CE CENTRE FOR EUROPEAN POLICY STUDIES WORKING DOCUMENT NO.156 NOVEMBER 2000 THE IMPACT OF ECONOMIC INTEGRATION ON FDI AND EXPORTS: A GRAVITY APPROACH FRANCESCA DI MAURO CEPS Working Documents are published to give an early indication of the work in progress within CEPS research programmes and to stimulate reactions from other experts in the field.Unless otherwise indicated,the views expressed are attributable only to the author in a personal capacity and not to any institution with which she is associated. ISBN92-9079-313-9 COPYRIGHT 2000,FRANCESCA DI MAURO
CENTRE FOR EUROPEAN POLICY STUDIES WORKING DOCUMENT NO. 156 NOVEMBER 2000 THE IMPACT OF ECONOMIC INTEGRATION ON FDI AND EXPORTS: A GRAVITY APPROACH FRANCESCA DI MAURO CEPS Working Documents are published to give an early indication of the work in progress within CEPS research programmes and to stimulate reactions from other experts in the field. Unless otherwise indicated, the views expressed are attributable only to the author in a personal capacity and not to any institution with which she is associated. ISBN 92-9079-313-9 © COPYRIGHT 2000, FRANCESCA DI MAURO
THE IMPACT OF ECONOMIC INTEGRATION ON FDI AND EXPORTS: A GRAVITY APPROACH FRANCESCA DI MAURO* CEPS WORKING DOCUMENT NO.156,NOVEMBER 2000 Abstract This paper uses the gravity-model approach to deal with two issues related to economic integration.The first concern is to analyse the impact on FDI stocks of specific variables denoting the will to integrate,and their relative impact on exports.Variables considered include tariffs,non-tariff barriers and exchange rate variability.The results show that the widespread opinion-and theoretical claim-of'tariff-jumping'FDI is not supported by the evidence.Moreover,non-tariff barriers have a negative impact on FDI,revealing the greater role of sunk costs for foreign investors as opposed to exporters.In contrast to the impact on exports,exchange rate variability does not have a negative impact on FDI,since it can partially be overcome by directly investing in the host country.The second concern deals with the debate on the complementarity vs.substitutability relationship between exports and FDI.At the aggregate level,the results show that a complementary relationship holds. Keywords:Foreign Direct Investment,Economic integration,Gravity Model. JEL classification codes:F15,F21,F23. PhD candidate at the Universite Libre de Bruxelles,ULB and Research Fellow at the Centre for European Policy Studies,CEPS(fdimauro @ceps.be).This work constitutes the first paper within her PhD programme at ULB.The author wishes to acknowledge useful comments and suggestions provided by her supervisor Prof.Andre Sapir,as well as by participants at a seminar held in CEPS: Paul Brenton,Daniel Gros,Jacques Pelkmans and Anna Maria Pinna.Ms.Di Mauro is also grateful for financial support given to this work by the European Commission under a TMR Grant of DG Research
THE IMPACT OF ECONOMIC INTEGRATION ON FDI AND EXPORTS: A GRAVITY APPROACH FRANCESCA DI MAURO* CEPS WORKING DOCUMENT NO. 156, NOVEMBER 2000 Abstract This paper uses the gravity-model approach to deal with two issues related to economic integration. The first concern is to analyse the impact on FDI stocks of specific variables denoting the will to integrate, and their relative impact on exports. Variables considered include tariffs, non-tariff barriers and exchange rate variability. The results show that the widespread opinion – and theoretical claim – of ‘tariff-jumping’ FDI is not supported by the evidence. Moreover, non-tariff barriers have a negative impact on FDI, revealing the greater role of sunk costs for foreign investors as opposed to exporters. In contrast to the impact on exports, exchange rate variability does not have a negative impact on FDI, since it can partially be overcome by directly investing in the host country. The second concern deals with the debate on the complementarity vs. substitutability relationship between exports and FDI. At the aggregate level, the results show that a complementary relationship holds. Keywords: Foreign Direct Investment, Economic integration, Gravity Model. JEL classification codes: F15, F21, F23. * PhD candidate at the Université Libre de Bruxelles, ULB and Research Fellow at the Centre for European Policy Studies, CEPS (fdimauro@ceps.be). This work constitutes the first paper within her PhD programme at ULB. The author wishes to acknowledge useful comments and suggestions provided by her supervisor Prof. André Sapir, as well as by participants at a seminar held in CEPS: Paul Brenton, Daniel Gros, Jacques Pelkmans and Anna Maria Pinna. Ms. Di Mauro is also grateful for financial support given to this work by the European Commission under a TMR Grant of DG Research
THE IMPACT OF ECONOMIC INTEGRATION ON FDI AND EXPORTS: A GRAVITY APPROACH FRANCESCA DI MAURO Introduction Economic integration between countries has continued to deepen over the past decade This is especially visible at the regional level,with the escalation of Regional Integration Agreements(RIAs)ranging from Free Trade Areas(FTAs)to Customs Unions(CUs),such as NAFTA,Mercosur or ASEAN.These developments have renewed interest in the economics of regional integration,first raised by Viner (1950).Since trade and Foreign Direct Investment(FDI)are generally recognised as the two main channels of economic integration,the most topical issues in the debate about RIAs relate to trade creation,trade diversion,the redirection'of FDI from non-members to members of the RIA and the phenomenon of 'tariff-jumping FDI'(the latter suggesting FDI is only substitute for trade). A typical example of the perceived threat'deriving from RIAs is the European Single Market,initially labelled Fortress Europe'.The European Union,however,provides also one of the most interesting laboratories to assess the impact of such deep integration. Indeed,several studies have been carried out in order to assess its economic impact,either ex-ante-see e.g.Baldwin (1989),Neven and Roller(1991)and Brenton and Winters (1992), or ex-post in the years following the implementation of the Single Market Programme (SMP) -Baldwin et al.(1995),Sapir (1996)and Brenton (1996).Amongst others,the European Commission study (1996)attempted to include all aspects of potential effects:on trade,on efficiency and competition,on FDI,on income and employment and on the growth and convergence of EU Member States. In this paper the main concern is not to assess the impact of the Single Market Programme specifically,but the emphasis is rather on the effect of economic integration upon FDI, relatire to that on exports.Is economic integration more beneficial to FDI or to exports? Which variables reflecting economic integration are more prone to FDI and to exports? Given that FDI is often associated with greater dynamic effects,such as the technology transfer,which in turn may lead to beneficial effects for the recipient country,the impact of economic integration on FDI is potentially more important and deserves close attention. Economic integration is proxied here through three main variables:exchange rate variability (ERV),tariff barriers and non-tariff barriers (NIBs),and included in gravity-type equations for FDI and for exports in turn.These variables are the most immediate policy instruments available to countries to reveal their will to integrate.The surge of currency boards in emerging countries is an example of what can be called 'monetary integration'. These economies commit to a stable exchange rate in order to gain the trust of investors and to boost their trade.The set-up of the European monetary system (EMS)in the EU pursued the primary objective of stabilising the EU currencies,with an expected benefit for the economies.Similar arguments hold for a reduction in tariffs and even more in NTBs, both measures of 'commercial integration'.I should stress here that this paper does not 1
1 THE IMPACT OF ECONOMIC INTEGRATION ON FDI AND EXPORTS: A GRAVITY APPROACH FRANCESCA DI MAURO Introduction Economic integration between countries has continued to deepen over the past decade. This is especially visible at the regional level, with the escalation of Regional Integration Agreements (RIAs) ranging from Free Trade Areas (FTAs) to Customs Unions (CUs), such as NAFTA, Mercosur or ASEAN. These developments have renewed interest in the economics of regional integration, first raised by Viner (1950). Since trade and Foreign Direct Investment (FDI) are generally recognised as the two main channels of economic integration, the most topical issues in the debate about RIAs relate to trade creation, trade diversion, the ‘redirection’ of FDI from non-members to members of the RIA and the phenomenon of ‘tariff-jumping FDI’ (the latter suggesting FDI is only substitute for trade). A typical example of the perceived ‘threat’ deriving from RIAs is the European Single Market, initially labelled ‘Fortress Europe’. The European Union, however, provides also one of the most interesting laboratories to assess the impact of such deep integration. Indeed, several studies have been carried out in order to assess its economic impact, either ex-ante - see e.g. Baldwin (1989), Neven and Röller (1991) and Brenton and Winters (1992), or ex-post in the years following the implementation of the Single Market Programme (SMP) – Baldwin et al. (1995), Sapir (1996) and Brenton (1996). Amongst others, the European Commission study (1996) attempted to include all aspects of potential effects: on trade, on efficiency and competition, on FDI, on income and employment and on the growth and convergence of EU Member States. In this paper the main concern is not to assess the impact of the Single Market Programme specifically, but the emphasis is rather on the effect of economic integration upon FDI, relative to that on exports. Is economic integration more beneficial to FDI or to exports? Which variables reflecting economic integration are more prone to FDI and to exports? Given that FDI is often associated with greater dynamic effects, such as the technology transfer, which in turn may lead to beneficial effects for the recipient country, the impact of economic integration on FDI is potentially more important and deserves close attention. Economic integration is proxied here through three main variables: exchange rate variability (ERV), tariff barriers and non-tariff barriers (NTBs), and included in gravity-type equations for FDI and for exports in turn. These variables are the most immediate policy instruments available to countries to reveal their will to integrate. The surge of currency boards in emerging countries is an example of what can be called ‘monetary integration’. These economies commit to a stable exchange rate in order to gain the trust of investors and to boost their trade. The set-up of the European monetary system (EMS) in the EU pursued the primary objective of stabilising the EU currencies, with an expected benefit for the economies. Similar arguments hold for a reduction in tariffs and even more in NTBs, both measures of ‘commercial integration’. I should stress here that this paper does not
FRANCESCA DI MAURO seek to reconsider the relationship between exchange rate variability and trade,a topic extensively treated in the literature (see e.g.De Grauwe and Bellefroid (1986),Peree and Steinherr(1989),Gagnon (1993)and Sapir and Sekkat(1995)).In contrast,this relationship has been largely neglected in the FDI literature,even though trade and FDI share very similar characteristics.My purpose is therefore to analyse the relative impact of exchange rate variability on exports as compared to the one on FDI. Economic integration among countries can also be a concern in terms of employment.An important example was the creation of NAFTA,with American workers perceiving a jobs threat from Mexican 'maquiladoras'.A similar concern arises in the EU,where European firms investing in Central and Eastern Europe are seen as replacing European labour with cheaper labour available in these countries.With either horizontal or vertical FDI,domestic firms investing abroad rather than exporting may leave the export sector at home worse- off.This constitutes the second concern of this paper,and the issue of whether exports and FDI are complements or substitutes will be addressed following Graham's (1996) approach. It should be stressed at the outset that the focus of the paper is the study of FDI,while the analysis on exports is used as a comparative tool.Thus the review in Section 1 only focuses on the theory of FDI and it is finalised to the derivation of the gravity model used in the empirical part.In Section 2 I discuss the expected impact of economic integration on exports and FDI and their potential inter-linkages.The empirical analysis will be presented in Section 3,followed by some concluding remarks in Section 4. 1. The theory of FDI In recent years FDI has received more and more interest from economists and policy- makers.On the one hand,this is probably due to its growing economic importance for both developed and developing countries.According to the 1999 World Inrestment Report,in the past decade both global output and global sales have grown faster than world GDP and world exports.Thus,sales of foreign affiliates are now greater than world total exports of goods,implying that firms use FDI more than they use exports to service foreign markets. Moreover,FDI inward flows represented in 1998 11%of Gross Fixed Capital Formation (UNCIAD (2000))revealing the importance that these flows can have for economic growth.On the other hand,given the many aspects of FDI,a wide range of economists has become involved in the research:from trade economists,for the close relationship between trade and FDI and development and growth economists,for its effects on the host economy,to regional economists,for its implications for RIAs and industrial economists, for the impact on industrial restructuring.In this section though,I am mainly concerned with the theories that explain the determinants of FDI. 1.1 The OLI framework The traditional theory of FDI tries to explain why firms produce abroad instead of simply servicing the markets via exports.After all,multinational companies (MNCs)experience additional costs in producing abroad:higher costs in placing personnel abroad, communication costs (international phone calls,travel expenses for executives or even time costs due to mail delays),language and cultural differences,informational costs on local tax laws and regulations,costs of being outside domestic networks;they also incur higher risks, 2
FRANCESCA DI MAURO 2 seek to reconsider the relationship between exchange rate variability and trade, a topic extensively treated in the literature (see e.g. De Grauwe and Bellefroid (1986), Perée and Steinherr (1989), Gagnon (1993) and Sapir and Sekkat (1995)). In contrast, this relationship has been largely neglected in the FDI literature, even though trade and FDI share very similar characteristics. My purpose is therefore to analyse the relative impact of exchange rate variability on exports as compared to the one on FDI. Economic integration among countries can also be a concern in terms of employment. An important example was the creation of NAFTA, with American workers perceiving a jobs threat from Mexican ‘maquiladoras’. A similar concern arises in the EU, where European firms investing in Central and Eastern Europe are seen as replacing European labour with cheaper labour available in these countries. With either horizontal or vertical FDI, domestic firms investing abroad rather than exporting may leave the export sector at home worseoff. This constitutes the second concern of this paper, and the issue of whether exports and FDI are complements or substitutes will be addressed following Graham's (1996) approach. It should be stressed at the outset that the focus of the paper is the study of FDI, while the analysis on exports is used as a comparative tool. Thus the review in Section 1 only focuses on the theory of FDI and it is finalised to the derivation of the gravity model used in the empirical part. In Section 2 I discuss the expected impact of economic integration on exports and FDI and their potential inter-linkages. The empirical analysis will be presented in Section 3, followed by some concluding remarks in Section 4. 1. The theory of FDI In recent years FDI has received more and more interest from economists and policymakers. On the one hand, this is probably due to its growing economic importance for both developed and developing countries. According to the 1999 World Investment Report, in the past decade both global output and global sales have grown faster than world GDP and world exports. Thus, sales of foreign affiliates are now greater than world total exports of goods, implying that firms use FDI more than they use exports to service foreign markets. Moreover, FDI inward flows represented in 1998 11% of Gross Fixed Capital Formation (UNCTAD (2000)) revealing the importance that these flows can have for economic growth. On the other hand, given the many aspects of FDI, a wide range of economists has become involved in the research: from trade economists, for the close relationship between trade and FDI and development and growth economists, for its effects on the host economy, to regional economists, for its implications for RIAs and industrial economists, for the impact on industrial restructuring. In this section though, I am mainly concerned with the theories that explain the determinants of FDI. 1.1 The OLI framework The traditional theory of FDI tries to explain why firms produce abroad instead of simply servicing the markets via exports. After all, multinational companies (MNCs) experience additional costs in producing abroad: higher costs in placing personnel abroad, communication costs (international phone calls, travel expenses for executives or even time costs due to mail delays), language and cultural differences, informational costs on local tax laws and regulations, costs of being outside domestic networks; they also incur higher risks
THE IMPACT OF ECONOMIC INTEGRATION ON FDI AND EXPORTS such as the risks of exchange rate changes or even of expropriation by the host country. One theoretical approach,introduced by Dunning (1977,1981),the "OLI framework", considers FDI as determined by Ownership,Location and Internalisation advantages which the MNC holds over the foreign producer;when these advantages outweigh the above costs,FDI arises.The ownership adrantage includes a product or a production process to which other firms do not have access,such as a patent,blueprint or trade secret,to more intangible advantages such as reputation for quality.The location adrantage stems directly from the foreign market,such as low factor prices or customer access,together with trade barriers or transport costs that make FDI more profitable than exporting.Finally,the internalisation adrantage is a more abstract concept to explain why licensing may not be practised;it derives from the firm's interest in maintaining its knowledge assets (such as highly skilled workers who know the firm's technology)internally.This avoids "defection" once the licensee has come to understand the technology and sets up his own firm,in competition with the MNC.1 Informational asymmetries may also push MNCs to prefer foreign production over licensing,such as better knowledge of the domestic market by the licensee.The fear of being substituted with direct production in the presence of highly selling markets would provide incentive for the licensee to under-declare the potential absorption capacity of a market.Finally,advantages derive from the reduction of transaction costs (for contracting,quality assurance,etc.)that arise in case of licensing. The main problem of this framework is that although it does explain the existence of MNCs,it has had difficulty explaining the recent trends in FDI,namely their surge among similar countries (horizontal FDD);further,no sound empirical models have been generated in order to compare real data with the theory. 1.2The“New Theory of FDI” The so-called "New Theory of FDI"refers mainly to the ownership and location advantage and introduces MNCs in general equilibrium models,where they arise endogenously. Helpman(1984)and Helpman and Krugman(1985)-exponents of the early literature- derive the activity of MNCs when they try to explain intra-firm trade,that is,an additional component of international trade.The models are based on two main assumptions:(1) there is product differentiation and economies of scale,and(2)there are some firm inputs that behave like public goods2.Moreover,it is assumed that transport costs are zero and the MNCs will split their production process between a headquarter'activity,often skill or capital-intensive,and the plant production abroad.In other terms,the factor proportions in the two activities of the MNE differ,which is the rationale for multinational activity to arise at all.This is recognisable as vertical FDI,since firms separate their production process in order to take advantage of factor price differentials across countries.The implications of these models for a potential derivation of the gravity model are that only differences in relative factor endowments across countries (often proxied by GDP per capita)matter for the location of MNCs abroad.This also implies that the 'type'of FDI Of course,the problem of"moral hazard"can also appear within a subsidiary,but its activity is more subject to the MNC's control. 2 These are inputs such as management,marketing,R&D,that are specific to the firm and that can be easily transferred from one plant to another,at virtually no cost,hence the denomination of public good'. 3
THE IMPACT OF ECONOMIC INTEGRATION ON FDI AND EXPORTS 3 . such as the risks of exchange rate changes or even of expropriation by the host country. One theoretical approach, introduced by Dunning (1977, 1981), the “OLI framework”, considers FDI as determined by Ownership, Location and Internalisation advantages which the MNC holds over the foreign producer; when these advantages outweigh the above costs, FDI arises. The ownership advantage includes a product or a production process to which other firms do not have access, such as a patent, blueprint or trade secret, to more intangible advantages such as reputation for quality. The location advantage stems directly from the foreign market, such as low factor prices or customer access, together with trade barriers or transport costs that make FDI more profitable than exporting. Finally, the internalisation advantage is a more abstract concept to explain why licensing may not be practised; it derives from the firm’s interest in maintaining its knowledge assets (such as highly skilled workers who know the firm’s technology) internally. This avoids “defection” once the licensee has come to understand the technology and sets up his own firm, in competition with the MNC.1 Informational asymmetries may also push MNCs to prefer foreign production over licensing, such as better knowledge of the domestic market by the licensee. The fear of being substituted with direct production in the presence of highly selling markets would provide incentive for the licensee to under-declare the potential absorption capacity of a market. Finally, advantages derive from the reduction of transaction costs (for contracting, quality assurance, etc.) that arise in case of licensing. The main problem of this framework is that although it does explain the existence of MNCs, it has had difficulty explaining the recent trends in FDI, namely their surge among similar countries (horizontal FDI); further, no sound empirical models have been generated in order to compare real data with the theory. 1.2 The “New Theory of FDI” The so-called “New Theory of FDI” refers mainly to the ownership and location advantage and introduces MNCs in general equilibrium models, where they arise endogenously. Helpman (1984) and Helpman and Krugman (1985) – exponents of the early literature - derive the activity of MNCs when they try to explain intra-firm trade, that is, an additional component of international trade. The models are based on two main assumptions: (1) there is product differentiation and economies of scale, and (2) there are some firm inputs that behave like public goods2. Moreover, it is assumed that transport costs are zero and the MNCs will split their production process between a ‘headquarter’ activity, often skill or capital-intensive, and the plant production abroad. In other terms, the factor proportions in the two activities of the MNE differ, which is the rationale for multinational activity to arise at all. This is recognisable as ‘vertical FDI’, since firms separate their production process in order to take advantage of factor price differentials across countries. The implications of these models for a potential derivation of the gravity model are that only differences in relative factor endowments across countries (often proxied by GDP per capita) matter for the location of MNCs abroad. This also implies that the ‘type’ of FDI 1 Of course, the problem of “moral hazard” can also appear within a subsidiary, but its activity is more subject to the MNC’s control. 2 These are inputs such as management, marketing, R&D, that are specific to the firm and that can be easily transferred from one plant to another, at virtually no cost, hence the denomination of ‘public good’
FRANCESCA DI MAURO that can arise from the theory is limited to 'vertical'FDI,also called 'efficiency-seeking' FDI. Instead,what is observed among developed countries is mainly horizontal FDI,because similar types of production activities,owned by MNCs,take place in different countries,as for example in the car industry.Brainard's (1993)and Markusen and Venables'(1998) models account for this type of FDI and they arrive at empirically testable hypothesis about the activity of MNCs.Their work starts from the observation that indeed most FDI is motivated by "market-access"reasons,rather than by differences in factor prices;this is confirmed for example by the fact that in 1999 over 90%of FDI is between North-North countries (UNCTAD (2000))rather than North-South,as the Helpman-Krugman theory of MNCs would predict.The main idea is that the firm faces a trade-off between advantages of proximity (to the foreign market)vs.advantages of concentration (of the plant),given the presence of firm-level economies of scale (again,special inputs such as R&D, management,etc.)as well as the usual plant-level economies of scale.With transport costs, whenever the former advantage outweighs the latter,a firm will go multinational and replace exports with FDI. In Brainard's model (1993)three types of equilibria can arise:a pure multinational equilibrium,when proximity advantages are greater than concentration ones,a pure trade equilibrium,when the opposite is true,and a'mixed equilibrium'when the two advantages are equal and MNCs coexist with single-plant,single-country firms.It is important to note at this stage that the model assumes symmetry of factor endowments,which is the reason why 'vertical MNCs'cannot arise,in contrast to the Helpman-Krugman model.The empirical hypotheses are derived from the costs structure of the firms,who face additional variable costs in exporting and additional fixed costs of opening up a plant abroad.For example,the number of firms that will export increases with the fixed plant cost (because then they can benefit from economies of scale by concentrating production)and decreases with the level of transport costs and trade barriers.Strictly speaking,from the theoretical model can be derived only those variables that are linked to plant-level vs.firm-level economies of scale,plus the 'trade costs'variables.This implies for example the presence of distance (a proxy for transport costs)or trade costs (tariffs).However,Brainard's empirical work(Brainard (1997),in addition to these mentioned variables,introduces the income per-worker differential',justified as a control variable for factor-proportions differences and the host country GDP,because,in her words,this variable is more likely to be an important determinant of the presence of MNCs. In the model of Markusen and Venables (1998)MNCs arise endogenously in a general equilibrium framework,even though their focus is only on horizontal direct investment. The model includes two countries,two homogenous goods and two factors.Firms in each country can be of two types:national or multinational,which gives four firm-types in total. From the firms'different cost structures and with the assumption of Cournot competition and free entry,the model can explicitly solve for production regimes',i.e.the combination of firm types that operate in equilibrium.Again,the key variables for determining the presence of MNCs are transport costs,plant and firm-level economies of scale and market size.Asymmetry of countries in terms of relative factor endowments does not lead to vertical MNCs,since they are excluded by assumption.In contrast,the general result is that MNCs become more and more important as countries become more similar in size,relative endowments and as world income grows
FRANCESCA DI MAURO 4 that can arise from the theory is limited to ‘vertical’ FDI, also called ‘efficiency-seeking’ FDI. Instead, what is observed among developed countries is mainly horizontal FDI, because similar types of production activities, owned by MNCs, take place in different countries, as for example in the car industry. Brainard’s (1993) and Markusen and Venables’ (1998) models account for this type of FDI and they arrive at empirically testable hypothesis about the activity of MNCs. Their work starts from the observation that indeed most FDI is motivated by “market-access” reasons, rather than by differences in factor prices; this is confirmed for example by the fact that in 1999 over 90% of FDI is between North-North countries (UNCTAD (2000)) rather than North-South, as the Helpman-Krugman theory of MNCs would predict. The main idea is that the firm faces a trade-off between advantages of proximity (to the foreign market) vs. advantages of concentration (of the plant), given the presence of firm-level economies of scale (again, special inputs such as R&D, management, etc.) as well as the usual plant-level economies of scale. With transport costs, whenever the former advantage outweighs the latter, a firm will go multinational and replace exports with FDI. In Brainard’s model (1993) three types of equilibria can arise: a pure multinational equilibrium, when proximity advantages are greater than concentration ones, a pure trade equilibrium, when the opposite is true, and a ‘mixed equilibrium’ when the two advantages are equal and MNCs coexist with single-plant, single-country firms. It is important to note at this stage that the model assumes symmetry of factor endowments, which is the reason why ‘vertical MNCs’ cannot arise, in contrast to the Helpman-Krugman model. The empirical hypotheses are derived from the costs structure of the firms, who face additional variable costs in exporting and additional fixed costs of opening up a plant abroad. For example, the number of firms that will export increases with the fixed plant cost (because then they can benefit from economies of scale by concentrating production) and decreases with the level of transport costs and trade barriers. Strictly speaking, from the theoretical model can be derived only those variables that are linked to plant-level vs. firm-level economies of scale, plus the ‘trade costs’ variables. This implies for example the presence of distance (a proxy for transport costs) or trade costs (tariffs). However, Brainard’s empirical work (Brainard (1997)), in addition to these mentioned variables, introduces the ‘income per-worker differential’, justified as a control variable for factor-proportions differences and the host country GDP, because, in her words, this variable is more likely to be an important determinant of the presence of MNCs. In the model of Markusen and Venables (1998) MNCs arise endogenously in a general equilibrium framework, even though their focus is only on horizontal direct investment. The model includes two countries, two homogenous goods and two factors. Firms in each country can be of two types: national or multinational, which gives four firm-types in total. From the firms’ different cost structures and with the assumption of Cournot competition and free entry, the model can explicitly solve for ‘production regimes’, i.e. the combination of firm types that operate in equilibrium. Again, the key variables for determining the presence of MNCs are transport costs, plant and firm-level economies of scale and market size. Asymmetry of countries in terms of relative factor endowments does not lead to vertical MNCs, since they are excluded by assumption. In contrast, the general result is that MNCs become more and more important as countries become more similar in size, relative endowments and as world income grows
THE IMPACT OF ECONOMIC INTEGRATION ON FDI AND EXPORTS In Markusen(1997)the model is refined even further,with the formal introduction of both types of MNCs:horizontal and vertical,plus the usual national single-plant firms.There are therefore six firm-types:single-plant NE in each country,two-plant horizontal MNCs and single-plant vertical MNCs3.We are still in a theoretical setting and empirical representations of the theory are always conducted via numerical simulations.One of the messages of the paper is that MNCs benefit from (or produce)some sort of knowledge- capital',that allows headquarter services to provide a crucial input for production abroad.It is only with a series of papers in late 1998 and 1999 that an empirical test of the theory is carried out (Carr,Markusen and Maskus (1998),Markusen and Maskus (1999a)and Markusen and Maskus(1999b)),as well as a synthesis of past theoretical advances.In the last of these papers,three models are tested simultaneously:the knowledge-capital model (KK),the horizontal MNCs model (HOR)and the vertical MNCs one (VER).A simulation analysis is used to identify the impact on each model of the recurrent variables:the size of the economy (measured by GDP),world GDP(the sum of each country's GDP)and the relative factor endowments(measured by the difference in the ratio of skilled workers over total employment),as well as some interaction term between them.Their results give strong support to the HOR model,while both the VER and KK do poorly. 1.3 From theory to practice:the Gravity Model When I get to the empirical analysis,and I want to be able to compare 'attractiveness' across countries and explain the geographic distribution of FDI,I need a model that can pick up its common determinants.In order to synthesise the two approaches discussed above,i.e.Helpman and Krugman's treatment of vertical FDI and Brainard's horizontal one,I will include in the model the following variables4:relative factor endowments,an index of countries'similarity in size,geographic distance between the partner countries and a measure of the 'economic space'between the two countries,given by the sum of the two GDPs.The last variable is included to catch the market-seeking'aspect of FDI,ie.when investors produce abroad to sell in the host market and increase their market shares there. Additional variables,such as a common language,a common border,or preferential trade agreements,that may reduce the costs of locating abroad,can be introduced via dummy variables. This simple specification can easily be recognised as the gravity model',which I will use throughout the empirical analysis.In contrast to the common view among economists,the gravity model rests on a sound theoretical basis.Maurel (1998)carries out a thorough investigation on its origins,and shows its evolution across the trade theories of Linnemann (1966),Helpman and Krugman (1985)and the empirical studies by Helpman (1987)and Brainard(1997).She also shows how the gravity model applied to trade can be compatible with both the traditional Heckscher-Ohlin and the Helpman and Krugman framework, 3 This set of firm-types is still not complete,as there may be two-plant vertical MNCs,where only a part of the production chain is carried out abroad and the re-imported goods are intermediary.The cost structure in this case may be different. 4 A similar version of this model specification was first introduced by Helpman(1987)for a trade equation;more recently,Egger(2000)has applied a refined version of it to both exports and FDI data. 5
THE IMPACT OF ECONOMIC INTEGRATION ON FDI AND EXPORTS 5 . In Markusen (1997) the model is refined even further, with the formal introduction of both types of MNCs: horizontal and vertical, plus the usual national single-plant firms. There are therefore six firm-types: single-plant NE in each country, two-plant horizontal MNCs and single-plant vertical MNCs3. We are still in a theoretical setting and empirical representations of the theory are always conducted via numerical simulations. One of the messages of the paper is that MNCs benefit from (or produce) some sort of ‘knowledgecapital’, that allows headquarter services to provide a crucial input for production abroad. It is only with a series of papers in late 1998 and 1999 that an empirical test of the theory is carried out (Carr, Markusen and Maskus (1998), Markusen and Maskus (1999a) and Markusen and Maskus (1999b)), as well as a synthesis of past theoretical advances. In the last of these papers, three models are tested simultaneously: the knowledge-capital model (KK), the horizontal MNCs model (HOR) and the vertical MNCs one (VER). A simulation analysis is used to identify the impact on each model of the recurrent variables: the size of the economy (measured by GDP), world GDP (the sum of each country’s GDP) and the relative factor endowments (measured by the difference in the ratio of skilled workers over total employment), as well as some interaction term between them. Their results give strong support to the HOR model, while both the VER and KK do poorly. 1.3 From theory to practice: the Gravity Model When I get to the empirical analysis, and I want to be able to compare ‘attractiveness’ across countries and explain the geographic distribution of FDI, I need a model that can pick up its common determinants. In order to synthesise the two approaches discussed above, i.e. Helpman and Krugman’s treatment of vertical FDI and Brainard’s horizontal one, I will include in the model the following variables4: relative factor endowments, an index of countries’ similarity in size, geographic distance between the partner countries and a measure of the ‘economic space’ between the two countries, given by the sum of the two GDPs. The last variable is included to catch the ‘market-seeking’ aspect of FDI, i.e. when investors produce abroad to sell in the host market and increase their market shares there. Additional variables, such as a common language, a common border, or preferential trade agreements, that may reduce the costs of locating abroad, can be introduced via dummy variables. This simple specification can easily be recognised as the ‘gravity model’, which I will use throughout the empirical analysis. In contrast to the common view among economists, the gravity model rests on a sound theoretical basis. Maurel (1998) carries out a thorough investigation on its origins, and shows its evolution across the trade theories of Linnemann (1966), Helpman and Krugman (1985) and the empirical studies by Helpman (1987) and Brainard (1997). She also shows how the gravity model applied to trade can be compatible with both the traditional Heckscher-Ohlin and the Helpman and Krugman framework, 3 This set of firm-types is still not complete, as there may be two-plant vertical MNCs, where only a part of the production chain is carried out abroad and the re-imported goods are intermediary. The cost structure in this case may be different. 4 A similar version of this model specification was first introduced by Helpman (1987) for a trade equation; more recently, Egger (2000) has applied a refined version of it to both exports and FDI data
FRANCESCA DI MAURO without becoming a meaningless black box.Given the similarity between trade and FDI in terms of trends,it has also been employed to estimate bilateral FDI flows(see Brenton (1996),Eaton and Tamura(1996),and Brenton and Di Mauro (1999)).The general form of the gravity equation that I estimate is the following: InY =a+B,SUMGDP,+B2 SIMSIZE +B,RELENDOW,+B,DIST,+YDu+ (1) with the following variable definitions:Y is the value of FDI or exports from country i (home country)to country /(host country); SUMGDP,=In(GDP +GDP) (②) SIMSIZE;In1 GDP GDP (3) GDP+GDP GDP,+GDP RELENDOW In- GDP-In GDP Popi Popi (④ DIST is the relative distance between countries i and j.Following the advice of Polak (1996),I use here a measure of relative distance rather than absolute,i.e.the logarithm of actual distance divided by the average distance of the investing country from its partners, weighted by the shares of the latter in world GDP.This correction prevents the gravity model from producing biased results:a downward bias for far-away countries and an upward one for close-in countries. Dkii are dummy variables (mostly country dummies)used when appropriate and 8 is the usual error term. A precision needs to be made at this point concerning the dependent variable FDI.In fact, the models a la Brainard and Markusen are mainly concerned with MNC activity,as opposed to that of the national exporter.With this respect,they correctly identify as the principal variable of interest-especially for the empirical analysis-the level of affiliate production abroad.This requires availability of such data,which in practice is limited-in a consistent way-to Sweden and the USA,typical case-studies of the literature (see e.g. Blomstrom et al.(1997),Brainard and Riker (1997),Markusen and Maskus (1999b)and Svensson (1996)).As my variable of interest is FDI,at least I make use of FDI stock,with the justification that these are used in the production process abroad,and hence constitute a good proxy for foreign affiliate production5.I now briefly discuss the impact of what I will call the gravity variables'. The 'economic space'variable (SUMGDP)is expected to have a positive impact in both the FDI and exports equation.The index of size similarity (SIMSIZE)takes values between -oo(i.e.the log of a number near zero)in case of perfect dissimilarity and-0.69 (the log of 0.5)for perfect similarity.Similarity in size should have a positive effect on exports: 5A strong correlation certainly exists between FDI stock and foreign affiliates production,and I am implicitly assuming here that the relationship is linear,but it may not be.Further inquiry in this direction could prove useful and I envisage it for the future. 6
FRANCESCA DI MAURO 6 without becoming a meaningless black box. Given the similarity between trade and FDI in terms of trends, it has also been employed to estimate bilateral FDI flows (see Brenton (1996), Eaton and Tamura (1996), and Brenton and Di Mauro (1999)). The general form of the gravity equation that I estimate is the following: Yij = + SUMGDPij + SIMSIZEij + RELENDOWij + DISTij + ∑ kDkij + ij α β β β β γ ε 1 2 3 4 ln (1) with the following variable definitions: Yij is the value of FDI or exports from country i (home country) to country j (host country); ( ) ij GDPi GDPj SUMGDP = ln + (2) + − + = − 2 2 ln 1 i j j i j i ij GDP GDP GDP GDP GDP GDP SIMSIZE (3) = − j j i i ij Pop GDP Pop GDP RELENDOW ln ln (4) DISTij is the relative distance between countries i and j. Following the advice of Polak (1996), I use here a measure of relative distance rather than absolute, i.e. the logarithm of actual distance divided by the average distance of the investing country from its partners, weighted by the shares of the latter in world GDP. This correction prevents the gravity model from producing biased results: a downward bias for far-away countries and an upward one for close-in countries. Dkij are dummy variables (mostly country dummies) used when appropriate and εij is the usual error term. A precision needs to be made at this point concerning the dependent variable FDI. In fact, the models à la Brainard and Markusen are mainly concerned with MNC activity, as opposed to that of the national exporter. With this respect, they correctly identify as the principal variable of interest - especially for the empirical analysis - the level of affiliate production abroad. This requires availability of such data, which in practice is limited - in a consistent way - to Sweden and the USA, typical case-studies of the literature (see e.g. Blomstrom et al. (1997), Brainard and Riker (1997), Markusen and Maskus (1999b) and Svensson (1996)). As my variable of interest is FDI, at least I make use of FDI stock, with the justification that these are used in the production process abroad, and hence constitute a good proxy for foreign affiliate production5. I now briefly discuss the impact of what I will call the ‘gravity variables’. The ‘economic space’ variable (SUMGDP) is expected to have a positive impact in both the FDI and exports equation. The index of size similarity (SIMSIZE) takes values between -∞ (i.e. the log of a number near zero) in case of perfect dissimilarity and -0.69 (the log of 0.5) for perfect similarity. Similarity in size should have a positive effect on exports: 5 A strong correlation certainly exists between FDI stock and foreign affiliates production, and I am implicitly assuming here that the relationship is linear, but it may not be. Further inquiry in this direction could prove useful and I envisage it for the future
THE IMPACT OF ECONOMIC INTEGRATION ON FDI AND EXPORTS countries similar in size will trade more,as the Helpman and Krugman theory of increasing returns predicts,trade is of intra-industry nature.I also expect a positive coefficient in the FDI equation,if the New FDI Theory holds true,as those models were motivated by the observation that FDI arises more among similar countries. Differences in relative endowments (RELENDOW)are measured here by the simple difference in GDP per capita;one could question the validity of this proxy,since,as noted in Helpman(1987),this method is accurate when there are only two factors of production (capital and labour)and all goods are freely traded.Better measures would be:GDP per worker,the ratio of capital(gross fixed capital formation)over working population,or that of skilled workers in total employment,as advocated by Wood (1994)for assessing the factor content of trade.However,given the poor reliability in employment and capital stocks data,especially for developing countries,I decided to stick with this definition of factor endowments.As far as the impact is concerned,a negative coefficient in the exports equation is a sign that Helpman and Krugman's theory of intra-industry trade (II'T) prevails:trade is not determined by differences in factor composition,as foreseen by traditional Heckscher-Ohlin inter-industry type trade models.For FDI,the story is similar: vertical FDI (equivalent to inter-industry trade)emerges as countries greatly differ in their factor composition -hence showing a positive coefficient,while horizontal FDI (comparable to IIT)is determined by similarity in factor composition,therefore displaying a negative coefficient.One cannot know a-priori which type prevails and the answer is an empirical one. As for distance,the effect on exports is clearly negative,being it a proxy for transport costs.On the other hand,FDI theory suggests that firms will invest abroad rather than export provided that trade costs are high.However,this variable may also have a negative coefficient in the FDI equation since the costs of operating overseas affiliates are still likely to rise the further they are from the MNCs headquarters.Overall though,I expect the coefficient in the exports equation to be higher than that in the FDI equation. Having defined the general gravity equation,I can now turn more specifically to the objective of the paper,ie.assessing the impact that economic integration has on FDI and on exports.The possible misspecifications in the gravity equation identified by Polak (1996)and Matyas (1997),especially when preferential trading blocs hypotheses are tested among a group of countries,are directly tackled here.By introducing some specific factors that characterise economic integration,rather than by just using dummy variables for countries belonging to the same regional integration area,I precisely try to avoid the problems identified by these authors.If a proper specification of the model is ensured then, it looks like the gravity equation represents a good way to proceed empirically. 6 Measuring this difference in absolute terms-as done by Helpman(1987)or Egger(2000)equates in practice to restricting the impact of this variable only to poritire differences,while in fact the rationale for it is just that amy large difference in factor endowments causes vertical FDI or inter- industry trade
THE IMPACT OF ECONOMIC INTEGRATION ON FDI AND EXPORTS 7 . countries similar in size will trade more, as the Helpman and Krugman theory of increasing returns predicts, trade is of intra-industry nature. I also expect a positive coefficient in the FDI equation, if the New FDI Theory holds true, as those models were motivated by the observation that FDI arises more among similar countries. Differences in relative endowments (RELENDOW) are measured here by the simple difference in GDP per capita6; one could question the validity of this proxy, since, as noted in Helpman (1987), this method is accurate when there are only two factors of production (capital and labour) and all goods are freely traded. Better measures would be: GDP per worker, the ratio of capital (gross fixed capital formation) over working population, or that of skilled workers in total employment, as advocated by Wood (1994) for assessing the factor content of trade. However, given the poor reliability in employment and capital stocks data, especially for developing countries, I decided to stick with this definition of factor endowments. As far as the impact is concerned, a negative coefficient in the exports equation is a sign that Helpman and Krugman’s theory of intra-industry trade (IIT) prevails: trade is not determined by differences in factor composition, as foreseen by traditional Heckscher-Ohlin inter-industry type trade models. For FDI, the story is similar: vertical FDI (equivalent to inter-industry trade) emerges as countries greatly differ in their factor composition – hence showing a positive coefficient, while horizontal FDI (comparable to IIT) is determined by similarity in factor composition, therefore displaying a negative coefficient. One cannot know a-priori which type prevails and the answer is an empirical one. As for distance, the effect on exports is clearly negative, being it a proxy for transport costs. On the other hand, FDI theory suggests that firms will invest abroad rather than export provided that trade costs are high. However, this variable may also have a negative coefficient in the FDI equation since the costs of operating overseas affiliates are still likely to rise the further they are from the MNCs headquarters. Overall though, I expect the coefficient in the exports equation to be higher than that in the FDI equation. Having defined the general gravity equation, I can now turn more specifically to the objective of the paper, i.e. assessing the impact that economic integration has on FDI and on exports. The possible misspecifications in the gravity equation identified by Polak (1996) and Matyas (1997), especially when preferential trading blocs hypotheses are tested among a group of countries, are directly tackled here. By introducing some specific factors that characterise economic integration, rather than by just using dummy variables for countries belonging to the same regional integration area, I precisely try to avoid the problems identified by these authors. If a proper specification of the model is ensured then, it looks like the gravity equation represents a good way to proceed empirically. 6 Measuring this difference in absolute terms - as done by Helpman (1987) or Egger (2000) equates in practice to restricting the impact of this variable only to positive differences, while in fact the rationale for it is just that any large difference in factor endowments causes vertical FDI or interindustry trade
FRANCESCA DI MAURO 2. Expected impact of economic integration Typically economic integration will be characterised by a few important changes affecting integrating countries,at various degrees,and they can be stylised as follows: 1.Reduction of tariffs(abolition of tariffs and adoption of a common external tariff,in the customs union case); 2.Reduction of Non-'Tariff Barriers (NI'Bs)-e.g.this was the aim of the Single Market Programme (SMP); 3.Potential greater exchange rate stability,as witnessed in the European Monetary System (EMS); 4.Greater efficiency in the allocation of resources,due to increased competition; The first two effects can be called commercial changes',since they are directly connected with trade policy.The third is related to 'monetary integration',while the last can be called 'market integration'.My focus will be on the first three,mainly because quantitative measures are readily available for these three effects and also because I have not investigated yet an aggregated measure of market integration.Nevertheless,reduction of NIBs can also be viewed as a general reduction of 'regulatory barriers,which in turn increases competition.The effect of market integration is therefore indirectly picked up, and what remains is the role of competition policy. 2.1 Impact of commercial changes Let's put ourselves in the case of a Regional Integration Area (RIA),as an extreme exemplification of what happens when countries deeply integrate among themselves.When analysing the impact of commercial changes,a main distinction should be made between members and non-members of the RIA.On the one hand we have the fact that a reduction of trade costs will directly benefit exports.If we assume that the tariff-jumping'argument holds (which in fact should be tested first)we would expect a reduction in FDI from the members of the RIA,relative to exports.That is,exports should become more profitable as a means of servicing the foreign market within the RIA.On the other hand though,but only in the case of a customs union,the opposite may become true for non-members.The trade diversion effect of the RIA creation leads previous exporters to directly invest in the RIA, in order to both avoid the tariff and access a larger market that is then free of tariffs. Smith (1987)analyses the effect on FDI of trade policy adjustments within a game- theoretical approach.MNCs act in a strategic way in concentrated markets,and may decide to invest rather than export into a country in order to avoid entry of domestic firms,in the presence of sunk costs.In his model,the impact of a reduction in tariffs is not clear a priori,and depending on the initial equilibrium,it may also increase FDI,thereby reversing the 'tariff-jumping'argument7. Blonigen and Feenstra(1996)look at the dynamic impact of a creation of a RIA,where it may act as a protectionist 'threat'against non-members and push MNCs to anticipate establishment into the bloc in order to avoid potential higher tariffs when exporting into it. A similar result is reached by Donnenfeld (1998),who constructs a model based on non- cooperative games,where the existence of MNCs may modify the optimal trade policy of 7 See Motta(1992)for an extension of Smith's model. 8
FRANCESCA DI MAURO 8 2. Expected impact of economic integration Typically economic integration will be characterised by a few important changes affecting integrating countries, at various degrees, and they can be stylised as follows: 1. Reduction of tariffs (abolition of tariffs and adoption of a common external tariff, in the customs union case); 2. Reduction of Non-Tariff Barriers (NTBs) – e.g. this was the aim of the Single Market Programme (SMP); 3. Potential greater exchange rate stability, as witnessed in the European Monetary System (EMS); 4. Greater efficiency in the allocation of resources, due to increased competition; The first two effects can be called ‘commercial changes’, since they are directly connected with trade policy. The third is related to ‘monetary integration’, while the last can be called ‘market integration’. My focus will be on the first three, mainly because quantitative measures are readily available for these three effects and also because I have not investigated yet an aggregated measure of market integration. Nevertheless, reduction of NTBs can also be viewed as a general reduction of 'regulatory barriers', which in turn increases competition. The effect of market integration is therefore indirectly picked up, and what remains is the role of competition policy. 2.1 Impact of commercial changes Let’s put ourselves in the case of a Regional Integration Area (RIA), as an extreme exemplification of what happens when countries deeply integrate among themselves. When analysing the impact of commercial changes, a main distinction should be made between members and non-members of the RIA. On the one hand we have the fact that a reduction of trade costs will directly benefit exports. If we assume that the ‘tariff-jumping’ argument holds (which in fact should be tested first) we would expect a reduction in FDI from the members of the RIA, relative to exports. That is, exports should become more profitable as a means of servicing the foreign market within the RIA. On the other hand though, but only in the case of a customs union, the opposite may become true for non-members. The trade diversion effect of the RIA creation leads previous exporters to directly invest in the RIA, in order to both avoid the tariff and access a larger market that is then free of tariffs. Smith (1987) analyses the effect on FDI of trade policy adjustments within a gametheoretical approach. MNCs act in a strategic way in concentrated markets, and may decide to invest rather than export into a country in order to avoid entry of domestic firms, in the presence of sunk costs. In his model, the impact of a reduction in tariffs is not clear a priori, and depending on the initial equilibrium, it may also increase FDI, thereby reversing the 'tariff-jumping' argument7. Blonigen and Feenstra (1996) look at the dynamic impact of a creation of a RIA, where it may act as a protectionist ‘threat’ against non-members and push MNCs to anticipate establishment into the bloc in order to avoid potential higher tariffs when exporting into it. A similar result is reached by Donnenfeld (1998), who constructs a model based on noncooperative games, where the existence of MNCs may modify the optimal trade policy of 7 See Motta (1992) for an extension of Smith's model