1 2

Georgia Institute of Technology. [email protected] Georgia Institute of Technology. [email protected]

1 Introduction While the increased importance of foreign direct investment (FDI) in the world economy generally has been well recognized,3it is a somewhat less appreciated fact that a large proportion of FDI occurs via international mergers and acquisitions (M&A) that frequently involve large multinational enterprises (MNEs). In recent years, the bulk of cross-border M&A transactions have been in services such as finance, transportation and telecommunications. In fact, the service sector has become a major driving force behind global flows of foreign direct investment (FDI): In 2013, services continued to account for the largest shares of both announced greenfield projects and M&A deals (UNCTAD 2014). Particularly, a growing volume of service-related FDI is concentrated in e-commerce sector. For instance, the online retailer giant, Amazon has been doing business global wise since 1998, either by acquiring local major platforms or by establishing their own localized domains. China as one of the key emerging markets has attracted FDI in e-commerce from a bunch of major companies including Amazon, Ebay and Expedia. Several factors explain this rise in FDI in services. Market driven changes have been complemented by policy initiatives: until fairly recently, many countries prohibited FDI in important services such as banking, telecommunications and transportation. At present, while many obstacles to FDI in services have been removed, several important ones remain. Policy restrictions that limit the degree of foreign ownership in services such as telecommunications and finance still persist in many countries. One of the most frequently observed policy restrictions in telecommunications services and e-commerce are those on the extent of foreign ownership allowed. The pattern of the restrictions varies across countries with different degrees of market competition. Take the e-commerce sector as an example. Starting from June 19 of 2015, China has allowed e-commerce online data processing and transaction processing businesses to be opened to 100 percent foreign ownership across the country.4 Noticeably, its current e-commerce business is dominated by the local giant, Alibaba. India and Indonesia, on the other hand, with much less

3

The importance of FDI is evident in the fact that sales of subsidiaries of multinational firms now exceed worldwide exports of goods and services. In 2013, the total sales of foreign affiliate in the world was 4.7 trillion dollars, whereas the value of global exports was 4.1 trillion dollars (UNCTAD World Investment Report 2014, Table 2). 4 http://www.usito.org/news/china‐lifts‐restrictions‐e‐commerce‐foreign‐investment

concentrated markets 56 , reiterated its limits on FDI in retail e-commerce in 2014 and 2015, respectively.78 An important characteristic of many services markets is the presence of network effects. In a growing number of industries, buyers and sellers interact and transact through (on-line or offline) intermediating platforms with network effects generated both within and across the sides of such a two-sided market (i.e., the buyers’ side and the sellers’ side). For example, the benefit to a buyer of being a customer of an on-line retail platform depends upon how many other buyers and sellers belong to the network of the same on-line platform. Given the frequent prevalence of network effects in services sectors and the importance of M&As as a vehicle for FDI, the effects of the entry of multinationals in e-commerce markets (as well as of the policy restrictions faced by them) on buyers, sellers and on-line platforms as well as the aggregate welfare of the host country deserve investigation. This paper thus has two objectives. First, it develops a theoretical model that examines an MNE platform’s choice between de novo entry and the acquisition of the rival platform in the host country when the market is two-sided and the degree of technology transfer by the MNE is endogenously determined. Second, it explores the welfare impact of such entry and asks whether the potential clash between the market equilibrium and the host country welfare can shed light on some policy restrictions that confront foreign investors in e-commerce. In particular, we examine restrictions on the degree of foreign ownership permitted by the host government. Under de novo entry, the foreign firm obtains a license to operate a new e-commerce platform (a wholly owned subsidiary in the host country market) that competes with the local incumbent. Under acquisition, the MNE takes over the existing license and operates as a monopolist. The technological know-how controlled by the MNE allows it to choose the degree of technology transfer under either entry mode. Thus, the MNE’s entry affects the degree of competition in the host country market and the extent to which the MNE’s affiliate in the host country benefits from the transfer of the more advanced technology. The competition effect of foreign entry is present only under de novo entry. It reduces the incentive to transfer technology under de novo entry as this will induce the incumbent to compete 5

http://www.pwc.in/assets/pdfs/publications/2015/ecommerce‐in‐india‐accelerating‐growth.pdf http://www.researchandmarkets.com/research/rr5fgm/malaysia_b2c 7 http://www.wsj.com/articles/india‐wont‐allow‐foreign‐investment‐in‐supermarkets‐online‐retail‐1410182468 8 http://www.loc.gov/law/foreign‐news/article/indonesia‐limits‐on‐foreign‐investment‐in‐e‐commerce‐proposed/ 6

more aggressively, which in turn hurts the MNE’s profits (strategic effect). Moreover, the relatively larger number of users that the MNE attracts under acquisition increases its incentive for transferring costly technology (scale effect). Both effects point to the result that the foreign entrant will transfer more cost-reducing technology when entering through acquisition. In effect, the MNE’s preferences over the two entry modes and the welfare comparison of the two modes also depend on the extent of competition effect. We show that when the two platforms are close substitutes, or when cross-side network externalities are high, the intense competition under de novo entry will induce the foreign entrant to prefer entry through acquisition while the welfare-maximizing host country government favors the opposite. The divergence between the foreign firm’s preferences the welfare interest of the host country creates a basis for policy intervention. We show that restrictions on the foreign ownership, when applied asymmetrically to the two modes of entry, can accommodate the differences in preferences. The model considered in this paper sheds light on the experience of many developing countries that seek to obtain access to more advanced technological know-how and managerial practices in industries which have two-sided market characteristics. In this context, many developing country governments face a critical policy choice between promoting actual (de novo) entry of foreign firms that have critical expertise in operating two-sided markets platforms, and entry of these firms through acquisition of the incumbent domestic platforms. Since our analysis considers the effects of the foreign firm’s entry mode and equity restrictions on FDI in industries with two-sided network externalities, it can be useful for guiding policy decisions faced by the regulators overseeing internet services and e-commerce. Some of the issues addressed here have been studied separately before, but we know of no analytical study of the relationship between technology transfer and mode of FDI (as in de novo entry versus acquisition) in the setting with two-sided markets.9 The papers closest to ours are Matto et. al. (2003) and Klimenko and Saggi (2007). While the former paper studies the preferences of a foreign firm and a welfare-maximizing host country government over de novo entry and acquisition in a Cournot oligopoly in the absence of network effects, the later paper 9

The literature has tended to focus on licensing and de novo entry where the foreign firm seeks to prevent the dissipation of its technological advantage (see, for example, Ethier and Markusen, 1996, and Markusen, 2001).

considers the clash between the entry mode preferences of a foreign firm and the host government in the duopoly setting with single-side network externalities and price-competing firms. Other related papers on international mergers and joint ventures include Svejnar and Smith (1984), AlSaadon and Das (1996), Roy et al. (1999), Horn and Persson (2001) and Norbäck and Persson (2006). We add value to this line of research by examining the role two-sided network externalities and endogenous technology transfer play in determining the effects of foreign entry on domestic welfare.10 For example, unlike us, Svejnar and Smith (1984) focus on the interaction of transfer pricing and local policy. Similarly, Al-Saadon and Das (1996) model an international joint venture in which ownership shares are determined via bargaining while Horn and Persson (2001) provide a novel model of endogenous mergers in which firms can merge both nationally as well as internationally. On the other hand, the literature on two-sided markets is burgeoning in industrial organization. Rochet and Tirole (2003, 2006), Armstrong (2006) and Rysman (2009) are among the seminal papers in this area. Our paper builds on the framework by Armstrong (2006) where users are heterogeneous in membership benefits (costs) and platforms charge membership fees on the two sides. A key insight of this line of research is that one side of the market will be targeted aggressively by the platforms if it generates large cross-side network externalities, i.e., large positive externalities on members of the other side. As mentioned earlier, in our model, the extent of cross-side network externality plays an important role in shaping the FDI entry mode decisions by the MNE and the host country government. A recent working paper, Jeon, Jullien and Klimenko (2014) also studies the impacts of a foreign two-sided platform on host country welfare, but with a focus on how bilingualism together with cross-side network externalities affects platform competition. The remainder of the paper is organized as follows. Section 2 describes the model setup and the timing of the game. Section 3 derives the Nash equilibrium demand, prices and amounts of technology transfer under both de novo entry and entry through acquisition. Then in Section 4, we compare the preferences of entry mode by the foreign platform and the domestic government. 10

Although this is the first paper to consider FDI policies in an environment with network externalities, several existing papers study the effects of trade policies and compatibility standards in open economies with the single‐ side network externalities. See, for example, Krishna (1988), Gandal and Shy (2001) and Barrett and Yang (2001).

In cases of a clash between the two preferences, we examine restrictions on foreign ownership as an example of policy intervention in Section 5. Section 6 concludes.

2 Model setup Consider a country with a mass one of buyers and a continuum of sellers, interacting through a domestic platform. Each transaction between a buyer and a seller generates surplus 0 to the buyer and

0 to the seller. A foreign platform is considering entering the market. The

two platforms are horizontally differentiated in terms of membership benefits to buyers, but are symmetric in all other features (except that the foreign platform may incur lower marginal costs by implementing cost-reducing technology, as described in later paragraphs). In what follows, let platform 1, 2 represent the foreign and domestic platform, respectively.A buyer subscribes to only one of the platforms (i.e., single-homes) due to time constraints and habit formation. The whole mass of buyers are uniformly distributed on the Hotelling interval between zero and one, with platform 1 and 2 located at zero and one, respectively. A seller will join both platforms (i.e., multihome) as long as each generates a positive profit. Assume that there is a mass of fixed cost of using a platform is below , where membership fees,

and

sellers whose

is a positive constant density. Platform charges

on each buyer and each seller, respectively. In what follows, let ,

represent the buyers’ and sellers’ side. The game proceeds as follows. In the first stage, the platform 1 chooses its mode of entry, i.e., entry through acquisition ( ) or de novo entry ( ). Under mode , it makes a take-it-or-leaveit offer to the incumbent, specifying both a fixed acquisition price ( ) and a share (

) of the new

platform’s profits. By accepting the offer, the incumbent will get the acquisition price and 1 of the profits, with the rest accruing to the entrant. The entrant will then operate as a monopoly to serve buyers and sellers.11 Otherwise, platform 1 will enter the country by establishing its own subsidiary and competing directly with the incumbent. It will choose to hold a share (

) of the

total profits. In section 5, we show that the foreign platform never chooses partial equity share unless the domestic government constraints the degree of foreign ownership. Therefore, for

11

In our setup, the domestic platform stops providing intermediary services after the acquisition. This is because to the best of our knowledge, we are not aware of a real counter example in the e‐commerce sector.

1 and compare full

sections 3 and 4 where there are no such restrictions, we must have acquisition with de novo entry.

In the second stage, platform 1 decides the amount of cost-reducing technology transfer under each entry mode. Without any transfer, the foreign entrant incurs the same constant marginal costs

and

as the domestic incumbent to serve each buyer and seller. By transferring

and

units of technology to the two sides, respectively, platform 1 can reduce its marginal costs to and

. We assume that such transfer is costly and the costs are

and

, where

,

0.

In the third stage, platform determines its membership charges

on buyers and

on

sellers. Then in the fourth stage, each buyer forms the habit of using one of the two platforms. Simultaneously, each seller decides which platform(s) to join. By subscribing to platform that has a mass of surplus

buyers and

sellers, each buyer gets surplus

, before the deduction of fixed charges and costs. Here,

, and each seller obtains is the stand-alone value of a

platform to each buyer. 3 Subscription, prices and the amount of technology transfer In what follows, we solve the game under each entry mode of FDI using backward induction. 3.1 De novo entry In the last stage, users subscribe, forming the demand for each platform. The location of the marginal buyer who is indifferent between platforms 1 and 2 is given by: 1

, (1)

where is the “transportation cost” in the Hotelling model. By assumption, the marginal seller who is indifferent between joining platform or not incurs a fixed cost of

, which is determined by:

0, 1

where

1,2, (2)

.

Equations (1) and (2) together lead to the following demand functions: ,

,

,

, (3)

,

,

,

, ,

1, 2. (4)

As we can see from equations (1) and (2), a decrease in charge on sellers increase in sellers’ demand

and thus a larger cross-side network externality from sellers to

buyers. Consequently, buyers’ demand sellers, and

will lead to a direct

will increase, exerting larger positive externalities to

will increase even further, so on and so forth. Similar analysis goes for changes in

other prices. This implies that the law of demand holds on each side of the market, i.e., 0,

0, which requires the following assumption on the parameters: . (A1) Bearing those demand functions in mind, in stage 3, platforms 1 and 2 determine the

membership charges to maximize their own profits: , , where

1

.

The first-order conditions are: 0, (5) 0, (6)

0. (7) 0. (8)

In equation (5),

represents the marginal benefit of raising

.

is the corresponding marginal cost. As we learn from the last stage, a higher price on sellers,

not only reduces the demand on side S, but also exerts lower cross-side

network externality, resulting in lower demand on side B. Therefore, the platform needs to subsidize sellers’ contribution to its profits on the other side in price determination. Similar interpretations go for equations (6)-(8). Solving equations (5)-(8) yields: ,

,

,

In the second stage, given transfer,

,

. See appendix for the details. ,

,

,

, platform 1 chooses the level of technology

to maximize its profit. By envelope theorem, the first-order conditions are: 0, (9) 0. (10)

where

,

,

,

. In order to interpret the above equations, we first show that:

Lemma 1

0.

Proof. See appendix. Lemma 2

0,

0,

0,

0.

Proof. See appendix. 0 and

Lemma 1 and 2 together implies that,

0 in equation (9) capture the

strategic effects of technology transfer on the buyers’ side: an increase in

induces platform 2

to act more aggressively by cutting prices on both sides, thus reducing the profit earned by platform 1. This strategic consideration reduces the foreign entrant’s incentive to transfer technology under a duopoly setting. On the other hand, unit of increase in

represents the scale effect of technology transfer: one

corresponds to one unit of increase in the price margin on buyers, and so

units of increase in profits from buyers. This implies that the more buyers on platform 1, the stronger will be its incentive to transfer cost-reducing technology. The remaining term,

is

simply the marginal cost of technology transfer. The same analysis applies to technology transfer on the sellers’ side in equation (10). The results are similar to Matto and Saggi (2004)’s finding for one-sided markets. 3.2 Entry through acquisition After acquisition, platform 1 becomes a monopolist in connecting buyers and sellers. We 0, and

assume that the buyers’ market is still fully covered, i.e.,

1.

Then the mass of sellers on platform 1 satisfies: 0. An immediate observation is that the demand on the two sides do not affect each other in this case. In the third stage, the merged entity determines

,

to maximize its profit: .

Under full market cover on side B, platform 1 maximizes its profit by setting: . In the meantime, the first-order condition for

yields:

. In the second stage, by envelope theorem, the levels of technology transfer chosen by platform 1 satisfy the following first-order conditions: 1

0, (11) 0. (12)

Since platform 1 has no competitor after acquisition, there are no negative strategic effects of technology transfer. 1 and

are respectively the scale effects for the buyers’ and sellers’ side.

Obviously, the scale effects are larger under acquisition than under de novo entry. Taking both effects into account, the following comparison result is immediate: Proposition 1 The foreign entrant will transfer more cost-reducing technology to both sides of platform 1 under acquisition than under de novo entry. 4 Preferences of entry mode As we mentioned in section 2, in the absence of any policy restrictions, it is sufficient to focus on the comparison between full acquisition and de novo entry. 4.1 Foreign platform’s choice of entry mode To determine the foreign platform’s choice of entry mode, we first need to pin down its equilibrium acquisition offer . The domestic incumbent will be willing to accept any offer that leaves it with a payoff no less than the amount it can earn by refusing the offer and competing directly with the foreign entrant. Assuming that the entrant has the full bargaining power, we have: .

The foreign platform will opt for direct entry if and only if it is more profitable: ∆ 0 . We will analyze the sign of ∆

≡

in section 4.3, together with the welfare

comparison. 4.2 Host country government’s preference of entry mode The host country’s welfare under de novo entry is given by:

, where the first square bracket represents buyers’ surplus and the second one represents sellers’ surplus. The last term is the domestic platform’s profit. Similarly, the country’s welfare under acquisition is given by: , where the last term is the acquisition price accepted by the domestic platform, according to our analysis in the previous subsection. Define ∆

≡

. Since full foreign ownership, platform 2 obtains the same

amount of payoffs under the two entry modes, the comparison of domestic welfare will be equivalent to comparing the sum of buyers’ and sellers’ surplus. The host country government will prefer direct entry if and only if it generates higher welfare: ∆

0.

4.3 Results of the profit and welfare comparisons As might be expected, the equilibrium expression for both ∆

and ∆

are quite

cumbersome. However, we can plot it and analyze graphically by assigning values to some of the parameters. 12 Consider

1,

5,

1. Therefore, the signs of ∆

and ∆

12

The parameter values must satisfy the following conditions for both entry modes to be considered feasible in our analysis: 1) profits for both platforms are non‐negative; 2) the amount of users on both platforms are non‐negative; and 3) each buyer obtains non‐negative utility from joining a platform.

depend on three parameters: “transportation cost” , and per-transaction cross-group network externalities

and . [Insert Figure 1 here.]

The three graphs in Figure 1 (see appendix) demonstrate how the signs of ∆

and ∆

change as varies. The shaded areas are feasible regions in our analysis, as defined in footnote 10. As increases, the foreign entrant becomes more profitable by entering directly than by entering through acquisition. The opposite holds for welfare comparison. When is low, the two platforms are close substitutes to buyers. On one hand, the intense competition under de novo entry will enforce low charges on both sides, thus benefiting the end-users while leading to low platform profits and a low acquisition price. On the other hand, the monopolist after acquisition will enjoy high market power as buyers pay low “transportation costs” to join the platform. Both point to the result that the foreign entrant will obtain higher profit through acquisition while the host country government will favor de novo entry. As increases, platform competition becomes less intense and so the preferences will be reversed when gets high enough. The graph in the middle shows that when is intermediate, ∆ positive or negative. ∆

tends to be positive if given ,

relatively large. The reverse holds for ∆ . Given , when

and ∆

can be either

is relatively large; or given ,

is

is higher, each buyer generates higher

positive externality to the sellers. This makes the platform more attractive to the sellers and so can extract higher profits from them. Therefore, platforms will compete more aggressively under the duopoly setting, resulting in lower profits and higher consumer surplus. The analysis for fixing is similar. 4.4 Comparison of the mode preferences The above analysis shows that the signs of ∆

and ∆

can be different in many cases.

This divergence in the preferences over entry modes implies room for government intervention. In the next section, we consider restrictions on the degree of foreign ownership as an example of such policies. 5 Restrictions on the degree of foreign ownership

Let ̅ represent the maximum degree of foreign ownership permitted by the domestic government, which can be implemented in one of two ways. One is to apply it symmetrically wherein a newly established foreign platform and an acquired entity are subject to the same cap. The other is to asymmetrically restrict the foreign equity share under acquisition, leaving the newly established platform to be fully owned by the entrant. 5.1 Technology transfer under equity restrictions Before analyzing the effect of equity restrictions on payoffs, we first look at the effect on as the equity

the level of technology transfer. Suppose that the foreign entrant will choose

share to hold when it enters by launching a new subsidiary (through acquisition), where

is

subject to the equity restrictions imposed by the host government (either symmetric or asymmetric). Given that information, platform 1 will select

,

(

,

) to maximize its profit under de

novo entry (acquisition): ≡

∙

, ≡

∙

. The first-order conditions for the sellers’ side now become13:

0, 0. Therefore, marginal benefits of the transfer (the sum of strategic effects and scale effects) are proportional to the degree of ownership under either entry mode. This imply that

13

The results for the buyers’ side are symmetric.

0,

0,

, .

Lemma 3 Equity restrictions will lower the levels of technology transfer on both sides under both entry modes. 5.2 Equilibrium ownerships under equity restrictions Under de novo entry, differentiating ,

theorem with regard to ∙

with respect to

and applying the envelope

, we have:

0. (13)

Therefore, the foreign entrant will choose the maximum degree of ownership permitted by the host government under de novo entry. When it enters through acquisition, we need to pin down both the takeover price equity share

and the

. As discussed in section 4.1, the domestic platform will be willing to accept any

acquisition offer that leaves it with as least as much payoff as it can earn by refusing the offer and competing directly with the entrant. Assuming that the foreign entrant has all the bargaining power, it will set 1

such that: ∙

.

Therefore, the entrant chooses ∙

to maximize: . (14)

,

Since

were selected to maximize

objective function (14) with respect to ∙

1

where

∙

0and

∙

0,

∙

, then differentiating the

by applying envelope theorem, we have: 0, (15)

, .

Combining this result with the one under de novo entry, we conclude:

Lemma 4 No matter which entry mode it selects, the foreign entrant will hold the maximum degree of ownership permitted by the host government. 5.2 Asymmetric equity restrictions Consider an asymmetric cap

which only restricts the maximum degree of foreign 1,

ownership in an acquired platform. By lemma 4, we know

. How does

affect

the foreign entrant’s and the host government’s preference of entry mode, respectively? Will this induce the entrant’s choice to be in line with the government’s? First, we look at the effect on the choice of the entrant. ∆

0,

according to inequality (15). Thus, an asymmetric equity restriction makes direct entry more attractive to the foreign platform . Therefore, a sufficiently stringent equity restriction (i.e.,

small enough) can induce direct entry

by the foreign platform. Now, we consider the effect on the domestic welfare. Recall that , then ∆

0.

Therefore, an asymmetric equity restriction will make de novo entry more preferable to the host government. Hence, Proposition 2 A stringent asymmetric restriction on foreign degree of ownership can induce the foreign platform to enter directly, which is preferred by the host government. 5.3 Symmetric equity restrictions

Consider ̅ as the equity restriction that applies symmetrically to the two entry modes. By lemma 4, we know

.̅ Then with similar arguments for Proposition 1, we can show

that: Lemma 5 Under symmetric equity restrictions, the foreign entrant will transfer more cost-reducing technology to both sides of platform 1 under acquisition than under de novo entry. Now let’s analyze the equilibrium profits for the foreign entrant. Under de novo entry, the , which means the entrant suffers a loss of 1

payoff will be ̅ ∙

it enters through acquisition, however, the payoff will be ∙ can offset the giveaway of 1

̅ ∙

̅ ∙ . If , where it

by manipulating its takeover offer . As might be expected,

this will make the foreign entrant more likely to prefer acquisition to direct entry. 6 Conclusion This paper has explored an MNE’s choice between de novo entry and entry through acquisition in two-sided markets and the impacts on the host country welfare when the degree of technology transfer is endogenously determined. We confirm that in two-sided markets, the competition effect under de novo entry still plays a key role in shaping the decision of how much technology to transfer and the preferences over the two entry modes. More specifically, it reduces the MNE’s incentive to transfer technology under de novo entry due to strategic considerations. When the two platforms are close substitutes, or when cross-side network externalities are high, the intense competition under de novo entry will induce the foreign entrant to prefer entry through acquisition while the welfare-maximizing host country government favors the opposite. This divergence in the preferences of entry mode between the foreign firm and the home government leaves room for policy intervention. We find that restrictions on foreign ownership imposed asymmetrically on the two modes will help to accommodate the difference. The analysis here can be useful for guiding policy decisions faced by regulators of many developing countries that seek to obtain access to more advanced technological know-how and managerial practices in industries with two-sided market characteristics such as internet services and e-commerce.

References

Al-Saadon,

Y.

and

S.P.

Das,

1996,

“Host-country

policy,

transfer

pricing

and

ownership distribution in international joint ventures: a theoretical analysis”, International Journal of Industrial Organization 14, 345–64. Armstrong, M., 2006, “Competition in Two-Sided Markets,” Rand Journal of Economics, 37(3), 669–691. Horn, H., and Persson, L., 2001, “The equilibrium ownership of an international oligopoly”, Journal of International Economics 53, 307–33. Jeon, D.-S., B. Jullien and M. Klimenko, 2014, “Language, Internet and Platform Competition”, working paper. An older version is available at: https://ideas.repec.org/p/cpr/ceprdp/9144.html. Klimenko, M. and K. Saggi, 2007, “Technical Compatibility and the Mode of Foreign Entry with Network Externalities”, Canadian Journal of Economics, 40 (1), 176-206. Matto, A., M. Olarreaga and K. Saggi, 2004, “Mode of Foreign Entry, Technology Transfer, and FDI Policy”, Journal of Development Economics, 75, 95-111. Markusen, James R., 1995, “The boundaries of multinational enterprises and the theory of international trade”, Journal of Economic Perspectives, 9, 169–89. Rochet, J.-C. and J. Tirole, 2003, “Platform Competition in Two-Sided Markets,” Journal of the European Economic Association, 1(4), 990–1029. Rochet, J.-C. and J. Tirole, 2006, “Two-sided markets: a progress report”, RAND Journal of Economics, 37(3), 645-667. Roy, P., et al, 1999, “Technology transfer, merger, and joint venture: a comparative welfare analysis”, Journal of Economic Integration 14, 442–66. Rysman, M., 2009, “The Economics of Two-Sided Markets”, Journal of Economic Perspectives, 23(3), 125–43. Svejnar, Jan, and Stephen Smith (1984) ‘The economics of joint ventures in less developed countries,’ Quarterly Journal of Economics 99, 149–68

UNCTAD, 2014, World Investment Report- Investing in the SDGs: an action plan. Wang, X. 2012. “Foreign direct investment and innovation in China’s e-commerce sector”. Journal of Asian Economics, 23, 288-301.

Appendix A A.1 Proof of Lemma 1 ,

,

,

, From the demand functions (3)-(4) and assumption (A1), we have: 0,

0,

0,

0.

Meanwhile, since platforms earn positive profits in equilibrium, then 0, (A.1.1) 0. (A.1.2) This implies each platform gains positive profits on at least one side of the market. We take platform 1 as an example in the following analysis. The situation on platform 2 is similar. Case 1: Suppose

0.

Then equation (A.1.1) is equivalent to

On one hand,

.

by equation (2).

On the other hand, by equations (2)-(4),

.

Therefore,

,

0.

which is equivalent to 0.

Case 2: Suppose

On one hand, by equations (2)-(4),

.

On the other hand, equation (5) implies

Since

.

by assumption (A1), then

Thus,

.

,

0.

which is equivalent to

Similarly,

0,

0 and

0.

A.2 Proof of Lemma 2 2

12 3

/12.

Therefore,

/4,

9

3

6

2

2

4

,

,

,

.

To determine the sign of the derivative, we first show the following claim: Claim 1 Under de novo entry, ,

strategic complement to

is strategic complement to when 4

3

,

when

(A3), where ,

1, 2 and

(A2);

is

.

Proof. The first-order conditions (5) and (6) can be re-written as: ,

,

,

,

and

0,

,

,

where to

,

0,

is best response to

,

. Differentiating the above expressions with respect

, we get: ,

,

,

0, (C.1.1)

,

,

,

0, (C.1.2)

,

,

,

0, (C.1.3)

,

,

,

0. (C.1.4)

(C.1.1) and (C.1.2) can be transformed into:

, (C.1.5)

Similarly, (C.1.3) and (C.1.4) can be transformed into:

, (C.1.6)

Define ∆ ≡

,

∆ ≡

∆ .

Using Cramer’s rule, (C.1.5) and (C.1.6) give:

/ ∆

,

∆

/ ∆

/ ∆

/ ∆

A sufficient condition for 0, or 8

Then ∆

6 ∆

,

∆

∆

∆

,

∆

0.

,

. defined by equations (5)-(6) to be equilibrium maximizers is: 0 (A4)

0,

Thus, 0,

0 if and only if 0 if and only if

3

0 (A2) 4

0 (A3)

Note that (A2) and (A3) imply (A1) and (A4). By imposing the assumptions (A2) and (A3), we can immediately get the signs of the derivatives as shown in Lemma 2.

Appendix B Figure 1