Foreign Direct Investment Modes and Local Vertical Linkages

Transcription

1 Foreign Direct Investment Modes and Local Vertical Linkages Chrysovalantou Milliou and Apostolis Pavlou June 2013 Abstract This paper studies a MNE s choice of FDI mode, Acquisition of a domestic rm or Green eld Investment, in the context of a vertically related industry with local input sourcing. We show that the Acquisition mode can prevail in markets with stronger downstream competition as well as in markets in which the input suppliers hold more bargaining power. We also show that the presence of vertical linkages in uences the FDI mode choice in favor of Acquisition. Our welfare analysis suggests that a host country s FDI policy should be designed in such as way that distinguishes among the di erent FDI modes and among di erent industry sectors, such as sectors with more or less competitive upstream and downstream markets. Keywords: foreign direct investment; green eld investment; acquisition; vertical relations; two-part tari s JEL classi cation: L13; F12; F23 Milliou: Department of International and European Economic Studies, Athens University of Economics and Business, Athens 10434, Greece, cmilliou@aueb.gr; Pavlou: Department of Economics, Athens University of Economics and Business, Athens 10434, Greece, pavlou@aueb.gr. This research has been co- nanced by the European Union (European Social Fund - ESF) and Greek national funds through the Operational Program "Education and Lifelong Learning" of the National Strategic Reference Framework (NSRF) - Research Funding Program: Thalis - Athens University of Economics and Business - "New Methods in the Analysis of Market Competition: Oligopoly, Networks and Regulation". Full responsibility for all shortcomings is ours.

2 1 Introduction When a multinational enterprise (MNE) considers expanding into a foreign market through Foreign Direct Investment (FDI), one of the main strategic decisions it has to make, is the mode of its FDI. Extensive empirical evidence indicates that in the last few years, most FDI is carried out through the acquisition of foreign rms assets rather through the establishment of new rms abroad, known as Green eld Investment. In fact, in 2008, the value of cross-border mergers and acquisitions was about 70% of the global FDI ows (UNCTAD, 2008). 1 Empirical evidence (see e.g., Belderbos et al., 2001, Giroud, 2007, and Jordaan, 2008 and 2011) also indicates that FDI often occurs in vertically related industries and results into local sourcing. That is, many MNEs enter into the downstream segments of a foreign markets and produce their goods using intermediate products that they source from local upstream suppliers. 2 As argued by Qiu and Tao (2001) and Lin and Saggi (2007), local sourcing arises because of technological reasons, such as high transportation costs, or because of policy restrictions, such as local content requirements imposed by the host country s government. In this paper, we endogenize the FDI mode: Acquisition of a foreign rm or Green eld Investment. A key and novel aspect of our approach is that we take into account the above empirical observations: We consider a MNE which operates in a vertically related industry and sources its inputs locally. This is important because in a vertically related industry, the FDI modes can a ect not only the competition between the MNE and the domestic rms, but also the trading with the input suppliers. We examine how the input prices respond to the di erent FDI modes, and, in turn, how this a ects the MNE s choice of FDI mode and its subsequent impact on the host country s welfare. Moreover, allowing for product di erentiation and for bargaining over the contract terms, we investigate, the role of the intensity of downstream competition and of the relative bargaining power in the FDI s mode selection. Finally, considering, as a benchmark, the case of a one-tier industry, 1 Most of the growth in international production between 1990 and 2000 has been through cross-border mergers and acquisitions, and not through Green eld Investments (UNCTAD, 2000). Moreover, in 2011 cross-border mergers and acquisitions rose 53% in value terms, while Green eld Investments remained at (UNCTAD, 2012). 2 Belderbos et al. (2001), in a survery of Japanese electronics manufacturing a liates in 24 countries, provide evidence of local backward linkages. For similar evidence in the case of Malaysia and Vietnam see Giroud (2007), and for Mexico see Jordaan (2008 and 2011). 1

3 we also examine how the presence of vertical linkages in uences the mode of FDI. In our model, a MNE considers expanding to a host country in which there are, initially, a domestic upstream rm and two domestic downstream rms. It can expand either by creating a new downstream rm (Green eld Investment) or by acquiring one of the domestic downstream rms (Acquisition). In the former case, the MNE incurs a xed cost. In the case of Acquisition instead, it pays the acquisition price, which is determined endogenously. Upon entry, the MNE sources the input from the domestic upstream supplier through bargaining over the terms of a two-part tari contract and competes in quantities with the domestic downstream rm(s). In this environment, we nd that a MNE pays a lower input price when it undertakes Green eld Investment than when it acquires a domestic rm. This results from the inability of the domestic upstream monopolist to commit not to impose a negative externality to one downstream rm by o ering better terms to another downstream rm - its "commitment problem". 3 Because of the commitment problem, the upstream monopolist subsidizes the downstream production. When the MNE expands through Green eld Investment, the number of downstream rms increases; hence, downstream competition intensi es. This increases the severeness of the commitment problem by reinforcing the incentives of the upstream monopolist to behave opportunistically and results into larger subsidization compared to that under Acquisition. In view of the above, when the MNE makes its FDI mode choice it faces the following trade-o s: Under the Acquisition mode, downstream competition is weaker, but under the Green eld Investment mode, downstream e ciency is higher. Moreover, under the Green eld Investment mode, it incurs the xed cost of setting up a new production plant while under the Acquisition mode, it pays the acquisition price which equals a domestic rm s pro ts under the alternative expansion mode. Not surprisingly, the MNE prefers the Acquisition mode only when the xed cost of Green eld Investment is too high. This occurs because the fact that, as we know from Salant et al. (1983), the sum of two rms pro ts in triopoly exceeds the pro ts of a duopolist together with the increases e ciency under the Green eld Investment mode, o set the impact of the weaker downstream competition under the Acquisition mode. 3 For more on the upstream monopolist s commitment problem see e.g., O Brien and Sha er (1992), McAfee and Schwartz (1994) and (1995), Rey and Vergé (2004), de Fontenay and Gans (2005), and Rey and Tirole (2006). 2

4 Importantly, we nd that FDI is more likely to take the form of Acquisition, instead of the Green eld Investment form, in markets with stronger downstream competition as well as in markets with more powerful domestic upstream suppliers. We reach this result because when products are closer substitutes and/or when the downstream rms have less bargain power, the MNE s pro ts with Green eld Investment decrease. Since the acquisition price equals the gross pro ts with Green eld Investment mode, a direct implication of this is that the Acquisition becomes cheaper then, and hence, more attractive. On the basis of this nding, one could claim that our paper provides support to the empirically observed tendency of the MNEs to expand through Green eld Investment in less developed countries and through cross-border mergers and acquisitions in developed countries (UNCTAD, 2012), given that in developed countries the MNEs face more e cient downstream competitors and more powerful input providers. Furthermore, we nd that the vertical market structure plays an important role for the FDI mode choice. In particular, we note, rst, that a MNE faces, due to the subsidization, a lower marginal cost in a vertically related industry with a concentrated upstream sector than in a one-tier industry, and second, that in a vertically related industry, the Green eld Investment mode has an advantage relative to the Acquisition mode which is absent in a one-tier industry: it results into lower wholesale prices, and thus, into a higher e ciency. 4 In light of these, we would expect that the MNE is more likely to choose the Green eld Investment in a vertically related industry than in a one-tier industry. Interestingly, we nd that the opposite holds. The main force behind this is that in a vertically related industry, the MNE, even when it has all the bargaining power, does not extract all the surplus. It compensates the upstream monopolist for its "outside option"; hence, the upstream monopolist gets part of the surplus generated by FDI through the xed fee. As a result, even though under the Green eld Investment mode, the MNE is more e cient in a vertically related industry, its pro ts can be higher in a one-tier industry. The above nding suggests that we should observe more often cross-border acquisitions when the MNEs expand into host countries in which they source inputs from a local concentrated upstream market than when they produce the inputs in house or when they obtain them from perfectly competitive upstream markets. Moreover, it suggests that the recent increase in cross-border acquisitions relative to Green eld Investment, could be due to the 4 Note that the case of a one-tier industry coincides with that of a vertically related industry with a perfectly competitive upstream market. 3

5 fact that the rms recently undertake more often FDI in sectors in which they rely heavily on input producers. The two entry modes of FDI di er also in terms of their impact on the host country s welfare. Green eld Investment, by increasing both downstream competition and e ciency, leads to higher consumer surplus than Acquisition or autarky. In addition, while Green eld Investment, is always welfare-enhancing, Acquisition is welfare-detrimental. Combining these with our ndings regarding the MNE s choice of FDI mode, we are in the position to draw some policy conclusions. For instance, we conclude that when the xed cost of setting up a production plant is su ciently high, the host country s government should undertake measures that encourage Green eld Investment and discourage cross-border acquisitions. Further, since Green eld Investment is less likely to be chosen in vertically related industries with concentrated upstream markets than in one-tier industries or in vertically related industries with competitive upstream markets, we also conclude that the government should vary its policy among di erent industry sectors accordingly. Extending our model, we allow for trading through linear wholesale price contracts and nd that when the domestic upstream supplier is powerful enough, the use of two-part tari contracts favors the Acquisition mode over the Green eld Investment mode more than the use of wholesale price contracts. Therefore, we identify another factor, besides the intensity of downstream competition, the vertical linkages and the distribution of bargaining power, that can in uence the choice of the FDI mode: the contract type used. Our paper is clearly related to the literature on FDI with strategic interactions. Within this literature, a number of papers (see e.g., Borvatn, 2004, Eicher and Kang, 2005, Mattoo et al., 2004, Muller, 2007, Ra et al., 2009, Qiu and Wang, 2011) have tried to explain, similarly to us, when an why a given entry mode is preferred relative to others. More speci cally, Mattoo et al. (2004), Eicher and Kang (2005), Muller (2007), and Qiu and Wang (2011), have shown that one of the factors that in uences the expansion mode choice is the di erence in e ciency between the MNE and the domestic rms. Allowing for exports, as an alternative expansion mode, Borvatn (2004), Eicher and Kang (2005), Muller (2007), and Ra et al. (2009) have identi ed the transportation/trade costs as also being an important factor for the expansion mode choice. Qiu and Wang (2011) have pointed out that the host country s FDI policy can also a ect the entry mode patterns. All of the above mentioned papers, in line with ours, have shown that the xed cost of Green eld Investment is a crucial determinant of the FDI mode choice. Importantly though, in contrast to our 4

6 paper, they have restricted their analysis to one-tier industries. Recently, a new strand of the literature on FDI with strategic interactions, focusing on vertically related industries has developed. Papers in this strand by Pack and Saggi (2001), Goh (2005), Lin and Saggi (2007), and Balsvik (2010) have studied issues related to vertical technology transfer from the entering MNE to the local upstream suppliers and the subsequent free-riding of the domestic upstream or downstream rms. Ma (2011 and 2012) and Milliou (2013) have studied a MNE s location choice among di erent potential host countries. Lin and Saggi (2011) have analyzed the incentives and the implications of both upstream and downstream FDI by two separate rms. Finally, Carluccio and Fally (2012) have examined the extent of technology adoption in the host country, in a setting where there is initially incompatibility between the foreign and the domestic technologies, and upon foreign downstream entry, the domestic upstream suppliers decide whether to adopt the foreign technology and produce compatible inputs for the MNEs or not. 5 Importantly, in all these papers, FDI takes only the form of Green eld Investment. 6 Obviously, this implies, rst, that the existing literature on FDI in vertically related industries, has ignored the recent dominance of cross-border mergers and acquisitions relative to Green eld Investment, and second, that it has not dealt at all with the choice of the FDI mode. 7 Our paper, attempts to ll this gap in the literature. Furthermore, our paper, unlike the papers in this strand of the existing literature, incorporates bargaining in the trading of the MNE with the local upstream rms and explores its role. The rest of the paper is organized as follows. In Section 2, we describe our main model. In Section 3, we analyze the MNE s FDI mode choice when it operates in a onetier industry. In Section 4, we analyze its choice when it operates in a vertically related industry and explore how the market structure, one-tier vs. two-tier structure, in uences the MNE s preferred FDI mode. In Section 5, we analyze the implications of FDI and its 5 Rodriguez-Clare (1996) and Markusen and Venables (1999) also study FDI in a vertically related industry. However, they consider a setting with monopolistic competition. Therefore, they abstract from strategic interactions among the MNE and the domestic rms. 6 In a recent paper, by Beladi et al. (2013), FDI takes, instead, only the form of a horizontal cross-border merger in a vertically related industry. Beladi et al. (2013), unlike us, assume that pre-merger, the MNE competes with the domestic rms in an international market and that the merger causes a decrease in the number of rms in the market. 7 Note that Balsvik (2010) has considered a MNE which operates in the downstream market of a host country and has examined its choice among importing its input from its home market, vertically integrating with a local upstream rm, or buying the input from a local upstream rm. In other words, he has analyzed the MNE s input sourcing choice, but not the MNE s FDI mode choice since he assumes throughout that the MNE enters through Green eld Investment. 5

7 di erent modes on the host country s welfare. In Section 6, we extend our analysis to the case in which trading takes place through wholesale price contracts. Finally, in Section 7, we conclude. All the proofs are relegated to the Appendix. 2 The Model We consider a domestic market of a host country in which there is, initially, a domestic upstream monopolist, denoted by U, and two domestic downstream rms, denoted by D 1 and D 2. The downstream rms produce a di erentiated nal good using, in a one-to-one proportion, an intermediate good, which they obtain from U. A MNE from another country, contemplates undertaking FDI in the host country in order to sell in its domestic market. 8 It can choose among two FDI modes. First, it may enter by establishing a new production plant in the downstream market of the host country -this is the Green eld Investment (GI) mode. Second, it may acquire one of the host country s domestic downstream rms -this is the Acquisition (AQ) mode. Since the domestic downstream rms are symmetric, without loss of generality, we assume that in the case of AQ, the MNE acquires D 1. The cost of GI consists of the xed cost G > 0 for setting up the production plant. The cost, instead, of AQ consists of the endogenously determined acquisition price, A, i.e., the price for acquiring D 1. We assume that the MNE s subsidiary in the host country, denoted by D m, under both GI and AQ, sources the intermediate product locally, i.e., it obtains it from U. 9 D m competes with the host country s domestic downstream rm(s). In particular, the (inverse) demand faced by each D i, with i = 1; 2; m, in the host country is p i (q i ; Q i ) = a q i Q i, where p i and q i are respectively D i s price and quantity, and Q i is the total quantity of its rival(s). In particular, Q i = q under the AQ mode with i; = m; 2, and Q i = q + q k under the GI mode, with i; ; k = 1; 2; m and i 6= 6= k. The parameter, with 0 < 1, measures product substitutability; namely, the higher is, the closer 8 Note that we implicitly assume that local presence is essential and exclude other ways of serving the host market such as exports. This allows us to focus on the choice of the mode of FDI. The choice between FDI and exports has been the focus of many studies in one-tier industries (see e.g., Ethier, 1986; Ethier and Markusen, 1996; Horstmann and Markusen, 1987, 1992; Motta, 1992; Saggi, 1996). An alternative usti cation for the exclusion of exports could be that the tari s that the host country imposes on imports are prohibitive high or that transportation costs are too high. 9 As mentioned in the Introduction, local sourcing in the context of FDI is empirically relevant. Moreover, local sourcing is a common assumption in the theoretical literature on FDI in vertically related markets, see e.g., Lin and Saggi (2007), Blasvik (2010), and Carluccio and Fally (2012). 6

8 substitutes the products of the downstream rms are. U s marginal cost of production is u, with 0 u < a. Each D i faces no other production cost than the cost of obtaining the input from U. The latter consists of a per-unit of input price, w i, and a xed fee, F i, i.e., rms trade through non-linear two-part tari contracts. 10 Firms play a three-stage game with observable actions. At stage one, the MNE chooses its FDI mode: GI or AQ. In case of AQ, it makes an acquisition o er to D 1, i.e., it speci es A. In turn, D 1 accepts or reects the o er. If the o er is reected, the MNE chooses among GI or no entry at all. At stage two, the domestic upstream rm bargains simultaneously and separately with the downstream rms which operate in the host country s market over their contract terms. Lastly, at stage three, the downstream rms compete in quantities. We model bargaining over the contract terms by invoking the Nash equilibrium of simultaneous generalized Nash bargaining problems, in which the upstream and the downstream bargaining power is given respectively by and 1, with 0 < < 1. This implies that during the negotiations of a bargaining pair (i.e., between U and D i ) each of its agents takes as given the outcome of the simultaneously-run negotiations of the other bargaining pair(s). A key assumption that underlies this modeling approach is that U bargains with the competing downstream rms simultaneously and separately. This assumption is standard in situations with multilateral contracting (see e.g., Cremer and Riordan, 1987, Horn and Wolinsky, 1988, Hart and Tirole, 1990, O Brien and Sha er, 1992, McAfee and Schwartz, 1994 and 1995, Marshall and Merlo, 2004, Rey and Vergé, 2004, Milliou and Petrakis, 2007). 11 As it has been noted by the literature, multiple equilibria can arise in such settings due to the multiplicity of the beliefs that the downstream rms can form when they receive out-of-equilibrium o ers. 12 Following Horn and Wolinsky (1988), Cremer and Riordan (1987), and O Brien and Sha er (1992), we obtain a unique equilibrium by imposing pairwise proofness on the equilibrium contracts. That is, we require that a contract between U and D i is immune to a bilateral deviation of U with a rival downstream rm, holding the contract with D i constant In Section 6, we discuss what happens when trading takes place, instead, through linear wholesale price contracts. 11 The rationale for the separate or private negotiations could simply be that the upstream monopolist has various representatives, each negotiating at the same time with a di erent downstream rm. Two additional usti cations are, rst, that third parties may be unable to observe others dealings and verify them in court, and second, that making one contract contingent on the speci cities of other contracts may be quite complex and thus di cult to implement (see McAfee and Schwartz, 1995). 12 For a detailed discussion see McAfee and Schwartz (1995). 13 Pairwise proofness is closely related to the passive beliefs assumption (see e.g. Hart and Tirole, 1990, 7

9 An additional key assumption of our bargaining game is that the contract terms of a bargaining pair are not contingent on the disagreement of a rival pair. That is, as in Horn and Wolinsky (1988), O Brien and Sha er (1992), McA e and Schwartz (1994 and 1995), Caprice (2006), and Milliou and Petrakis (2007), we assume that a bargaining pair is unable to write and implement a contract specifying di erent contract terms in the event of a successful termination of another pair s negotiations and in the event instead of a breakdown in another pair s negotiations. This assumption captures the idea that bargaining parties cannot commit to a permanent and irrevocable breakdown in their negotiations. We make the following assumption throughout the paper: Assumption 1: () = (5 2) :14 Assumption 1 is a necessary and su cient condition in order to avoid the non-existence of pure strategy pairwise proof equilibria. Non-existence may occur because pairwise proofness leads to negative pro ts for the upstream monopolist, violating thus its individual rationality condition (McAfee and Schwartz, 1995, Rey and Vergé, 2004, and Milliou and Petrakis, 2007). Our notational convention will be as follows. The rst superscript, G or A, will denote respectively whether the MNE has expanded through GI or through AQ. The second superscript will denote the contract type. We will use T, when trading is conducted through two-part tari s, and W when it is conducted through wholesale price contracts. Finally, we will use B in order to denote the benchmark case in which rms operate in a one-tier industry. 3 Benchmark Case: FDI Modes in a One-tier Industry We present here the MNE s FDI mode choice in the benchmark case in which it operates in a one-tier industry. More speci cally, we assume that initially there are ust two competing domestic rms in the host country, which face the potential entry of the MNE. This case coincides with the case in which rms are vertically integrated and face a marginal cost equal to u as well as with the case in which the upstream sector is perfectly competitive. Solving each rm s maximization problem under both FDI modes, we obtain the equi- McAfee and Schwartz, 1994 and 1995, Rey and Vergé, 2004, and de Fontenay and Gans, 2005 and 2006). 14 Note that () is increasing in, with (0) = 0 and (0:7807) = 1. 8

10 librium quantities and (gross) pro ts: q A i = a u 2 + ; A i = [q A i ] 2 ; q G i = a u 2(1 + ) ; G i = [q G i ] 2 ; where i = m; 2 under the AQ mode and i = m; 1; 2 under the GI mode. Clearly, GI is a viable option for the MNE only when its net pro ts under GI, GB m G i G; are positive. This holds when G < G B (a u)2 4(1+) 2 : Otherwise, the xed cost is prohibitively high and GI never occurs. Similarly, AQ is a viable option when the acquisition price does not exceed the MNE s gross pro ts under the AQ mode. The acquisition price is determined in the following way: When the MNE sets the acquisition price, it takes into account the outside option (reservation utility) of the target rm, that is, it takes into account that if the domestic rm reects the acquisition o er, it will face entry through GI, as long as G < G B, and thus, that the pro ts of the domestic downstream rm will be equal to the ones under the GI mode. It follows that when G < G B, the MNE optimally sets A B = G i : Given this, it can be easily veri ed that the MNE s net pro ts under the AQ mode, AB m A i A B ; are positive when G < G B. In order to identify the FDI mode that prevails in equilibrium, we take the di erence GB m AB m. Its sign leads to the following Proposition. Proposition 1 In a one-tier industry with G G B, the MNE chooses GI over AQ if and only if G < G B cr; where G B cr() (2 2 )(a u) 2 2( ) 2 ; with 0 < G B cr G B < 0: Proposition 1 informs us that the MNE prefers to set up a new production plant in the host country rather than to acquire an already existing one, when the cost of doing the former is su ciently low. The MNE s preference is driven by the fact that in a one-tier industry the two FDI modes di er only in terms of the number of rms in the market: While the AQ mode does not alter the number of rms -duopoly remains, the GI mode creates a triopoly in the market. Because of this and also because the acquisition price equals the gross pro ts under the GI mode, the choice between the two modes is reminiscent of the one in Salant et al. (1983). We know from Salant et al. that merger incentives are absent when rms operate in a symmetric Cournot triopoly and their merger results into a duopoly. This 9

11 is driven by the so called "business stealing e ect" according to which the outsider to the merger responds with an increase in its output. We should mention that our above nding is in line with Eicher and Kang (2005), Muller (2007) and Qiu and Wang (2011) which have demonstrated that GI is the preferred entry mode in a one-tier industry when its cost is low. Borvatn (2004) has also shown that for a very low xed cost, GI is the chosen FDI mode. At the same time though, he has shown that if the xed cost of GI is high enough then a reduction in it may trigger AQ by reducing the business stealing e ect as well as the reservation price of the target rm. One might wonder what happens when G G B, and thus, GI is not a pro table option. The MNE chooses then among either entering through AQ in the host market or not entering at all. In this situation, when a domestic rm receives an acquisition o er, its outside option equals its pro ts under no entry, i.e., equals its pro ts under duopoly in the market. This, in turn, means that when G G B, the acquisition price will be equal to the MNE s gross pro ts under the AQ mode, and thus, that the MNE will make no pro ts. It follows that when G G B, the MNE is indi erent between undertaking FDI through AQ in the host market and not expanding there at all. 4 FDI Modes in a Vertically Related Industry In the downstream market competition stage, each D i chooses its quantity q i in order to maximize its (gross) pro ts, Di = p i (q i ; Q i )q i w i q i. The rst order conditions give rise to the following reaction functions: R i (Q i ; w i ) = a w i Q i : (1) 2 Obviously, a reduction in the wholesale price faced by D i shifts its reaction function upwards and increases its aggressiveness in the nal market. It is straightforward to derive the equilibrium quantities and the (gross) pro ts in terms of the wholesale prices with AQ: q A i (w i ; w ) = a(2 ) 2w i + w 4 2 ; A D i (w i ; w ) = [q A i (w i ; w )] 2 ; (2) A U(w i ; w ) = (2 )[(a + u)(w i + w ) 2au] 2(w 2 i w i w + w 2 ) 4 2 ; (3) 10

12 as well as with GI: qi G (w i ; w ; w k ) = a(2 ) 2w i + (w + w k w i ) ; (4) 2(2 )(1 + ) G D i (w i ; w ; w k ) = [qi G (w i ; w ; w k )] 2 ; (5) 8 9 < (2 )[(a + u)(w i + w + w k ) 3au]+ = : 2(w i w + w i w G k + w w k ) (2 + )(wi 2 U (w i ; w ; w k ) = + w2 + w2 k ) ; : (6) 2(2 )(1 + ) (i) Green eld Investment Under the GI mode, U bargains with three downstream rms. When U bargains with D i over (w i ; F i ), it takes as given the outcomes of its simultaneously run negotiations with the other downstream rms, (w GT ; F GT ) and (w GT k in order to solve the following generalized Nash product: ; Fk GT ). In particular, w i and F i are chosen max [ G U (w i ; w GT ; wk GT ) + F i + F GT + Fk GT w i ;F i [ G D i (w i ; w GT ; wk GT ) F i ] 1 ; d(w GT ; F GT ; wk GT ; Fk GT )] (7) where d(w GT F GT +F GT k ; F GT ; w GT k ; Fk GT ) = (w GT u)q A (wgt ; wk GT ) + (wk GT u)qk A(wGT k ; w GT ) + is U s disagreement payo, i.e., its pro ts when in the downstream market there are only D and D k and face the equilibrium contract terms (w GT Maximizing (7) in terms of F i and rewriting it, we have: ; F GT ) and (wk GT ; Fk GT ). max w i G U (w i ; w GT ; wk GT )+ [(w GT w i G D i (w i ; w GT u)q A (w GT ; wk GT ) (wk GT ; w GT k ) (8) u)qk A (wgt k ; w GT )]: One can observe from (8) that w i is chosen in order to maximize the oint surplus of U and D i minus U s disagreement pro ts -their "extra" oint surplus. w GT i = (2 )(1 + )u a2 : (9) 2 + (1 2) The resulting equilibrium wholesale price under the GI mode (9) is lower than the upstream marginal cost. The reason that the upstream monopolist subsidizes the downstream production is the so called "commitment problem" (see e.g., McAfee and Schwartz, 1995, Rey 11

13 and Vergé, 2004, and de Fontenay and Gans, 2005). This refers to U s inability to commit to D i that it will not behave opportunistically and make D an aggressive competitor in the nal product market, via a lower wholesale price w. U has incentives to behave opportunistically because it can then use the xed fee F in order to transfer upstream the higher gross pro ts of D. By inspection of (10), we note that the subsidy increases as the nal goods become closer GT =@ < 0, and that the equilibrium wholesale prices are independent of the bargaining power distribution. Substituting, we obtain the equilibrium rms net pro ts under the GI mode: GT D m = (1 )(2 )2 (a u) 2 4(2 + )[2 + (1 2)] G; GT D 1 = GT D 2 = (1 )(2 )2 (a u) 2 4(2 + )[2 + (1 2)] ; (10) GT U = 3(2 )[4 52 2(2 ) 3 ](a u) 2 4(2 + )[2 + (1 2)] 2 : (11) It follows from (10) that GI is a pro table option only when G < G T (1 )(2 )2 (a u) 2 4(2+)[2+(1 2)]. One can easily conclude that G T decreases with and. This is very intuitive: The (gross) pro ts of the MNE decrease when products become closer substitutes because competition gets stronger then. Moreover, they decrease when the MNE s bargaining power gets weaker because the upstream supplier extracts then a larger share of their surplus. (ii) Acquisition Under the AQ mode, U bargains with D i over (w i ; F i ), taking as given the outcome of its simultaneously-run two-part tari negotiations with D, (w AT product now is: ; F AT ). The generalized Nash max [ A U(w i ; w AT ) + F i + F AT w i ;F i d(w AT ; F AT )] [ A D i (w i ; w AT ) F i ] 1 ; (12) where d(w AT ; F AT ) = (w AT u)q m(wat ) + F AT with q m(wat ) = (a w AT )=2, is U s disagreement payo when D acts as a monopolist in the nal good market facing (w AT ; F AT ). Maximizing (12), rst in terms of F i, we obtain: F i = A D i (w i ; w AT ) (1 ) A U(w i ; w AT ) (w AT u)q mon (w AT ) : (13) It follows again that the generalized Nash product reduces to an expression proportional to 12

14 the "extra" oint surplus of U and D i, that w i in turn is chosen to maximize: max w i [ A U(w i ; w AT ) + A D i (w i ; w AT ) (w AT u)q mon (w AT ): (14) For the same reason as the one explained under the GI mode, we nd that the resulting equilibrium wholesale prices with AQ (14) are below the upstream marginal cost. w AT i = (4 2 )u a 2 2(2 2 : (15) ) Proposition 2 In a vertically related industry with G < G T, the equilibrium wholesale price is lower under the GI mode than under the AQ mode, w GT i < w AT i. The MNE pays a lower input price under the GI mode than under the AQ mode. Intuitively, when downstream competition becomes more erce, U s incentives to behave opportunistically get reinforced, and in turn, its commitment problem gets more severe (see e.g., Rey and Tirole, 2006, Rey and Vergé, 2004). Since under the GI mode, downstream competition is stronger, the commitment problem is more severe then, and as a result there is more subsidization of the downstream production than under the AQ mode. When the MNE makes its acquisition o er to D 1, it knows that the target rm will accept the o er if and only if the acquisition price is at least as high as the pro ts that it will obtain under the alternative entry mode, i.e., under the GI mode when G < G T. Taking this into account the MNE optimally sets A T = GT D 1 when G < G T. It is important to note that the acquisition price decreases with and since as explained above the (gross) pro ts of a downstream rm under GI are lower then. The resulting equilibrium net pro ts under the AQ mode when G < G T are: D m = 1 8 (1 )(2 )2 (a u) ( 2 2 ); 4 + [4 (3 + 2)] (16) D 2 = (1 )(2 )2 (a u) 2 8(2 2 ; ) (17) AT AT AT U = (2 )(a u)2 [(2 )(2 2 ) 3 ] 4(2 2 ) 2 : (18) One can easily check from (16) that as long as G < G T, the MNE makes positive pro ts under AQ. Note though that when G > G T, the MNE makes zero pro ts under the AQ mode for the same reason as the one explained in Section 3. 13

15 (ii) Green eld Investment versus Acquisition Having analyzed what happens under both FDI modes, we are ready now to examine which of them will be chosen by the MNE in equilibrium. Proposition 3 In a vertically related industry with G < G T, the MNE chooses GI over AQ if and only if G < G T cr, where G T cr(; ) (1 )(2 )2 [4 (4 + (1 2))](a u) 2 8(2 + )(2 2 ; )[2 + (1 2)] with 0 < G T cr < G < 0; < 0. The MNE prefers to acquire an existing rm rather than to set up a new production unit unless the latter is not too costly. Why is that? When the MNE chooses its FDI mode, it takes into account the following. First, the MNE pays a lower input with GI than with AQ. As a result, it faces a lower marginal cost when it chooses GI. This clearly a ects its choice in favor of GI. Second, with GI, it faces stronger downstream competition than with AQ. This a ects its choice in favor of AQ. However, under the AQ mode, the MNE has to give part of its pro ts to the acquired domestic rm. Since the acquisition price equals a rm s pro ts under the GI mode, i.e., under triopoly, the MNE s net pro ts with AQ equal the di erence between a rm s pro ts under duopoly and triopoly. At the same time, the MNE s net pro ts under the GI mode equal a rm s pro ts under triopoly minus the xed cost of GI. We know from Salant et al. (1983) that the sum of pro ts of two rms in triopoly exceed a rm s pro ts in duopoly. Therefore, as long as the xed cost of GI is su ciently low, GI is more pro table for the MNE. Proposition 3 informs us that as products become closer substitutes, and thus, as downstream competition gets more erce, it becomes more probable to observe AQ instead of GI in equilibrium. This happens, mainly, because, as explained above, the gross pro ts under the GI mode decrease then and AQ becomes cheaper. For the same reason, it also becomes more probable to observe AQ instead of GI when the downstream rms bargaining power decreases. These ndings suggest that we should observe more FDI through AQ than through GI in industries with stronger downstream competition as well as in industries with more powerful upstream suppliers. Combining Propositions 1 and 3, we reach the following conclusion. 14

16 Proposition 4 AQ is more likely to take place relative to GI in a vertically related industry than in a one-tier industry. A MNE faces a lower marginal cost in a vertically related industry than in a onetier industry, due to the subsidization of the downstream production in the former case. Moreover, in a vertically related industry, GI has an advantage relative to AQ which is absent in a one-tier industry: GI results into lower wholesale prices, and thus, into a lower marginal cost. In view of these, we would expect that FDI takes more often the form of GI in a vertically related industry than in a one-tier industry. Proposition 4, though, informs us that the opposite holds. This result can be understood as follows. In a vertically related industry with two-part tari s, downstream rms, even when they have all the bargaining power ( = 0) in their hands, they still pay a positive xed fee. In other words, the downstream rms do not extract all the surplus - they compensate the domestic upstream monopolist for its "outside option". Because of this, and although the MNE is more e cient under the GI mode in a vertically related industry with a concentrated upstream market, its net pro ts can be lower then than its respective pro ts in a one-tier industry or in a vertically related industry with a competitive upstream market. The above result suggests that we should observe more often FDI through AQ when the MNEs enter into host countries in which they source inputs from a local concentrated upstream market than when they produce the inputs in house or when they obtain them from host counties with perfectly competitive upstream markets. The same result also suggests that the recent increase in cross-border acquisitions relative to GI, could be due to the fact that the rms recently undertake FDI in sectors where they depend heavily on powerful input suppliers. 5 Domestic Welfare and FDI Modes In this Section, we examine the implications of FDI, and its di erent modes, from the host country s perspective. We start by noting that the market structure is the same with FDI through AQ and without FDI - autarky (AU). This implies that AT U not, however, imply that AT D 1 = AUT U and AT D 2 = AUT D 2. It does = AUT D 1. In fact, when G < G T, while the pro ts of D 1 with FDI (either through AQ or GI) correspond to a rm s pro ts in a downstream market 15

17 with tripoli, without FDI, they correspond to a rm s pro ts in a duopolistic downstream market. Keeping this in mind, we nd the following. Proposition 5 In a vertically related industry with G < G T : (i) U is better o under GI rather than under AQ/AU if and only if is su ciently low, (ii) D 2 is always better o under AQ/AU rather than under GI, (iii) D 1 is better o under AU than under AQ/GI. When the MNE enters into the host country by creating a new production plant, the local upstream supplier enoys increased input demand. Still, when the downstream market is su ciently competitive ( high), U is better o instead when the MNE enters through acquisition or when it does not enter at all. explained in Section 4, GI has on its commitment problem. This is due to the adverse e ect that, as Stated in di erent words, from U s perspective, GI results, on the one hand, into higher demand, and on the other hand, into lower input prices. When downstream competition intensi es, the commitment problem gets even larger and its impact dominates. The entry of the MNE always hurts D 1. It also hurts D 2 only when it takes place through GI. This suggests that from a downstream rm s viewpoint, the weaker downstream competition under AU/AQ, always o sets the higher, due to the input prices, e ciency under GI. Next, we ask how the two FDI entry patterns di er in terms of their impact on the host country s consumer and total welfare, as well as whether FDI makes the host country better o relative to autarky. Since there is duopoly in the downstream market under both autarky and acquisition, it follows that CS AUT = CS AT. Under GI instead, there are three rms in the downstream market. The higher downstream competition in the latter case along with the increased downstream e ciency clearly result into higher output; hence, into larger consumer surplus. This is formally stated in Proposition 6(i) below. Turning to total welfare, we should clarify that in its measurement, we exclude the pro ts of the MNE. Obviously, this means that our subsequent welfare discussion is di erent from the welfare discussion in a model without MNEs. In turn, prior to the entry of the MNE, the domestic total welfare is W AUT = CS AT + AT U +2AT D 2, while under FDI through GI and AQ, it is respectively W AT = CS AT + AT U + GT D 1 + AT D 2 16 and W GT = CS GT + GT U + 2GT D 2.

18 Proposition 6 In a vertically related industry with G < G T : (i) CS GT > CS AUT = CS AT, (ii) W GT > W AUT > W AT : From the above welfare ranking, it follows that the two entry modes of FDI have di erent implications on the host country s welfare; the host country is strictly better o with GI than with AQ. Further, it follows that while FDI is welfare-enhancing when it is through GI, it is welfare-detrimental when it is, instead, through AQ. The logic for these results is as follows. GI intensi es downstream competition and e ciency and generates vertical linkages in the host country s domestic industry as compared with AQ and autarky. This is good for the consumer surplus. This is also good for the domestic upstream pro ts as long as competition is not too strong. However, this can be bad for the domestic downstream rms which even though they enoy higher e ciency, they loose part of their market shares by the MNE. GI gives higher welfare because the expansion that it causes in the market demand, both downstream and upstream, bene ts more the consumers and the upstream supplier than it hurts the downstream rms. The above results place an important quali cation for policy. More speci cally, they suggest that the host country s government does not need to set any policy when the xed cost of setting up a new plant is very low; the MNE s choice then, GI, is fully aligned with the host country s preferences. When, instead, the xed cost of GI is not su ciently low (G > G T cr), there is room for policy intervention: the government should undertake measures that at the same time encourage GI and discourage AQ. Such measures could be subsidies o ered to the MNEs when the latter set up production plants in the local market. Also, they could be restrictions imposed on the maximum pro t shares that the MNEs could obtain from the acquired domestic rms so that they are driven away from AQ. Such restrictions vary across countries as well as across industries within a country and are the most popular FDI policy among many countries (UNCTAD, 2000). Note that Proposition 6 refers to the case that G < G T, and thus, GI is a viable option. When instead, G > G T, GI never occurs without policy intervention; hence, the welfare comparison is restricted then to that among autarky and AQ with the acquisition price being equal to the pro ts of the domestic downstream rm under autarky. Obviously, the latter means that when G > G T, consumer and total welfare are the same with AQ and with autarky. We should stress though that policy intervention could alter this conclusion. 17

19 In particular, the government could undertake measures that induce GI even when G > G T, as long as the burden of such measures on the country s budget is smaller than the bene ts generated by GI. When FDI occurs in a one-tier industry, GI results again into more competition compared to AQ and autarky. Not surprisingly thus, consumers are again always better o under GI than under AQ/AU. However, GI is not necessary always welfare-enhancing in a one-tier industry. Indeed, in a one-tier industry, although welfare is always higher with GI than with AQ, it is lower with GI than with autarky when the market is highly competitive ( > 0:464). 15 In other words, when market competition is already erce, the loss in domestic pro ts from the further intensi cation of competition with GI is so strong that it o sets its the positive impact on consumer surplus. This happens because in a one-tier industry, as opposed to what happens in a vertically related industry, GI does not lead to more linkages or to increased e ciency (through the lower input prices). Merging these with our respective ndings in the presence of vertical linkages, we can conclude that the host country s government should vary its policy among di erent industry sectors. While it should always use measures that promote the entry of MNEs through GI in vertically related industry sectors, it should avoid doing so in very competitive one-tier industries. Moreover, since a MNE is less likely to choose the GI mode over the AQ mode in a vertically related industry than in one-tier industry, the host country should provide stronger incentives for GI in vertically related industries with concentrated upstream markets rather than in vertically related industries with competitive upstream markets or industries where rms do not depend heavily on input suppliers. 6 Extension: Wholesale Price Contracts In this Section, we extend our analysis and examine what happens when vertical trading takes place through wholesale price contracts. This allows us to check whether our main ndings are robust to the contract type used, as well as to extract some additional insights. Under the GI mode, in the last stage of the game, the quantities of the downstream rms are given again by (4). In the previous stage, U bargains with D i over the wholesale price w i, taking as given the outcome of its simultaneously run negotiations with the other 15 The detailed analysis is available from the authors upon request. 18

20 downstream rms, w GW and wk GW. The generalized Nash product is now: max w i G U (w i ; w GW ; w GW k ) d(w GW ; w GW k ) G Di (w i ; w GW ; wk GW ) 1 ; (19) where d(w GW ; wk GW ) = (w GW u)q d (wgw ; wk GW ) + (wk GW u)q d k (wgw k ; w GW ) is U s disagreement payo when D and D k act as duopolists with input prices w GW resulting equilibrium wholesale prices and net pro ts are: and wk GW. The wi GW = 1 (a u) + u; (20) 2 GW D m = (2 )2 (a u) 2 16(1 + ) 2 G; GW D 1 = GW D 2 = (2 )2 (a u) 2 16(1 + ) 2 (21) GW U = One can easily observe that GW D m G < G W (2 )2 (a u) 2. 16(1+) 2 3(2 )(a u)2 : (22) 8(1 + ) > 0 only when the xed cost of GI is not too high; when Under the AQ mode, the quantities that the downstream rms set in the last stage of the game are given by (2). In stage two, U and D i maximize the generalized Nash product: max w i A U (w i ; w AW ) d(w AW ) A Di (w i ; w AW ) 1 ; (23) where w AW is the equilibrium wholesale price charged to D and d(w AW ) = (w AW u)q m(waw ) is U s disagreement payo, in case that D acts as a downstream monopolist with input price w AW. The equilibrium wholesale prices are: wi AW = 1 (a u) + u. (24) 2 In contrast to what happens under trading through two-part tari contracts, note from (21) and (25), that the equilibrium wholesale prices, rst, exceed the upstream marginal cost - double marginalization is present - and second, that they are the same under the two FDI modes. The latter means that under trading through wholesale price contracts, the equilibrium input prices are independent of the number of downstream rms. This nding is not new and can also be found in Dhillon and Petrakis (2002). We move next to the determination of the equilibrium acquisition price. When G < G W, the MNE will optimally set A W = GW D 1. The resulting equilibrium net pro ts when 19

21 G < G W are: AW D m = 1 16 (a u)2 (2 ) [ (2 + ) 2 ]; (25) (1 + ) 2 AW D 2 = (2 )2 (a u) 2 4(2 + ) 2 ; AW U = (2 )(a u)2 : (26) 2(2 + ) From (26), it follows that, when G < G W ; the pro ts of the MNE under the AQ mode are always positive. What happens though when G > G W? For similar reasons, as the ones explained in Section 3, the MNE will be indi erent then between entering in the host country through AQ and not entering at all. Comparing the net pro ts of the MNE under the two FDI modes ((22) and (26)), we nd that similarly to what happens with two-part tari contracts, the MNE chooses to expand through the GI mode only when the xed cost of doing so is su ciently low. More precisely, this holds when G < G W cr, where: G W cr (; ) (2 )2 (2 2 )(a u) 2 8( ) 2 ; (27) with 0 < G W cr < G W. One can con rm < 0 < 0, i.e., under wholesale price contracts too, it becomes more probable to observe AQ instead of GI when products are closer substitutes, as well as when the bargaining power of the downstream rms decreases. Since, as mentioned above, when rms trade through wholesale price contracts, the input prices are the same under the two FDI modes, the two modes di er now only in terms of their intensity of downstream competition. Downstream competition is clearly stronger under the GI mode. Since this, in turn, means that there are more backward linkages under the GI mode and higher input demand, the domestic upstream rm is always better o when the MNE enters through GI rather than through AQ. At the same time, the opposite holds for the domestic downstream rm which under the GI mode, it su ers from a decrease in its market share. We con rm that under wholesale price contracts too, AQ is more likely to take place relative to GI in a vertically related industry than in a one-tier industry. The intuition behind this di ers from the respective one under two-part tari contracts. In a vertically related industry with wholesale price contracts, double marginalization results into a wholesale price which exceeds the upstream marginal cost. This means that the MNE faces a 20