Modes of Delivery in Services

by

Elisabeth CHRISTEN*)

Joseph FRANCOIS

Working Paper No. 1008 June 2010

DDEEPPAARRTTMMEENNTT OOFF EECCOONNOOMMIICCSSJJOOHHAANNNNEESS KKEEPPLLEERR UUNNIIVVEERRSSIITTYY OOFF

LLIINNZZ

Johannes Kepler University of LinzDepartment of Economics

Altenberger Strasse 69 A-4040 Linz - Auhof, Austria

www.econ.jku.at

*)elisabeth.christen@jku.at phone +43 (0)70 2468 -8229, -8238 (fax)

Modes of Delivery in Services ∗

Elisabeth ChristenJohannes Kepler Universitat Linz

Joseph FrancoisJohannes Kepler Universitat Linz,

Tinbergen Institute, and CEPR

June 22, 2010

Abstract

We develop a new analytical framework for both cross-border servicestrade and services trade through foreign affiliates, based on heteroge-neous firms operating under oligopoly. This leads to direct predictionsabout choice of services delivery (mode of delivery) at the firm level, andabout the pattern of bilateral trade at the industry level. We examinethe industry-level predictions, working with a panel of U.S. data. Unlikethe recent literature that works with FDI as a proxy for affiliate servicessales, we work directly with data on bilateral U.S. trade through affiliates.These data feature more sector detail than in the recent literature. Wealso directly compare observed patterns of services trade and affiliate saleswith the corresponding indicators of patterns of cross-border and affiliatesales for manufacturing sectors. In contrast to mixed results in manufac-turing, in services overseas multinational activities consistently increaserelative to direct exports the further away are host countries. Languageand the presence of manufacturing FDI are also important. The impactof factors like corporate tax rates and relative stocks of human capital onmodes of service delivery varies across sectors. The evidence on interde-pendence across modes and the importance of local affiliates implies thatthe impact of policy in any one mode is likely to depend on the mix ofdomestic regulation and policy across all modes of supply.

Keywords: International Trade in Services, Modes of Supply, FDI, For-eign Affiliates Trade, GATSJEL codes: F10, F14, F23, L80

∗We would like to thank Eddy Bekkers, Bernard Hoekman, Marc Melitz, Gianmarco Otta-viano, and the participants of the Midwest Trade Meeting, ETSG and the first meeting of theEU-funded GIST network for very helpful comments. Address for correspondence: JosephFrancois, Johannes Kepler University of Linz, Department of Economics, Altenbergerstr. 69,4040 Linz, Austria. email: joseph.francois@jku.at

1 Introduction

Because services are a flow and not a stock, direct proximity and interactionbetween user and supplier are more important for trade in services than fortrade in goods. Historically, this has hampered growth in international servicestrade relative to commodities trade. However, because of technical change, theproximity burden has progressively weakened in recent decades for some (butnot all) service activities. This has led both to dramatic growth in servicestrade and foreign investment, and to a nascent empirical and theoretical liter-ature on trade in services (Francois and Hoekman, 2010). In this paper, wedevelop an analytical framework for both cross-border services trade and ser-vices trade through foreign affiliates, based on heterogeneous firms operatingunder oligopoly. This leads to direct predictions about choice of services deliv-ery at the firm level, and about the pattern of bilateral trade at the industrylevel. We then examine the industry-level predictions with a panel of U.S. dataon affiliate sales and cross border sales. Our data on affiliate sales allows moresector detail than found in the recent literature, which relies instead on FDI asa proxy for affiliate sales. In addition, and again in contrast to the recent liter-ature, we also directly compare observed patterns of services trade and affiliatesales with the corresponding indicators of patterns of cross-border and affiliatesales for manufacturing sectors.

Proximity and jointness in production has important implications for thenormative and positive aspects of trade and foreign investment in services.Bhagwati (1984) emphasized the implications of a decline in the cost of dis-tance, highlighting mechanisms through which services are ”disembodied” or”splintered” from goods or people as ”carriers.” Trade in services may thenexpand as a result of the incentive to ”splinter” the production chain geograph-ically, not just in terms of tangible inputs but also services. The subsequentliterature has called this process fragmentation. In both goods and servicesectors, fragmentation can lead to basic changes in the structure and patternof trade, as low-wage activities can be sliced away and outsourced (Francois,1990; Baldwin and Robert-Nicoud, 2007). However, the significance of under-lying proximity constraints for service transactions to be feasible means that”trade” may require a heavier dose of local presence of suppliers in the mix ofcross-border and local supplied services than is the case with goods .1 The local

1Horn and Shy (1996) argue that once account is taken of the fact that many servicesare also bundled with goods, and that the associated services-input bundle is non-tradablein the sense it must be provided locally, in direct proximity to the consumer/buyer of thegoods, the impact of liberalization of trade in goods can be limited because of differences in

2

presence component of services trade may be foreign or domestic. In general,services provision will often have an element of jointness in production, in thesense that complementary inputs - including other services - are needed to al-low effective exchange (trade) of a service to occur. This is recognized in thepolicy community, where the cross-border and local presence (or commercialestablishment) components of international service transactions are referred toas modes of supply (Francois and Hoekman, 2010).

Questions raised in the recent literature on services trade and investment areclosely related to the large body of empirical evidence regarding determinants ofmultinational activity with respect to goods production and trade. In general,the literature on goods finds more support for horizontal FDI motives (Brainard,1997; Carr et al., 2001; Bloningen et al., 2003), although some studies also findevidence for the vertically integrated multinational firm (Hanson et al., 2001).Due to data limitations, this literature is largely based on aggregate data2.The data issues are even more severe for services investment than for goods,placing even more constraints on scope for empirical analysis of services tradeand FDI linkages. Indeed, because of data issues the recent literature alongthese lines uses FDI flows or stocks as a proxy for affiliate sales. For example,Grunfeld and Moxnes (2003) explore the determinants of services trade andforeign affiliate sales (they use FDI stocks as proxy for foreign affiliate sales) ina gravity model, finding that trade barriers and distance have a strong negativeimpact on exports and FDI (a proxy for foreign affiliate sales), while GDP andsimilar income levels have a significant positive impact. Kolstad and Villanger(2008) study the determinants of service FDI with panel analysis for the wholeservice sector and a small number of sub-sectors. They conclude that FDI inservices tends to be more market seeking and find strong correlation betweenmanufacturing FDI and FDI in producer services as well as an important impactof institutional quality and democracy on services FDI. Mirza and Nicoletti(2004) develop an extended gravity model and explore whether services tradediffers from trade in goods, but they do not look at FDI and foreign affiliatesales. In addition, Kimura and Lee (2006) and Lennon (2008) explore thedifferences and complementarities between trade in goods and trade in services,whereby Lennon (2008) uses disaggregated data classified in four IMF BOPSsub-sectors. Fillat-Castejon et al. (2008) examine more service sectors, based

the prices/costs of the ancillary local services that make up the ”product bundle”. For moreon this see (Fillat-Castejon et al., 2008) and (Francois and Wooton, 2010)

2See for example Bloningen (2005) for a detailed literature review on FDI determinants.Also see Markusen and Strand (2009) for extension of the knowledge capital model to the caseof business services

3

on more detailed IMF BOPS categories and stressing long-run linkages, butagain using FDI as a proxy for affiliate sales.

Globalization in services has yielded a set of sectors dominated by multina-tional companies and high profile investments, as well as a governing institu-tional structure for service trade (the GATS) that emphasizes multiple deliverymodes in the structure of trade negotiations and commitments. This in turn hasgiven rise to emphasis in the empirical literature on the determinants and therelationship between trade and FDI in services. Analytically, the literature onFDI in services is largely guided by the body of empirical evidence and relatedtheoretical literature on patterns of trade and FDI in goods. In some ways,the emphasis of the analytical literature is on factors that should apply to bothgoods and services sectors. However, given the greater role of proximity andcoordination costs between provider and buyer, we can also expect importantdifferences to emerge as the literature on services matures. Furthermore, giventhe heavy degree of industrial concentration in service industries (the top 8firms in U.S. service sectors, for example, typically account for 60 to 80 percentof sales), the interaction of market power and regulation with openness is likelyto prove quite important in future theoretical and empirical work in this area.

For service firms, there are parallels to the proximity mechanisms stressedin the literature for goods.3,4 In particular, because of what we call the prox-imity burden, we can expect increased costs linked to the coordination betweenprovider and customer as the distance increases from the firm to its customers.To the extent this holds true, increased distance may then provide increasedincentive to engage in FDI instead of cross-border trade in services, much astransport costs may encourage FDI in goods. Similarly, to the extent such FDIinvolves fixed costs, large markets offer a better opportunity to spread fixedcosts linked to FDI, so that size may play a role in the balance between cross-border and establishment based trade. At the same time, the knowledge capitalmodel implies that, for some sectors, the choice between local establishment anddirect trade may also be linked to risk of appropriation of firm-specific assets

3This includes Horstmann and Markusen (1992); Brainard (1993); Helpman (1984); Help-man and Krugman (1985); Markusen et al. (1996); Markusen (1997, 2002).

4More recently, the theoretical literature on multinational firms has highlighted hetero-geneity with respect to important characteristics such as productivity (Melitz, 2003; Helpmanet al., 2004). According to the model by Helpman et al. (2004), the decision of firms to becomemultinational depends on their productivity. Thus, setting up an affiliate in foreign countriesonly pays for the most productive firms. Firms with intermediate levels of productivity serveforeign markets through exports, while low-productivity firms produce only for the home mar-ket. It is important to note that the coexistence of firms operating through different modes ofdelivery is not limited to models of cost heterogeneity. Markusen (2002), for example, demon-strates that a mixture of MNEs and exporting firms can coexist under imperfect competitionwithout heterogeneity, depending on relative endowments and trade costs.

4

like business models (where the risk is linked to skill levels and institutionalfeatures of the market). In addition, if service firms are selling locally with amix of locally produced and home produced inputs, FDI restrictions are alsolikely to affect the relative patterns of affiliate and direct sales.

We proceed in this paper as follows. In Section 2, we provide an overviewof special characteristics of services and patterns of service trade and FDI. InSection 3 we develop a new theoretical model based on services trade and FDIinvolving Cournot competition between heterogeneous firms. We use this frame-work for mapping firm choice of modes of delivery to the bilateral patterns ofdirect trade and foreign affiliates sales observed in the data. The next section,Section 4, describes the data set in more detail and highlights the motives ofbecoming multinational. Bilateral data on foreign affiliate sales and unaffiliatedand partly affiliated cross-border sales for services as well as manufacturing sec-tors come from the Bureau of Economic Analysis. We also work with the recentIIDE Trade in Services Database, which offers a panel of bilateral total tradeflows by service sector from the 1990s to 2006. Taken together, these data allowus to recover affiliated (intra-firm) services trade by sector for several OECDsource and destination markets. Trade data for total manufacturing and sevensub-sectors are drawn from the WITS database. Our estimation strategy andthe estimation results for service sectors are also discussed in the empirical sec-tion on patterns of trade in Services. We focus in the empirics on the relativeimportance of direct cross-border trade and indirect sales through local estab-lishments, developing and exploiting a data set that merges information from anumber of sources on sector level U.S. inward and outward sales. We apply amixture of estimation methods to explore the relationship between cross-borderand FDI based modes, while avoiding simultaneity problems between affiliateand cross-border activity. A contrasting study on determinants in manufactur-ing sectors is presented in Section 5. Focusing on the U.S. as both a sourceand destination market provides insight into sector-level variation in modes ofentry (foreign affiliate sales and cross-border), including the impact of standardmeasures of economic distance and relative stocks of human capital. Our re-sults highlight sector-level variation in modes of entry (sales through foreignaffiliate and cross-border sales). While manufacturing FDI and trade patternsare mixed, for services we find that overseas multinational activity consistentlyincreases relative to direct exports as a function of distance. In addition, the rel-ative importance of establishment trade increases with more liberal FDI regimes(i.e. with lower investment barriers) and with higher manufacturing FDI. Com-mon language and market size are also significant factors. The impact of some

5

factors, like like corporate tax rates and relative stocks of human capital, alsovaries across sectors. This result highlights the heterogeneous nature of ser-vices across sectors, suggesting that the core factors to emphasize in developinga full analytical picture for trade and FDI in services may vary in importantways from the relevant set of factors for goods. Finally, given evidence of inter-dependence across modes and the importance of local affiliates, the impact ofpolicy in any one mode is likely to depend on the mix of domestic regulationand policy across modes. We offer a brief summary and concluding remarks inSection 6. The appendix provides all details on data and estimation results.

2 Patterns of services trade and FDI

The WTO distinguishes four different modes of supply5, which have also beenadopted for the General Agreement of Trade in Services (GATS):

• Mode 1 - cross-border trade: when suppliers of services in one countrysupply services to consumers in another country without either supplieror consumer moving into the territory of the other

• Mode 2 - consumption abroad: process by which a consumer resident inone country moves to another country to obtain a service

• Mode 3 - commercial presence: enterprises in an economy supply servicesinternationally through the activities of their foreign affiliates abroad

• Mode 4 - movement of natural persons: process by which an individualmoves to the country of the consumer in order to provide a service.

Multinationals are obviously important for Mode 3 trade. They are alsoimportant for Mode 4 (which includes movement of technical personnel), andMode 1 (as MNEs also engage in direct exports). The significant role of multi-national firms in trade in services is depicted in Table 1 and 2 and will bediscussed more below. From Table 1 and 2 it is clear that local presence isan important dimension of trade in services. In recent years the majority ofboth U.S. international sales and purchases of services was through affiliates(see Figure 1). In contrast to the persistent deficit in goods trade, the UnitedStates runs surpluses in trade in services. Over the period 1999-2005, bothU.S. cross-border exports and imports increased in all major service categories.The largest increase in cross-border exports and imports was in other private

5This typology for modes was developed by Sampson and Snape (1985) and was largelyadopted as a framework for the GATS.

6

services, mainly reflecting increases in business, professional and technical ser-vices.6 However, the majority of U.S. international services transactions wasthrough foreign affiliate sales, which marked strong sales growth over the pe-riod 1999-2005 (Figure 1). The largest increase for sales by U.S. multinationalsthrough their foreign affiliates was attributable to affiliates in finance and in-surance services and in ”professional, scientific and technical services”.7 Salesby foreign multinationals through U.S. affiliates were largest in finance andprofessional, scientific and technical services, but have not experienced growthin insurance services. In PST services, the largest increase was attributableto affiliates in computer and information services, management and consultingservices as well as in other PST services, including legal services, advertisingservices and architectural, engineering and other technical services.

Interestingly, in finance services the majority of sales of services is throughforeign affiliates, rather than direct cross-border trade, although the availabil-ity of online financial services is rising rapidly. This prevailing role of foreignaffiliate sales stresses the importance of location of production and proximityconstraints regarding the supply of services through multinationals. Data ontrade and foreign affiliate sales in insurance services reflect the effects of dereg-ulation. Insurance services have experienced a tremendous increase in outwardsales (sales by U.S. multinationals through foreign affiliates) and cross-bordertrade, while inward sales (sales by foreign multinationals through U.S. affili-ates) decreased slightly over the period 1999-2005. However, the dominant roleof supply through local establishments can also be seen in insurance services.

Regarding the share of unaffiliated and affiliated trade the dominant role ofaffiliated trade in business, professional and technical services is apparent. Theshare of affiliated trade for this sectoral category is much grater than the sharesof affiliate trade for other service classes, which illustrates the importance ofintra-firm trade in this service category. Trade within multinational companies(affiliate trade) accounted for 25.9 percent of U.S. exports of private services in2005 and for 22 percent of U.S. imports of private services. In contrast, affiliatedtrade in business, professional and technical services accounted for 50.1 percentof total exports and for 69.6 percent of total imports in 2005. In addition, data

6Other private services include education, financial services, insurance, telecommunica-tions, ”business, professional and technical services” and ”other services”. ”Business, profes-sional and technical services ” (BPT) consist of a variety of services, such as computer and in-formation services, management and consulting services, research and development and testingservices, operational leasing and ”other BPT services” (for instance legal services, advertisingservices, accounting services and architectural, engineering and other technical services).

7Data on foreign affiliate sales are collected separately by BEA, which explains differentnames. However, ”professional, scientific and technical services” (PST) cover the same serviceindustries as in ”business, professional and technical services” and are comparable.

7

on U.S. cross-border trade and commercial presence support a complementaryrelationship between local establishments and direct cross-border trade. For theindustries included, trade flows and affiliate sales show a positive correlation inboth directions, but the correlation is much stronger for outward activities andexports (67.71 percent) than for inward activities and imports (39.71 percent).Over the last decade, both cross-border exports and imports as well as foreignaffiliate sales increased significantly and suggest a clear interdependence acrossmodes. Depending on the mix of domestic regulations and FDI policies U.S.multinationals tend to supply foreign markets through local establishments andvia cross-border trade. The determinants of entry modes (commercial presenceversus cross-border trade) and patterns of service delivery will be examinedmore below.

3 Distance, Market Potential and Modes of Delivery

In this section, we develop a theoretical framework mapping firm choice ofmodes of delivery to the bilateral patterns of direct trade and foreign affiliatesales observed in the data. We start by a characterization of market structure.Like manufacturing, individual service sectors are typically characterized by ahandful of large firms representing a relatively large share of the market. Thispoint can be lost when looking at the full population of firms across all servicesectors collectively. However, when one focused on individual sectors, the im-portance of concentration in the services landscape is striking. For example, inthe United States, the largest 8 software publishers accounted for 46 percent ofreceipts in 2002, the 8 largest largest theater chains accounted for 60 percent ofreceipts, the eight largest wireless telecommunications carriers accounted for 82percent of receipts, the eight largest music publishers accounted for 72 percentof receipts, and the eight largest wired communications carriers accounted for 81percent of receipts. In financial services, the top 8 consumer lenders accountedfor 75 percent of receipts, the top 4 international trade financing companiesaccounted for 70 percent of receipts, the top 8 securities firms accounted for50 percent of receipts, the top four direct life insurance carriers accounted for81 percent of receipts, the top 8 commodity brokers accounted for 45 percentof receipts, and the top 8 commercial banks accounted for 42 percent of re-ceipts. There are sectors where the combined market shares of the top firmsare relatively small – including mortgage brokers – but this is the exceptionand not the rule. Even in the pawn shop sector, the top four firms accountedfor 21 percent of receipts. The general pattern in the service sectors is one ofconcentration, with large firms playing a dominant role. They are also clearly

8

the most efficient firms, with their importance in terms of output (revenues)outweighing their importance in employment. In international trade, the pat-tern is one of even greater concentration. This is clear on inspection of the U.S.FATS data for service firms. Data points are frequently suppressed in publisheddata because they represent the data of a single firm, and as such the data re-veal confidential business information. Indeed, the importance of larger firms,limited competition, and variations in oligopoly behavior characterizes serviceindustries from banking to telecommunications and transport in a broad rangeof countries (Francois and Hoekman, 2010).

Given our characterization of the market structure of service sectors, wemodel service markets as oligopolistic. In a representative service sector, werepresent each country b ∈ B as having a set of heterogeneous domestic serviceproducers i characterized by the cost vectors (impacting both fixed and marginalcosts) γbi ∈ Γb. Foreign service providers in country b may also sell services indestination market d. They have two options. One is arms-length (cross-border)transactions with production in country b and sales in country d. The otheris exporting through a foreign affiliate based in country d, so that cross-bordertransactions takes place within the firm. An important factor in the cost ofdelivery is the proximity burden. We represent this here as a transaction costthat is increasing in the international distance between the service providerand the customer. Referring back to (Hill, 1997), these costs are because of thechallenge of meeting potential problems with coordination between provider andclient. This leads to a rising marginal cost τx that interacts with the distancebetween the home market b and the destination market d. Representing distancefrom b to d as δb,d the marginal cost of a country b service exporter whenselling in country d is then (γb,dx τxδb,d). We also assume some establishmentcosts (γb,dx fdx) have to be paid as well, to manage sales, communications, andcontracting for arms-lenth exporters. The second option is for the country b firmto establish a full-scale affiliate office in country d at cost (γb,dx fdm) where fdm >

fdx . In this case, the advantages is that they can export services through theaffiliate rather than at arms length, which we assume saves on coordination costslinked to distance because the cross-border transactions are internal to the firm,while the affiliate office is assumed to facilitate better real time coordinationbetween provider and customer. This means distance costs are τm < τx. Intotal, marginal costs for a country b services MNE operating in country d arethen (γb,dm τmδb,d).

9

Under a Cournot-Nash equilibrium, for country b service exporting firmsmarket share in country d is a function of price and marginal cost. From thefirst order conditions for profit maximization, we have

θb,dx = max[−ε(P d − γb,dx τxδb,d

)/P d, 0

](1)

where ε is the elasticity of demand. Firms will only export if it is profitable,which means we also have the following export condition that holds for allexporting firms.

P d∣∣∣θb,dx >0

≥ γb,dx τxδb,d + γb,dx fdx

(θdxQ

d)−1

(2)

We have a similar set of conditions for services MNEs.

θdm = max[−ε(P d − γb,dm τmδb,d

)/P d, 0

](3)

P d∣∣∣θb,dm >0

≥ γb,dm τmδb,d + γb,dm fdm

(θdmQ

d)−1

(4)

We can rearrange equations (2) and (4), yielding the following necessary con-ditions for the export and MNE decision. This involves substitution of ourdefinitions of θdm and θdx. For exporters, we have the condition in equation (5):

P d − γb,dx τxδb,d ≥ γb,dx fdx

(θdxQ

d)−1

P d − γb,dx τxδb,d ≥ −γb,dx P dfdx

(P d − γb,dx τxδb,d

)−1 (εQd

)−1

QdP d ≥ −

(γb,dx fdxε

) P d(P d − γb,dx τxδb,d

)2

(5)

A similar condition holds for multinational firms.

QdP d ≥ −

(γb,dm fdmε

) P d(P d − γb,dm τmδb,d

)2

(6)

Equations (5) and (6) lead directly to the following propositions.

Proposition 1 In larger markets, we will observe more firms meet the cutoffconditions for cross-border exporters and MNEs.

Proposition 2 In more distant markets, we will observe fewer firms that canmeet the cutoff conditions for cross-border exporters and MNEs.

10

Propositions 1 and 2 follow from the left hand side of both equations (5) and(6), as in both cases this increase the range of distance and marginal costswhere trade is profitable. The relative efficiency ranking of MNE and serviceexporters, however, is more complex. It depends on the ranking of fdm > fdx andalso τm < τx. It is the case that because fixed costs are higher for MNEs, theyneed to have sufficient market share to cover the fixed costs involved. This ismore easily met when marginal costs are lower, so that like the export decisionthere is a strict ranking of firm efficiency and cutoff in the option to be an MNE.There will be cases where firms that find acting as an MNE profitable will alsofind direct exporting profitable. Which business model they follow depends onrelative profits from both modes. To sort out this aspect of the decision process,we first need to define profits as follows:

πm =(P d − γb,dm τmδb,d

)θdmQ

d − γb,dm fdm

= −εQd(P d − γb,dm τmδb,d

)(P d − γb,dm τmδb,d

)/P d − γb,dm fdm (7)

πx =(P d − γb,dx τmδb,d

)θdxQ

d − γb,dx fdx

= −εQd(P d − γb,dx τxδb,d

)(P d − γb,dx τxδb,d

)/P d − γb,dx fdx (8)

The choice of mode follows from the condition that firms choose establishmentwhen πm − πx > 0.

πm − πx = γb,dj

(fdx − fdm

)− εQ

d

P d

[(P d − γb,dj τmδb,d

)2−(P d − γb,dj τxδb,d

)2]

= γb,dj

(fdx − fdm

)− εEd

[µ2m,j − µ2

x,j

](9)

In the first version of equation (9), the last term on the right hand side is in-creasing in the distance coefficient δb,d. This follows from as long as τx > τm. Itis also increasing in the firm cost coefficient γi. This means we can characterizethe aggregate pattern of trade across modes (aggregated across firms), thoughat the firm level there is a level of ambiguity. At the same time, market sizewill drive a move to establishment trade, as the second term on the right handside is clearly increasing in market size. In the second version of equation (9),Ed = P dQd is the total size of the market d in terms of expenditure. Expressedthis way, the revenues from higher markups need to be enough to cover higherfixed costs.

11

The conditions expressed in equations (2), (4), and (9) can be manipulatedto relate entry conditions to the choice of modes. With some manipulation,equations (2) and (4) can be re-written as follows.

kdx = −εEd(P d − γb,dx τxδb,d

)P d

2

− γb,dx fdx ≥ 0 (10)

kdm = −εEd(P d − γb,dm τmδb,d

)P d

2

− γb,dm fdm ≥ 0 (11)

From (10) and (11), we can re-write equation (9) in terms of the new entryconditions kx and km.

zdj = πm − πx = γb,dj

(fdx − fdm

)−εEd

(P d − γb,dj τmδb,d

P d

)2

−

(P d − γb,dj τxδb,d

P d

)2

= kdm − kdx (12)

From inspection of equations (10) and (11), the zero (k = 0) for establishmenttrade is at a lower cost parameter γ than is the case for direct trade. Thismeans only the most efficient firms will be able to opt for establishment tradefor sufficiently high fixed cost differences. However, the location of the zerosalso means kx > km in the region of threshold condition for establishment trade.When will firms opt for establishment trade? From equation (12), for firms tochoose establishment it has to be the case that kx < km. Differentiation ofequation (12) yields the following:

∂zdj∂γj

=(fdx − fdm

)− 2εEdδb,d

[(τx − τm)−

(τ2m − τ2

x

) γjP d

](13)

From equation (13), the distance between km and kx falls with increasing valuesof γ.

We illustrate the two entry conditions with a numeric example in Figure 2.8

For sufficiently high fixed cost differences, the cut-off cost coefficient parameteris lower for establishment trade than it is for direct trade.9 However, at the sametime, from equation (13), at lower value of γ, the km curve will eventually cut

8For the example in the figure the coefficients are τx=0.13 τm=0.06, fx=2, fm=80, ε=-2.0,Ed=2000, P d=2, and δb,d=1.

9Without this condition, firms would never choose direct trade, and all firms would chooseestablishment trade, with the km curve lying strictly above the kx curve.

12

above the kx curve. Firms to the left of the intersection find both establishmentand export trade profitable, but will choose establishment trade. Firms to theright of the intersection will choose direct exporting, even if they would alsofind establishment trade profitable. Also, controlling for both distance andthe revenue size of the market, less competitive markets with higher pricesalso move both curves to the right while moving the intersection to the right.We illustrate the relationship of larger equilibrium market size to the observedpattern of modes of delivery in Figure 3, using the numeric value from Figure2 but with a 10 percent increase in total market revenues. What can be seenis that both z curves move to the right, along with a shift in the cutoff point.Basically, we expect to see more firms operating in larger markets, while atthe same time a larger number of firms will opt for establishment based trade(with the zm curve lying above the zx curve). Finally, in Figure 4 we use thesame numeric values to highlight the impact of rising distance and so distance-related costs (17.5 percent increased distance in the figure), again leading to ashift in the relevant zm and zx curves. The intersection of the two curves shiftsincreasingly close to the horizontal axis as distance rises, eventually crossing,at which point establishment dominates as the only viable option, and so direct(unaffiliated) trade vanishes. All these observations follow from the propertiesof equation (9). We summarize them formally in the following propositions.

Proposition 3 At the firm level, for sufficiently high fixed costs linked to es-tablishment, only the most efficient firms will choose establishment. Firms withintermediate levels of efficiency will choose direct exports, while the least effi-cient firms will not export.

Proposition 4 At the firm level, there is a range of costs where firms wouldfind establishment trade profitable, but would still engage in direct trade.

At the industry level, increasing market size in terms of revenue moves bothcurves to the right, while moving the intersection to the right as well. Thismeans the incentive to engage in FDI instead of unaffiliated sales increaseswith the size of the destination market Qd for all firms. Therefore we will havean unambiguous shift from observed direct cross border sales to affiliate salesin aggregate, as a function of size of the destination market.

Proposition 5 Increasing size of the destination market will lead to an unam-biguous shift from observed direct cross border sales to affiliate sales in aggre-gate.

At the industry level, increasing distance moves both curves in Figure 4 down,but also moves the intersection to the right. The incentive to engage in FDI

13

instead of unaffiliated sales increases with distance to the destination marketQd for all firms. This implies an unambiguous shift from observed direct crossborder sales to affiliate sales in aggregate, as a function of distance.10 Thisleads us to Proposition 6.

Proposition 6 The incentive to engage in FDI instead of unaffiliated salesincreases with discrete changes in distance to the destination market Qd.

At the industry level, higher price in a market moves both curves to the right,while moving the intersection to the right as well. The incentive to engage inFDI instead of unaffiliated sales increases with price levels in the destinationmarket Qd for all firms. This implies an unambiguous shift from observed directcross border sales to affiliate sales in aggregate, as a function of prices. Thisleads us to Proposition 7.

Proposition 7 The incentive to engage in FDI instead of unaffiliated salesincreases with price levels in the destination market Qd for all firms.

Next we offer some observations on the role of distance in total bilateral tradeand establishment trade. From equation (3), market share of firms already op-erating as multinationals is declining with distance. However, from Proposition5 we have a shift from trade to multinational operations with rising distance.This implies a potential ambiguity for the impact of distance on observed MNEactivity. For direct trade, Proposition 5 means some share of firms shift outof exporting, either into establishment trade or out of the market entirely. Forremaining firms, equation (1) means they will have falling market share and wewill have falling direct trade volumes. We summarize these points as follows.

Proposition 8 At the industry level, greater distance means an unambiguousfall in total direct cross-border bilateral sales.

Proposition 9 At the industry level, greater distance means an ambiguouschange in total bilateral establishment sales. If the switching of firms fromcross-border sales to establishment sales is minimal (i.e. if their market shareis sufficiently small) total bilateral establishment sales are declining with dis-tance.

10While Proposition 6 holds strictly for large changes in distance, there is a second ordereffect for marginal firms with low market shares that follows from the squeezing out of bothtypes of providers. In the region of marginal firms (those close to the threshold of droppingout) there may be a shift from establishment back to direct exporting because of decliningmarket size. Because these are those with the smallest market share, these firms drop outentirely for discrete (non-marginal) increases in distance. As such, we should still expect thetotal share of establishment sales to rise at industry level. We can offer numeric examples ofsuch effects on request.

14

Finally, we are also interested in the impact of policies restricting foreign invest-ment that increase fixed costs on total bilateral trade and establishment trade.From equation (12), increasing the fixed costs of entry is qualitatively the sameas a fall in market size in the destination market.

∂zdj = γb,dx

(∂fdx − ∂fdm

)(14)

Where we have an x% increase in fixed costs, then as long as establishmentrelated fixed costs are greater than cross-border related fixed costs, limits onFDI will induce some firms to shift from establishment trade to cross-bordertrade. 11 From equations (3) and (1), market share of remaining firms sellingthrough establishment will fall, while the sales of existing firms selling throughdirect exports will also fall. Unless the switch of firms from establishment tocross-border trade dominates, we should therefore also see a complementaritybetween cross-border and establishment-based sales in the face of policies thatraise fixed costs for both modes. We summarize these points as follows:

Proposition 10 At the industry level, policies that raise fixed costs in similarproportions for establishment and cross-border trade will induce a shift by somefirms out of establishment trade and into cross-border trade.

Proposition 11 Policies that raise fixed costs in similar proportions for estab-lishment and cross-border trade will lead to lower observed market shares for allindividual firms.

Proposition 12 There is scope for complementarity in aggregate, where bi-lateral cross-border and establishment trade rise and fall together in the face ofpolicy-induced cost increases, even as individual firms substitute between modes.However, a substantive shift in firm population from establishment to cross-border trade can lead to apparent substitution in aggregate flows across modes,even as market shares for all firms and total sales in the destination marketfall.

Basically, we have a clean ranking of the aggregate pattern of modes, in termsof market size or potential, degrees of competition, and distance. We may alsohave firms that export directly to some markets, and sell through affiliates inothers. At the same time, among the subset of potential MNE firms, the leastefficient of these firms may be the ones that engage in establishment based trade.In the empirics that follow, we focus on bilateral patterns at the industry level,following directly from Propositions 1,2,5,6,7,8, 9, and 12.

11In terms of Figure 3, this is analogous to changes in destination market size, with both zcurves move to the right, along with a shift in the cutoff point.

15

4 Empirics - Patterns of Trade in Services

We now turn to empirics. In order to give a better understanding of the inter-actions between various modes we focus on the linkages between cross-bordertrade and local sales of services through affiliates at a sectoral level - distinguish-ing between 13 major service categories. The choice of the estimating framworkis guided by our propositions in Section 3. Our empirical results provide in-sight into sector-level variation in modes of entry (foreign affiliate sales andcross-border). We employ a mixture of GLM models, three-stage least squaresmethods and seemingly unrelated equations to explore the relationship betweencross-border and FDI based modes. In order to avoid simultaneity problems be-tween affiliate activity and cross-border trade we use shares of affiliate sales andcross-border sales in total foreign sales as dependent variables. Distinguishingbetween inward and outward activities our empirical analysis is based on thefollowing two shares first introduced by Brainard (1993, 1997):

• Outward affiliate sales share: Share of outward affiliate sales in totaloutward sales

• Inward affiliate sales share: Share of inward affiliate sales in total inwardsales

4.1 Data

For the purpose of this paper we work with a multi-sourced data set on sectorlevel U.S. exports and imports. These are summarized in Table 3. Bilat-eral data on foreign affiliate sales as well as unaffiliated and partly affiliatedcross-border sales come from the Bureau of Economic Analysis. Cross-bordertransactions include both affiliated and unaffiliated transactions between U.S.companies and foreign residents. Affiliated cross-border trade indicates intra-firm trade within multinational companies and consists of trade between U.S.parent companies and their foreign affiliates and transactions between U.S. affil-iates and their foreign parent groups. In order to describe Mode 3, commercialpresence, we make use of U.S. Foreign Affiliates Trade Statistics (FATS) pub-lished by BEA, which illustrates the importance of services in affiliate activities.These data are drawn from benchmark and annual sample surveys of U.S. directinvestment abroad and of foreign investment in the United States. By usingthe FATS data we gather information on sales of services by majority-ownedforeign affiliates and on sales of majority-owned U.S. affiliates. One advan-tage of this data is that it has been classified by destination of sales through

16

affiliates. However, data on sales through affiliates are published by primaryindustry of the affiliate and not by type of service which asks for reclassification.Our dataset on foreign affiliates sales finally covers 13 service sectors over theyears 1997 to 2005. In addition, we also work with the recent IIDE Trade inServices Database, which offers a panel of bilateral total trade flows by sectorfrom the 1990s to 2006. Taken together, these data allow us to recover affiliated(intra-firm) services trade by sector for several OECD source and destinationmarkets. Data on foreign affiliate sales in manufacturing sectors are also takenfrom the BEA’s publications. Additionally, trade data for total manufacturingand seven sub-sectors are drawn from the WITS database. The final datasetincludes bilateral U.S. trade and foreign affiliate sales for 10 partner countriesin total. They include: Australia, Brazil, Canada, France, Germany, Japan,Mexico, The Netherlands, Switzerland and the United Kingdom.

To identify the determinants of entry modes we use several explanatory vari-ables suggested by the recent theoretical and empirical literature. The size ofthe host country markets are captured through GDP (measured in billions ofcurrent U.S. dollars). According to previous literature, market size is expectedto have a positive impact on services trade and foreign affiliate sales. GDP,income and population data are taken from the IMF and World DevelopmentIndicators database respectively. In addition, we employ a similarity index forper capita income to proxy for skill and human capital differences. The simi-larity index ranges from 0 to 1, whereby a higher score means a higher degreeof per-capita income similarity, implying a similar per capita stock of skills andhuman capital. The similarity index is used to help identify whether multina-tional activity is motivated by horizontal FDI strategies (trading partner aremore similar) or vertical FDI strategies (trading partner are dissimilar in theirfactor endowments and factor prices). Hence, a positive coefficient on the simi-larity index can be interpreted in favor of horizontal FDI motives. In addition,per capita income inequalities can also be used to test for the convergence hy-pothesis (Markusen, 1995; Markusen and Venables, 1998) which suggests thatmultinational firms become more important relative to domestic firms the moresimilar are countries in size and endowments. Next, to reflect the proximityburden, we include geographic distance12 between the United States and therespective partner countries as a proxy for transportation costs (variable dis-tance costs). Hence, we expect a positive coefficient on geographic distanceif local establishment sales and cross-border trade act as substitutes (support-ing the horizontal FDI model), since variable distance cost make cross-border

12Geographic distance is calculated following the great circle formula, which uses latitudesand longitudes of the relevant capital cities.

17

trade more expensive so that affiliates in the host country could pay for firms.A negative coefficient on distance indicates a complementary relationship be-tween trade and foreign affiliate sales and supports the vertical FDI model. Inorder to capture some cultural influences we include a language dummy, whichindicates if home and host country share a common language familiarity andgenerally share the same cultural heritage. Since a common language plays animportant role in services trade we expect a positive coefficient on language,fostering the establishment of affiliates in the host market to a greater extentthan cross-border trade. Geographic distance, together with data on culturalfamiliarity are taken from Mayer and Zignago (2006)13.

All determinants of affiliate activity described so far, such as market size,geographic distance to trade partners, and to a lesser extent economic devel-opment and education levels, are beyond the influence of trade policy. Nev-ertheless, economic and trade policies are used to influence the activities ofmultinational firms through various channels (Blomstrom and Kokko, 2003),and so these need to be accounted for in the economic analysis. To capture theimpact of FDI and trade policies on multinational activity we include severalindices designed to quantify the underlying trade and investment climate ofhost and source markets. We include the measures of the OECD’s FDI Reg-ulatory Restrictiveness Index14. The index captures deviations from nationaltreatment in order to identify discrimination against foreign investment and ismeasured on a 0-to-1 scale, with 0 representing full openness and 1 prohibition.Hence, we expect a negative coefficient on the FDI regulatory restrictiveness in-dex. The advantage of the OECD’s FDI restrictiveness index is that it displayssector-specific levels of restrictiveness and covers important main sectors andseveral sub-sectors. In addition, we make use of the Heritage Foundation in-dex of economic freedom which comprises 10 components of economic freedom,such as trade freedom and investment freedom as well as an averaged overallscore. The indices are scaled from 0 to 100, with 100 indicating the highestlevel of freedom. Hence, we expect a positive coefficient on this variable. TheHeritage Foundation index does not account for sectoral freedom differencesand is therefore only used for sensitivity analysis15. In addition to indices dis-playing the investment and trade climate, we include the destination-marketcorporate income tax rate for obvious reasons regarding location decisions ac-cording to tax rate advantages. Central and sub-central corporate income tax

13http://www.cepii.com/anglaisgraph/bdd/distances.htm14http://www.oecd.org/dataoecd/1/40/40476272.pdf15Results using the Heritage Foundation index can be obtained upon request from the

authors.

18

rates are taken from the OECD’s tax database16 and KPMG’s Corporate andIndirect tax rate Survey 200717. The establishment of local affiliates shouldbe preferred towards cross-border trade the lower is the corporate tax rate inthe destination market relative to the home market, since affiliate profits aretaxed at the foreign tax rate while cross-border trade profits are taxed at thehome-market rate. Furthermore, to address the discussion on fragmentationand the increased importance of traded services in the fragmentation processwe include manufacturing FDI18 in our empirical analysis. This allows us tostudy the role of services as inputs in the manufacturing process and accountsfor indirect exports of services. Thus, we expect a positive relationship betweenmanufacturing FDI and affiliate activity. In addition, given the linkage betweenmanufacturing and services trade, both FDI streams are influenced by invest-ment regulation and policies across modes which needs to be considered in theactual policy environment.

4.2 Preliminaries

Before we turn to the relationship between trade across the border and pro-ducing abroad, we discuss the patterns of cross-border and affiliate sales us-ing a gravity type approach. From Propositions 1, 2, 8, and 9, both marketsize and distance should be important for total observed trade. In its originalformulation, the gravity model originally predicted bilateral trade flows as afunction of distance between any two countries and their size. The approachhas been widely applied in international trade literature. Recently, the originalmodel specification (Tinbergen, 1962) has been augmented by the inclusion ofadditional variables which are thought to effect trade flows, such as dummyvariables for language familiarities, trade barriers or historical linkages betweenthe countries. In addition, better controls have been introduced for country-specific factors in the standard model of bilateral flows (Baldwin and Taglioni,2006; Feenstra, 2002). Since trade flows between countries change over time,the empirical estimation of gravity models is increasingly conducted using paneldata specifications which is also used in this paper. In formal terms, we use anaugmented standard gravity model, which can be written as

lnTradejit = α0 + β1Xit + β2Xi + β3Xij + εjit, (15)

16http://www.oecd.org/ctp/taxdatabase17http://www.kpmg.com/SiteCollectionDocuments/2007CorporateandIndirectTaxRateSurvey.pdf18Data on manufacturing FDI are taken from the Bureau of Economic Analysis and comprise

FDI in the United States and U.S. direct investment positions abroad.

19

where i, t and j index countries, time and service sectors. The dependentvariable Tradejit represents either direct cross-border trade volumes (exportsor imports) or affiliate sales (outward or inward affiliate sales). While vectorXit represents time-varying explanatory variables for country i (GDP, similar-ity index, tax rates, freedom indices, etc.), vector Xi comprises time invariantexplanatory variables for country i (distance, common language). Vector Xij

displays sector-varying explanatory variables for country i, like the FDI restric-tiveness index. Concerning the interpretation of the results it is important tonote that the gravity equations for outward flows and inward flows differ. Whilethe outward gravity equation depicts how characteristics of the host markets ef-fect the volume of trade flows, given that the home market is the United States,the inward equations display how characteristics of the source country influencethe volume of trade, given that the destination market is the United States. Weapply a system of simultaneous system estimations, in which exports and af-filiate sales are treated as endogenous. Table 4 reports the patterns for levelsof outward affiliate sales and cross-border exports, applying three-stage leastsquares (3SLS) methods. As hypothesized in Proposition 6 before, increaseddistance between suppliers and consumers makes exports less attractive andleads to more affiliate activity, while exports decrease with increasing variabledistance cost. The coefficient of the restrictiveness index has the predicted neg-ative sign for affiliate sales. Affiliate activity decreases the higher are FDI flowrestrictions and U.S. multinationals serve foreign markets through exports ifdiscriminatory barriers are high. Moreover, the United States tend to exportmore the bigger are host markets. Interestingly, the impact of the endogenousvariables affiliate sales and exports are both positive, indicating a complemen-tary relationship between cross-border exports and commercial presence. FromProposition 12 above, this means that the drop in sales across firms in bothmodes dominates any switching that takes place between modes. In addition,our results suggest a positive relationship between manufacturing FDI and af-filiate sales for some service sectors. Moreover, a common language just seemsto be important for some service sectors, like telecommunication, finance, in-surance and R&D services. The other gravity variables such as the corporatetax rate or the similarity index vary across sectors, indicating the heterogenousnature of services.

Turning to the inward gravity equations we consider that patterns of cross-border imports and affiliate sales depict the characteristics of the home market.Table 5 reports the gravity estimation results applying 3SLS estimation pro-cedures. We find a less consistent impact of the distance variable on inward

20

side, which varies by sector. Since our setup just allows us to control for homemarket characteristics, the results on the inward side are in general less consis-tent than what we have seen before on the outward side. Common languagefamiliarities seem to have an important impact on inward affiliate sales and im-ports to the United States. Moreover, our gravity pattern suggest that inwardaffiliate sales increase the bigger is the home market (except for insurance andfinancial services), indicating that bigger home markets tend to import moreto the United States. However, no clear conlusion can be drawn from the re-strictiveness index. The coefficient on manufacturing FDI is not reporting anysignificant linkages between investment in manufacturing and service sectors inalmost all sectors, apart from business services.

The basic gravity type patterns of cross-border and affiliate sales highlightthe importance of factors like distance, market size, manufacturing FDI anddiscriminatory barriers for affiliate sales. Since direct cross-border trade andaffiliates sales may act as alternative modes of foreign market penetration, weexamine the impact of distance, FDI restrictions, language familiarity, skilland human capital differences, FDI in manufacturing and additional factors indetermining the motive of overseas production relative to direct cross-bordersales.

4.3 Outward Shares

Given that the United States is always the home market, outward shares analyzehow characteristics of the destination markets determine the choice betweenexporting and affiliate sales. The baseline econometric model can be written as

OUTSHjit = α0 + β1Xit + β2Xi + β3Xij + υjit, (16)

where i, t and j index countries, time and service sectors. While vector Xit rep-resents time-varying explanatory variables for country i (GDP, similarity index,tax rates, freedom indices, etc.), vector Xi comprises time invariant explana-tory variables for country i (distance, common language). Vector Xij displayssector-varying explanatory variables for country i, like the FDI restrictivenessindex. υjit represents the respective error term. As proposed by Papke andWooldridge (1996) we use GLM-based estimators as preferred econometric ap-proach. As a measure of robustness of the results we also use the share ofexports in total outward sales as dependent variable. Table 6 reports the re-sults using OUTSH, the share of outward affiliates sales in total outward sales,as dependent variable applying GLM estimation approaches. We find a veryconsistent positive impact of distance on the likelihood of U.S. firms to establish

21

affiliates in foreign countries across all service categories. The results suggestthat proximity between consumers and suppliers of services is still needed ordesired, so that multinational activity relative to exports increases the furtheraway are destination markets (see Proposition 6). The impact of FDI restric-tions on the share of outward affiliate sales is consitently negative, indicatingthat discriminatory barriers have a strong interfering aspect on affiliate salesexcept for legal services. The coefficient of the common language variable showsgenerally the expected sign, although our results support sector level differences.In addition, regarding characteristics between the United States as the homemarket and various destination markets the coefficients on the similarity indexshow up differences between sectors. While we stated above that vertical FDImainly takes place between countries which are dissimilar in their factor en-dowments and horizontal FDI takes place between similar countries, the resultsare ambiguous and once again point out that services differ across sectors.

In a first step, we estimated equation (16) alone using GLM specifications.In a second step, we also test for linkages between total outward sales (affiliatesales+cross border sales) and the outward affiliate sales share. Consequently,we estimate both influencing variables jointly using the seemingly unrelated re-gression (SUR) procedure. Since both variables - total sales and FDI intensity- are influenced by the same determinants, the residuals of both single equa-tions might contain some information of omitted variables. By imposing a jointvariance covariance structure SUR takes contemporaneous correlations into ac-count. Table 7 reports results from the baseline equation using SUR estimation.Applying SUR supports the results found in the base equation and related lit-erature and improves the results and significance levels for the basic results.Again, our results show a consistent strong impact of distance, discriminatorybarriers and language familiarities on total sales and affiliate sales. Moreover,the coefficient on manufacturing FDI supports the previous findings in relatedliterature that manufacturing FDI is intertwined with trade in services, sinceservices are used as inputs in the manufacturing process.

4.4 Inward Shares

Turning to imports and foreign multinational activity in the United States it isimportant to note that equation regarding the inward share is not equivalentto the equation depicting the outward share. In particular, while the outwardequation explains how characteristics of the host markets determine the choicebetween exporting and local presence abroad, given that the home market is theUnited States, the inward equation displays how characteristics of the source

22

country influence the mode of entry, given that the destination market is theUnited States. The baseline econometric model can be written as

INSHjit = α0 + β1Xit + β2Xi + β3Xij + ψjit, (17)

where i, t and j index countries, time and service sectors. While vector Xit rep-resents time-varying explanatory variables for country i (GDP, similarity index,tax rates, freedom indices, etc.), vector Xi comprises time invariant explanatoryvariables for country i (distance, common language). Vector Xij displays sector-varying explanatory variables for country i, like the FDI restrictiveness index.ψjit represents the respective error term. Similar to the outward equation weapply GLM estimation procedures and use the share of imports as alternativein order to check for the robustness of the results. Estimates on the share ofinward affiliate sales in total inward sales are reported in Table 8. The impactof distance on the share of inward affiliate sales is ambiguous and varies bysector. While some sectors, such as financial, information and R&D services,show a significant positive effect, the FDI intensity decreases with increasingdistance in other sectors, like insurance and computer services. We find a con-sistent positive impact of the language dummy, indicating the enhancing impactof a common cultural heritage on affiliate sales except for insurance services.Moreover, it seems that bigger markets and manufacturing FDI offer a betteropportunity to undertake FDI in services in the United States, though the im-pact differs by sector. The impact of the restrictiveness index on affiliate salesremains ambiguous. In the same way, our results on the impact of tax ratedifferences as well as the similarity index do not allow a clear interpretation ofthe results.

Going one step further we reestimate the base equation (17) together withtotal inward sales (affiliate sales+cross border sales) in a SUR system to accountfor linkages between total inward sales and the inward affiliate sales share.The results on the SUR estimation are presented in Table 9. We find similarresults as in the baseline estimation, indicating the importance of distance, FDIrestrictions, manufacturing FDI and language familiarities for affiliate activity.

5 Empirics - Comparison to Manufacturing

Fragmentation of production processes in order to increase efficiency and prof-its has accelerated in manufacturing sectors over the last decade. Technologicalchanges as well as trade and investment liberalization foster this fragmentationprocess, which is characterized by increasing complexity and international ori-

23

entation. Since services are increasingly used in manufacturing processes - asintermediate inputs but also as stand-alone production components - the inter-twined linkage between services trade and manufacturing is apparent. Includingmanufacturing FDI in our regressions, we find a positive effect of manufacturingFDI on affiliate activity for some service sectors in inward and outward sales.Our results support the results on manufacturing and services linkages previ-ously found in the economic literature (Francois and Woerz, 2008; Gage andLesher, 2005). Going one step further, we aim at identifying whether motivesfor investing abroad are similar between manufacturing and service sectors.Although there exists a growing literature analyzing the determinants of FDIempirically, mainly based on aggregate data, but there also some studies focus-ing on the determinants of services FDI, little attention is paid to the issue ofcontrasting manufacturing and service sectors. Riedl (2008) contrasts the dy-namic patterns of manufacturing and service FDI in transition countries usingthe data from the WIIW database on Foreign Direct Investment. She finds thatinvestment in services adjust much faster to its desired level than manufacturingFDI.

Our analysis here addresses the open question by examining the motives forgoing abroad in total manufacturing, as well as seven sub-sectors. By applyingthe same econometric specifications we will then compare our results on thedeterminants in manufacturing sectors to our results on the motives for affiliateactivity in service sectors, which are discussed in Section 4. Accordingly to theunique characteristics of services, which require the proximity between supplierand consumer, we find a consistent positive impact of distance on inward andoutward affiliate sales. In contrast, we expect manufacturing FDI to take placebetween countries located next to each other, in order to guarantee proximityto large markets and to exploit agglomeration advantages. Moreover, manu-facturing services are likely to be affected by efficiency motives, rather thanmarket-seeking motives hypothesized for FDI in service sectors.

In order to examine the motives for affiliate activity in manufacturing sec-tors we perform similar estimation procedures as discussed above. Hence, weuse the share of outward affiliate sales in total outward sales and the share ofinward affiliate sales in total inward sales and apply GLM specifications. Theestimation results for the outward affiliate sales share in manufacturing sectorsbased on equation (16) are presented in Table 10. Table 10 shows the reversedimpact of distance on affiliate activity and cross-border trade in manufacturingservices. In contrast to our results found in service sectors, affiliate sales de-crease the further away are host countries, while direct exports increase with

24

distance. These results are consistent across all manufacturing sectors withsignificant impacts of the distance variable, except for machinery and the foodsub-sector. But it’s significantly negative for the transport equipment sector,which is mainly dominated by trade in automotive parts. In the case of mar-ket size, proxied by GDP, we observe a significant positive impact on affiliateactivity of the proximity to large markets hypothesis. Our results support theidea and also previous findings, that manufacturing FDI is positively and sig-nificantly affected by market size. To test whether manufacturing FDI is drivenby horizontal FDI motives or vertical FDI motives, we find a consistent nega-tive impact on manufacturing FDI. Thus, we find evidence in favor of verticalFDI motives for manufacturing sectors. Furthermore, the estimation results forthe share of outward affiliate sales in manufacturing sectors suggest a positiveimpact of language familiarity on affiliate activity, indicating that a commonlanguage heritage fosters affiliate activity significantly in 5 out of 8 manufac-turing sectors. The impact of FDI restrictions shows the predicted negativesign. However, manufacturing sectors are relatively less restricted. Using theHeritage Foundation index instead, we find a significantly positive impact ofthe investment climate on affiliate sales19.

Turning to the inward side it is important to note that specifications re-garding the inward shares are not equivalent to the specification for the out-ward shares. The inward shares display the characteristics of the home market,given that the United Sates is the destination market. Table 11 reports theestimation results, based on equation (17) applying GLM specifications. Theestimation results for the inward share shows similar decisive motives for affili-ate activities, as found for the outward share. Looking at the distance variable,the importance of proximity to host countries is apparent. Thus, affiliate activ-ity decreases the further away are the United States as host market, except fortotal manufacturing. Moreover, affiliate activity is more likely the larger is themarket size of the home country. Moreover, we find a strong negative impactof FDI restrictions in the respective home country on foreign direct investmentin the United States. Again, we use the Heritage Foundation index and thecoefficient on the investment freedom index of the home market suggests thata less restricted investment climate allows more affiliate activity in the UnitedStates20.

19Results using the Heritage Foundation index can be obtained upon request from theauthors.

20Results using the Heritage Foundation index can be obtained upon request from theauthors.

25

6 Conclusions

In this paper we focus on the relationship between direct cross-border tradeand indirect sales through affiliates as alternative modes of services delivery.We offer a parsimonious analytical framework for the choice of mode. This isbased on a model of heterogenous service firms operating under oligopoly. Theframework yields predictions about the choice of service delivery at the firmlevel, and about the pattern of bilateral trade and MNE activity at the industrylevel. While we then focus on industry-level predictions in the empirical sectionof this paper, the analytical model developed here could also be applied to firmlevel data as well. At the industry level, the importance of proximity betweensupplier and consumer, what we call the proximity burden, appears empiricallyrobust in explaining increased affiliate activity relative to cross-border sales withincreased distance. Multinational activity in services increases relative to directexports the further away are host countries, the lower are investment barriersand the higher is manufacturing FDI. Common language familiarities and biggermarkets also foster affiliate activity. The impact of factors like corporate taxrates and relative stocks of human capital on modes of service delivery variesacross sectors. For manufacturing firms, the impact of the key factors identifiedhere differ in important ways from our results for service sectors, mainly withrespect to distance and the underlying proximity burden for services trade.The empirical results suggesting that core factors to emphasize in developinga full analytical picture for trade and FDI in services may vary in importantways from the relevant set of factors for goods. In addition, the impact of anyparticular policy on one mode of services delivery is likely to interact with thebroad mix of domestic regulation and policy across all modes of supply.

26

References

Baldwin, R. E., Robert-Nicoud, F., 2007. Offshoring: General equilibrium ef-fects on wages, production and trade. NBER Working Paper No. 12991.

Baldwin, R. E., Taglioni, D., 2006. Gravity for dummies and dummies for grav-ity equations. CEPR Discussion Paper No. 5850.

Bhagwati, J. N., 1984. Splintering and disembodiment of services and develop-ing nations. The World Economy 7, 133 – 144.

Blomstrom, M., Kokko, A., 2003. The economics of foreign direct investmentincentives. NBER Working Paper No. 9489.

Bloningen, B. A., 2005. A review of the empirical literature on fdi determinants.NBER Working Paper No. 11299.

Bloningen, B. A., Davies, R. B., Head, K., 2003. Estimating the knowledgecapital model of the multinational enterprise: comment on carr, david l.American Economic Review 93(3), 980 – 994.

Brainard, S. L., 1993. A simple theory of multinational corporations and tradewith a trade-off between proximity and concentration. NBER Working PaperNo. 4269.

Brainard, S. L., 1997. An empirical assessment of the proximity-concentrationtrade-off between multinational sales and trade. American Economic Review87, 520 – 544.

Carr, D. L., Markusen, J. R., Maskus, K. E., 2001. Estimating the knowledgecapital model of the multinational enterprise. American Economic Review91(3), 693 – 708.

Feenstra, R., 2002. Border effects and the gravity equation: Consistent methodsfor estimation. Scottish Journal of Political Economy 49(5), 491 – 506.

Fillat-Castejon, C., Francois, J. F., Woerz, J., 2008. Cross-border trade andfdi in services. Department of Economics, Johannes Kepler University Linz.Working Paper No. 0812.

Francois, J. F., 1990. Trade in nontradables: Proximity requirements and thepattern of trade in services. Journal of Economic Integration 5 (1), 31–46.

Francois, J. F., Hoekman, B., 2010. Services trade and policy. Journal of Eco-nomic Literature forthcoming.

Francois, J. F., Woerz, J., 2008. Producer services, manufacturing linkages, andtrade. Journal of Industry, Competition and Trade 8 (3), 199–229.

Francois, J. F., Wooton, I., 2010. Market structure and market access. WorldEconomy forthcoming.

27

Gage, J., Lesher, M., 2005. Intertwined: Fdi in manufacturing and trade inservices. OECD Trade Policy Working Papers No. 25.

Grunfeld, L. A., Moxnes, A., 2003. The intangible globalization: Explaining thepatterns of international trade in services. Norwegian Institute for Interna-tional Affairs. Working Paper No. 657 - 2003.

Hanson, G. H., Mataloni, R. J. J., Slaughter, M. J., 2001. Expansion strategiesof u.s. multinational firms. Brookings Trade Forum, 245–282.

Helpman, E., 1984. A simple theory of international trade with multinationalcorporations. Journal of Political Economy 92 (3), 451 – 471.

Helpman, E., Krugman, P. R., 1985. Market structure and Foreign Trade. MITPress, Cambridge.

Helpman, E., Melitz, M. J., Yeaple, S. R., 2004. Export versus fdi. AmericanEconomic Review 94(1), 300–316.

Hill, T. P., 1997. On goods and services. The Review of Income and Wealth 23,315–338.

Horstmann, I. J., Markusen, J. R., 1992. Endogenous market structures ininternational trade (natura facit saltum). Journal of International Economics32, 109 – 129.

Kimura, F., Lee, H.-H., 2006. The gravity equation in international trade inservices. Review of World Economics 142(1), 92 – 121.

Kolstad, I., Villanger, E., 2008. Determinats of foreign direct investment inservices. European Journal of Political Economy 24, 518 – 533.

Lennon, C., 2008. Trade in services and trade in goods. Paris-Jordan SciencesEconomiques. Working Paper No. 2008-52.

Markusen, J., Strand, B., 2009. Adapting the knowledge-capital model of themultinational enterprise to trade and investment in business services. TheWorld Economy 32(1), 6–29.

Markusen, J. R., 1995. The boundary of multinational enterprises and the the-ories of international trade. Journal of Economic Perspectives 9(2), 169 –189.

Markusen, J. R., 1997. Trade versus investment liberalization. NBER WorkingPaper No. 6231.

Markusen, J. R., 2002. Multinational Firms and the Theory of InternationalTrade. MIT Press, Cambridge.

Markusen, J. R., Venables, A. J., 1998. Multinational firms and the new tradetheory. Jounral of International Economics 46(2), 183–203.

28

Markusen, J. R., Venables, A. J., Konan, D. E., Zhang, K. H., 1996. A unifiedtreatment of horizontal direct investment, vertical direct investment, and thepattern of trade in goods and services. NBER Working Paper No. 5696.

Mayer, T., Zignago, S., 2006. Notes on cepii’s distances measures. CEPII, Paris.

Melitz, M. J., 2003. The impact of trade on intra-industry reallocations andaggregate industry productivity. Econometrica 71(6), 1695–1725.

Mirza, D., Nicoletti, G., 2004. What is so special about trade in services. Uni-versity of Nottingham Research Paper No. 2004/02.

Papke, L. E., Wooldridge, J. M., 1996. Econometric methods for fractionalresponse variables with an application to 401(k) plan participation rates.Journal of Applied Econometrics 11(6), 619–632.

Riedl, A., 2008. Contrasting the dynamic patterns of manufacturing and servicefdi: Evidence from transition economies. Department of Economics, ViennaUniversity of Economics B.A.. Working Paper No. 117.

Sampson, G. P., Snape, R. H., 1985. Identifying th issues in trade in services.The World Economy 8, 171–182.

Tinbergen, J., 1962. Shaping the World Economy: Suggestions for an Interna-tional Economic Policy. New York: Twentieth Century Fund.

29

A Appendix

Table 1: Sales through Affiliates

Sales of services

to foreign persons by U.S. MNCs through their foreign affiliates

All countries 1999 2005Total private services 353,207 528,481

Education n.a. n.a.Financial services 31,641 42,912Insurance services 52,855 94,438

Telecommunications n.a. 21,483Professional, scientific and technical services 63,898 95,412

Computer and information services 14,708 n.a.Management and consulting services n.a. 12,405

Research and development n.a. 3,600Architectural, engineering and other technical services 11,939 12,059

Legal services 821 2,402Advertising n.a. 10,080

to U.S. persons by foreign MNCs through their U.S. affiliates

All countries 1999 2005Total private services 293,485 389,030

Education 355 403Financial services 15,318 24,916Insurance services 78,479 77,168

Telecommunications 13,095 n.a.Professional, scientific and technical services 15,421 48,590

Computer and information services 4,022 8,815Management and consulting services 585 2,079

Research and development 658 882Architectural, engineering and other technical services 3,880 6,175

Legal services 21 n.a.Advertising 5,219 20,327

n.a. Not available.Professional, scientific and technical services are comparable to business, professional and technicalservices.All data are in millions of US Dollars.

30

Table 2: U.S. Cross-border Trade

Cross-border trade - Exports Cross-border trade - Imports

All countries 1999 2005 1999 2005

Total private services 265,106 367,813 183,034 281,607unaffiliated 203,081 272,724 147,137 219,688

affiliated 62,025 95,088 35,897 61,920

Education 9,616 14,076 1,808 3,962unaffiliated 9,616 14,076 1,808 3,962

affiliated n.a. n.a. n.a. n.a.

Financial services 17,410 7,787 9,418 12,620unaffiliated 13,410 7,787 3,418 6,720

affiliated 4,000 n.a. 6,000 5,900

Insurance services 3,053 7,787 9,389 28,540unaffiliated 3,053 7,787 9,389 28,540

affiliated n.a. n.a. n.a. n.a.

Telecommunications 4,549 5,231 6,602 4,527unaffiliated 4,549 5,231 6,602 4,527

affiliated n.a. n.a. n.a. n.a.

BPT 53,517 83,990 27,635 48,765unaffiliated 27,700 41,874 8,588 14,824

affiliated 25,817 42,116 19,047 33,941

Computer&information services 6,643 9,782 4,494 9,048unaffiliated 5,443 7,482 1,494 2,748

affiliated 1,200 2,300 3,000 6,300

Management&consulting services n.a. 6,864 n.a. 6,070unaffiliated 1,832 2,564 842 1,870

affiliated n.a. 4,300 n.a. 4,200

Research&development n.a. 10,191 n.a. 6,744unaffiliated 994 1,291 749 2,244

affiliated n.a. 8,900 n.a. 4,500

AET n.a. n.a. n.a. n.a.unaffiliated 2,620 3,430 19 181

affiliated n.a. n.a. n.a. n.a.

Legal services n.a. n.a. n.a. n.a.unaffiliated 2,465 4,274 742 897

affiliated n.a. n.a. n.a. n.a.

Advertising n.a. n.a. n.a. n.a.unaffiliated 481 574 881 982

affiliated n.a. n.a. n.a. n.a.

n.a. Not available.BPT - Business, professional and technical services.AET - Architectural, engineering and other technical services.All data are in millions of US Dollars.

31

Figure 1: U.S. International sales and purchases of services of Private Services,1987-2005

Table 3: Descriptive Statistics

Variable Obs Mean Std. Dev. Min MaxOutsh 559 0.7044478 0.2575423 0 0.9938525

Insh 538 0.4262909 0.3845714 0 0.9986651Manufacturing outsh 456 0.5316677 0.444042 0 0.9987003

Manufacturing insh 415 0.4685912 0.4496828 0 0.9997787Log distance 1590 8.577451 0.854291 6.306995 9.680893

Log GDP 1590 6.871722 0.8164167 5.522241 8.448654Language 1590 0.4 0.4900521 0 1

Similarity index 1590 0.6186601 0.3077173 0 0.9991627Restrictiveness Index 1300 0.1218231 0.1034626 0.011 0.56

Log outward manufacturing FDI 1270 9.864045 0.6884036 8.138272 11.33833Log inward manufacturing FDI 1446 9.993362 1.554714 4.624973 11.35117

Corporate tax rate 1558 35.35746 7.122973 21.3 57.5EU Dummy 1590 0.4 0.4900521 0 1

NAFTA Dummy 1590 0.2 0.4001258 0 1Log outward affiliate sales 647 6.192992 1.735329 0 10.10316

Log inward affiliate sales 492 5.183279 2.707609 0 10.31088Log exports 1407 5.291403 1.551603 1.098612 9.671997Log imports 1346 4.275433 1.874804 0 9.192685

Log total outward sales 559 6.770904 1.316524 3.78419 10.29008Log total inward sales 538 5.379661 2.322071 0 10.31454

32

!"##$

#$

"##$

%###$

%"##$

&###$

#$

&$

'$

($

)$

%#$

%&$

%'$

%($

%)$

&#$

k

!!

*+,-.$/0121$324

5$5/

,.60$$

72+4$.2891$.+$15.8:;,5041-.$

kx

km

*+,-.$/0121$324

5$

08<1$-+$7+21,=-$58;15$

Figure 2: Threshold levels for establishment and direct trade

33

!"##$

#$

"##$

%###$

%"##$

&###$

#$

&$

'$

($

)$

%#$

%&$

%'$

%($

%)$

&#$

k

!!

kx

km

k' m

k' x

Figure 3: Increasing size and the threshold levels for establishment and directtrade

34

!"##$

#$

"##$

%###$

%"##$

&###$

#$

&$

'$

($

)$

%#$

%&$

%'$

%($

%)$

&#$

k

!!

kx

km

k' m

k' x

Figure 4: Increasing distance and the threshold levels for establishment anddirect trade

35

Tab

le4:

Reg

ress

ion

resu

lts:

3SL

Sou

twar

dsa

les

(1)

(2)

(1)

(2)

(1)

(2)

(1)

(2)

(1)

(2)

exp

ort

saffi

liate

sale

sexp

ort

saffi

liate

sale

sexp

ort

saffi

liate

sale

sexp

ort

saffi

liate

sale

sexp

ort

saffi

liate

sale

s

Log

Dis

tance

Language

Rest

ricti

veness

Index

Log

affi

liate

sale

sL

og

FD

IM

anufa

ctu

ring

Educati

on

-0.2

83***

0.1

30

0.0

45

0.0

97

7.5

86***

-14.4

63***

0.3

29***

0.1

25

(-3.5

04)

(0.4

70)

(0.2

80)

(0.2

57)

(4.8

75)

(-7.5

80)

(3.4

85)

(0.3

78)

Tele

com

munic

ati

on

-0.3

51***

0.8

42***

0.7

21***

0.2

17

0.6

45

-1.6

55**

0.2

73***

1.1

20***

(-5.4

95)

(5.3

58)

(3.2

52)

(0.6

20)

(1.1

47)

(-2.2

00)

(3.1

45)

(3.7

99)

Insu

rance

-0.3

48***

-0.1

45

1.0

13***

1.5

80***

0.1

31

-5.7

97***

0.2

18*

-0.0

44

(-5.2

73)

(-0.8

85)

(3.4

25)

(6.0

57)

(0.1

20)

(-5.6

78)

(1.6

82)

(-0.1

93)

Fin

ancia

lse

rvic

es

-0.3

07***

0.2

45**

-0.0

01

1.7

90***

-1.0

38

-4.9

98***

0.3

32***

-0.1

65

(-5.3

47)

(1.9

63)

(-0.0

0358)

(7.2

06)

(-0.9

64)

(-3.9

04)

(3.3

48)

(-0.7

38)

Com

pute

r&In

form

ati

on

-0.1

37

0.2

48

-0.2

50

0.1

79

4.6

46

-5.0

65

0.3

64**

0.2

81

(-1.3

40)

(1.0

23)

(-0.5

91)

(0.2

79)

(1.5

83)

(-1.0

24)

(2.2

83)

(0.6

78)

Com

pute

r-0

.011

-0.4

15

1.1

89

-1.0

40

1.0

51

-10.0

32

0.4

44**

0.3

49

(-0.0

117)

(-0.2

82)

(1.2

28)

(-0.7

80)

(0.1

07)

(-0.7

24)

(2.3

02)

(0.3

25)

Info

rmati

on

-0.7

57

0.2

18

0.1

70

2.2

20

-4.2

28

-5.8

70

0.3

50

-1.2

04

(-0.9

23)

(0.1

44)

(0.1

46)

(1.5

24)

(-0.7

13)

(-0.4

25)

(1.0

79)

(-0.9

98)

Op

era

tional

leasi

ng

-0.5

02**

-0.1

62

-0.0

79

-0.1

41

-0.0

69

-1.1

98

0.2

34

0.1

92

(-2.3

50)

(-0.3

93)

(-0.3

27)

(-0.3

98)

(-0.0

330)

(-0.4

00)

(0.8

80)

(0.3

21)

Busi

ness

serv

ices

-0.3

53***

0.4

54**

0.1

55

-0.1

66

0.5

45

-1.2

52

0.4

97***

0.4

07*

(-3.3

72)

(2.2

11)

(0.7

94)

(-0.5

35)

(0.3

56)

(-0.5

87)

(3.3

15)

(1.7

67)

Legal

serv

ices

-0.4

02***

1.1

68***

-0.0

00

-0.7

12***

-1.7

91**

0.3

71

0.1

26**

1.7

06***

(-6.7

67)

(6.7

27)

(-0.0

00304)

(-2.9

18)

(-2.3

94)

(0.3

54)

(1.9

81)

(6.4

79)

Consu

ltin

g-0

.253***

0.0

26

0.5

31***

0.1

77

-3.0

42***

-1.5

50

0.4

41***

0.0

20

(-4.4

28)

(0.2

24)

(4.1

95)

(0.7

68)

(-3.2

93)

(-0.9

79)

(5.0

97)

(0.1

10)

Advert

isin

g-0

.647***

0.3

77**

-0.1

99

-0.0

05

4.6

64***

-6.7

49***

0.4

31***

0.5

41**

(-9.7

65)

(2.0

48)

(-1.4

36)

(-0.0

210)

(3.4

51)

(-4.3

53)

(3.7

59)

(2.2

97)

R&

D-0

.546***

0.3

85*

-0.6

51***

1.1

18***

3.6

81***

-6.2

53***

0.6

89***

-0.2

34

(-7.3

58)

(1.7

30)

(-3.9

45)

(3.5

09)

(3.3

68)

(-3.2

62)

(8.2

97)

(-0.7

18)

Log

GD

PSim

ilari

tyIn

dex

Corp

ora

teta

xL

og

Exp

ort

sO

bse

rvati

ons

Educati

on

1.0

11***

-1.2

62***

0.6

26

0.3

79

0.0

32*

-0.0

07

1.1

36***

554

554

(7.8

13)

(-2.9

04)

(1.1

23)

(0.5

21)

(1.8

12)

(-0.2

56)

(3.4

27)

R-s

quare

dT

ele

com

munic

ati

on

0.3

46***

0.0

31

0.2

30

-1.0

38***

0.0

01

-0.0

34

0.3

56*

0.9

96

0.9

96

(2.7

97)

(0.1

80)

(0.8

58)

(-2.9

96)

(0.0

373)

(-1.6

01)

(1.7

25)

Insu

rance

1.0

09***

1.3

28***

0.9

11***

-1.7

96***

-0.0

48***

0.0

48***

0.3

04**

(4.2

13)

(6.4

04)

(2.8

31)

(-5.7

99)

(-4.0

83)

(3.0

58)

(2.4

26)

Fin

ancia

lse

rvic

es

0.3

71***

0.2

95*

0.1

78

-1.8

26***

-0.0

57***

0.0

43**

0.8

73***

(3.3

46)

(1.7

16)

(0.4

48)

(-4.0

72)

(-5.4

69)

(2.2

82)

(3.5

38)

Com

pute

r&In

form

ati

on

0.0

22

-0.0

52

0.5

76**

-0.2

95

0.0

58*

-0.0

80*

1.3

44***

(0.1

40)

(-0.2

56)

(2.0

41)

(-0.5

79)

(1.8

95)

(-1.7

78)

(3.0

03)

Com

pute

r0.3

21

0.1

93

2.6

18*

-5.6

46***

-0.0

06

-0.0

14

1.0

65**

(0.5

13)

(0.2

34)

(1.6

77)

(-3.6

01)

(-0.1

25)

(-0.2

51)

(2.2

22)

Info

rmati

on

0.5

20

-0.0

55

1.3

12

-1.7

56

-0.0

22

0.0

31

0.5

72

(1.2

25)

(-0.0

648)

(1.1

10)

(-0.8

72)

(-0.4

64)

(0.5

25)

(0.7

35)

Op

era

tional

leasi

ng

0.4

27***

-0.2

04

-0.7

41

2.0

80***

-0.1

40***

-0.0

18

0.2

90

(2.8

18)

(-0.9

44)

(-1.0

71)

(3.4

70)

(-4.3

10)

(-0.3

42)

(1.1

58)

Busi

ness

serv

ices

0.3

52***

-0.1

62

-0.8

71***

1.7

66***

-0.0

32***

0.0

21

0.9

92***

(2.5

91)

(-0.7

27)

(-2.7

38)

(6.8

61)

(-2.5

99)

(1.1

39)

(3.8

36)

Legal

serv

ices

0.5

71***

1.2

79***

1.2

20*

-4.5

63***

-0.0

30**

-0.0

91***

0.1

36

(3.0

80)

(4.3

90)

(1.9

01)

(-6.6

77)

(-2.1

53)

(-4.9

75)

(0.5

69)

Consu

ltin

g-0

.030

0.2

72***

0.4

73***

0.5

96***

0.0

36***

-0.0

56***

0.5

33***

(-0.3

72)

(2.7

20)

(2.6

10)

(2.7

59)

(3.5

52)

(-4.7

52)

(4.9

81)

Advert

isin

g0.7

62***

-0.1

05

0.6

29

-2.2

20***

-0.0

34***

-0.0

33**

0.5

45***

(7.3

03)

(-0.5

93)

(1.1

45)

(-3.8

67)

(-2.7

57)

(-2.2

75)

(4.6

38)

R&

D0.2

20*

0.2

23

-0.7

75

1.5

66**

-0.0

32**

0.0

17

0.7

95***

(1.9

53)

(1.0

18)

(-1.2

71)

(2.1

47)

(-2.4

59)

(1.0

07)

(5.0

58)

Est

imati

ons

base

don

equati

on

(14).

Dep

endent

vari

able

sare

log

of

exp

ort

sand

log

of

outw

ard

affi

liate

sale

s.E

stim

ati

on

meth

od:

3SL

S.

zst

ati

stic

sin

pare

nth

ese

s.Secto

rdum

my

vari

able

sare

inclu

ded

inth

ere

gre

ssio

n,

but

are

not

rep

ort

ed

inth

eta

ble

.*,

**

and

***

indic

ate

stati

stic

al

signifi

cance

at

the

10-p

erc

ent

level,

5-p

erc

ent

level,

and

1-p

erc

ent

level

resp

ecti

vely

.

36

Tab

le5:

Reg

ress

ion

resu

lts:

3SL

Sin

war

dsa

les

(1)

(2)

(1)

(2)

(1)

(2)

(1)

(2)

(1)

(2)

imp

ort

saffi

liate

sale

sim

port

saffi

liate

sale

sim

port

saffi

liate

sale

sim

port

saffi

liate

sale

sim

port

saffi

liate

sale

s

Log

Dis

tance

Language

Rest

ricti

veness

Index

Log

affi

liate

sale

sL

og

FD

IM

anufa

ctu

ring

Educati

on

1.7

95***

1.5

65**

0.2

43

0.4

32

9.2

99

11.9

87

-1.1

71***

-0.0

82

(3.0

25)

(1.9

89)

(0.4

82)

(0.4

45)

(1.2

75)

(1.1

53)

(-3.8

97)

(-0.0

870)

Tele

com

munic

ati

on

-0.4

65***

-1.1

70***

1.2

55***

2.2

02***

-1.2

79*

-6.6

11***

-0.2

44***

-0.3

85

(-3.9

21)

(-3.2

67)

(5.0

42)

(4.0

23)

(-1.7

68)

(-3.8

89)

(-2.6

83)

(-1.0

33)

Insu

rance

1.1

13***

0.1

04

1.6

12***

-0.1

77

18.6

37***

1.7

68

0.7

78***

0.4

55

(5.4

22)

(0.2

23)

(6.9

50)

(-0.3

20)

(7.7

33)

(0.2

92)

(8.2

35)

(1.1

64)

Fin

ancia

lse

rvic

es

0.4

21

1.9

21***

1.5

77***

1.4

66

-0.2

75

21.4

73***

0.1

36

0.2

17

(1.3

64)

(3.1

90)

(5.2

93)

(1.4

75)

(-0.0

851)

(3.4

81)

(1.2

91)

(0.4

57)

Com

pute

r&In

form

ati

on

-2.6

41***

-1.8

10

-1.9

36**

4.6

75

13.2

31

43.3

20**

-0.0

69

1.1

70

(-8.4

46)

(-0.6

87)

(-2.0

67)

(1.5

80)

(1.5

44)

(2.5

74)

(-0.4

84)

(0.7

81)

Com

pute

r-2

9.3

90**

16.4

99

-25.1

45*

16.9

49

-414.9

37*

235.6

47

0.0

96

0.6

54

(-2.0

22)

(0.2

32)

(-1.7

37)

(0.2

73)

(-1.9

40)

(0.2

33)

(0.1

11)

(0.4

84)

Info

rmati

on

7.1

75**

16.2

35***

2.5

86*

6.8

00***

-63.9

25**

-121.5

11**

-0.0

75

0.3

66

(1.9

94)

(2.9

45)

(1.8

54)

(3.8

58)

(-2.3

69)

(-2.5

73)

(-0.3

90)

(0.4

47)

Op

era

tional

leasi

ng

2.6

38***

4.9

46**

5.2

09***

9.3

38***

25.7

44***

30.6

80

-0.4

91**

-2.2

06

(3.7

96)

(2.0

27)

(3.2

06)

(4.6

86)

(4.3

98)

(1.3

48)

(-2.1

02)

(-0.3

44)

Busi

ness

serv

ices

-0.6

09***

-1.5

07***

1.1

61***

2.6

03***

-2.0

63**

-5.3

31**

-0.0

21

1.1

51***

(-9.1

94)

(-5.3

12)

(9.6

04)

(3.7

83)

(-2.2

66)

(-2.1

37)

(-0.5

84)

(4.5

44)

Legal

serv

ices

-0.5

46

-9.9

50

0.0

00

-198.5

99

0.0

00

0.0

00

0.3

12

11.8

93

(-0.3

36)

(-0.2

34)

()(-

0.3

30)

()()

(0.2

03)

(0.2

62)

Consu

ltin

g-0

.627***

-0.0

26

1.0

99***

0.6

40

-6.5

33***

-11.6

62**

-0.0

37

0.5

29

(-5.0

39)

(-0.0

737)

(6.2

12)

(1.1

65)

(-3.9

46)

(-2.4

37)

(-0.6

56)

(1.2

77)

Advert

isin

g5.2

16**

-16.2

72***

7.4

54***

-20.4

93***

77.3

77**

-245.7

26***

0.2

07*

-0.1

05

(2.5

72)

(-5.4

74)

(3.0

24)

(-5.0

80)

(2.5

00)

(-5.0

89)

(1.6

46)

(-0.1

65)

R&

D-1

.484***

1.5

28**

0.3

22

2.5

28***

-16.6

04***

29.2

23***

0.0

26

0.1

28

(-7.4

85)

(2.3

77)

(1.3

13)

(5.5

98)

(-7.0

53)

(4.8

79)

(0.5

22)

(0.2

33)

Log

GD

PSim

ilari

tyIn

dex

Corp

ora

teta

xL

og

Imp

ort

sO

bse

rvati

ons

Educati

on

1.3

61*

1.5

37

-1.4

73

-2.4

08

-0.0

25

0.0

07

-0.5

71

396

396

(1.6

55)

(1.3

99)

(-0.7

17)

(-0.6

37)

(-0.9

65)

(0.1

25)

(-1.2

62)

R-s

quare

dT

ele

com

munic

ati

on

0.9

22***

2.4

27***

-0.9

71

-2.0

63

0.0

38*

0.0

80

-0.9

60***

0.9

95

0.9

84

(4.3

61)

(6.9

02)

(-1.2

96)

(-1.2

73)

(1.7

78)

(1.5

15)

(-2.5

96)

Insu

rance

1.0

61***

-1.0

79***

3.0

84***

-2.9

07**

-0.0

46***

0.0

28

0.3

42**

(7.6

27)

(-4.4

71)

(4.6

48)

(-2.2

22)

(-4.0

81)

(1.1

59)

(2.2

77)

Fin

ancia

lse

rvic

es

0.4

35***

-0.9

15***

-0.0

29

-0.2

86

-0.0

36***

0.0

51**

0.2

65

(3.1

23)

(-3.4

15)

(-0.0

531)

(-0.2

52)

(-3.0

86)

(1.9

87)

(0.5

41)

Com

pute

r&In

form

ati

on

2.5

15

9.8

50**

1.7

78

-11.7

76

-0.4

22***

-0.5

18

-0.5

61

(1.2

51)

(2.0

61)

(0.4

99)

(-1.4

92)

(-3.4

11)

(-0.9

89)

(-0.6

62)

Com

pute

r7.1

15**

-2.3

07

-7.9

38***

3.3

43

0.3

50**

-0.1

27

0.0

31

(2.4

62)

(-0.1

34)

(-2.6

07)

(0.1

75)

(2.0

04)

(-0.1

49)

(0.0

138)

Info

rmati

on

-0.3

45

-0.9

50

-1.4

54

3.9

28

0.0

06

-0.0

78

-0.1

71

(-1.0

28)

(-1.2

44)

(-0.8

33)

(0.9

97)

(0.1

93)

(-1.4

93)

(-0.2

85)

Op

era

tional

leasi

ng

1.3

08***

1.3

72

-0.8

03

-0.3

87

0.2

01***

0.3

35***

-1.4

47

(3.0

07)

(0.8

84)

(-0.4

97)

(-0.1

10)

(2.6

16)

(2.6

10)

(-1.4

08)

Busi

ness

serv

ices

0.7

26***

1.8

69***

0.6

98

1.1

95

-0.0

20**

-0.0

64***

-1.7

68***

(8.9

56)

(5.3

58)

(1.5

73)

(1.2

28)

(-2.4

17)

(-3.3

04)

(-3.6

99)

Legal

serv

ices

2.4

70

43.1

88

-2.6

26

-81.1

19

-0.0

65

-0.8

46

-14.8

90

(0.1

85)

(0.2

80)

(-0.1

69)

(-0.2

74)

(-0.0

832)

(-0.1

75)

(-0.2

46)

Consu

ltin

g0.6

13***

-0.1

41

0.0

46

0.3

29

-0.0

43***

-0.0

85***

-0.2

59

(4.3

13)

(-0.4

03)

(0.0

722)

(0.2

46)

(-3.7

36)

(-3.1

07)

(-0.9

29)

Advert

isin

g1.1

27***

-0.9

56

-0.3

59

1.0

76

0.0

45*

-0.1

30***

0.8

66

(4.2

47)

(-1.0

81)

(-0.3

07)

(0.3

78)

(1.8

10)

(-2.8

61)

(1.6

17)

R&

D0.9

05***

0.4

49

1.9

93***

0.1

71

-0.0

72***

0.0

26

0.1

23

(7.4

45)

(1.2

01)

(3.4

82)

(0.1

22)

(-6.7

98)

(0.8

85)

(0.4

26)

Est

imati

ons

base

don

equati

on

(14).

Dep

endent

vari

able

islo

gof

imp

ort

sand

log

of

inw

ard

affi

liate

sale

s.E

stim

ati

on

meth

od:

3SL

S.

zst

ati

stic

sin

pare

nth

ese

s.Secto

rdum

my

vari

able

sare

inclu

ded

inth

ere

gre

ssio

n,

but

are

not

rep

ort

ed

inth

eta

ble

.*,

**

and

***

indic

ate

stati

stic

al

signifi

cance

at

the

10-p

erc

ent

level,

5-p

erc

ent

level,

and

1-p

erc

ent

level

resp

ecti

vely

.

37

Tab

le6:

Reg

ress

ion

resu

lts:

outw

ard

affilia

tesa

les

shar

e

Log

Dis

tance

Language

Rest

ricti

veness

Index

Log

FD

IM

anufa

ctu

ring

Educati

on

0.3

56*

-0.1

89

-10.2

46***

0.5

60

(1.8

22)

(-0.4

01)

(-3.6

07)

(1.5

54)

Tele

com

munic

ati

on

0.7

57***

0.0

27

-2.1

55*

0.5

85

(3.9

97)

(0.0

461)

(-1.7

91)

(1.5

80)

Insu

rance

0.4

01**

0.4

43*

-3.8

21***

0.4

91*

(2.2

19)

(1.9

56)

(-3.9

27)

(1.7

15)

Fin

ancia

lse

rvic

es

0.2

21*

1.7

73***

-5.0

64***

-0.2

58

(1.6

50)

(9.3

51)

(-5.7

92)

(-1.0

92)

Com

pute

r&In

form

ati

on

0.2

29

0.0

14

-3.0

24

0.3

83

(1.1

44)

(0.0

262)

(-0.7

41)

(1.0

25)

Com

pute

r-0

.590

-0.6

96

-12.5

77

0.0

51

(-0.4

65)

(-0.7

29)

(-1.0

30)

(0.0

746)

Info

rmati

on

0.6

56

1.6

68***

-1.8

08

-0.8

46

(0.9

46)

(4.4

52)

(-0.2

78)

(-1.3

94)

Op

era

tional

leasi

ng

-0.2

25

0.2

96

-2.9

02

-0.6

66*

(-0.8

18)

(1.2

64)

(-1.5

55)

(-1.7

25)

Busi

ness

serv

ices

0.4

91***

-0.2

41

-0.8

93

0.4

88***

(4.7

01)

(-1.4

59)

(-0.7

31)

(3.7

61)

Legal

serv

ices

1.4

95***

-0.6

43**

5.6

88***

2.0

06***

(6.0

23)

(-2.4

55)

(4.0

83)

(4.8

95)

Consu

ltin

g0.2

42**

-0.2

07

1.5

02

0.1

61

(2.0

22)

(-0.9

06)

(0.8

96)

(0.7

57)

Advert

isin

g0.9

37***

-0.3

06

-4.9

27***

1.0

54***

(8.4

78)

(-1.1

21)

(-3.5

40)

(4.8

65)

R&

D0.6

66***

0.6

75**

-3.6

48**

0.2

62

(3.8

89)

(2.3

49)

(-2.2

94)

(1.2

17)

Log

GD

PSim

ilari

tyIn

dex

Corp

ora

teta

xO

bse

rvati

ons

Educati

on

-1.1

92***

1.1

15

-0.0

07

559

(-8.1

29)

(1.2

92)

(-0.2

48)

Tele

com

munic

ati

on

-0.2

12

-0.2

68

-0.0

17

(-1.1

69)

(-0.3

07)

(-0.5

08)

Insu

rance

0.1

28

-2.1

68***

0.0

80***

(0.9

21)

(-5.9

51)

(3.8

66)

Fin

ancia

lse

rvic

es

0.1

97

-1.9

12***

0.0

47**

(1.4

86)

(-5.5

30)

(2.5

16)

Com

pute

r&In

form

ati

on

-0.1

37

0.0

41

-0.0

54*

(-0.8

12)

(0.2

16)

(-1.9

22)

Com

pute

r0.2

42

-5.9

04***

-0.0

08

(0.3

66)

(-5.8

76)

(-0.1

72)

Info

rmati

on

-0.3

64

-1.8

13

0.0

38

(-0.9

85)

(-1.5

38)

(1.3

81)

Op

era

tional

leasi

ng

-0.3

68**

2.6

34***

0.0

56

(-2.3

04)

(6.1

18)

(1.5

67)

Busi

ness

serv

ices

-0.1

97***

1.7

93***

0.0

22**

(-2.8

42)

(10.3

5)

(2.3

48)

Legal

serv

ices

1.0

85***

-5.7

98***

-0.0

76***

(4.5

18)

(-6.7

32)

(-3.4

44)

Consu

ltin

g0.1

84

0.2

32

-0.0

56***

(1.3

16)

(1.0

02)

(-3.4

85)

Advert

isin

g-0

.654***

-1.6

71***

-0.0

13

(-4.5

27)

(-2.8

07)

(-1.1

68)

R&

D-0

.074

1.5

16**

0.0

20

(-0.7

14)

(2.1

94)

(0.9

95)

Est

imati

ons

base

don

equati

on

(15).

Dep

endent

vari

able

issh

are

of

outw

ard

affi

liate

sale

sin

tota

loutw

ard

sale

s.E

stim

ati

on

meth

od:

GL

Mm

odel.

Robust

zst

ati

stic

sin

pare

nth

ese

s.Secto

rdum

my

vari

able

sare

inclu

ded

inth

ere

gre

ssio

n,

but

are

not

rep

ort

ed

inth

eta

ble

.*,

**

and

***

indic

ate

stati

stic

al

signifi

cance

at

the

10-p

erc

ent

level,

5-p

erc

ent

level,

and

1-p

erc

ent

level

resp

ecti

vely

.

38

Tab

le7:

Reg

ress

ion

resu

lts:

SUR

outw

ard

sale

s

(1)

(2)

(1)

(2)

(1)

(2)

(1)

(2)

tota

lsa

les

affi

liate

sale

sto

tal

sale

saffi

liate

sale

sto

tal

sale

saffi

liate

sale

sto

tal

sale

saffi

liate

sale

s

Log

Dis

tance

Language

Rest

ricti

veness

Index

Log

FD

IM

anufa

ctu

ring

Educati

on

-0.5

13***

0.0

99

0.5

55**

0.0

75

-1.3

99

-13.6

75***

-0.3

07

0.1

79

(-4.3

43)

(0.5

63)

(2.3

78)

(0.2

14)

(-0.9

85)

(-6.4

37)

(-1.4

93)

(0.5

81)

Tele

com

munic

ati

on

0.5

62***

0.8

16***

0.5

79**

-0.1

34

-1.1

58**

-2.1

90**

1.1

22***

0.6

82**

(4.5

05)

(4.3

73)

(2.2

56)

(-0.3

49)

(-1.9

83)

(-2.5

08)

(5.2

04)

(2.1

16)

Insu

rance

-0.2

90**

0.2

58

1.9

73***

0.5

17**

-5.8

26***

-4.5

99***

-0.0

69

0.1

70

(-2.4

31)

(1.4

48)

(13.9

5)

(2.4

46)

(-7.4

11)

(-3.9

14)

(-0.3

82)

(0.6

26)

Fin

ancia

lse

rvic

es

0.0

15

0.2

02

1.7

87***

1.7

78***

-5.9

42***

-5.0

42***

0.0

83

-0.3

21

(0.1

54)

(1.3

77)

(11.0

6)

(7.3

67)

(-6.5

81)

(-3.7

36)

(0.5

14)

(-1.3

23)

Com

pute

r&In

form

ati

on

0.0

31

0.2

10

-0.2

46

0.0

65

2.5

55

-3.2

50

0.3

84

0.3

38

(0.1

64)

(0.7

50)

(-0.4

96)

(0.0

880)

(0.7

36)

(-0.6

26)

(1.1

61)

(0.6

85)

Com

pute

r-0

.014

-0.5

33

0.4

00

-0.9

33

-7.5

80

-11.3

47

1.3

53*

0.2

72

(-0.0

119)

(-0.3

11)

(0.4

28)

(-0.6

68)

(-0.6

94)

(-0.6

95)

(1.8

32)

(0.2

47)

Info

rmati

on

-0.5

42

0.6

89

2.4

37***

1.6

75

-10.3

54

-1.4

29

-1.3

29

-0.8

54

(-0.5

28)

(0.4

49)

(2.8

50)

(1.3

10)

(-1.1

14)

(-0.1

03)

(-1.5

51)

(-0.6

67)

Op

era

tional

leasi

ng

-0.2

15

-0.1

64

-0.2

49

0.1

56

-0.3

58

-2.8

29

0.3

97

-0.5

19

(-0.6

58)

(-0.3

36)

(-0.9

20)

(0.3

84)

(-0.1

53)

(-0.8

09)

(0.8

67)

(-0.7

58)

Busi

ness

serv

ices

-0.1

01

0.4

94**

0.4

50**

-0.2

30

-3.3

03**

-0.8

02

0.2

60

0.4

82*

(-0.7

66)

(2.5

12)

(2.0

24)

(-0.6

93)

(-2.0

76)

(-0.3

37)

(1.4

12)

(1.7

50)

Legal

serv

ices

-0.1

30

1.5

12***

-0.2

68

-0.8

00***

-1.3

78*

2.0

96*

0.4

20**

1.7

55***

(-1.0

91)

(8.4

71)

(-1.3

88)

(-2.7

78)

(-1.8

76)

(1.9

09)

(2.0

00)

(5.5

89)

Consu

ltin

g-0

.200**

0.2

46*

0.6

75***

-0.3

22

-4.6

22***

1.5

74

-0.0

51

0.1

70

(-2.3

12)

(1.8

96)

(4.1

33)

(-1.3

19)

(-4.0

38)

(0.9

20)

(-0.3

44)

(0.7

71)

Advert

isin

g-0

.188*

1.0

36***

0.1

53

-0.5

35*

-6.8

64***

-3.9

72**

0.3

20*

1.1

87***

(-1.9

17)

(7.0

66)

(0.8

04)

(-1.8

87)

(-5.7

56)

(-2.2

29)

(1.8

52)

(4.5

93)

R&

D-0

.716***

0.6

97***

1.7

02***

0.7

31**

-9.0

30***

-4.1

05**

-1.2

62***

0.2

49

(-6.5

27)

(4.2

52)

(8.2

10)

(2.3

58)

(-6.9

97)

(-2.1

28)

(-6.7

58)

(0.8

92)

Log

GD

PSim

ilari

tyIn

dex

Corp

ora

teta

xO

bse

rvati

ons

Educati

on

0.6

59***

-1.1

48***

1.3

53**

0.5

82

0.0

58***

-0.0

00

554

554

(4.6

13)

(-5.3

74)

(2.4

64)

(0.7

10)

(3.5

23)

(-0.0

0305)

R-s

quare

dT

ele

com

munic

ati

on

0.1

98*

-0.1

89

-1.0

90***

-0.9

04**

-0.0

30*

-0.0

22

0.9

98

0.9

38

(1.6

53)

(-1.0

55)

(-4.0

34)

(-2.2

39)

(-1.8

47)

(-0.8

98)

Insu

rance

1.7

34***

0.3

23***

-1.4

56***

-2.2

21***

0.0

30***

0.0

82***

(20.9

9)

(2.6

16)

(-6.1

96)

(-6.3

24)

(2.6

24)

(4.7

70)

Fin

ancia

lse

rvic

es

0.7

41***

0.2

49**

-1.4

57***

-1.8

88***

-0.0

27***

0.0

49***

(9.0

32)

(2.0

33)

(-4.1

71)

(-3.6

15)

(-2.6

58)

(3.1

85)

Com

pute

r&In

form

ati

on

-0.0

01

-0.0

48

0.9

36***

0.0

20

0.0

08

-0.0

61

(-0.0

0530)

(-0.2

02)

(3.9

18)

(0.0

561)

(0.2

73)

(-1.3

92)

Com

pute

r0.7

09

0.2

83

-3.2

91***

-5.7

12***

-0.0

30

-0.0

16

(1.1

26)

(0.3

00)

(-2.7

79)

(-3.2

28)

(-0.6

90)

(-0.2

53)

Info

rmati

on

0.4

40

-0.3

70

-0.8

00

-1.9

76

0.0

11

0.0

38

(0.8

55)

(-0.4

81)

(-0.5

21)

(-0.8

61)

(0.2

31)

(0.5

62)

Op

era

tional

leasi

ng

-0.0

44

-0.4

00*

1.2

84***

2.6

20***

-0.0

71**

0.0

60

(-0.2

81)

(-1.6

91)

(2.8

17)

(3.8

45)

(-2.1

36)

(1.1

97)

Busi

ness

serv

ices

0.5

41***

-0.1

90

0.9

96***

1.7

81***

-0.0

24**

0.0

20

(4.8

55)

(-1.1

42)

(4.9

63)

(5.9

37)

(-2.0

48)

(1.1

38)

Legal

serv

ices

0.8

73***

0.4

89**

-0.0

39

-4.9

02***

-0.0

63***

-0.0

51***

(5.4

26)

(2.0

32)

(-0.0

734)

(-6.2

07)

(-5.4

88)

(-2.9

51)

Consu

ltin

g0.3

20***

0.1

88

1.0

84***

0.1

76

-0.0

41***

-0.0

65***

(4.1

48)

(1.6

33)

(7.1

97)

(0.7

84)

(-4.4

50)

(-4.7

67)

Advert

isin

g0.5

00***

-0.6

96***

-2.3

82***

-1.8

54***

-0.0

62***

-0.0

10

(5.7

26)

(-5.3

31)

(-5.3

59)

(-2.7

91)

(-6.2

08)

(-0.6

97)

R&

D1.0

79***

-0.0

55

1.5

62***

1.4

02*

-0.0

14

0.0

18

(10.4

4)

(-0.3

56)

(2.7

77)

(1.6

68)

(-1.0

73)

(0.9

24)

Est

imati

ons

base

don

equati

on

(15).

Dep

endent

vari

able

sare

log

of

tota

loutw

ard

sale

sand

share

of

outw

ard

affi

liate

sale

s.E

stim

ati

on

meth

od:

SU

R.

zst

ati

stic

sin

pare

nth

ese

s.Secto

rdum

my

vari

able

sare

inclu

ded

inth

ere

gre

ssio

n,

but

are

not

rep

ort

ed

inth

eta

ble

.*,

**

and

***

indic

ate

stati

stic

al

signifi

cance

at

the

10-p

erc

ent

level,

5-p

erc

ent

level,

and

1-p

erc

ent

level

resp

ecti

vely

.

39

Tab

le8:

Reg

ress

ion

resu

lts:

inw

ard

affilia

tesa

les

shar

e

Log

Dis

tance

Language

Rest

ricti

veness

Index

Log

FD

IM

anufa

ctu

ring

Educati

on

0.0

69

0.8

93

28.8

59***

0.5

68

(0.1

56)

(1.3

15)

(3.6

48)

(0.8

60)

Tele

com

munic

ati

on

-0.5

36

0.4

63

-6.0

54**

-0.0

96

(-1.1

73)

(1.0

92)

(-2.3

36)

(-0.1

47)

Insu

rance

-0.6

35**

-1.1

73***

-14.4

30***

-0.2

72

(-2.0

88)

(-4.4

39)

(-3.9

20)

(-0.7

37)

Fin

ancia

lse

rvic

es

1.6

90***

0.4

76

19.0

57***

-0.6

83***

(4.3

83)

(1.3

89)

(4.6

12)

(-2.6

33)

Com

pute

r&In

form

ati

on

-0.1

89

4.7

20***

40.7

11**

-1.0

68

(-0.2

60)

(3.9

83)

(2.4

89)

(-1.4

62)

Com

pute

r-0

.780***

-0.8

16**

24.3

52***

0.2

14

(-3.1

48)

(-2.5

34)

(3.3

11)

(0.6

01)

Info

rmati

on

7.5

13***

5.1

04***

-43.8

45**

0.9

80

(3.5

45)

(7.6

50)

(-2.3

74)

(1.2

82)

Op

era

tional

leasi

ng

0.5

24

5.4

39***

9.7

13

6.0

36**

(0.5

83)

(10.9

5)

(1.3

85)

(2.0

63)

Busi

ness

serv

ices

-0.0

07

-0.2

29

-4.5

87

0.6

57**

(-0.0

364)

(-0.5

40)

(-1.3

96)

(2.0

36)

Legal

serv

ices

-0.0

84

28.2

21**

18.9

51

2.2

38

(-0.0

522)

(2.5

04)

(0.9

87)

(1.4

92)

Consu

ltin

g0.4

99

-0.0

62

10.2

20*

1.5

27***

(1.4

13)

(-0.1

03)

(1.9

36)

(2.5

91)

Advert

isin

g-0

.084

2.6

76***

13.4

20

0.8

29

(-0.1

81)

(3.5

51)

(1.4

33)

(1.2

73)

R&

D2.3

27***

2.0

57***

33.7

44***

-0.3

73

(4.1

11)

(5.3

59)

(5.3

84)

(-0.2

79)

Log

GD

PSim

ilari

tyIn

dex

Corp

ora

teta

xO

bse

rvati

ons

Educati

on

5.0

48***

-7.7

00**

0.1

56***

538

(4.6

70)

(-2.1

22)

(4.6

37)

Tele

com

munic

ati

on

1.1

36***

-3.2

30

0.0

69

(3.0

24)

(-1.3

06)

(0.8

26)

Insu

rance

-1.2

66***

-4.7

39***

0.0

57***

(-7.5

90)

(-4.5

60)

(4.2

64)

Fin

ancia

lse

rvic

es

-1.1

28***

-0.4

21

0.0

69***

(-5.0

72)

(-0.5

85)

(4.2

41)

Com

pute

r&In

form

ati

on

13.5

82***

-25.7

70***

-0.5

89**

(3.6

90)

(-3.4

10)

(-2.2

11)

Com

pute

r6.1

13***

2.8

89*

-0.3

31***

(8.7

75)

(1.8

44)

(-8.9

92)

Info

rmati

on

-1.2

75**

10.8

78**

-0.0

46

(-2.1

37)

(2.4

64)

(-1.1

08)

Op

era

tional

leasi

ng

0.6

87

-0.5

59

0.2

15***

(0.8

51)

(-0.5

94)

(3.7

88)

Busi

ness

serv

ices

-0.2

18

1.6

40

-0.0

05

(-0.8

42)

(1.2

01)

(-0.2

12)

Legal

serv

ices

10.1

30

-14.2

21**

0.3

34

(1.1

85)

(-2.2

42)

(0.6

92)

Consu

ltin

g0.3

67

0.7

86

-0.0

37

(0.8

00)

(0.3

86)

(-0.9

34)

Advert

isin

g1.3

74***

-3.1

15

-0.1

39***

(2.6

95)

(-0.7

88)

(-2.8

80)

R&

D-0

.409

0.6

71

0.0

76***

(-0.7

42)

(0.5

49)

(2.8

25)

Est

imati

ons

base

don

equati

on

(16).

Dep

endent

vari

able

issh

are

of

inw

ard

affi

liate

sale

sin

tota

lin

ward

sale

s.E

stim

ati

on

meth

od:

GL

Mm

odel.

Robust

zst

ati

stic

sin

pare

nth

ese

s.Secto

rdum

my

vari

able

sare

inclu

ded

inth

ere

gre

ssio

n,

but

are

not

rep

ort

ed

inth

eta

ble

.*,

**

and

***

indic

ate

stati

stic

al

signifi

cance

at

the

10-p

erc

ent

level,

5-p

erc

ent

level,

and

1-p

erc

ent

level

resp

ecti

vely

.

40

Tab

le9:

Reg

ress

ion

resu

lts:

SUR

inw

ard

sale

s

(1)

(2)

(1)

(2)

(1)

(2)

(1)

(2)

tota

lsa

les

affi

liate

sale

sto

tal

sale

saffi

liate

sale

sto

tal

sale

saffi

liate

sale

sto

tal

sale

saffi

liate

sale

s

Log

Dis

tance

Language

Rest

ricti

veness

Index

Log

FD

IM

anufa

ctu

ring

Educati

on

0.8

75

1.7

74*

0.5

36

1.6

85

0.0

57

34.1

31***

0.0

32

-0.6

81

(1.6

31)

(1.9

34)

(0.8

70)

(1.6

00)

(0.0

102)

(3.5

55)

(0.0

496)

(-0.6

27)

Tele

com

munic

ati

on

-0.4

54**

-0.2

80

1.1

26***

0.5

90

-2.6

03**

-5.1

43***

-0.2

11

0.0

53

(-2.0

93)

(-0.7

57)

(3.6

21)

(1.1

10)

(-2.2

69)

(-2.6

23)

(-0.8

47)

(0.1

24)

Insu

rance

0.6

64**

-0.8

38*

0.6

61**

-1.3

82**

11.7

29***

-16.0

17***

0.7

97***

-0.2

23

(2.3

54)

(-1.7

39)

(2.0

22)

(-2.4

73)

(3.8

08)

(-3.0

43)

(3.2

33)

(-0.5

30)

Fin

ancia

lse

rvic

es

1.9

56***

1.3

31**

1.9

19***

0.1

38

20.9

55***

16.7

98***

0.4

22*

-0.2

65

(6.2

54)

(2.4

91)

(6.2

90)

(0.2

65)

(5.7

04)

(2.6

75)

(1.6

95)

(-0.6

22)

Com

pute

r&In

form

ati

on

-1.2

58

2.2

80

2.9

86**

8.3

46***

26.7

35***

26.6

51

0.8

18

1.3

04

(-1.2

99)

(1.3

77)

(1.9

99)

(3.2

69)

(2.7

57)

(1.6

08)

(0.7

99)

(0.7

46)

Com

pute

r-1

1.0

94

43.1

86

-7.5

29

39.4

90

-140.4

33

611.9

67

0.4

17

0.5

71

(-0.4

55)

(1.0

36)

(-0.3

23)

(0.9

93)

(-0.3

90)

(0.9

94)

(0.4

56)

(0.3

65)

Info

rmati

on

7.8

25***

9.5

18*

4.2

78***

4.4

31***

-62.6

22**

-61.1

95

0.5

79

0.2

17

(2.6

63)

(1.8

95)

(4.9

06)

(2.9

74)

(-2.5

72)

(-1.4

71)

(1.0

38)

(0.2

27)

Op

era

tional

leasi

ng

1.9

19**

-0.2

30

6.2

83***

5.2

30***

16.6

02

10.5

21

2.8

13

8.0

19

(2.5

05)

(-0.1

76)

(9.2

49)

(4.5

05)

(1.1

52)

(0.4

27)

(0.8

59)

(1.4

33)

Busi

ness

serv

ices

-0.6

00***

-0.1

87

0.4

21**

-0.9

98***

-3.0

96*

-4.5

98

0.3

39**

0.5

91**

(-5.2

23)

(-0.9

52)

(2.0

75)

(-2.8

80)

(-1.8

18)

(-1.5

80)

(2.1

13)

(2.1

52)

Legal

serv

ices

-0.6

82

1.2

15

-14.7

54

-44.6

10

0.0

00

0.0

00

0.7

70

0.0

10

(-0.3

56)

(0.3

71)

(-0.1

72)

(-0.3

05)

()()

(0.2

78)

(0.0

0215)

Consu

ltin

g-0

.423*

0.6

50*

0.5

41*

-0.8

75*

-6.9

21**

-6.1

42

0.2

68

0.2

92

(-1.9

33)

(1.7

38)

(1.8

12)

(-1.7

16)

(-2.1

90)

(-1.1

37)

(0.9

53)

(0.6

08)

Advert

isin

g-1

0.3

96***

-16.6

33***

-12.7

67***

-21.0

90***

-159.4

71***

-251.1

37***

-0.1

09

-0.1

25

(-5.7

90)

(-5.4

21)

(-5.6

63)

(-5.4

74)

(-5.3

48)

(-4.9

28)

(-0.2

50)

(-0.1

69)

R&

D-0

.686**

3.0

17***

1.0

56***

2.3

20***

-1.5

19

42.6

84***

0.5

56

-0.5

57

(-2.3

86)

(6.1

44)

(3.4

63)

(4.4

50)

(-0.5

51)

(9.0

63)

(1.6

03)

(-0.9

40)

Log

GD

PSim

ilari

tyIn

dex

Corp

ora

teta

xO

bse

rvati

ons

Educati

on

-0.1

12

3.6

00***

2.4

69

-8.8

98**

-0.0

54**

0.1

67***

396

396

(-0.1

77)

(3.3

35)

(1.0

99)

(-2.3

18)

(-2.2

05)

(4.0

09)

R-s

quare

dT

ele

com

munic

ati

on

0.9

85***

1.3

69***

-1.8

74*

-0.2

67

-0.0

04

0.0

97

0.9

94

0.8

56

(4.9

72)

(4.0

41)

(-1.7

30)

(-0.1

44)

(-0.1

23)

(1.5

79)

Insu

rance

-0.8

97***

-1.2

22***

-1.8

19**

-3.9

71***

0.0

10

0.0

47*

(-5.4

87)

(-4.3

71)

(-2.0

67)

(-2.6

42)

(0.6

38)

(1.7

05)

Fin

ancia

lse

rvic

es

-0.7

22***

-1.0

29***

-0.1

21

-0.4

46

0.0

34**

0.0

66**

(-4.1

32)

(-3.4

46)

(-0.1

57)

(-0.3

38)

(2.1

00)

(2.3

66)

Com

pute

r&In

form

ati

on

6.1

93**

6.8

53

-7.9

87

-15.9

63*

-0.2

99

0.1

11

(2.0

16)

(1.3

05)

(-1.5

46)

(-1.8

07)

(-1.1

02)

(0.2

38)

Com

pute

r4.4

37

-8.9

13

-3.2

80

10.6

37

0.1

34

-0.4

52

(0.7

88)

(-0.9

26)

(-0.5

19)

(0.9

85)

(0.4

78)

(-0.9

44)

Info

rmati

on

-0.1

45

-0.5

43

3.1

27

5.5

93

-0.0

73**

-0.0

92

(-0.2

87)

(-0.6

31)

(1.1

89)

(1.2

45)

(-2.0

61)

(-1.5

32)

Op

era

tional

leasi

ng

1.0

27

1.0

96

0.0

40

0.4

65

0.2

40***

0.2

51*

(0.9

70)

(0.6

06)

(0.0

168)

(0.1

14)

(2.8

41)

(1.7

44)

Busi

ness

serv

ices

0.8

59***

0.1

09

-0.7

38

-0.3

97

-0.0

51***

-0.0

38*

(7.3

74)

(0.5

47)

(-1.1

56)

(-0.3

64)

(-3.9

56)

(-1.7

16)

Legal

serv

ices

3.0

65

2.9

64

-5.3

02

-3.6

69

-0.0

42

0.4

03

(0.2

59)

(0.1

47)

(-0.5

49)

(-0.2

22)

(-0.0

616)

(0.3

45)

Consu

ltin

g0.1

60

-0.8

67**

0.0

57

0.2

66

-0.0

64***

-0.0

43

(0.7

54)

(-2.3

89)

(0.0

625)

(0.1

70)

(-3.6

70)

(-1.4

16)

Advert

isin

g0.3

82

-1.1

18

0.1

11

1.1

53

-0.0

72**

-0.1

33**

(0.9

12)

(-1.5

60)

(0.0

572)

(0.3

48)

(-2.3

86)

(-2.5

65)

R&

D0.7

92***

-0.4

01

3.7

66***

-1.9

30

-0.0

19

0.0

80***

(4.6

47)

(-1.3

78)

(4.5

10)

(-1.3

53)

(-1.2

32)

(2.9

57)

Est

imati

ons

base

don

equati

on

(16).

Dep

endent

vari

able

sare

log

of

tota

lin

ward

sale

sand

share

of

inw

ard

affi

liate

sale

s.E

stim

ati

on

meth

od:

SU

R.

zst

ati

stic

sin

pare

nth

ese

s.Secto

rdum

my

vari

able

sare

inclu

ded

inth

ere

gre

ssio

n,

but

are

not

rep

ort

ed

inth

eta

ble

.*,

**

and

***

indic

ate

stati

stic

al

signifi

cance

at

the

10-p

erc

ent

level,

5-p

erc

ent

level,

and

1-p

erc

ent

level

resp

ecti

vely

.

41

Table 10: Regression results: Manufacturing outward affiliate sales share

Log Distance Language Restrictiveness IndexTotal manufacturing -0.170 2.006*** -13.428***

(-1.231) (2.999) (-2.691)Food 1.735*** 9.857*** -42.955***

(4.834) (4.658) (-4.197)Chemicals 0.351 1.026 -4.128

(1.397) (1.520) (-0.863)Metals -0.816*** 1.545** -23.640***

(-3.403) (2.373) (-4.464)Machinery 0.462*** 3.819*** -18.512***

(2.958) (4.164) (-2.671)Computer -2.497 -0.036 24.086

(-1.023) (-0.0450) (1.552)Electrical equipment -0.948** -1.345* 7.859

(-2.337) (-1.804) (1.499)Transport equipment -1.456*** 2.314*** 13.785

(-2.818) (2.772) (1.356)

Log GDP Similarity Index Corporate tax ObservationsTotal manufacturing 0.759*** -7.281*** -0.050 456

(3.356) (-3.674) (-1.315)Food -6.141*** 6.382 0.413***

(-5.080) (1.543) (4.186)Chemicals 2.353*** -15.037*** -0.186***

(3.885) (-5.873) (-3.199)Metals 0.212 -3.728 0.010

(0.496) (-1.279) (0.224)Machinery 0.151 -3.805 -0.017

(0.287) (-1.244) (-0.335)Computer 2.099*** 2.756 -0.094

(4.024) (0.737) (-0.891)Electrical equipment 0.995 -0.817 -0.148*

(1.472) (-0.245) (-1.666)Transport equipment 2.718*** -14.539*** -0.138

(3.819) (-4.729) (-1.556)

Estimations based on equation (15). Dependent variable is share of outward affiliate sales in total outward sales.Estimation method: GLM model. Robust z statistics in parentheses.Sector dummy variables are included in the regression, but are not reported in the table.*, ** and *** indicate statistical significance at the 10-percent level, 5-percent level, and 1-percent level respectively.

42

Table 11: Regression results: Manufacturing inward affiliate sales share

Log Distance Language Restrictiveness IndexTotal manufacturing 0.165*** -0.103 -18.285***

(2.759) (-0.516) (-4.986)Food -1.049 -1.312 14.314

(-1.503) (-1.322) (0.672)Chemicals -0.717*** -2.430*** -20.771***

(-3.211) (-5.227) (-3.218)Metals 0.190 5.758*** -24.809***

(0.659) (6.913) (-2.690)Machinery -1.217*** -0.573 -39.851***

(-4.838) (-0.516) (-7.420)Computer -0.554 -1.455* -15.410

(-1.595) (-1.682) (-1.418)Electrical equipment -0.175 0.067 -12.613***

(-0.556) (0.0617) (-2.675)Transport equipment -1.676*** -1.217 -43.264***

(-3.191) (-1.365) (-5.632)

Log GDP Similarity Index Corporate tax ObservationsTotal manufacturing -0.701* -0.192 -0.050*** 415

(-1.798) (-0.169) (-3.417)Food 1.873 11.584** 0.034

(1.338) (2.315) (0.595)Chemicals -0.393 2.958* -0.084**

(-0.689) (1.729) (-2.399)Metals 2.031*** -2.160 0.152***

(3.658) (-0.317) (4.448)Machinery -1.005** 4.876 -0.010

(-2.345) (1.197) (-0.238)Computer 0.588 -2.502 -0.003

(0.678) (-0.641) (-0.0407)Electrical equipment 0.078 -0.984 -0.315**

(0.123) (-0.244) (-2.211)Transport equipment 3.110*** -23.413*** -0.070

(5.144) (-3.739) (-1.279)

Estimations based on equation (16). Dependent variable is share of inward affiliate sales in total inward sales.Estimation method: GLM model. Robust z statistics in parentheses.Sector dummy variables are included in the regression, but are not reported in the table.*, ** and *** indicate statistical significance at the 10-percent level, 5-percent level, and 1-percent level respectively.

43