Multinational Enterprises and New Trade Theory: Evidence for the Convergence Hypothesis1 by Salvador Barrios* Holger Görg** Eric Strobl*** DOCUMENTO DE TRABAJO 2000-28

December, 2000

* University of Manchester

** University of Nottingham

*** University College Dublin

Los Documentos de trabajo se distribuyen gratuitamente a las Universidades e Instituciones de Investigación que lo solicitan. No obstante están disponibles en texto completo a través de Internet: http://www.fedea.es/hojas/publicaciones.html#Documentos de Trabajo

These Working Documents are distributed free of charge to University Department and other Research Centres. They are also available through Internet: http://www.fedea.es/hojas/publicaciones.html#Documentos de Trabajo

¹ We are grateful to participants of the ETSG 2000 conference in Glasgow for helpful comments on an earlier draft. All remaining errors are, of course, our own. This research has benefited from financial support through the “Evolving Macroeconomy” programme of the Economic and Social Research Council (ESRC Grant No. L138251002), the European Commission Fifth Framework Programme (Grant No. HPSE-CT-1999-00017) and the Leverhulme Trust (Grant No. F114/BF).

Abstract

According to the ‘convergence hypothesis’ multinational companies will tend to displace national firms and trade as total market size increases and as countries converge in relative size, factor endowments, and production costs. Using a recent model developed by Markusen and Venables (1998) as a theoretical framework, we explicitly develop, and address the properties of, empirical measures to proxy displacement of national by multinational firms between two countries. These empirical measures are then used to test the convergence hypothesis for a panel of data of country pairs over the years 1985-96. Our results provide some empirical support for the convergence hypothesis.

Classification codes: F21, F23

Introduction

High-income industrialised countries are both the most important source and destination for foreign direct investment (FDI). For example, in 1997, the total outward FDI stock of the Triad was equal to 2,441 billion US dollars and 63% of this amount concerned FDI stock within the Triad. In view of the substantial rise of both intra- and inter industry trade between industrialised countries, this feature may not seem that startling, but the important fact to consider is that FDI has grown far more than trade: in real terms, the total outward FDI stock rose at an average of 10.5% between 1988 and 1997 against an increase of 2.6% in trade². This evidence is in line with the observation by Markusen (1998, p.753), that "most of the growth in North Atlantic economic activity since the early 1980s has been in investment, not trade". These trends, thus, mean that in the industrialised world multinational production is replacing national production, at least in the sense of supplying goods between countries. For example between 1984 and 1995 UNCTAD (1997) estimated that sales by multinational enterprises (MNE) were higher than the total exports of goods and services.

In looking for a theoretical rationale for these patterns one observes that much of the New Trade Theory (NTT) has expended its efforts on providing support for the increased importance of trade between industrialised countries and the prevalence of intra-industry specialization between them, rather than the growing importance of multinationals relative to trade (Markusen and Venables, 1998).³ The theoretical challenge in terms of the observed pattern of production of MNEs, however, lies in attempting to explain the existence of MNE within the general equilibrium theory of trade. Put differently, one needs models to explain why some firms choose to invest abroad rather than export. To achieve this task trade economists have mainly relied on Dunning's OLI framework (1977) as a starting point. Accordingly, MNEs are seen as firms which internalise a specific ownership advantage that provides them with some market power. Firms are willing to exploit this through FDI instead of exports in order to benefit from some location advantage and to avoid possible asset dissipation that may occur with, for example, licensing.

This line of reasoning has resulted in a (relatively) small number of theoretical models within the NTT framework that can explain some of the observed pattern of multinational production: see, for example, the pioneering analyses of Markusen (1984), Ethier (1986), Helpman (1984 and 1985) and Brainard (1993). In these models firms are seen as being willing to engage in direct investment instead of alternatives such as exporting if firm-level economies of scale are important relative to plant-level economies. This may be the case if, for example, R&D activity is important for the firm, as R&D has some of the characteristics of a public good; in particular, the output of R&D can be transferred between different plants within the firm at low or zero costs (Markusen, 1995). In addition, FDI may occur to avoid prohibitive transport costs related to trade and to benefit from proximity advantages. FDI may then displace trade when countries are relatively similar in both size and factor endowments, as pointed out by Markusen and Venables (1998).⁴

² These figures are taken from the UN International Trade Statistics Yearbook (1997) and the World Investment Report (1999). Trade figures are for the period 1990-1997 and concern merchandises only. The Triad refers to the EU, the US and Japan.

³ For example, Krugman (1979) and Helpman and Krugman (1985) suggest that trade between countries with similar factor proportions is likely to be mostly in differentiated products and increasing return to scale activities, which contrasts clearly with the Heckscher-Ohlin (HO) neoclassical model where trade occurred as a consequence of factor proportion differences.

The model developed by Markusen and Venables (1998) probably provides the most coherent framework within which to analyse the increasing importance of FDI relative to trade in the world economy. Using a two-country model they show that the convergence of countries in size and relative endowments shifts the regime from national to multinational firms, a phenomena termed the ‘convergence hypothesis’ according to which “multinational production will tend to displace national firms and trade as the two countries converge in (a) relative size, (b) relative factor endowments, and (c) relative production costs” (Markusen and Venables, 1996, p. 172)5.

Our paper utilises the Markusen-Venables (1998) model to define and discuss empirical equations with which to analyse this convergence hypothesis. Specifically, using these equations, we investigate how total market size, differences in the market size and factor endowments of host and home countries, transport costs and plant level scale economies (relative to firm level scale economies) impact on the level of two way multinational activity between the two countries. To this end we make use of a panel dataset for the period 1985 to 1996, which allows us to analyse changes in bilateral investment behaviour between a set of OECD countries over time.

There have been a number of recent papers which are related to our work. Ekholm (1995) examines the level and determinants of foreign production by home i firms in host j using data on FDI stocks for a number of OECD countries and also analyses the determinants of two-way multinational activity between countries i and j. Ekholm (1997, 1998) extends this work using employment data for the US and Sweden. She finds that foreign production and the level of two-way MNE activity are positively related to similarities in GDP and relative endowments of human capital between the host and home country, while similarities in the endowments of physical capital do not seem to affect foreign production to any great extent. Also, total market size is found to be a positive determinant of multinational activity. Using data on US multinationals, Brainard (1997) investigates the determinants of exports vs. sales by multinationals in the host country. She finds that multinational production abroad is found to increase with higher transport costs and trade barriers, and the less important are plant level scale economies. Complementary work by Markusen and Maskus (1999) and Carr et al. (2000), also using US data, furthermore highlights the importance of market size, size differences, and differences in factor endowments between host and home country for the multinationals’ decision to produce abroad for the host country market or for exports to third country markets.⁶

⁴ These models refer to horizontal FDI where foreign affiliates produce similar goods to those produced in the home country in order to exploit firm specific economies of scale.

⁵ As the authors point out this is not simply due to trade disappearing with the convergence in relative factor endowments and production costs

We extend this work in at least two ways. Firstly, we analyse bilateral multinational activity for a number of OECD country pairs. Secondly, we link our indices explicitly to the Markusen and Venables (1998) model and the convergence hypothesis, i.e., the displacement of indigenous firms by multinationals when countries become more similar in terms of size and endowments. The remainder of this paper is structured as follows. In Section 2 we outline the theoretical framework upon which our empirical analysis is based. We discuss the empirical measurement of the convergence hypothesis in Section 3 and describe the dataset used in Section 4. Section 5 contains an outline of our empirical methodology and Section 6 presents the results of the econometric estimations. Concluding remarks are provided in the final section.

2 Theoretical Framework

The present section gives a brief review of the structure and the main results of the model by Markusen and Venables (1998), hereafter referred to as MV. We do not provide a detailed account of MV model; for further details, the reader should refer to the original model and to the outline provided by Barrios et al. (2000). Our objective is to provide a basis for the test of the convergence hypothesis interpreting the MV main findings in terms of MNE’s employment. Hence, we focus on the main equations necessary to allow us to derive an index of displacement based on employment data.

⁶ There have, of course, been numerous other empirical studies of FDI and the activities of multinational companies in the literature. These studies, however, mainly examine one way multinational activity only, i.e., the dependent variable is the unilateral activity of multinationals from home to host country, and they do not explicitly examine the effect of size and endowment differences between host and home country. See, for example, Kravis and Lipsey (1982), Culem (1988), Wheeler and Mody (1992), Head et al. (1995), Barrell and Pain (1996, 1999).

The MV model derives the conditions necessary for multinationals to dominate within a general equilibrium framework of trade with increasing return to scale and imperfect competition (the NTT framework) in two countries: h and . Each economy is perfectly identical and consists of two industries producing homogenous goods (X,Y) and using two production factors: L (labour) and R (resources). L is mobile between X and Y industries but internationally immobile. L is used in both sectors but R is used in Y-sector only. Both goods are traded but only X entails a positive transport cost between h and which is represented by a variable quantity of L used in transport activities. The Y-industry is perfectly competitive and Y is produced under constant return to scale while X-industry is characterized by Cournot-type competition and the equilibrium is defined by free entry and exit of firms with zero profits. FDI only occurs in the X-sector which is hence the main focus of the analysis. The variable under scrutiny is the total number of multinational firms, m, and national firms, n, in equilibrium, where subscripts are added to identify the number of these each due to each source country.

In a Cournot model with homogenous products, each firm produces the same quantity and behaves exactly the same way as the other firms within the same category. Firms sell their product in both countries and behave as follows:

firms sell in market j through exports. These firms incur an additional variable cost, the cost related to the transportation of X from i to j represented by Fixed costs may be decomposed into a firm-level and a plant-level fixed cost. The firm-level fixed cost is represented by an amount of labour F needed for organizational activities, R&D etc., and the plant-specific fixed cost is derived to the use of an additional quantity of labour G, needed for productive activities. All these costs are incurred in country i.

firms sell in market j through FDI. These firms incur an additional fixed cost through FDI, where the additional fixed cost relates to the new plant in j. As a consequence, multinationals have the same firm-specific fixed labour requirement F and a plant-specific fixed labour requirement equal to both employed in the home country i plus an additional fixed labour requirement G in the host country j where it locates its affiliate to sell X.

Each firm-type is identified with a particular cost function according to the preceding features. The different cost functions between type-n and type-m firms determine in turn different demands for labour. Type-n firms located in i, have the following demand for labour:

⁷ Symbols i and j are used throughout the paper in order to avoid the repetition of each equation for each country h and f since both of these countries is considered to be identical.

⁸ We assume that the transport cost is the same both for shipping goods from i to j and from j to i.

\[l _ {i} ^ {n} = c X _ {i i} ^ {n} + (c + t) X _ {i j} ^ {n} + G + F \qquad i, j = h, f, i \neq j\tag{1}\]

where and denote the quantities sold in countries i and respectively, the c represents the constant marginal labour requirement.

Multinationals have a different demand for labour since their productive activity is divided between h and f. The term represents the total demand for labour of a multinational headquartered in country i and it can be decomposed into and , that represent the demand in home and host country respectively:

\[l _ {i} ^ {m} = l _ {i i} ^ {m} + l _ {i j} ^ {m} \quad i, j = h, f a n d i \neq j\tag{2}\]

\[l _ {i i} ^ {m} = c X _ {i i} ^ {m} + G + F \quad i = h, f\tag{3}\]

\[l _ {i j} ^ {m} = c X _ {i j} ^ {m} + G \quad i, j = h, f \quad a n d i \neq j\tag{4}\]

Accordingly, multinationals avoid transportation costs but incur twice the plant-fixed cost of national firms since they use a fixed quantity of labour equal to 2G instead of G, the quantity of labour used by type-n firms. The differences in cost functions in turn determine the differences in the relative profitability of each firm type. With free entry of firms in the market of X, profits are zero so that the relative performance of each firm-type falls on the number of active firms. All other variables of interest, such as trade or employment can be derived from the equilibrium number of firms.

Using this theoretical framework, Markusen and Venables show that MNEs, as outlined in Appendix 1, will have an advantage relative to type-n firms when:

1. The overall market is large.

2. The markets are of similar size.

3. Labour costs are similar.

4. Firm-level scale economies are large relative to plant-level scale economies.

5. Transport costs are high.

These results are just a re-statement of the convergence hypothesis since, as noted by Markusen and Venables (1998, p.196) "convergence of countries h and f in either size or relative endowments shifts the regime from national to multinational firms". We now turn to the practicalities of the empirical examination of the convergence hypothesis using this model.

(L_iY).

ni.lin)

⁹ In each country, labour is also employed by type-n firms (with a total demand represented by and by firms in Y-sector

3 Measuring the convergence hypothesis

From an empirical perspective, the preceding results raise several issues. The first one is a problem common to many empirical analyses of general equilibrium trade models: for theoretical tractability these models are usually designed to describe the relationships between two countries, while in the real world, trade and foreign investment concern a large number of countries (see Bowen et al., 1998). In the special case of MNE activities, an investor in one country may be primarily concerned by the accession to a third (neighbouring) market. This is especially true in the case of FDI in the European Union where the reduction of market fragmentation and barriers to trade between countries led firms to reorganize their activity within the area in order to benefit from the potential gain of an integrated market.¹⁰ In our empirical analysis we take account of this by estimating the model not only for our total sample, but also separately for EU countries only.

The second issue is a data related problem. As Markusen and Venables point out, displacement of national by multinational firms could be expressed in terms of the following index:

\[I _ {M - V} = \frac {m _ {h} + m _ {f}}{m _ {h} + m _ {f} + n _ {h} + n _ {f}}\tag{5}\]

that is, the share of multinationals in the total number of firms. This measure will depend on the values of the exogenous parameters that determine the relative profitability of each firm-type. Hence, the MV model and its results are specifically concerned with the number of multinational firms (relative to indigenous firms). This may prove problematic in terms of measuring the convergence hypothesis for several reasons. Firstly, the number of active firms may not reflect the relative importance of MNE. In the real world, firms have different production scales and the number of MNE is not a reliable measure of MNE importance. In contrast, employment data are likely to be better able to reflect the importance of MNE, because they take into account the relative size of MNE. Secondly, employment data are a better measure of MNE presence than FDI flows or stocks since a growing share of investment is made with funds raised locally (Markusen, 1998).

Ideally, therefore, the equivalent of the share of multinational firms in terms of employment should be:

¹⁰ See European Commission (1996) for evidence on the impact of the Single Market Programme on the location of industries in the EU.

\[I _ {M - V} ^ {\text {empirical}} = \frac {(m _ {h} l _ {h h} ^ {m} + m _ {h} l _ {h f} ^ {m} + m _ {f} l _ {f f} ^ {m} + m _ {f} l _ {f h} ^ {m})}{(n _ {h} l _ {h} ^ {n} + n _ {f} l _ {f} ^ {n} + m _ {h} l _ {h h} ^ {m} + m _ {h} l _ {h f} ^ {m} + m _ {f} l _ {f f} ^ {m} + m _ {f} l _ {f h} ^ {m})}\tag{6}\]

that is, the proportion of total employment between two countries h and f that is due to MNEs. However, the data typically used in empirical studies relate to total employment in multinationals in the host country (and not in the home country) only, so that one cannot calculate the numerator in (6) because there is no information concerning and The closest approximation to (6) that one instead may be able to calculate is therefore:

\[Y _ {1} = \frac {\left(m _ {h} l _ {h f} ^ {m} + m _ {f} l _ {f h} ^ {m}\right)}{\left(n _ {h} l _ {h} ^ {n} + n _ {f} l _ {f} ^ {n} + m _ {h} l _ {h h} ^ {m} + m _ {h} l _ {h f} ^ {m} + m _ {f} l _ {f f} ^ {m} + m _ {f} l _ {f h} ^ {m}\right)}\tag{7}\]

which is simply the share of MNEs’ employment in the host country of total employment in both countries. However, this index raises two problems. The first relates to the fact that it can take on high values even when there is little bilateral FDI. Consider, for example, that one country is much larger than the other, and because of this size effect, in presence of increasing returns to scale, MNEs will preferably locate in the larger country. We allow for this problem by introducing a second index which only considers the minimum of the share of FDI of both countries, i.e.,

\[Y _ {2} = \min \left(\frac {\left(m _ {f} l _ {f h} ^ {m}\right)}{\left(n _ {h} l _ {h} ^ {n} + m _ {h} l _ {h h} ^ {m} + m _ {f} l _ {f h} ^ {m}\right)}, \frac {\left(m _ {h} l _ {h f} ^ {m}\right)}{\left(n _ {f} l _ {f} ^ {n} + m _ {f} l _ {f f} ^ {m} + m _ {h} l _ {h f} ^ {m}\right)}\right)\tag{8}\]

The second problem arises from the hypothesis made concerning the demand for labour described by (2)-(4). A closer look at the theoretical model reveals that we cannot directly use the MV results to interpret either or

To clarify this point one may consider a simple case. From the model one knows that each MNE uses a fixed number of workers equal to in their home country and a (lower) quantity of labour equal to in the host country. This implies that when multinationals dominate, the demand for labour in the home country increases by more than the demand for labour in the host country. The immediate consequence is that we cannot draw clear-cut conclusions on the behaviour of and from the MV results. When the equilibrium number of multinationals rises, the denominator of the preceding indices can rise more than its numerator depending on how MNE employment in their home country behaves with respect to national firms’ employment. The model does not provide a clear answer on this point and one has to rely on the empirical estimates provided in the next section to shed light on the issue.

Alternatively we consider a third index that can be more directly interpreted following the basic model:

\[Y _ {3} = 1 - \frac {\left| m _ {f} l _ {h f} ^ {m} - m _ {h} l _ {f h} ^ {m} \right|}{m _ {f} l _ {h f} ^ {m} + m _ {h} l _ {f h} ^ {m}}\tag{9}\]

This index refers to the notion of cross-country FDI and it can be directly interpreted using the MV framework since foreign employment by multinationals will rise according to the increase in the equilibrium number of type-m firms.¹¹ Utilising the MV model we show elsewhere (Barrios et al., 2000) that will increase if total income increases, the distribution of income falls, the difference in labour costs falls, and if transport costs rise. However, the effect of changes in the firm-specific to plant-specific cost ratio is theoretically ambiguous. Most importantly, it must realised that this third index only can be considered to be a direct test of the convergence hypothesis if the number of national firms is taken as given by the relative profitability of each firm-type.

4 Data Set

In order to assess the convergence hypothesis empirically, our primary variable is the extent of FDI between country pairs. It is constructed from the OECD data base Measuring Globalisation: The Role of Multinationals in OECD Economies 1999 Edition.¹² This data source provides a set of detailed statistical data for assessing and analysing the role played by multinationals in the industrial sectors of 16 OECD countries constructed from national sources. For the purpose of this paper, we need to construct measures of bilateral FDI, in terms of employment, between country pairs and thus need bilateral foreign employment between any two countries in the same year. Given the nature of the data, this was only possible for the manufacturing sector as a whole, rather than individual subsectors. We were able to do so for 27 country pairs over a number of years, providing us with a data sample size of 118 for which we have provided summary statistics of our three proxies of displacement in Table 1. The denominators of and were computed using data on countries’ total employment in manufacturing from the Stan Database (OECD). This provides data on a basis that is compatible with the OECD database on Multinational activities.

As can be seen, there is a considerable variety of country pairs, all of which are between developed countries. The sample years and size also differ for each country pair. In terms of our displacement proxies, varies considerably across these, from 1.66 to 0.01, for the country pairs Netherlands-Sweden and Japan-Sweden, respectively.¹³ The correlation coefficient between and is 0.96, implying that both may be a good proxy for both displacement of indigenous industry by MNEs and bilateral FDI. In contrast, the correlation coefficients between and and are 0.69 and 0.55, respectively. Thus may not be a particularly good proxy of displacement.

¹¹ This index is in fact similar to the intra-industry trade index proposed by Grubel and Lloyd (1975). A similar index was used by Ekholm (1995, 1997, 1998).

¹² This data was previously published as ‘Activities of Foreign Affiliates in OECD Countries’.

Table 1: Summary Statistics of Y1, Y2, and Y3 (means) for Country Pairs

*Multiplied by 100

Y1

Y2

¹³ For both the summary statistics and the regression results we have multiplied both Y₁ and Y₂ by 100.

5 Econometric Methodology

In order to estimate empirically the effect of the above factors on the intensity of cross-direct investment we propose the following basic model,¹⁴

\[\begin{array}{l} Y _ {i j} = \beta_ {0} + \beta_ {1} [ \ln (G D P _ {i}) + \ln (G D P _ {j}) ] + \beta_ {2} (a b s (G D P _ {i} - G D P _ {j})) + \beta_ {3} ((a b s (S E N D _ {i} - S E N D _ {j})) + \\ + \beta_ {4} ((a b s (C E N D _ {i} - C E N D _ {j})) + \beta_ {5} [ \ln (R D _ {i}) + \ln (R D _ {j}) ] + \beta_ {6} (D I S T _ {i j}) + \beta_ {7} (L A N G _ {i j}) + e _ {i j} \\ \text {(16)} \end{array}\]

where is the two-way multinational activity between countries i and j measured in terms of employment using the three indices as discussed in Section 3, is gross domestic product of country is country relative endowment of skilled labour, is country relative endowment of physical capital, is a proxy of country research intensity, is the distance between the capitals of i and and is a dummy variable set equal to one if countries i and j have a common language.

The first four terms on the right hand side of the equation relate to the issue of size and relative endowments. They are included to test the predictions that multinational employment between two countries increases as (i) total market size increases and countries grow more similar in (ii) size and (iii) relative endowments. We measure relative endowments of skilled workers using two alternative proxies. As a first measure, we calculate the share of employment in R&D activity relative to total employment in the country. We also calculate the percentage of students enrolled in secondary school education in the total population in the country as a second proxy. Relative physical capital endowments are measured as per capita capital stock in the country.

The MV model also predicts that multinational employment can be expected to be more important if firm-level scale economies are large. Therefore, R&D intensity is included in the empirical model to proxy the importance of firm-level scale economies. A country pair’s R&D intensity is calculated as the sum of each country’s per capita R&D expenditures. This in some way captures the "knowledge-capital model" referred to by Markusen (1995) in which multinationals are seen as firms exploiting some ownership advantage through investment abroad. The ability to exploit such advantages is more likely in industries in which knowledge intensive production is important. As a consequence, multinationals are associated with high ratios of R&D relative to sales and employ a large proportion of qualified workers. From an empirical perspective this implies that multinationals are more likely to exist in countries were industry is R&D-intensive. In the MV model this variable corresponds to F, the firm-specific fixed cost.

¹⁴ The definition of the explanatory variables and their data sources are summarised in appendix 3.

Furthermore, the MV model suggests that multinational companies can be assumed to be important relative to national firms if transport costs and trade barriers are high. We include the distance between the two countries as a rough proxy in the empirical formulation to take account of this. Finally, the equation also includes a dummy variable which is set equal to one if countries share a common language, since a common language can be assumed to reduce transaction costs of setting up subsidiaries abroad and should, therefore, favour multinational production.

It should be noted that equation (16) is similar to a gravity equation as used in empirical work in international trade (see, for example, Bergstrand, 1985, McCallum, 1995, Frankel et al., 1998) and, more recently, in the analysis of the activities of multinational companies (Ekholm, 1995, 1997, 1998, Brainard, 1997, Markusen and Maskus, 1999, Carr et al., 2000). While the theoretical foundations for using gravity equations in trade are, however, debatable (see Deardorff, 1984 and Evenett and Keller, 1998) the MV model appears to provide a coherent theoretical framework for the use of gravity equations for the analysis of the activities of multinational companies.

6 Econometric Results

Our results for estimating (10) using standard OLS techniques for our total sample for the three different indices and are given in Table For all three indices we find that total market size (GDP) and the absolute difference in relative market size (ABSGDP) are significantly positive and negative determinants of bilateral multinational activity respectively, although in the case of ABSGDP does not turn out to be statistically significant. Despite this flaw, these results are in line with the convergence hypothesis, i.e., the share of bilateral FDI of total manufacturing employment of country pairs increases as total market size increases and as countries become more similar in size.

The evidence on the effect of differences in relative endowments is less clear-cut, however. While the first proxy of endowments of skilled workers, SEND1, is statistically insignificant in all specifications, SEND2 turns out to be significantly negative in the case of but positive and significant in the case of using as the dependent variable. Differences in physical capital intensity turn out positive in all specifications but are only statistically significant in the specifications that utilise SEND2. We thus find evidence for the convergence hypothesis in terms of relative endowments only for the specification using and SEND2 as proxy of relative human capital endowments, and even in that case, although for this specification the result on the proxy of physical capital endowments is not as one would expect under the convergence hypothesis.^16,17

¹⁵ One could argue that given the panel nature of our data set we should, perhaps, have controlled for country pair fixed effects. However, we chose not to do so for two reasons. Firstly, given our sample size and the large number of country pairs this would have considerably reduced the degrees of freedom. Secondly, given that for each country pair the time period covered is generally only a few years and many of our explanatory variables, such as relative market size or relative factor endowments, will not have varied to any great extent over short time periods, controlling for country pair fixed effects would have purged from our equation or randomised exactly what we are trying to measure. Hence, our estimation always assumes that country pair specific fixed unobservables are not correlated with the explanatory variables.

Table 2: Total Sample

Notes: 1. Heteroskedasticity consistent standard error (White, 1980) in parentheses 2. *, **, *** signify ten, five, and one per cent significance levels, respectively.

In all specifications, we find a positive and significant coefficient on the proxy for firm level economies of scale, RD. As noted earlier, the nature of our independent variable did not allow us, within the framework of the MV model, to predict a priori what effect this proxy for the importance of firm-level fixed costs would have on displacement. However, our results indicate that the relationship is clearly positive and statistically significant.

¹⁶ We also re-ran the regressions including only one measure of resource endowments, i.e., either SEND1, SEND2 or CEND. The results, which are not reported here, are qualitatively similar to the ones reported herein, however.

¹⁷ One should note, however, that Markusen (1998) in his review of the literature also states that, "a high volume of outward direct investment is positively related to a country's endowment of skilled labour and insignificantly or negatively related to its physical capital endowment" (p. 736).

Our results also show that countries with a common language experience greater displacement. In contrast, the greater the distance between two country’s capitals the lower the degree of displacement will be. If distance is an appropriate proxy for transportation costs, then this latter result is contrary to the convergence hypothesis. However, Markusen (1998, p. 736) notes that there is weak evidence that FDI is primarily motivated by the avoidance of tariffs or measurable transport costs. This may particularly be the case for vertical FDI, where the production of intermediate goods is located in low-cost locations but the final good is assembled in the home country. In this case, the multinational firm may wish to locate as near as possible to the home country in order to avoid transportation costs for shipping the intermediate good between host and home country. Also, Brenton et al. (1999) note that distance may be negatively related to FDI since the costs of operating overseas are likely to rise with distance because of, for example, higher communication costs and higher costs of placing personnel abroad.

In the present study, the opposite sign for distance could also possibly be due to the fact that our sample contains mainly European firms with strong cross-investment during the period under study.¹⁸ Moreover, the distance between countries may not be a perfect proxy for transport costs. For instance, it may also reflect cultural differences between countries - countries that are further apart may also be culturally more distinct from each other and, as Kumar (2000) finds for US and Japanese FDI, foreign investment is positively related to cultural proximity.

Part of the drawback of estimating (10) for our total sample of country pairs is that we are pooling within EU, across the EU and outside the EU data. The nature of displacement may, however, be intrinsically different for country pairs within these three groupings. To examine how this may affect our results we also sub-divided our sample into those with bilateral FDI within the EU and those for which at least one country was not in the EU at the time. Note, however, that particularly for the EU sample, all results must be viewed with some caution given the relatively small sample size.

¹⁸ The increased level of integration has translated into a significant decrease in barriers for goods, services and factor movements within the Union, lowering the segmentation of markets and enhancing the rationalization of production as noted by Muchielli and Burgenmeier (1991). The fall in trade barriers may act in favour of vertical FDI since intermediate goods are more easily traded within the firm. The fact that investment occurs between developed countries means that there is still a scope for production rationalization and cost minimization within the firm even if labour costs are similar between locations. Labour skills may be diversified across countries or regions and this in turn may provide a rationale for relocation of productive activities when trade barriers are being removed. On the other hand, factor costs may still be different between industrialized countries, even if those differences are lower than comparing to developing countries. The combination of low internal barriers to trade and still high barriers to external trade may provide a valid reason for cost reduction through FDI in low-wages countries when factor intensities between stages of production differ.

The results for estimating (10) for the sample of EU country pairs only are given in Table 3. Accordingly, reducing our sample to bilateral FDI within the EU does not alter our results in terms of relative market size, distance or R&D intensity, although the coefficients on distance and R&D intensity turn out to be statistically insignificant in the case of .¹⁹ Total market size also turns out only to be statistically significant and positive in two out of six cases. In terms of relative factor endowments we find that both SEND1 and CEND are statistically significant for the dependent variable but both have a positive effect. Differences in physical capital endowments also turn out positive and statistically significant for although the proxies for human capital endowments are statistically insignificant in that case.

Table 3: EU-country pairs

Notes: 1. Heteroskedasticity consistent standard error (White, 1980) in parentheses 2. *, **, *** signify ten, five, and one per cent significance levels, respectively.

For the non-EU sample, the results of which are reported in Table 4, the proxy for differences in physical capital endowments are found be statistically insignificant in all specifications, while differences in country size and the proxy for R&D intensity are only significant in one and three specifications, respectively, although with similar signs as in the overall sample. Our findings are much in line with the overall sample, however, for the total market size, human capital endowments, language and distance variables.

¹⁹ The language dummy is dropped for the EU sample because none of the country pairs share a common language.

Table 4: Non-EU country pairs

Notes: 1. Heteroskedasticity consistent standard error (White, 1980) in parentheses

2. *, **, *** signify ten, five, and one per cent significance levels, respectively. ， 2

As discussed above, distance may not be a perfect proxy for transport costs. In the absence of detailed data on the actual level of transport costs between countries, we, therefore, in an extension to the model suggested in equation (10), replace the distance variable with an indicator of the importance of trade flows between two countries. This may serve as a proxy for transport costs if we assume that countries trade more the lower are transport costs. Specifically, we calculate the variable as , where are exports from country i to j, and are imports into country i from country j.

The results of this exercise are reported in Table 5. While the overall results are similar to the ones reported previously, we find that the trade variable has a positive sign and is statistically significant in four out of six cases. This is again in contrast to the predictions by the MV model, according to which high transport costs favour displacement. The positive sign then provides further evidence that this relationship is more complex for a number of reasons – not the least because trade and FDI may be complements rather than substitutes (Markusen, 1983).

Table 5: Extensions – Openness Indicator

Notes: 1. Heteroskedasticity consistent standard error (White, 1980) in parentheses

2. *, **, *** signify ten, five, and one per cent significance levels, respectively.

In the estimations thus far we have assumed that it is the size of the individual country that matters. Arguably, however, it may be the case that for EU countries it is not the size of the individual EU member country, but the size of the total EU that attracts FDI (both from extra-EU and intra-EU sources) to serve the large EU market (see Görg and Ruane, 1999 for a discussion). In an alternative specification we re-calculated both GDP and ABSGDP to take account of this. GDP is calculated as total EU GDP if we analyse a pair of EU countries, the sum of total EU GDP and individual country GDP in the case of one EU country and one non-EU country, and the sum of individual country GDP for two non-EU countries. ABSGDP is calculated similarly.

The results of the estimations using these measures of GDP and ABSGDP are presented in Table 6. While the coefficients on ABSGDP are negative and statistically significant as expected in all cases the results on the GDP variable are rather disappointing. While it turns out to be statistically significant and positive for the total sample, breaking the sample up into EU country pairs and non-EU country pairs yields statistically insignificant results in all cases. Assuming the convergence hypothesis is correct, this casts doubt on the assertion that it is the total EU market size, rather than individual country size that matters for investment decisions, although again we have to be cautious in drawing conclusions due to the small sample sizes, particularly for the EU sample.

Table 6: Extensions – EU GDP

Notes: 1. Heteroskedasticity consistent standard error (White, 1980) in parentheses 2. *, **, *** signify ten, five, and one per cent significance levels, respectively. 2 2

7 Conclusion

In this paper we investigate whether multinational companies tend to displace indigenous firms and trade as countries become more similar in size, factor endowments and production costs and as total market size increases as predicted by the convergence hypothesis. We set out to test the convergence hypothesis using a panel of OECD country pairs over the period 1985-96.

Our results support the convergence hypothesis to some extent. Overall market size tends to increase, while differences in market size tend to reduce bilateral MNE activity. While the role of differences in relative endowments of human or physical capital skilled workers is not clear from our results, R&D intensity, which serves to proxy the importance of firm level scale economies, and a common language in home and host country significantly increase displacement. We also find that for many cases transportation costs, contrary to the convergence hypothesis, are negative determinants, although these findings are in line with similar findings in the literature. Breaking down our sample into EU and non-EU pairs we find that a large number of our results in aggregate still hold, although, given the small sample size, particularly for EU country pairs, these results must be viewed with some caution.

There are, of course, a number of factors that may weaken our findings. Firstly, our dependent variables are exclusively calculated for the manufacturing sector in its entirety. We are thus very likely aggregating over very heterogeneous sub-sectors. Secondly, our definition of market size, in absolute or relative terms, may not always be appropriate. Particularly, in view of trade agreements, the market in which a MNE locates may no longer be the appropriate one to consider, but may merely serve as a base to provide goods to other countries within these arrangements. Thirdly, our definitions of resource endowments cannot be considered optimal, which may explain at least partly our inconclusive results regarding the role of differences in factor endowments. Finally, our data set refers exclusively to developed countries. It would prove insightful to either include or separately estimate the determinants of bilateral FDI with regard to developing countries. These issues will hopefully be addressed as better and more comparable data become available in the future.

Appendix 1: Further details concerning MV results.

Following MV, we can derive separately the impact of the exogenous variables on the equilibrium number of firms holding the other endogenous variables constant. Accordingly, increases when:

The overall market is large since FDI is a better option than exporting to reach foreign markets given the increasing return nature of the model. Multinationals avoid transport costs but have to cope with higher fixed costs, they have then to reach higher production levels than national firms to obtain non-negative profits. This constraint is being relaxed when market size allow multinational firms to cover their fixed costs with markup revenue. This can be formally represented by the impact of an increase in income for both countries as stated by MV:

\[d M _ {h} = d M _ {f} > 0 \Rightarrow d m _ {h} = d m _ {f} > d n _ {h} = d n _ {f} \geq 0\]

The markets are of similar size, because different market sizes would provide a clear advantage for firms to locate in the large market and export a marginal part of their product to the disadvantaged (small) country. Type-n firms advantage would translate into type-m firms relative disadvantage and some (or all) of them would have to leave the market. From MV this means that:

\[d M _ {h} = - d M _ {f} > 0 \Rightarrow d n _ {h} > d m _ {h} = d m _ {f} = 0 > d n _ {f}\]

Labour costs are similar, then the factor proportion differences and H-O arguments becomes less important for countries’ specialization. With significant factor proportions differences countries would specialize according to their relative endowment in L and R and national firms would locate in the advantaged (i.e. labour abundant) country. Export is a better option in this case since multinationals are constraint to produce in high cost countries by definition. The impact of a larger labour costs difference could be represented as:

\[d w _ {f} = - d w _ {h} > 0 \Rightarrow d n _ {h} > d m _ {h} > 0 > d m _ {f} > d n _ {f}\]

Firm-level scale economies are large relative to plant-level scale economies, i.e. F is large relative to G. Since multinationals incur twice the plant-fixed cost w.G, then they benefit from lowers G, i.e., relatively high F while national firms are less affected by this. Multinational profitability would then rise relatively to national firms. This can be represented formally by the impact of a symmetric change in plant vs. firm-specific cost:

\[d F = - d G > 0 \Rightarrow d m _ {h} = d m _ {f} > 0 > d n _ {f} = d n _ {h}\]

Transport costs are high, in this case export is simply a costly option compared to FDI. Following MV, we have:

\[d t > 0 \Rightarrow d m _ {h} = d m _ {f} = 0 > d n _ {h} = d n _ {f}\]

Appendix 2: Definitions of the Variables and data sources:

References

1. Barrell, R. and Pain, N. 1996, “An Econometric Analysis of U.S. Foreign Direct Investment”, Review of Economics and Statistics, Vol. 78, pp. 200-207.
2. Barrell, R. and Pain, N. 1999, “Domestic Institutions, Agglomerations and Foreign Direct Investment in Europe”, European Economic Review, Vol. 43, pp. 925-934.
3. Barrios, S., Görg, H. and Strobl, E. 2000, “Multinational Enterprises and New Trade Theory: Evidence for the Convergence Hypothesis”, GLM Research Paper, University of Nottingham.
4. Bergstrand, J.H. 1985, “The Gravity Equation in International Trade: Some Microeconomic Foundations and Empirical Evidence”, Review of Economics and Statistics, Vol. 67, pp. 474-481.
5. Bowen H.P., Holander, A. and Viaene, J.M. (ed.) 1998, Applied International Trade Analysis, London: MacMillan.
6. Brainard, S.L. 1993, A simple theory of multinational corporations and trade with a trade-off between proximity and concentration, NBER working paper 4269.
7. Brainard, S.L. 1997, “An Empirical Assessment of the Proximity-Concentration Trade-off between Multinational Sales and Trade”, American Economic Review, Vol. 87, pp. 520-544.
8. Brenton, P., Di Mauro, F. and Lücke, M. 1999, “Economic Integration and FDI: An Empirical Analysis of Foreign Investment in the EU and in Central and Eastern Europe”, Empirica, Vol. 26, pp. 95-121.
9. Carr, D.L., Markusen, J.R. and Maskus, K.E. 2000, “Estimating the Knowledge-Capital Model of the Multinational Enterprise”, American Economic Review, forthcoming.
10. Culem, C.G. 1988, “The Locational Determinants of Direct Investments among Industrialized Countries”, European Economic Review, Vol. 32, pp. 885- 904.
11. Deardorff, A.V. 1984, “Testing Trade Theories and Predicting Trade Flows”, in: Jones, R. and Kenen, P. (eds), Handbook of International Economics Vol. 1 (Amsterdam: Elsevier), pp. 467-517.
12. Dunning, J.H. 1977, “Trade, Location of Economic Activity and MNE: A search for an Eclectic Approach”, in: Ohlin, B. Hesselborn, P.O. and Wijkman, P.M. (eds), The International Allocation of Economic Activity (London, MacMillan), pp. 395-418.
13. Ekholm, K. 1995, “Multinational Production and Trade in Technical Knowledge”. Lund Economic Studies No. 58, Lund.
14. Ekholm, K. 1997, “Factor Endowments and the Pattern of Affiliate Production by Multinational Enterprises”, CREDIT Research Paper 97/19, University of Nottingham.
15. Ekholm, K. 1998, “Proximity Advantages, Scale Economies, and the Location
16. of Production”, in: Pontus Braunerhjelm and Karolina Ekholm (eds.), The Geography of Multinationals (Kluwer Academic Publishers, Dordrecht), pp. 59-76.
17. Ethier, W.J. 1986, “The multinational firm”, Quarterly Journal of Economics, Vol. 80, pp. 805-834.
18. European Commission 1996, Economic Evaluation of the Internal Market, European Economy no4, Directorate General for Economic and Financial Affairs, Brussels.
19. Evenett, S.J. and Keller, W. 1998, On Theories Explaining the Success of the Gravity Equation, NBER working paper 6529.
20. Frankel, J.; Stein, E. and Wei, S. 1998, “Continental trading blocs: are they natural or supernatural?”, in J. Frankel (ed.), The Regionalization of the World Economy. Chicago University Press, Chicago, pp. 91-113.
21. Görg, H. and Ruane, F. 1999, „US Investment in EU Member Countries: The Internal Market and Sectoral Specialization”, Journal of Common Market Studies, Vol. 37, pp. 333-348.
22. Grubel, H.G. and Lloyd, P.J. 1975, Intra Industry Trade, Macmillan: London.
23. Head, K.; Ries, J. and Swenson, D. 1995, “Agglomeration Benefits and Location Choice: Evidence from Japanese Manufacturing Investments in the United States”, Journal of International Economics, Vol. 38, pp. 223- 247.
24. Helpman, E.M. 1984, “A simple theory of international trade with multinational corporations”, Journal of Political Economy, Vol. 92, pp. 451-471.
25. Helpman, E.M. 1985, “Multinational corporations and trade structure”, Review of Economic Studies, Vol. 52, pp. 443-457.
26. Helpman, E.M. and Krugman, P.R. 1985, Market Structure and Foreign Trade, Cambridge, MIT Press (ed.).
27. Kravis, I.B. and Lipsey, R.E. 1982, “The Location of Overseas Production and Production for Export by U.S. Multinational Firms”, Journal of International Economics, Vol. 12, pp. 201-223.
28. Krugman, P.R. 1979, “Increasing returns, Monopolistic Competition, and International Trade”, Journal of International Economics, Vol. 9, pp.469- 480.
29. Kumar, N. 2000, “Explaining the Geography and Depth of International Production: The Case of US and Japanese Multinational Enterprises”, Weltwirtschaftliches Archiv, Vol. 136, pp. 442-477.
30. Markusen, J.R. 1983, "Factor Movements and Commodity Trade as Complements". Journal of International Economics. Vol. 14. pp. 341- 356.
31. Markusen, J.R. 1984, “Multinationals, multi-plant economies and the gains from trade”, Journal of International Economics, Vol. 16, pp.205-226.
32. Markusen, J.R. 1995, “The boundaries of Multinational Enterprises and the Theory of International Trade”, Journal of Economic Perspectives, Vol. 9, pp.169-189.
33. Markusen, J.R. 1998, “Multinational Firms, Location and Trade”, World Economy, Vol. 21, pp. 733-56.
34. Markusen, J.R. and Maskus, K.E. 1999, “Multinational Firms: Reconciling Theory and Evidence”, NBER Working Paper 7163.
35. Markusen, J.R. and Venables, A.J. 1996, “The Increased Importance of Direct Investment in North Atlantic Economic Relationships: A Convergence Hypothesis”, in Cazoneri, M., Ethier, W., and Grilli, V. (eds.), The New Transatlantic Economy (Cambridge University Press: Cambridge).
36. Markusen, J.R. and Venables, A.J. 1998, “Multinational Firms and the New Trade Theory”, Journal of International Economics, Vol. 46, pp. 183- 203.
37. McCallum, J. 1995, “National borders matter: Canada-U.S. regional trade patterns”, American Economic Review, Vol. 85, pp. 615-623.
38. Muchielli, J.L. and Burgenmeier, B. 1991, Multinationals and Europe 1992: Strategies for the future, London, Routledge.
39. Organisation for Economic Co-Operation and Development, 1999(a), Measuring Globalisation: The role of multinationals in OECD economies, OECD, Paris.
40. Organisation for Economic Co-Operation and Development, 1999(b), Science, Technology and Industry Scoreboard, Benchmarking Knowledge based economies, OECD, Paris.
41. Organisation for Economic Co-Operation and Development 1999(c), Research and Development Expenditure in Industry, OECD, Paris.
42. United Nations 1997, International Trade Statistics Yearbook, U.N. New York. United Nations 1999, World Investment Report, U.N. New York.
43. Wheeler, D.and Mody, A. 1992, “International Investment Location Decisions”, Journal of International Economics, Vol. 33, pp. 57-76.
44. White, H. 1980, “A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity”, Econometrica, Vol. 48, pp. 817- 830.

RELACION DE DOCUMENTOS DE FEDEA

COLECCION RESUMENES

98-01: “Negociación colectiva, rentabilidad bursátil y estructura de capital en España”, Alejandro Inurrieta.

TEXTOS EXPRESS

2000-03: “Efectos sobre la inflación del redondeo en el paso a euros”, Mario Izquierdo y Simón Sosvilla-Rivero.

2000-02: “El tipo de cambio Euro/Dolar. Encuesta de FEDEA sobre la evolución del Euro”, Simón Sosvilla-Rivero y José A. Herce.

2000-01: “Recomendaciones para controlar el gasto sanitario. Otra perspectiva sobre los problemas de salud”, José A. Herce.

DOCUMENTOS DE TRABAJO

References

1. 2000-28: “Multinational Enterprises and New Trade Theory: Evidence for the Convergence Hypothesis”, Salvador Barrios, Holger Görg y Eric Strobl.

References

1. 2000-27: “Obsolescence Vs modernization in a Schumpeterian vintage capital model”, Raouf Boucekkine, Fernando del Río y Omar Licandro.

References

1. 2000-26: “Provisión de servicios públicos y localización industrial”, Luis Lanaspa, Fernando Pueyo y Fernando Sanz.

References

1. 2000-25: “Labor Force Participation and Retirement of Spanish Older Men: Trends and Prospects”, Namkee Ahn y Pedro Mira.

References

1. 2000-24: “Paridad del poder adquisitivo y provincias españolas, 1940-1992”, Irene Olloqui y Simón Sosvilla-Rivero.

References

1. 2000-23: “Optimal Growth under Endogenous Depreciation, Capital Utilization and Maintenance Costs”, Omar Licandro, Luis A. Puch y J. Ramón Ruiz-Tamarit.

References

1. 2000-22: “Expectativas, Aprendizaje y Credibilidad de la Política Monetaria en España”, Jorge V. Pérez-Rodríguez, Francisco J. Ledesma-Rodríguez, Manuel Navarro-Ibáñez y Simón Sosvilla-Rivero.

References

1. 2000-21: “Población y salud en España. Patrones por género, edad y nivel de renta”, José Alberto Molina y José A. Herce.

References

1. 2000-20: “Integration and Growth in the EU. The Role of Trade”, José A. Herce y Mª Luz García de la Vega.

References

1. 2000-19: “Foreign Direct Investment and Productivity Spillovers”, Salvador Barrios.

References

1. 2000-18: “Female Employment and Occupational Changes in the 1990s: How is the EU Performing Relative to the US?, Juan J. Dolado, Florentino Felgueroso y Juan F. Jimeno.

References

1. 2000-17: “Do tobacco taxes reduce lung cancer mortality?, José Julián Escario y José Alberto Molina.

References

1. 2000-16: “Solution to Non-Linear MHDS arising from Optimal Growth Problems”, J. R. Ruiz-Tamarit y M. Ventura-Marco.

References

1. 2000-15: “El sistema de pensiones contributivas en España: Cuestiones básicas y perspectivas en el medio plazo”, Juan Francisco Jimeno.

References

1. 2000-14: “Assessing the Credibility of the Irish Pound in the European Monetary System”, Francisco Ledesma-Rodríguez, Manuel Navarro-Ibáñez, Jorge Pérez-Rodríguez y Simón Sosvilla-Rivero

References

1. 2000-13: “La utilidad de la econometría espacial en el ámbito de la ciencia regional”, Esther Vayá Valcarce y Rosina Moreno Serrano.