# Cross-border mergers and acquisitions: does the exchange rate matter? Some evidence for Canada

←

**Page content transcription**

If your browser does not render page correctly, please read the page content below

Cross-border mergers and acquisitions: does the exchange rate matter? Some evidence for Canada George J. Georgopoulos Economics, Atkinson Faculty of Liberal and Professional Studies, York University Abstract. Theoretical and empirical studies investigating the relationship between the ex- change rate and FDI have generated mixed results. Using bilateral Canadian-U.S. industry level count data on cross-border mergers and acquisitions (M&As) and conditioning on industry tariff rates, value added share of industries, industry M&A trend activity, and the number of establishments, we find evidence that a real dollar depreciation of the home currency leads to an increase in the probability of foreign M&As but only in high R&D industries. These empirical results are consistent with Blonigen’s asset acquisition hypoth- esis. Results on European M&As of Canadian firms also lean towards this result. JEL classification: F21, F41 Fusions et acquisitions trans-frontalières: est-ce que le taux de change a de l’importance? Quelques résultats pour le Canada. Les études théoriques et empiriques de la relation entre taux de change et investissement direct à l’étranger donnent des résultats mixtes. A l’aide de données bilatérales canado-américaines au niveau de l’industrie, pour les fusions et acquisitions (F&A) transfrontalières, et de certaines conditions imposées aux tarifs douaniers au niveau industriel, à la part de la valeur ajoutée des industries, à la tendance des F&A au niveau industriel, et au nombre des établissements, on montre qu’une dépréciation de la devise nationale en termes de dollars réels entraı̂ne un accroissement de la probabilité de F&A étrangères mais seulement dans les industries à forte intensité de R&D. Ces résultats empiriques sont compatibles avec l’hypothèse d’acquisition d’actifs de Blonigen. Les résultats quant aux F&A européennes de firmes canadiennes tendent à supporter ce résultat. I am grateful to Bruce Blonigen, Bernardo Blum, John D. Daniels, and Walid Hejazi for useful comments. Of course, all remaining errors are mine. Email: georgop@yorku.ca Canadian Journal of Economics / Revue canadienne d’Economique, Vol. 41, No. 2 May / mai 2008. Printed in Canada / Imprimé au Canada 0008-4085 / 08 / 450–474 / Canadian Economics Association C

Cross-border mergers and acquisitions 451 1. Introduction The decline in the value of the Canadian dollar relative to the U.S. dollar over the late 1980s and 1990s raised concerns that Canadian firms were vulnerable for takeover by U.S. firms at ‘fire sale’ prices. Descriptive data on the number of U.S. mergers and acquisitions (hereafter M&As) in Canada lend support to this claim. While the value of the Canadian dollar declined over 1994–2000, the number of U.S. M&As increased from 168 in 1994 to 300 in 2000, with associated real transaction values of CDN $5.3 billion and CDN $24.1 billion, respectively (Investment Review, Statistics Canada). Theories on a link between exchange rates and cross-border M&As provide ambiguous predictions. One view is that there is no link. While a Canadian dol- lar depreciation relative to the U.S. dollar reduces the cost for U.S. firms of acquiring Canadian assets, the revenues from these assets operating in Canada are denominated in Canadian dollars. Upon repatriating the revenues, conver- sion into U.S. dollars at the lower rate of exchange of U.S. dollars for Canadian dollars offsets the cost reduction of the purchase. Thus there is no net impact of a Canadian dollar depreciation on cross-border takeovers. This view treats the asset as financial in nature, whereby costs and revenues are generated in only one currency. A second view identifies the asset as a factor input that is transferable, such as technology, whereby it can generate revenues in another currency. Blonigen (1997) proposes such a theory, where the target firm has an innovation or a firm specific asset. This ‘asset acquisition’ theory relies on the assumption of market segmentation or imperfect markets for goods. Blonigen applied this theory to the case of Japanese M&As of U.S. firms, partly motivated by the fact that the Japanese market was relatively insulated from foreign penetration. The level of imports from the U.S. and the number of U.S. multinationals in Japan over the 1980s were relatively low. Blonigen’s assumption of market segmentation would thus well apply to the U.S.-Japan case. This paper tests Blonigen’s asset acquisition hypothesis by assessing the em- pirical connection between the exchange rate and cross-border M&As, focusing on Canadian inward and outward M&A data. No other similar study has used Canadian data. In using Canadian data, this study will look at whether Blonigen’s theory holds generally, or holds exclusively only in the case of pronounced mar- ket segmentation, as in the U.S.-Japan case. While in the U.S.-Japan case market, imperfect markets were mainly in the form of import restrictions and restrictions on foreign multinational presence, there are other forms of market segmentation such as established distributional or marketing networks in the acquiring firm’s country or agglomeration advantages that the acquiring firm has in its country. These other forms of market segmentation yield greater profitable opportunities for the acquiring firm in its country relative to a foreign firm. Any such form of market segmentation will produce the Blonigen effect. We thus test whether the Blonigen result holds more broadly by using Canadian M&A data. Furthermore,

452 G.J. Georgopoulos whereas Blonigen uses Japanese M&As in the U.S., this study looks at bilateral Canada-U.S. M&A data, along with European M&As in Canada in testing the asset acquisition hypothesis. Random and fixed-effects negative binomial approaches are employed to model Canada-U.S. bilateral M&A count data. Explanatory variables include the industry-specific real exchange rate, along with traditional determinants of M&As. To further lend support to the acquisition hypothesis, we also look at the role of the exchange rate on European M&As of Canadian firms. As a robustness check, we investigate whether the exchange rate has a similar role on greenfield investments. The following results emerge. There is support for Blonigen’s asset acquisition hypothesis, as results from Canadian-U.S. bilateral data show that a home country currency deprecation will increase the probability of a foreign M&A only in high R&D intensity industries. Results from greenfield investments show that the exchange rate does not play a role for high R&D industries. The organization of the paper is as follows. Section 2 reviews the literature on the theoretical link between exchange rates and foreign direct investment (hereafter FDI). Section 3 presents data trends on Canadian and U.S. M&A and greenfield counts. Section 4 outlines the econometric methodology and presents the determinants of cross-border M&As and the data used. Section 5 provides the empirical results. Section 6 concludes. 2. Exchange rate - FDI link and related literature Motives for FDI are organized around a framework that outlines the advantages and conditions under which FDI will occur in light of the inherent disadvantages and higher costs of foreign production. Early conceptualization of this framework includes Dunning’s development of the ownership-location-internationalization (OLI) framework (1988, 2001). Models that formalize these motives include Markusen (1984; developing a horizontal model of FDI) and Helpman (1984; model of vertical FDI). Markusen, Venables, and Zhang (1996) and Markusen (1997) developed a ‘knowledge-capital’ model, which provides a framework that integrates vertical and horizontal motives for FDI; Carr, Markusen, and Maskus (2001) provide the first empirical investigation of the predictions from this model. The models mentioned above attempt the difficult task of deriving general equi- librium models of FDI behaviour. 1 In this paper we employ a partial equilibrium framework in empirically mod- elling FDI, focusing on the role of the exchange rate. Furthermore we focus on one form of FDI, that being cross-border M&As. Cross-border M&As offer two main advantages over start-up or greenfield investments. One advantage is greater speed in gaining market presence. A second advantage is access to proprietary as- sets. The target firm may have an asset that is not tangible or licensable, such as a 1 See Blonigen, Davies and Head (2003) on issues concerning the empirical model of Carr et al. (2001).

Cross-border mergers and acquisitions 453 scientific technology, a management or organizational skill and/or marketing ex- pertise. This asset acquisition motivation underlies the exchange rate cross-border M&A link. Explanations for exchange rate effects on FDI typically divide into two cate- gories, relative wage effects (Caves 1989) and relative wealth effects (asymmetric information in capital markets: Froot and Stein 1989; stock market performances: Klein and Rosengreen 1994 and Dewenter 1995). Both sets of theories predict that a currency depreciation will lead to increased inward FDI. A shortcoming of the relative wage and wealth explanations is their inherent assumption that only the price of the asset matters in locating abroad, whereas the relevant factor is the rate of return. Blonigen (1997) overcomes this issue, provid- ing both theoretical and empirical evidence in support of an inverse relationship between a domestic country’s currency value and the amount of foreign acquisi- tions of domestic firms. His theory assumes that FDI is motivated by acquiring firm-specific assets that are transferable. Blonigen’s theory relies on the assump- tion of imperfect goods markets, where the domestic firm has limited access to the foreign market to sell its product. For example, suppose a U.S. firm and a Canadian firm have equal opportunity to purchase a Canadian firm possessing a transferable technology. A depreciation of the Canadian dollar will increase the present discounted profits of the U.S. firm as the cost of acquiring the Canadian firm decreases. The present discounted value of profits for the Canadian acquir- ing firm does not change or changes marginally because it has no or limited access to the U.S. market as a result of imperfect goods markets. This may be due to superior or established networks of the U.S. firm in the U.S. in the form of buyer and seller relationships and distributional networks (Greaney 2003; Rauch and Watson 2005; Helliwell 2002; Head, Ries, and Spencer 2002). It may also be that only the U.S. firm is part of a cluster or an agglomeration, where the U.S. firm’s potential profitability from the acquired Canadian firm is higher, owing to knowledge spillovers from localized related industries and research centres. In general, a U.S. firm is more familiar with the workings of the U.S. market relative to a Canadian firm and will have more profitable opportunities. Therefore, the potential Canadian target firm will be more valuable to the U.S. firm than to the Canadian firm following a Canadian dollar depreciation. Using count data on the number of Japanese M&As of U.S. firms over 1975–92, Blonigen finds that a U.S. dollar depreciation increases the likelihood of Japanese M&As in U.S. industries, particularly in those that are R&D intensive. The acquisition hypothesis presumes that the technology is transferred to the U.S. after the Canadian firm is acquired. This seems intuitive and is supported by McFetridge (1987), Caves (1996), and Van Pottelsberghe and Lichtenberg (2001) who show that technology is quickly transferred back to the parent company’s market. The literature concerning Canadian studies is sparse. Schembri (2002) and Aba and Mintz (2002) plot the net annual value of acquisition flows (value of foreign acquisitions of Canadian assets less value of Canadian acquisition of foreign assets) over 1975–2000. The data show no overwhelming evidence of a

454 G.J. Georgopoulos positive trend over the 1990s. Neither study, however, investigates the role of the Canada-U.S. exchange rate to U.S. cross-border M&As while conditioning on other determinants; that is, neither employs a formal econometric model. 2 Lafrance and Tessier (2001) employ Vector Autoregression (VAR) analysis, where the VAR contains Canadian inward FDI flows, the real exchange rate, and a measure of exchange rate volatility. Using the sample period 1970:1-2000:1, they find that the level of the real exchange rate Granger causes inward FDI, although the result was not robust to the introduction of the ratio of undistributed corporate profits to GNP. A shortcoming of the Granger technique is that it does not provide an estimate of the direction or magnitude of the exchange rate-FDI relationship. Furthermore their paper does not distinguish between M&A and greenfield investments. An extensive study on the role of the exchange rate on cross-border M&As in involving Canada has not been done. This paper fills this void. Using bilateral count data on the number of cross-border M&As between Canada and the U.S., along with European M&As of Canadian firms, Blonigen’s asset acquisition hypothesis is tested. Robustness checks are also conducted by looking at the role of the exchange rate on greenfield acquisitions. 3. Merger and acquisition trends in Canada According to the World Investment Report (UNCTAD 2000), cross-border M&As were the major driving force in the substantial increase in FDI over the late 1990s. During the span of 1999, cross-border M&As increased by 3%, rep- resenting over one-third of total FDI for that year. Of all M&A transactions, approximately 3% were in the form of mergers and 97% were in the form of ac- quisitions. Among developed countries, approximately 85% of total FDI is in the form of cross-border M&As. Table 1 presents bilateral data of U.S. and Canadian FDI, by M&As and greenfield investments, where it shows FDI data in all industries, high R&D manufacturing and low R&D manufacturing. 3 A few patterns emerge from the data. First, there are considerably more M&A investments than greenfield invest- ments by both countries. Second, acquisitions fluctuate more from year to year than the start-up of new establishments. Third, there are more transactions in low 2 Moreover, Aba and Mintz (2002) do not consider the possibility that Canadian acquisitions of foreign firms may have resulted from a Canadian dollar appreciation relative to the associated countries. For example, the Canadian dollar generally appreciated relative to the English Pound, the German Mark and the French Franc and Japanese Yen over the latter part of the 1990s, thus potentially explaining Canadian M&As in these countries. Also, in using values of transactions, this may lead to one or two large value transactions dominating the total value of transactions, which would then misrepresent net M&A activity between the two countries. This issue is avoided when looking at the number of transactions (count data). 3 In this study all data on M&As and greenfields exclude governmental service industries. High and Low R&D manufacturing groupings are defined in Section 5.

Cross-border mergers and acquisitions 455

456 G.J. Georgopoulos R&D manufacturing industries than high R&D industries, partly because there are more industries in the former group. This analysis will determine whether the exchange rate has differential effects between these two groups. The table also shows that the Canadian dollar depreciated significantly from 1991 to 1992 and continued to do so over the latter part of the 1990s up to 2001. The data roughly suggest an association between the value of the Canadian dol- lar and the number of U.S. M&As, as the number of M&As generally increased over the periods of depreciations. U.S. greenfield investments seem less respon- sive, generally showing a slight decline over the 1990s relative to the 1980s. This indicates possible differential responses of M&As and greenfields to exchange rate changes. Both forms of Canadian FDI seem more stable relative to the U.S. numbers, although Canadian greenfields have declined noticeably since the late 1990s. Table 2 reports the top 30 Canadian industries experiencing the highest num- ber of U.S. M&As. The wholesale petroleum products sector experienced the most M&A activity, where 185 transactions took place over 1985–2001. The data show that a fair amount of M&A activity occurred in the wholesale indus- try, which consists of SIC codes 5011–5999. Of the 30 reported codes, 11 are in the wholesale industry. Of the top 13 sectors, 6 are in wholesale. The high fre- quency of M&As in the wholesale industry may partially be explained by the high number of establishments in each SIC code. The last two columns in table 2 show that for most of the wholesale industries there were more than 2,000 es- tablishments in 2001, the industry average for that year being 1,204 (bottom of table 2). Looking more closely at the data in table 2, we see that the number of estab- lishments is not the only factor driving the frequency of M&As. There were 67 M&A transactions in industries with codes 3192 and 7512, yet the number of establishments in each differed substantially. The same can be said of industries in codes 7771 and 3799, which experienced 51 and 46 M&As, respectively. Thus, there are other determinants of cross-border M&As beyond the number of es- tablishments. The exchange rate may be one such determinant, along with other factors. In an overview of Canada-U.S. bilateral M&A activity in all industries, fig- ures 1 and 2 present the distribution of the number of M&As by major industry groupings over the sample periods. For U.S. M&As over 1985–2001 there were a total of 4,376, predominantly in the manufacturing and wholesale and retail trade industries. Most of the Canadian M&As were in manufacturing, with a substantially smaller proportion of M&As of firms in the wholesale industry relative to U.S. M&As of Canadian firms. The data from tables 1 and 2 suggests that the exchange rate may be a de- terminant of M&As, along with other determinants such as the number of es- tablishments and perhaps M&A trend activity. These and other factors will be investigated in the empirical model.

Cross-border mergers and acquisitions 457

458 G.J. Georgopoulos FIGURE 1 Number of U.S. acquisitions of Canadian firms by major industry, groupings, 1985– 2001 (total number: 4,376) FIGURE 2 Number of Canadian acquisitions of U.S. firms by major industry, groupings, 1987– 2001 (total number: 939) 4. Empirical model and data The random effects negative binomial (hereafter RENegbin) model proposed by Hausman, Hall, and Griliches (1984) is employed to model the determinants of the number of cross-border M&As. As opposed to the Poisson model, the RENegbin model allows for the conditional expected value and the variance of the number of M&As to differ, that is, allowing for overdispersion. Given the panel nature of the data used here, the RENegbin model allows dispersion to vary randomly across industries. Hausman tests are presented to determine whether the fixed effects negative binomial model is more appropriate than the random effects model. 4 Appendix 1 discusses the RENegbin model in more detail. We begin with U.S. M&As of Canadian firms. The dependent variable is the number of U.S. M&As of Canadian firms at the 4-digit SIC level, excluding government services industries. There are 400 balanced panels of industry clas- sifications. 5 See appendix B for details about the collection of these and other data. 4 Hausman, et al. (1984) state that the RENegbin model yields asymptotically more efficient estimators than their fixed effects model. For these results to be consistent, the industry-specific effects need to be uncorrelated with the regressors. 5 The number of panels used was constrained by the fixed effects estimation methodology. This methodology removes the fixed effects by conditioning on the number of positive outcomes within each panel. Therefore if all the outcomes for the dependent variable are zero (or all outcomes of the dependent variable are the same) for a particular panel, then that panel does not contribute to the likelihood function and it is dropped from the sample set. For direct

Cross-border mergers and acquisitions 459 The regressors chosen generally follow Blonigen (1997). The industry-specific real exchange rate is constructed using the nominal Canadian dollars per U.S. dollar exchange rate, and 2-digit SIC price indexes for the U.S. and Canada. Concordance tables from Statistics Canada were used to match Canadian 1980 SIC codes with U.S. 1987 SIC codes. An increase in the industry-specific real exchange rate reflects a real Canadian dollar depreciation. According to the asset acquisition hypothesis, such a depreciation should be positively correlated with the dependent variable. Cross-border M&As may result from foreign firms avoiding tariffs and other restrictions that nations impose on imports. Thus, cross-border M&As are alter- native strategies to penetrate a domestic market. This strategy effect is captured by including effective Canadian tariff rates on imports from the U.S. at the 4-digit SIC level. A positive correlation between this variable and the dependent variable is expected. To account for factors affecting U.S. demand for Canadian firms, the growth rate of U.S. real GDP and the share of value added output to total GDP of U.S. industry i at the 2-digit SIC level were employed. Concordance ta- bles were used for the latter variable. These demand factors may reflect hor- izontal mergers as firms seek to achieve greater market presence and power. The demand variables should positively affect the number of U.S. M&As in Canada. For supply variables, as is illustrated in table 2, the larger the sector size mea- sured by the number of establishments, the higher the probability of a U.S. ac- quisition, ceteris paribus. Thus, we include the total number of Canadian es- tablishments at the 4-digit SIC level. To capture the overall climate of M&A activity in Canada, the number of M&As of Canadian firms by Canadian firms at the 4-digit SIC level is included. A positive climate of M&A opportunities in Canada should be reflected by a high amount of M&As by Canadian firms. A positive climate should also increase the probability of an acquisition by a U.S. firm. Thus, there should be a positive correlation between the number of Canadian M&As and U.S. M&As. This supply variable should be positively cor- related with the dependent variable. 6 Finally, to capture wealth effects from the stock market, we include Morgan Stanley’s growth in market capitalization for the U.S. comparison, we use the same fixed effects sample set as for the random effects estimation. Although not reported here, the random effects results from the unrestricted (larger) sample set are similar to the random effects results from the restricted sample set reported in Tables 3-5; the results are robust. The results from the unrestricted random effects model are available upon request. 6 As does this paper, Blonigen (1997) uses a similar real exchange rate, along with the two demand variables and the number of domestic M&As. A notable difference is that Blonigen’s acquisition data are at the 3 digit level, whereas the U.S. acquisition of Canadian firms used in this study are at the 4 digit level.

460 G.J. Georgopoulos Thus, the following is the testing specification in the context of U.S. M&As of Canadian firms: Prob(USM&Asit ) = f (rerit , tariff it , USgrowtht , USsharei t , CDNM&Asit , CDNestabit , USMCgrowtht ), (1) where USM&As it is the number of U.S. M&As in Canada at the 4-digit Canada SIC level; rer it is the Canada-U.S. real exchange rate at the 2-digit SIC level; tariff it is the effective tariff rate on imports from the U.S. at the 4-digit SIC level; USgrowth t is the growth rate of U.S. real GDP; USshare it is share of value added of U.S. industry i at the 2-digit SIC level; CDNM&As it is the number of Canadian M&As at the 4-digit SIC level; CDNestab it is the number of Canadian establishment at the 4-digit SIC level; and USMCgrowth t is the growth in market capitalization in the U.S. The data reflect the highest available level of disaggre- gation. The data are at annual frequency (t) and cover the period 1985–2001. 5. Empirical results 5.1. U.S. M&As of Canadian firms Table 3 presents estimation results involving U.S. M&As of Canadian firms. Columns 1 and 2 show the random- and fixed-effects estimates from the sample of all industries. To determine whether the industry-specific effects are correlated with the regressors, the Hausman (1978) test statistic is reported at the bottom of column 1, which tests the specification of the random effects model in column 1 and the fixed effects model in column 2. This test is distributed as χ 2 under the null hypothesis of no correlation between the industry-specific effects and the regressors. The p-value, reported in brackets, is 0.090, suggesting the random ef- fects is appropriate at the 5% level of significance. The results in column 1 thus are appropriate, although the fixed-effects results are relatively similar. The bottom of column 1 also reports a likelihood ratio test which compares the RENegbin or panel estimator with the Negbin model or pooled estimator. The test statistic rejects the null hypothesis of the pooled estimator or the Negbin estimator with constant dispersion. 7 The random-effects results in column 1 show that all the coefficients have expected signs. Focusing on the real exchange rate reveals that the sign is what is expected, suggesting that a rise in the real exchange rate – a real depreciation 7 Although not reported, the likelihood ratio test on the alpha parameter is used to determine whether the data can be modeled by a Poisson process. A value of zero for the alpha parameter implies that the observations are generated by a Poisson process, or no overdispersion, which is a special case of the Negbin model. For this and all specifications, the Poisson model is rejected in favour of the Negbin model. The test results and estimates of the alpha parameter are available upon request.

Cross-border mergers and acquisitions 461 TABLE 3 U.S. mergers and acquisitions of Canadian firms, 1985–2001; dependent variable: number of U.S. mergers and acquisitions of Canadian firms All industries High R&D mfgc Low R&D mfg Random Fixed Random Fixed Random Fixed Variables effects effects effects effects effects effects rer (real exchange rate) 0.002 0.005 1.600 1.432 −0.401 −0.344 (0.861) (0.604) (0.030) (0.049) (0.307) (0.393) tariff (average tariff rate on 0.0252 0.0264 −0.0045 −0.0085 0.036 0.0388 imports from the U.S.) (0.002) (0.002) (0.856) (0.740) (0.001) (0.001) USgrowth (growth rate of 0.0005 0.00058 −0.0015 0.00004 0.0006 0.00064 U.S. real GDP) (0.001) (0.001) (0.784) (0.931) (0.040) (0.031) USshare (U.S. industry value 0.00067 −0.00029 0.0187 0.0165 0.0019 0.00069 added share) (0.075) (0.525) (0.001) (0.002) (0.113) (0.803) CDNacquisitions (number of 0.003 0.005 −0.057 −0.074 0.034 0.048 Canadian domestic (0.628) (0.520) (0.304) (0.230) (0.027) (0.005) acquisitions) CDNestab (number of 0.00002 0.00002 0.0007 0.0006 0.0003 0.0002 Canadian establishments) (0.001) (0.001) (0.001) (0.001) (0.001) (0.003) USmcgrowth (U.S. market 0.0023 0.0024 −0.0064 −0.0067 0.0049 0.0047 capitalization growth rate) (0.103) (0.085) (0.124) (0.107) (0.069) (0.054) constant 0.287 0.496 −4.533 −3.782 0.960 1.122 (0.008) (0.001) (0.002) (0.009) (0.085) (0.073) Hausman testa (0.090) (0.040) (0.847) Likelihood ratio testb (0.001) (0.001) (0.001) Observations 6800 6800 374 374 2142 2142 NOTE: p-values are in parenthesis. a Hausman (1978) test, where the null hypothesis is no correlation between the industry-specific events and the regressors. b The likelihood ratio test compares the panel or RENegbin estimator with the pooled or Negbin estimator, where the null hypothesis is the Negbin model. c High R&D manufacturing industries are Machinery (2-digit SIC 31), Aircraft and Parts (3digit SIC 321), Electrical and Electronic Products (33), and Pharmaceutical and Medicine (374). SOURCE: Industrial Research and Development, Catalogue 88-202-XIE, Statistics Canada of the Canadian dollar – will increase U.S. M&As in Canada. However, the real exchange rate is statistically insignificant. This result appeals to the view that there is no link between the exchange rate and FDI; the depreciation of the domestic currency will leave the rate of return of a foreign investor equal to that of a domestic investor, as the profits are repatriated at a higher exchange rate. The tariff rate, the U.S. real GDP growth, and the number of establishments in each sector are significant at the 5% level, while the U.S. industry share and the market capitalization growth rate of the U.S. stock market are statistically significant at the 10% level. Overall, for the full sample, the traditional determinants of FDI perform well in this empirical specification. To test the asset acquisition hypothesis, along the lines of Blonigen (1997) we focus on acquisitions in manufacturing as firm-specific assets would be important

462 G.J. Georgopoulos for this industry. We examine whether manufacturing firms with high R&D in- tensities are more likely to be acquired after a domestic currency depreciation relative to low R&D intensity firms. High R&D industries are more likely to have technology-related firm-specific assets. According to the asset acquisition hypothesis proposed by Blonigen (1997) these industries are expected to ex- perience a higher number of cross-border M&A activity in the presence of a domestic currency depreciation, as domestic firm-specific assets that are trans- ferable become less costly to acquire. We split manufacturing into high R&D industries and low R&D industries, where high R&D industries are classified as those having ratios of R&D expenditures to sales above the manufactur- ing average. The industries with ratios consistently greater than the average are reported at the bottom of table 3. Most are the traditional R&D indus- tries such as machinery, electrical products, and pharmaceutical and medicine industries. 8 Columns 3 and 4 present the results from high R&D manufacturing. At the bottom of column 3 the Hausman test shows that the fixed-effects model is the correct empirical model, although again the random effects results are similar. In column 4, of importance is that the real exchange rate is now statistically significant. Furthermore, the value of the coefficient is considerably larger than in column 1. These results lend support to Blonigen’s (1997) asset acquisition hypothesis. In addition the tariff rate went from being significant in the full sample case to insignificant. This result is intuitively appealing, as a tariff barrier is not a primary factor for a foreign firm whose primary motive is to acquire transferable assets. Industry share is still significant and has the expected sign, along with the number of establishments. Columns 5 and 6 show the results from the low R&D manufacturing industries. The Hausman test accepts the null hypothesis of no correlation between industry specific effects and the regressors. Thus, in column 5, of importance is that the real exchange rate is now statistically insignificant. For those firms having a relatively low level of technology intensities, there is little if any M&A activity in their associated industries for the purpose of asset acquisition; thus, the exchange rate does not matter. Note that the tariff rate is now significant. With respect to these industries, the suggested motive for cross-border M&As is to gain access to a new market or exploit economies of scale, where this mode of entry substitutes for trade in the presence of tariffs. Industry share is not significant, whereas the remaining variables have expected signs and are statistically significant. These results also hold for the fixed-effects estimates. The results show that the exchange rate is a statistically significant factor only for cross-border M&As of firms with high R&D intensities. Tariff rates are not 8 R&D expenditures relative to sales data at the 2 and 3 digit SIC level were provided by the Industrial Research and Development Department, Catalogue 88-202-XIE, Statistics Canada.

Cross-border mergers and acquisitions 463 statistically significant for high R&D firms. This is expected when the primary motive for an M&A is to acquire a firm-specific asset that is transferable. The exchange rate does not play a role for cross-border M&As for industries with low levels of R&D intensity, whereas tariff rates do. 5.2. Canadian M&As of U.S. firms Further evidence that would support the asset acquisition hypothesis would be to find a reduction in the probability of Canadian M&As of U.S. high R&D firms after a Canadian dollar real depreciation. We investigate this using count data on the number of Canadian M&As of U.S. firms. These data were provided by the Bureau of Economic Analysis, where it was available only at the 3-digit U.S. SIC level beginning in 1987. Table 4 presents the results, where again similar explanatory variables were used but in the context of determinants of Canadian M&As of U.S. firms. For data sources on these variables see appendix B. For the full sample case, the Hausman test shows that the fixed-effects estimation results in column 2 are appropriate. The estimated coefficient on the real exchange rate is statistically significant, but has an unexpected sign, although the coefficient is relatively small in magnitude. The growth rate of Canadian real GDP is statistically significant and positively affects the probability of a Canadian M&A in the U.S. A questionable result is the statistically significant negative effect of the market capitalization rate in Canada. The estimation was again taken for the high and low R&D manufacturing industries, where R&D shares were provided by the U.S. National Science Foun- dation. The high R&D U.S. industries are reported at the bottom of table 4. For high R&D firms, the fixed effect results of column 4 show evidence sup- porting the asset acquisition hypothesis. Specifically, the real exchange rate now has a negative affect, which is consistent with the asset acquisition theory; a real depreciation of the Canadian dollar decreases the probability of a Canadian firm acquiring a U.S. firm, although this effect is statistically significant at the 10% level. The random-effects model provides similar evidence, where the real exchange rate is significant at the 1% level. U.S. domestic M&A activity has a positive effect and is significant at the 10% level. The result on tariffs is unclear, as the coefficient is negative and significant. For low R&D industries, the exchange rate results from the random effects model differ from the high R&D in that the sign has changed and is statistically insignificant, further supporting the asset acquisition hypothesis. U.S. domestic M&A activity and the number of establishments in an industry are the main factors driving Canadian M&As of low R&D U.S. firms. The results from tables 3 and 4 show support of the asset acquisition effect holding bilaterally. Given a Canadian dollar depreciation, the probability of U.S. M&As of Canadian high R&D firms increases, while the probability of Canadian M&As of U.S. high R&D firms decreases.

464 G.J. Georgopoulos TABLE 4 Canadian mergers and acquisitions of U.S. firms, 1987–2001; dependent variable: number of Cana- dian mergers and acquisitions of U.S. firms All industries High R&D mfgc Low R&D mfg Random Fixed Random Fixed Random Fixed Variables effects effects effects effects effects effects rer (real exchange rate) 5.38e–07 0.00001 −4.322 −3.409 0.788 0.888 (0.575) (0.008) (0.021) (0.105) (0.386) (0.510) tariff (average tariff rate −0.0939 −0.0266 −0.5093 −0.4971 −0.0154 0.07494 on imports from (0.087) (0.675) (0.011) (0.039) (0.787) (0.325) Canada) CDNgrowth (growth rate 0.0527 0.0622 0.0736 0.101 0.002 0.0129 of Canadian real (0.019) (0.007) (0.320) (0.217) (0.962) (0.780) GDP) CDNshare (Canadian −0.0283 0.2174 0.2102 0.1859 −0.563 −0.684 industry value added (0.694) (0.242) (0.007) (0.904) (0.360) (0.738) share) USacquisitions (number 0.001 0.0005 0.008 0.009 0.015 0.013 of U.S. domestic (0.001) (0.123) (0.010) (0.085) (0.003) (0.046) acquisitions) USestab (number of U.S. −1.23e-07 1.09e-06 0.00004 0.00001 0.0001 0.0001 establishments) (0.233) (0.669) (0.440) (0.859) (0.012) (0.007) CDNmcgrowth −0.0045 −0.0050 −0.0036 −0.0043 −0.0011 −0.0014 (Canadian market (0.002) (0.001) (0.369) (0.313) (0.640) (0.582) capitalization growth rate) constant 0.298 0.201 3.528 3.605 0.536 −0.037 (0.149) (0.421) (0.069) (0.122) (0.700) (0.982) Hausman testa (0.019) (0.001) (0.078) Likelihood ratio testb (0.001) (0.001) (0.001) Observations 1886 1886 234 234 640 640 NOTE: p-values are in parenthesis. a Hausman (1978) test, where the null hypothesis is no correlation between the industry-specific events and the regressors. b The likelihood ratio test compares the panel or RENegbin estimator with the pooled or Negbin estimator, where the null hypothesis is the Negbin model. c High R&D Manufacturing industries are Chemical and Allied Products (US SIC 28), Office, Computing, and Accounting Machines (357), Electrical Equipment (36), Transportation Equipment (37), Professional and Scientific Instruments (38). SOURCE: National Science Foundation, Survey of Industrial Research and Development: 1998 5.3. European M&As of Canadian firms The asset acquisition hypothesis should apply to other countries’ M&A activities in Canada. We look for further support by investigating total M&As of Canadian firms by the U.K., France, and Germany (Euro3 henceforth) as a whole. 9 They account for a total of 67% of total non-U.S. foreign M&As of Canadian firms. 9 Comparable data on Canadian M&As in these 3 European countries and the determinants are not available.

Cross-border mergers and acquisitions 465 FIGURE 3 Number of Euro3 acquisitions of Canadian firms by major industry, groupings, 1985– 2001 (total number: 1,289) NOTES: Euro3: France, Germany, U.K.; Comm: Commodities; Mfg: manufacturing; CTU: Construction, Transportation and Storage, and Utilities; WRT: Wholesale and Retail Trade; FR- Finance, Real Estate, and Insurance SOURCE: Investment Review, Industry Canada Owing to the relative infrequency of M&As for each of these countries separately across all 4-digit SIC categories, we used the total amount of M&As from these three countries as the dependent variable. 10 Figure 3 shows the distribution of Euro3 M&As of Canadian firms over 1985–2001. The majority of M&As were in manufacturing and wholesale and retail trade industries. Table 5 shows regressions similar to those used in table 3, but in the context of Euro3 determinants of M&As in Canada. The acquisition data again were taken from the Investment Review, Statistics Canada, and are at the 4-digit SIC level. The real exchange rate measure is the simple unweighted average of the three Canada-country-specific real exchange rates, where the individual real exchange rates were positively correlated. 11 Data sources of the remaining variables are outlined in appendix B. We again estimate over all industries and over high and low R&D manu- facturing industries. Over all industries, the random effects model shows that the average real exchange rate has a negative sign and is statistically signifi- cant. Real GDP growth of the three European countries, industry share, and the number of Canadian establishments are positive and statistically signif- icant. Tariff rates and the number Canadian domestic acquisitions are in- significant, whereas the average stock market capitalization growth is signif- icant but has a negative sign. The interpretation of the latter effect is not clear. For high R&D manufacturing industries, the fixed-effects model shows that the real exchange rate effect is insignificant, but of interest is that the coefficient is now positive, a sign consistent with the asset acquisition hypothesis. 12 The 10 The U.S. accounts for 72% of foreign M&As in Canada from 1985-2001, with the remaining countries accounting for 28%. The total number of M&As by France, Germany, and the U.K. over this period were 279, 200, and 810 respectively. 11 The correlation between the Canadian-German and Canadian- French real exchange rates was 0.937, 0.5142 for the Canadian-German and Canadian-U.K. real exchange rates, and 0.3145 for Canadian-French and Canadian-U.K. real exchange rates. 12 Blonigen (1997) found that results on German M&As of U.S. firms showed the exchange rate

466 G.J. Georgopoulos TABLE 5 Euro3 mergers and acquisitions of Canadian firms, 1985–2000; dependent variable: number of Euro3 (U.K., France, Germany) M&As of Canadian firms All industries High R&D mfgc Low R&D mfg Random Fixed Random Fixed Random Fixed Variables effects effects effects effects effects effects rer (average real exchange −0.539 −0.395 1.313 1.397 0.594 0.733 rate of Euro3) (0.052) (0.169) (0.214) (0.191) (0.344) (0.246) tariff (average tariff rate on −0.0062 −0.0054 −0.033 0.0410 −0.0015 −0.0138 imports from Euro3) (0.382) (0.564) (0.422) (0.389) (0.835) (0.264) Euro3growth (growth rate of 0.186 0.192 −0.0517 0.0383 0.1501 0.1620 Euro real GDP) (0.001) (0.001) (0.630) (0.724) (0.025) (0.016) Euro3share (Euro3 industry 0.0331 0.0413 0.1794 0.0770 0.001 0.1094 value added share) (0.001) (0.009) (0.007) (0.287) (0.002) (0.003) CDNacquisitions (number −0.006 −0.008 0.067 0.030 −0.0459 −0.012 of Canadian domestic (0.631) (0.555) (0.141) (0.596) (0.315) (0.777) acquisitions) CDNestab (number of 0.00001 −1.05e-06 0.0009 0.0003 0.0003 0.00007 Canadian establishments) (0.015) (0.905) (0.001) (0.322) (0.004) (0.688) EUROmcgrowth (European −0.0056 −0.0056 −0.0091 −0.0073 −0.0077 −0.0076 market capitalization (0.001) (0.001) (0.053) (0.127) (0.006) (0.007) rate) constant −0.419 0.572 −4.887 −2.534 −3.224 −2.982 (0.314) (0.293) (0.055) (0.328) (0.019) (0.033) Hausman testa (0.336) (0.001) (0.154) Likelihood ratio testb (0.001) (0.001) (0.001) Observations 4743 4743 374 374 1683 1683 NOTE: p-values are in parenthesis. a Hausman (1978) test, where the null hypothesis is no correlation between the industry-specific events and the regressors. b The likelihood ratio test compares the panel or RENegbin estimator with the pooled or Negbin estimator, where the null hypothesis is the Negbin model. c High R&D Manufacturing industries are Machinery (Canada SIC 31), Aircraft and Parts (321), Electrical and Electronic Products (33), and Pharmaceutical and Medicine (374). SOURCE: Industrial Research and Development, Catalogue 88-202-XIE, Statistics Canada insignificant coefficient may reflect more disparate or heterogeneous firms and production processes between Canada and Europe relative to Canada and the U.S., thus leading to fewer opportunities for accessing complimentary assets. For low R&D industries, the average real exchange rate is insignificant. Of note is that its magnitude is nearly half the size of the high R&D coefficient. Overall, while the real exchange rate is insignificant in the high R&D case, the effect becomes positive relative to the full sample case and is noticeably larger than the low R&D effect, evidence leaning towards the asset acquisition hypothesis. From the three cases above, in the Canada-U.S. bilateral case there is support for the asset acquisition hypothesis, where the European case leans towards this having an expected sign but being statistically insignificant.

Cross-border mergers and acquisitions 467 FIGURE 4 Number of U.S. greenfield investments in Canada by major industry, Groupings, 1985– 2001 (total number: 1,244) FIGURE 5 Number of Canadian greenfield investments in the U.S. by major industry groupings, 1987–2001 (total number: 532) hypothesis. For M&As in low R&D manufacturing industries the exchange rate does not play a role. 5.4. Greenfield investments Access to proprietary assets certainly is not relevant when investigating the fac- tors for greenfield investments. We would thus not expect the exchange rate to play a role in this form of FDI. We next look at the role of the exchange rate on U.S. greenfield investments in Canada, Canadian greenfield investments in the U.S., and Euro3 greenfield investments in Canada. Figures 4–6 show the distribu- tion of each, respectively, by major industry groupings. All three source countries show a significantly lower proportion of greenfield investment in manufacturing relative to M&A investment in manufacturing. For U.S. and Euro3 greenfield investment in Canada, the dominant proportion is in the wholesale and retail in- dustries, whereas for Canadian greenfield investments in the U.S., the proportion of finance, insurance, and real estate dominates. We run specifications similar to those used in equation (1) for greenfield in- vestments, but excluding the number of domestic acquisitions, as there is no theoretical evidence suggesting that this variable affects greenfield investments. In table 6 we report for all countries only the results from the key variables of interest, the exchange rate, and the tariff rate. 13 For the Canada-U.S. bilateral 13 The results relate to the appropriate fixed or random effects model. The results for the remaining

468 G.J. Georgopoulos FIGURE 6 Number of Euro3 greenfield investments in Canada by major industry, groupings, 1985– 2001 (total number: 362) NOTES: Euro3: France, Germany, U.K.; Comm: Commodities; Mfg; manufacturing; CTU: Construction, Transportation and Storage, and Utilities; WRT: Wholesale and Retail Trade; FR: Finance, Real Estate, and Insurance SOURCE: Investment Review, Industry Canada TABLE 6 Cross-border greenfield investments and the exchange rate; dependent variable: number of inward greenfield investments All industries High R&D mfg∗ Low R&D mfg (a) U.S. greenfields in Canada rer (real exchange rate) −0.133 0.876 −0.135 (0.319) (0.574) (0.262) tariff (average tariff rate on imports 0.191 0.153 0.082 from the U.S.) (0.001) (0.011) (0.003) (b) Canadian greenfields in the U.S. rer (real exchange rate) 9.12e-06 0.502 3.889 (0.254) (0.866) (0.163) tariff (average tariff rate on imprts 0.245 0.200 0.304 from Canada) (0.018) (0.466) (0.052) (c) Euro3 greenfields in Canada rer (average real exchange rate of Euro3) −1.71 −1.432 3.39 (0.001) (0.593) (0.051) tariff (average tariff rate on imports −0.066 −0.091 −0.014 from Euro3) (0.009) (0.336) (0.728) NOTES: p-values are in parenthesis. The results relate to the appropriate fixed- or random-effects model. ∗ See tables 3 and 4 for the industry sectors with high R&D intensities. case (results (a) and (b)), the exchange rate does not play a role in determining inward greenfield investments. Of note is that the exchange rate is not signifi- cant in the high R&D manufacturing case, a result unlike the M&A case of high R&D manufacturing industries. Overall, tariffs play a greater role in determining greenfields relative to the M&A results of tables 3–5, reflecting the conventional result of foreign firms jumping tariffs through greenfield investments to access regressors are similar to the results in Tables 3-5 and are available upon request.

Cross-border mergers and acquisitions 469 markets. Furthermore in the case of Canadian greenfields in the U.S., the tariff rate now has the expected sign unlike the situation in the M&A case. For the European results, the exchange rate coefficient in the high R&D case is negative and insignificant. The exchange rate results in the full sample and low R&D sample cases are mixed. The tariff rate for all three samples have unexpected signs, with the results for manufacturing industries being insignificant. Overall, there is no role for the exchange rate in determining greenfield invest- ments in high R&D industries, a result different from the Canada-U.S. bilateral M&As in high R&D industries. In general, the role for the exchange rate for greenfields is weak. 6. Concluding remarks While most previous studies on the effect of the exchange rate on FDI have produced ambiguous results, this study contributes to the understanding of this relationship by focusing on one form of FDI: mergers and acquisitions. Using U.S. and Canadian bilateral M&A data, the results showed that a real currency depreciation leads to an increase in the probability of a cross-border M&As but only in high R&D intensive industries, and evidence on European M&As of Canadian firms lean towards this result. These results are consistent with Blonigen’s asset acquisition hypothesis; however, this is the first study using Canadian data on cross-border M&As. The results lend support to the asset acquisition holding more generally or beyond the U.S.-Japan setting, where the degree of market segmentation is high. The exchange rate affect is not present in the case of greenfield investments, a result that further strengthens the asset acquisition hypothesis, as access to proprietary assets is not relevant in this form of FDI. Appendix A: Random effects negative binomial model The Poisson distribution is widely used in analyzing count data where the depen- dent variable is discrete and defined for non-negative integers corresponding to the number of events occurring in a given interval (Hausman, Hall, and Griliches 1984). The objective is to empirically model the determinants of the number of U.S. mergers and acquisitions. The Poisson probability function may be expressed as follows: y Prob(yit ) = e−λi t λitit yit ! where y it is the number of acquisitions of M&As of firms in the SIC code i in period t, and λ it is the conditional mean and variance. Standard procedure is to

470 G.J. Georgopoulos make λ it to be an exponential function of the explanatory variables: λit = exp(Xit β). The coefficients are estimated by maximizing the log-likelihood function of the Poisson model. As noted, this model has the property that the conditional ex- pected value and variance are equal. This condition is violated in many data sets. To overcome this condition, the negative binomial (hereafter Negbin) model has been developed for cross-sectional data (Cameron and Trivedi 1986). It allows for the second conditional moment to differ from the first. In many economic applications it is not uncommon to find that the variance exceeds the mean. This implies ‘overdispersion’ in the data. Overdispersion may have two explanations. First, it may be caused by unobserved heterogeneity. Second, the data-generating process may be such that there is interdependence between the occurrence of suc- cessive events. In the presence of overdispersion, employing the Poisson model will lead to an estimated variance-covariance matrix that is biased downwards, yielding incorrectly small estimated standard errors of the parameter estimates and overstated t-statistics. The Negbin model allows for heterogeneity in the mean function by introduc- ing an additional stochastic component to λ it : λit = exp(Xit β + εit ), where εit captures unobserved heterogeneity and is uncorrelated with the explana- tory variables, and exp (ε) follows a gamma distribution. A concern with the standard Negbin model is that it does not take into ac- count the panel nature of the data used here. The Negbin model pools the data where it constrains the dispersion to be constant across all panels. In contrast, heterogeneity across panels or industries is allowed by employing the random ef- fects Negbin model proposed by Hausman, Hall, and Griliches (1984). 14 In this model, the dispersion varies randomly from industry to industry. Specifically, the Poisson parameter λ it follows a gamma distribution with parameters (γ , δ). For the variation across industries, we allow each industry i to have its own δ i such that 1/(1 + δ i ) is distributed as a Beta random variable with the shape parameters (a, b). 15 For any industry i, the dispersion (variance divided by mean) is equal to 1 + δ i . As is standard, we set γit = exp(X it β). Hausman test results and fixed-effects results are also presented. 14 Hausman, et al. (1984) states that the RENegbin model yields asymptotically more efficient estimators than their fixed effects model. For these results to be consistent, the industry-specific effects need to be uncorrelated with the regressors. 15 The Beta distribution will be symmetric if a = b, and can be U-shaped or hump-shaped.

Cross-border mergers and acquisitions 471 Appendix B: Data sources B.1. U.S. M&A and Greenfield Investments in Canada Count data on U.S. mergers and acquisitions of Canadian firms are from the Investment Review Division, Industry Canada. The data record the transactions involving U.S.-owned businesses merging or acquiring Canadian-owned business in Canada. According to the Investment Canada Act, section 26, a merger or ac- quisition relates to a change in control, where control is defined as acquiring at least 50% of common voting stocks or assets. The act requires every non- Canadian to file a Notification or an Application for Review with Investment Review, Industry Canada, each time a non-Canadian plans to or actually com- mences a new business activity in Canada (greenfield investment) or acquires a Canadian business. Applications for Review are necessary for relatively large transactions (in excess of $218 million for WTO members and $5 million for non-WTO members), and, upon receipt, the minister of finance will determine whether the investment is permitted. Authorities at Investment Review contact the potential investors within a specified period after the submission of forms to determine whether the transaction was implemented. The recorded acquisition data reflect those applications where the investment was actually carried out. All the Notifications are recorded in the data, but there is no follow up process for the Notifications. Thus, there may be an upward bias on the number of invest- ments in this category. However, authorities at Investment Review estimate that the vast majority of Notifications are submitted after the investment has taken place. The data do not distinguish between mergers and acquisitions. The full sample size ‘All industries’ in table 3 consists of 400 panels of balanced data over the years 1985–2001 (6,800 observations), where the panels used were determined by the fixed-effects estimation methodology (see footnote 5 on the choice of the restricted sample set). For the Canada-U.S. real exchange rate, the nominal exchange rate data e are from CANSIM I and are annual averages. The 2-digit U.S. industry price data are from the Bureau of Labour Statistics, Industry Prices. The Canadian prices are from CANSIM I, Industrial Product Indices. Concordance tables were used to match Canadian 1980 SIC codes with U.S. 1987 SIC codes. The tariff variable is the effective tariff rate at the 4-digit SIC level. This is equal to the ratio of the value duties collected on U.S. imports at the 4-digit SIC level to the value of U.S. imports at the 4-digit SIC level. The source is the International Trade Division, Statistics Canada. Data for the growth rate of U.S. real GDP are from CITIBASE. U.S. Industry share is the share of value added of U.S. industry i at the 2-digit SIC level, which was constructed using the share of GDP of industry i to total GDP. U.S. Bureau of Economic Analysis, GDP by Industry. Concordance tables were used to match Canadian 1980 SIC codes with U.S. 1987 SIC codes (Statistics Canada 1990). Data on the number of Canadian establishments are at the 4-digit SIC level from the Canadian Business Patterns, Statistics Canada. The number of

You can also read