represent the valuation shocks or I construct a productivity index from the ?rst principal component of the six productivity measures and construct a valuation index from the three valuation measures.6
6
It might be a concern that indices constructed from WorldScope variables might not be an accurate
proxy of industry shock because private ?rms in SDC are not covered by WorldScope. To the extent that shocks to ?rms in the same industry in the same country are correlated, the ?uctuations in the average productivity of the listed ?rms can be used as a proxy of the ?uctuations in the average productivity of all ?rms. To further address this concern, I use the United Nations Industrial Development Organization (UNIDO) database, which is a census-type dataset, to construct an alternative measure for industry shocks - value added per worker and output per worker. Again, I ?nd that the industry shocks have positive coe?cients. However, the new coe?cients are less signi?cant than those of the WorldScope shocks. This result might be due to the fact that UNIDO uses ISIC industry classi?cation, but WorldScope and SDC use SIC classi?cation and that the mapping between ISIC and SIC introduces noise into the UNIDO measures. Yet another possibility is that the universe of WorldScope (large/listed ?rms) is more relevant to mergers compared to the small manufacturing ?rms in UNIDO.
59
I regress M&A volume of industry i in country c on the conditions of industry i in country c and the macroeconomic conditions of country c. I also control for the year ?xed-e?ects. M ergeri,c,t = ?0 + ?1 Ii,c,t?1 + ?2 Xc,t?1 +
i,c,t
where Ii,c is the condition of industry i in country c and Xc is the condition of country c. [INSERT TABLE 2.7 HERE] The results are reported in Table 2.7. In Table 2.7A, I regress cross-border mergers on acquirer country and industry indicators. In Table 2.7B, I regress cross-border mergers on target country and industry indicators. In Table 2.7C, I regress domestic mergers on the country and industry indicators. In columns 1-3 and columns 7-9, I use the representative measures as my regressors: industry ROA as the real industry indicator, gross value added as the real economy indicator, industry M/B as the industry valuation indicator, and stock market capitalization as the ?nancial market indicator. In columns 4-6 and columns 1012, I use the principal component indices as my regressors. The country-level principal component indicators are similar to the ones in section 2.4. The coe?cients of the industry-level real indicators are positive and statistically signi?cant. In most speci?cations, the country-level real indicators are driven down to zero or less signi?cant than the industry-level real indicators. From the third columns of Table 2.7A2.7C (the speci?cation in which I include the real indicators at the country level and control for the year ?xed-e?ect), the industry real indicator’s coe?cients are estimated at 0.06 for an acquirer country, 0.03 for a target country, and 0.03 for domestic mergers. The coe?cients of the country-level real indicators in the same regressions are 0.02, 0.03, and 0, respectively. The coe?cients of the industry-level valuation indicators and the country-level ?nancial indicators are positive and statistically signi?cant in all speci?cations. In most speci?cations, the country-level ?nancial indicators outperform the industry-level valuation shocks. From the ninth column of Table 2.7A-2.7C (the speci?cation in which I include the country-level ?nancial indicators and control for the year ?xed-e?ect), industry valuation’s coe?cients are estimated at 0.09 for an acquirer country, 0.02 for a target country, 60
and 0.04 for domestic mergers. The coe?cients of the country-level ?nancial indicators in the same regressions are 0.12, 0.02, and 0.10, respectively. The country-level patterns that I document in section 2.4 are still preserved at the industry level. All the coe?cients in Table 2.7 are positive, indicating that there are more mergers during booms. The coe?cients of the valuation/?nancial market indicators are higher than those of the real/ productivity indicators. The coe?cients in Table 2.7A are larger than the coe?cients in Tables 2.7B and 2.7C, suggesting that cross-border mergers are more pro-cyclical than domestic mergers and that the conditions of acquirer countries are more important than the conditions of target countries. In sum, shocks that e?ect mergers have both signi?cant industry components and signi?cant country components. Productivity shocks are mostly signi?cant at the industry level. On the other hand, the country-level ?nancial indicators are still important after controlling for the industry-level valuation shocks. These ?ndings support the literature on U.S. merger waves (e.g., Harford, 2005; Eisfeldt and Rampini, 2006; Maksimovic, Phillips, and Yang, 2009) in which mergers are a?ected by both productivity shocks and macro ?nancial shocks. Fact 4: Shocks that e?ect mergers have both signi?cant industry and signi?cant country components.
2.6
Additional Results
This section provides two sets of cross-sectional evidence to complement the time-series results. At the most aggregated level, I examine the long-run country e?ect on the 20-year aggregate of M&A activities. At the most disaggregated level, I examine the deal-level characteristics of domestics and cross-border M&As. The results in this section will give more information about the motives behind M&As and help to identify the issues that might be particularly pertinent to cross-border mergers. Country-Level I study the cross-country determinants of M&As by estimating the gravity model. The 61
gravity model is one of the most popular empirical models used in international trade. Instead of using di Giovanni (2005)’s panel version of the gravity model, which combines the time-series and the cross-sectional e?ects together, I use a cross-sectional version. M ergerc1 ,c2 = ?0 + ?1 Distancec1 ,c2 + ?2 Xc1 ,t0 + ?3 Xc2 ,t0 +
c1 ,c2 ,
where the variable M ergerc1 ,c2 is the 20-year aggregate volume of M&A ?ow from country c1 to country c2 . I de?ate the annual mergers data with the GDP de?ators and aggregate them from 1989 to 2008.7 I use two variables to measure the distance between the acquirer and the target countries: the geographical distance and the common language dummy. The geographical distance captures various aspects of the a?nity between the two countries. Examples include the volume of international trade, the presence of regional associations (such as EU and NAFTA), and the transportation costs. The common language dummy is a proxy for informational distance or the degree of information asymmetry between the two countries. The X s are the likely determinants of mergers in the long run. I use total population to measure country size, GDP to measure the level of income, and stock market capitalization to measure the degree of ?nancial development. The X s are measured in the base year (t0 =1988, outside of the M&As 20-year sample) to avoid the endogeneity problem. The descriptive statistics of the X s are reported in Table 2.9. [INSERT TABLE 2.8 AND TABLE 2.9 HERE] The results of the gravity model are reported in Table 2.8. From column 1 and column 2, the log of distance has negative and signi?cant coe?cients and the common language dummy has positive and signi?cant coe?cients. In other words, most mergers are between countries that are close together in terms of geographical and informational distance. If the distance of a country-pair is 1% smaller, then mergers will increase by 0.5-0.6%. If a country pair that does not share a common language adopts a common language, then mergers will increase by 1.8%. From columns 3 and 4, all four coe?cients of populations and GDPs are positive and statistically signi?cant. This result means that most M&A activities are between large
7
The de?ator used is the average GDP de?ator of the acquirer and the target countries.
62
and rich countries. A 1% increase in the 1988 population is associated with a 0.2% increase in M&As for the acquirer country and a 0.2% increase for the target countries. Keeping total population constant, a 1% increase in 1988 GDP is associated with a 1.1% increase in M&As for the acquirer country and a 0.5% increase for the target countries. The coe?cients of the acquirer countries are higher than the coe?cients of the target countries, implying that acquirer countries are on average larger and richer than the target countries. The results so far are consistent with the neoclassical theories and the standard results from international trade literature. That is, although ?rms tend to trade with partners in closer countries with larger markets, acquirers tend to seek targets in closer countries with a higher level of economic activities. Interestingly, the log of stock market capitalization in 1988 is highly signi?cant. From column 6, the coe?cient of acquirer market capitalization is estimated at 0.65, driving out the signi?cance of the GDP variable. The coe?cients of the target market capitalization is also statistically signi?cant but with a much smaller magnitude (0.15). This result suggests deep ?nancial markets, especially in the acquirer countries, are very important for M&As. Deal-level Here, I examine the characteristics of deals in my sample and compare the characteristics of domestic deals to the characteristics of cross-border deals. The characteristics I study are deal size, payment method, and the acquirer’s and target’s listing status. I am also interested in whether these deals are in high-tech, tradable, or related industries. “Deal Size” is the transaction value in millions of dollars. The “Cash-based Dummy” is a dummy variable taking the value of one if the percentage of cash is higher than the percentage of stock, and zero otherwise. The “Listed Acquirer” is a dummy taking the value of one if the acquirer is listed. The “Listed Target” is a dummy taking the value of one if the target is listed. The “Tradable” dummy is equal to one if the acquirer and the target are in tradable industries as de?ned by Aguiar and Gopinath (2005). The “High-tech” dummy is equal to one if the acquirer and the target are in the high-tech industry according to the American Electronic Association (http://www.aeanet.org/Publications/IDMK de?nition.asp). “Relatedness” is the absolute value of the di?erence between an acquirer’s four-digit SIC
63
and target’s four-digit SIC, as Alfaro and Charlton (2006) argue that related industries tend to have closer SIC codes. The results are reported in Table 2.10. Columns 1, 2, and 3 show the average characteristics of all deals, the domestic deals, and the cross-border deals, respectively. In column 4, I report the t-statistics of the di?erences between the domestic deals and the cross-border deals. On average, the cross-border deals are larger than the domestic deals and more likely to involve listed acquirers. This is consistent with the ?nding in section 2.4 that, among the listed acquirers, the cross-border acquirers tend to be larger than the domestic acquirers. I also ?nd that cross-border deals are more likely to be cash-based compared to the domestic deals. These comparisons suggest that ?nancial constraints might be more relevant to cross-border deals than to domestic deals. For the industry comparisons, cross-border deals are more likely to be in the tradable, hightech, and related industries. This is circumstantial evidence that cross-border mergers are more likely to be driven by neoclassical motives than domestic mergers; mergers in tradable industry are likely to be driven by comparative advantage and trade costs; ?rms in the high-tech industry are more likely to have ?rm speci?c assets that can be redeployed in another country; and mergers in related industries are more likely to generate synergies. A concern might be that the results in column 4 are driven by the di?erences in the compositions of the domestic deals and cross-border deals along country, industry, and year dimensions. For example, cross-border deals might cluster in certain countries or certain time periods compared to domestic deals. I address this problem by regressing deal characteristics on a cross-border dummy and controlling for country, industry, and year ?xed-e?ects (only country and year ?xed-e?ects for the industry comparisons). To ensure that the reported numbers are comparable to the t-statistics in column 4, I use a linear model instead of a logit or a probit. The coe?cients of the cross-border dummy are reported in column 5. These numbers are roughly similar to the ones in column 4.
64
2.7
Conclusion and Discussion
In this section, I discuss the results from sections 2.4-2.6 in light of the popular theories on mergers and then conclude. The Fire Sale Theory Krugman (1998), Aguiar and Gopinath (2005), and Acharya, Shin, and Yorulmazer (2009) propose the ?re sale theory of FDI in which foreign investors acquire ?rms in countries facing bad shocks such as ?nancial crises in order to take advantage of the liquidity constrained targets. The ?re sale theory has received broad empirical support. For example, Aguiar and Gopinath (2005) document an increase in foreign acquisitions during the East Asian ?nancial crisis. Desai, Foley, and Forbes (2007) ?nd that multinationals increase their investment in foreign a?liates when the host countries are facing currency crises. Using the data from 50 countries over the last 20 years, my results suggest the opposite: there are more mergers when the target economy is booming. Even after controlling for the acquirer’s boom and the global boom, there are still more foreign acquisitions when the target economy receives good shocks. In section 2.4, I ?nd that the correlations between cross-border mergers and the target’s macroeconomic conditions are approximately the same as the correlations between domestic mergers and domestic conditions. In other words, the capital in?ow through acquisitions is as pro-cyclical as domestic mergers. While M&As driven by the ?re sale motive might be present in a speci?c country at a speci?c time, most mergers in my sample are more consistent with the theory that ?rms invest in other countries to gain access to new markets and new investment opportunities and that it is better to enter the target countries when the demand is strong, the productivity is high, and the business environment is good. The Agency Theory Jensen (1986) indicates that M&As can be driven by agency problems: that is, the acquirer’s CEO might value mergers excessively. Although M&As might destroy ?rm value, corporate diversi?cation tends to reduce the risk of managerial human capital and enhance the CEO’s career prospects. The agency theory is strengthened by the fact that
65
numerous authors ?nd that conglomerates in the U.S. are traded at a discount, compared to single-segmented ?rms. The agency theory of mergers focuses on the acquirer’s problems and provides no speci?c predictions about the targets. However, in the international context, I ?nd that the characteristics of the target ?rms and the target countries have signi?cant e?ects on merger decisions. Moreover, Schenzler, Gande, and Senbet (2009) ?nd that global diversi?cations enhance ?rm values as measured by Tobin’s q. Their paper, combined with my ?ndings in section 2.5, shows that the acquirers are the more productive ?rms and the targets are the less productive ?rms. This ?nding suggests that most cross-border M&As tend to be driven by the value-enhancing neoclassical motives, rather than by value-destroying agency motives. The Misvaluation Theory Shleifer and Vishny (2003) and Rhodes-Kropf and Viswanathan (2004) propose a theory in which mergers are driven by stock market misvaluations. In their papers, targets accept the overpriced stock of the acquirers because they have a short time horizon or because they overestimate the synergies from the mergers. Rhodes-Kropf, Robinson, and Viswanathan (2005) ?nd that merger waves in the U.S. coincide with high M/B ratios and argue in favor of the misvaluation theory: when the market valuation is high, there are more M&As because acquirers will try to sell their overpriced stocks to targets. In the international context, Baker, Foley, and Wurgler (2009) use the U.S. foreign direct investment data to show that FDI ?ow is large when the source country stock market valuation is high. The authors attribute this ?nding to the misvaluation theory. In section 2.4, I study the behavior of ten macroeconomic indicators during the seven-year period around merger waves. I ?nd that most indicators, including the ones that are typically associated with the misvaluation theory like M/B and exchange rate, are highly correlated with one another and exhibit strong cyclical patterns. This correlation implies that M&As might not react directly to these indicators. Instead, all variables might be a part of a larger business cycle model in which M&As, market capitalization, and exchange rates are driven by common factors like productivity shocks. Moreover, my data do not re?ect the many predictions of the misvaluation theory. One 66
example is that misvaluation theory predicts that merger waves coincide with a strong acquirer currency. If the main motive of mergers is to take advantage of temporary exchange rate ?uctuations, then most mergers should occur when the acquirer’s currency is at its strongest relative to a target’s currency. While a one-year lagged exchange rate can predict mergers, I ?nd that M&As do not peak when the acquirer’s currency is at its strongest. The peak appreciation is three years prior to merger waves. Another example is the prediction about the method of payment. In section 2.6, I ?nd that the cross-border deals are less likely to be stock-based compared to domestic deals. The misvaluation theory predicts that domestic mergers are more pro-cyclical and more responsive to stock prices. However, my ?ndings in sections 2.4 and 2.5 are the opposite: cross-border mergers are much more pro-cyclical than domestic mergers.8 In this paper, I present key facts about international mergers. Speci?cally, I answer these four main questions: (1) How do cross-border mergers behave over a business cycle? International mergers come in waves and are very pro-cyclical. (2) Where do shocks that e?ect cross-border mergers originate? Most mergers occur when both the acquirer and the target economies are booming. (3) What type of shocks (real or ?nancial) e?ect crossborder mergers? Merger booms have industry-level (productivity shock) and country-level (?nancial shock) components. (4) What types of ?rms engage in cross-border mergers? Acquirers tend to be more productive than average ?rms and targets tend to be less productive than average ?rms. In the next chapter, guided by the key facts above, I propose and estimate a dynamic structural model that is built on the neoclassical theory of mergers.
8
9
I do not mean to suggest that misvaluations do not occur. It is possible that stock market booms
re?ect bubbles or other irrationalities in the stock market. Because the merging ?rms are not paying for such irrationalities, as shown in Shleifer and Vishny (2003) and Rhodes-Kropf and Viswanathan (2004), it does not matter from my modeling standpoint whether the low cost of capital comes from rational sources or irrational sources like bubbles. 9 Even though my ?ndings suggest that neither the agency nor the misvaluation theories are the main motive for cross-border mergers, the distinctions between these two theories and the neoclassical theories are not crucial from my modeling standpoint. The dynamic structural model in Chapter 3 can be reinterpreted from the misvaluation or agency angles conveniently. For example, the liquidity shocks in my model can be interpreted as misvaluation shocks and the productivity shocks in my model can be interpreted as
67
Table 2.1 Merger Activities over Time Year All Mergers (Frequency) 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 Total 6,473 8,918 9,451 12,964 12,616 13,214 15,603 18,945 20,188 22,233 24,541 26,689 29,138 22,049 19,270 20,393 22,450 25,163 27,223 28,766 26,523 412,810 Cross-border (Frequency) 1,450 2,177 2,378 2,706 2,408 2,654 3,271 3,950 4,239 4,836 5,673 6,504 7,731 5,358 3,937 3,960 4,682 5,547 6,274 7,033 5,857 92,625 All Mergers (Volume) 503 545 399 329 363 463 603 934 1,084 1,571 2,385 3,110 3,227 1,529 1,078 1,223 1,743 2,401 3,155 3,422 1,669 31,736 Cross-border (Volume) 99 122 125 66 73 85 108 186 197 289 559 971 972 431 268 232 441 608 828 1,068 524 8,252
The table reports the aggregate volume and the aggregate frequency of M&A activities from Thomson’s Securities Data Corporation (SDC). The sample covers all deals whose acquirers and targets are from 25 developed countries and 25 developing countries with the most M&A’s. Frequency is measured by the number of deals. Volume is the transaction value in billion of current dollars.
in?ated productivity shocks contaminated by managers’ private bene?ts of control.
68
Table 2.2 Merger Activities by Country Country Argentina Australia Austria Belgium Brazil Bulgaria Canada Chile China Colombia Czech Republic Denmark Egypt Finland France Germany Greece Hong Kong Hungary India Indonesia Ireland-Rep Israel Italy Japan Lithuania Malaysia Mexico Netherlands New Zealand Norway Peru Philippines Poland Portugal Number of Acquirers 1,001 15,715 2,408 3,292 2,543 378 20,643 561 4,099 213 818 3,546 236 5,101 18,229 21,070 1,102 5,630 927 4,095 624 2,287 1,126 7,771 18,595 184 7,424 970 8,207 2,582 3,291 219 786 1,277 1,415 Number of Targets 2,023 16,950 2,152 3,208 3,918 685 19,781 967 6,220 458 1,668 3,360 342 4,991 18,253 21,939 1,028 5,248 1,669 4,780 1,186 1,899 1,191 8,811 16,716 372 7,151 1,853 6,749 3,253 3,314 494 1,120 2,378 1,737 Volume Acquired 78 794 68 292 234 4 1,104 37 174 18 15 136 16 155 1,738 1,243 50 332 8 106 42 100 67 1,024 1,211 1 174 186 807 77 180 10 35 20 89 Volume Sold 136 800 80 262 338 15 1,151 68 255 45 49 136 32 132 1,309 1,298 56 270 33 128 64 82 59 1,025 1,107 5 150 224 701 97 187 26 53 58 94
69
Table 2.2 (Continued) Merger Activities by Country Country Romania Russian Fed Singapore Slovak Rep South Africa South Korea Spain Sweden Switzerland Thailand Turkey Ukraine United Kingdom United States Venezuela Number of Acquirers 215 3,342 4,618 204 2,841 2,891 7,764 7,400 5,162 1,494 518 271 49,040 158,492 193 Number of Targets 641 3,922 3,672 381 3,090 3,167 8,933 6,687 4,274 2,035 859 521 46,171 150,225 368 Volume Acquired 1 283 233 2 143 219 720 358 623 29 27 2 3,613 14,800 14 Volume Sold 20 302 129 11 150 259 582 417 432 47 73 17 3,305 15,400 22
The table reports the aggregate volume and the aggregate frequency of M&A activities from Thomson’s Securities Data Corporation (SDC). The sample covers all deals announced and completed between 1988 and 2008. Frequency is measured by the number of deals. Volume is the transaction value in billion of current dollars.
70
Table 2.3 Descriptive Statistics of Firms in WorldScope Variable Productivity Return on Assets Pro?t Margin Labor Productivity Sale Growth Employment Growth Dividend Valuation Market-to-Book Past Return-1 year Past Return-3 year Other Size Age1 Age2 Leverage 5.07 3.25 2.45 24.4 2.37 1.05 1.09 23.05 -0.5 0.69 0 0 10.13 4.9 4.41 93.03 1.41 -11.48 -17.69 1.77 67.14 110.88 0.03 -208.7 -339.56 8.91 129.88 206.23 2.22 -3.04 0.04 11.43 4.85 1.1 29.26 91.43 0.15 35.72 25.06 1.86 -133.27 -476.68 -0.31 -80.49 -61.47 0 34.68 88.09 0.67 124.66 84.73 8.32 Mean Standard Deviation Minimum Maximum
The table reports descriptive statistics of ?rms in the WorldScope database. Return on Assets is pro?t (EBITDA) divided by total assets; Pro?t Margin is pro?t divided by sales; Labor Productivity is pro?t per worker; Sales Growth is the percentage change in annual sales; Employment Growth is the percentage change in number of workers; Payout Ratio is dividends divided by total assets; M/B is market capitalization divided by (total assets less total debts); Past Return is the percentage change in market prices; Size is log of total assets (book value); Age1 is number of years since the incorporation date; Age2 is number of years since the listing date; and Leverage is total debts divided by total assets. All Percentage changes are calculated using the log formula.
71
Table 2.4A Correlations between Cross-Border Mergers and Acquirer’s Macroeconomic Conditions Cor(cross border mergers, acquirer’s Xt+j ) Real Economy Gross Value Added Gross Domestic Product Capital Formation Financial Markets Stock Market Capitalization Average M/B Ratio Average P/E Ratio Domestic Credit International Trade and Investment Current Account Balance Exchange Rate Net Foreign Portfolio Investment 0.08** -0.11*** -0.16 0.01 -0.15*** -0.28** 0.04 -0.11*** -0.06 -0.02 0 0.07 -0.04 0.09*** 0.12* 0 0.16*** 0.11 0.10*** 0.10*** -0.01 -0.11*** -0.17*** -0.12** -0.08* 0.12*** 0.02 0 0 0.31*** 0.26*** 0.15*** 0.06* 0.42*** 0.50*** 0.31*** 0.16*** 0.18*** 0.44*** 0.18*** 0.11*** -0.04 0.10* 0.01 0 -0.13*** -0.27*** -0.12** -0.10*** -0.09** -0.10** -0.08** 0.02 -0.02 0.05 0.14*** 0.11*** 0.15*** 0.26*** 0.25*** 0.24*** 0.23*** 0.22*** 0.19*** 0.07** 0.07** 0.04 -0.07* -0.06* -0.11*** t-3 t-2 t-1 t t+1 t+2 t+3
The table reports the correlations between the volume of cross-border mergers at time t and the acquirer country’s conditions at time t+j, where j is from -3 to +3. Gross Value Added, Gross Domestic Product, Capital Formation, Stock Market Capitalization, Domestic Credit, Current Account Balance, and Exchange Rate (local currency unit /U.S. dollar) are from the World Development Indicator Database. Average M/B Ratio and Average P/E Ratio are equal-weighted averages from Kenneth French’s website. Net Foreign Portfolio Investment (out?ow - in?ow) is from the Coordinated Portfolio Investment Survey Database. All variables are standardized by the Mendoza and Terrones’ procedure. The *, **, and *** indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
72
Table 2.4B Correlations between Cross-Border Mergers and Target’s Macroeconomic Conditions Cor(cross border mergers, target’s Xt+j ) Real Economy Gross Value Added Gross Domestic Product Capital Formation Financial Markets Stock Market Capitalization Average M/B Ratio Average P/E Ratio Domestic Credit International Trade and Investment Current Account Balance Exchange Rate Net Foreign Portfolio Investment 0.02 -0.12*** -0.03 -0.03 -0.10*** -0.04 -0.07** -0.02 -0.26*** -0.07** 0.09** 0.1 -0.13*** 0.13*** 0.13* -0.07** 0.17*** 0.20** 0.02 0.17*** -0.13 -0.04 -0.17*** -0.12** -0.06 0.09** 0.03 0.03 -0.02 0.23*** 0.24*** 0.18*** 0.08** 0.26*** 0.37*** 0.20*** 0.10*** 0.09*** 0.38*** 0.25*** 0.08** -0.11*** 0.09* 0.04 0.01 -0.20*** -0.24*** -0.13** -0.09*** -0.04 -0.06 0.02 0.01 0.02 0.11*** 0.11*** 0.12*** 0.17*** 0.18*** 0.18*** 0.18*** 0.17*** 0.18*** 0.13*** 0.06 0.04 -0.01 -0.11*** -0.13*** -0.15*** t-3 t-2 t-1 t t+1 t+2 t+3
The table reports the correlations between the volume of cross-border mergers at time t and the target country’s conditions at time t+j, where j is from -3 to +3. Gross Value Added, Gross Domestic Product, Capital Formation, Stock Market Capitalization, Domestic Credit, Current Account Balance, and Exchange Rate (local currency unit /U.S. dollar) are from the World Development Indicator Database. Average M/B Ratio and Average P/E Ratio are equal-weighted averages from Kenneth French’s website. Net Foreign Portfolio Investment (out?ow - in?ow) is from the Coordinated Portfolio Investment Survey Database. All variables are standardized by the Mendoza and Terrones’ procedure. The *, **, and *** indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
73
Table 2.4C Correlations between Domestic Mergers and Macroeconomic Conditions Cor(domestic mergers, Xt+j ) Real Economy Gross Value Added Gross Domestic Product Capital Formation Financial Markets Stock Market Capitalization Average M/B Ratio Average P/E Ratio Domestic Credit International Trade and Investment Current Account Balance Exchange Rate Net Foreign Portfolio Investment 0.09** -0.07* 0.21* 0.06* -0.05 -0.15 0.03 -0.04 -0.12 -0.09*** 0.02 -0.05 -0.13*** 0.08** 0.24*** -0.04 0.15*** -0.09 0.01 0.18*** 0.1 -0.09** -0.10* 0.01 -0.06 0.05 -0.03 -0.06 -0.08** 0.21*** 0.10** 0.09* -0.03 0.27*** 0.32*** 0.20*** 0.04 0.12*** 0.33*** 0.12** 0.08** 0 0.10** 0.03 0.06* -0.14*** -0.08 0 -0.01 -0.11*** -0.15*** -0.11*** -0.08* -0.09** -0.04 0.01 0.02 0.07** 0.07** 0.12*** 0.13*** 0.13*** 0.17*** 0.17*** 0.08** 0.10*** 0.08** -0.01 -0.03 -0.01 t-3 t-2 t-1 t t+1 t+2 t+3
The table reports the correlations between the volume of domestic mergers at time t and the macroeconomic conditions at time t+j where j is from -3 to +3. Gross Value Added, Gross Domestic Product, Capital Formation, Stock Market Capitalization, Domestic Credit, Current Account Balance, and Exchange Rate (local currency unit /U.S. dollar) are from the World Development Indicator Database. Average M/B Ratio and Average P/E Ratio are equal-weighted averages from Kenneth French’s website. Net Foreign Portfolio Investment (out?ow - in?ow) is from the Coordinated Portfolio Investment Survey Database. All variables are standardized by the Mendoza and Terrones’ procedure. The *, **, and *** indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
74
Table 2.5A Mergers and Country-Pair Conditions Y= Volume of Mergert Exchange Ratet?1 Real Economy Indicatoracquirer,t?1 Real Economy Indicatortarget,t?1 Financial Market Indicatoracquirer,t?1 Financial Market Indicatortarget,t?1 Year Fixed E?ect Observations R-squared No 23450 0 Yes 16838 0.02 1 -0.02** [0.01] 2 -0.01 [0.01] 0.03*** [0.01] 0.02** [0.01] 0.06*** [0.02] 0.04** [0.02] Yes 6905 0.03 Yes 14784 0.02 3 -0.02 [0.01] 4 0 [0.01] 0.03*** [0.01] 0.02*** [0.01] 0.05*** [0.01] 0.03*** [0.01] Yes 6457 0.03 5 -0.02 [0.01]
The table reports the coe?cient estimates from the country-pair regressions. The dependent variable is M&A volume. The explanatory variables are lagged Exchange Rate, lagged conditions of the acquirer country, and lagged conditions of the target country. Exchange Rate is in acquirer currency unit / target currency unit. In column 2-3, Real Economy Indicator is Gross Value Added and Financial Market Indicator is Stock Market Capitalization. In column 4-5, Real Economy Indicator is the ?rst principal component of {Gross Value Added, Gross Domestic Product, Capital Formation} and Financial Market Indicator is the ?rst principal component of {Stock Market Capitalization, Average M/B Ratio, Average P/E Ratio, Domestic Credit}. Also estimated but not reported are a constant term and the year ?xede?ects. Numbers in the brackets are the standard errors. All variables are standardized by the Mendoza and Terrones’ procedure. The *, **, and *** indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
75
Table 2.5B Mergers and Global Economic Conditions Y= Volume of Mergert Exchange Ratet?1 Real Economy Indicator acquirert?1 Real Economy Indicator targett?1 Financial Market Indicator acquirert?1 Financial Market Indicator targett?1 World Equity Markett?1 World Interest Ratet?1 Credit Spreadt?1 Observations R-squared 23450 0 0.09*** [0.02] 0.05*** [0.01] -0.05** [0.03] 13418 0.02 1 -0.02** [0.01] 2 -0.01 [0.01] 0.04*** [0.01] 0.03*** [0.01] 0.07*** [0.01] 0.03*** [0.01] 0.07*** [0.01] 0.03** [0.01] -0.04 [0.02] 17474 0.02 0.10*** [0.02] 0.05*** [0.01] -0.04 [0.03] 12390 0.01 3 -0.01 [0.01] 4 -0.01 [0.01] 0.03*** [0.01] 0.02*** [0.01] 0.05*** [0.01] 0.05*** [0.01] 0.16*** [0.02] 0.01 [0.03] -0.04 [0.04] 4946 0.02 5 -0.02 [0.02]
The table reports the coe?cient estimates from regressions of the MA volume between country c1 and c2 on the lagged conditions of the acquirer, c1, the lagged conditions of the target, c2, and the lagged global economic conditions. Exchange Rate is in acquirer currency unit / target currency unit. In column 2-4, real sector indicator is gross-value added and ?nancial sector indictor is market capitalization. In column 5-7, real sector indicator is the ?rst principal component of Gross Value Added, Gross Domestic Product, Capital Formation and ?nancial sector indictor is the ?rst principal component of stock market capitalization, Average M/B Ratio, Average P/E Ratio, Domestic Credit. World Equity Market is measured as the return on Morgan Stanley World Capital Index. World Interest Rate is measured as the average of US, Japan, and Germany three month treasury-bill rates. Credit Spread is computed as Moody’s AAA bond rate minus Moody’s BAA bond rate. Numbers in the brackets are the t-statistics. All variables are standardized by the Mendoza and Terrones’ procedure. *, **, and *** indicate statistical signi?cant at 10, 5, and 1 percent levels, respectively.
76
Table 2.6A Characteristics of Acquirers and Targets Compared to Their Peers Acquirer minus Target All Mergers 1 2 3 4 5 6 Cross-Border All Mergers Cross-Border All Mergers Cross-Border Acquirer vs Target (Kolmogorov-Smirnov) Acquirer minus Non-Acquirer Target minus Non-Target All Mergers 7 Cross-Border 8
Characteristics
Productivity 0.37*** [58.174] 0.3*** [46.983] 0.27*** [33.212] 0.36*** [46.303] 0.48*** [51.972] -0.03*** [-3.493] [-0.222] 0 0.06*** 0.06*** [19.327] 0.28*** 0.22*** 0.15*** [15.407] 0.18*** 0.19*** 0.14*** [18.31] [53.784] 0.27*** [84.944] 0.39*** [112.805] 0.09*** [27.822] 0.24*** 0.19*** 0.21*** 0.18*** [27.296] [56.433] 0.27*** 0.2*** 0.2*** 0.17*** [35.25] [65.26] 0.35*** 0.2*** 0.2*** 0.2*** 0.23*** [51.838] 0.18*** [41.868] 0.18*** [39.268] 0.16*** [35.525] 0.27*** [55.329] 0.13*** [29.15] -0.1*** [-19.972] -0.06*** [-12.154] -0.03*** [-4.921] -0.05*** [-8.569] -0.05*** [-8.376] -0.06*** [-11.774] -0.13*** [-12.059] -0.09*** [-8.255] -0.04*** [-3.552] 0.01 [0.602] 0.02* [1.466] -0.09*** [-8.638]
Return on Assets
Pro?t Margin
Labor Productivity
Sale Growth
77
0.03*** [4.005] 0.22*** [28.991] 0.19*** [26.887] [18.356] 0.21*** [14.402] 0.17*** 0.17*** 0.18*** 0.12*** [-0.824] 0.11*** -0.01 0.06*** 0.08***
Employment Growth
Dividend
Valuation 0.05*** [16.145] 0.12*** [37.103] 0.24*** [67.907] 0.07*** [14.857] 0.09*** [20.79] 0.21*** [42.01] 0.01*** [2.788] 0.09*** [17.639] 0.03*** [4.944] 0.07*** [6.017] 0.09*** [8.921] 0.03*** [2.922]
Market-to-Book
Past Return-1 year
Past Return-3 year
Table 2.6A (Continued) Characteristics of Acquirers and Targets Compared to Their Peers Acquirer vs Target (Kolmogorov-Smirnov) All Mergers 3 4 5 6 7 Cross-Border All Mergers Cross-Border All Mergers Acquirer minus Non-Acquirer Target minus Non-Target Cross-Border 8
Characteristics Cross-Border 2
Acquirer minus Target
All Mergers 1
Other 0.68*** [57.784] 0.26*** [14.708] 0.32*** [15.931] -0.03*** [-2.531] 0.06*** 0.06*** 0.08*** [12.437] [43.355] 0.09*** 0.14*** 0.21*** [46.002] 0.09*** 0.13*** 0.2*** 0.37*** [59.24] 0.39*** [56.154] 0.07*** [16.947] [252.874] [242.22] 0.28*** 0.29*** 0.76*** 1.04*** 0.22*** [42.124] 0.05*** [6.659] 0.09*** [11.078] 0.1*** [19.823] 0.41*** [37.821] 0.11*** [7.886] 0.07*** [4.313] 0.13*** [12.437]
Size
0.64***
[89.504]
Age1
0.17***
[15.742]
Age2
0.16***
[13.8]
Leverage
-0.04***
[-5.985]
78
Observations are grouped by country-industry-year. Z-scores are computed from the mean and the standard deviation of each group. Column 1
reports the di?erence between the average Z-scores of all acquirers and the average Z-score of all targets. Column 2 reports the di?erence between
the cross-border acquirers and the cross-border targets. The numbers in the parenthesis are the t-statistics from the t-tests. Column 3 reports the
distributional distances (the Kolmogorov-Smirnov’s D-statistics) between all acquirers and all targets. Column 4 reports the distributional distances
between the cross-border acquirers and the cross-border targets. Column 5-6 compares the acquirers with the non-acquirers and column 7-8 compares
the targets with the non-targets. The numbers in the parenthesis are the t-statistics from the t-tests. *, **, and *** indicate statistical signi?cant
at 10, 5, and 1 percent levels, respectively.
Table 2.6B Comparisons of Acquirers and Targets Acquirer minus Target All Mergers 1 2 3 4 5 6 Cross-Border All Mergers Cross-Border All Mergers Cross-Border Acquirer vs Target (Kolmogorov-Smirnov) Acquirer minus Non-Acquirer Target minus Non-Target All Mergers 7 Cross-Border 8
Characteristics
Productivity 0.28*** [46.262] 0.2*** [32.282] 0.19*** [23.513] 0.37*** [47.329] 0.53*** [54.454] 0.16*** [22.034] [16.303] 0.2*** 0.24*** 0.11*** [22.795] 0.35*** 0.18*** 0.2*** [17.591] 0.21*** 0.16*** 0.14*** [14.271] [39.805] 0.29*** [90.633] 0.45*** [127.438] 0.08*** [26.586] 0.18*** 0.15*** 0.2*** 0.13*** [21.076] [45.881] 0.18*** 0.1*** 0.15*** 0.14*** [32.445] [67.963] 0.28*** 0.17*** 0.18*** 0.21*** 0.26*** [59.93] 0.17*** [38.101] 0.16*** [34.316] 0.18*** [40.218] 0.32*** [64.735] 0.08*** [18.828] -0.02*** [-3.053] -0.01 [-1.969] -0.01 [-0.969] -0.04*** [-7.3] -0.04*** [-5.673] -0.03*** [-6.534] -0.02** [-2.114] -0.01 [-0.961] 0 [-0.381] 0 [0.188] 0.02 [1.229] -0.03*** [-2.675]
Return on Assets
Pro?t Margin
Labor Productivity
Sale Growth
79
0.07*** [10.159] 0.01* [1.566] 0.21*** [28.831] [16.861] 0.2*** [-2.906] 0.15*** 0.19*** -0.03*** 0.11*** [9.741] 0.13*** 0.11*** 0.06*** 0.08***
Employment Growth
Dividend
Valuation 0.12*** [36.681] 0.28*** [78.027] 0.11*** [36.742] 0.09*** [19.878] 0.24*** [47.424] 0.19*** [44.245] 0.11*** [21.396] 0.08*** [13.725] -0.04*** [-7.248] 0.11*** [10.77] 0.04*** [3.728] 0 [-0.4]
Market-to-Book
Past Return-1 year
Past Return-3 year
Table 2.6B (Continued) Comparisons of Acquirers and Targets Acquirer vs Target (Kolmogorov-Smirnov) All Mergers 3 4 5 6 7 Cross-Border All Mergers Cross-Border All Mergers Acquirer minus Non-Acquirer Target minus Non-Target Cross-Border 8
Characteristics Cross-Border 2
Acquirer minus Target
All Mergers 1
Other 0.75*** [63.407] 0.29*** [16.128] 0.36*** [17.461] -0.04*** [-3.456] 0.06*** 0.06*** 0.03*** [10.904] [47.744] 0.1*** 0.17*** 0.23*** [44.363] 0.09*** 0.16*** 0.2*** 0.44*** [69.099] 0.44*** [63.559] 0.01* [1.545] [240.488] [242.725] 0.28*** 0.32*** 0.73*** 1.05*** 0.23*** [42.82] 0.08*** [11.507] 0.07*** [8.735] 0.06*** [11.681] 0.36*** [32.972] 0.17*** [12.079] 0.09*** [5.605] 0.07*** [6.83]
Size
0.63***
[86.414]
Age1
0.14***
[12.889]
Age2
0.19***
[15.729]
Leverage
-0.04***
[-6.255]
80
Observations are grouped by country-year. Z-scores are computed from the mean and the standard deviation of each group. Column 1 reports
the di?erence between the average Z-scores of all acquirers and the average Z-score of all targets. Column 2 reports the di?erence between the
cross-border acquirers and the cross-border targets. The numbers in the parenthesis are the t-statistics from the t-tests. Column 3 reports the
distributional distances (the Kolmogorov-Smirnov’s D-statistics) between all acquirers and all targets. Column 4 reports the distributional distances
between the cross-border acquirers and the cross-border targets. Column 5-6 compares the acquirers with the non-acquirers and column 7-8 compares
the targets with the non-targets. The numbers in the parenthesis are the t-statistics from the t-tests. *, **, and *** indicate statistical signi?cant
at 10, 5, and 1 percent levels, respectively.
Table 2.6C Comparisons of Merging Firms during Booms and Busts Real Booms minus Real Busts All Acq 1 2 3 4 5 6 7 Cross-Border Acq All Targets Cross-Border Targets All Acq Cross-Border Acq All Targets Financial Booms minus Financial Busts Cross-Border Targets 8
Characteristics
Productivity 0.04*** [5.108] 0.04*** [5.429] 0.13*** [13.696] 0.23*** [21.791] 0.3*** [24.95] 0.02* [1.569] [-1.44] [8.791] [4.855] -0.02* 0.16*** 0.17*** [19.379] [10.006] [6.648] 0.31*** 0.22*** 0.31*** [17.015] [12.041] [5.896] 0.23*** 0.23*** 0.24*** 0.39*** [35.682] 0.39*** [31.18] 0.08*** [8.548] [12.279] [8.846] [1.51] [11.264] 0.16*** 0.18*** 0.07* 0.12*** [3.938] [7.057] [-0.145] [2.106] [1.57] 0.13*** [8.745] 0.34*** [21.986] 0.42*** [22.81] 0.04*** [2.646] 0.04*** 0.14*** -0.01 0.02** 0.02* [4.891] [9.793] [1.512] [5.298] [3.158] 0.05*** 0.2*** 0.07* 0.05*** 0.04*** 0.04** [1.793] 0.03 [1.159] 0.1*** [4.225] 0.33*** [15.756] 0.32*** [13.174] 0.14*** [7.699] 0.04 [0.883] -0.06* [-1.34] 0.05 [1.086] 0.23*** [5.419] 0.38*** [7.51] 0.11*** [2.894]
Return on Assets
Pro?t Margin
Labor Productivity
Sale Growth
81
0.03*** [3.166] -0.17*** [-17.634] 0.06*** [6.012] [-5.934] -0.08*** [-18.474] [-6.113] 0.17*** [8.398] -0.26*** -0.11*** [-0.029] [5.543] 0 0.1*** 0.22*** [5.509] -0.23*** [-5.702] 0.07** [1.73]
Employment Growth
Dividend
Valuation 0.31*** [29.014] -0.08*** [-7.499] 0.17*** [16.7] 0.38*** [21.306] -0.16*** [-9.579] 0.13*** [8.524] 0.31*** [15.207] -0.01 [-0.669] 0.26*** [11.948] 0.45*** [9.65] 0.01 [0.118] 0.26*** [5.912]
Market-to-Book
Past Return-1 year
Past Return-3 year
Table 2.6C (Continued) Comparisons of Merging Firms during Booms and Busts Real Booms minus Real Busts Cross-Border Acq 2 3 4 5 6 7 All Targets Cross-Border Targets All Acq Cross-Border Acq All Targets Financial Booms minus Financial Busts Cross-Border Targets 8
Characteristics
All Acq 1
Other -0.12*** [-8.604] 0.07*** [3.091] 0.05** [2.059] -0.08*** [-6.883] [-7.583] [-3.844] [2.01] -0.13*** -0.14*** 0.02** [-1.579] [-4.355] [-3.084] -0.05* -0.27*** -0.05*** -0.04* [-1.644] 0 [0.375] [-0.779] [-0.831] [-8.209] [-5.758] -0.02 -0.04 -0.13*** -0.14*** [0.644] [-1.952] [0.331] [-0.397] 0.01 -0.07** 0 -0.01 0.02 [1.222] -0.07*** [-2.878] -0.01 [-0.279] -0.09*** [-4.857] 0.02 [0.553] -0.12*** [-2.371] -0.08* [-1.291] -0.09** [-2.247]
Size
-0.11***
[-11.108]
Age1
-0.02
[-1.211]
Age2
-0.06***
[-3.62]
Leverage
-0.03***
[-3.145]
82
Observations are grouped by country-industry. Z-scores are computed from the mean and the standard deviation of each group. Columns 1 and
5 report the di?erence between the average Z-scores of acquirers during booms and the average Z-score of acquirers during busts. Columns 2 and
6 report the di?erence between the cross-border acquirers during booms and the cross-border acquirers during busts. Columns 3 and 7 report
the di?erence between the targets during booms and the targets during busts. Columns 4 and 8 report the di?erence between the cross-border
targets during booms and the cross-border targets during busts. Booms (Busts) in Columns 1-4 are the periods in which HP detrended Gross Value
Added is above (below) one (minus one) standard deviation. Booms (Busts) in Columns 5-8 are the periods in which HP detrended Stock Market
Capitalization is above (below) one (minus one) standard deviation. The numbers in the parenthesis are the t-statistics from the t-tests. *, **, and
*** indicate statistical signi?cant at 10, 5, and 1 percent levels, respectively
Table 2.6D Post-Merger Operating Performance Unmatched Cross-Border Acq 2 -3.5*** [-7.36] -7.13*** [-12.55] -9.86*** [-15.4] -10.95*** [-15.53] [1.48] [1.44] [0.93] 1.52* 2.83* 1.03 [1.25] [1.34] [0.39] [-2.23] -6.26*** [-4.51] 1.15* 2.39* 0.38 -2.84** [-0.64] [-3.2] [1.3] [-0.14] -0.51 -5.02*** 1.09* -0.16* [-1.74] [-1.67] [1.01] [1.01] [0.05] 0.66 [0.44] 2.61* [1.54] 1.14 [0.58] -1.13** -2.14** 0.72 0.95 0.06 3 4 5 6 7 All Targets Cross-Border Targets All Acq Cross-Border Acq All Targets Propensity Score Matched Cross-Border Targets 8 -0.7 [-0.32] -6.92** [-2.3] 0.75 [0.21] 3.71 [0.99]
Time Windows
All Acq 1
t-1 to t
-3.17***
[-8.59]
t-1 to t+1
-4.78***
[-10.81]
t-1 to t+2
-6.07***
[-12.18]
t-1 to t+3
-6.05***
[-10.94]
83
Pro?tability is measured as ROA. t denotes the year of acquisition. Column 1 reports the di?erence between the average pro?tability changes of all
acquirers and the average changes of the non-acquirers. Column 2 reports the di?erence between the cross-border acquirers and the non-acquirers.
Column 3 reports the di?erence between the targets and the non-targets. Column 4 reports the di?erence between the cross-border targets and
the non-targets. Columns 5-8 reports propensity-score-matched e?ects (average treatment e?ects on the treated). Propensity scores are computed
from the Probit model using the following covariates: lagged pro?tability, sales, total assets, and ?rm age. The numbers in the parenthesis are the
t-statistics from the t-tests. *, **, and *** indicate statistical signi?cant at 10, 5, and 1 percent levels, respectively.
Table 2.7A Cross-Border Mergers and Industry Shocks in Acquirer Countries 1 0.08*** [0.01] 0.04*** [0.01] 0.15*** [0.01] [0.01] 0.15*** [0.01] Yes 10537 0.01 0.01 0.04 0.01 0.01 0.04 8519 8519 9029 7198 7198 Yes 10261 0.02 10198 0.04 0.10*** [0.01] [0.01] [0.01] 0.09*** [0.01] 0.12*** [0.01] Yes 10198 0.05 8515 0.01 4739 0.02 0.09*** [0.01] 0.08*** [0.01] 0.08*** [0.01] 0.08*** [0.01] 0.05*** [0.02] Yes 4739 0.05 0.02 0.03*** 0.01 [0.01] [0.01] [0.01] [0.01] [0.01] 0.09*** 0.06*** 0.07*** 0.06*** 0.04*** 2 3 4 5 6 7 8 9 10 11 12
Y=Cross-border Mergers
Real Industry Indicatori,c,t?1
Real Economy Indicatorc,t?1
Industry Valuation Indicatori,c,t?1
Financial Market Indicatorc,t?1
Year Fixed E?ects
Observations
R-squared
84
The table reports the coe?cient estimates from regressions of the cross-border M&A volume on the lagged conditions of the acquirer industry and
country. In column 1-3 and column 7-9, the industry productivity shock is average ROA, the real sector indicator is gross value added, the industry
valuation shock is average M/B, and the ?nancial market indicator is stock market capitalization. In column 4-6 and column 10-12, the industry
productivity shock is the ?rst principal component of 6 productivities measures in Table 2.6, the real sector indicator is the ?rst principal component
measure from section 2.4, the industry valuation shock is the ?rst principal component of 3 valuation measures in Table 6, and the ?nancial market
indicator is the ?rst principal component measure from section 2.4. Also estimated but not reported are a constant term and the year ?xed-e?ects.
Numbers in the brackets are the standard errors. All variables are standardized by the Mendoza and Terrones’ procedure. The *, **, and ***
indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
Table 2.7B Cross-Border Mergers and Industry Shocks in Target Countries 1 0.05*** [0.01] 0.05*** [0.01] 0.06*** [0.01] 0.09*** [0.01] Yes 11557 0 0 0.03 0 0 0.03 9329 9329 9621 7597 7597 Yes 11180 0 11106 0.01 [0.01] 0.03** [0.01] [0.01] [0.01] 0.02 [0.01] 0.02** [0.01] Yes 11106 0.03 9236 0 4770 0.02 0.02*** [0.01] 0.06*** [0.01] 0.07*** [0.01] 0.06*** [0.02] 0.04** [0.02] Yes 4770 0.04 0.03** 0.03*** 0.02** [0.01] [0.01] [0.01] [0.01] [0.01] 0.05*** 0.03*** 0.03*** 0.02** 0.01 2 3 4 5 6 7 8 9 10 11 12
Y=Cross-border Mergers
Real Industry Indicatori,c,t?1
Real Economy Indicatorc,t?1
Industry Valuation Indicatori,c,t?1
Financial Market Indicatorc,t?1
Year Fixed E?ects
Observations
R-squared
85
The table reports the coe?cient estimates from regressions of the cross-border M&A volume on the lagged conditions of the target industry and
country. In column 1-3 and column 7-9, the industry productivity shock is average ROA, the real sector indicator is gross value added, the industry
valuation shock is average M/B, and the ?nancial market indicator is stock market capitalization. In column 4-6 and column 10-12, the industry
productivity shock is the ?rst principal component of 6 productivities measures in Table 6, the real sector indicator is the ?rst principal component
measure from section 2.4, the industry valuation shock is the ?rst principal component of 3 valuation measures in Table 2.6, and the ?nancial market
indicator is the ?rst principal component measure from section 2.4. Also estimated but not reported are a constant term and the year ?xed-e?ects.
Numbers in the brackets are the standard errors. All variables are standardized by the Mendoza and Terrones’ procedure. The *, **, and ***
indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
Table 2.7C Domestic Mergers and Industry Shocks 1 0.04*** [0.01] 0.01 [0.01] 0.07*** [0.01] 0.10*** [0.01] Yes 11403 0 0 0.01 0 0 0.02 0 9197 9197 9533 7528 7528 11054 Yes 10981 0.01 [0.01] 0.04*** 0.04*** [0.01] 0.10*** [0.01] Yes 10981 0.02 9152 0 4749 0.01 [0.01] [0.01] [0.01] 0.05*** [0.01] 0.06*** [0.01] 0.04*** [0.01] 0.08*** [0.01] 0.02 [0.02] Yes 4749 0.03 0 0.01 0 [0.01] [0.01] [0.01] [0.01] [0.01] 0.04*** 0.03** 0.04*** 0.04*** 0.03*** 2 3 4 5 6 7 8 9 10 11 12
Y= Domestic Mergers
Industry Productivityi,c,t?1
Real Economy Indicatorc,t?1
Industry Valuationi,c,t?1
Financial Market Indicatorc,t?1
Year Fixed E?ects
Observations
R-squared
86
The table reports the coe?cient estimates from regressions of the domestic M&A volume on the lagged conditions of the domestic industry and
country. In column 1-3 and column 7-9, the industry productivity shock is average ROA, the real sector indicator is gross value added, the industry
valuation shock is average M/B, and the ?nancial market indicator is stock market capitalization. In column 4-6 and column 10-12, the industry
productivity shock is the ?rst principal component of 6 productivities measures in Table 2.6, the real sector indicator is the ?rst principal component
measure from section 2.4, the industry valuation shock is the ?rst principal component of 3 valuation measures in Table 6, and the ?nancial market
indicator is the ?rst principal component measure from section 2.4. Also estimated but not reported are a constant term and the year ?xed-e?ects.
Numbers in the brackets are the standard errors. All variables are standardized by the Mendoza and Terrones’ procedure. The *, **, and ***
indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
Table 2.8 Cross-Sectional Gravity Model Y= Log of Aggregate Volume of Merger (from 1989-2008) Log of Distance 1 -0.58*** [0.08] Common Language Dummy 2 -0.53*** [0.08] 1.77*** [0.16] Log of Populationacquirer,1988 Log of Populationtarget,1988 Log of Real GDPacquirer,1988 Log of Real GDPtarget,1988 Log of Market Capitalizationacquirer,1988 Log of Market Capitalizationtarget,1988 Constant Observations R-squared Yes 1052 0.04 Yes 1052 0.14 Yes 1052 0.16 Yes 1015 0.38 3 -0.62*** [0.08] 1.64*** [0.16] 0.19*** [0.05] 0.18*** [0.06] 4 -0.61*** [0.07] 1.21*** [0.15] 0.73*** [0.05] 0.45*** [0.06] 1.11*** [0.06] 0.53*** [0.06] 5 -0.84*** [0.07] 1.23*** [0.14] 0.69*** [0.05] 0.56*** [0.06] 0.09 [0.12] 0.32*** [0.12] 0.65*** [0.07] 0.15** [0.06] Yes 875 0.46
The table reports the coe?cient estimates from regressions of the 1989-2008 aggregate volume of M&A ?ow from country c1 to country c2 on the conditions of the acquirer, c1 , and the conditions of the target, c2 , in 1988. Distance and Common Language dummy are from Di Giovanni (2005). Population, GDP, and Stock Market Capitalization are from the World Development indicator Database. Numbers in the brackets are the standard errors. The *, **, and *** indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
87
Table 2.9 1988 Country Characteristics Variable Distance Common Language Dummy Population Real GDP Stock Market Capitalization Mean 8.696378 0.207841 16.99215 8.732422 7.200922 Standard Deviation 0.911802 0.405871 1.383335 1.245854 2.339167 Minimum 5.41272 0 14.86143 5.692476 0.419264 Maximum 9.895177 1 20.82006 10.35871 10.60626
The table reports the summary statistics of country characteristics in 1988. Distance and Common Language dummy are from di Giovanni (2005). Population, GDP, and Stock Market Capitalization are from the World Development Indicator Database. All the variables except the common language dummy are in log form.
88
Table 2.10 Characteristics of Domestic versus Cross-Border Deals Deal Characteristics All Deal Average 1 Deal Size 165.39 Domestic Deal Average 2 157.27 Cross-Border Deal Average 3 193.86 Cross-Border - Domestic 4 36.58*** [4.89] Prob(Cash Deals) 0.51 0.49 0.55 0.06*** [4.39] Prob(Listed Acquirer) 0.45 0.43 0.53 0.11*** [57.45] Prob(Listed Target) 0.14 0.15 0.13 -0.02*** [-18.81] Prob(Tradable Industry) 0.2 0.17 0.29 0.13*** [85.32] Prob(High-Tech Industry) 0.11 0.1 0.13 0.03*** [25.55] Relatedness 1106.69 1122.11 1053.35 -68.75*** [-11.46] Controlled for Fixed E?ects 5 40.11*** [4.7] 0.05*** [3.18] 0.11*** [58.8] 0 [1.27] 0.12*** [76.29] 0.05*** [41.02] -94.89*** [-14.24]
Deal Size is the transaction value in million of current dollars. Cash Deals is a dummy variable taking the value of one if the percentage of cash is higher than the percentage of stock. Listed Acquirer is a dummy taking the value of one if the acquirer is listed. Listed Target is a dummy taking the value of one if the target is listed. Tradable is equal to one if the acquirer and the target are in the tradable industries as de?ned by Aguiar and Gopinath (2005). High-Tech is equal to one if the acquirer and the target are in the high-tech industry according to the American Electronic Association. Relatedness is the absolute value of the di?erence between the acquirer’s 4-digit SIC and the target’s 4-digit SIC. Columns 1, 2 and 3 shows the average characteristics of all deals, the domestic deals, and the cross-border deals. Column 4 shows the di?erences between the domestic deals and the cross-border deals. Column 5 shows coe?cients of the cross-border dummy after controlling for the ?xed-e?ects. Numbers in the brackets are t-statistics. The *, **, and *** indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
89
Figure 2.1 Volume of Aggregate Mergers
The ?gure shows the aggregate volume of M&A activities from the top 50 countries in trillion of current dollars.
90
Figure 2.2A US Acquisition of Foreign Firm
Figure 2.2B US Acquisition of Foreign Firm: Hodrick-Proscott Detrended
The ?gures show an example of the raw and the detrended series of US ?rm acquisitions of assets in other countries.
91
Appendix 2.1: The Country Coverage of WorldScope
Full-Coverage Developed countries include Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Hong Kong, Ireland, Italy, Japan, the Netherlands, Norway, Portugal, Singapore, Spain, Sweden, Switzerland, the United Kingdom, and the United States. Emerging markets include Brazil, China, Indonesia, Korea, Malaysia, Mexico, Philippines, South Africa, and Thailand. Targeted Coverage Countries include Argentina, Chile, Colombia, Czech Republic, Egypt, Hungary, India, Israel, Jordan, New Zealand, Peru, Poland, Russia, Slovakia, Turkey, and Venezuela.
92
Chapter 3
A Dynamic Model of International Mergers and Acquisitions
3.1 Introduction
In the past two decades, 26% of worldwide M&A activities involve acquirers and targets from di?erent countries. The aggregate volume of cross-border mergers from 1989 to 2008 adds up to above 8 trillion dollars. In spite of such a large volume, much of the M&A literature focuses on domestic mergers. Moreover, the amount of cross-border mergers varies greatly from year to year. For example, the volume of worldwide M&A deals dropped by 62% from 2000 to 2003 but bounced back by 158% in 2006. Despite such a large year-toyear ?uctuation, most existing papers on cross-border M&As study the e?ects of long-run determinants like corporate governance and capital market development. These gaps in the literature motivate the research questions that are at the core of this paper: what are the dynamic patterns of cross-border mergers, and what are the factors that drive them? In the previous chapter, I present key facts about international mergers. Speci?cally, I answer these four main questions: (1) How do cross-border mergers behave over a business cycle? International mergers come in waves and are very pro-cyclical. (2) Where do shocks that e?ect cross-border mergers originate? Most mergers occur when both the acquirer and
93
the target economies are booming. (3) What type of shocks (real or ?nancial) e?ect crossborder mergers? Merger booms have industry-level (productivity shock) and country-level (?nancial shock) components. (4) What types of ?rms engage in cross-border mergers? Acquirers tend to be more productive than average ?rms and targets tend to be less productive than average ?rms. In this chapter, I use the four empirical facts mentioned earlier as a guideline and build a dynamic structural model of cross-border mergers. The dynamic structural approach o?ers two major advantages. First, by construction, it solves the identi?cation problem inherent in reduced-form estimation. Using simulated data, I can quantify the e?ects of productivity and ?nancial shocks on endogenous variables. Second, the dynamic structural model provides me with an analytical framework to investigate the impacts of various government policies. As an example of such policy analyses, I examine the impact of multinational corporation taxation which has long been the subject of heated policy debates. My model is related to Gomes and Livdan (2004) and Yang (2008) in that ?rms make investment and merger decisions based on the productivity shocks they received. To investigate the e?ects of ?nancial shocks, I incorporate external ?nancing cost similar to the ones in Gomes (2001) and Whited (2006) and allow the cost to ?uctuate along a business cycle. The distinguishing features of my model are that there are two countries and that a local ?rm has an option of engaging in cross-border mergers in order to become a multinational corporation. I also assume that the productivity shock has two components: ?rm-speci?c and location-speci?c. With this setup, productive ?rms will seek assets in booming locations. Recently, President Obama proposed a 200 billion dollar tax increase on multinational corporations. As a consequence, the largest US corporations are concerned that the tax raise will put American ?rms at a disadvantage overseas and leave them vulnerable to foreign acquisitions.1 Others are concerned that the tax will primarily impact the productive sectors, such as technology and pharmaceutical industries.2 Clearly, there is an urgent need to understand the e?ects of multinational taxation. The simulation results from my tax
1 2
BusinessWeek May 4, 2009 The New York Times May 5, 2009
94
experiments con?rm that foreign operation tax can be very distortionary for cross-border mergers and has larger e?ects on more productive ?rms. The model also provides a policy implication: when analyzing the e?ects of multinational corporation taxation, we should be careful not to focus only on multinational ?rms that already have overseas operations. Since cross-border mergers are sensitive to tax rate, we must also take into consideration its e?ects on productive local ?rms for whom the tax is a disincentive for future mergers. This paper joins the growing literature on dynamic corporate ?nance (e.g., Whited, 2006; Gomes and Livdan, 2004; Yang, 2008). The dynamic structural approach is particularly appropriate for my context, since merger waves are, by nature, dynamic phenomena. The dynamic simulations in this paper are also related to the recent work on the impact of multinational corporation taxation. While the literature on taxation is voluminous, identifying the e?ect of taxes can pose a challenge since the tax policies are likely to be endogenous. Even if the tax policies are exogenous, their e?ects, as measured by the reduced-form coe?cients, might still be endogenous according to the Lucas’ critique. Dharmapala, Foley, and Forbes (2009) and Faulkender and Peterson (2009) use the Homeland Investment Act of 2004 as a natural experiment and analyze the e?ects of this one-time tax break on U.S. ?rms. In this paper, I o?er structural estimation as an alternative approach to address the identi?cation problem.
3.2
Conceptual Framework
The goal of chapter 3 is to integrate the ?ndings from chapter 2 into a dynamic structural model and policy analysis. Because implications from a structural model are, to a great extent, driven by its structural assumptions, the strength of my analysis lies in the fact that my model is consistent with the key facts derived from a large amount of data. Guided by the reduced-form evidence, I develop a dynamic structural model of crossborder mergers. I assume that ?rm decisions are driven by productivity shocks under the presence of ?nancial frictions. Merger gain comes from access to the target country’s markets and resources as well as the utilization of acquirer ?rm-speci?c assets. With these assumptions, productive ?rms will seek assets in booming locations. Then, I prove that a 95
solution to the problem indeed exists and characterize the properties of the model. These properties can provide insights into ?rms’ merger and investment decisions. Next, I solve the model numerically using the value function iteration algorithm. Given the value functions and the policy functions, I construct a panel of ?rms, generate structural shocks, and observe how ?rms react to these shocks. These exercises allow me to quantify the e?ects of productivity and ?nancial shocks on the endogenous variables. In addition, I use the model to perform policy experiments on taxation. The issue of multinational corporation taxation frequently captures public attention. On May 4, 2009, President Obama proposed a 200 billion dollar tax increase on multinational corporations. As a consequence, the largest U.S. corporations have launched a vigorous lobbying e?ort against the plan. One of the arguments against the tax increase is that it will put American ?rms at a competitive disadvantage overseas and leave them vulnerable to foreign acquisitions. There is also some concern that the tax will primarily impact the productive sectors, such as the technology and pharmaceutical industries. Given my structural model, I can investigate how the tax on foreign operations might in?uence ?rm investment and merger decisions. I can also verify whether the concerns about the tax proposal above are valid.
3.3
The Model and its Basic Properties
I build a neoclassical model of cross-border investments. Firms make investment and production decisions in the presence of productivity shocks. The model is related to the domestic investment models in Gomes and Livdan (2004), Cooper and Haltiwanger (2006), as well as in Yang (2008). The distinguishing feature of this model is that I allow domestic ?rms to acquire establishments in another country and become multinationals. Posit that there are two countries, A and B . In each country, there are a large number of ?rms so that each ?rm is a price-taker in the market for corporate assets. The model is in discrete time, and one period is de?ned as one year. Technology 96
The pro?t function is described by ?( , K ). In particular, the function is ?( , K ) = e K ? , where is the level of the productivity shock and ? is the curvature of the pro?t
function. The ? is assumed to be less than one so that the production function exhibits the decreasing-return-to-scale property.3 This property captures the concept that local resources and local markets are limited and that there is an incentive for local ?rms to expand to another country. The productivity shock ( =
i
+
S)
has two components. The
i,
which is ?rm-speci?c,
captures the ?rm-level shocks that cannot easily be traded or transferred outside of the ?rm, such as patents, know-how, managerial skills, and reputation. The
S
?{
A , B },
which is speci?c to the location where ?rms operate, captures any country-speci?c factors that can e?ect ?rm pro?ts, such as local input prices, proximity to customers, and other institutional environments. Given the structure of the shocks, more productive ?rms (high acquire assets, and less productive ?rms (low
i) i)
are more likely to
are more likely to sell their assets. Firms
S,
will also want to invest in a country where the country shock, be so.
is high and expected to
Firms are not certain about their future productivities.4 For the calibration, I assume that each follows an AR(1) process: =? + ei,t and ei,t ? N (0, ?i ); + eS,t and eS,t ? N (0, ?S ).
i,t S,t
i,t?1 S,t?1
=?
Firm Organization Firms are risk neutral and maximize the expected present value of dividend streams over an in?nite time horizon. There are two types of ?rms: single-country ?rms and multinational ?rms. A single-country ?rm operates an establishment in one country, S ? {A, B }, but
3
This pro?t function is a shorthand version of a large class of production processes. For example,
Cooper and Haltiwanger (2006) show that it can be derived from the production processes that involve more than one type of inputs. It also allows imperfect competitions in the product markets. 4 For the proof, I only assume that the transition matrix governing the dynamic of has the Feller property.
97
has the option of acquiring ?rms in another country and becoming a multinational. At the beginning of each period, single-country ?rms observe the productivity shocks and choose whether to stay local or not. Then, they decide how much capital they are going to buy or sell in that period. A single-country ?rm i in country S that chooses to remain local has a value function de?ned as VSS (X, KS ) = max [dSS + ?E [VS (X , KS )]], (eq1)
{KS }
where dSS = ?( i +
S , KS )
? (KS ? (1 ? ? )KS ) ? ?(KS , KS ),
i.e., dividend = operating pro?t - asset purchase - adjustment cost. In this equation, d stands for dividend. Subscript SS denotes a single-country ?rm in country S that decides to stay in country S . Firm i’s establishment in country S has productivity
i
+
S.
The exogenous state variable is X = { i ,
A , B }.
The value function
has the productivity of the foreign country as an argument even though it does not have an establishment there. This is because the foreign country’s productivity e?ects the option value of becoming a multinational. The prime variables represent the future values, while other variables represent current values. The 0 < ? < 1 is a discount factor. The capital stock depreciates at the exogenous rate 0 < ? < 1. The cost of investment has two components: the direct cost of capital goods and the quadratic adjustment cost. The direct capital expenditure is given by K ? (1 ? ? )K . The
?? )K 2 ) K . The parameter ? re?ects quadratic adjustment cost is ?(K , K ) = ?/2( K ?(1 K
imperfections in the market for real assets, such as the transaction cost of purchasing and liquidating capital, as well as other real costs associated with change in the level of capital stocks, such as the disruption in the production processes. A multinational has the value function: VM (X, KA , KB ) = where dM = max [dM + [VM (X , KA , KB )]], (eq2) +
S , KS )
{KA ,KB }
S ={A,B } (?( i
? (KS ? (1 ? ? )KS ) ? ?(KS , KS )).
Subscript M denotes a multinational ?rm. A multinational ?rm has two establishments, 98
one in country A and one in country B . At the beginning of each period, multinational ?rms observe the productivity shocks, X , and decide how much capital they are going to buy or sell in each country. Merger Process When single-country ?rms decide whether to remain local or to go abroad, they compare the expected net bene?ts of each alternative. Therefore, the value function of a singlecountry ?rm is: VS (X, KS ) = max [VSS (X, KS ), VSM (X, KS )]. (eq3) Subscript S denotes a single-country ?rm. Subscript SM denotes a single-country ?rm in country S that decides to become a multinational in the next period. The value function of the single-country ?rm that chooses to remain local, VSS (X, KS ), is de?ned by (eq1). The single-country ?rm that chooses to acquire production capacity in another country has the value function of: If S = A, then VAM (X, KA ) = max [dAM + ?E [VM (X , KA , f )]], (eq4A)
{KA }
dAM = ?( i +
A , KA )
? (KA ? (1 ? ? )KA ) ? ?(KA , KA ) ? F , or
{KB }
If S = B , then VBM (X, KB ) = max [dBM + ?E [VM (X , f, KB )]], (eq4B) dBM = ?( i +
B , KB )
? (KB ? (1 ? ? )KB ) ? ?(KB , KB ) ? F .
In order to become a multinational, a single-country ?rm has to pay a one-time ?xed cost, F . After paying F at time t, the single-country ?rm will become a multinational at time t+1. The F captures the idea that investing in a foreign country is more di?cult than investing domestically. The new multinational ?rm will start with toe-hold capital f , 0 < f < F , in the foreign country. Therefore, F re?ects the price of the toe-hold capital combined with other costs of international mergers such as costs of due diligence, costs of setting up new headquarters, and fees for foreign consultants. Under these assumptions, FDIs can be thought of as cross-border mergers. This is consistent with the existing evidence that most FDIs are in the form of cross-border M&As.5
5
According to UNCTAD’s FDI database, from 1988-2006, 62% of global FDIs are in the form of cross-
border M&As.
99
[INSERT FIGURE 3.1 HERE] The timeline of ?rm investment and merger decisions is given in Figure 3.1. Before analyzing and calibrating the model, I need to ensure that the dynamic programming problems (eq1) and (eq2) have a solution and that VM (X, KA , KB ), VSS (X, KS ), and VSM (X, KS ) exist. Let C (X × K ) and C (X × K × K ) be the space of all bounded and continuous functions in (X × K ) and (X × K × K ), respectively. Existence Proposition 1: There exists a unique continuous function VM (X, KA , KB ) that solves the dynamic programming problem (eq2). See the Appendix for the proof The proof is a direct application of Blackwell’s su?cient conditions for a contraction mapping. From theorem 9.7 and 9.11 in Stokey, Lucas, and Prescott (1989), VM (X, KA , KB ) is also increasing in all its arguments. The solution VM (X, KA , KB ) produces the policy function KA and KB , which determine a multinational’s optimal level of investment in country A and country B . Proposition 2: There exists a unique continuous function VS (X, KS ) that solves the dynamic programming problem (eq1), the maximization problems (eq3), and (eq4). See the Appendix for the proof The proof of Proposition 2 is more complicated than Proposition 1’s because VS (X, KS ) can be mapped to either VSS (X, KS ) or VSM (X, KS ), depending upon the values of the state variables. The outline of the proof is as follows: (1) From Proposition 1, there exists a unique function VM (X, KA , KB ) in C (X × K × K ) that solves the multinational dynamic programming problem (eq2). (2) Because the maximization problem of the single-country ?rm that chose to become multinational, (eq4) only involves the function VM (X, KA , KB ), there exists a function VSM (X, KS ) in C (X × K ) that solves (eq4). (3) Next, I apply Blackwell’s su?cient conditions for a contraction mapping for the dynamic programming problem of the single-country ?rm that chose to remain local (eq5).
100
Therefore, there exists a unique function VSS (X, KS ) in C (X × K ) that solves (eq 5). (4) Finally, VS (X, KS ) = max [VSS (X, KS ), VSM (X, KS )] exists and is in C (X × K ), because both VSS (X, KS ) and VSM (X, KS ) are in C (X × K ). The solutions VSS (X, KS ) and VSM (X, KS ) also produce the policy function KS , which determines the domestic ?rm’s optimal level of investment in country S . Investment Euler’s Equation From (eq1) to (eq4), I can characterize the optimal level of investment by deriving the ?rst order conditions and applying the envelope theorem. Proposition 3: The optimal levels of investment in country S of multinationals and singlecountry ?rms are governed by the following Euler’s equation: 1 + ?KS ?(KS , KS ) = [?KS ?( i + See the Appendix for the proof At the optimum, ?rms equate the marginal cost and marginal bene?t of investment. Investing an additional unit of capital costs one plus the marginal adjustment costs. The gain from that additional unit of capital consists of the expected present value of the marginal product of capital, the value of capital left from depreciation, and the marginal e?ect that capital has on next period’s adjustment cost. From Euler’s equation, multinationals operate establishments in two locations as if they are two independent ?rms. Gains from entering another country will depend upon the acquirer’s ?rm-speci?c productivity and the location-speci?c productivity of the target country. This proposition shows that the merger gains in this model come from the utilization of an acquirer’s ?rm-speci?c assets and access to goods and factor markets in the target country. Costly External Financing External ?nancing is more costly than internal ?nancing. In particular, when ?rms raise external capital (dividend is less than zero), it has to pay the cost of external ?nance ?(d) = ?0 + ?1 d, where d stands for dividend. This linear speci?cation is frequently seen in the ?nance and macroeconomic literature. For example, Gomes (2001) assumes that ?(d) is 0.08 ? 0.028d and Whited (2006) assumes that ?(d) is 0.04 ? 0.0264d. The function ?(d) can be thought of as the transaction costs of accessing external equity markets, 101
S , KS )
+ (1 ? ? ) ? ?KS ?(KS , KS )].
such as the cost of an IPO, as well as the premium for agency problems or asymmetric information problems associated with external ?nancing, such as the cost of monitoring the ?rms.6 With costly external ?nancing, Proposition 3 does not hold: investment and mergers in one country might depend on productivity shocks of another country. To illustrate the importance of ?nancial constraint, I consider the extreme case in which only internal ?nancing is possible. Proposition 4: If external ?nancing is prohibitively costly (i.e., ?(d) approaches ?), the optimal level of investment of a multinational satis?es the following conditions:
E [?K ?( i +
A A ,KA )+(1?? )??K ?(KA ,KA )]+cov (?K d,? A A 1+?K ?(KA ,KA ) A
˜)
=
˜)
E [?K ?( i +
B
B ,KB )+(1?? )??K ?(KB ,KB )]+cov (?K d,? B B 1+?K ?(KB ,KB ) B
,
˜ = 1+? . where ? is the shadow value of relaxing the ?nancial constraint: dM ? 0 and ? E [1+? ] See the Appendix for the proof This condition implies that ?rms invest in such a way that, at the optimum, the cost/bene?t ratios are equalized across the two countries. The denominator is the marginal cost of investing from the left hand side of Euler’s equation in Proposition 3. The numerator is the marginal bene?t from the right hand side of Euler’s equation, plus the covariance term. The reasoning behind the covariance terms is that ?rms value an establishment that can generate internal cash ?ow when the ?nancial constraint is binding (high ?) more than an establishment generating cash ?ow when ?nancial constraint is less or not binding (low or zero ?).
6
Gomes (2001) proves that the ?rm dynamic optimization with costly external ?nancing ?(d) has a
unique solution in the technical appendix. With the cost of external ?nancing, the ?rm’s decision can be decomposed into two stages: (1) whether or not to incur the cost of external ?nancing and (2) conditioned on decision in (1) how much to invest and whether or not to merge. For example, multinational’s optimization problem becomes:
InternalF inance ExternalF inance VM (X, KA , KB ) = max[VM (.), VM (.)], InternalF inance where VM ( .) = ExternalF inance VM ( .) =
max
{KA ,KB }?{dM ?0}
[dM + ?E [VM (X , KA , KB )]] and
max
{KA ,KB }?{dM
The International Finance Corporation (IFC) is an international financial institution which offers investment, advisory, and asset management services to encourage private sector development in developing countries.
ABSTRACT
Title of dissertation:
ESSAYS IN INTERNATIONAL FINANCE TANAKORN MAKAEW, Doctor of Philosophy, 2010
Dissertation directed by:
Professor VOJISLAV MAKSIMOVIC Department of Finance
The dissertation consists of three essays on international capital ?ows.
In the ?rst essay, titled “Do small ?rms bene?t more from foreign portfolio investment? Evidence from a Natural Experiment,” I test whether an increase in the supply of foreign portfolio capital bene?ts small ?rms by using the Thai government’s unique restriction on capital in?ows as a natural experiment. The Thai government imposed a very stringent capital control on December 19, 2006, and then quickly abandoned it one day later. Although many other studies have been plagued with the di?culty of separating the impact of foreign capital from the impact of other concurrent events, this experiment helps me solve the time-series identi?cation problem. My results suggest that foreign portfolio investment helps large ?rms the most, contrary to existing evidence, which ?nds a bene?t in foreign portfolio investment for small ?rms. I also investigate the importance of other ?rm characteristics correlated with size, which includes a ?rm’s exchange rate exposure, foreign ownership, and political connection.
The next two essays are on the dynamic patterns of international mergers and acquisitions.
In the second essay, I uncover key facts about international M&As by estimating a variety of reduced form models. I ?nd that: (1) Cross-border mergers come in waves that are highly correlated with business cycles. (2) Most mergers occur when both the acquirer and the target economies are booming. (3) Merger booms have both an industry-level component (productivity shocks) and a country-level component (?nancial shocks). (4) Across over one million observations, acquirers tend to be more productive and targets tend to be less productive, compared to their industry peers. These facts are consistent with the neoclassical theory of mergers in which productive ?rms expand overseas to seize new investment opportunities, but not with the widely held views that most crossborder mergers occur when the target economies are in a recession or face a ?nancial crisis.
In the third essay, I construct a dynamic structural model of cross-border mergers and integrate the important facts above into the model. This dynamic structural approach allows me to quantify the e?ects of productivity and ?nancial shocks on M&A decisions. In addition, this approach provides a proper analytical framework for conducting policy experiments. As an example of such analyses, I investigate the impact of President Obama’s proposal on multinational corporation taxation. My simulation results suggest that the foreign operation tax has economically signi?cant e?ects on productive ?rms and can be very distortionary for cross-border mergers.
ESSAYS IN INTERNATIONAL FINANCE by TANAKORN MAKAEW
Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park in partial ful?llment of the requirements for the degree of Doctor of Philosophy 2010
Advisory Committee: Professor Vojislav Maksimovic, Chair Professor Albert ‘Pete’ Kyle Professor Gordon Phillips Professor Lemma Senbet Professor Peter Murrell
ACKNOWLEDGEMENTS
I owe the deepest gratitude to my advisor, Vojislav Maksimovic. This dissertation would not have been possible without him. I am also deeply indebted to my committee members, Albert “Pete’ Kyle, Gordon Phillips, and Lemma Senbet, for their help and guidance throughout my doctoral studies.
I thank Jerry Hoberg and Mark Loewenstein for introducing me to the ?nance faculties. It was a pleasure working for Gurdip Bakshi, Steve Heston, Alex Triantis, and Haluk Unal. My thanks also go to my coauthors, Minwen Li and Juan Contreras.
I thank Peter Murrell for serving on my dissertation committee and Charle Lahaie for data supports. I appreciate valuable job market advices from Ethan Cohen-Cole, Shawn Cole, Michael Faulkender, Dalida Kadyrzhanova, Anna Obizhaeva, N.R. Prabhala, Georgios Skoulakis, and Russ Wermers. I thank Elinda Kiss for her teaching advices.
I thank Mara Faccio, Nandini Gupta, Anton Korinek, Andrew Karolyi, Carmen Reinhart, Antoinette Schoar, and Shangjin Wei, as well as seminar participants at FMA, EFA, and SWFA for comments and suggestions on Chapter 1.
I thank Rui Albuquerque, Sudipto Dasgupta, Antonio Felato, Amar Gande, Jarrad Harford, Pab Jotikastira, Sandy Klasa, Anton Korinek, Yrjo Koskinen, Michael Lemmon, Peter MacKay, Robert Marquez, Darius Miller, Kanda Naknoi, Mark Seascholes, Karin Thorburn, Missaka Warusawitharana, and Shangjin Wei, as well as seminar participants at Boston University, Federal Reserve Board of Governors, HKUST, Norwegian School of Economics, Southern Methodist University, University of Maryland, University of South Carolina, and University of Utah for many helpful suggestions on Chapters 2 and 3.
ii
Contents
1 Do small ?rms bene?t more from foreign portfolio investment? Evidence from a Natural Experiment 1.1 1.2 1.3 1.4 1.5 1.6 1.7 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Related Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Natural Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1 4 8
Data and Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Robustness Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Conclusion and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2 Waves of International Mergers and Acquisitions 2.1 2.2 2.3 2.4 2.5
40
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 Empirical Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 Merger Activities and Macroeconomic Conditions . . . . . . . . . . . . . . . 48 Firm Characteristics and Industry Merger Waves . . . . . . . . . . . . . . . 54
iii
2.6 2.7
Additional Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 Conclusion and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3 A Dynamic Model of International Mergers and Acquisitions 3.1 3.2 3.3 3.4 3.5 3.6
93
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 Conceptual Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 The Model and its Basic Properties . . . . . . . . . . . . . . . . . . . . . . . 96 Solution Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 Simulations and Policy Experiments . . . . . . . . . . . . . . . . . . . . . . 104 Conclusion and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
iv
Chapter 1
Do small ?rms bene?t more from foreign portfolio investment? Evidence from a Natural Experiment
1.1 Introduction
Small ?rms play an important role in emerging market economies since they are often associated with employment generation, economic diversity, balanced income distribution as well as being a source of entrepreneurship, innovation, and economic growth. While it is apparent that foreign portfolio investment has a signi?cant impact on ?rms in emerging markets, it is less clear whether ?rms of di?erent sizes are a?ected by foreign portfolio investment symmetrically. In this paper, I test whether an increase in the supply of foreign portfolio capital bene?ts small ?rms by using the Thai government’s unique restriction on capital in?ows as a natural experiment. I ?nd that foreign portfolio investment helps large ?rms the most, contrary to existing evidence, which ?nds a bene?t in foreign portfolio investment for small ?rms.
1
Existing papers tend to argue that small ?rms bene?t from foreign portfolio investment more than large ?rms (for example, Gelos and Werner 1999, Knill 2005, and Patro and Wald 2005). A number of authors document a positive correlation between foreign portfolio investment and small ?rms’ growth, both in terms of capital accumulation and the ability to access external capital markets. Others study the impacts of ?nancial liberalization, the event that leads to a large increase in foreign portfolio investment. They ?nd that, after ?nancial liberalization, small ?rms have lower investment-cash ?ow sensitivities, face lower cost of capital, and invest more. Another important ?nding is that, during the time of liberalizations, small ?rms, on average, experience higher stock returns compared to large ?rms. However, from the existing literature, it is ambiguous whether small ?rms bene?t from foreign portfolio investment or from other factors correlated with the surge in foreign capital. Foreign portfolio investment is potentially correlated with a number of macroeconomic variables. Stock market and capital account liberalizations are usually concurrent with other major changes such as trade liberalizations, reforms in stock market regulations, and reforms in banking supervisions. I believe that this time-series identi?cation problem is severe since it is virtually impossible to list all the events that a?ect ?rm value. Even if I can identify all the relevant factors, it is still hard to identify the exact time these changes took place (in order to control for them in a panel data study) or the exact time the market learned about them (in order to control for them in an event-study). I analyze the stock market impacts of Thailand’s unique restriction on portfolio capital in?ow. The Thai government imposed a very stringent capital control on December 19, 2006 and then quickly abandoned it on December 20, 2006. The fact that the control only lasted for one day provides an excellent framework for a natural experiment study. It is di?cult to come up with another factor that is unrelated to capital control, has as dominant of an e?ect on ?rms compared to capital control, and changes back and forth overnight like capital control. For example, one might argue the stock return on the capital control day may re?ect both changes in foreign capital and changes in investor’s perception about the Thai government’s ability to run the economy e?ectively. However, it is not likely that this perception was largely reversed in a day at the same time that
2
the government reversed its decision about the capital control. Natural experiments have recently become popular in social sciences, especially in economics. A recent search using the term “natural experiment” on Google Scholar yields more than one million results. The 2002 Nobel Laureate in Economics, Vernon L. Smith, also stated in the Journal of Economic Perspective, “Natural experiments occur all the time and it would be desirable to develop a professional readiness to seize upon these occasions (p.155).” This paper joins a growing literature in ?nance that uses a natural experiment as a solution to the identi?cation problems. I show that large ?rms experienced more negative abnormal returns on the capital control day (December 19) and more positive abnormal returns on the liberalization day (December 20), suggesting that foreign portfolio investment by itself bene?ts large ?rms the most. Compared to small ?rms, large ?rms have higher fractions of foreign ownership and are more likely to have political and business connections. In order to examine how much of the size e?ects are due to the di?erence in other ?rm characteristics correlated with size, I control for (1) ?rm ?nancial characteristics (pro?tability, investment opportunities, leverage, accounting liquidity, and industry dummy), (2) ?rm international involvements (exchange rate exposure, foreign ownership, and foreign control), and (3) ?rm connections (both political and business connections). I ?nd that size still has a large and signi?cant explanatory power after including these variables in our regressions. In the full speci?cation, ?rms that are one standard deviation larger earn 83.35 basis points less on the capital control day and 84.78 basis points more on the liberalization day. I further show that size is correlated with visibility to foreign investors and past capital market activities - large ?rms tend to be included in key stock market indices, to be rated by rating agencies, and to have issued securities in international markets. In addition to my ?ndings about size, I ?nd that ?rms with higher pro?tability, exportoriented ?rms, and ?rms with foreign directors are less a?ected by the capital control. Most interestingly, I ?nd that the stock prices of ?rms connected to the former Prime Minister Thaksin Shinawatra, the major opponent of the incumbent coup government, reacted more strongly to the capital control and the subsequent liberalization. Consistent with Johnson and Mitton (2003) which views Malaysian capital control as a way to support
3
?rms connected to the incumbent government, here investors view Thai capital control as a way to punish ?rms connected to the opponent of the government in power. One might be concerned about the ?ndings that large ?rms are a?ected by the capital control more are due to the market microstructure e?ects, not changes in ?rms’ fundamental value. In other words, since small ?rms are less liquid and small ?rms’ stocks are more closely held, stock prices of large ?rms might be more sensitive to any negative news that has an impact on the macroeconomy. To address this concern, I use the day the market learned about the September 2006 coup as a placebo test. I regress abnormal returns on the coup date on ?rm size together with other ?rm characteristics and ?nd that while the market return was negative on that day, small ?rm returns were signi?cantly more negative, ruling out the claim that large ?rms are more sensitive to any bad news. My results are robust to various econometric speci?cations and variable de?nitions. To correct the potential problems from the non-normality in error terms such as heteroscedasticity and cross-sectional correlations, I (1) use Huber/White/Sandwich standard errors, (2) cluster standard errors at ?rm-level, and (3) cluster standard errors at industry-level. I also (4) compute the empirical standard errors from bootstrapping and (5) compute the empirical standard errors from historical data. Finally, I use alternative de?nitions of size, industry classi?cation, pro?tability, and liquidity and use raw returns instead of abnormal returns. All of the results I ?nd are qualitatively the same. The paper proceeds as follows. Section 2 summarizes related literature. Section 3 describes the natural experiment. Section 4 outlines the empirical strategy and provides the data sources. Section 5 estimates the e?ects that ?rm size and other control variables have on the bene?ts from foreign portfolio investment and analyzes the results. Section 6 performs robustness checks. Section 7 concludes.
1.2
Related Literature
Foreign portfolio investment has become an important part of international capital ?ow. According to Bosworth et al.(1999), the composition of capital ?ow has shifted away from
4
foreign direct investment and bank loans to portfolio investment; the fraction of foreign portfolio investment in emerging markets has increased from 9% in 1978-1981 to 44% in the 1990’s. Consequently, costs and bene?ts of foreign portfolio investment are usually at the heart of any ?erce debates on ?nancial globalization. Bene?ts of Foreign Portfolio Investment Foreign portfolio investment is believed by many to have large potential bene?ts. The in?ow of foreign fund increases the supply of capital in a domestic economy. With more capital, ?rms can expand their existing capacities and undertake more projects. From the ?nancial markets perspective, foreign portfolio investment increases market liquidity and hence improves asset-pricing e?ciency (Levine and Zervos 1998). Li et al. (2006) provide evidence that capital account liberalization lowers the co-movement and raises idiosyncratic variation in stock prices, suggesting that stock prices contain more ?rmspeci?c information and the stock market becomes more e?cient. Some policy experts (for example, Evan 2003) additionally argue that foreign portfolio investors have superior technologies to value ?rms compared to domestic investors. Therefore, they create informational externalities that help domestic investors identify the best place to invest. The Importance of Small Firms Small ?rms have long been a center of attention in academia and policy circles. In the Federal Reserve’s Economic Quarterly, Weinberg (1994) stated, “It seems that a necessary part of the debate over any proposed public policy action, from healthcare to tax policy, is the question of how it will a?ect small ?rms (p.1).” Internationally, the World Bank has approved more than $10 billion support to the small and medium business enterprises during 1998-2002 (Beck et al. 2005). The attention that small ?rms received comes as no surprise since the growth of small ?rms and the growth of large ?rms are perceived to have very di?erent impacts on an economy. Small ?rms are often associated with employment generation, economic diversity, balanced income distribution, and being a source of entrepreneurship, innovation, and economic growth. While small ?rms are looked fondly upon, large ?rms are often associated with entrenchment and economic ine?ciencies. For example, Fogel, Morck, and Yeung (2005) ?nd that in countries whose large ?rms are doing well, overall economic growth, productivity growth and capital accumulation is lower.
5
They interpret this evidence as a support of Schumpeter’s theory of creative destruction, in which growth comes from small creative ?rms destroying large old ?rms. Small ?rms in emerging markets are generally considered the ones that su?er more from informational problems (Kang and Stulz 1997; Dahlquist and Robertsson 2001; and others) since large ?rms tend to be better known, older and have a longer track record of past performance. Analysts and the media also tend to cover large ?rms more frequently, making it harder for executives of large ?rms to hide mistakes or overstate pro?ts. Given that small ?rms su?er more from informational problems, I could easily deduce from the classical theories of corporate ?nance that small ?rms will be more ?nancially constrained. For example, small ?rms will face more credit rationing according to Stiglitz and Weiss (1981) and will face a higher cost of equity according to Jensen (1976) and Jensen and Meckling (1986). Small Firms and the Bene?ts from Foreign Portfolio Investment While foreign portfolio investment mechanically increases the aggregate supply of capital and hence should bene?t all ?rms in the domestic economy, it remains an empirical question whether or not small ?rms bene?t more than large ?rms. Theoretically, if foreign portfolio investment does help alleviate asymmetric information and agency problems for all ?rms, then small ?rms should bene?t more since they are the ones who su?er more from these problems and starve for capital in the ?rst place (see Section 1 in Evan (2003) and Section 4 in Forbes (2005) for the detailed arguments how foreign capital might solve informational problems). Even though it is well-documented empirically that foreign portfolio investors prefer to invest in larger ?rms (Kang and Stulz 1997; Dahlquist and Robertsson 2001; and others), many researchers in corporate ?nance note that small ?rms do not have to be the direct recipients of foreign portfolio investment in order to bene?t more. In one example, Knill (2005) suggests the monetary transmission mechanism through bank loans (Bernanke and Blinder 1988; Kashyap and Stein 1995; Kashyap, Rajan and Stein 2002); small ?rms and large ?rms are competing for the same pool of bank loans. When additional supply of capital ?ows to large ?rms, small ?rms receive more bank loans since large ?rms have less demand for loans. In another example, Gallego and Hernandez (2003) suggest the
6
trade credit channel; small ?rms and large ?rms are competing for the same pool of trade credits. When foreign portfolio investment ?ows to large ?rms, small ?rm receive more trade credit since large ?rms demand less trade credit. Gallego and Hernandez give the 1998 ?nancial market turmoil in Chile as an anecdotal example: “When interest rates in Chile (and in other emerging market economies) reached extremely high levels. During this period a group of large ?rms arbitrarily extended the payment period to suppliers from 90 to 180 days, forcing smaller ?rms to assume the increase in the cost of funds (p.17).” Existing empirical evidence strongly supports the hypothesis that small ?rms, not large ?rms, bene?t more. These works come in a variety of forms, including panel data studies, event studies, single-country studies, and cross-country studies. Examples that represent each genre of work are summarized in the appendix. The ?rst example is a cross-country panel data study by Knill (2005). She studies a panel of ?rms from 53 countries during 1996 to 2005 and ?nds that foreign portfolio investment is associated with an increase in the ability to issue securities for small ?rms. Additionally, she ?nds that foreign portfolio investment increases the maturity of bank loans, leading her to conclude that small ?rms bene?t more because they rely more on bank loans than large ?rms. The next group of papers (Harris, Schiantarelli, and Siregar 1994; Jaramillo, Schiantarelli, and Weiss 1996; Gelos and Werner 2002; Laeven 2003; Koo and Shin 2004; Contreras and Makaew 2007) analyzes ?rm behaviors before and after ?nancial liberalization, the event that leads to a large change in foreign investment. Most of these papers ?nd that ?nancial liberalization has di?erent impacts on small and large ?rms: compared to large ?rms, it a?ects small ?rms by further relaxing ?nancial constraints, lowering investment-cash ?ow sensitivity, increasing investment, and lowering purchasing price of capital. The last example, which is closest to my work, is Patro and Wald (2005). They study the impacts of stock market liberalization on small and large ?rms by extending the event study framework of Henry (2000) to a cross-sectional event study. Using the stock market data from 18 developing countries, they ?nd that small ?rms earn signi?cantly higher abnormal returns when stock markets are liberalized. 7
Even though the amount of the existing evidence supporting the “small-?rms-bene?tmore” hypothesis is overwhelming, it is not clear whether small ?rms bene?t from foreign portfolio investment or from other factors correlated with the surge in foreign capital. Stock market and capital account liberalizations are often concurrent with (1) banking deregulations (reduction in reserve requirements and credit controls; privatizations of state banks; allowing foreign bank entries), (2) reforms in stock market regulations and banking supervisions, and (3) trade liberalization. For instance, the Korean stock market liberalization was concurrent with interest rate deregulations and a strengthening of prudential regulations. The Colombian capital account liberalization was also concurrent with constitutional reforms and banking deregulations. I believe that identi?cation of the bene?ts from foreign portfolio investment is di?cult, given that countries’ economic prospects are changing rapidly along with their capital account policies. It is virtually impossible to list all the events that a?ect the value of small ?rms. Even if I could identify all the relevant factors, the liberalization process is still complicated and dynamic by nature. It is di?cult to identify the exact date these changes took place (in order to control for them in a panel data study) or identify the exact time the market learned about them (in order to control for them in an event study). In this paper, I propose a cross-sectional analysis of a unique event in Thailand. The Thai government imposed a draconian capital control on December 19, 2006 and then quickly abandoned it on December 20, 2006. This experiment-like event helps me separate the impact of foreign capital from the impact of other concurrent events.
1.3
Natural Experiment
In this section, I describe the capital control and liberalization which is the event of interest but I will ?rst discuss the political and economic situations in Thailand that lead up to the event. Thaksin Shinawatra and the September 2006 Coup From January 2001 to September 2006, Thailand was under the administration of Prime
8
Minister Thaksin Shinawatra who was also a successful businessperson and one of the richest people in the country. He and his family were major shareholders of many listed ?rms in the Stock Exchange of Thailand including Advanced Info Service - the largest mobile phone operators in Thailand, Shin Satellite - the only operator of Thailand’s commercial satellites, and ITV - a television station. Even though Prime Minister Thaksin swept the elections in 2001 and 2005, his popularity started to decline in late 2005 when he was accused of fraud, human rights o?enses and lese-majeste. On September 19, 2006, the Thai Military staged a coup against Prime Minister Thaksin and overthrew his government while he was attending the United Nation Assembly in New York. The new Prime Minister, as well as the new cabinet, and the new governor of the Thai central bank were appointed in October 2006 and November 2006, respectively. Prime Minister Thaksin is currently in exile. The One-Day Capital Control On December 19, 2006, the Thai central bank had decided to implement a reserve requirement on short-term capital in?ows. Under this new regime, foreigners bringing portfolio capital into Thailand had to deposit 30% of the funds into an account at the central bank which would earn no interest. This meant that only 70% of the funds would be available for investment in the Thai market. Moreover, if foreign investors wished to withdraw their money within one year, they would be ?ned 1/3 of the amount. The control was targeted straight at future portfolio capital in?ows. The central bank stated clearly that foreign direct investment was not subjected to this reserve requirement.1 Any foreign exchange transactions which had been traded before the announcement were also exempted. As anyone would expect, the Thai stock market reacted to this surprising news immediately; the Stock Exchange of Thailand (SET) Index dropped from 730.55 to 622.14 (i.e. 14.84% reduction in one day). On December 20, 2006, the central bank announced that in?ows for the investment in the Stock Exchange of Thailand, the Thai Market for Alternative Investment, the Thai Futures Exchange, and the Agricultural Futures Exchange of Thailand, which are basically most of
1
Initially, foreign direct investment was also required to place 30% of the in?ow as a reserve requirement,
but after submitting the relevant documents to support the claim of legitimacy, the central bank would refund the reserve amount.
9
the portfolio ?ows into Thailand, were no longer subjected to the 30% reserve requirement. Again, this announcement took the market by surprise and the SET Index bounced back from 622.14 to 691.55 (or 11.16% increase in one day).
[INSERT FIGURE 1.1 HERE]
The fact that the capital control restriction only lasted for one day provides a proper framework for a natural experiment study. It is di?cult to come up with another factor that is unrelated to capital control, has as dominant of an e?ect on ?rms compared to capital control, and changes back and forth overnight like capital control. For example, one might argue validly that the stock return on the capital control day may re?ect both changes in foreign capital and changes in investor’s perception about the Thai government’s ability to run the economy e?ectively. However, it is not likely that this perception was largely reversed back to normal when the government reversed its decision about the capital control the next day.
1.4
Data and Methodology
In this section, I describe the sample ?rms, the variables used and how they are constructed. The dataset consists of all Thai ?rms listed in the Stock Exchange of Thailand. All of the trading data are from the Reuters Database. All of the ?nancial statement data are from Reuters and the Stock Exchange of Thailand Market Analysis and Reporting Tool (SETSMART) Database. Daily stock prices are the last reported trade prices. Other ?rm characteristics are measured at the end of 2005 since most ?rms in Thailand report their ?nancial status at the end of December and ?rm characteristics measured at the end of 2006 might be contaminated by the e?ects of the experiment already. The details how each ?rm characteristic is constructed are in the appendix. The summary statistics are provided in Table 1.1A. I compute abnormal returns using the market model as the benchmark: Ri = ?i + ?i RM + 10
where RM is the percentage change in the MSCI Emerging Markets Asia Index. The market model is estimated by the daily returns from September 29, 2005 to August 31, 2006 (a 261-trading-day period). Abnormal returns, ARi , are de?ned as: ˆi RM ) ARi = Ri ? (? ˆi + ? ˆi are stock i’s estimated market model coe?cients. I also exclude all the where ? ˆi and ? ?rms that are not traded on the capital control day or the liberalization day. My empirical strategy is to link the abnormal returns on the capital control day and on the liberalization day with ?rm sizes and other control variables. Firms that experienced larger reduction in value on the capital control day and ?rms that experienced larger gain on the liberalization day should be the ones that bene?t more from foreign portfolio investment. [INSERT TABLE 1.1 HERE]
1.5
Analysis
Table 1.1B presents a univariate comparison of the abnormal returns and ?rm characteristics across four size quartiles. On average, ?rms in the largest quartile earn 2.94% less on the capital control day and 3.27% more on the liberalization day, compared to ?rms in the smallest quartile. The di?erences are statistically signi?cant at a 95% con?dence level. This suggests that large ?rms are a?ected by foreign portfolio investment more than small ?rms. Table 1.1B also suggests that there are systematic di?erences in the characteristics of small and large ?rms. Therefore, multiple regression analysis will be performed in the next section, but for now I use the propensity score method as a preliminary analysis to get a better feel of the data. The propensity score method matches treated ?rms with control ?rms that have the nearest propensity scores. I assign the largest size quartile as the treatment group and the smallest quartile as the control group. Propensity scores are computed from the probit model predicting the probability of being in the treatment
11
group using four covariates: pro?tability, exchange rate exposure, foreign director, and Thaksin connection. (These four variables are chosen because in the next section, I ?nd that they are indeed the most relevant variables.) Table 1.1C compares the abnormal returns of ?rms in the largest quartile to the abnormal returns of propensity-score matched ?rms from the smallest quartile (average treatment e?ect on the treated). After matching, the di?erences in abnormal returns between small and large ?rms change slightly; the magnitude of the t-statistic drops from 2.78 to 2.18 on the capital control day and from 3.25 to 2.72 on the liberalization day. The E?ects of Firm Size and Other Financial Characteristics In this section, I analyze the e?ects of ?rm size on the bene?t from foreign portfolio investment by regressing the abnormal return on size and other ?rm characteristics. I use the least square method with robust standard errors. Firm Size is measured by log of total asset. Besides ?rm size, other ?nancial characteristics might also determine how ?rms will be a?ected when the supply of capital decreases so I have to add these variables to the regressions. The ?rst set of control variables I include are the 9 Industry Dummies classi?ed by the Stock Exchange of Thailand. These industry dummies capture any industry-level changes in goods and capital market conditions. They also ?lter out any e?ects of the cross-sectional co-movement in stock returns that are driven by industry-level factors. Next, I include accounting pro?tability, market-to-book ratio and cash ?ow growth to capture ?rm investment opportunities. Pro?tability is measured as pre-tax pro?t or loss scaled by lagged total asset. Market-to-Book is measured as market value of equity divided by book value of equity. Cash Flow Growth is measured as lagged annual growth in operating cash ?ow. Theoretically, ?rms with better investment opportunities should have higher demand for fund and hence should be a?ected by the capital control more. Unlike market-to-book that captures investment opportunities in the future, accounting pro?tability and cash ?ow growth measured in the previous year are also proxies for ?rm ability to generate internal fund in the short-run. Firms that can generate more internal fund, and hence rely on external ?nancing less, should be a?ected by the capital control less. Finally, I include accounting liquidity and leverage. Liquidity is measured as net working capital (current asset minus current liability) scaled by lagged
12
total asset. Leverage is measured as total debt scaled by lagged total asset. When the supply of capital decreases, ?rms that have less current asset compared to current debt are more likely to face a liquidity problem. At the same time, ?rms that have higher leverage might have trouble paying interests and are more likely to be bankrupt. Therefore, ?rms with lower liquidity and higher leverage should be a?ected by the capital control more. [INSERT TABLE 1.2 HERE] Results Table 1.2A provides correlations between ?rm size and other characteristics. From the size column, I can see that size is not strongly correlated with any other variables. Large ?rms tend to be slightly less liquid and have higher leverage. No variables are strongly correlated with one another; the highest correlation of 24% is between liquidity and pro?tability. Therefore, multi-colinearity should not be a problem in my analysis. The regression results are reported in Table 1.2B. In Panel A, the dependent variable is the abnormal return on the capital control day. In Panel B, the dependent variable is the abnormal return on the liberalization day. I ?nd that the coe?cients on size are statistically signi?cant at a 95% or 99% con?dence levels in all speci?cations. The coe?cients on size are uniformly negative on the capital control day and uniformly positive on the liberalization day. This means that large ?rms lose more when a capital control is imposed and bene?t more when a capital control is lifted. In other words, large ?rms bene?t from foreign portfolio investment more. The estimated size coe?cients indicate that the size e?ect is economically large. From the full speci?cation (Model 7), one standard deviation increase in size leads to 0.97% reduction in ?rm value on the capital control day and 1.13% increase on the liberalization day. Other results suggest that ?rms with higher pro?tability bene?t less from foreign portfolio investment. This is consistent with the hypothesis that ?rm with higher pro?tability are less sensitive to changes in external capital markets. The economic signi?cance of pro?tability is comparable to size’s. From the full speci?cation, one standard deviation increase in pro?tability leads to 1.27% increase in ?rm value on the capital control day and 1.05% decrease in ?rm value on the liberalization day. In panel A, market-to-book also 13
has statistically negative coe?cients, suggesting that ?rms that have better investment opportunities are a?ected more from the capital control. The coe?cients on cash ?ow growth, liquidity and leverage are not statistically signi?cant and these variables do not increase the statistical ?t of my models. Firm Size and International Involvement In this section, I examine the e?ects of ?rm’s international involvement. In particular, I analyze the relationship between ?rm size, exchange rate exposure, foreign ownership, and foreign control. I then include these variables in the regressions in order to examine how much of the size e?ects found in the previous section are due to the di?erent degrees of international involvement between small and large ?rms. A ?rm’s Exchange Rate Exposure is proxied by exchange rate beta calculated from a factor model (see details in the appendix). Firms that have a positive exchange rate beta are likely to be ?rms that earn income in US dollars and have expenditure in the local currency (Thai Bahts) such as export-oriented ?rms and ?rms that own income-generating assets abroad. The high-beta ?rms should su?er less or even pro?t from the capital control. Since a control on capital in?ow automatically reduces the demand for Thai Bahts relative to US dollars, these ?rms’ cash ?ow in Thai Bahts will increase as a result of exchange rate depreciation. The next variable is Foreign Ownership which is measured as the fraction of ?rm’s equity owned by non-Thai citizens. Firms with a higher foreign ownership fraction tend to rely on foreign capital more and hence should be a?ected by the capital control more. The last control variable is the Foreign Director Dummy. This dummy takes the value of one if a ?rm has at least one non-Thai citizen as a director and zero otherwise. I have to include Foreign Director dummy because foreign control and foreign ownership sometimes do not go hand in hand as foreign ownership might be di?used. Theoretically, ?rms that have a foreign director should be a?ected by the capital control less since (1) foreign directors might provide better corporate governance. Firms with better corporate governance are able to attract more external capital when needed and hence are a?ected by the capital control less. (2) The existence of a foreign director might re?ect the fact that foreign investment in that ?rm is a non-di?used direct investment rather than portfolio investment and foreign direct investment is exempt from the December 19 capital control
14
in the ?rst place. [INSERT TABLE 1.3 HERE] Results Table 1.3A reports correlations between ?rm size and international involvement variables. Size is strongly and positively correlated with foreign ownership with the correlation coe?cient of 40.5%. This is consistent with the ?ndings of Kang and Stulz (1997) as well as Dahlquist and Robertson (2001) that foreign institutional investors tend to invest in larger ?rms. Size is also negatively correlated with exchange rate exposure which is partially due to the industry e?ects; ?rms in export-oriented industries (such as textile and food processing) tend to be smaller than ?rms domestic-oriented industries (such as real estates and telecommunication). As expected, ?rms that have higher foreign ownership fractions are more likely to have foreign directors. Therefore, foreign director dummy is strongly correlated with foreign ownership (51.24%). However, the correlation between size and foreign director is much weaker, only 9.64%. The regression results are reported in Table 3B. In Panel A, the dependent variable is the abnormal return on the capital control day. In Panel B, the dependent variable is the abnormal return on the liberalization day. I ?nd that the coe?cients on exchange rate exposure are signi?cant at a 95% or 99% con?dence level. As anticipated, these coe?cients are positive on the capital control day and negative on the liberalization day. It is likely that this ?nding re?ects the fact that ?rms with income in foreign currencies should be a?ected by the capital control less. The economic signi?cance of exchange rate exposure is large; from Model 2, one standard deviation increase in exchange rate beta leads to 1.45% increase in ?rm value on the capital control day and 1.16% decline on the liberalization day. From Model 3, I ?nd that the coe?cients on foreign ownership fraction are not statistically signi?cant. One of the plausible explanations is that ?rm size is a better proxy for the bene?t from future foreign investment, compared to foreign ownership fraction which re?ects past investment. The insigni?cance of foreign ownership also rules out another alternative explanation for the size e?ect: one might claim that the size e?ect found the 15
previous section is simply a result of foreign investors getting panic and liquidating their positions (which are mostly large ?rms) on the capital control day. If this explanation were valid, the foreign ownership variable would have driven out the signi?cance of ?rm size. From Model 4, I ?nd that ?rms that have a foreign director are a?ected less by the capital control. On average, ?rms with foreign directors earn 1.16% more on the day of the capital control and 2.19% less on the liberalization day. The e?ect of ?rm size is still large and signi?cant even after controlling for a ?rm’s international involvement. Comparing before and after including the control variables, on the capital control day, the magnitude of the size coe?cients drops slightly from -0.74 in the baseline model (Model 1) to -0.60 in the full speci?cation (Model 4) but remains statistically signi?cant at a 95% level. Similarly, on the liberalization day, the magnitude drops from 0.70 to 0.55 but remains statistically signi?cant at a 95% level. The ?ndings that size and foreign ownership is positively correlated and that size might be better as a proxy for the bene?t from future foreign investment are interesting. Therefore, I further investigate the relationship between ?rm size and other capital market activities. Table 3C reports the ?rms’ activities classi?ed into four size quartiles. From the ?rst three columns, I ?nd that large ?rms are more visible to foreign investors in the sense that most ?rms that are constituents of key national indices and have credit ratings are from the largest quartile. I also ?nd that large ?rms are more likely to engage in international capital market activities: In the secondary market, 9 out of 12 ?rms in the sample that are cross-listed or have over-the-counter ADRs are from the largest quartile. In the primary market, 75% of ?rms that have issued equities or debts in international capital market (from 1990 to 2006) are from the largest quartile. In short, I con?rm that size is highly correlated with international capital market activities in general. Firm Size and Political / Family Connections In this section, I analyze the relationship between ?rm size and its political and family business group connections. A number of studies have documented that (1) large ?rms are more likely to have political connections (Faccio 2006), (2) drastic government policies tend to a?ect connected ?rms and unconnected ?rms di?erently (Johnson and Mitton 16
2003; Faccio, McConnell and Masulis 2006; and others), and (3) a ?rm’s international capital market activities are in?uenced by its political connections (Leuz and OberholzerGee 2006). The event of interest, the capital control, was imposed during the tenure of the coup government, shortly after throwing out Prime Minister Thaksin. Therefore, it is important that I include these political and family business group connections in the regressions in order to examine how much of the size e?ect is due to the di?erent degrees of connectedness. The ?rst variable captures the direct political connection to the former Prime Minister Thaksin who is the rival of the coup government. Thaksin Connection is a dummy variable that takes the value of one if the ?rm’s major shareholder is, or is blood-related to, a member of Thaksin’s Cabinet and zero otherwise. Firms that are politically connected to Prime Minister Thaksin should su?er from the capital control more since these ?rms are more likely to have di?culties obtaining funds from domestic ?nancial institutions when the coup government is in power. Next, I include the family business group variable. In Thailand, like in many other East Asian countries, business groups consist of ?rms whose major shareholders are relatives. Family Business Group dummy takes the value of one if the ?rm’s major shareholder is from the 50 largest family business groups in Thailand and zero otherwise. Theoretically, ?rms in large business group can internalize many capital allocation functions and hence rely less on external capital markets. Therefore, ?rms that belong to a business group should su?er less from the capital control. Finally, I also allow for a more general de?nition of political connections. Faccio (2006) de?nes politically connected ?rms broadly as ?rms that (1) have a major shareholder or a top executive who is a parliament member, minister or head of the state, or (2) have a major shareholder or a top executive who is related to a top politician or a political party. Political Connection is a dummy variable that takes the value of one if the ?rm is politically connected according to Faccio (2006) and zero otherwise. I note that Political Connection captures political connections to anyone in Thai politics and Thaksin Connection should be a subset of Political Connection. [INSERT TABLE 1.4 HERE]
17
Results Table 1.4A reports correlations between ?rm size and the connection variables. I con?rm that ?rms that are connected to Thaksin, ?rms that are connected to a top family business group and ?rms with any political connections tend to be larger. The correlation between ?rm size and the political connection dummy is 30.19% suggesting that large ?rms tend to have some sorts of connections, either with Thaksin or with other people in high o?ces. The correlation between Thaksin connection and family business group connection is very strong (42.31%) suggesting that many of the top family ?rms in Thailand have someone representing them in the Thaksin administration. The regression results are reported in Table 4B. As usual, the dependent variable in Panel A is the abnormal return on the capital control day. The dependent variable in Panel B is the abnormal return on the liberalization day. From Panel B Model 1, I ?nd that the coe?cient on Thaksin Connection is signi?cant while all other connection coe?cients are not. (Since only 5% of ?rms in the sample are directly connected to Thaksin and the connected ?rms are concentrated in a few industries, it is natural that the level of signi?cance is not high.) This supports the belief that political connection does a?ect ?rm value. In fact, my ?nding here is closely related to the ?ndings of Johnson and Mitton (2003). They view the 1998 capital control in Malaysia as a way to support ?rms connected to the incumbent government. Here the market views Thai capital control as a way to punish ?rms that are connected to the opponent of the people in power. After taking a closer look at the ?rms connected with Thaksin, I ?nd that these ?rms are also the ones that heavily engage in international capital market activities. For example, 7 out of 12 ?rms in the sample that have ADRs are connected to Thaksin and approximately half of ?rms that are connected to Thaksin have issued equities in international capital markets. I ?nd that the e?ect of ?rm size is still large and signi?cant after controlling for the connection e?ects. On the capital control day, the magnitude of the size coe?cients drops from -0.74 in the baseline model to -0.53 in the model with Thaksin Connection dummy but remains statistically signi?cant at a 95% level. Similarly, on the liberalization day, the coe?cient drops from 0.70 to 0.54 but remains statistically signi?cant at a 95% level.
18
1.6
Robustness Tests
Are large ?rms more susceptible to any bad news? One might argue large ?rms are more susceptible to any negative national news compared to small ?rms, possibly because small ?rms are less liquid and small ?rms’ stocks are more closely held. My ?rst response to this argument is that I have excluded all the stocks that are not traded on the capital control day or the liberalization day. Conditioned on being traded, simple microstructure models predict that less-liquid ?rms should su?er more on a bad day since the prices of illiquid stocks have to decline more in order to induce trade. My second response is to use the day the market learn about the September 2006 coup in Thailand as a placebo test. The coup is a good candidate for a placebo as it is a bad news that a?ects the entire economy within a short period of time before the capital control. If the hypothesis that large ?rms are more susceptible to any bad macro news is true, I should ?nd that, after controlling for other factors, large ?rms were a?ected more. [INSERT TABLE 1.5 HERE] Results Table 1.5 reports the placebo regression results. The dependent variable is the abnormal return on September 23, 2006 which is the ?rst trading day that the market learned about the coup. In all speci?cations, I ?nd that the coe?cient on size is positive and signi?cant at a 99% con?dence level. In other words, large ?rms are a?ected by the coup less. This is consistent with the ?ight to quality hypothesis: when bad things happen, large ?rms become more attractive relative to small ?rms. More importantly, this ?nding rules out the alternative explanation that large ?rms are more sensitive to any bad news. I also note that the coe?cient of Thaksin Connection is negative and signi?cant. From Model 6, ?rms that are connected to Thaksin earn 3.23% less than other ?rms when the Thaksin government lost power. I take this as a con?rmation that my measure of connections with Prime Minister Thaksin is legitimate. Econometric Issues I use a number of techniques to address econometric concerns regarding the non-normality 19
of the error terms. First, I note that in all of The previous regressions the t-statistics are computed from Huber/White/sandwich standard errors so they are already robust to the heteroscedasticity problem. Second, a more serious problem here is the cross-sectional correlation problem because the event dates are the same across all ?rms in my analysis. My solution is to cluster standard errors both at the ?rm-level and the industry-level to eliminate the biases from stock return co-movements. I ?nd that the signi?cance of the size coe?cient barely changes when clustered at ?rm-level and even improves when clustered at the industry-level. Bootstrapping Third, I address any non-normality problems in the error terms by using empirical standard errors. I use two methods to generate an empirical distribution. In the ?rst one, I create 1,000 synthetic samples from the original dataset and estimate the full speci?cation using these samples. I then compute the z-scores of the size coe?cient using the empirical standard errors from these 1,000 synthetic betas. Once again, I obtain a similar result the size coe?cient is still statistically signi?cant at a 95% con?dence level. In the second method, I use the historical data to generate the empirical distribution. I perform the cross-sectional regression everyday from January to September 2006. I then compute the z-score of the size coe?cient using the standard errors computed from these daily beta estimates. This time, the empirical distribution computed from historical data yields a very high signi?cance level because size is consistently a poor predictor of daily abnormal returns. The mean of the size coe?cient on a regular day is -0.0005 (compared to -0.53 and 0.54 on the event days). The empirical standard deviation is 0.0227. Therefore, the z-value of the size coe?cient becomes 23.28 which is statistically signi?cant at a 99.99% con?dence level. Alternative De?nitions I con?rm that my results are robust to alternative variable de?nitions. I use log of market capitalization and log of market capitalization adjusted for free-?oat as alternative proxies for size, GIC industry classi?cation for industry dummies, sales for pro?tability, and cash holdings for liquidity. I also use raw returns instead of abnormal returns. All the results are qualitatively similar to what I found earlier. 20
1.7
Conclusion and Discussion
In this paper, I examine the e?ect foreign portfolio investment has on ?rms of di?erent sizes using Thailand’s unique restriction on short-term capital in?ow as a natural experiment. The fact that this restriction was a surprise and only lasted one day makes it an appropriate set-up for an event-study. In contrast to the majority of existing literature, my evidence suggests that large ?rms bene?t from foreign portfolio investment more; I ?nd that large ?rms stock market valuations were hurt by the capital control and helped by the subsequent liberalization. Compared to small ?rms, large ?rms have a higher fraction of foreign ownership and are more likely to have political and business connections. After controlling for ?rm ?nancial characteristics, international involvements and connections, size still has a large and signi?cant explanatory power. Other results suggest that ?rms with higher accounting pro?tability, export-oriented ?rms, and ?rms with foreign directors are less a?ected by the capital control while ?rms with higher market-to-book and ?rms connected to Prime Minister Thaksin, the major opponent of the incumbent coup government, were a?ected more. My results are robust to various econometric speci?cations and variable de?nitions. This paper contributes to the international ?nance literature in several ways. First, I provide a clean framework to test the size e?ects and, from this, I ?nd that the results from the existing literature are reversed. Second, by including a wide range of control variables, I identify that certain types of ?rms are a?ected more when the supply of foreign fund declines. Third, I con?rm that, in emerging markets, political connection does a?ect ?rm value. Firms connected to Prime Minister Thaksin su?ered more when the market learned about the coup and when the capital control was announced. The main implication of my study is not that foreign portfolio investment does not bene?t small ?rms. Rather what I ?nd is that, after separating other events or factors that are typically correlated with foreign portfolio investment, foreign portfolio investment by itself does not favor small ?rms. Therefore, my results, taken together with the existing literature, imply that the “other factors” concurrent with stock market and capital account liberalizations are very important. Whatever governments were doing at the time of
21
liberalizations - deregulating the banking systems, improving capital market supervisions, liberalizing international trade etc. - is probably good for small ?rms. In sum, my study calls for more research on how domestic reforms could channel funds to small ?rms who need it the most, and how domestic market conditions interact with foreign portfolio investment.
22
Table 1.1A: Summary Statistics Quartile Variable Firm Size Pro?tability Market-to-Book Cash Flow Growth Liquidity Leverage Exchange Rate Exposure Foreign Ownership Foreign Director Thaksin Connection Family Business Group Political Connection Mean 22.2 2.07 2.59 -58.03 19.33 51.23 -0.63 19.32 0.34 0.05 0.10 0.08 SD 1.57 3.66 21.89 576.8 29.38 28.38 0.66 21.22 0.47 0.22 0.29 0.26 Min 19.46 -15.45 -5.03 -3567.23 -51.15 0 -2.81 0 0 0 0 0 0.25 21.1 0.42 0.65 -107.74 1 29.93 -1.07 1.45 0 0 0 0 Median 21.88 1.85 1.02 -21.77 14.97 50.67 -0.58 12.19 0 0 0 0 0.75 22.97 3.43 1.53 41.98 36.51 70.68 -0.13 30.6 1 0 0 0 Max 27.97 17.01 409.18 5304.19 161.81 150.39 0.75 95.56 1 1 1 1
Firm size is measured as the log of total asset; Pro?tability is measured as pre-tax pro?t or loss scaled by lagged total asset; Market-to-Book is measured as market value of equity divided by book value of equity; Cash Flow Growth is measured as lagged annual growth in operating cash ?ow; Liquidity is measured as net working capital (current asset - current liability) scaled by lagged total asset; Leverage is measured as total debt scaled by lagged total asset; Exchange Rate Exposure is measured as the exchange rate beta estimated from the multi-factor model; Foreign ownership is a fraction of the ?rm that is owned by non-Thai citizens; Foreign Director is a dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen as a director and zero otherwise; Thaksin Connection is a dummy variable that takes the value of one if the ?rm’s major shareholder is, or is related to (has the same last name as), a member of Thaksin’s Cabinet and zero otherwise; Family Business Group is a dummy variable that takes the value of one if the ?rm’s major shareholder is from the 50 largest family business groups in Thailand and zero otherwise; Political Connection is a dummy variable that takes the value of one if the ?rm is politically connected according to Faccio (2006). The details how each variable is constructed are in the appendix.
23
Table 1.1B: Univariate Analysis Size Quartile Smallest Capital Control Day Abnormal Return Liberalization Day Abnormal Return Firm Size Pro?tability Market-to-Book Cash Flow Growth Liquidity Leverage Exchange Rate Exposure Foreign Ownership Foreign Director Thaksin Connection Family Business Group Political Connection -10.38 7.29 20.56 1.18 6.31 -65.62 19.09 47.91 -0.43 7.61 0.22 0.02 0.01 0 Second -11.05 8.15 21.43 2.09 1.04 35.05 22.87 49.48 -0.57 13.32 0.27 0.02 0.05 0.02 Third -11.16 8.25 22.38 1.9 1.33 -83.31 17.15 58.32 -0.63 22.96 0.43 0.05 0.07 0.1 Largest -13.28 10.67 24.34 1.73 1.94 -117.68 12.13 54.62 -0.86 33.37 0.4 0.12 0.24 0.17
The table reports summary statistics of abnormal returns and ?rm characteristics classi?ed into four size quartiles. The capital control day is December 19, 2006 and the liberalization day is December 20, 2006. Abnormal returns are from market model estimated by the daily returns from September 29, 2005 to August 31, 2006 (a 261-trading-day period). All ?rms that were not traded on the capital control day and the liberalization day are excluded. Firm size is measured as the log of total asset; Pro?tability is measured as pre-tax pro?t or loss scaled by lagged total asset; Market-to-Book is measured as market value of equity divided by book value of equity; Cash Flow Growth is measured as lagged annual growth in operating cash ?ow; Liquidity is measured as net working capital (current asset - current liability) scaled by lagged total asset; Leverage is measured as total debt scaled by lagged total asset; Exchange Rate Exposure is measured as the exchange rate beta estimated from the multi-factor model; Foreign ownership is a fraction of the ?rm that is owned by non-Thai citizens; Foreign Director is a dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen as a director and zero otherwise; Thaksin Connection is a dummy variable that takes the value of one if the ?rm’s major shareholder is, or is related to (has the same last name as), a member of Thaksin’s Cabinet and zero otherwise; Family Business Group is a dummy variable that takes the value of one if the ?rm’s major shareholder is from the 50 largest family business groups in Thailand and zero otherwise; Political Connection is a dummy variable that takes the value of one if the ?rm is politically connected according to Faccio (2006).
24
Table 1.1C: Propensity Score Matching Treated (Largest) Capital Control Day Abnormal Return Unmatched -13.3067 -10.3577 -2.949 [-2.78] Average Treatment E?ect on the Treated -13.3067 -10.2044 -3.1023 [-2.18] Liberalization Day Abnormal Return Unmatched 10.6555 7.3836 3.2719 [3.25] Average Treatment E?ect on the Treated 10.6555 7.115 3.5405 [2.72] Controls (Smallest) Di?erence (Largest-Smallest)
The table compares the abnormal returns from the treatment group and the control group. The treatment group is the largest size quartile and the control group is the smallest quartile. The capital control day is December 19, 2006 and the liberalization day is December 20, 2006. Abnormal returns are from market model estimated by the daily returns from September 29, 2005 to August 31, 2006 (a 261-trading-day period). Both unmatched e?ects and propensity-score-matched e?ects (average treatment e?ects on the treated) are reported. Propensity scores are computed from the Probit model using four covariates: profitability, exchange rate exposure, foreign director dummy, and Thaksin connection. Firm size is measured as the log of total asset; Pro?tability is measured as pre-tax pro?t or loss scaled by lagged total asset; Foreign Director is a dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen as a director and zero otherwise; Thaksin Connection is a dummy variable that takes the value of one if the ?rm’s major shareholder is, or is related to (has the same last name as), a member of Thaksin’s Cabinet and zero otherwise. Numbers in the brackets are t-statistics.
25
Table 1.2A: Correlation Matrix: Size and Basic Firm Characteristics Firm Size Firm Size Pro?tability Market-to-Book Cash Flow Growth Liquidity Leverage 1 -0.0538 -0.0911 -0.0015 -0.1926 0.148 1 -0.197 -0.0107 0.2405 0.0536 1 0.0128 -0.0761 0.0413 1 -0.0079 0.0427 1 -0.1998 1 Pro?tability Market-to-Book Cash Flow Growth Liquidity Leverage
Firm size is measured as the log of total asset; Pro?tability is measured as pre-tax pro?t or loss scaled by lagged total asset; Market-to-Book is measured as market value of equity divided by book value of equity; Cash Flow Growth is measured as lagged annual growth in operating cash ?ow; Liquidity is measured as net working capital (current asset - current liability) scaled by lagged total asset; Leverage is measured as total debt scaled by lagged total asset.
26
Table 1.2B: Basic Firm Characteristics and the E?ects of Foreign Portfolio Investment Panel A Dependent Variable Firm Size Model 1 -0.8392*** [-4.24] Pro?tability Model 2 -0.7564*** [-3.3] Model 3 -0.7457*** [-3.31] 0.3619*** [3.65] Market-to-Book Model 4 -0.742*** [-3.08] 0.3042*** [3.01] -0.0274*** [-5.28] Cash Flow Growth Model 5 -0.7682*** [-3.14] 0.3038*** [3.1] -0.0282*** [-5.39] -0.0009 [-0.81] Liquidity Model 6 -0.6195** [-2.09] 0.346*** [3.18] -0.0259*** [-4.93] -0.0009 [-0.65] -0.0026 [-0.17] Leverage Model 7 -0.6148** [-2.08] 0.3481*** [3.18] -0.0257*** [-4.88] -0.0008 [-0.64] -0.0032 [-0.21] -0.0024 [-0.16] Constant Industry Dummy R-squared Observations Yes No 0.0367 312 Yes Yes 0.0927 309 Yes Yes 0.1232 306 Yes Yes 0.1326 294 Yes Yes 0.1374 293 Yes Yes 0.1479 265 Yes Yes 0.148 265
27
Table 1.2B (Continued): Basic Firm Characteristics and the E?ects of Foreign Portfolio Investment Panel B Dependent Variable Firm Size Model 1 0.7397*** [3.92] Pro?tability Model 2 0.7267*** [3.37] Model 3 0.7092*** [3.27] -0.2642*** [-2.64] Market-to-Book Model 4 0.7385*** [3.18] -0.2419** [-2.24] 0.003 [0.6] Cash Flow Growth Model 5 0.7804*** [3.38] -0.2398** [-2.32] 0.0042 [0.92] 0.0014 [1.42] Liquidity Model 6 0.753*** [2.63] -0.2731** [-2.47] 0.0029 [0.6] 0.0014 [1.1] -0.0011 [-0.07] Leverage Model 7 0.7204** [2.49] -0.2878*** [-2.61] 0.0015 [0.31] 0.0013 [1.05] 0.0031 [0.19] 0.0164 [1.12] Constant Industry Dummy R-squared Observations Yes No 0.0324 312 Yes Yes 0.0779 309 Yes Yes 0.0969 306 Yes Yes 0.1007 294 Yes Yes 0.1151 293 Yes Yes 0.127 265 Yes Yes 0.1313 265
The table reports the coe?cient estimates from regressions of abnormal returns on ?rm characteristics. The dependent variable in Panel A is the abnormal return on December 19, 2006 (the capital control day) and the dependent variable in Panel B is the abnormal return on December 20, (the liberalization day). Firm size is measured as the log of total asset; Pro?tability is measured as pre-tax pro?t or loss scaled by lagged total asset; Market-to-Book is measured as market value of equity divided by book value of equity; Cash Flow Growth is measured as lagged annual growth in operating cash ?ow; Liquidity is measured as net working capital (current asset - current liability) scaled by lagged total asset; Leverage is measured as total debt scaled by lagged total asset. All variables are from December 2004 and 2005 ?nancial statements. Also estimated but not reported are a constant term and 9-industry dummy variables. Numbers in the brackets are heteroscedasticity-robust t-statistics. *, **, and *** indicate statistical signi?cant at 10, 5, and 1 percent levels, respectively. R-squared and the number of observations are reported in the last two rows.
28
Table 1.3A: Correlation Matrix: Size and Firm’s International Involvement Firm Size 1 -0.0418 -0.2724 0.405 0.0964 0.0207 -0.0726 0.5124 1 0.0495 -0.1343 1 0.1234 1 1 Pro?tability Exchange Rate Exposure Foreign Ownership Foreign Director
Firm Size
Pro?tability
Exchange Rate Exposure
Foreign Ownership
Foreign Director
Exchange Rate Exposure is measured as the exchange rate beta estimated from the multi-factor model; Foreign ownership is a fraction of the ?rm
that is owned by non-Thai citizens; Foreign Director is a dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen as
a director and zero otherwise.
29
Table 1.3B: Firms International Involvement and the E?ects of Foreign Portfolio Investment Panel A Model 1 -0.7457*** [-3.31] 0.3619*** [3.65] 2.1986*** [3.18] 0.0172 [0.95] 1.1604 [1.18] Yes Yes 0.1232 306 306 302 302 0.159 0.1598 0.1642 Yes Yes Yes Yes 0.0969 306 Yes Yes Yes Yes Yes Yes 0.1229 306 Yes Yes 0.1216 302 [0.11] 0.0025 [3.13] [3.24] [-2.69] 2.2274*** 2.2872*** -1.7571*** [3.31] [3.24] [3.26] [-2.64] [-2.3] [-2.27] -1.7922*** [-2.68] -0.0129 [-0.6] 0.3266*** 0.316*** 0.3169*** -0.2642*** -0.236** -0.2291** [-2.35] [-2.62] [-2.42] [3.27] [2.5] [2.64] [2.3] -0.2308** [-2.3] -1.9052*** [-2.88] 0.0148 [0.54] -2.1936** [-2.12] Yes Yes 0.1394 302 -0.5426** -0.6438*** -0.6026** 0.7092*** 0.5469** 0.6273*** 0.5494** Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4 Panel B
Dependent Variable
Firm Size
Pro?tability
Exchange Rate Exposure
Foreign Ownership
Foreign Director
Constant
30
Industry Dummy
R-squared
Observations
The table reports the coe?cient estimates from regressions of abnormal returns on ?rm’s international involvement. The dependent variable in Panel
A is the abnormal return on December 19, 2006 (the capital control day) and the dependent variable in Panel B is the abnormal return on December
20, (the liberalization day). Firm size is measured as the log of total asset; Pro?tability is measured as pre-tax pro?t or loss scaled by lagged total
asset; Exchange Rate Exposure is measured as the exchange rate beta estimated from the multi-factor model; Foreign ownership is a fraction of the
?rm that is owned by non-Thai citizens; Foreign Director is a dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen
as a director and zero otherwise. Also estimated but not reported are a constant term and 9-industry dummy variables. Numbers in the brackets
are heteroscedasticity-robust t-statistics. *, **, and *** indicate statistical signi?cant at 10, 5, and 1 percent levels, respectively. R-squared and the
number of observations are reported in the last two rows.
Table 1.3C: Firm Size and Activities in International Capital Markets Number of Firms Are SET 100 Index Constituent 0 1 20 69 90 42 12 82 33 9 56 6 0 16 6 40 46 3 2 4 0 0 1 6 0 Ratings uity Abroad Abroad Have TRIS Credit Have ADR Have Issued EqHave Issued Bond
Size Quartile
Are SET 50 Index
Constituent
Smallest
0
Second
0
Third
2
Largest
44
All
46
Stock Exchange of Thailand selects the constituents of SET 50 and SET 100 indices by choosing the top 50 and top 100 listed ?rms in terms of
market capitalization, liquidity and compliance with requirements regarding the distribution of shares to minor shareholders. TRIS (Thailand Rating
Information Services Co.) is Thailand’s ?rst and most prominent rating agency. The list of Thai ?rms with ADRs (American Depository Receipts)
31
is from JP Morgan’s and Bank of New York’s Databases. Data on the issuances of equity and debt securities by Thai ?rms outside of Thailand (from
1990 to 2006) are from Thomson’s SDC Platinum Database.
Table 1.4A: Correlation Matrix: Size and Firms Political/Family Connections Firm Size Firm Size Thaksin Connection Family Business Group Political Connection 1 0.1327 0.3290 0.3019 1 0.4231 0.1248 1 0.1004 1 Thaksin Connection Family Business Group Political Connection
Thaksin Connection is a dummy variable that takes the value of one if the ?rm’s major shareholder is, or is related to (has the same last name as), a member of Thaksin’s Cabinet and zero otherwise. Family Business Group is a dummy variable that takes the value of one if the ?rm’s major shareholder is from the 50 largest family business groups in Thailand and zero otherwise; Political Connection is a dummy variable that takes the value of one if the ?rm is politically connected according to Faccio (2006): A ?rm is politically connected if (1) its major shareholder or top executive are parliament member, minister or head of the state or (2) its major shareholder or top executive are closely related to a top politician or a political party.
32
Table 1.4B: Political/Family Connections and the E?ects of Foreign Portfolio Investment Panel A Dependent Variable Firm Size Model 1 -0.5309** [-2.26] Pro?tability 0.3165*** [3.3] Exchange Rate Exposure 2.244*** [3.22] Foreign Director 1.1498 [1.49] Thaksin Connection -2.2718 [-1.24] Family Business Group -1.2385 [-0.95] Political Connection 0.5489 [0.31] Constant Industry Dummy R-squared Observations Yes Yes 0.1702 306 Yes Yes 0.1677 306 Yes Yes 0.1657 306 Yes Yes 0.1511 306 Yes Yes 0.1434 306 Model 2 -0.4991** [-2.04] 0.3069*** [3.11] 2.2457*** [3.24] 1.1881 [1.54] Model 3 -0.6117** [-2.41] 0.3185*** [3.27] 2.287*** [3.34] 1.1939 [1.55] Model 1 0.54** [2.41] -0.2213** [-2.23] -1.8366*** [-2.76] -1.8121** [-2.46] 3.1052* [1.91] 1.2429 [1.16] -1.7614 [-1.13] Yes Yes 0.1448 306 Panel B Model 2 0.5289** [2.23] -0.2138** [-2.18] -1.8530*** [-2.85] -1.8656** [-2.53] Model 3 0.6941*** [2.95] -0.2167** [-2.17] -1.9001*** [-2.93] -1.878** [-2.55]
The table reports the coe?cient estimates from regressions of abnormal returns on ?rms political/ family connections. The dependent variable in Panel A is the abnormal return on December 19, 2006 (the capital control day) and the dependent variable in Panel B is the abnormal return on December 20, (the liberalization day). Firm size is measured as the log of total asset; Pro?tability is measured as the pre-tax pro?t or loss scaled by lagged total asset; Exchange Rate Exposure is measured as the exchange rate beta estimated from the multi-factor model; Foreign Director is a dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen as a director and zero otherwise; Thaksin Connection is a dummy variable that takes the value of one if the ?rms major shareholder is, or is related to (has the same last name as), a member of Thaksins Cabinet and zero otherwise. Family Business Group is a dummy variable that takes the value of one if the ?rms major shareholder is from the 50 largest family business groups in Thailand and zero otherwise; Political Connection is a dummy variable that takes the value of one if the ?rm is politically connected according to Faccio (2006) and zero otherwise. Also estimated but not reported are a constant term and 9-industry dummy variables. Numbers in the brackets are heteroscedasticity-robust t-statistics. *,**, and *** indicate statistical signi?cant at 10, 5, and 1 percent levels, respectively. R-squared and the number of observations are reported in the last two rows.
33
Table 1.5: Placebo Test: Are large ?rms more susceptible to bad news? Dependent Variable Firm Size Model 1 0.4772*** [2.95] Pro?tability Model 2 0.4438*** [2.67] Model 3 0.4481*** [2.74] 0.3057*** [4.14] Exchange Rate Exposure Model 4 0.5496*** [3.25] 0.2881*** [4.19] 1.0988* [1.79] Foreign Director Model 5 0.5311*** [3.2] 0.2864*** [4.22] 1.1337* [1.83] 0.484 [1.14] Thaksin Connection Model 6 0.6125*** [3.64] 0.278*** [4.12] 1.0762* [1.8] 0.4255 [1.01] -3.232* [-1.87] Constant Industry Dummy R-squared Observations Yes No 0.0257 312 Yes Yes 0.0746 309 Yes Yes 0.1231 306 Yes Yes 0.1424 306 Yes Yes 0.1447 306 Yes Yes 0.1661 306
These placebo regressions examine the e?ects of ?rm size and other characteristics on abnormal returns when there is bad news. The dependent variable in is the abnormal return on September 22, 2006 (the ?rst day the market learned about the coup). Firm size is measured as the log of total asset; Pro?tability is measured as the pre-tax pro?t or loss scaled by lagged total asset; Exchange Rate Exposure is measured as the exchange rate beta estimated from the multi-factor model; Foreign Director is a dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen as a director and zero otherwise; Thaksin Connection is a dummy variable that takes the value of one if the ?rms major shareholder is, or is related to, a member of Thaksins Cabinet and zero otherwise. Also estimated but not reported are a constant term and 9-industry dummy variables. Numbers in the brackets are heteroscedasticity-robust t-statistics. *, **, and *** indicate statistical signi?cant at 10, 5, and 1 percent levels, respectively. R-squared and the number of observations are reported in the last two rows.
34
Figure 1.1: Stock Market Response to the Imposition and the Reversal of Capital Control
The Stock Exchange of Thailand (SET) Index is a composite calculated from the prices of all common stocks on the main board of the Stock Exchange of Thailand. The SET Index dropped from 730.55 to 622.14 on the capital control day (December 19, 2006) and bounced back from 622.14 to 691.55 on the liberalization day (December 20, 2006).
35
Appendix 1.1: Summary of Related Empirical Work Country Time Period Methodology Finding about Firm Size
Paper
Cross-Country Study 53 countries 1996-2005 Panel Data Regression Foreign portfolio investment is associated with an increase in the ability to issue securities for small ?rms; Foreign portfolio investment increases the maturity of bank loans. 13 countries 1988-1998 Euler Equation Method Financial liberalization relaxes ?nancial constraints for small ?rms.
Knill (2005)
Laeven (2003)
Single-Country Study Indonesia large ?rms. Ecuador 1983-1988 Euler Equation Method The 1986 liberalization did not relax ?nancial constraints, even for small ?rms. Mexico Korea 1980-2000 Tobin Q Mothod 1984-1994 Euler Equation Method The 1989 liberalization relaxed ?nancial constraints for small ?rms. The 1995 liberalization relaxed ?nancial constraints for small ?rms; small ?rms get much wider access to outside credit after the liberalization. Colombia 1989-1993 Structural Estimation The 1991 liberalization increased investment rate and lowered buying price of capital for small ?rms. 1981-1988 Euler Equation Method The 1983 liberalization relaxed ?nancial constraints for both small and
Harris et al. (1994)
36
18 countries rolling window Event Study
Jaramillo et al. (1996)
Gelos and Werner (2002)
Koo and Shin (2004)
Contreras
and
Makaew
(2007)
Event Study Small ?rms earn signi?cantly higher abnormal returns when stock markets are liberalized.
Patro and Wald (2005)
Appendix 1.2: De?nitions and Data Sources Description and Source The last reported trade price each day, Source: Reuters Percentage changes in daily prices, Source: Authors Calculation Everything that the company or group owns at the end of the ?nancial year, expressed in local currency. Is equal to total current assets plus total ?xed assets, Source: Reuters and SETSMART Log of total assets, Source: Authors Calculation Pre-tax pro?t or loss, expressed in local currency, Source: Reuters and SETSMART Dummy variables representing 9 industries: Agro and Food, Consumer Products, Financials, Industrials, Property and Construction, Resources, Services, Technology, and Others, Source: SETSMART Pro?t scaled by lagged total assets, Source: Authors Calculation Number of shares outstanding times market price, Source: Reuters Number of shares outstanding times book value per share, Source: Reuters Market equity scaled by book equity, Source: Authors Calculation The net cash in?ow from operating activity, expressed in local currency. Is equal to the sum of various operating activities., Source: Reuters and SETSMART Lagged annual growth in operating cash ?ows, Source: Authors Calculation
Variable Name
Daily Stock Price
Daily Stock Return
Total Assets
Firm Size
Pro?t
Industry Dummy
37
Pro?tability
Market Equity
Book Equity
Market to Book
Operating Cash Flows
Cash Flow Growth
Variable Name
Description and Source The cash and other assets that the company or group expects to turn into cash. Is usually equal to cash and equivalents plus receivables plus inventories plus any of the following items: short term investments; marketable securities; prepayments, Source: Reuters and SETSMART All of the liabilities that the company or group expects to have to meet within 12 months, Source: Reuters and SETSMART Current asset minus Current liability, Source: Authors Calculation Working capital scaled by lagged total assets, Source: Authors Calculation Current Liabilities plus Long-term liabilities, Source: Reuters and SETSMART Total debts scaled by lagged total assets, Source: Authors Calculation Local Currency (Thai Baht) / US Dollar Exchange Rate, Source: WRDS The beta coe?cient estimated from a multi-factor model regressing stock returns on percentage changes in the exchange rate. The factor model is estimated by the daily returns from September 29, 2005 to August 31, 2006 (a 261-trading-day period), Source: Authors Calculation A fraction of the ?rm that is owned by non-Thai citizens, Source: SETSMART A dummy variable that takes the value of one if the ?rm has at least one non-Thai citizen as a director and zero otherwise, Source: SETSMART A dummy variable that takes the value of one if the ?rms major shareholder is, or is related to, a member of Thaksins Cabinet and zero otherwise, Source:Ownership information is from SETSMART. The list of Thaksins Cabinet members is from the Thai government website: www.cabinet.thaigov.go.th/ A dummy variable that takes the value of one if the ?rms major shareholder is from the 50 largest family business groups in Thailand and zero otherwise, Source: Ownership information is from SETSMART. The list of family business groups is from a Brooker Group Publication: Thai Business Group: A Unique Guide to Who Owns What
Current Assets
Current Liabilities
Working Capital
Liquidity
Total Debts
Leverage
Exchange Rate
Exchange Rate Exposure
38
Foreign Ownership
Foreign Director
Thaksin Connection
Family Business Group
Variable Name
Description and Source A dummy variable that takes the value of one if the ?rm is politically connected according to Faccio (2006) and zero otherwise. Faccio (2006) de?nes politically connected ?rms broadly as ?rms that (1) have a major shareholder or a top executive who is a parliament member, minister or head of the state or (2) have a major shareholder or a top executive who is related to a top politician or a political party, Source: Faccio (2006)’s dataset is available on American Economic Review website: www.aeaweb.org/ Number of shares outstanding times market price, Source: Reuters The cash and cash equivalents value as shown in the company’s balance sheet. Cash and equivalents can include cash at bank and in hand, short-term deposits or other liquid assets, Source: Reuters and SETSMART The amount derived from the provision of goods and services falling within the company or group’s ordinary activities, otherwise known as turnover. The data is entered as reported in the pro?t and loss account of the annual report, Source: Reuters and SETSMART Stock Exchange of Thailand selects the constituents of SET 50 and SET 100 indices by choosing the top 50 and top 100 listed ?rms in terms of market capitalization, liquidity and compliance with requirements regarding the distribution of shares to minor shareholders, Source: Stock Exchange of Thailand website: www.set.or.th/ TRIS (Thailand Rating Information Services Co.), Thailands ?rst and most prominent rating agency, provides credit ratings of Thai ?rms, Source: TRIS website: www.trisrating.com/ American Depository Receipts of Thai ?rms, Source: JP Morgan website: www.adr.com/ and the Bank of New York website: www.adrbny.com/ Issuances of equity securities by Thai ?rms outside of Thailand from 1990 to 2006, Source: SDC Platinum Issuances of debt securities by Thai ?rms outside of Thailand from 1990 to 2006, Source: SDC Platinum
Political Connection
Market Capitalization
Cash Holdings
Sales
39
SET 50/SET 100 Index
TRIS Ratings
ADR
Equity Issuance Abroad
Debt Issuance Abroad
Chapter 2
Waves of International Mergers and Acquisitions
2.1 Introduction
In the past two decades, 26% of worldwide M&A activities involve acquirers and targets from di?erent countries. The aggregate volume of cross-border mergers from 1989 to 2008 totals more than 8 trillion dollars. In spite of such a large volume, most of the M&A literature focuses on domestic mergers. Moreover, the amount of cross-border mergers varies greatly from year to year. For example, the volume of worldwide M&A deals dropped by 62% from 2000 to 2003 but bounced back by 158% in 2006.1 Despite such a large year-to-year ?uctuation, most papers on cross-border M&As study the e?ects of long-run determinants like corporate governance and capital market development. These gaps in the literature motivate the research questions that are at the core of this paper: what are the dynamic patterns of cross-border mergers, and what are the factors that drive them? Using the data from 50 countries over the period of 1989-2008, I document the following facts about international M&As:
1
See table 2.1 and ?gure 2.1 for details
40
(1) International mergers come in waves that are highly correlated with business cycles. Merger booms coincide with booms in the real sector and in the ?nancial market. While the literature on merger waves shows that domestic mergers are pro-cyclical, I ?nd that cross-border mergers are even more pro-cyclical than domestic mergers. (2) Mergers are more likely to occur when both the acquirer and the target economies are booming. This is true even when I eliminate the e?ects of global booms. My ?nding refutes the widespread belief that most cross-border mergers occur when the target economies are in a recession or face a ?nancial crisis, and that acquirers are vulture investors taking advantage of liquidity-constrained targets (Krugman, 1998; Aguiar and Gopinath, 2005; Desai, Foley, and Forbes, 2007; Acharya, Shin, and Yorulmazer, 2009). Although such “?re sale” mergers can happen under speci?c circumstances, most mergers do not follow this pattern. (3) Merger booms have both an industry-level and a country-level component. Given that productivity shocks are better measured at the industry level and that ?nancial shocks such as a change in monetary policy are more of a country-wide phenomenon, this ?nding is consistent with the notion that M&As are driven by productivity shocks and facilitated by macro liquidity shocks (Harford, 2005; Eisfeldt and Rampini, 2006). (4) Across over one million ?rm-year observations, acquirers tend to be more productive than their industry peers and targets tend to be less productive than their industry peers. This ?nding supports the neoclassical theory of mergers in that high productivity ?rms acquire low productivity ?rms in order to redeploy their assets toward more pro?table uses (Maksimovic and Phillips, 2001). This ?nding is at odds with the “like-buys-like” theory in which high productivity acquirers seek high productivity targets to realize gains from asset complementarity (Rhodes-Kropf and Robinson, 2008). This paper makes contributions to several strands of merger literature. Most papers on cross-border M&As put their emphasis on long-run determinants of mergers. Starting in the 1980s, Errunza and Senbet (1984) develop a theory of why corporations diversify internationally based on capital and goods market imperfections. Rossi and Volpin (2004) ?nd that acquirers are more likely to be from countries with stronger investor protection than targets. Bris and Cabolis (2008) ?nd that the merger premium is higher when the acquirer 41
country’s investor protection is stronger than the target country’s. Ferreira, Massa, and Matos (2009) ?nd that cross-border mergers increase with foreign institutional ownership. Di Giovanni (2005) ?nds that mergers are more likely to originate from countries with developed ?nancial markets. Although corporate governance and capital market development are undeniably important for mergers, it is hard to imagine that the large year-to-year ?uctuations (i.e., waves) in M&A activities are driven by these long-run determinants. Variables such as investor protections are extremely persistent and can be traced back to colonial origins. The central contribution of this paper is to incorporate the dynamic dimensions from business cycle theories and help explain the cyclical ?uctuations of international mergers. My paper is also related to the neoclassical theory of mergers. This literature typically argues that merger waves are driven by productivity shocks (Maksimovic and Phillips, 2001; Ditmar and Ditmar, 2008; Yang, 2008). Some authors further argue that, for waves to be formed, productivity shocks must be accompanied by liquidity shocks (Eisfeldt and Rampini, 2003; Harford, 2005; Maksimovic, Phillips, and Yang, 2009). Although my international paper draws insights from relatively well-established domestic merger literature, I also address important issues unique to cross-border mergers, such as the role of exchange rates and government policies on multinational corporations. In relation to empirical work on domestic mergers, cross-border M&As are an excellent setting for out-of-sample tests and provide more comprehensive data. For instance, Dittmar and Dittmar (2008) suggest that U.S. domestic merger activities ?uctuate in response to GDP shocks. However, it is di?cult to identify the mechanisms by which GDP e?ects mergers using data from a single country. All acquirers, targets, and non-merging ?rms face the same macroeconomic shocks, and most macroeconomic shocks are highly correlated. In my international context, acquirers and targets are from di?erent countries so I can better identify where the shocks originate. Moreover, using the data from 50 di?erent countries, or 50x50 = 2,500 country pairs, will substantially increase the degrees of freedom of the analysis.
42
2.2
Empirical Framework
The goal of this paper is to document the important facts about cross-border mergers. More precisely, I ask the following four questions. (1) How do cross-border mergers behave over a business cycle? I ?rst present exploratory evidence on the cyclicality of international mergers. Without making any structural assumptions, I compute the correlations between merger activities and a number of macroeconomic indicators measuring the real sector, the ?nancial sector, and the external sector. The lead-lag correlations are computed within a seven-year window around mergers to o?er a clear picture of what happens before, during, and after merger waves. (2) Where do shocks that e?ect cross-border mergers come from? The results from the correlation analysis can be driven by the co-movements of the acquirer economy and the target economy or by global economic booms. Following Harford (2005) and Dittmar and Dittmar (2008), I assume that lagged macroeconomic indicators are proxies for exogenous shocks and that M&As ?uctuate in response to these shocks. Then, I regress mergers on acquirer country shocks and target country shocks, controlling for the year ?xed-e?ects to eliminate the e?ects of global booms. This exercise will help me identify where the cyclical nature of cross-border mergers originates. (3) What type of shocks (real or ?nancial) e?ect cross-border mergers? Mergers come in waves either because productivity shocks occur in waves or because ?nancial shocks occur in waves. Using industry-level data, I form indices of industry-level productivity and valuation shocks. Then, I regress mergers on the industry indices, controlling for country-level shocks and year ?xed-e?ects. Given that the macro ?nancial shocks (such as changes in lending rates, monetary policy, and security issuance cost) are likely to be highly correlated across di?erent industries within the same country, these regressions will allow me to distinguish the e?ects of productivity shocks from the e?ects of ?nancial shocks. (4) What types of ?rms engage in cross-border mergers? To complete the analysis, I examine the characteristics of the merging ?rms. Using data from WorldScope, I compute a number of productivity and valuation measures. Then, I
43
compare the characteristics of the merging ?rms with non-merging ?rms and compare the characteristics of acquirers with targets. Studying what types of ?rms engage in merger activities will shed some light on the motives behind M&A decisions.
2.3
Data
Mergers and Acquisitions Data The source of M&A data is Thomson’s Securities Data Corporation (SDC) database. My sample covers all deals announced and completed between 1988 and 2008. To ensure that my results represent a wide range of countries but are not driven by countries that rarely have mergers, I require that acquirers and targets must be from 25 developed countries with the most M&A deals and 25 developing countries with the most M&A deals. These 50 countries are listed in Table 2.2. There are 412,810 deals in my sample. The aggregate value of these deals is approximately 40 trillion dollars. Eight trillion dollars are from cross-border deals. [INSERT TABLE 2.1 AND FIGURE 2.1 HERE] Table 2.1 reports the aggregate volume and aggregate frequency of M&A activities. In Figure 2.1, aggregate volume exhibits both growing trends and large cyclical ?uctuations. The volume of all M&A deals grows from around 500 million dollars in the early 1990s to more than 3 trillion in 2006-2007. The cyclical component is very large, especially in recent years. For example, the volume of all M&A deals dropped by 62% from 2000 to 2003 but bounced back by 158% three years later. Cross-border M&A deals are more volatile than domestic deals. The standard deviation scaled by mean is 70% for all deals but 84% for cross-border deals. [INSERT TABLE 2.2 HERE] Table 2.2 shows the breakdown by country. Most M&As are between high-income countries. The countries that have a large number of acquirers also have a large number of targets. The developed countries that have the highest number of acquirers and targets 44
are the G-7 countries (except Italy): the U.S., UK, Germany, Canada, and Japan. The top developing countries are the BRIC countries (Brazil-Russia-India-China) plus Malaysia. SDC provides detailed information on deal characteristics. Aside from basic information such as country, industry, and year, the SDC data include deal size, percent acquired, method of payment, and acquirer/target public status. The SDC collects data from a number of sources including the SEC and international stock exchange ?lings, news wires, trade publications, as well as surveys of banks and advisory ?rms. It might be a concern that some deals do not have the size attached to them because ?rms are not required to report the transaction values to SDC. Di Giovanni (2005) ?nds no pattern in which industries, countries, or years have more missing values than others. As a precautionary measure, I also compare data from the SDC to the country-level FDI data from UNCTAD and the transaction-level data from Capital IQ. At the aggregate level, I observe similar cyclical patterns from these three sources. Macroeconomic Data Most of the macroeconomic data are from the World Bank’s World Development Indicator (WDI) database. I use the variables that capture the states of an economy in terms of the real sector, the ?nancial sector, and the external economy sector. For the real sector, I use the data on GDP, gross value added, gross capital formation, and total population. For the ?nancial sector, I use the data on domestic credit and stock market capitalization. For the external sector, I use current account balance and nominal exchange rate. I augment the WDI data with the foreign portfolio investment data (net foreign portfolio investment) from the IMF’s Coordinated Portfolio Investment Survey (CPIS) and stock price data (Average M/B and Average P/E ratios) from Kenneth French’s Website. The average M/B and average P/E are equal-weighted. The data from WDI cover all 50 countries over the 20-year sample period, but the CPIS and French’s data have less coverage. The CPIS data are available from 2001 and French’s data cover the entire period but are only available for 21 countries. The country-pair variables, geographical distance, and common language dummy are from Di Giovanni (2005). Micro Data I use the WorldScope database, which covers over 95% of world market capitalization. It 45
provides the ?nancial statement information and market price of ?rms around the world. From the 50 countries in Table 2.2, WorldScope provides full coverage of the listed ?rms in 31 countries, 10 of which are developing countries. WorldScope also provides targeted coverage (all listed ?rms with a market capitalization higher than 100 million dollars) for 16 countries. The missing countries are Slovakia, Lithuania, and Ukraine. The list of these countries is available in the appendix. I construct a 1988-2008 annual data set of all the public ?rms available. There are 1,104,516 observations in this data set.2 [INSERT TABLE 2.3 HERE] To ensure that my results are not speci?c to a particular variable de?nition, I compute six di?erent measures of productivity: return on assets, pro?t margin, labor productivity, sales growth, employment growth, and payout ratio. Return on assets is pro?t (EBITDA) divided by total assets; pro?t margin is pro?t divided by sales; labor productivity is pro?t per worker; sales growth is the percentage change in annual sales; employment growth is the percentage change in number of workers; and payout ratio is dividends divided by total assets. I compute three valuation measures: M/B, past one-year return, and past three-year returns. The M/B is market capitalization divided by total assets less total debts; past return is the percentage change in market prices. I also collect four other variables that might e?ect M&As: size as measured by log of total assets (book value), age calculated from incorporation date, age calculated from listing date, and leverage. Leverage is total debts divided by total assets. To ensure that my results are not driven by outliers or any mistakes in the original data set, I winsorize the data at 0.025. I report the descriptive statistics of WorldScope variables in Table 2.3.
2
I believe that WorldScope is an appropriate database for this research. Compared to practitioner-
oriented products like Reuters, WorldScope retains inactive ?rms but Reuters does not. M&As can result in the de-listing of target ?rms; therefore, information for many targets are not available in Reuters. Compared to other popular research-oriented products, such as S&P’s Research Insight or Compustat Global, which also cover inactive ?rms, WorldScope provides better coverage. For example, there are only 27,805 ?rms represented in Research Insight [but there are 52,596 ?rms represented in WorldScope]. Finally, many premium databases, such as Dun and Bradstreet’s WorldBase or CapitalIQ, cover larger cross-sections of ?rms. They provide limited or no historical data, which I deem necessary for a time-series study.
46
Filtering Procedure There are challenges associated with using raw data directly. First, most of the variables of interests such as merger activities, stock market capitalizations, and GDPs, are increasing over time. If I use the raw data to compute correlations or run regressions, then the results are likely to be spurious. Second, the focus of this paper is the cyclical properties of M&As rather than their crosscountry variations. If I run a panel-data regression using raw data as seen in Di Giovanni (2005), then the estimated coe?cients will combine the time-series and cross-sectional e?ects together. For example, Di Giovanni (2005) ?nds that that larger stock market capitalization in year t-1 leads to more acquisitions in year t. This ?nding could be driven by cross-country di?erences. For example, countries like the U.S. have larger stock markets than Sub-Saharan African countries do; therefore, his results might re?ect the fact that there are more American acquirers than Nigerian acquirers. My research question is di?erent: I ask whether there are more U.S. acquirers and targets when there is an economic boom in the U.S. Mendoza and Terrones (2008) develop an algorithm to identify and analyze credit booms. I use a minor variation of their algorithm to transform my data: (1) I de?ate all the nominal variables with GDP de?ators and scale them by total population. Because I try to measure various shocks to the economy, scaling by total population is more appropriate for my application than scaling by GDP. GDP itself is also a?ected by the shocks; thus, scaling by GDP will confound the e?ect of the shocks in the original variables. (2) I ?lter out trends in all variables by using the Hodrick-Prescott (HP) ?lter. Hodrick and Prescott (1997) propose a de-trending method, which is now commonly used in the business cycle literature. The HP ?lter decomposes the raw variable Xt into the trend component, trendt , and the cyclical component, shockt . Given the smoothing parameter ? , the ?lter will choose the trend component that minimizes the objective function:
T t=1 (Xt
? trendt )2 + ?
T ?1 t=2 ((trendt+1
? trendt ) ? (trendt ? trendt?1 ))2
47
The ?rst term in the objective function penalizes the deviations from the trend, while the second term penalizes the ?uctuation in the growth rate of the trend components, i.e., the non-smoothness of the trend. Following Mendoza and Terrones (2008), I apply the HP ?lter to the full sample period 1988-2008 and set the smoothing parameter equal to 100, which is commonly used with annual data.3 (3) I compute the standard deviations of shocks in each country and then scale the shocks with their standard deviations. Mendoza and Terrones (2008) de?ne “boom” as a situation in which the deviation from trend is unusually large relative to the country’s typical cycle. Scaling by the country’s standard deviation is necessary because some economies are more volatile than others. Moreover, such scaling will eliminate the cross-country di?erences in size and allow me to run a panel data regression in which all countries are treated equally. [INSERT FIGURE 2.2 HERE] Figure 2.2 shows an example of the raw and the de-trended series of U.S. ?rm acquisitions of assets in other countries. Consistent with Mendoza and Terrones (2008), I identify a “wave” as a situation in which the deviation from trend is unusually large. Instead of providing a speci?c cuto? and discretizing the wave variable, I use the HP-de-trended and standard deviation-normalized variables directly to preserve their information content. From the ?ltered variables, I observe many “merger waves” in the data. Out of 1,000 country-year observations, there are 165 observations in which total M&As activities exceed one standard deviation and 109 observations in which M&As exceed two standard deviations.
2.4
Merger Activities and Macroeconomic Conditions
Correlation Analysis Business cycle theories predict that there are systematic relations among macroeconomic
3
I examine the stationarity of the detrended data using the Levin-Lin-Chu test (a pooled Dickey-Fuller
test). The test con?rms that the panel is indeed stable.
48
variables. These variables might e?ect one another endogenously or be driven jointly by some unobservable factors. As a ?rst step, I do not make any causality or structural assumptions. To see which variables coincide or have lead-lag relations with M&As activities, I compute sample correlations at di?erent time periods in a seven-year window around mergers. This exercise o?ers a clearer picture of what happens before, during, and after merger waves. To examine how M&As ?uctuate over a business cycle, I compute the following correlations: Correlation(M ergerc,t , Xc,t+j ) j ? {?3, ?2, ?1, 0, 1, 2, 3}, where M ergerc is the aggregate volume of mergers in country c. The Xc s are the macroeconomic indicators capturing: (1) the real sector of the economy (gross value added, GDP, and capital formation); (2) the ?nancial sector of the economy (market capitalization, M/B, P/E, and domestic credit); as well as (3) the external sector of the economy (current account, exchange rate, and foreign portfolio investment). Both M ergerc and Xc are de-trended using the Mendoza and Terrones’ ?ltering procedure described earlier. [INSERT TABLE 2.4 HERE] The results are reported in Table 2.4. Because cross-border deals are associated with two countries, there are three sets of correlations: (1) between cross-border deals and the acquirer country’s characteristics, reported in Table 2.4A; (2) between cross-border deals and the target country’s characteristics, reported in Table 2.4B; and (3) between domestic deals and the corresponding country characteristics, reported in Table 2.4C. It is apparent from these tables that M&As are pro-cyclical and that merger waves coincide with macroeconomic booms. The correlations between mergers and real/?nancial indicators show a similar pattern across the board. The correlations between mergers at time t and the indicators at time t-3 are negative; then these correlations increase, become positive at time t-2 or t-1, and peak at time t; they then remain positive for a period of time and then turn negative at t+3. According to this cyclical pattern, the lagged values of the indicators will be able to predict mergers. 49
The ?uctuations of the real indicators should be highly correlated with aggregate productivity shocks facing an economy.4 The contemporaneous correlations between M&As and gross value added are positive and statistically signi?cant in all speci?cations. A one standard deviation shock in the acquirer’s value added is associated with a 0.26 standard deviation change in cross-border mergers; a one standard deviation shock in target’s value added is associated with a 0.18 standard deviation change in cross-border mergers; a one standard deviation shock in domestic value added is associated with a 0.07 standard deviation change in domestic mergers. I ?nd that the correlations between M&As and the ?nancial indicators are higher than the correlations between M&As and the real indicators. For example, stock market capitalization in the acquirer country has a 42% correlation with cross border M&As. A one-year lagged stock market capitalization has a 31% correlation with cross-border M&As, but gross value added only has correlations of 26% and 14%, respectively. This is not surprising given that the ?nancial indicators are forward-looking, but the real indicators are accounting numbers measuring past performance. As Harford (2005) points out, the ?uctuations in valuation measures can come from any source, including productivity shocks, liquidity shocks, and misvaluations. Comparing Tables 2.4A and 2.4B, real and ?nancial conditions of the acquirer country have a higher impact on M&As than the conditions in the target country. This ?nding can be explained by a variety of reasons. One example could be that acquirers have to raise funds for the acquisitions, and that the cost of ?nancing is lower when the acquirer country is booming. Another might be that the acquirers must take control of the targets, so it is more important that the acquirers receive high productivity shocks. Comparing Tables 2.4A and 2.4B with Table 2.4C, I ?nd that cross-border mergers are much more correlated with real and ?nancial conditions than domestic mergers. Harford (2005), Eisfeldt and Rampini (2006), Dittmar and Dittmar (2008), and Yang (2008) document that domestic M&As are pro-cyclical. The comparison between Table 2.4A-2.4B and Table 2.4C shows that cross-border mergers are even more pro-cyclical than domestic mergers.
4
Wurgler (2000) and Maksimovic and Phillips (2001) use value added as a proxy for productivity shocks.
50
Turning to the measures of the external sector, the dynamic pattern of correlations between cross-border mergers and the exchange rate is interesting. Merger waves do not occur when the domestic currency is strongest. The contemporaneous correlation between M&As and exchange rates is statistically zero. The domestic currencies are strong two to three years before the peak of the merger waves and become weak two to three years after the merger’s peak. This ?nding suggests that M&As do not react directly to exchange rate appreciations. Instead, both mergers and exchange rate movements are more likely to be part of a larger business cycle model in which symptoms such as the appreciation of local currencies and the run-up in real estate prices are typical during economic expansions. The correlations of other external indicators are less signi?cant. Cross-border mergers do not have a signi?cant relation with the acquirer’s current account. There is weak evidence that target countries run current account de?cits during the merger waves. This is expected because M&As are a part of the capital in?ow that might worsen the current account balance (current account de?cit = capital in?ow + change in foreign reserve). I detect a small correlation between mergers and foreign portfolio investments. This is probably due to the fact that the foreign portfolio investment data from CPIS have much less coverage than the domestic variables, which come from WDI. In sum, the correlation analysis reveals that M&As exhibit a strong cyclical pattern. Real and ?nancial indicators coincide with and predict mergers. Fact 1: Cross-border mergers are highly correlated with business cycles. Regression Analysis Because the correlation analysis is univariate in nature, it is possible that the results in Table 2.4 are driven by global economic booms or by the co-movements of the acquirer economy and the target economy. To answer this question, I move to a multivariate framework. Following Harford (2005) and Dittmar and Dittmar (2008), I assume that lagged macroeconomic indicators are proxies for exogenous shocks and that M&As ?uctuate in response to these shocks. Speci?cally, I regress mergers between country c1 and c2 on the lagged conditions of c1 and the lagged conditions of c2 , controlling for the year ?xed-e?ects. By
51
putting the acquirer conditions and the target conditions side by side, I can identify how much of the M&As are driven by the acquirer conditions and how much are driven by target conditions. Additionally, by including the year ?xed-e?ect, I can determine whether the variations beyond the global averages still have an e?ect on M&As. My speci?cation is: M ergerc1 ,c2 ,t = ?0 + ?1 Exc1 ,c2 ,t?1 + ?2 Xc1 ,t?1 + ?3 Xc2 ,t?1 +
c1 ,c2 ,t
where M ergerc1 ,c2 is the volume of deals with the acquirer in c1 and the target in c2 . Exc1 ,c2 is the exchange rate (acquirer currency per one unit of target currency). The Xc s are the real and ?nancial indicators I used earlier. All the variables are again de-trended by the Mendoza and Terrones’ ?ltering procedure. Putting all seven real and ?nancial indicators in a regression at the same time will result in a multi-collinearity problem. I solve this problem in two ways: (1) by picking a representative variable and (2) by using all the indicators to form economic shock indices. First, I pick the representative variable based on data availability. If data are equally available, then I try all of the indicators in a regression and select the horse-race winner. Second, I adopt Harford’s (2005) approach by forming indices using the ?rst principal component of all the indicators. (The real economy indicator is constructed from the gross value added, GDP, and capital formation; the ?nancial market indicator is constructed from market capitalization, M/B, P/E, and domestic credit.) [INSERT TABLE 2.5 HERE] The results are reported in Table 2.5. In column 1, the coe?cient of exchange rate is signi?cant and negative, which suggests that the acquirer currency is strong relative to the target currency prior to merger booms. In column 2, I use gross value added as a real economy indicator. The coe?cient of the acquirer country is estimated at 3%, and the coe?cient of the target country is estimated at 2%. These numbers are positive and statistically signi?cant even after controlling for the year ?xed-e?ect. These estimates suggest that mergers react to shocks both from acquirer countries and from the target countries. Because the year ?xed-e?ect has removed the global averages from all the 52
variables year by year, my result does not depended upon the worldwide merger booms in a particular time period. In column 3, I use market capitalization as a ?nancial market indicator. The coe?cient of the acquirer country is estimated at 6% and the coe?cient of the target country is estimated at 4%. Again, both numbers are positive and statistically signi?cant. Mergers are more likely to take place when stock markets in both the acquirer and the target country are booming. Consistent with the correlation analysis, ?nancial indicators are a better predictor of mergers than the real indicators. In addition, the conditions of the acquirer country are more signi?cant than the conditions of the target country. The speci?cations in columns 4 and 5 are similar to those in columns 2 and 3, except that I use ?rst principal component indicators. The results in columns 4 and 5 are similar to the ones in columns 2 and 3. In sum, the regression analysis con?rms that M&As react to shocks both in the acquirer country and in the target country. In other words, there are more M&As when both the acquirer and the target economies are booming. My results are also robust to the inclusion of the year ?xed-e?ects. Fact 2: There are more M&As when both the acquirer and the target economies are booming. Mergers and Global Economic Conditions From Table 2.5A, the year ?xed e?ect explains approximately 2% of the variations in cross-border merger activities. This ?nding suggests that there must be global factors driving mergers across di?erent countries. As an example of such factors, I replace the year dummies with three candidate measures of global economic conditions. Following Albuquerque, Loayza and Serven (2005), I use (1) World Equity Market, measured by the return on Morgan Stanley World Capital Index, (2) World Interest Rate, measured by the average of American, Japanese, and German three-month treasury rates, and (3) Credit Spread which is Moody’s AAA bond rate minus Moody’s BAA bond rate. I collect raw monthly data from Bloomberg, convert them into annual series, and detrend the series using the Mendoza and Terrones’ procedure.
53
The results are reported in Table 2.5B. The sign and the magnitude of the countrylevel indicators are similar to the ones in Table 2.5A. Even though the coe?cients of the acquirer and the target country indicators are still statistically signi?cant after controlling for the worldwide economic conditions, the e?ects of the global variables are relatively large - in some cases, larger than the e?ects of the acquirer and the target countries. For instance, the coe?cient of acquirer’s market capitalization is 0.07 and the coe?cient of target country is 0.03. The coe?cient of World Equity Market is estimated at 0.07.
2.5
Firm Characteristics and Industry Merger Waves
In this section, I examine the characteristics of merging ?rms and show how these characteristics change along with the merger waves. Characteristics of the Merging Firms In the previous section, I examined merger waves at the country level. In this section, I look inside each country and identify which ?rms engage in M&As activities. Analyzing the characteristics of the merging ?rms will shed some light on the main motives behind M&As in my sample. From the WorldScope data, I construct 13 measures of ?rm characteristics: 6 productivity measures, 3 valuation measures, and 4 other measures that might e?ect mergers. The productivity measures consist of return on assets (ROA), pro?t margin, labor productivity, sales growth, employment growth, and payout ratio.5 Although these measures are positively correlated, each represents di?erent concepts of productivity and has its own strength. For example, labor productivity captures technological shocks; pro?t margin captures demand conditions; the level measures, such as return on assets, capture productivity more directly; growth measures, such as sales growth, are less a?ected by ?rm-speci?c reporting practices or earning management. I examine all six to ensure that my results are not speci?c to a particular measure. For the valuation measures, I compute the market-to-book ratio, past one-year return, and past three-year returns, which
5
The term “productivity” is used loosely here- some of these should be labeled “pro?tability” measures
instead.
54
are similar to the measures used in Harford (2005). The other potential determinants of mergers are size, age based on incorporation date, age based on the listing date, and leverage. Because my sample consists of ?rms from di?erent countries, industries, and time periods, it is di?cult to interpret any di?erences in the unadjusted ?rm characteristics. For example, it is unclear whether a 2% ROA of a food factory in Thailand means the same thing as a 2% ROA of a car company in the U.S. To address this issue, I normalize each characteristic, labeled i, by: (1) Grouping all observations by country-industry-year and, for each group, computing the means and the standard deviations of i I use Fama-French’s 16 industries instead of the four-digit SIC code provided by WorldScope. The four-digit SIC industry is rather small, leaving some industries in small countries empty or sparsely populated. The de?nitions of Fama-French’s 16 industries and the
mapping from SIC code can be found on Kenneth French’s website:http://mba.tuck.dartmouth.edu/pages/ (2) Using the means and standard deviations to compute i’s Z-score (Z-score = (i-mean)/ standard deviation) for each observation. In other words, using the distribution of i in each country-industry-year, I convert the raw value of i into its position in the distribution. Mechanically, these Z-scores will have the mean of zero and the standard deviation of one. This adjustment will eliminate all cross-country and cross-industry di?erences. To investigate which ?rms are more likely to engage in M&As, I ?rst match the WorldScope data with the SDC data. For mergers that take place at time t, I compare the characteristic at time t-1 to avoid the reverse causality problem. Of all 173,357 deals that involve public acquirers, 122,118 deals can be matched with WorldScope ?rms. Of all 52,524 deals that involve public targets, 46,607 deals are matched. I then compare the Z-scores between the population of acquirers, targets, cross-border acquirers, cross-border targets, and other non-merging ?rms. [INSERT TABLE 2.6 HERE] The results are reported in Table 2.6. The ?rst four columns compare acquirers with 55
targets. Column 1 reports the di?erence between the average Z-scores of all acquirers and the average Z-score of all targets. Column 2 reports the di?erence between the crossborder acquirers and the cross-border targets. The numbers in the parenthesis are the t-statistics from the t-tests. Across all the productivity and valuation measures, I ?nd that the acquirers’ Z-scores are consistently higher than the targets’. The magnitudes of the di?erence are around 0.2-0.3 standard deviations. The economic signi?cance of these numbers is large but varies from country to country and from industry to industry. For example, in 2008, one standard deviation of ROA of the food industry in Thailand is 13% and one standard deviation of ROA of the American auto industry is 35%. The t-test only compares the averages of acquirers and targets. In columns 3 and 4, I compare the whole distributions using the Kolmogorov-Smirnov’s D-statistics. Column 3 reports the distributional distances between all acquirers and all targets. Column 4 reports the distributional distances between the cross-border acquirers and the cross-border targets. I ?nd that the di?erences are statistically signi?cant. This signi?cance con?rms that not only do the acquirers and the targets have di?erent means but they also come from di?erent distributions. In the last four columns, I compare the Z-score of the merging ?rms with the non-merging ?rms. Column 5 compares all of the acquirers with the non-acquirers, and column 6 compares cross-border acquirers with the non-acquirers. I ?nd that acquirers are more productive and have higher valuations than their industry peers. Column 7 compares all of the targets with the non-targets, and column 8 compares cross-border targets with non-targets. I ?nd that targets are less productive and have lower valuations than their industry peers. These results are virtually uniform across all measures. This is consistent with the neoclassical theory of mergers in which more productive ?rms purchase less productive ?rms to realize e?ciency gain. Turning to other characteristics, ?rms that participate in mergers are likely to be older and larger than non-merging ?rms. Among the merging ?rms, ?rms participating in cross-border mergers are older and larger than ?rms participating in domestic mergers. The acquirers are, generally, larger and older than the targets. For leverage, I ?nd some evidence that the targets tend to have higher leverage than the acquirers.
56
Because acquirers are larger and are more likely to be public than targets, acquirers ?nd more matches in WorldScope than targets. There is concern that the comparisons between acquirers and targets in columns 1-4 are potentially biased. However, Maksimovic, Phillips, and Yang (2009) report that non-listed ?rms are smaller and less productive than listed ?rms. This ?nding will bias my results toward zero. The fact that I still ?nd signi?cant di?erences between the acquirers and the targets implies that the unbiased di?erences must be very large and that my results in columns 1-4 can be thought of as a lower bound of the true di?erences. Fact 3: Acquirers are more productive ?rms. Targets are less productive ?rms. Comparisons between Acquirers and Targets In Table 2.6A, for cross-border mergers, the di?erence between the acquirers’ and the targets’ Z-scores might not be aligned with the di?erence in the original variable, i. In other words, the acquirers and targets are benchmarked by their industry peers in their own countries. It is unclear if the cross-border acquirers are more or less productive than their targets since the Z-scores come from two di?erent distributions. As a robustness check, I combine all countries together and assign a new Z-score based on industry-year grouping instead of country-industry-year grouping. The results are reported in Table 2.6B. I ?nd that, on average, the acquirers are still more productive and have higher valuation than the targets, but the signi?cance is not as strong as those of the original grouping. This is consistent with the country-level cross-sectional results in Section 2.5 that acquirers seek less productive ?rms in high income/productive countries, and not the least productive ?rms anywhere in the world. Characteristics of Merging Firms during Booms and Busts Motives behind mergers during booms and busts can be di?erent. Therefore, comparing the characteristics of the merging ?rms at di?erent points along the business cycle might shed some light on the factors driving merger activities. From the country- and the industry- level regressions, there are more mergers during booms. The standard neoclassical explanation is that productivity shocks are di?erent during booms and busts. Alternatively, more mergers during booms can be a result of the increase in participation by less productive ?rms. Less productive ?rms might engage in
57
MAs during booms due to higher capital liquidity, more free cash ?ows, or more intense product market competitions. In Table 2.6C, I compare the merging ?rms during booms and busts by computing another set of Z-scores based on country-industry grouping. I de?ne two types of booms (busts): (1) real booms (busts) as the periods in which HP detrended Gross Value Added is above (below) one standard deviation (2) ?nancial booms (busts) as the periods in which HP detrended Stock Market Capitalization is above (below) one standard deviation. I ?nd that more ?rms participate in MAs during booms and that the average acquirers during booms are smaller, younger, and have less leverage. However, I ?nd no evidence that economic booms lower the productivity threshold for mergers. The acquirers during booms are more productive than the acquirers during busts and the targets during booms are more productive than the targets during busts. Post-Merger Operating Performance To examine post-merger operating performance, I compute the percentage change in return on assets (Pro?t/Total Assets) over four windows: [t-1, t], [t-1, t+1], [t-1, t+2], and [t-1, t+3] where t denotes the year of the acquisition. The results are reported in Table 2.6D. Numbers in the table are the average performance of the treatment group (merging ?rms) minus the performance of the control group. From the ?rst four columns, without controlling for the selection issues, I ?nd that the acquirers and the targets underperform after mergers. Performance of the acquirers is worse than performance of the targets. Performance of the ?rms involving cross-border deals is worse than performance of the ?rms involving domestic deals. In Section 2.4, I ?nd that merging ?rms and non-merging ?rms have di?erent characteristics. Therefore, the unmatched results might re?ect di?erent initial characteristics rather than merger outcomes. To solve the problem, I use the propensity scores matching method. My probit selection model uses basic ?rm characteristics as covariates: lagged pro?tability, sales, total assets, and ?rm age. Columns 5-8 reports propensity-score-matched e?ects (average treatment e?ects on the treated). I ?nd that the most underperformance in the ?rst four columns disappears after controlling for basic ?rm characteristics. A potential explanation is that ?rms receive large productivity shocks prior to engaging in MA and the shocks revert in subsequent years.
58
Industry Shocks Next, I examine how the industry-level M&As are a?ected by the year-to-year ?uctuations in productivity and valuation measures. From section 2.4, it is not obvious whether the country-level indicators represent productivity shocks or other macroeconomic shocks because these shocks are highly correlated at the country level. Performing industry-level regressions will help me identify what types of shocks are driving the results. Although productivity shocks are probably best described at the industry level, ?nancial shocks (such as changes in lending rates, monetary policy, and security issuance cost) are more of an economy-wide phenomenon. If the industry-level regressors are signi?cant after controlling for the country-level regressors, then the productivity shocks are likely to be the factor driving M&As. At the same time, if the country-level regressors are signi?cant after controlling for the industry-level shocks, then the true shocks e?ecting M&As must have an economy-wide component. Using the six productivity measures and three valuation measures from the previous section, I construct the industry shock indices. I average the ?rm characteristics for each country-industry-year and normalize each series using the Mendoza and Terrones’ procedure. Similar to section 2.4, I either choose one variable (ROA) to represent productivity shocks and choose another variable (market-to-book) to represent the valuation shocks or I construct a productivity index from the ?rst principal component of the six productivity measures and construct a valuation index from the three valuation measures.6
6
It might be a concern that indices constructed from WorldScope variables might not be an accurate
proxy of industry shock because private ?rms in SDC are not covered by WorldScope. To the extent that shocks to ?rms in the same industry in the same country are correlated, the ?uctuations in the average productivity of the listed ?rms can be used as a proxy of the ?uctuations in the average productivity of all ?rms. To further address this concern, I use the United Nations Industrial Development Organization (UNIDO) database, which is a census-type dataset, to construct an alternative measure for industry shocks - value added per worker and output per worker. Again, I ?nd that the industry shocks have positive coe?cients. However, the new coe?cients are less signi?cant than those of the WorldScope shocks. This result might be due to the fact that UNIDO uses ISIC industry classi?cation, but WorldScope and SDC use SIC classi?cation and that the mapping between ISIC and SIC introduces noise into the UNIDO measures. Yet another possibility is that the universe of WorldScope (large/listed ?rms) is more relevant to mergers compared to the small manufacturing ?rms in UNIDO.
59
I regress M&A volume of industry i in country c on the conditions of industry i in country c and the macroeconomic conditions of country c. I also control for the year ?xed-e?ects. M ergeri,c,t = ?0 + ?1 Ii,c,t?1 + ?2 Xc,t?1 +
i,c,t
where Ii,c is the condition of industry i in country c and Xc is the condition of country c. [INSERT TABLE 2.7 HERE] The results are reported in Table 2.7. In Table 2.7A, I regress cross-border mergers on acquirer country and industry indicators. In Table 2.7B, I regress cross-border mergers on target country and industry indicators. In Table 2.7C, I regress domestic mergers on the country and industry indicators. In columns 1-3 and columns 7-9, I use the representative measures as my regressors: industry ROA as the real industry indicator, gross value added as the real economy indicator, industry M/B as the industry valuation indicator, and stock market capitalization as the ?nancial market indicator. In columns 4-6 and columns 1012, I use the principal component indices as my regressors. The country-level principal component indicators are similar to the ones in section 2.4. The coe?cients of the industry-level real indicators are positive and statistically signi?cant. In most speci?cations, the country-level real indicators are driven down to zero or less signi?cant than the industry-level real indicators. From the third columns of Table 2.7A2.7C (the speci?cation in which I include the real indicators at the country level and control for the year ?xed-e?ect), the industry real indicator’s coe?cients are estimated at 0.06 for an acquirer country, 0.03 for a target country, and 0.03 for domestic mergers. The coe?cients of the country-level real indicators in the same regressions are 0.02, 0.03, and 0, respectively. The coe?cients of the industry-level valuation indicators and the country-level ?nancial indicators are positive and statistically signi?cant in all speci?cations. In most speci?cations, the country-level ?nancial indicators outperform the industry-level valuation shocks. From the ninth column of Table 2.7A-2.7C (the speci?cation in which I include the country-level ?nancial indicators and control for the year ?xed-e?ect), industry valuation’s coe?cients are estimated at 0.09 for an acquirer country, 0.02 for a target country, 60
and 0.04 for domestic mergers. The coe?cients of the country-level ?nancial indicators in the same regressions are 0.12, 0.02, and 0.10, respectively. The country-level patterns that I document in section 2.4 are still preserved at the industry level. All the coe?cients in Table 2.7 are positive, indicating that there are more mergers during booms. The coe?cients of the valuation/?nancial market indicators are higher than those of the real/ productivity indicators. The coe?cients in Table 2.7A are larger than the coe?cients in Tables 2.7B and 2.7C, suggesting that cross-border mergers are more pro-cyclical than domestic mergers and that the conditions of acquirer countries are more important than the conditions of target countries. In sum, shocks that e?ect mergers have both signi?cant industry components and signi?cant country components. Productivity shocks are mostly signi?cant at the industry level. On the other hand, the country-level ?nancial indicators are still important after controlling for the industry-level valuation shocks. These ?ndings support the literature on U.S. merger waves (e.g., Harford, 2005; Eisfeldt and Rampini, 2006; Maksimovic, Phillips, and Yang, 2009) in which mergers are a?ected by both productivity shocks and macro ?nancial shocks. Fact 4: Shocks that e?ect mergers have both signi?cant industry and signi?cant country components.
2.6
Additional Results
This section provides two sets of cross-sectional evidence to complement the time-series results. At the most aggregated level, I examine the long-run country e?ect on the 20-year aggregate of M&A activities. At the most disaggregated level, I examine the deal-level characteristics of domestics and cross-border M&As. The results in this section will give more information about the motives behind M&As and help to identify the issues that might be particularly pertinent to cross-border mergers. Country-Level I study the cross-country determinants of M&As by estimating the gravity model. The 61
gravity model is one of the most popular empirical models used in international trade. Instead of using di Giovanni (2005)’s panel version of the gravity model, which combines the time-series and the cross-sectional e?ects together, I use a cross-sectional version. M ergerc1 ,c2 = ?0 + ?1 Distancec1 ,c2 + ?2 Xc1 ,t0 + ?3 Xc2 ,t0 +
c1 ,c2 ,
where the variable M ergerc1 ,c2 is the 20-year aggregate volume of M&A ?ow from country c1 to country c2 . I de?ate the annual mergers data with the GDP de?ators and aggregate them from 1989 to 2008.7 I use two variables to measure the distance between the acquirer and the target countries: the geographical distance and the common language dummy. The geographical distance captures various aspects of the a?nity between the two countries. Examples include the volume of international trade, the presence of regional associations (such as EU and NAFTA), and the transportation costs. The common language dummy is a proxy for informational distance or the degree of information asymmetry between the two countries. The X s are the likely determinants of mergers in the long run. I use total population to measure country size, GDP to measure the level of income, and stock market capitalization to measure the degree of ?nancial development. The X s are measured in the base year (t0 =1988, outside of the M&As 20-year sample) to avoid the endogeneity problem. The descriptive statistics of the X s are reported in Table 2.9. [INSERT TABLE 2.8 AND TABLE 2.9 HERE] The results of the gravity model are reported in Table 2.8. From column 1 and column 2, the log of distance has negative and signi?cant coe?cients and the common language dummy has positive and signi?cant coe?cients. In other words, most mergers are between countries that are close together in terms of geographical and informational distance. If the distance of a country-pair is 1% smaller, then mergers will increase by 0.5-0.6%. If a country pair that does not share a common language adopts a common language, then mergers will increase by 1.8%. From columns 3 and 4, all four coe?cients of populations and GDPs are positive and statistically signi?cant. This result means that most M&A activities are between large
7
The de?ator used is the average GDP de?ator of the acquirer and the target countries.
62
and rich countries. A 1% increase in the 1988 population is associated with a 0.2% increase in M&As for the acquirer country and a 0.2% increase for the target countries. Keeping total population constant, a 1% increase in 1988 GDP is associated with a 1.1% increase in M&As for the acquirer country and a 0.5% increase for the target countries. The coe?cients of the acquirer countries are higher than the coe?cients of the target countries, implying that acquirer countries are on average larger and richer than the target countries. The results so far are consistent with the neoclassical theories and the standard results from international trade literature. That is, although ?rms tend to trade with partners in closer countries with larger markets, acquirers tend to seek targets in closer countries with a higher level of economic activities. Interestingly, the log of stock market capitalization in 1988 is highly signi?cant. From column 6, the coe?cient of acquirer market capitalization is estimated at 0.65, driving out the signi?cance of the GDP variable. The coe?cients of the target market capitalization is also statistically signi?cant but with a much smaller magnitude (0.15). This result suggests deep ?nancial markets, especially in the acquirer countries, are very important for M&As. Deal-level Here, I examine the characteristics of deals in my sample and compare the characteristics of domestic deals to the characteristics of cross-border deals. The characteristics I study are deal size, payment method, and the acquirer’s and target’s listing status. I am also interested in whether these deals are in high-tech, tradable, or related industries. “Deal Size” is the transaction value in millions of dollars. The “Cash-based Dummy” is a dummy variable taking the value of one if the percentage of cash is higher than the percentage of stock, and zero otherwise. The “Listed Acquirer” is a dummy taking the value of one if the acquirer is listed. The “Listed Target” is a dummy taking the value of one if the target is listed. The “Tradable” dummy is equal to one if the acquirer and the target are in tradable industries as de?ned by Aguiar and Gopinath (2005). The “High-tech” dummy is equal to one if the acquirer and the target are in the high-tech industry according to the American Electronic Association (http://www.aeanet.org/Publications/IDMK de?nition.asp). “Relatedness” is the absolute value of the di?erence between an acquirer’s four-digit SIC
63
and target’s four-digit SIC, as Alfaro and Charlton (2006) argue that related industries tend to have closer SIC codes. The results are reported in Table 2.10. Columns 1, 2, and 3 show the average characteristics of all deals, the domestic deals, and the cross-border deals, respectively. In column 4, I report the t-statistics of the di?erences between the domestic deals and the cross-border deals. On average, the cross-border deals are larger than the domestic deals and more likely to involve listed acquirers. This is consistent with the ?nding in section 2.4 that, among the listed acquirers, the cross-border acquirers tend to be larger than the domestic acquirers. I also ?nd that cross-border deals are more likely to be cash-based compared to the domestic deals. These comparisons suggest that ?nancial constraints might be more relevant to cross-border deals than to domestic deals. For the industry comparisons, cross-border deals are more likely to be in the tradable, hightech, and related industries. This is circumstantial evidence that cross-border mergers are more likely to be driven by neoclassical motives than domestic mergers; mergers in tradable industry are likely to be driven by comparative advantage and trade costs; ?rms in the high-tech industry are more likely to have ?rm speci?c assets that can be redeployed in another country; and mergers in related industries are more likely to generate synergies. A concern might be that the results in column 4 are driven by the di?erences in the compositions of the domestic deals and cross-border deals along country, industry, and year dimensions. For example, cross-border deals might cluster in certain countries or certain time periods compared to domestic deals. I address this problem by regressing deal characteristics on a cross-border dummy and controlling for country, industry, and year ?xed-e?ects (only country and year ?xed-e?ects for the industry comparisons). To ensure that the reported numbers are comparable to the t-statistics in column 4, I use a linear model instead of a logit or a probit. The coe?cients of the cross-border dummy are reported in column 5. These numbers are roughly similar to the ones in column 4.
64
2.7
Conclusion and Discussion
In this section, I discuss the results from sections 2.4-2.6 in light of the popular theories on mergers and then conclude. The Fire Sale Theory Krugman (1998), Aguiar and Gopinath (2005), and Acharya, Shin, and Yorulmazer (2009) propose the ?re sale theory of FDI in which foreign investors acquire ?rms in countries facing bad shocks such as ?nancial crises in order to take advantage of the liquidity constrained targets. The ?re sale theory has received broad empirical support. For example, Aguiar and Gopinath (2005) document an increase in foreign acquisitions during the East Asian ?nancial crisis. Desai, Foley, and Forbes (2007) ?nd that multinationals increase their investment in foreign a?liates when the host countries are facing currency crises. Using the data from 50 countries over the last 20 years, my results suggest the opposite: there are more mergers when the target economy is booming. Even after controlling for the acquirer’s boom and the global boom, there are still more foreign acquisitions when the target economy receives good shocks. In section 2.4, I ?nd that the correlations between cross-border mergers and the target’s macroeconomic conditions are approximately the same as the correlations between domestic mergers and domestic conditions. In other words, the capital in?ow through acquisitions is as pro-cyclical as domestic mergers. While M&As driven by the ?re sale motive might be present in a speci?c country at a speci?c time, most mergers in my sample are more consistent with the theory that ?rms invest in other countries to gain access to new markets and new investment opportunities and that it is better to enter the target countries when the demand is strong, the productivity is high, and the business environment is good. The Agency Theory Jensen (1986) indicates that M&As can be driven by agency problems: that is, the acquirer’s CEO might value mergers excessively. Although M&As might destroy ?rm value, corporate diversi?cation tends to reduce the risk of managerial human capital and enhance the CEO’s career prospects. The agency theory is strengthened by the fact that
65
numerous authors ?nd that conglomerates in the U.S. are traded at a discount, compared to single-segmented ?rms. The agency theory of mergers focuses on the acquirer’s problems and provides no speci?c predictions about the targets. However, in the international context, I ?nd that the characteristics of the target ?rms and the target countries have signi?cant e?ects on merger decisions. Moreover, Schenzler, Gande, and Senbet (2009) ?nd that global diversi?cations enhance ?rm values as measured by Tobin’s q. Their paper, combined with my ?ndings in section 2.5, shows that the acquirers are the more productive ?rms and the targets are the less productive ?rms. This ?nding suggests that most cross-border M&As tend to be driven by the value-enhancing neoclassical motives, rather than by value-destroying agency motives. The Misvaluation Theory Shleifer and Vishny (2003) and Rhodes-Kropf and Viswanathan (2004) propose a theory in which mergers are driven by stock market misvaluations. In their papers, targets accept the overpriced stock of the acquirers because they have a short time horizon or because they overestimate the synergies from the mergers. Rhodes-Kropf, Robinson, and Viswanathan (2005) ?nd that merger waves in the U.S. coincide with high M/B ratios and argue in favor of the misvaluation theory: when the market valuation is high, there are more M&As because acquirers will try to sell their overpriced stocks to targets. In the international context, Baker, Foley, and Wurgler (2009) use the U.S. foreign direct investment data to show that FDI ?ow is large when the source country stock market valuation is high. The authors attribute this ?nding to the misvaluation theory. In section 2.4, I study the behavior of ten macroeconomic indicators during the seven-year period around merger waves. I ?nd that most indicators, including the ones that are typically associated with the misvaluation theory like M/B and exchange rate, are highly correlated with one another and exhibit strong cyclical patterns. This correlation implies that M&As might not react directly to these indicators. Instead, all variables might be a part of a larger business cycle model in which M&As, market capitalization, and exchange rates are driven by common factors like productivity shocks. Moreover, my data do not re?ect the many predictions of the misvaluation theory. One 66
example is that misvaluation theory predicts that merger waves coincide with a strong acquirer currency. If the main motive of mergers is to take advantage of temporary exchange rate ?uctuations, then most mergers should occur when the acquirer’s currency is at its strongest relative to a target’s currency. While a one-year lagged exchange rate can predict mergers, I ?nd that M&As do not peak when the acquirer’s currency is at its strongest. The peak appreciation is three years prior to merger waves. Another example is the prediction about the method of payment. In section 2.6, I ?nd that the cross-border deals are less likely to be stock-based compared to domestic deals. The misvaluation theory predicts that domestic mergers are more pro-cyclical and more responsive to stock prices. However, my ?ndings in sections 2.4 and 2.5 are the opposite: cross-border mergers are much more pro-cyclical than domestic mergers.8 In this paper, I present key facts about international mergers. Speci?cally, I answer these four main questions: (1) How do cross-border mergers behave over a business cycle? International mergers come in waves and are very pro-cyclical. (2) Where do shocks that e?ect cross-border mergers originate? Most mergers occur when both the acquirer and the target economies are booming. (3) What type of shocks (real or ?nancial) e?ect crossborder mergers? Merger booms have industry-level (productivity shock) and country-level (?nancial shock) components. (4) What types of ?rms engage in cross-border mergers? Acquirers tend to be more productive than average ?rms and targets tend to be less productive than average ?rms. In the next chapter, guided by the key facts above, I propose and estimate a dynamic structural model that is built on the neoclassical theory of mergers.
8
9
I do not mean to suggest that misvaluations do not occur. It is possible that stock market booms
re?ect bubbles or other irrationalities in the stock market. Because the merging ?rms are not paying for such irrationalities, as shown in Shleifer and Vishny (2003) and Rhodes-Kropf and Viswanathan (2004), it does not matter from my modeling standpoint whether the low cost of capital comes from rational sources or irrational sources like bubbles. 9 Even though my ?ndings suggest that neither the agency nor the misvaluation theories are the main motive for cross-border mergers, the distinctions between these two theories and the neoclassical theories are not crucial from my modeling standpoint. The dynamic structural model in Chapter 3 can be reinterpreted from the misvaluation or agency angles conveniently. For example, the liquidity shocks in my model can be interpreted as misvaluation shocks and the productivity shocks in my model can be interpreted as
67
Table 2.1 Merger Activities over Time Year All Mergers (Frequency) 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 Total 6,473 8,918 9,451 12,964 12,616 13,214 15,603 18,945 20,188 22,233 24,541 26,689 29,138 22,049 19,270 20,393 22,450 25,163 27,223 28,766 26,523 412,810 Cross-border (Frequency) 1,450 2,177 2,378 2,706 2,408 2,654 3,271 3,950 4,239 4,836 5,673 6,504 7,731 5,358 3,937 3,960 4,682 5,547 6,274 7,033 5,857 92,625 All Mergers (Volume) 503 545 399 329 363 463 603 934 1,084 1,571 2,385 3,110 3,227 1,529 1,078 1,223 1,743 2,401 3,155 3,422 1,669 31,736 Cross-border (Volume) 99 122 125 66 73 85 108 186 197 289 559 971 972 431 268 232 441 608 828 1,068 524 8,252
The table reports the aggregate volume and the aggregate frequency of M&A activities from Thomson’s Securities Data Corporation (SDC). The sample covers all deals whose acquirers and targets are from 25 developed countries and 25 developing countries with the most M&A’s. Frequency is measured by the number of deals. Volume is the transaction value in billion of current dollars.
in?ated productivity shocks contaminated by managers’ private bene?ts of control.
68
Table 2.2 Merger Activities by Country Country Argentina Australia Austria Belgium Brazil Bulgaria Canada Chile China Colombia Czech Republic Denmark Egypt Finland France Germany Greece Hong Kong Hungary India Indonesia Ireland-Rep Israel Italy Japan Lithuania Malaysia Mexico Netherlands New Zealand Norway Peru Philippines Poland Portugal Number of Acquirers 1,001 15,715 2,408 3,292 2,543 378 20,643 561 4,099 213 818 3,546 236 5,101 18,229 21,070 1,102 5,630 927 4,095 624 2,287 1,126 7,771 18,595 184 7,424 970 8,207 2,582 3,291 219 786 1,277 1,415 Number of Targets 2,023 16,950 2,152 3,208 3,918 685 19,781 967 6,220 458 1,668 3,360 342 4,991 18,253 21,939 1,028 5,248 1,669 4,780 1,186 1,899 1,191 8,811 16,716 372 7,151 1,853 6,749 3,253 3,314 494 1,120 2,378 1,737 Volume Acquired 78 794 68 292 234 4 1,104 37 174 18 15 136 16 155 1,738 1,243 50 332 8 106 42 100 67 1,024 1,211 1 174 186 807 77 180 10 35 20 89 Volume Sold 136 800 80 262 338 15 1,151 68 255 45 49 136 32 132 1,309 1,298 56 270 33 128 64 82 59 1,025 1,107 5 150 224 701 97 187 26 53 58 94
69
Table 2.2 (Continued) Merger Activities by Country Country Romania Russian Fed Singapore Slovak Rep South Africa South Korea Spain Sweden Switzerland Thailand Turkey Ukraine United Kingdom United States Venezuela Number of Acquirers 215 3,342 4,618 204 2,841 2,891 7,764 7,400 5,162 1,494 518 271 49,040 158,492 193 Number of Targets 641 3,922 3,672 381 3,090 3,167 8,933 6,687 4,274 2,035 859 521 46,171 150,225 368 Volume Acquired 1 283 233 2 143 219 720 358 623 29 27 2 3,613 14,800 14 Volume Sold 20 302 129 11 150 259 582 417 432 47 73 17 3,305 15,400 22
The table reports the aggregate volume and the aggregate frequency of M&A activities from Thomson’s Securities Data Corporation (SDC). The sample covers all deals announced and completed between 1988 and 2008. Frequency is measured by the number of deals. Volume is the transaction value in billion of current dollars.
70
Table 2.3 Descriptive Statistics of Firms in WorldScope Variable Productivity Return on Assets Pro?t Margin Labor Productivity Sale Growth Employment Growth Dividend Valuation Market-to-Book Past Return-1 year Past Return-3 year Other Size Age1 Age2 Leverage 5.07 3.25 2.45 24.4 2.37 1.05 1.09 23.05 -0.5 0.69 0 0 10.13 4.9 4.41 93.03 1.41 -11.48 -17.69 1.77 67.14 110.88 0.03 -208.7 -339.56 8.91 129.88 206.23 2.22 -3.04 0.04 11.43 4.85 1.1 29.26 91.43 0.15 35.72 25.06 1.86 -133.27 -476.68 -0.31 -80.49 -61.47 0 34.68 88.09 0.67 124.66 84.73 8.32 Mean Standard Deviation Minimum Maximum
The table reports descriptive statistics of ?rms in the WorldScope database. Return on Assets is pro?t (EBITDA) divided by total assets; Pro?t Margin is pro?t divided by sales; Labor Productivity is pro?t per worker; Sales Growth is the percentage change in annual sales; Employment Growth is the percentage change in number of workers; Payout Ratio is dividends divided by total assets; M/B is market capitalization divided by (total assets less total debts); Past Return is the percentage change in market prices; Size is log of total assets (book value); Age1 is number of years since the incorporation date; Age2 is number of years since the listing date; and Leverage is total debts divided by total assets. All Percentage changes are calculated using the log formula.
71
Table 2.4A Correlations between Cross-Border Mergers and Acquirer’s Macroeconomic Conditions Cor(cross border mergers, acquirer’s Xt+j ) Real Economy Gross Value Added Gross Domestic Product Capital Formation Financial Markets Stock Market Capitalization Average M/B Ratio Average P/E Ratio Domestic Credit International Trade and Investment Current Account Balance Exchange Rate Net Foreign Portfolio Investment 0.08** -0.11*** -0.16 0.01 -0.15*** -0.28** 0.04 -0.11*** -0.06 -0.02 0 0.07 -0.04 0.09*** 0.12* 0 0.16*** 0.11 0.10*** 0.10*** -0.01 -0.11*** -0.17*** -0.12** -0.08* 0.12*** 0.02 0 0 0.31*** 0.26*** 0.15*** 0.06* 0.42*** 0.50*** 0.31*** 0.16*** 0.18*** 0.44*** 0.18*** 0.11*** -0.04 0.10* 0.01 0 -0.13*** -0.27*** -0.12** -0.10*** -0.09** -0.10** -0.08** 0.02 -0.02 0.05 0.14*** 0.11*** 0.15*** 0.26*** 0.25*** 0.24*** 0.23*** 0.22*** 0.19*** 0.07** 0.07** 0.04 -0.07* -0.06* -0.11*** t-3 t-2 t-1 t t+1 t+2 t+3
The table reports the correlations between the volume of cross-border mergers at time t and the acquirer country’s conditions at time t+j, where j is from -3 to +3. Gross Value Added, Gross Domestic Product, Capital Formation, Stock Market Capitalization, Domestic Credit, Current Account Balance, and Exchange Rate (local currency unit /U.S. dollar) are from the World Development Indicator Database. Average M/B Ratio and Average P/E Ratio are equal-weighted averages from Kenneth French’s website. Net Foreign Portfolio Investment (out?ow - in?ow) is from the Coordinated Portfolio Investment Survey Database. All variables are standardized by the Mendoza and Terrones’ procedure. The *, **, and *** indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
72
Table 2.4B Correlations between Cross-Border Mergers and Target’s Macroeconomic Conditions Cor(cross border mergers, target’s Xt+j ) Real Economy Gross Value Added Gross Domestic Product Capital Formation Financial Markets Stock Market Capitalization Average M/B Ratio Average P/E Ratio Domestic Credit International Trade and Investment Current Account Balance Exchange Rate Net Foreign Portfolio Investment 0.02 -0.12*** -0.03 -0.03 -0.10*** -0.04 -0.07** -0.02 -0.26*** -0.07** 0.09** 0.1 -0.13*** 0.13*** 0.13* -0.07** 0.17*** 0.20** 0.02 0.17*** -0.13 -0.04 -0.17*** -0.12** -0.06 0.09** 0.03 0.03 -0.02 0.23*** 0.24*** 0.18*** 0.08** 0.26*** 0.37*** 0.20*** 0.10*** 0.09*** 0.38*** 0.25*** 0.08** -0.11*** 0.09* 0.04 0.01 -0.20*** -0.24*** -0.13** -0.09*** -0.04 -0.06 0.02 0.01 0.02 0.11*** 0.11*** 0.12*** 0.17*** 0.18*** 0.18*** 0.18*** 0.17*** 0.18*** 0.13*** 0.06 0.04 -0.01 -0.11*** -0.13*** -0.15*** t-3 t-2 t-1 t t+1 t+2 t+3
The table reports the correlations between the volume of cross-border mergers at time t and the target country’s conditions at time t+j, where j is from -3 to +3. Gross Value Added, Gross Domestic Product, Capital Formation, Stock Market Capitalization, Domestic Credit, Current Account Balance, and Exchange Rate (local currency unit /U.S. dollar) are from the World Development Indicator Database. Average M/B Ratio and Average P/E Ratio are equal-weighted averages from Kenneth French’s website. Net Foreign Portfolio Investment (out?ow - in?ow) is from the Coordinated Portfolio Investment Survey Database. All variables are standardized by the Mendoza and Terrones’ procedure. The *, **, and *** indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
73
Table 2.4C Correlations between Domestic Mergers and Macroeconomic Conditions Cor(domestic mergers, Xt+j ) Real Economy Gross Value Added Gross Domestic Product Capital Formation Financial Markets Stock Market Capitalization Average M/B Ratio Average P/E Ratio Domestic Credit International Trade and Investment Current Account Balance Exchange Rate Net Foreign Portfolio Investment 0.09** -0.07* 0.21* 0.06* -0.05 -0.15 0.03 -0.04 -0.12 -0.09*** 0.02 -0.05 -0.13*** 0.08** 0.24*** -0.04 0.15*** -0.09 0.01 0.18*** 0.1 -0.09** -0.10* 0.01 -0.06 0.05 -0.03 -0.06 -0.08** 0.21*** 0.10** 0.09* -0.03 0.27*** 0.32*** 0.20*** 0.04 0.12*** 0.33*** 0.12** 0.08** 0 0.10** 0.03 0.06* -0.14*** -0.08 0 -0.01 -0.11*** -0.15*** -0.11*** -0.08* -0.09** -0.04 0.01 0.02 0.07** 0.07** 0.12*** 0.13*** 0.13*** 0.17*** 0.17*** 0.08** 0.10*** 0.08** -0.01 -0.03 -0.01 t-3 t-2 t-1 t t+1 t+2 t+3
The table reports the correlations between the volume of domestic mergers at time t and the macroeconomic conditions at time t+j where j is from -3 to +3. Gross Value Added, Gross Domestic Product, Capital Formation, Stock Market Capitalization, Domestic Credit, Current Account Balance, and Exchange Rate (local currency unit /U.S. dollar) are from the World Development Indicator Database. Average M/B Ratio and Average P/E Ratio are equal-weighted averages from Kenneth French’s website. Net Foreign Portfolio Investment (out?ow - in?ow) is from the Coordinated Portfolio Investment Survey Database. All variables are standardized by the Mendoza and Terrones’ procedure. The *, **, and *** indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
74
Table 2.5A Mergers and Country-Pair Conditions Y= Volume of Mergert Exchange Ratet?1 Real Economy Indicatoracquirer,t?1 Real Economy Indicatortarget,t?1 Financial Market Indicatoracquirer,t?1 Financial Market Indicatortarget,t?1 Year Fixed E?ect Observations R-squared No 23450 0 Yes 16838 0.02 1 -0.02** [0.01] 2 -0.01 [0.01] 0.03*** [0.01] 0.02** [0.01] 0.06*** [0.02] 0.04** [0.02] Yes 6905 0.03 Yes 14784 0.02 3 -0.02 [0.01] 4 0 [0.01] 0.03*** [0.01] 0.02*** [0.01] 0.05*** [0.01] 0.03*** [0.01] Yes 6457 0.03 5 -0.02 [0.01]
The table reports the coe?cient estimates from the country-pair regressions. The dependent variable is M&A volume. The explanatory variables are lagged Exchange Rate, lagged conditions of the acquirer country, and lagged conditions of the target country. Exchange Rate is in acquirer currency unit / target currency unit. In column 2-3, Real Economy Indicator is Gross Value Added and Financial Market Indicator is Stock Market Capitalization. In column 4-5, Real Economy Indicator is the ?rst principal component of {Gross Value Added, Gross Domestic Product, Capital Formation} and Financial Market Indicator is the ?rst principal component of {Stock Market Capitalization, Average M/B Ratio, Average P/E Ratio, Domestic Credit}. Also estimated but not reported are a constant term and the year ?xede?ects. Numbers in the brackets are the standard errors. All variables are standardized by the Mendoza and Terrones’ procedure. The *, **, and *** indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
75
Table 2.5B Mergers and Global Economic Conditions Y= Volume of Mergert Exchange Ratet?1 Real Economy Indicator acquirert?1 Real Economy Indicator targett?1 Financial Market Indicator acquirert?1 Financial Market Indicator targett?1 World Equity Markett?1 World Interest Ratet?1 Credit Spreadt?1 Observations R-squared 23450 0 0.09*** [0.02] 0.05*** [0.01] -0.05** [0.03] 13418 0.02 1 -0.02** [0.01] 2 -0.01 [0.01] 0.04*** [0.01] 0.03*** [0.01] 0.07*** [0.01] 0.03*** [0.01] 0.07*** [0.01] 0.03** [0.01] -0.04 [0.02] 17474 0.02 0.10*** [0.02] 0.05*** [0.01] -0.04 [0.03] 12390 0.01 3 -0.01 [0.01] 4 -0.01 [0.01] 0.03*** [0.01] 0.02*** [0.01] 0.05*** [0.01] 0.05*** [0.01] 0.16*** [0.02] 0.01 [0.03] -0.04 [0.04] 4946 0.02 5 -0.02 [0.02]
The table reports the coe?cient estimates from regressions of the MA volume between country c1 and c2 on the lagged conditions of the acquirer, c1, the lagged conditions of the target, c2, and the lagged global economic conditions. Exchange Rate is in acquirer currency unit / target currency unit. In column 2-4, real sector indicator is gross-value added and ?nancial sector indictor is market capitalization. In column 5-7, real sector indicator is the ?rst principal component of Gross Value Added, Gross Domestic Product, Capital Formation and ?nancial sector indictor is the ?rst principal component of stock market capitalization, Average M/B Ratio, Average P/E Ratio, Domestic Credit. World Equity Market is measured as the return on Morgan Stanley World Capital Index. World Interest Rate is measured as the average of US, Japan, and Germany three month treasury-bill rates. Credit Spread is computed as Moody’s AAA bond rate minus Moody’s BAA bond rate. Numbers in the brackets are the t-statistics. All variables are standardized by the Mendoza and Terrones’ procedure. *, **, and *** indicate statistical signi?cant at 10, 5, and 1 percent levels, respectively.
76
Table 2.6A Characteristics of Acquirers and Targets Compared to Their Peers Acquirer minus Target All Mergers 1 2 3 4 5 6 Cross-Border All Mergers Cross-Border All Mergers Cross-Border Acquirer vs Target (Kolmogorov-Smirnov) Acquirer minus Non-Acquirer Target minus Non-Target All Mergers 7 Cross-Border 8
Characteristics
Productivity 0.37*** [58.174] 0.3*** [46.983] 0.27*** [33.212] 0.36*** [46.303] 0.48*** [51.972] -0.03*** [-3.493] [-0.222] 0 0.06*** 0.06*** [19.327] 0.28*** 0.22*** 0.15*** [15.407] 0.18*** 0.19*** 0.14*** [18.31] [53.784] 0.27*** [84.944] 0.39*** [112.805] 0.09*** [27.822] 0.24*** 0.19*** 0.21*** 0.18*** [27.296] [56.433] 0.27*** 0.2*** 0.2*** 0.17*** [35.25] [65.26] 0.35*** 0.2*** 0.2*** 0.2*** 0.23*** [51.838] 0.18*** [41.868] 0.18*** [39.268] 0.16*** [35.525] 0.27*** [55.329] 0.13*** [29.15] -0.1*** [-19.972] -0.06*** [-12.154] -0.03*** [-4.921] -0.05*** [-8.569] -0.05*** [-8.376] -0.06*** [-11.774] -0.13*** [-12.059] -0.09*** [-8.255] -0.04*** [-3.552] 0.01 [0.602] 0.02* [1.466] -0.09*** [-8.638]
Return on Assets
Pro?t Margin
Labor Productivity
Sale Growth
77
0.03*** [4.005] 0.22*** [28.991] 0.19*** [26.887] [18.356] 0.21*** [14.402] 0.17*** 0.17*** 0.18*** 0.12*** [-0.824] 0.11*** -0.01 0.06*** 0.08***
Employment Growth
Dividend
Valuation 0.05*** [16.145] 0.12*** [37.103] 0.24*** [67.907] 0.07*** [14.857] 0.09*** [20.79] 0.21*** [42.01] 0.01*** [2.788] 0.09*** [17.639] 0.03*** [4.944] 0.07*** [6.017] 0.09*** [8.921] 0.03*** [2.922]
Market-to-Book
Past Return-1 year
Past Return-3 year
Table 2.6A (Continued) Characteristics of Acquirers and Targets Compared to Their Peers Acquirer vs Target (Kolmogorov-Smirnov) All Mergers 3 4 5 6 7 Cross-Border All Mergers Cross-Border All Mergers Acquirer minus Non-Acquirer Target minus Non-Target Cross-Border 8
Characteristics Cross-Border 2
Acquirer minus Target
All Mergers 1
Other 0.68*** [57.784] 0.26*** [14.708] 0.32*** [15.931] -0.03*** [-2.531] 0.06*** 0.06*** 0.08*** [12.437] [43.355] 0.09*** 0.14*** 0.21*** [46.002] 0.09*** 0.13*** 0.2*** 0.37*** [59.24] 0.39*** [56.154] 0.07*** [16.947] [252.874] [242.22] 0.28*** 0.29*** 0.76*** 1.04*** 0.22*** [42.124] 0.05*** [6.659] 0.09*** [11.078] 0.1*** [19.823] 0.41*** [37.821] 0.11*** [7.886] 0.07*** [4.313] 0.13*** [12.437]
Size
0.64***
[89.504]
Age1
0.17***
[15.742]
Age2
0.16***
[13.8]
Leverage
-0.04***
[-5.985]
78
Observations are grouped by country-industry-year. Z-scores are computed from the mean and the standard deviation of each group. Column 1
reports the di?erence between the average Z-scores of all acquirers and the average Z-score of all targets. Column 2 reports the di?erence between
the cross-border acquirers and the cross-border targets. The numbers in the parenthesis are the t-statistics from the t-tests. Column 3 reports the
distributional distances (the Kolmogorov-Smirnov’s D-statistics) between all acquirers and all targets. Column 4 reports the distributional distances
between the cross-border acquirers and the cross-border targets. Column 5-6 compares the acquirers with the non-acquirers and column 7-8 compares
the targets with the non-targets. The numbers in the parenthesis are the t-statistics from the t-tests. *, **, and *** indicate statistical signi?cant
at 10, 5, and 1 percent levels, respectively.
Table 2.6B Comparisons of Acquirers and Targets Acquirer minus Target All Mergers 1 2 3 4 5 6 Cross-Border All Mergers Cross-Border All Mergers Cross-Border Acquirer vs Target (Kolmogorov-Smirnov) Acquirer minus Non-Acquirer Target minus Non-Target All Mergers 7 Cross-Border 8
Characteristics
Productivity 0.28*** [46.262] 0.2*** [32.282] 0.19*** [23.513] 0.37*** [47.329] 0.53*** [54.454] 0.16*** [22.034] [16.303] 0.2*** 0.24*** 0.11*** [22.795] 0.35*** 0.18*** 0.2*** [17.591] 0.21*** 0.16*** 0.14*** [14.271] [39.805] 0.29*** [90.633] 0.45*** [127.438] 0.08*** [26.586] 0.18*** 0.15*** 0.2*** 0.13*** [21.076] [45.881] 0.18*** 0.1*** 0.15*** 0.14*** [32.445] [67.963] 0.28*** 0.17*** 0.18*** 0.21*** 0.26*** [59.93] 0.17*** [38.101] 0.16*** [34.316] 0.18*** [40.218] 0.32*** [64.735] 0.08*** [18.828] -0.02*** [-3.053] -0.01 [-1.969] -0.01 [-0.969] -0.04*** [-7.3] -0.04*** [-5.673] -0.03*** [-6.534] -0.02** [-2.114] -0.01 [-0.961] 0 [-0.381] 0 [0.188] 0.02 [1.229] -0.03*** [-2.675]
Return on Assets
Pro?t Margin
Labor Productivity
Sale Growth
79
0.07*** [10.159] 0.01* [1.566] 0.21*** [28.831] [16.861] 0.2*** [-2.906] 0.15*** 0.19*** -0.03*** 0.11*** [9.741] 0.13*** 0.11*** 0.06*** 0.08***
Employment Growth
Dividend
Valuation 0.12*** [36.681] 0.28*** [78.027] 0.11*** [36.742] 0.09*** [19.878] 0.24*** [47.424] 0.19*** [44.245] 0.11*** [21.396] 0.08*** [13.725] -0.04*** [-7.248] 0.11*** [10.77] 0.04*** [3.728] 0 [-0.4]
Market-to-Book
Past Return-1 year
Past Return-3 year
Table 2.6B (Continued) Comparisons of Acquirers and Targets Acquirer vs Target (Kolmogorov-Smirnov) All Mergers 3 4 5 6 7 Cross-Border All Mergers Cross-Border All Mergers Acquirer minus Non-Acquirer Target minus Non-Target Cross-Border 8
Characteristics Cross-Border 2
Acquirer minus Target
All Mergers 1
Other 0.75*** [63.407] 0.29*** [16.128] 0.36*** [17.461] -0.04*** [-3.456] 0.06*** 0.06*** 0.03*** [10.904] [47.744] 0.1*** 0.17*** 0.23*** [44.363] 0.09*** 0.16*** 0.2*** 0.44*** [69.099] 0.44*** [63.559] 0.01* [1.545] [240.488] [242.725] 0.28*** 0.32*** 0.73*** 1.05*** 0.23*** [42.82] 0.08*** [11.507] 0.07*** [8.735] 0.06*** [11.681] 0.36*** [32.972] 0.17*** [12.079] 0.09*** [5.605] 0.07*** [6.83]
Size
0.63***
[86.414]
Age1
0.14***
[12.889]
Age2
0.19***
[15.729]
Leverage
-0.04***
[-6.255]
80
Observations are grouped by country-year. Z-scores are computed from the mean and the standard deviation of each group. Column 1 reports
the di?erence between the average Z-scores of all acquirers and the average Z-score of all targets. Column 2 reports the di?erence between the
cross-border acquirers and the cross-border targets. The numbers in the parenthesis are the t-statistics from the t-tests. Column 3 reports the
distributional distances (the Kolmogorov-Smirnov’s D-statistics) between all acquirers and all targets. Column 4 reports the distributional distances
between the cross-border acquirers and the cross-border targets. Column 5-6 compares the acquirers with the non-acquirers and column 7-8 compares
the targets with the non-targets. The numbers in the parenthesis are the t-statistics from the t-tests. *, **, and *** indicate statistical signi?cant
at 10, 5, and 1 percent levels, respectively.
Table 2.6C Comparisons of Merging Firms during Booms and Busts Real Booms minus Real Busts All Acq 1 2 3 4 5 6 7 Cross-Border Acq All Targets Cross-Border Targets All Acq Cross-Border Acq All Targets Financial Booms minus Financial Busts Cross-Border Targets 8
Characteristics
Productivity 0.04*** [5.108] 0.04*** [5.429] 0.13*** [13.696] 0.23*** [21.791] 0.3*** [24.95] 0.02* [1.569] [-1.44] [8.791] [4.855] -0.02* 0.16*** 0.17*** [19.379] [10.006] [6.648] 0.31*** 0.22*** 0.31*** [17.015] [12.041] [5.896] 0.23*** 0.23*** 0.24*** 0.39*** [35.682] 0.39*** [31.18] 0.08*** [8.548] [12.279] [8.846] [1.51] [11.264] 0.16*** 0.18*** 0.07* 0.12*** [3.938] [7.057] [-0.145] [2.106] [1.57] 0.13*** [8.745] 0.34*** [21.986] 0.42*** [22.81] 0.04*** [2.646] 0.04*** 0.14*** -0.01 0.02** 0.02* [4.891] [9.793] [1.512] [5.298] [3.158] 0.05*** 0.2*** 0.07* 0.05*** 0.04*** 0.04** [1.793] 0.03 [1.159] 0.1*** [4.225] 0.33*** [15.756] 0.32*** [13.174] 0.14*** [7.699] 0.04 [0.883] -0.06* [-1.34] 0.05 [1.086] 0.23*** [5.419] 0.38*** [7.51] 0.11*** [2.894]
Return on Assets
Pro?t Margin
Labor Productivity
Sale Growth
81
0.03*** [3.166] -0.17*** [-17.634] 0.06*** [6.012] [-5.934] -0.08*** [-18.474] [-6.113] 0.17*** [8.398] -0.26*** -0.11*** [-0.029] [5.543] 0 0.1*** 0.22*** [5.509] -0.23*** [-5.702] 0.07** [1.73]
Employment Growth
Dividend
Valuation 0.31*** [29.014] -0.08*** [-7.499] 0.17*** [16.7] 0.38*** [21.306] -0.16*** [-9.579] 0.13*** [8.524] 0.31*** [15.207] -0.01 [-0.669] 0.26*** [11.948] 0.45*** [9.65] 0.01 [0.118] 0.26*** [5.912]
Market-to-Book
Past Return-1 year
Past Return-3 year
Table 2.6C (Continued) Comparisons of Merging Firms during Booms and Busts Real Booms minus Real Busts Cross-Border Acq 2 3 4 5 6 7 All Targets Cross-Border Targets All Acq Cross-Border Acq All Targets Financial Booms minus Financial Busts Cross-Border Targets 8
Characteristics
All Acq 1
Other -0.12*** [-8.604] 0.07*** [3.091] 0.05** [2.059] -0.08*** [-6.883] [-7.583] [-3.844] [2.01] -0.13*** -0.14*** 0.02** [-1.579] [-4.355] [-3.084] -0.05* -0.27*** -0.05*** -0.04* [-1.644] 0 [0.375] [-0.779] [-0.831] [-8.209] [-5.758] -0.02 -0.04 -0.13*** -0.14*** [0.644] [-1.952] [0.331] [-0.397] 0.01 -0.07** 0 -0.01 0.02 [1.222] -0.07*** [-2.878] -0.01 [-0.279] -0.09*** [-4.857] 0.02 [0.553] -0.12*** [-2.371] -0.08* [-1.291] -0.09** [-2.247]
Size
-0.11***
[-11.108]
Age1
-0.02
[-1.211]
Age2
-0.06***
[-3.62]
Leverage
-0.03***
[-3.145]
82
Observations are grouped by country-industry. Z-scores are computed from the mean and the standard deviation of each group. Columns 1 and
5 report the di?erence between the average Z-scores of acquirers during booms and the average Z-score of acquirers during busts. Columns 2 and
6 report the di?erence between the cross-border acquirers during booms and the cross-border acquirers during busts. Columns 3 and 7 report
the di?erence between the targets during booms and the targets during busts. Columns 4 and 8 report the di?erence between the cross-border
targets during booms and the cross-border targets during busts. Booms (Busts) in Columns 1-4 are the periods in which HP detrended Gross Value
Added is above (below) one (minus one) standard deviation. Booms (Busts) in Columns 5-8 are the periods in which HP detrended Stock Market
Capitalization is above (below) one (minus one) standard deviation. The numbers in the parenthesis are the t-statistics from the t-tests. *, **, and
*** indicate statistical signi?cant at 10, 5, and 1 percent levels, respectively
Table 2.6D Post-Merger Operating Performance Unmatched Cross-Border Acq 2 -3.5*** [-7.36] -7.13*** [-12.55] -9.86*** [-15.4] -10.95*** [-15.53] [1.48] [1.44] [0.93] 1.52* 2.83* 1.03 [1.25] [1.34] [0.39] [-2.23] -6.26*** [-4.51] 1.15* 2.39* 0.38 -2.84** [-0.64] [-3.2] [1.3] [-0.14] -0.51 -5.02*** 1.09* -0.16* [-1.74] [-1.67] [1.01] [1.01] [0.05] 0.66 [0.44] 2.61* [1.54] 1.14 [0.58] -1.13** -2.14** 0.72 0.95 0.06 3 4 5 6 7 All Targets Cross-Border Targets All Acq Cross-Border Acq All Targets Propensity Score Matched Cross-Border Targets 8 -0.7 [-0.32] -6.92** [-2.3] 0.75 [0.21] 3.71 [0.99]
Time Windows
All Acq 1
t-1 to t
-3.17***
[-8.59]
t-1 to t+1
-4.78***
[-10.81]
t-1 to t+2
-6.07***
[-12.18]
t-1 to t+3
-6.05***
[-10.94]
83
Pro?tability is measured as ROA. t denotes the year of acquisition. Column 1 reports the di?erence between the average pro?tability changes of all
acquirers and the average changes of the non-acquirers. Column 2 reports the di?erence between the cross-border acquirers and the non-acquirers.
Column 3 reports the di?erence between the targets and the non-targets. Column 4 reports the di?erence between the cross-border targets and
the non-targets. Columns 5-8 reports propensity-score-matched e?ects (average treatment e?ects on the treated). Propensity scores are computed
from the Probit model using the following covariates: lagged pro?tability, sales, total assets, and ?rm age. The numbers in the parenthesis are the
t-statistics from the t-tests. *, **, and *** indicate statistical signi?cant at 10, 5, and 1 percent levels, respectively.
Table 2.7A Cross-Border Mergers and Industry Shocks in Acquirer Countries 1 0.08*** [0.01] 0.04*** [0.01] 0.15*** [0.01] [0.01] 0.15*** [0.01] Yes 10537 0.01 0.01 0.04 0.01 0.01 0.04 8519 8519 9029 7198 7198 Yes 10261 0.02 10198 0.04 0.10*** [0.01] [0.01] [0.01] 0.09*** [0.01] 0.12*** [0.01] Yes 10198 0.05 8515 0.01 4739 0.02 0.09*** [0.01] 0.08*** [0.01] 0.08*** [0.01] 0.08*** [0.01] 0.05*** [0.02] Yes 4739 0.05 0.02 0.03*** 0.01 [0.01] [0.01] [0.01] [0.01] [0.01] 0.09*** 0.06*** 0.07*** 0.06*** 0.04*** 2 3 4 5 6 7 8 9 10 11 12
Y=Cross-border Mergers
Real Industry Indicatori,c,t?1
Real Economy Indicatorc,t?1
Industry Valuation Indicatori,c,t?1
Financial Market Indicatorc,t?1
Year Fixed E?ects
Observations
R-squared
84
The table reports the coe?cient estimates from regressions of the cross-border M&A volume on the lagged conditions of the acquirer industry and
country. In column 1-3 and column 7-9, the industry productivity shock is average ROA, the real sector indicator is gross value added, the industry
valuation shock is average M/B, and the ?nancial market indicator is stock market capitalization. In column 4-6 and column 10-12, the industry
productivity shock is the ?rst principal component of 6 productivities measures in Table 2.6, the real sector indicator is the ?rst principal component
measure from section 2.4, the industry valuation shock is the ?rst principal component of 3 valuation measures in Table 6, and the ?nancial market
indicator is the ?rst principal component measure from section 2.4. Also estimated but not reported are a constant term and the year ?xed-e?ects.
Numbers in the brackets are the standard errors. All variables are standardized by the Mendoza and Terrones’ procedure. The *, **, and ***
indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
Table 2.7B Cross-Border Mergers and Industry Shocks in Target Countries 1 0.05*** [0.01] 0.05*** [0.01] 0.06*** [0.01] 0.09*** [0.01] Yes 11557 0 0 0.03 0 0 0.03 9329 9329 9621 7597 7597 Yes 11180 0 11106 0.01 [0.01] 0.03** [0.01] [0.01] [0.01] 0.02 [0.01] 0.02** [0.01] Yes 11106 0.03 9236 0 4770 0.02 0.02*** [0.01] 0.06*** [0.01] 0.07*** [0.01] 0.06*** [0.02] 0.04** [0.02] Yes 4770 0.04 0.03** 0.03*** 0.02** [0.01] [0.01] [0.01] [0.01] [0.01] 0.05*** 0.03*** 0.03*** 0.02** 0.01 2 3 4 5 6 7 8 9 10 11 12
Y=Cross-border Mergers
Real Industry Indicatori,c,t?1
Real Economy Indicatorc,t?1
Industry Valuation Indicatori,c,t?1
Financial Market Indicatorc,t?1
Year Fixed E?ects
Observations
R-squared
85
The table reports the coe?cient estimates from regressions of the cross-border M&A volume on the lagged conditions of the target industry and
country. In column 1-3 and column 7-9, the industry productivity shock is average ROA, the real sector indicator is gross value added, the industry
valuation shock is average M/B, and the ?nancial market indicator is stock market capitalization. In column 4-6 and column 10-12, the industry
productivity shock is the ?rst principal component of 6 productivities measures in Table 6, the real sector indicator is the ?rst principal component
measure from section 2.4, the industry valuation shock is the ?rst principal component of 3 valuation measures in Table 2.6, and the ?nancial market
indicator is the ?rst principal component measure from section 2.4. Also estimated but not reported are a constant term and the year ?xed-e?ects.
Numbers in the brackets are the standard errors. All variables are standardized by the Mendoza and Terrones’ procedure. The *, **, and ***
indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
Table 2.7C Domestic Mergers and Industry Shocks 1 0.04*** [0.01] 0.01 [0.01] 0.07*** [0.01] 0.10*** [0.01] Yes 11403 0 0 0.01 0 0 0.02 0 9197 9197 9533 7528 7528 11054 Yes 10981 0.01 [0.01] 0.04*** 0.04*** [0.01] 0.10*** [0.01] Yes 10981 0.02 9152 0 4749 0.01 [0.01] [0.01] [0.01] 0.05*** [0.01] 0.06*** [0.01] 0.04*** [0.01] 0.08*** [0.01] 0.02 [0.02] Yes 4749 0.03 0 0.01 0 [0.01] [0.01] [0.01] [0.01] [0.01] 0.04*** 0.03** 0.04*** 0.04*** 0.03*** 2 3 4 5 6 7 8 9 10 11 12
Y= Domestic Mergers
Industry Productivityi,c,t?1
Real Economy Indicatorc,t?1
Industry Valuationi,c,t?1
Financial Market Indicatorc,t?1
Year Fixed E?ects
Observations
R-squared
86
The table reports the coe?cient estimates from regressions of the domestic M&A volume on the lagged conditions of the domestic industry and
country. In column 1-3 and column 7-9, the industry productivity shock is average ROA, the real sector indicator is gross value added, the industry
valuation shock is average M/B, and the ?nancial market indicator is stock market capitalization. In column 4-6 and column 10-12, the industry
productivity shock is the ?rst principal component of 6 productivities measures in Table 2.6, the real sector indicator is the ?rst principal component
measure from section 2.4, the industry valuation shock is the ?rst principal component of 3 valuation measures in Table 6, and the ?nancial market
indicator is the ?rst principal component measure from section 2.4. Also estimated but not reported are a constant term and the year ?xed-e?ects.
Numbers in the brackets are the standard errors. All variables are standardized by the Mendoza and Terrones’ procedure. The *, **, and ***
indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
Table 2.8 Cross-Sectional Gravity Model Y= Log of Aggregate Volume of Merger (from 1989-2008) Log of Distance 1 -0.58*** [0.08] Common Language Dummy 2 -0.53*** [0.08] 1.77*** [0.16] Log of Populationacquirer,1988 Log of Populationtarget,1988 Log of Real GDPacquirer,1988 Log of Real GDPtarget,1988 Log of Market Capitalizationacquirer,1988 Log of Market Capitalizationtarget,1988 Constant Observations R-squared Yes 1052 0.04 Yes 1052 0.14 Yes 1052 0.16 Yes 1015 0.38 3 -0.62*** [0.08] 1.64*** [0.16] 0.19*** [0.05] 0.18*** [0.06] 4 -0.61*** [0.07] 1.21*** [0.15] 0.73*** [0.05] 0.45*** [0.06] 1.11*** [0.06] 0.53*** [0.06] 5 -0.84*** [0.07] 1.23*** [0.14] 0.69*** [0.05] 0.56*** [0.06] 0.09 [0.12] 0.32*** [0.12] 0.65*** [0.07] 0.15** [0.06] Yes 875 0.46
The table reports the coe?cient estimates from regressions of the 1989-2008 aggregate volume of M&A ?ow from country c1 to country c2 on the conditions of the acquirer, c1 , and the conditions of the target, c2 , in 1988. Distance and Common Language dummy are from Di Giovanni (2005). Population, GDP, and Stock Market Capitalization are from the World Development indicator Database. Numbers in the brackets are the standard errors. The *, **, and *** indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
87
Table 2.9 1988 Country Characteristics Variable Distance Common Language Dummy Population Real GDP Stock Market Capitalization Mean 8.696378 0.207841 16.99215 8.732422 7.200922 Standard Deviation 0.911802 0.405871 1.383335 1.245854 2.339167 Minimum 5.41272 0 14.86143 5.692476 0.419264 Maximum 9.895177 1 20.82006 10.35871 10.60626
The table reports the summary statistics of country characteristics in 1988. Distance and Common Language dummy are from di Giovanni (2005). Population, GDP, and Stock Market Capitalization are from the World Development Indicator Database. All the variables except the common language dummy are in log form.
88
Table 2.10 Characteristics of Domestic versus Cross-Border Deals Deal Characteristics All Deal Average 1 Deal Size 165.39 Domestic Deal Average 2 157.27 Cross-Border Deal Average 3 193.86 Cross-Border - Domestic 4 36.58*** [4.89] Prob(Cash Deals) 0.51 0.49 0.55 0.06*** [4.39] Prob(Listed Acquirer) 0.45 0.43 0.53 0.11*** [57.45] Prob(Listed Target) 0.14 0.15 0.13 -0.02*** [-18.81] Prob(Tradable Industry) 0.2 0.17 0.29 0.13*** [85.32] Prob(High-Tech Industry) 0.11 0.1 0.13 0.03*** [25.55] Relatedness 1106.69 1122.11 1053.35 -68.75*** [-11.46] Controlled for Fixed E?ects 5 40.11*** [4.7] 0.05*** [3.18] 0.11*** [58.8] 0 [1.27] 0.12*** [76.29] 0.05*** [41.02] -94.89*** [-14.24]
Deal Size is the transaction value in million of current dollars. Cash Deals is a dummy variable taking the value of one if the percentage of cash is higher than the percentage of stock. Listed Acquirer is a dummy taking the value of one if the acquirer is listed. Listed Target is a dummy taking the value of one if the target is listed. Tradable is equal to one if the acquirer and the target are in the tradable industries as de?ned by Aguiar and Gopinath (2005). High-Tech is equal to one if the acquirer and the target are in the high-tech industry according to the American Electronic Association. Relatedness is the absolute value of the di?erence between the acquirer’s 4-digit SIC and the target’s 4-digit SIC. Columns 1, 2 and 3 shows the average characteristics of all deals, the domestic deals, and the cross-border deals. Column 4 shows the di?erences between the domestic deals and the cross-border deals. Column 5 shows coe?cients of the cross-border dummy after controlling for the ?xed-e?ects. Numbers in the brackets are t-statistics. The *, **, and *** indicate statistical signi?cance at the 10, 5, and 1 percent levels, respectively.
89
Figure 2.1 Volume of Aggregate Mergers
The ?gure shows the aggregate volume of M&A activities from the top 50 countries in trillion of current dollars.
90
Figure 2.2A US Acquisition of Foreign Firm
Figure 2.2B US Acquisition of Foreign Firm: Hodrick-Proscott Detrended
The ?gures show an example of the raw and the detrended series of US ?rm acquisitions of assets in other countries.
91
Appendix 2.1: The Country Coverage of WorldScope
Full-Coverage Developed countries include Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Hong Kong, Ireland, Italy, Japan, the Netherlands, Norway, Portugal, Singapore, Spain, Sweden, Switzerland, the United Kingdom, and the United States. Emerging markets include Brazil, China, Indonesia, Korea, Malaysia, Mexico, Philippines, South Africa, and Thailand. Targeted Coverage Countries include Argentina, Chile, Colombia, Czech Republic, Egypt, Hungary, India, Israel, Jordan, New Zealand, Peru, Poland, Russia, Slovakia, Turkey, and Venezuela.
92
Chapter 3
A Dynamic Model of International Mergers and Acquisitions
3.1 Introduction
In the past two decades, 26% of worldwide M&A activities involve acquirers and targets from di?erent countries. The aggregate volume of cross-border mergers from 1989 to 2008 adds up to above 8 trillion dollars. In spite of such a large volume, much of the M&A literature focuses on domestic mergers. Moreover, the amount of cross-border mergers varies greatly from year to year. For example, the volume of worldwide M&A deals dropped by 62% from 2000 to 2003 but bounced back by 158% in 2006. Despite such a large year-toyear ?uctuation, most existing papers on cross-border M&As study the e?ects of long-run determinants like corporate governance and capital market development. These gaps in the literature motivate the research questions that are at the core of this paper: what are the dynamic patterns of cross-border mergers, and what are the factors that drive them? In the previous chapter, I present key facts about international mergers. Speci?cally, I answer these four main questions: (1) How do cross-border mergers behave over a business cycle? International mergers come in waves and are very pro-cyclical. (2) Where do shocks that e?ect cross-border mergers originate? Most mergers occur when both the acquirer and
93
the target economies are booming. (3) What type of shocks (real or ?nancial) e?ect crossborder mergers? Merger booms have industry-level (productivity shock) and country-level (?nancial shock) components. (4) What types of ?rms engage in cross-border mergers? Acquirers tend to be more productive than average ?rms and targets tend to be less productive than average ?rms. In this chapter, I use the four empirical facts mentioned earlier as a guideline and build a dynamic structural model of cross-border mergers. The dynamic structural approach o?ers two major advantages. First, by construction, it solves the identi?cation problem inherent in reduced-form estimation. Using simulated data, I can quantify the e?ects of productivity and ?nancial shocks on endogenous variables. Second, the dynamic structural model provides me with an analytical framework to investigate the impacts of various government policies. As an example of such policy analyses, I examine the impact of multinational corporation taxation which has long been the subject of heated policy debates. My model is related to Gomes and Livdan (2004) and Yang (2008) in that ?rms make investment and merger decisions based on the productivity shocks they received. To investigate the e?ects of ?nancial shocks, I incorporate external ?nancing cost similar to the ones in Gomes (2001) and Whited (2006) and allow the cost to ?uctuate along a business cycle. The distinguishing features of my model are that there are two countries and that a local ?rm has an option of engaging in cross-border mergers in order to become a multinational corporation. I also assume that the productivity shock has two components: ?rm-speci?c and location-speci?c. With this setup, productive ?rms will seek assets in booming locations. Recently, President Obama proposed a 200 billion dollar tax increase on multinational corporations. As a consequence, the largest US corporations are concerned that the tax raise will put American ?rms at a disadvantage overseas and leave them vulnerable to foreign acquisitions.1 Others are concerned that the tax will primarily impact the productive sectors, such as technology and pharmaceutical industries.2 Clearly, there is an urgent need to understand the e?ects of multinational taxation. The simulation results from my tax
1 2
BusinessWeek May 4, 2009 The New York Times May 5, 2009
94
experiments con?rm that foreign operation tax can be very distortionary for cross-border mergers and has larger e?ects on more productive ?rms. The model also provides a policy implication: when analyzing the e?ects of multinational corporation taxation, we should be careful not to focus only on multinational ?rms that already have overseas operations. Since cross-border mergers are sensitive to tax rate, we must also take into consideration its e?ects on productive local ?rms for whom the tax is a disincentive for future mergers. This paper joins the growing literature on dynamic corporate ?nance (e.g., Whited, 2006; Gomes and Livdan, 2004; Yang, 2008). The dynamic structural approach is particularly appropriate for my context, since merger waves are, by nature, dynamic phenomena. The dynamic simulations in this paper are also related to the recent work on the impact of multinational corporation taxation. While the literature on taxation is voluminous, identifying the e?ect of taxes can pose a challenge since the tax policies are likely to be endogenous. Even if the tax policies are exogenous, their e?ects, as measured by the reduced-form coe?cients, might still be endogenous according to the Lucas’ critique. Dharmapala, Foley, and Forbes (2009) and Faulkender and Peterson (2009) use the Homeland Investment Act of 2004 as a natural experiment and analyze the e?ects of this one-time tax break on U.S. ?rms. In this paper, I o?er structural estimation as an alternative approach to address the identi?cation problem.
3.2
Conceptual Framework
The goal of chapter 3 is to integrate the ?ndings from chapter 2 into a dynamic structural model and policy analysis. Because implications from a structural model are, to a great extent, driven by its structural assumptions, the strength of my analysis lies in the fact that my model is consistent with the key facts derived from a large amount of data. Guided by the reduced-form evidence, I develop a dynamic structural model of crossborder mergers. I assume that ?rm decisions are driven by productivity shocks under the presence of ?nancial frictions. Merger gain comes from access to the target country’s markets and resources as well as the utilization of acquirer ?rm-speci?c assets. With these assumptions, productive ?rms will seek assets in booming locations. Then, I prove that a 95
solution to the problem indeed exists and characterize the properties of the model. These properties can provide insights into ?rms’ merger and investment decisions. Next, I solve the model numerically using the value function iteration algorithm. Given the value functions and the policy functions, I construct a panel of ?rms, generate structural shocks, and observe how ?rms react to these shocks. These exercises allow me to quantify the e?ects of productivity and ?nancial shocks on the endogenous variables. In addition, I use the model to perform policy experiments on taxation. The issue of multinational corporation taxation frequently captures public attention. On May 4, 2009, President Obama proposed a 200 billion dollar tax increase on multinational corporations. As a consequence, the largest U.S. corporations have launched a vigorous lobbying e?ort against the plan. One of the arguments against the tax increase is that it will put American ?rms at a competitive disadvantage overseas and leave them vulnerable to foreign acquisitions. There is also some concern that the tax will primarily impact the productive sectors, such as the technology and pharmaceutical industries. Given my structural model, I can investigate how the tax on foreign operations might in?uence ?rm investment and merger decisions. I can also verify whether the concerns about the tax proposal above are valid.
3.3
The Model and its Basic Properties
I build a neoclassical model of cross-border investments. Firms make investment and production decisions in the presence of productivity shocks. The model is related to the domestic investment models in Gomes and Livdan (2004), Cooper and Haltiwanger (2006), as well as in Yang (2008). The distinguishing feature of this model is that I allow domestic ?rms to acquire establishments in another country and become multinationals. Posit that there are two countries, A and B . In each country, there are a large number of ?rms so that each ?rm is a price-taker in the market for corporate assets. The model is in discrete time, and one period is de?ned as one year. Technology 96
The pro?t function is described by ?( , K ). In particular, the function is ?( , K ) = e K ? , where is the level of the productivity shock and ? is the curvature of the pro?t
function. The ? is assumed to be less than one so that the production function exhibits the decreasing-return-to-scale property.3 This property captures the concept that local resources and local markets are limited and that there is an incentive for local ?rms to expand to another country. The productivity shock ( =
i
+
S)
has two components. The
i,
which is ?rm-speci?c,
captures the ?rm-level shocks that cannot easily be traded or transferred outside of the ?rm, such as patents, know-how, managerial skills, and reputation. The
S
?{
A , B },
which is speci?c to the location where ?rms operate, captures any country-speci?c factors that can e?ect ?rm pro?ts, such as local input prices, proximity to customers, and other institutional environments. Given the structure of the shocks, more productive ?rms (high acquire assets, and less productive ?rms (low
i) i)
are more likely to
are more likely to sell their assets. Firms
S,
will also want to invest in a country where the country shock, be so.
is high and expected to
Firms are not certain about their future productivities.4 For the calibration, I assume that each follows an AR(1) process: =? + ei,t and ei,t ? N (0, ?i ); + eS,t and eS,t ? N (0, ?S ).
i,t S,t
i,t?1 S,t?1
=?
Firm Organization Firms are risk neutral and maximize the expected present value of dividend streams over an in?nite time horizon. There are two types of ?rms: single-country ?rms and multinational ?rms. A single-country ?rm operates an establishment in one country, S ? {A, B }, but
3
This pro?t function is a shorthand version of a large class of production processes. For example,
Cooper and Haltiwanger (2006) show that it can be derived from the production processes that involve more than one type of inputs. It also allows imperfect competitions in the product markets. 4 For the proof, I only assume that the transition matrix governing the dynamic of has the Feller property.
97
has the option of acquiring ?rms in another country and becoming a multinational. At the beginning of each period, single-country ?rms observe the productivity shocks and choose whether to stay local or not. Then, they decide how much capital they are going to buy or sell in that period. A single-country ?rm i in country S that chooses to remain local has a value function de?ned as VSS (X, KS ) = max [dSS + ?E [VS (X , KS )]], (eq1)
{KS }
where dSS = ?( i +
S , KS )
? (KS ? (1 ? ? )KS ) ? ?(KS , KS ),
i.e., dividend = operating pro?t - asset purchase - adjustment cost. In this equation, d stands for dividend. Subscript SS denotes a single-country ?rm in country S that decides to stay in country S . Firm i’s establishment in country S has productivity
i
+
S.
The exogenous state variable is X = { i ,
A , B }.
The value function
has the productivity of the foreign country as an argument even though it does not have an establishment there. This is because the foreign country’s productivity e?ects the option value of becoming a multinational. The prime variables represent the future values, while other variables represent current values. The 0 < ? < 1 is a discount factor. The capital stock depreciates at the exogenous rate 0 < ? < 1. The cost of investment has two components: the direct cost of capital goods and the quadratic adjustment cost. The direct capital expenditure is given by K ? (1 ? ? )K . The
?? )K 2 ) K . The parameter ? re?ects quadratic adjustment cost is ?(K , K ) = ?/2( K ?(1 K
imperfections in the market for real assets, such as the transaction cost of purchasing and liquidating capital, as well as other real costs associated with change in the level of capital stocks, such as the disruption in the production processes. A multinational has the value function: VM (X, KA , KB ) = where dM = max [dM + [VM (X , KA , KB )]], (eq2) +
S , KS )
{KA ,KB }
S ={A,B } (?( i
? (KS ? (1 ? ? )KS ) ? ?(KS , KS )).
Subscript M denotes a multinational ?rm. A multinational ?rm has two establishments, 98
one in country A and one in country B . At the beginning of each period, multinational ?rms observe the productivity shocks, X , and decide how much capital they are going to buy or sell in each country. Merger Process When single-country ?rms decide whether to remain local or to go abroad, they compare the expected net bene?ts of each alternative. Therefore, the value function of a singlecountry ?rm is: VS (X, KS ) = max [VSS (X, KS ), VSM (X, KS )]. (eq3) Subscript S denotes a single-country ?rm. Subscript SM denotes a single-country ?rm in country S that decides to become a multinational in the next period. The value function of the single-country ?rm that chooses to remain local, VSS (X, KS ), is de?ned by (eq1). The single-country ?rm that chooses to acquire production capacity in another country has the value function of: If S = A, then VAM (X, KA ) = max [dAM + ?E [VM (X , KA , f )]], (eq4A)
{KA }
dAM = ?( i +
A , KA )
? (KA ? (1 ? ? )KA ) ? ?(KA , KA ) ? F , or
{KB }
If S = B , then VBM (X, KB ) = max [dBM + ?E [VM (X , f, KB )]], (eq4B) dBM = ?( i +
B , KB )
? (KB ? (1 ? ? )KB ) ? ?(KB , KB ) ? F .
In order to become a multinational, a single-country ?rm has to pay a one-time ?xed cost, F . After paying F at time t, the single-country ?rm will become a multinational at time t+1. The F captures the idea that investing in a foreign country is more di?cult than investing domestically. The new multinational ?rm will start with toe-hold capital f , 0 < f < F , in the foreign country. Therefore, F re?ects the price of the toe-hold capital combined with other costs of international mergers such as costs of due diligence, costs of setting up new headquarters, and fees for foreign consultants. Under these assumptions, FDIs can be thought of as cross-border mergers. This is consistent with the existing evidence that most FDIs are in the form of cross-border M&As.5
5
According to UNCTAD’s FDI database, from 1988-2006, 62% of global FDIs are in the form of cross-
border M&As.
99
[INSERT FIGURE 3.1 HERE] The timeline of ?rm investment and merger decisions is given in Figure 3.1. Before analyzing and calibrating the model, I need to ensure that the dynamic programming problems (eq1) and (eq2) have a solution and that VM (X, KA , KB ), VSS (X, KS ), and VSM (X, KS ) exist. Let C (X × K ) and C (X × K × K ) be the space of all bounded and continuous functions in (X × K ) and (X × K × K ), respectively. Existence Proposition 1: There exists a unique continuous function VM (X, KA , KB ) that solves the dynamic programming problem (eq2). See the Appendix for the proof The proof is a direct application of Blackwell’s su?cient conditions for a contraction mapping. From theorem 9.7 and 9.11 in Stokey, Lucas, and Prescott (1989), VM (X, KA , KB ) is also increasing in all its arguments. The solution VM (X, KA , KB ) produces the policy function KA and KB , which determine a multinational’s optimal level of investment in country A and country B . Proposition 2: There exists a unique continuous function VS (X, KS ) that solves the dynamic programming problem (eq1), the maximization problems (eq3), and (eq4). See the Appendix for the proof The proof of Proposition 2 is more complicated than Proposition 1’s because VS (X, KS ) can be mapped to either VSS (X, KS ) or VSM (X, KS ), depending upon the values of the state variables. The outline of the proof is as follows: (1) From Proposition 1, there exists a unique function VM (X, KA , KB ) in C (X × K × K ) that solves the multinational dynamic programming problem (eq2). (2) Because the maximization problem of the single-country ?rm that chose to become multinational, (eq4) only involves the function VM (X, KA , KB ), there exists a function VSM (X, KS ) in C (X × K ) that solves (eq4). (3) Next, I apply Blackwell’s su?cient conditions for a contraction mapping for the dynamic programming problem of the single-country ?rm that chose to remain local (eq5).
100
Therefore, there exists a unique function VSS (X, KS ) in C (X × K ) that solves (eq 5). (4) Finally, VS (X, KS ) = max [VSS (X, KS ), VSM (X, KS )] exists and is in C (X × K ), because both VSS (X, KS ) and VSM (X, KS ) are in C (X × K ). The solutions VSS (X, KS ) and VSM (X, KS ) also produce the policy function KS , which determines the domestic ?rm’s optimal level of investment in country S . Investment Euler’s Equation From (eq1) to (eq4), I can characterize the optimal level of investment by deriving the ?rst order conditions and applying the envelope theorem. Proposition 3: The optimal levels of investment in country S of multinationals and singlecountry ?rms are governed by the following Euler’s equation: 1 + ?KS ?(KS , KS ) = [?KS ?( i + See the Appendix for the proof At the optimum, ?rms equate the marginal cost and marginal bene?t of investment. Investing an additional unit of capital costs one plus the marginal adjustment costs. The gain from that additional unit of capital consists of the expected present value of the marginal product of capital, the value of capital left from depreciation, and the marginal e?ect that capital has on next period’s adjustment cost. From Euler’s equation, multinationals operate establishments in two locations as if they are two independent ?rms. Gains from entering another country will depend upon the acquirer’s ?rm-speci?c productivity and the location-speci?c productivity of the target country. This proposition shows that the merger gains in this model come from the utilization of an acquirer’s ?rm-speci?c assets and access to goods and factor markets in the target country. Costly External Financing External ?nancing is more costly than internal ?nancing. In particular, when ?rms raise external capital (dividend is less than zero), it has to pay the cost of external ?nance ?(d) = ?0 + ?1 d, where d stands for dividend. This linear speci?cation is frequently seen in the ?nance and macroeconomic literature. For example, Gomes (2001) assumes that ?(d) is 0.08 ? 0.028d and Whited (2006) assumes that ?(d) is 0.04 ? 0.0264d. The function ?(d) can be thought of as the transaction costs of accessing external equity markets, 101
S , KS )
+ (1 ? ? ) ? ?KS ?(KS , KS )].
such as the cost of an IPO, as well as the premium for agency problems or asymmetric information problems associated with external ?nancing, such as the cost of monitoring the ?rms.6 With costly external ?nancing, Proposition 3 does not hold: investment and mergers in one country might depend on productivity shocks of another country. To illustrate the importance of ?nancial constraint, I consider the extreme case in which only internal ?nancing is possible. Proposition 4: If external ?nancing is prohibitively costly (i.e., ?(d) approaches ?), the optimal level of investment of a multinational satis?es the following conditions:
E [?K ?( i +
A A ,KA )+(1?? )??K ?(KA ,KA )]+cov (?K d,? A A 1+?K ?(KA ,KA ) A
˜)
=
˜)
E [?K ?( i +
B
B ,KB )+(1?? )??K ?(KB ,KB )]+cov (?K d,? B B 1+?K ?(KB ,KB ) B
,
˜ = 1+? . where ? is the shadow value of relaxing the ?nancial constraint: dM ? 0 and ? E [1+? ] See the Appendix for the proof This condition implies that ?rms invest in such a way that, at the optimum, the cost/bene?t ratios are equalized across the two countries. The denominator is the marginal cost of investing from the left hand side of Euler’s equation in Proposition 3. The numerator is the marginal bene?t from the right hand side of Euler’s equation, plus the covariance term. The reasoning behind the covariance terms is that ?rms value an establishment that can generate internal cash ?ow when the ?nancial constraint is binding (high ?) more than an establishment generating cash ?ow when ?nancial constraint is less or not binding (low or zero ?).
6
Gomes (2001) proves that the ?rm dynamic optimization with costly external ?nancing ?(d) has a
unique solution in the technical appendix. With the cost of external ?nancing, the ?rm’s decision can be decomposed into two stages: (1) whether or not to incur the cost of external ?nancing and (2) conditioned on decision in (1) how much to invest and whether or not to merge. For example, multinational’s optimization problem becomes:
InternalF inance ExternalF inance VM (X, KA , KB ) = max[VM (.), VM (.)], InternalF inance where VM ( .) = ExternalF inance VM ( .) =
max
{KA ,KB }?{dM ?0}
[dM + ?E [VM (X , KA , KB )]] and
max
{KA ,KB }?{dM