WPS6905 Policy Research Working Paper 6905 Cross-Border Mergers and Acquisitions in Services The Role of Policy and Industrial Structure Alessandro Barattieri Ingo Borchert Aaditya Mattoo The World Bank Development Research Group Trade and International Integration Team June 2014 Policy Research Working Paper 6905 Abstract This paper presents evidence on the determinants of but cultural factors affect mergers and acquisitions in cross-border mergers and acquisitions in services sectors. services more than in manufacturing. (2) Controlling It develops a stylized model of mergers and acquisitions for these bilateral factors, restrictive investment policies that predicts that the incidence of merger and acquisition reduce the probability of merger and acquisition inflows deals depends, inter alia, on the target economy’s size, but this negative effect is mitigated in countries with industrial structure and investment policies, as well as relatively large shares of manufacturing and (to a lesser on bilateral transactions costs. These predictions are extent) services in gross domestic product. The same examined with bilateral merger and acquisition flow results hold for the number of merger and acquisition data and detailed information on policy barriers from a deals received. These findings suggest that the impact of new database of restrictions on services investment. The policy is state-dependent and related to the composition analysis finds that: (1) geographical factors affect mergers of gross domestic product in the target economy. and acquisitions in services and manufacturing similarly This paper is a product of the Trade and International Integration Team, Development Research Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at barattieri.alessandro@uqam.ca, I.Borchert@sussex.ac.uk, and amattoo@worldbank.org. . The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Cross-Border Mergers and Acquisitions in Services: The Role of Policy and Industrial Structure∗ Alessandro Barattieri† Ingo Borchert ‡ Aaditya Mattoo§ JEL classification: F13, F21, L80 Keywords: Cross-border mergers and acquisitions, state dependency in policy effectiveness ∗ We would like to thank our discussants, Jean-Claude Cosset and Arthur Blouin, as well as Ari Van- Assche, Lucian Cernat, Barry Reilly, and seminar participants at HEC Montreal, University of Sussex, and ESG UQAM and conference participants at GIST 2012, ETSG 2012, ITFA 2013, RECODE 2013, and NEUDC 2013 for useful comments and suggestions. Research for this paper has been supported in part by the governments of Norway, Sweden, and the United Kingdom through the Multidonor Trust Fund for Trade and Development, and by the UK Department for International Development (DFID). † ´ ESG UQAM and CIRPEE. Mail: Case Postale 8888, succursale Centre-ville Montreal (Que- bec) H3C 3P8, Canada. Tel: +1-514-987-3000 (0850#). E-mail: barattieri.alessandro@uqam.ca ‡ University of Sussex, Department of Economics, School of Business, Management and Eco- nomics. Mail: Jubilee Building, Falmer, Brighton, BN1 9SL, UK. E-mail: I.Borchert@sussex.ac.uk § World Bank. Mail: 1818 H Street, N.W., Washington, DC 20433. E-mail: amattoo@worldbank.org 1 Introduction Cross-border mergers and acquisitions (M&A) have been among the most striking inter- national economic phenomena of the last two decades. While the determinants of M&A have been studied at the aggregate level, there is little analysis of the determinants of M&A in services sectors. This is surprising, given that these sectors accounted for 65 percent of global cross-border M&A deals in the period 1990-2012 (UNCTAD, 2013). This paper presents evidence on the determinants of the M&A in services seen through the lens of a model featuring a role for both policy and inter-sectoral linkages. The empirical analysis is based on a large sample of developed and developing economies, and a new database on services policy measures. We proceed in three steps. First, we present some broad patterns emerging from our data. Aggregating our transaction-level M&A data into a country-level and bilateral database, we show that the fraction of country-pairs with non-zero M&A flows is small. We also present cross-country evidence on policy restrictiveness in trade in services drawing on a new policy database. Detailed information on restrictions on the forms of entry, licensing, operations and regulation are aggregated into a Service Trade Restrictiveness Index (STRI), computed for five service sectors in each of 103 countries. The country-level STRI is broadly declining in the GDP per capita, i.e. developing countries tend to have more restrictive policies than developed countries. Restrictiveness in sectors such as transport and legal services seems to be much higher than for other services in all countries, irrespective of their level of development. Second, we propose a stylized model of M&A featuring heterogeneous firms and a role for inter-sectoral linkages. The manufacturing sector is perfectly competitive, and it produces a homogenous good using labor and intermediate services. This assumption reflects evidence from the OECD STAN Input-Output tables that services are important intermediate inputs in the production of manufacturing across a wide range of countries. In our model, services are characterized by product differentiation and they are produced by heterogeneous firms under monopolistic competition. M&As in services are presented as a way of serving foreign 2 markets, and they are subject to fixed costs depending on country-pair specific costs (cultural or physical distance) and the policy environment of the host country. We show that the model can rationalize the existence of zero bilateral M&As flows. Moreover, the impact of policy liberalization that reduces the costs of M&As on the probability of observing M&A flows (and on the number of M&A deals) depends on the overall size of the market, the composition of the GDP, and the extent of inter-sectoral linkages. We label the possibility that the effect of a given policy on a certain outcome variable might depend on the state of the world in which the policy is applied state-dependency in policy effectiveness. Third, inspired by the model, we undertake a two-stage empirical analysis. In the first stage, we estimate a Probit model using bilateral M&A data. The Probit includes host coun- try and home country fixed effects, physical distance and proxies for the cultural distance between countries. In the second stage, we investigate the determinants of host country fixed effects, in particular the role of policy barriers towards M&A. We repeat the analysis also using the total number of M&A deals flowing to different service sectors in individual countries, thus exploring the intensive margin. Following Santos-Silva and Tenreyro (2006), we use in this case a pseudo-maximum likelihood (PPML) estimator for the first stage. Our findings are the following: (1) Geographical barriers affect M&A in services and manufac- turing similarly, but cultural barriers affect M&A in services more than in manufacturing. Both types of barriers have heterogeneous effects across services sectors: a shared border matters more in transportation, as expected, but also in retail; and a common language is more important for M&A in telecommunications, as well as in banking and insurance. (2) Across countries, the probability of receiving investment through cross-border M&A is strongly positively correlated with market size. Restrictive policy dampens M&A inflows but the negative effect of policy is mitigated in countries with a relatively large shares of manufacturing and (to a lesser extent) services in GDP. The same results hold for the total number of M&A deals received. These findings suggest that the impact of policy is state- dependent and related to the composition of the GDP of the target country. The results may help explain why policy restrictions have inhibited services investment in the industrializing 3 economies of South-East Asia less than in other parts of the world. (3) We try to identify the individual policy measures that account for results obtained using the aggregate index. Restrictions on the nationality of employees and lack of transparency in the licensing process are among the measures deterring investment. This paper is connected to the literature on the determinants of cross-border M&A. Early empirical studies of aggregate cross-border mergers and acquisitions include Rossi and Volpin (2004) and Di Giovanni (2005). Head and Ries (2008) present a model of bilateral flows of M&A based on a theory of optimal corporate control, and test it on the same data used by Di Giovanni. Hijzen et al (2008) analyze cross-border M&A among OECD countries and focus on the number of deals and the distinction between horizontal and vertical M&A. Coeurdiacer et al (2009) analyze the impact of European integration on bilateral cross- border M&A and the impact of product market deregulation in services within the EU. Hyun and Kim (2010) extend the analysis of determinants of M&A to a large sample of countries using aggregate data. Ahern et al. (2012) analyze the effects of cultural values on M&As. Boudier and Lochard (2013) explore the impact of deregulation in services on cross-border M&A focusing on the OECD economies. Ahern and Harford (2014) analyze the importance of inter-sectoral linkages on waves of M&As within the U.S. We extend this literature by offering a perspective, through the lenses of an economic model, on the importance of policy barriers on cross-border M&A in a variety of service sectors and for a large sample of developed and developing countries. This paper is also connected to the literature on the measurement of policy restrictiveness in services. We use a novel policy database described in Borchert, Gootiiz and Mattoo (2014). Borchert et al (2012) use this data to explore the implication of policy restrictiveness for landlocked economies in the telecommunication and air transport sectors.1 Our findings also speak to the literature on inter-sectoral linkages in FDI patterns. De- vereux and Griffith (1998) show that previous FDI in manufacturing attracts additional manufacturing FDI for the U.S. case. Head et al. (1995, 1999) analyze the case of Japan, 1 Van der Marel and Shepherd (2013) use the data to explore the implication of policy restrictiveness for cross-border trade in services. 4 with similar results. Gross et al. (2005) analyze the case for a sample of Japanese outward investment projects to Europe. They demonstrate how Japanese FDI in manufacturing at- tracted other Japanese FDI in services, among others, but they also show that the pattern of inter-sectoral linkages in FDI changed substantially over time. Finally, the paper is also conceptually connected to a series of papers containing findings of what we would label state-dependency in policy effectiveness. Aghion et al (2008) find that the effect of de-licensing in India depends on the institutional framework of different states. Auerbach and Gorodnichenko (2012) claim that the fiscal policy multiplier (the impact of fiscal policy on output) is highly dependent on whether the economy is experiencing a period of expansion or a recession. Aghion et al (2013) find that the effects of product market reform on innovation depends on the importance of the level of protection of intellectual property rights. The paper is structured as follows. Section 2 describes the data and in particular discusses the new policy information drawn from the Services Trade Restrictions Database. Section 3 outlines a stylized model of cross-border M&A. Section 4 describes the empirical strategy, while Section 5 contains the empirical results. Finally, Section 6 concludes. 2 Data description 2.1 Cross-border Mergers and Acquisitions Data We use a comprehensive dataset on global mergers and acquisitions from ThomsonReuters Platinum database, spanning the period 2003-2009.2 The dataset consists of individual cross- border equity deals between the home country of the acquirer and the host country where the target firm is domiciled. A deal’s sector affiliation is determined based on the target firm’s SIC classification. 2 We focus on the M&A part of investment flows for two reasons. First, modeling the choice between M&A and Greenfield as a mode of investment goes beyond the scope of this paper, though we plan to investigate this important aspect in future work. Second, the best Greenfield data with global coverage that is currently available are not of a quality comparable to the M&A data used here. In addition, its sectoral breakdown is less detailed and often is defined differently than in the M&A data. 5 Table 1 provides basic summary statistics on our investment data. Aggregating informa- tion on individual deals across years, we have a total of roughly 19,000 M&As. The total value of investment covered amounts to 5 trillion USD, of which 2.8 trillion are in services sectors. Half of these investments (1.4 trillion) is concentrated in the services sub-sectors for which we have policy information.3 Banking is quantitatively the most important sector in our sample, followed by Telecommunications. The distribution of M&A flows between countries, although aggregated across years and sectors, still exhibits a large mass point at zero. This is made clear by the third column, where we report the percentage share of non-zero observations over the 21,525 possible country-pair observations . Even considering the totality of the sectors, we observe positive M&A flows only in 14% of the possible cases. This probability is naturally lower in the case of each single sector. We characterize the M&A data further by looking at the profile of M&A inflows into host countries. The distribution of inbound investment is highly skewed. At the top end, one country receives a total of 588 inflows (the U.S.). Figure 1 shows that the attractiveness for M&A, which underpins the data’s skewness at the extensive margin, is closely linked to the host economy’s GDP A similar relationship holds between the number of deals received and a country’s GDP per capita, though it is not as tight as with market size, see Figure 2. Highlighted in a lighter color are seven East Asian economies that appear to be particularly sought-after locations for services M&A inflows, at least beyond levels their per capita income would suggest in this unconditional scatter plot.4 The extensive margin displays similar features when sliced along the home country dimension, indicating that a few economies account for the majority of outbound M&A activity. 3 The category “Other services” includes Construction Services, Gas and Electricity, Business Services and Personal Services such as Health Services and Education Services. 4 Interestingly, firms from India, China and Malaysia are also initiating more outward services M&A deals than their country’s income per capita would suggest. 6 2.2 A New Policy Database 2.2.1 Information on Policy Barriers to Trade in Services Borchert, Gootiiz and Mattoo (2014) describe a project to collect primary data on policies affecting international trade in services. The resulting Services Trade Restrictions Database contains information on legal provisions affecting services trade and investment, including by establishing a commercial presence abroad. It is thus ideally suited to be matched with data on actual cross-border investment flows, variation in which could be expected to reflect the impact of policy barriers. The new database covers the following five major services sectors: financial services (banking and insurance), telecommunications, retail distribution, transportation5 and pro- fessional services, with each of these broad sectors further disaggregated into subsectors. It covers a total of 103 economies, of which 79 are developing countries and 24 OECD coun- tries, representing all the world’s regions and income groups. First-hand information from developing countries was collected by administering a survey instrument whereas informa- tion for OECD countries was obtained from publicly available sources.6 To the best of our knowledge, no other data source provides comparable information on barriers to services trade in a consistent manner for such a wide range of services sectors and countries. The primary focus of the database is to gather information on policies and regulations that potentially constitute a discriminatory barrier for foreign services providers, as well as certain key aspects of the regulatory environment.7 Regulatory measures affecting foreign investment are organized along the following broad categories: 5 Regarding policies governing cross-border trade in international air passenger transportation services, the Database draws on the WTO’s QUASAR database since it represents the most comprehensive source currently available on bilateral air services agreements, covering over 2000 agreements. 6 To ensure data accuracy, all policy information has been reviewed by government officials, though not all countries eventually responded to the vetting request. 7 For every service sector included, the database covers the most important mode(s) of supplying that particular service, i.e. cross-border delivery or the movement of a natural person, in addition to establishing commercial presence. In this paper we focus predominantly on measures affecting foreign investment and include other measures only to the extent that they can be expected to have a bearing on M&A capital flows. 7 • Requirements on the legal form of entry and restrictions on foreign equity; • Limits on licenses and discrimination in the allocation of licenses; • Transparency and accountability of licensing; • Restrictions on ongoing operations; • Relevant aspects of the regulatory environment. This core set of variables, which is available for every subsector, is supplemented with sector-specific variables, for instance whether in telecommunications foreign providers are allowed to operate their own international gateways or to offer voice-over-IP services. 2.2.2 Quantification of Policy Information It is notoriously difficult to gauge the restrictiveness of policies affecting services trade and investment because of their variety and complexity (see the survey by Deardoff and Stern 2008). In this paper we use the Services Trade Restrictiveness Index (STRI) developed by Borchert, Gootiiz and Mattoo (2014). The STRI is a scalar measure of overall openness for a given subsector-mode combination, e.g. for accepting bank deposits (subsector) by establishing commercial presence abroad (mode). All applicable measures within each such combination are evaluated and the overall policy regime is judged to be one of five possible “types”: completely open, i.e. no restrictions at all; completely closed, i.e. no foreign entry allowed at all; virtually open but with minor restrictions; virtually closed but with very limited opportunities to enter and operate; and a final residual ”middle” category of regimes which allow entry and operations but impose restrictions that are neither trivial nor virtually prohibitive. Each of these five regimes is assigned a value on an openness scale from 0 to 1 with intervals of 0.25. Once a score has been attached to each category, STRI values can be aggregated across sectors using weights that reflect the relative importance of constituent services sectors in domestic value added for an average industrialized country. More detailed 8 information about the data and the construction of the STRI can be found in Borchert, Gootiiz and Mattoo (2012b).8 The index number approach adopted here contrasts with methods of econometrically estimating the restrictiveness of policies based on their impact on some outcome variable of interest, controlling for other determinants. A measure of restrictiveness thus derived can obviously not be employed in an analysis of policy impact, for the variation in the outcome variable has already been used to pin down the relative effect of policy measures. Since in this paper we are interested in the relative effect of policy barriers on investment flows in services sectors, our measure of policy restrictiveness needs to be based on exogenous judgment that is not by construction linked to the dependent variable of interest. The STRI measure is relatively simple and transparent, and the method builds on a long tradition of restrictiveness indices, ranging from the pioneering work in this area of the Australian Productivity Commission to more complex weighted averages (see OECD 2009, 2011). We acknowledge the subjectivity of this approach, but given data constraints as well as the wide range of sectors covered, there is no obviously superior method of quantification. The subjectivity is somewhat mitigated by the extensive consultations conducted with the private sector and regulators on how scores are best assigned. We would argue that on balance the STRI approach is better equipped than any fixed algorithm to turn the rich and difficult-to-quantify aspects of policy information into a broadly plausible restrictiveness score. 2.2.3 Patterns of Services Trade Policy Based on the approach laid out in the previous section, we begin by mapping out patterns of services trade protection for the sample of countries and sectors for which we are able to match services M&A data. Figure 3 presents each country’s overall index of services trade restrictions as it relates to 8 For the present analysis, we assume policy restrictiveness to be exogenous data. In their study of telecommunications and transport sector policies, Borchert et al. (2012) go further and account for the endogeneity of policy choices. 9 the establishment of commercial presence (mode 3), plotted against that country’s per capita income, plus a simple linear fit of the relationship. Figure 3 reveals a great deal of variation in the overall restrictiveness of services trade policies. On the one hand, most OECD countries are clustered together at the bottom-right corner, reflecting their general overall openness (notwithstanding some rather restricted subsectors, an aspect we will return to below). On the other hand, some fast-growing dynamic economies in East Asia such as Thailand, Malaysia, Indonesia, the Philippines and China appear to have relatively significant services trade barriers. The same is true for India and some countries in the Middle East, including Iran, Egypt and Gulf Cooperation Council (GCC) countries. Some of Africa’s poorest nations also have rather restrictive services policies. In particular, Ethiopia and Zimbabwe turn out to be among the least open countries in the sample (top-left corner). However, other African nations, such as Ghana, Mozambique and Senegal appear to be relatively open.9 Thus, the restrictiveness of applied policies varies widely among developing countries. Figure 4 provides a more detailed breakdown of STRI scores by world region and by service sector. It is evident again that countries in the Middle East as well as in South and East Asia impose on average the highest barriers to investment. But the relative restrictive- ness across sectors is surprisingly similar in developing and industrial countries. Figure 4 shows that even those OECD and ECA (Eastern Europe and Central Asia) countries that are widely known for their open policies regarding the establishment of commercial presence still maintain substantial barriers to investment in transportation and professional services.10 Thus, we find that countries in South and East Asia are characterized both by large M&A inflows as well as by relatively restrictive investment policies. This is one puzzle we rationalize by suggesting that the impact of investment policies is likely to be state dependent. In the next section, we present a stylized model to formalize this idea. 9 We interpret the apparent openness of some poor developing countries with caution, though, as low STRI scores may in part reflect the absence of any sectoral regulation, in which case the resulting openness is qualitatively different from the predictable market access in countries that formally institute open policies. 10 By focussing on mode 3 STRI scores, the Figure is likely to even understate the true degree of restric- tiveness as barriers to the international movement of professionals (mode 4) are critical in these sectors and mode 4 is often thought to be complementary to commercial presence. 10 3 A Simple Model with Inter-Sectoral Linkages In this section we outline a simple model featuring heterogeneous firms and inter-sectoral linkages in order to illustrate the notion of state-dependency in policy effectiveness. The main innovation in the model is the structure of production. 3.1 Set-up Suppose there are N countries in the world. In a generic country i a representative consumer enjoys utility from consuming Agricultural goods, Manufacturing goods and Services. The utility function is assumed to be Cobb-Douglas (the subscripts i are omitted for simplicity): 1−α−β α β U = Ca Cm Cs (1) The utility maximization problem implies the following demand functions: PC Ca = (1 − α − β ) (2) Pa PC Cm = α (3) Pm PC Cs = β (4) Ps Where P is the aggregate price index, a Cobb-Douglas aggregator of the price indexes in the three sectors: 1−α−β α β P = Pa Pm Ps (5) ¯. Labor is the only primary factor of production, and the total endowment of labor is L Agriculture is a perfectly competitive sector. An homogeneous agricultural good is produced using only labor under constant return to scale: Ya = La , where La is the labor employed in the agricultural sector. Agriculture is also the numeraire, hence Pa = 1, from which it follows that the nominal wage is also one in every sector (we assume free labor mobility across sectors). 11 Manufacturing is a perfectly competitive sector as well. A homogenous manufacturing output is produced using labor and intermediate services (Ysm ): Ym = (Lm )γ (Ysm )1−γ (6) Equation (6) is the only non-standard assumption in the model. In order to motivate this assumption empirically, we present in Figure (5) the importance of the services covered by the STRI index as inputs for manufacturing production in several countries. We use data from the OECD STAN Input-Output tables for the mid-2000 for the OECD countries and some developing countries.11 We report in Figure (6) the importance of manufacturing as intermediate input in the production of the services covered in the STRI index. Looking at Figures (5) and (6) two main features emerge. First, the average input share of STRI services into manufacturing production (32%) is more than two times higher than the importance of manufacturing in the production of services (14%).12 Second, there is a lot of heterogeneity in the Input-Output structure of different countries.13 From equation (6) it also follows that Pm = wγ Ps1−γ . Services, finally, is a sector characterized by product differentiation and monopolistically competitive firms. The output of services is a C.E.S. aggregator of individual service varieties 1 with elasticity of substitution σ = 1−ρ >1: [∫ 1 ]ρ 1 ρ Ys = ys (ω ) dω (7) 0 11 Unfortunately, only about a third of the countries for which the STRI index is built are represented in the OECD-STAN database. 12 Obviously, manufacturing itself is a more important input into the production of manufacturing than services, and services are a more important input into the production of services than manufacturing. We are not explicitly incorporating these features into the model, and choosing instead to write ”net-output” production functions for each sector. 13 This also calls for some caution about the practice, common in the literature, of using the input-output structure of the U.S. as a proxy for the input-output structures of other countries. 12 The price index is the standard C.E.S. ideal price index: [∫ 1 ] 1− 1 σ 1 −σ Ps = (ps (ω )) (8) 0 Firms in the service sector are heterogeneous in the sense of Melitz (2003). The produc- tion function for each service variety is ys = aLs , where a is a measure of labor productivity, drawn from a distribution with a cumulative distribution function G(a) over a support [aL , aH ]. Naturally, the most productive firm in a given country i is aiH . Profit maximization in the manufacturing sector implies the following demand function for labor in manufacturing and intermediate services: Lm = γPm Ym (9) Pm s Ym = (1 − γ ) Ym (10) Ps Profit maximization in the service sector implies a standard optimal pricing rule as a markup over marginal costs: 1 ps (a) = (11) aρ Profit maximization in the manufacturing sector and utility maximization implies the fol- lowing demand for each service variety: ( )−σ ps (a) ys (a) = (Cs + Ysm ) (12) Ps To close the model, we need to impose market clearing. The three goods market clearing conditions read: Ya = Ca (13) Ym = Cm (14) Ys = Cs + Ysm (15) 13 While the labor market clearing condition is: ¯ = La + Lm + Ls L (16) 3.2 M&As in Services We introduce the possibility of M&As in services from a given country i to a country j .14 Under the assumption that the target firm inherit the productivity level of the parent firm, the additional profits obtainable from this operations can be written, using equations (4) (10) and (11) as : 1 Πij = psj (ai )ysj (ai ) −ysj (ai ) (17) ai ( ) 1 σ −1 Pj Cj Pmj Ymj = (1 − ρ) 1−σ −σ ai β + (1 − γ ) (18) ρ Psj Psj Psj Importantly, from equation (18) we can see how these extra-profits are a linear function −1 of aσ i , which can be taken as a positive proxy for productivity (since σ > 1). Further manipulating equation (18) by using equations (3) and (14), we can express equation (18) as: −1 Πij = Υj · Pj Cj (β + (1 − γ ) α) · aσ i (19) where Υj = (1 − ρ) ρ1−σ1 P 1−σ > 0. From equation (19) it is clear how the slope of the profit sj function depends essentially on three things: i) the size of the target economy (Pj Cj ), ii) the structural composition of the target economy (parameters β and α), and iii) the extent of the inter-sectoral linkages present (the parameter γ ). 14 Technically we are not distinguishing here between M&As and greenfield FDI. See Nockle and Yeaple (2007) for a model explaining possible determinants of these different foreign markets entry modes. Also, we are purposely ignoring trade in any of the three sector as well as M&As in Agriculture and Manufacturing. While all these features could be added to the model, at the cost of adding complexity, they would not modify the key insights we want to focus on. See Helpman, Melitz, and Yeaple (2004) for a model featuring both trade and FDI. 14 We assume that the M&A implies fixed costs, which we model in a very flexible way as depending both on bilateral factors, such as the distance (physical and cultural) between countries, and on source and host country specific factors, including the policy environment Φj : Cij = Cij (τij , Φj , Xi , Xj ) (20) Naturally, we observe an M&A from a given country i to a given country j i.i.f.: Πij Πij > Cij or Πij − Cij > 0 or >1 (21) Cij The three conditions expressed in equation (21) are identical.15 It is possible to draw −1 16 the middle one using equations (19) and (20) as a function of aσ i . Figure (7) reports the extra-profits and the cost for an M&A from a country i to a country k and to a country j . The two cases are different. In the case of country j , the most productive firm in country i has a productivity high enough to insure a profitable investment. This is not the case of country k , where no M&As from country i is profitable. A formal way of seeing this, is to realize that from equation (21) it is possible to derive for each pair of countries, for example for i and j , the cut-off productivity level a∗ ij above which a firm from country i will find profitable to acquire a firm in country j . This will be the productivity level of a firm whose extra-profit are just enough to recover the fixed cost of the investment. Using equations (19) and (20) this is equal to: Υj · Pj Cj (β + (1 − γ ) α) · (a∗ ij ) σ −1 = Cij (22) So we get ( ) σ− 1 Cij 1 a∗ = (23) ij Υj · Pj Cj (β + (1 − γ ) α) 15 This approach based on a zero-profit condition is similar to Santos Silva et al. (2014), including the fact that firm productivity is the source of randomness while other covariates (such as policy barriers) are taken as given. 16 We use a figure originally proposed by Helpman, Melitz, and Yeaple (2004). 15 A further alternative way of stating the conditions expressed in equation (21) is to say that in order to observe an M&A from a country i to a country j , the productivity of the most productive firm in country i must exceed a∗ ∗ ij . In Figure (7) we have aiH > aij but aiH < a∗ ik . Finally, we can formally derive the number of M&A deals as M &AN ij um = Ni Vij , where Ni are the number of firms in country i and Vij represents the fraction of firms in the country i who are productive enough to be able to acquire a firm in country j . Vij is defined by   1 − G(a∗ ) if a > a∗ ij iH ij Vij =  0 otherwise Assuming a particular functional form for G(a) allows to get an explicit expression for M &AN ij um . 3.3 State Dependency in Policy Effectiveness The framework presented in this section allows considering what we call state-dependency in policy effectiveness. In Figure (8) the solid lines report the additional profits and the cost for an M&A operation from a country i to country j and k when the cost of the operations are 1 1 Cij and Cik . Obviously, we would not observe any M&A from country i to either country j or k because even the most productive firm in country i would make losses. Now suppose that the same liberalization policy is implemented in country k and j , with the result of reducing 1 2 1 2 the cost of investing for firms from country i from Cik to Cik and from Cij to Cij . The new situation is described by the dashed lines. Crucially, the profit function for investing in country j is now making it profitable for some firms located in country i to invest in country j . However, in country k , even with a more favorable policy environment, we still do not observe investment flows from country i. The fact that the same policy change can generate two different outcomes depending on other conditions prevailing in the host country is what we call state-dependent policy effectiveness. Importantly, in the picture the different slopes were key in delivering the results. The slopes of the curves, in turn, as underlined in equation 16 (19), depend on the size of the target economy, the structural composition of the GDP, and the extent of the inter-sectoral linkages present in the economy. The simplest way to express this argument more formally, is assuming that the support of productivity draws is itself stochastic, and in particular that the productivity of the most productive firm in a given country i, aiH is uniformly distributed on an interval [aiH , aiH ]. Then, we can express the probability of observing an M&A from country i to country j as P rob(M &Aij = 1) = P rob(aiH > a∗ ∗ ij ) = 1 − F (aij ) (24) aiH −aiH where F (aiH ) = aiH −aiH is the distribution of aiH .17 Taking the derivative of (24) with respect to the host country policy environment, and using equation (23), gives: ( ) σ− 1 ′ ∂P r(M &Aij = 1) 1 1 Cij 1 Cij =− <0 (25) ∂ Φj σ − 1 aiH − aiH Υj · Pj Cj (β + (1 − γ ) α) Cij ′ ∂Cij where Cij = ∂ Φj > 0. We see from (25) that restrictive policies have a negative impact on the probability of observing M&As. However, from (25) we also see that the cross derivatives w.r.t Φj and α, γ and Pj Cj are positive. For instance: ( ) σ− 1 −1 ∂ 2 P r(M &Aij = 1) Cij 1 Υj · Pj Cj (1 − γ ) =∆ >0 ∂ Φj ∂α Υj · Pj Cj (β + (1 − γ ) α) [Υj · Pj Cj (β + (1 − γ ) α)]2 (26) ′ Cij 1 1 where ∆ = (σ −1) aiH −aiH Cij 2 > 0. From (26) we see that a larger market and a composition of the GDP more skewed towards the sectors who demand services, dampens the negative effect of restrictive policies on the probability of observing M&As.18 17 We naturally operate under the assumption that aiH > aiL . 18 An analogous analysis can be performed for the numbers of M&As between the two countries. 17 4 Empirical Strategy In this section we outline our empirical strategy, which is inspired by the model presented in the previous section. Since our policy variable is country-specific, but not country-pair specific, we adopt a two-stage empirical strategy. In the first stage, we analyze the impact of geographical and cultural factors on the intensive and extensive margin of the bilateral M&As in services. As for the intensive margin, we evaluate the probability of observing bilateral M&As between a country i and j . We assume that τij is proportional to a vector of trade frictions β −ϵij Tij which is stochastic due to unmeasured bilateral frictions (ϵij ), so that τij = Tij e . Unobserved frictions ϵij are i.i.d. unit normal distributed. As a proxy for Tij we use the physical distance between two countries and dummy variables for the presence of a border, the fact of sharing a common language and sharing the same legal origin. Under these assumptions, equations (19) and (20) and (21) suggest a Probit model with a full set of home and host country fixed effects.19 In our first stage, we also analyze the intensive margin of M&A activity by using the number of M&A deals. In this case, we use a Poisson Pseudo Maximum Likelihood estimator (discussed in Santos-Silva and Tenreyro, 2006), including both home country and host country fixed effects (we will call this model PPML-N). In the second stage, we use the estimated host country fixed effects from the Probit model and the PPML-N model and relate them to variation in policy restrictiveness, conditional on appropriate covariates Xj . Thus the second stage’s estimable equation takes the form: δj = β0 + β1 yj + β2 ϕj + β3 (ϕj Xj ) + β4 Xj + ϵj (27) In equation (27) we introduce interaction effects between the measure of policy restric- tiveness and some country characteristics in Xj so as to explore state dependency in policy 19 In general, the Probit model with fixed effects suffers from the so-called incidental parameters problem. However, Heckman (1981) provides Monte Carlo based evidence that in a panel context, the Probit model with fixed effect performs relatively well when the number of periods analyzed exceeds eight. In a cross- sectional context with bilateral dependent variables, as ours, the correspondent concept to the number of periods in a panel context is the number of trading partners, which in our case is above 100. 18 effectiveness, i.e. the possibility that the effect of a given policy depends on other factors. Inspired by our model, in particular by equation (19), we will explore the size and the composition of the GDP in the host economy as possible relevant factors. 5 Results 5.1 First Stage: Gravity-type Determinants Table 2 reports the results of the first-stage Probit model for aggregate data (recall from Section 2 that the M&A data are cumulated from deal level data over the period 2003-09). We report separately the results for total manufacturing (column two) and total services (column three). The effect of distance on the probability of observing cross-border M&A is negative and statistically significant for both the manufacturing sector and the service sector, and the coefficient on the border dummy is roughly the same for services as for 20 manufacturing. Sharing a common language increases the probability of observing M&A in services more than it does in manufacturing. Given the critical role of communication in the delivery of intangible services, this finding is hardly surprising. As another proxy for cultural proximity, we use a common religion21 . As table 2 shows, the coefficient attached to common religion is positive and highly statistically significant, and the coefficient for services is much larger than the one for manufacturing. A common colonial origin has a positive impact on the probability of observing M&A in services but not in manufacturing. In contrast, a common legal system is quantitatively slightly less important for services M&A than for manufacturing. Trade in goods has a positive and statistically significant impact on both manufacturing and services M&A, and so does the existence of a Bilateral Investment Treaty (BIT) between the two countries. However, sharing a regional trade agreement seems to affect the M&A in services, but not in manufacturing. 20 See Keller and Yeaple (2013) for a model where the negative relation between distance and multinational activity is explained through the presence of barriers to the diffusion of knowledge. 21 Common religion is the probability that randomly extracting two people from the two countries, they belong to the same religion. Formally, it is the sum of the products of the shares of population belonging to the same religions in the two countries. 19 We move to the analysis of the determinants of the numbers of deals in Table 3. Let PPML-N denote the results for the numbers of deals using the PPML estimator. Observing Table 3 we see how sharing a common language, a common religion and a common legal origin have a much stronger impact on services M&A than on manufacturing M&As. The coefficients for distance and the border dummy are similar for manufacturing and services. We conclude that geographical barriers have similar effects on M&A in manufacturing and services, while cultural barriers are more important for services than for manufacturing when considering the aggregate data. However, given the heterogeneity of different services sectors, we explore in Table 4 and 5 whether our results differ significantly across different service sectors. Hence, we run our first-stage models (Probit and PPML-N) using six different service sectors: Accounting, Banking, Insurance, Retail, Telecommunications, and Transport.22 Considering first the Probit results (Table 4), the coefficient on distance is negative and statistically significant for every sector. A shared border, instead, seems to be a relevant determinant of cross-border M&A only in the case of Transport and Retail services. In both these sectors the need to establish cross-border networks for delivery and distribution are likely to drive firms to establish a commercial presence in neighboring countries. The effect of sharing a common language seems to affect more M&As in Telecommunications as well as Banking and Insurance. Common religion appear to be an important determinant of bilateral M&A flows in all sectors, except Transportation services. A common origin for the legal system appears to have a positive and highly significant effect on the probability of observing cross-border M&A in Banking, Insurance and Retail. The presence of a BIT, on the other hand, displays a statistically significant coefficient only in the case of Banking. The results obtained using the Probit model are broadly confirmed moving to the number of deals (PPML-N), as Table 5 shows.23 22 The Professional Services sector in our sample includes the Accounting, Engineering and Research sector, and Legal Services. Since out of the 804 deals classified under Professional Services in Table 2, only 4 transactions arise from the legal sector, we decided to exclude it from the analysis. Engineering and research services are included under the category ”Accounting.” 23 The border is now found to be a significant determinant only in Retail. The coefficient on common language is larger for Banking, Telecom and Accounting services than for other sectors. Common religion 20 5.2 Second Stage: Aggregate Policy Restrictiveness Following our empirical strategy, in the second stage of the analysis we take the host country fixed effects and we relate them to a set of explanatory variables, including a measure of 24 market size and our policy restrictions measure. Before reporting the results based on a regression framework, it is instructive to visualize some relationships of first-stage fixed effects with variables of interest. Figure 9 reports the scatter plot of the host country fixed effects, obtained from the probit regressions of the existence of positive bilateral flows of services M&A in all sectors, against the log of the country GDP, averaged other the period 1998-2002, which we take as a first proxy of the market size.25 As the picture shows, there is a strong positive association between these two variables. Figure 10 reports the scatter plot of the same host country fixed effects against our index of restrictiveness in services. The fact that there is only a weak negative relation between the two measures appears to be essentially due to a set of countries that are characterized by both a relatively closed policy regime and high values for the fixed effect. This group of countries is not random. It includes China, India, Indonesia, Thailand, Malaysia, Vietnam and Philippines. In these countries a relatively high level of policy restrictiveness co-exists with a high level of M&A inflows. Following the insight of the model we presented on the importance of inter-sectoral link- ages, we correlate the Probit host fixed effect with the share of manufacturing sector in total value-added (VA). The results are shown in Figure 11, in which a strong positive correla- tion is found between the share of the manufacturing sector in value added in 2002 and the services sector fixed effects from the first-stage Probit. displays positive and statistically significant coefficients in all sectors, except Transportation. 24 In the first stage, probit host country fixed effects cannot be estimated for eight economies (Bolivia, Cameroon, Algeria, Ethiopia, Lesotho, Madagascar, Mozambique and Nepal. On average these countries are similar to the rest of the sample in terms of log income and log per capita income, and exhibit only slightly more restrictive policies. Thus we believe that there is no bias introduced to second stage results by losing these countries. 25 In order to avoid presenting contemporaneous correlations, we chose to report all the results using covariates from 2002 and before, naturally except for the STRI variable. 21 We repeat the same exercise using the fixed effects obtained with the PPML-N model, and we obtain virtually identical results.26 In order to validate the visual intuition within a regression framework, we present corre- sponding econometric results in Tables 6 and 7. In Table 6 we report the results from the Probit model. In the first specification, we only include the log of GDP and the Service Trade Restrictiveness Index (STRI). Consistent with Figures 10 and 11, we find a positive and highly significant effect of GDP (which turns out to be stable across all specifications), and an almost zero coefficient for the STRI. In the second to the fourth column, we add the interaction terms of the STRI with the log of GDP, the value-added share of manufac- turing in GDP, and the value-added share of services in GDP. It is only when we include the interaction terms between the STRI and the shares of manufacturing and services in value added that the coefficient on the STRI becomes negative and statistically significant, while the interaction terms are positive and statistically significant. This finding points towards state-dependency in policy effectiveness: the effect of restrictive policies is damp- ened in countries were the share of manufacturing and service sectors are larger. In the fifth column we simultaneously use the interactions with manufacturing and services shares, respectively. We find that the interaction term with the share of manufacturing is positive and statistically significant, and larger than the interaction with the service share. In the subsequent specification, we control also for a host of additional country characteristics such as a measure of financial development (total Credit to Private sector as a fraction of GDP), the log of per capita GDP, the cost of starting a business, and indicators of political stability and government effectiveness. The results are qualitatively unchanged.27 Lastly, we check whether our results are driven by particular countries. In column (7) we exclude China and India as well as the U.S. and the UK and we obtain virtually identical results.28 In Table 7, 26 The results are available upon request. 27 The negative and statistically significant coefficient on logged GDP per capita in Tables 6 and 7 reflects more the impact of few outliers (India, Saudi Arabia and Venezuela) rather than a genuinely strong negative correlation between the level of development and the attractiveness for M&As in Services. The same holds true in Tables 8 and 9 as well. We also considered a specification in which the STRI was interacted with logged GDP per capita (rather than the shares of manufacturing and services in GDP). Neither in the probit nor in the PPML specification is the interaction term significant, though. 28 The same happens if we exclude China and Korea, the countries with the lowest input shares of services 22 we report the results for the PPML-N model, which are qualitatively similar. Since the Services Trade Restrictions Database also describes in detail how policy restric- tiveness varies across sectors within any given country, we go one step further and exploit the cross sectoral variation in the STRI (Tables 8 and 9). Importantly, this approach em- ploys sectoral fixed effects obtained from the first stage which are then pooled across services sectors. For our policy measure, we use a pooled vector of disaggregated country-by-sector STRI scores, which allows a considerable increase in the number of observations. We follow the structure of tables 6 and 7 but in addition include sector fixed effects in every specifi- cation. Table 8 reporxts the result for the Probit model. The results are broadly consistent with those reported in Tables 6 and 7. In this case, however, the interaction term between the STRI and GDP also appears to be positive and highly statistically significant when inserted alone in the regression (column 2).29 The results are also qualitatively unchanged when additional control variables are included (column 6) or when China, India, the U.S. and the UK are excluded (column 7). In Table 9 we report the corresponding results with respect to the incoming number of M&A deals. Results are analogous to the ones presented in Table 8. However, unlike in the case of the Probit model, now the coefficient on the interaction term between the STRI and the manufacturing share is appreciably larger than the coefficient on the interaction term between the STRI and the services share. Thus manufacturing and services shares exert a roughly similar effect in terms of attenuating policy restrictiveness as far as the overall probability to attract foreign investment is concerned. However, when the number of deals as a metric of the “intensive margin” is considered, the manufacturing share is relatively more important in enhancing attractiveness for FDI. While we are aware of the limitations of a cross-section approach, dictated by the cross- sectional nature of the policy information available, the results presented in Tables 6 to 9 into manufacturing production. 29 Again, we also considered a specification in which the STRI was interacted with logged GDP per capita (rather than the shares of manufacturing and services in GDP). The interaction term turns out to be positive and statistically significant in both the pooled probit and the pooled PPML-N specification, respectively. These results would suggest that restrictive policies have a larger impact in poorer countries. 23 lend empirical support to the hypothesis that policy restrictiveness is a significant factor in determining bilateral M&A flows in services sectors. As we would expect, we find a baseline negative effect of policy restrictions on foreign investment. At the same time, though, we find intriguing evidence of policy effectiveness to be state-dependent. Specifically, relatively high shares of manufacturing and services in value added seem to allow countries to maintain a more restrictive regime without deterring M&A in service sectors. This effect might be sizeable. Take for example the cases of Vietnam and Botswana, two countries which share a roughly similar STRI score. The industrial structure of the two countries, however, is quite different. Vietnam has a manufacturing share of 20% of GDP and a services share of about 38%. In Botswana, the service share is 43%, while the manufacturing share is only 3.8%. Suppose both countries would engage in policy reforms that lowered their STRI scores in an identical manner. Using as reference point the results displayed in column (7) of Tables 6 and 7, that same liberalization policy would in Botswana have an impact about five times as large as in Vietnam in terms of both the probability of observing an M&A flows and of number of M&A deals.30 5.3 Individual Policy Measures (Second Stage) While the STRI is an composite index, the Services Trade Restriction Database also makes available a range of individual policy measures.31 While there are econometric problems related to multicollinearity and degrees of freedom, one would ideally want to identify the individual policy measures that most affect M&A decisions. We divide the policy measures into different categories, namely restrictions affecting market entry, licensing and operations. We present results by pooling the different sectors, conditioning on GDP and including sector fixed effects. Table 10 reports the results.32 In the first four columns, we report the results obtained 30 Alternatively, the same policy restriction would have much larger negative effects if imposed in Botswana than in Vietnam. 31 These include measures that underpin the STRI score plus additional variables and contextual informa- tion. 32 We pool different sectors in order to increase the degrees of freedom in our regressions. However, we 24 using the Probit host country fixed effects. In the first column, we only include a dummy indicating the presence of a restriction in setting up a branch and the maximum amount of capital that a foreign investor can hold when acquiring a domestic company. The positive and significant coefficient on the restriction to open a branch is likely to reflect some substitution among different forms of entry: firms engage more in M&A when they cannot enter a foreign market simply by setting up a branch.33 Equity restrictions do not seem to be significantly correlated with the probability of a particular country receiving any cross-border M&A inflow in services. In the second column, we include several restrictions concerning licensing: a difference in criteria for domestic versus international firms in applying for a license, transparency of criteria to obtain a license, the automatic nature of the license renewal, and transparency in obtaining reasons why a license is denied. We find a negative and statistically significant coefficient on the transparency about the reason for a license denial. Surprisingly, we find a positive and significant coefficient on the dummy for differential criteria for licensing for domestic and foreign firms. A way to interpret this result is that in the presence of discrimi- natory licensing, firms might prefer to acquire a local firm (rather than to establish a branch or a subsidiary), and thus be able to apply for licences as a “domestic” firm. Among the restrictions affecting operations, we explore the role of nationality limits on employees and on the board of directors. We find a negative and significant coefficient on the nationality restrictions for the general employees, but not on the restrictions specific to the board of directors which, in practice, often does not add any new information that would not already be embodied in the variable capturing foreign equity participation limits. Column (4) proposes a model where we include the key restrictions from each of the different types of measures, retaining only those measures that exhibit significant coefficients in the previous specifications. In this joint specification the restriction on setting up a branch, the lack of transparency on the reason for a license denial, and limits to the nationality of are limited by the fact that not all our individual policy variables are available for all the countries in our sample. This is the reason why estimations in Table 10 have fewer observations than in Tables 8 and 9. 33 The difference between a branch and a subsidiary is that the former is legally still part of the parent firm, while the second is a separate legal entity. 25 employees stand out as significant determinants of the Probit host fixed effects. Columns (5) to (8) of Table 10 report the results obtained for the PPML-N model, which broadly confirm what emerged from the Probit model. The only exceptions are that we find a positive and weakly significant coefficient on the maximum equity share that a foreign company can hold in an acquired domestic company (column (5)), and a negative and weakly significant coefficient on the lack of clarity on the criteria to get a license (column (6)). When we include different types of restrictions in the same specification (column (8)), the policies that still display a significant coefficient are the restriction in setting up a branch, equity restrictions on M&As, lack of transparency on the reason for a license denial, and limits to the nationality of the employees. Considering the overall results displayed in columns (4) and (8) of Table 10, we conclude that the policies that seem to matter the most in restricting the inflows of M&As in services are restrictions on setting up branches, the rules concerning the nationality of the employees, and the lack of transparency in the denial of licenses. The first indicator may reflect substi- tution possibilities between modes of entry; the second, the sensitivity of investors to their ability to bring in relevant skills regardless of nationality, which is what makes locations such as Singapore particularly attractive; and the last indicator may reflect the importance investors attach to transparent decision-making by governments. 6 Conclusions In this paper we present evidence on the determinants of cross-border mergers and acqui- sitions (M&A) in services sectors. We develop a stylized model of M&A that incorporates firm heterogeneity and inter-sectoral linkages. The model predicts that the incidence of M&A deals depends, inter alia, on the target economy’s size, industrial structure and in- vestment policies, as well as on bilateral transactions costs. We examine these predictions using comprehensive bilateral M&A flow data and detailed information on policy barriers, drawing upon a new database of investment restrictions in services sectors. Detailed gravity estimates show, among other things, that geographical factors affect M&A in services and 26 manufacturing similarly but institutional and cultural factors such as a common legal system or religion affect M&A in services more than in manufacturing. Our main interest, however, lies in exploring the impact of policy barriers and other country characteristics after controlling for these bilateral factors. Indeed, our results show that restrictive investment policies dampen the probability of M&A inflows but this negative effect is mitigated in countries with relatively large shares of manufacturing and (to a lesser extent) services in GDP. The same results hold for the number of M&A deals received. These findings suggest that the impact of policy is state-dependent and related to the composition of GDP in the target economy. The effect induced by different industrial structures can be substantial. When considering countries that exhibit roughly the same level of policy restrictiveness but differ in their industrial structure, the cost of protectionism or, conversely, the gains from policy reform vary by a wide margin. There are several avenues for future research. First, it would be interesting to explore more fully the concept of state-dependency in policy effectiveness, which has general ap- plicability. Recognizing this phenomenon also has implications for empirical analysis. Any analysis aimed at assessing the effectiveness of a certain policy could (and maybe should) check whether its impact is state-independent, i.e. explore relevant dimensions along which the effect on the outcome of interest is being dampened or magnified. In so doing, the present study continues a tradition of other policy studies that have employed interaction terms between the policy variable and factors that are conjectured to affect its potency. Second, it would be interesting to study the dynamics of M&A in Services and how they relate to the dynamics of M&As in Manufacturing, given that recent evidence indicates services FDI tends to follow manufacturing FDI (Nefussi and Schwellnus 2010; Kolstad and Villanger 2008). Such an analysis would help us understand the spillover effects of policy restrictions on both manufacturing and services investment and assess the impact of destination country industrial structure on each type of investment. Finally, the analysis undertaken here would clearly be improved by the availability of better data. For instance, regarding FDI flows an effort needs to be made to improve the quality of data on greenfield investment and to harmonize that data (e.g. in terms of classification of sectors) with the data on M&A. 27 References [1] Aghion, P. Burgess, R. Redding, S. and Zilibotti, F. (2008) “The Unequal Effects of Lib- eralization: Evidence from Dismantling the Licence Raj in India” The American Economic Review, 98(4): 1397-1412. [2] Aghion, P. Howitt, P. and Prantl, S. (2013) “Patent Rights, Product Market Reforms, and Innovation”, mimeo [3] Ahern, K. R., Daminelli, D. and Fracassi, C. (2012), “Lost in Translation? The Ef- fect of Cultural Values on Mergers around the world.” Journal of Financial Economics, forthcoming. [4] Ahern, K. and Harford, J. (2014), “The Importance of Industry Links in Merger Waves.” Journal of Finance, forthcoming. [5] Auerbach, A. and Gorodnichenko, Y. (2012), “Measuring the Output Responses to Fiscal Policy,’ American Economic Journal: Economic Policy, 4(2), pages 1- 27. [6] Borchert, I., Gootiiz, B., Grover, A. and Mattoo, A. (2012), “Landlocked or Policy Locked? How Service Trade Protection Deepens Economic Isolation”, World Bank Policy Research Working Paper No. 5942. [7] Borchert, I., Gootiiz, B. and Mattoo, A. (2014), “Policy Barriers to International Trade in Services: Evidence from a New Database,” World Bank Economic Review 28(1): 162-188. [8] Borchert, I., Gootiiz, B. and Mattoo, A. (2012b), “Guide to the Services Trade Restric- tions Database”, World Bank Policy Research Working Paper No. 6108. [9] Boudier F. and Lochard, J (2013), “How do Cross-border Mergers and Acquisitions Answer to Deregulation in Services?”, The World Economy, 36(11): 1424-1441. [10] Coeurdiacer, N. De Santis, R.A. and Aviat, A. (2009), “Cross Border mergers and acquisitions and European integration”, Economic Policy, 57, 55-106. [11] Deardorff, A.V. and Stern, R.M. (2008), “Empirical Analysis of Barriers to International Services Transactions and the Consequences of Liberalizatio”, Chapter 5 in: Mattoo, A., Stern, R.M. and Zanini, G. (eds.), A Handbook of International Trade in Services, Oxford University Press, pp. 169-220. [12] Devereux, M.P. and Griffith, R. (1998), “Taxes and the location of production: evidence from a panel of US multinationals.” Journal of Public Economics, 68, 335-367. [13] Di Giovanni, J. (2005), “What drives capital flows? The case of cross-border M&A activity and financial deepening”, Journal of International Economics, 65, 127-149. 28 [14] Gross, D. M. Raff, H. and Ryan, M (2005), “Inter- and intra-sectoral linkages in foreign direct investment: evidence from Japanese investment in Europe”, Journal of the Japanese and International Economies, 19, 110-134. [15] Head, K. and Ries, J. (2008), “FDI as an outcome of the market for corporate control: Theory and Evidence”, Journal of International Economics, 74(1), 2-20. [16] Head, K. Ries, J. and Swenson, D. (1995) “Agglomeration benefits and location choice: evidence from Japanese manufacturing investments in the US”. Journal of International Economics, 38, 223-247. [17] Head, K. Ries, J. and Swenson, D. (1999) “Attracting foreign manufacturing: Invest- ment Promotion and agglomeration”. Regional Science and Urban Economics, 29, 197-218. [18] Heckman, J.J.(1981) “The incidental parameters problem and the problem of initial conditions in estimating a discrete-time discrete-data stochastic process” in C. Manski and D. McFadden, eds., Structural Analysis of Discrete Data with Econometric Applications, MIT Press (1981) [19] Helpman, H. Melitz, M. and Rubinstein, Y. (2008) “Estimating trade flows: trading partners and trading volumes”. Quarterly Journal of Economics, 123(2), 441-487. [20] Helpman, H. Melitz, M. and Yeaple, S.R. (2004), “Export vs. FDI with heterogenous firms”. The American Economic Review, 94(1), 300-317. [21] Hijzen, A., G¨org, H. and Manchin, M. (2008), “Cross-border mergers and acquisitions and the role of trade costs”, European Economic Review, 52, 849-866. [22] Hyun, H.J. and Kim, H.H. (2010), “The Determinants of Cross-border M&A: The Role of Institutions and Financial Development in the Gravity Model”, The World Economy, 33(2), 292-310. [23] Keller, W. and Yeaple, S. (2013), “Gravity in the Knowledge Economy,” American Economic Review, 103(4), 1414:1444. [24] Kolstad, I. and Villanger, E. (2008), Determinants of foreign direct investment in ser- vices European Journal of Political Economy, 24, 518533. [25] Nefussi, B. and Schwellnus, C. (2010), Does FDI in manufacturing cause FDI in business services? Evidence from French firm-level data, Canadian Journal of Economics 43(1), 180-203. [26] Nockle, V. and Yeaple, S. (2007) “Cross-Corder Merger and Acquisitions versus Green- field Foreign Direct Investment: The Role of Firms Heterogeneity,” The Journal of Inter- national Economics, 72(2), 336-365. 29 [27] OECD (2011), “Services Trade Restrictiveness Index: Opening the Way for Services Trade Liberalization”, downloadable at http://www.oecd.org/trade/stri. [28] OECD (2009), “Methodology for Deriving the STRI”, paper presented at the OECD Experts Meeting on the Services Trade Restrictiveness Index (STRI), 2-3 July 2009, Paris. [29] Rossi, S. and Volpin, P. (2004) “Cross-country Determinants of Mergers and Acquisi- tions.” Journal of Financial Economics, 74, 277-304. [30] Santos Silva, J.M.C. Tenreyro, S. (2006) “The Log of Gravity” The Review of Economics and Statistics, Vol. 88, No. 4: 641-658. [31] Santos Silva, J.M.C., Tenreyro, S., and Wei, K. (2014) “Estimating the Extensive Mar- gin of Trade” Journal of International Economics, forthcoming. [32] UNCTAD (2013) “World Investment Report 2013”, UNCTAD, Geneva. [33] Van der Marel, E. and B. Shepherd (2013) “Services Trade, Regulation and Regional Integration: Evidence from Sectoral Data” The World Economy, June 2013. 30 Table 1: M&A Investment: Descriptive Statistics (2003-09, cumulated) Total. Num. Total Value Share of country pairs (%) with non-zero values All Sectors 19792 5004656 14 STRI Services 3944 1409067 7 –Banking 1032 585309 3 –Insurance 347 139796 2 –Retail 612 144694 2 –Telecom 787 372757 3 –Transport 433 122758 2 –Professional 733 43752 2 Other Services 7201 1443917 7 Manufacturing 6254 1602664 8 Other 2393 549008 4 31 Table 2: First Stage: Probit Estimates Sector Total Manufacturing Services Distance -0.698*** -0.650*** -0.595*** (0.049) (0.065) (0.060) Contiguity 0.297** 0.384** 0.318** (0.132) (0.166) (0.147) RTA 0.206** 0.068 0.247** (0.091) (0.113) (0.111) Comm legal system 0.237*** 0.311*** 0.189** (0.060) (0.080) (0.074) Colony 0.132 -0.213 0.406*** (0.132) (0.159) (0.141) Comm language 0.325*** 0.315*** 0.336*** (0.087) (0.119) (0.106) Comm religion 0.773*** 0.644*** 1.014*** (0.120) (0.157) (0.153) Goods trade (2002) 0.076*** 0.082*** 0.112*** (0.014) (0.024) (0.020) BIT (2002) 0.226*** 0.159** 0.180** (0.062) (0.079) (0.076) Host FE Yes Yes Yes Home FE Yes Yes Yes Obs 11341 6345 7662 Log-L -2015.599 -1236.104 -1290.71 Robust standard errors in parenthesis. ***,**,* statistically significant at 1%, 5% and 10% respectively. 32 Table 3: First Stage: PPML Estimates - Numbers of Deals Sector Total Manufacturing Services Distance -0.606*** -0.589*** -0.566*** (0.058) (0.068) (0.072) Contiguity -0.184 -0.133 -0.146 (0.128) (0.136) (0.119) RTA 0.231* -0.009 0.352** (0.129) (0.163) (0.148) Comm legal system 0.051 -0.026 0.216** (0.088) (0.102) (0.089) Colony 0.569*** 0.451*** 0.573*** (0.092) (0.110) (0.106) Comm language 0.390*** 0.376*** 0.507*** (0.113) (0.121) (0.111) Comm religion 1.528*** 1.209*** 1.857*** (0.169) (0.217) (0.213) Goods trade (2002) 0.212*** 0.256*** 0.284*** (0.031) (0.041) (0.041) BIT (2002) -0.116 0.038 -0.083 (0.085) (0.097) (0.090) Host FE Yes Yes Yes Home FE Yes Yes Yes Obs 11341 6345 7662 Log-L -6971.544 -3209.426 -2705.399 R-squared 0.9307 0.9187 0.8953 Robust standard errors in parenthesis. ***,**,* statistically significant at 1%, 5% and 10% respectively. 33 Table 4: First Stage: Disaggregate Probit Estimates Sector Services Acc. Bank Ins Ret Tel Transp Distance -0.595*** -0.305* -0.668*** -0.521*** -0.516*** -0.488*** -0.415*** (0.060) (0.159) (0.081) (0.129) (0.114) (0.080) (0.104) Contiguity 0.318** -0.123 -0.059 -0.077 0.526** 0.038 0.391* (0.147) (0.268) (0.176) (0.247) (0.218) (0.183) (0.228) RTA 0.247** 0.460 0.288* 0.222 0.394* 0.235 0.129 (0.111) (0.331) (0.154) (0.235) (0.213) (0.154) (0.217) Comm legal system 0.189** 0.131 0.222** 0.259* 0.272** 0.092 0.169 (0.074) (0.180) (0.096) (0.153) (0.137) (0.101) (0.138) Colony 0.406*** 0.479* 0.395** 0.459* 0.199 0.320* 0.262 (0.141) (0.273) (0.162) (0.236) (0.209) (0.168) (0.219) Comm language 0.336*** 0.341 0.353** 0.676*** 0.313 0.343** 0.298 (0.106) (0.253) (0.139) (0.222) (0.192) (0.140) (0.185) Comm religion 1.014*** 1.174*** 0.761*** 1.235*** 1.212*** 0.955*** 0.413 (0.153) (0.422) (0.211) (0.361) (0.315) (0.205) (0.289) Goods trade (2002) 0.112*** 0.259*** 0.111*** 0.218*** 0.076 0.086*** 0.111*** (0.020) (0.092) (0.030) (0.065) (0.052) (0.030) (0.041) BIT (2002) 0.180** 0.046 0.307*** 0.074 0.110 -0.046 0.016 (0.076) (0.197) (0.104) (0.174) (0.146) (0.100) (0.134) Host FE Yes Yes Yes Yes Yes Yes Yes Home FE Yes Yes Yes Yes Yes Yes Yes Obs 7662 1654 4971 2284 2559 4700 2418 Log-L -1290.71 -287.8649 -726.275 -337.1912 -409.0344 -715.6322 -433.5455 Robust standard errors in parenthesis. ***,**,* statistically significant at 1%, 5% and 10% respectively. 34 Table 5: First Stage: Disaggregate PPML Estimates - Number of Deals Sector Services Acc. Bank Ins Ret Tel Transp Distance -0.566*** -0.401*** -0.658*** -0.388** -0.561*** -0.553*** -0.610*** (0.072) (0.148) (0.121) (0.152) (0.124) (0.112) (0.151) Contiguity -0.146 -0.417** -0.217 -0.312 0.448** -0.310 0.234 (0.119) (0.210) (0.205) (0.268) (0.214) (0.194) (0.233) RTA 0.352** 0.118 0.255 0.538* 0.544* 0.586** 0.167 (0.148) (0.349) (0.222) (0.314) (0.281) (0.249) (0.304) Comm legal system 0.216** -0.279 0.236* 0.498** 0.274 0.334** 0.051 (0.089) (0.206) (0.139) (0.209) (0.204) (0.166) (0.201) Colony 0.573*** 0.608*** 0.759*** 0.448* 0.199 0.518*** 0.265 (0.106) (0.230) (0.188) (0.250) (0.223) (0.175) (0.234) Comm language 0.507*** 0.830*** 0.422** 0.905*** 0.227 0.248 0.501** (0.111) (0.233) (0.178) (0.282) (0.227) (0.190) (0.242) Comm religion 1.857*** 2.609*** 1.971*** 2.484*** 1.641*** 1.990*** 0.766 (0.213) (0.534) (0.336) (0.576) (0.419) (0.360) (0.475) Goods trade (2002) 0.284*** 0.565*** 0.291*** 0.402*** 0.255*** 0.211*** 0.293*** (0.041) (0.118) (0.062) (0.080) (0.089) (0.060) (0.094) BIT (2002) -0.083 -0.407** 0.220 -0.192 0.020 -0.177 0.036 (0.090) (0.204) (0.136) (0.228) (0.151) (0.144) (0.186) Host FE Yes Yes Yes Yes Yes Yes Yes Home FE Yes Yes Yes Yes Yes Yes Yes Obs 7662 1654 4971 2284 2559 4700 2418 Log-L -2705.399 -525.6424 -1221.727 -483.879 -640.1007 -1098.198 -621.1776 R-squared 0.8953 0.9329 0.7314 0.6470 0.7832 0.6332 0.6880 Robust standard errors in parenthesis. ***,**,* statistically significant at 1%, 5% and 10% respectively. 35 Table 6: Second Stage: Host Probit Fixed Effects and STRI, Serv aggregate (1) (2) (3) (4) (5) (6) (7) Log GDP (Avg 98-02) 0.2132*** -0.0657 0.1884*** 0.2510*** 0.2161*** 0.2841*** 0.2484*** (0.0301) (0.2090) (0.0348) (0.0377) (0.0424) (0.0572) (0.0618) Log STRI -0.0742 -2.2514 -1.1164*** -1.3096*** -2.1320*** -1.9791*** -2.1032*** (0.1465) (1.6137) (0.3062) (0.4532) (0.4465) (0.6054) (0.6041) (Log STRI)x(Avg GDP) 0.0914 (0.0677) Share Manuf VA (2002) -0.2091*** -0.2083*** -0.1502** -0.1603*** (0.0494) (0.0484) (0.0575) (0.0562) (Log STRI)x(Manuf VA) 0.0691*** 0.0681*** 0.0498*** 0.0545*** (0.0139) (0.0138) (0.0172) (0.0168) Share Serv VA (2002) -0.0762*** -0.0616** -0.0603** -0.0660** (0.0263) (0.0234) (0.0294) (0.0306) (Log STRI)x(Serv VA) 0.0219*** 0.0184*** 0.0205** 0.0217** (0.0078) (0.0069) (0.0088) (0.0091) Credit PrivSec (Avg 98-02) -0.0006 -0.0014 (0.0020) (0.0021) Cost Start Busi (2005) -0.0007 -0.0006 (0.0017) (0.0016) Govt Effect (2002) 0.2785* 0.2461 (0.1426) (0.1498) Political Stab (2002) 0.0847 0.1100 (0.1034) (0.1079) Log GDP PC -0.3255*** -0.2729*** (0.0919) (0.0996) Obs 91 91 91 91 91 88 84 R-squared 0.284 0.291 0.394 0.297 0.400 0.452 0.367 Robust standard errors in parenthesis. ***,**,* statistically significant at 1%, 5% and 10% respectively. Column (7): Excluding China, India, UK and the US 36 Table 7: Second Stage: Host PPMLN Fixed Effects and STRI, Serv aggregate (1) (2) (3) (4) (5) (6) (7) Log GDP (Avg 98-02) 0.3202*** -0.2834 0.2818*** 0.3831*** 0.3266*** 0.4491*** 0.3849*** (0.0500) (0.3648) (0.0512) (0.0640) (0.0646) (0.0920) (0.0995) Log STRI -0.2729 -4.9850* -2.2937*** -2.3619*** -4.0157*** -3.9678*** -4.2045*** (0.2806) (2.8280) (0.5753) (0.7271) (0.7896) (1.0147) (0.9764) (Log STRI)x(Avg GDP) 0.1978* (0.1178) Share Manuf VA (2002) -0.4117*** -0.4107*** -0.3421*** -0.3573*** (0.0879) (0.0864) (0.0996) (0.0984) (Log STRI)x(Manuf VA) 0.1334*** 0.1318*** 0.1099*** 0.1165*** (0.0250) (0.0248) (0.0302) (0.0298) Share Serv VA (2002) -0.1286*** -0.1038** -0.1069** -0.1180** (0.0451) (0.0448) (0.0513) (0.0511) (Log STRI)x(Serv VA) 0.0370*** 0.0312** 0.0365** 0.0393** (0.0133) (0.0128) (0.0156) (0.0156) Credit PrivSec (Avg 98-02) -0.0022 -0.0036 (0.0034) (0.0038) Cost Start Busi (2005) -0.0016 -0.0013 (0.0027) (0.0026) Govt Effect (2002) 0.3390 0.2849 (0.2344) (0.2435) Political Stab (2002) 0.2606 0.2953* (0.1640) (0.1701) Log GDP PC -0.4892*** -0.3871** (0.1618) (0.1708) Obs 91 91 91 91 91 88 84 R-squared 0.230 0.247 0.372 0.243 0.378 0.463 0.391 Robust standard errors in parenthesis. ***,**,* statistically significant at 1%, 5% and 10% respectively. Column (7): Excluding China, India, UK and the US. 37 Table 8: Second Stage: Host Probit Fixed Effects and STRI, pooled (1) (2) (3) (4) (5) (6) (7) Log GDP (Avg 98-02) 0.1276*** 0.0754** 0.1275*** 0.1724*** 0.2243*** 0.2440*** 0.2073*** (0.0190) (0.0298) (0.0191) (0.0197) (0.0407) (0.0272) (0.0302) Log STRI 0.0435** -0.6165** -0.1638*** -0.3499*** -0.5944** -0.7285*** -0.7421*** (0.0219) (0.2637) (0.0612) (0.1210) (0.2547) (0.1554) (0.1556) (STRI)x(Avg GDP) 0.0261** -0.0130 (0.0106) (0.0137) Share Manuf VA (2002) -0.0289*** -0.0531*** -0.0339*** -0.0289*** (0.0097) (0.0114) (0.0106) (0.0106) (STRI)x(Manuf VA) 0.0118*** 0.0170*** 0.0102*** 0.0103*** (0.0031) (0.0037) (0.0032) (0.0032) Share Serv VA (2002) -0.0292*** -0.0442*** -0.0243*** -0.0266*** (0.0063) (0.0090) (0.0074) (0.0074) (STRI)x(Serv VA) 0.0060*** 0.0104*** 0.0089*** 0.0092*** (0.0020) (0.0027) (0.0021) (0.0022) Credit Priv Sec (Avg 98-02) 0.0017* 0.0005 (0.0011) (0.0011) Cost Start Busi (2005) -0.0006 -0.0005 (0.0010) (0.0009) Govt Effect (2002) 0.0004 -0.0427 (0.0825) (0.0842) Political Stab (2002) 0.0590 0.1045* (0.0589) (0.0574) Log GDP PC -0.2641*** -0.2114*** (0.0506) (0.0541) Sector Fixed Effects Yes Yes Yes Yes Yes Yes Yes Obs 339 339 339 339 339 332 308 R-squared 0.207 0.220 0.230 0.246 0.289 0.338 0.235 Robust standard errors in parenthesis. ***,**,* statistically significant at 1%, 5% and 10% respectively. Column (7): Excluding China, India, UK and the US. 38 Table 9: Second Stage: Host PPML-N Fixed Effects and STRI, pooled (1) (2) (3) (4) (5) (6) (7) Log GDP (Avg 98-02) 0.1294*** 0.0355 0.1282*** 0.1920*** 0.2524*** 0.3462*** 0.2884*** (0.0301) (0.0482) (0.0286) (0.0308) (0.0584) (0.0414) (0.0455) Log STRI 0.0374 -1.1505*** -0.4262*** -0.4357** -1.1217*** -1.2868*** -1.2750*** (0.0347) (0.4182) (0.0950) (0.1906) (0.3870) (0.2363) (0.2371) (STRI)x(Avg GDP) 0.0470*** -0.0121 (0.0166) (0.0193) Share Manuf VA (2002) -0.0629*** -0.0956*** -0.0735*** -0.0653*** (0.0153) (0.0163) (0.0157) (0.0157) (STRI)x(Manuf VA) 0.0263*** 0.0329*** 0.0250*** 0.0245*** (0.0048) (0.0053) (0.0047) (0.0048) Share Serv VA (2002) -0.0387*** -0.0608*** -0.0349*** -0.0370*** (0.0096) (0.0129) (0.0109) (0.0109) (STRI)x(Serv VA) 0.0071** 0.0137*** 0.0132*** 0.0133*** (0.0031) (0.0038) (0.0031) (0.0032) Credit PrivSec (Avg 98-02) 0.0004 -0.0013 (0.0015) (0.0016) Cost Start Busi (2005) -0.0010 -0.0008 (0.0014) (0.0013) Govt Effect (2002) -0.0001 -0.0706 (0.1196) (0.1215) Political Stab (2002) 0.1954** 0.2568*** (0.0913) (0.0904) Log GDP PC -0.4382*** -0.3567*** (0.0788) (0.0839) Sector Fixed Effects Yes Yes Yes Yes Yes Yes Yes Obs 339 339 339 339 339 332 308 R-squared 0.183 0.201 0.235 0.213 0.284 0.355 0.300 Robust standard errors in parenthesis. ***,**,* statistically significant at 1%, 5% and 10% respectively. Column (7): Excluding China, India, UK and the US 39 Table 10: Individual Policy Variables Model (1) (2) (3) (4) (5) (6) (7) (8) FE used Probit Probit Probit Probit PPMLN-N PPMLN-N PPMLN-N PPMLN-N Restriction Affecting: Form of Entry Licensing Operations Overall Form of Entry Licensing Operations Overall Acq. Max. Equity 0.003 0.005* 0.006* (0.002) (0.003) (0.004) No branch entry 0.340*** 0.216** 0.447*** 0.297* (0.082) (0.089) (0.136) (0.174) Lic. diff. ri. 0.374** 0.283* 0.539** 0.372 (0.147) (0.162) (0.237) (0.265) Lic Crit non-public -0.330 -0.904* -0.413 (0.385) (0.539) (0.749) Lic not automatic -0.007 -0.144 (0.100) (0.174) No reasons lic rejec -0.237** -0.282*** -0.436** -0.539*** (0.111) (0.101) (0.181) (0.170) Nat. req. emp. -0.220** -0.256*** -0.399** -0.456*** (0.094) (0.091) (0.166) (0.162) Nat. req. bod. 0.062 0.073 (0.136) (0.252) Log GDP (Avg 98-02) 0.126*** 0.094*** 0.115*** 0.086*** 0.126*** 0.099* 0.122** 0.089* (0.021) (0.033) (0.030) (0.031) (0.034) (0.056) (0.052) (0.051) Constant -3.337*** -2.100*** -2.499*** -1.794** -3.422*** -1.925 -2.408* -2.125 (0.578) (0.795) (0.729) (0.763) (0.913) (1.365) (1.259) (1.298) Sector Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.228 0.276 0.272 0.349 0.190 0.176 0.170 0.238 N 287 149 147 141 287 149 147 141 Robust standard errors in parenthesis. ***,**,* statistically significant at 1%, 5% and 10% respectively. 40 Figure 1: Total number of M&A deals in services sectors, by GDP of host economy 600 Total number of deals 200 400 CHN IND IDN MYSTHA VNM PHL 0 20 22 24 26 28 30 Log GDP (Avg 98−02) Note: top five recipients: USA (588), GBR (456), AUS (193), CHN (188), DEU (166). Figure 2: Total number of M&A deals in services sectors, by per capita GDP of host economy 600 Total number of deals 200 400 CHN IND IDN THA MYS VNM PHL 0 4 6 8 10 Log GDP p.c. (Avg 98−02) Note: top five recipients: USA (588), GBR (456), AUS (193), CHN (188), DEU (166). 41 Figure 3: Global Services Trade Restrictiveness 100 ETH 80 Services trade restrictiveness index (STRI) IND ZWE IRN 60 QAT PHL EGY DRC BHR KWT IDN JOR THA PAN OMN MYS BGD TUN NPL VNM LBN BWA SAU 40 LKA NAM DZA CHN MWI UGA BLR ZAFVEN TZAKEN YEM CRI MEX MLI LSONGAPAK UKR URY CIVCMR RUS ITA FRA FIN RWA KHMUZB TUR BRA CHL KOR JPN PRT BEL ZMB HND MAR CAN DNK 20 BDI MDG GHASEN ALB MOZ GTM COL GRC AUT DEU USA PRY PER KAZARG HUN MUS CZE AUS ESP KGZ BGR ROM SWE GBR BOL DOM NICMNG LTU IRL GEO ARM NZL NLD POL TTO ECU 0 5 6 7 8 9 10 11 12 GDP per capita 2007, PPP Source: Services Trade Restrictions Database Figure 4: Services Investment Restrictiveness, by Region and Sector 80 Services trade restrictiveness index 60 40 20 0 GCC SAR MENA EAP AFR LAC OECD ECA Banking Insurance Telecom Retailing Transportation Legal Note: 103 countries included Source: Services Trade Restrictions Database 42 mean of MANUF_INPUT_SHARE mean of STRI_INPUT_SHARE 0 .1 .2 .3 .4 0 .1 .2 .3 .4 .5 CHN GRC IND LVA BRA EST TUR BEL KOR CYP TWN CHL POL NOR ZAF IRL MEX AUT IDN SVN ESP ISR BGR SWE ITA POL SVK MEX JPN HUN LTU ITA CHE NLD ROM FIN CAN LTU FIN AUS 43 CHL FRA SWE DEU NOR TUR GBR GBR AUS MLT LVA CHE HUN PRT USA DNK AUT CZE GRC BGR NLD SVK PRT ESP FRA CAN CYP USA ISR IND MLT ZAF DNK ROM DEU IDN LUX BRA CZE TWN BEL LUX SVN JPN IRL CHN EST KOR Figure 5: Input shares of STRI Services in Manufacturing, Simple Averages Figure 6: Input Shares of Manufacturing in STRI Services, Simple Averages Figure 7: Extra-Profits and Costs of M&As: Two Possibilites Figure 8: Example of State-Dependency of Policy Effectiveness 44 Figure 9: Probit Host Country Fixed Effects and SIZE 3 2 IND CHN IDN VNM THA MYS 1 MNG KHM PHL 0 −1 20 22 24 26 28 30 Log GDP (Avg 98−02) Figure 10: Probit Host Country Fixed Effects and STRI 3 2 CHN IND IDN VNM THA Services FE MYS 1 MNG KHM PHL 0 −1 0 20 40 60 80 STRI M3 Note: Solid line represents linear fit without East Asian economies (highlighted). 45 Figure 11: Probit Host Country Fixed Effects and Manufacturing Share 3 2 IND CHN IDN VNM THA Services FE MYS 1 MNG KHM PHL 0 −1 0 10 20 30 Manufact share VA 46