The Location of American and Japanese Multinationals in Europe

 

 

 

 

 

 

Hideki Yamawaki

 

 

 

August 2005

 

 

 

 

 

 

 

 

 

Jean-Marc Thiran and Luca Barbarito helped greatly in the development of the data used in this paper.  The author gratefully acknowledges the valuable comments and suggestions of two anonymous referees.

 

 

 

 

 

 

 

Peter F. Drucker and Masatoshi Ito Graduate School of Management

Claremont Graduate University

1021 North Dartmouth Avenue

Claremont, CA 91711

USA

 

Tel: +1 909 607 8494

Fax: +1 909 621 8543

E-mail: hideki.yamawaki@cgu.edu

 

 


 

 Abstract

 

This paper examines the determinants of location choice of American and Japanese foreign direct investment in European manufacturing industries in the early 1990s.  The new data used in this study covers 340 Japanese-owned affiliates and 2312 US-owned affiliates distributed across 38 regions and 7 member states in the European Union.  The statistical analysis finds, most importantly, that the location decisions of US and Japanese MNEs are not the same.  The Japanese results are consistent with the behavior that firms consider production-cost factors more important than demand-side factors, suggesting Japanese firms’ motive to establish local production capacity to export within the EU market.  On the contrary, both cost-side and demand-side factors are found important determinants of location choices for US firms.  The results also suggest that location decisions are industry- specific.

 

JEL codes: F23, F15, R32

 

Key words: firm location, multinational firms, European integration


 

1.  Introduction

 

               When the European Communities completed the 1992 program to form the European Union (EU), it was expected that the creation of a single enlarged market provides incentives for European as well as non-European firms to organize pan-European manufacturing activities and locate their manufacturing facilities in the regions that suit best to serve the single market.  The questions of where multinational enterprises (MNEs) establish production platforms within the EU and what factors determine their location choices were hence particularly relevant for several reasons.  From a MNE’s point of view, location strategy obviously determines the performance of its European and global operations.  For the government of potential host country, the location decision of MNEs is considered important because foreign direct investments influence the host country’s prospects for regional development.  Despite this preeminence, there is a scarcity of empirical research attempting to identify the determinants of location decisions within the EU (see Yamawaki, 1993; Thiran and Yamawaki, 1995; Devereux and Griffith, 1998; Mayer and Mucchielli, 1998; and Ferrer, 1998)[1]. 

This paper is unique in two ways.  First, we focus on regional characteristics as determinants of MNEs’ location decisions in the EU.  While state characteristics were used as the determinants of foreign direct investment in the United States in the existing literature (e.g. Bartik, 1985; Glickman and Woodward, 1988; Coughlin, Terza, and Arromdee, 1991; Woodward, 1992; Head, Ries, and Swenson, 1995 and 1999; and Head, Ries and Ruckman, 1998), they were seldom used in previous research that examines MNEs’ location decisions in Europe (Hill and Munday, 1992; and Thiran and Yamawaki, 1995; and Ferrer, 1998).  Second, we use data for both American and Japanese MNEs in the EU.  Previous work examines location decisions of MNEs originating from one country (Devereux and Griffith, 1998 for US firms; Yamawaki, 1993, and Mayer and Mucchielli, 1998 for Japanese firms; and Ferrer 1998 for French firms) and does not consider how location decisions are different between MNEs originating from different countries.

After the hypothesis and statistical model are presented in the next section, the data are described in the third section of the paper.  A conditional logit model estimating the choice of locations in the EU for both US and Japanese samples is presented in the fourth section of the paper.  Finally, in the last section, a summary and conclusion from the empirical results is provided. The Japanese results are consistent with the behavior that firms consider production-cost factors more important than regional demand-side factors, suggesting Japanese firms’ motive to establish local production capacity to export within the EU market.  On the contrary, both cost and regional demand factors are found important determinants of location choices for US firms.  The statistical analysis also shows that MNEs’ location decisions differ between industries, suggesting they are industry-specific.   

 

2. Empirical Model

 Following the existing literature on location decisions (Head et al. 1995 and 1999; and Devereux and Griffith, 1998), we assume that a firm chooses a specific region to invest when it yields the highest profits.  Since profits P are latent variables and not observed, the chosen alternative is the one with the highest value.  If the disturbances associated with each choice is assumed independent, the probability that region i is chosen by the investor has the form[2]

Prob (i is chosen) = exp(Zib) / ĺj exp(Zjb)                     (1)

where b is the vector of parameters to be estimated, Zi are conditional variables that determine profits P = Zb + u, and j = 1,….,n.  Non-exporting affiliates compare the profitability of each of the regions.  Their choice depends on the region-specific demand factor and region-specific cost-side factors and taxes.  On the contrary, exporting affiliates that chooses a location to export to the EU market compare the profitability of each of the regions mainly based on cost-side factors.  To the extent that their revenue from exporting to the EU market does not vary across different production locations, the demand factor is common to all the exporting regions and does not affect the location decision of exporting affiliates.  In this case conditional variables Z include only cost-related factors, taxes, and transportation costs. 

   Our dependent variable is the location choices, and the independent variables are the characteristics of the EU regions classified at the NUTS 1 level of classification.  In the case of pure exporting affiliates, we expect that the coefficient for the region-specific demand factor is insignificant and dominated by the influence of cost-side factors.

 

Cost Factors

Labor Market Conditions

Previous research on MNEs’ location decisions within countries has used mainly three variables, wages, the unionization rate, and the unemployment rate, to measure labor market conditions (Glickman and Woodward, 1988; Coughlin et al., 1991; Woodward, 1992; Friedman et al., 1992; and Head et. al, 1999).  The extent of unionization is found to have conflicting influences on the choice of regions.  While Bartik (1985) argues that high union activity will deter foreign direct investment because of the restrictions imposed by union contracts, Coughlin et al. (1991) found evidence that higher level of unionization attracts foreign direct investment in the United States.  They explain this unexpected result by the fact that higher levels of unionization are associated with higher productivity efficiency (Beeson and Husted, 1989) .  And, Coughlin et al. (1991) found that a high unemployment rate gives a signal to the potential MNE of the availability of labor and thus attracts it to invest in the region.

The effects of labor costs on MNEs’ location decisions were found erratic in previous research.  Bartik (1985) and Coughlin et al. (1991) found a negative influence of labor costs on the choice of regions in the United States, but Hill and Munday (1992) did not confirm this effect for British regions.  Head et al. (1999) finds the effects of labor costs on Japanese investors’ location choices in the United States are sensitive to the inclusion of agglomeration variables in their econometric specifications.[3]     

The effects of labor conditions on MNEs’ location choices are examined in this study by four variables: wages (WAGES), unemployment rate (UNEMP), level of education of labor force (EDUC), and unionization rate (UNION).  The three variables, WAGES, UNEMP, and EDUC are defined at the regional-level, while UNION is constructed at the national-level due to the unavailability of regional data.  The coefficients of UNEMP might have a positive sign if MNEs choose a region where a large labor pool of potential workers, and hence they can screen their qualifications better in the hiring process.  Regions with high unemployment rates may engage in some measures to attract investment (Coughlin et al., 1991; Head et al., 1999).  The coefficient for UNION may be negative if unions insist on restrictive work rules that lower labor productivity (Bartik, 1985; Head et al, 1999).  We expect the coefficient for WAGES has a negative sign.  And, the general qualification of potential labor force in the region is controlled by EDUC.

 

Technological Capability

Another cost-side factor that determines MNEs’ location decisions is technological capability of the host country.  For MNEs that manufacture technologically sophisticated products, it is crucial for local affiliates to employ skilled workers and engineers, and procure technologically advanced parts and components from local suppliers. They may also organize local research and development to develop new processes and products and adapt existing products to local standards and tastes.  Host countries and regions well endowed with skilled labor and engineers and clustered with technologically capable suppliers, and hence those with high R&D capacity in general, will therefore attract foreign MNEs (Cantwell, 1989; Kogut and Chang, 1991; Neven and Siotis, 1996; and Blonigen, 1997).

Since a region’s technological capability is not easily measured, a proxy variable (TECH) is constructed by assuming that R&D output defined by the number of patents granted measures the level of technological capability.  As data on patents for European regions are not available, the number of patents granted at the national level is used by assuming that the national difference in technological capability dominates the regional difference as the determinant of MNEs’ location decisions.  The coefficient for TECH is hence expected to be positive.

 

Transportation Costs

The existence of highly developed transport network in the region has been found as a significant factor to attract foreign direct investment (Bartik, 1985; Glickman and Woodward, 1988; Couglin et al, 1991; and Hill and Munday, 1992).  In a study examining the trade-off between achieving proximity to customers and concentrating production to achieve scale economies, Brainard (1997) found that overseas production by US-owned affiliates abroad increases relative to exports the higher are transport costs and the lower are scale economies at the plant level.  This result suggests that MNEs are likely to concentrate production if transportation costs are low and plant-level scale economies are important and hence that a region with low transportation costs attracts foreign investment by export-oriented MNEs.  This hypothesis is tested by a variable (TRANS) that identifies the density of transport network.  TRANS is constructed from principal component analysis by utilizing the correlations within a set of three observed variables that measure the access to sea transport (SEA), the access to airport (AIR), and the access to highway network (ROAD).  Assuming each variable is represented as a vector, we derived a factor that represents “density” of transportation network with strong positive loadings on SEA, AIR, and ROAD.  If a region’s accessibility to transport network is negatively related to transportation costs of shipping goods from that region to the rest of the EU, the coefficient of TRANS is expected to have a positive sign.

 

Agglomeration

Previous research on location choice proposes the hypothesis that investors are attracted to the region where manufacturing activity is dense because of the existence of a cluster of potential customers, workers, and suppliers and thus of the agglomeration economies (Krugman, 1991; Fujita, Krugma, and Venables, 1999)[4].  Previous study that examines Japanese direct investments showed that Japanese investors are likely to be attracted to US states and European countries with other Japanese plants in the same industry and other Japanese affiliates in the same corporate group (Woodward, 1992; Smith and Florida, 1994; Head et al., 1995 and 1999; Belderbos and Sleuwaegen, 1996; and Mayer and Mucchielli, 1997).  While the agglomeration effects are found important, they are highly correlated with the effects of labor market conditions and other conditional variables (Head et al., 1999; and Mayer and Mucchielli, 1997).  We use a variable that measures the density of manufacturing activity in the region, AGGLO, to control for the agglomeration effects in the statistical analysis.  To the extent that the agglomeration effects reduce production costs, the coefficient for AGGLO is expected to have a positive sign.

 

Demand Factors

  We expect that the income of consumers in the host region should have a positive effect on the MNE’s location choice if it sets up an affiliate for local sales.  On the other hand, the host region’s income level should not affect the location choice if the local affiliate is established for exporting.  In this case, cost-side factors should dominate demand-side factors.  To test this hypothesis we use a variable that measures GDP per capita of the region, GDP[5].

 

Government Policy

 Government intervention in the form of subsidy was found most important stimuli for foreign direct investment in the US states (Coughlin et al., 1991; and Head et al., 1999) as well as in the British regions (Hill and Munday, 1992).  The effects of corporate taxes on the location decision of US MNEs in Europe were found to be highly significant (Devereux and Griffith, 1998).  To control for the effect of tax rate, we include the corporate tax rate defined at the national level (TAX) in the empirical specification[6].  The coefficient for TAX is expected to have a negative sign.

 

3. Data

Japanese investors’ location choices are identified from the database constructed from a corporate directory published by Toyo Keizai.  This source lists the manufacturing subsidiaries more than 10 % controlled by Japanese firms that were distributed among 15 European (EU members and non-EU) countries in 1993.  From this list, we eliminated subsidiaries located outside the EU (16% of all the European manufacturing subsidiaries listed in Toyo Keizai[7]) and those that were unable to match with a region at the NUTS-1 level of EU regions due to the lack of some specific information[8].  

Because of the unavailability of comparable data source for US MNEs, we are obliged to identify the locations of US MNEs in European manufacturing industries by coding the corporate directories compiled and provided by the American Chamber of Commerce in each member state of the EU.   The Chambers of Commerce in only seven member states (Belgium, Germany, France, Ireland, Italy, the Netherlands, and the United Kingdom) provided us with the lists of US MNEs at the time of data collection[9]. We then classified the locations of European affiliates of US MNEs according to the NUTS-1 level of regional classification as in the case of Japanese MNEs.  Since this data source does not provide any information on sales and employment, we are unable to construct any sorts of quantitative measures.

 These two separate data sets provided us with the database from which our data for statistical analysis is constructed.  In this original database the total counts of European affiliates of US MNEs are 3528, while the total counts for Japanese MNEs are 432.  The final sample for statistical analysis became smaller in size because we had to deal with two issues on data.  First, in an attempt to make these two data sets comparable in terms of industry composition of MNEs, we matched the two data sets according to the two-digit level of the European industry classification system (NACE).  The industrial distribution of Japanese investment is highly skewed and concentrated in several industries, particularly in the machinery sector.  We thus had to drop 631 US subsidiaries and 35 Japanese subsidiaries from the original samples to mach the US industry with the Japanese industry.  Second, we further eliminated 585 US subsidiaries and 57 Japanese subsidiaries from the samples due to the unavailability of regional-specific variables for several regions[10].    This procedure left us with 2312 American subsidiaries and 340 Japanese subsidiaries in the sample for statistical analysis.  The lists of regions included in the samples for statistical analysis are provided in Table 1.

 

4. Statistical Results

Table 2 presents the estimation results of a conditional logit model for all manufacturing industries.  For the US sample, all the coefficients for the independent variables are significant.  Among the cost-side factors, the coefficients for WAGES, TECH, and TRAN have all the expected signs, suggesting that US MNEs are likely to choose a region where wages are lower, transport network is more easily accessible, and the country where technological capability is higher.  The coefficient for GDP is significant and has a positive sign, suggesting that for the US MNEs the size of regional market is an important determinant for their location choices. The effect of TAX on the location choice is highly significant and is negative as expected.

On the contrary, in the Japanese sample, the coefficients for TECH, WAGES, and GDP are all insignificant, while the coefficients for TAX, UNEMP, UNION, and TRAN are statistically significant.  The coefficient for TAX has a negative sign, and the coefficient for TRAN has a positive sign as expected.  It appears that for the Japanese MNEs the size of regional market is not an important determinant.[11]

Table 3 presents the industry-specific estimation results for the US and Japanese samples.  In our conditional logit model constant is not allowed as a conditional variable and not identified since it does not vary across choices.  For this reason, it is not feasible to include industry dummies to control for industry-specific effects.  Instead, we estimate the model by splitting the sample into industries.  Since Japanese and US foreign direct investments in Europe are most common in the machinery sector (NACE 32, 33, 34, and 35), we first estimate the logit model for this sector.  The numbers of observations are 1368 for the US sample and 246 for the Japanese sample.  The US result (equation 3) remains virtually unchanged from the result using the full sample of Table 2.  On the contrary, the Japanese result (equation 1) changed importantly in Table 3 from the results of Table 2.  While the effects of TAX, UNEMP, UNION, and TRAN on location decisions remain unchanged, the coefficient for TECH became statistically significant for the Japanese firms in the machinery sector.  The coefficient for TECH has a positive sign, suggesting that the Japanese machinery firms are attracted to the country where technological capability is high.  The coefficient for GDP remains insignificant for the group of Japanese machinery firms in equation 1 in Table 3.  This result further implies that the location choice of the Japanese firm is industry-specific. 

The results in Table 3 reinforce our earlier findings that location decisions are different between the US and Japanese firms.  It is likely that US firms in machinery determine a location in the EU based on both the cost-side factors and the demand-side factors.  On the contrary, Japanese firms in the same industry are likely to choose a location based mainly on the cost-side factors.  They prefer a region where transportation costs are lower, technological capability is higher, the level of unionization is lower, and taxes are lower. 

Equations 2 and 4 in Table 3 add the agglomeration variable, AGGLO, to the specification.  As expected, the coefficients for AGGLO have positive signs and are statistically significant in both the US and Japanese samples.  This result suggests that MNEs are attracted to regions where manufacturing activities are concentrated.  While the coefficient for AGGLO suggests strong agglomeration effects, this variable dampens the technology factor in the Japanese sample.  In Equation 2 the coefficient for TECH is no longer significant, suggesting that the agglomeration factor is compounded of the technology factor.  Similarly, AGGLO dampens the effects of technology and labor costs in the US sample as shown in Equation 4.  The effect of local demand remains significant and positive for US firms even after AGGLO is added in the specification.

Since US firms are also highly concentrated in chemicals and pharmaceuticals after non-electrical machinery, equations 5 and 6 in Table 3 presents the estimation results for chemicals and pharmaceuticals (NACE 25) for the US sample.  The result for the Japanese sample is not shown since their sample size is too small (29) to have a statistically significant result.  Comparing equations 3 and 5 for the US sample, the most important industry-specific difference appears in the effects of labor costs and technological capability.  While low labor costs and high technological capability of the host region attracts US investors in machinery, that is not the case for chemicals. 

 

 

 

5. Conclusions

This paper examined if the location decisions of MNEs originating form different source countries are the same.  The statistical results found evidence that supports the hypothesis that Japanese-owned affiliates in the EU are export-oriented and hence choose production locations based on regions’ cost-side characteristics.  On the other hand, US-owned affiliates in the EU are attracted to large local demands as well as low-cost sites.      

The findings of this paper provide some implications.  The difference in location decisions between US and Japanese firms may be reflecting the underlying difference in their motives of foreign direct investment.  Japanese firms may choose a location to establish an export base from which they export to other regions in Europe.  US firms have chosen their production locations near the customers.  There are at least two reasons for such difference in investment behavior.  US MNEs were historically attracted to low-information cost European countries such as United Kingdom where common language is used (Davidson, 1980; Veuglers, 1991).  Table 1 shows the tendency of US firms to rank English-spoken countries as major destination countries in Europe.[12]

The second explanation is given by the policy measures that the European Communities imposed in the pre-1992 period.  While the appreciation of the Japanese yen against the US dollar is considered triggered massive outflow of investment from Japan to the United States (Blonigen, 1997), Japanese direct investment in Europe in the late 1980s may be caused by additional factors.  Trade policy measures imposed by the European Communities such as antidumping and other trade restrictions motivated Japanese firms exporting from Japan to Europe to shift their production to Europe to circumvent such trade restriction (Belderbos and Sleuwaegen, 1998).  To the extent that such “tariff jumping” investment substituted for exports from Japan, their location choices should focus on the production-cost factors.  Yet another factor is the 1992 program itself.  The formation of a common market in Europe motivated Japanese firms, latecomers in foreign direct investment, to set up production capacity suited for exporting in the enlarged market (Heitger and Stehn, 1990; Jacquemin and Buigues, 1991).  On the contrary, the history of US foreign direct investments in Europe goes back much earlier period of the Common Market (Vernon, 1971).

Finally, the statistical analysis of this paper shows importantly that MNEs’ location decisions differ between industries and they are different between different groups of firms within the same industry.  


 

References

 

Bartik, T.J., 1985, “Business Location Decisions in the United States: Estimates of the Effects of Unionization, Taxes, and Other Characteristics of States”, Journal of Business and Economic Statistics, pp. 14-22.

 

Beeson, P.E. and S. Husted, 1989, “Patterns and Determinants of Productive Efficiency in State Manufacturing”, Journal of Regional Science, pp. 15-28.

 

Belderbos, R.A., and L. Sleuwaegen, 1996, “Japanese Firms and the Decision to Invest Abroad: Business Groups and Regional Core Networks,” Review of Economics and Statistics, pp. 203-231.

 

Belderbos, R.A., and L. Sleuwaegen, 1998, “Tariff Jumping DFI and Export Substitution: Japanese Electronics Firms in Europe,” International Journal of Industrial Organization, pp. 601-638.

 

Blonigen, B.A., 1997, “Firm-Specific Assets and the Link Between Exchange Rates and Foreign Direct Investment,” American Economic Review, pp. 447-465.

 

Brainard, S.L., 1997, “An Empirical Assessment of the Proximity-Concentration Trade-Off Between Multinational Sales and Trade,” American Economic Review, pp. 520-544.

 

Buijink, W., B. Janssen, and Y. Schols, 2000, Corporate Effective Tax Rates in the EU and the OECD: Effective Tax Rates for Listed Companies in OECD-Countries, MARC, Faculty of Economics and Business, Maastricht University.

 

Cantwell, J., 1989, Technical Innovations in Multinational Corporations, London: Basil Blackwell.

 

Carlton, D.W., 1983, “The Location and Employment Choices of New Firms: An Econometric Model with Discrete and Continuous Endogenous Variables”, The Review of Economics and Statistics, pp. 440-449.

 

Coughlin, C.C., J.V. Terza, and V. Arromdee, 1991, “State Characteristics and the Location of Foreign Direct Investment Within the United States”, The Review of Economics and Statistics, pp. 675-683.

 

Culem, C.G., 1988, “The Locational Determinants of Foreign Direct Investments Among Industrialized Countries”, European Economic Review, pp. 885-904.

 

Davidson, W., 1980, “The Location of Foreign Direct Investment Activity, Country Characteristics and Experience Effects”, Journal of International Business Studies, pp. 9-22.

 

Devereux, M., and R. Griffith, 1998, “Taxes and the Location of Production: Evidence from a Panel of US Multinationals,” Journal of Public Economics, pp. 335-367.

 

Dunning, J.H., 1980, “Toward an Eclectic Theory of International Production: Some Empirical Tests”, Journal of International Business Studies, pp. 9-31.

 

Ferrer, C., 1997, “Pattern and Determinants of Location Decisions by French Multinationals in European Regions,” in A.M. Rugman and J.-L. Mucchielli (eds.), Research in Global Strategic Management, London: JAI Press.

 

Friedman, J., D.A. Gerlowski, and J. Silberman, 1992, “What Attracts Foreign Multinational Corporations? Evidence from Branch Plant Location in the United States”, Journal of Regional Science, pp. 403-418.

 

Fujita, M, P. Krugman, and A.J.  Venables, 1999, The Spatial Economy:  Cities, Regions, and International Trade, Cambridge, MA: MIT Press.

 

Glickman, N.J. and D.P. Woodward, 1988, “The Location of Foreign Direct Investment in the United States: Patterns and Determinants”, International Regional Science Review, pp. 137-154.

 

Head, K.R., J. C. Ries, and D. Swenson, 1995, “Agglomeration Benefits and Location Choice: Evidence from Japanese Manufacturing Investment in the United States”, Journal of International Economics, pp. 223-247.

 

Head, K.R., J.C. Ries, and D. Swenson, 1999, “Attracting Foreign Manufacturing: Investment Promotion and Agglomeration,” Regional Science and Urban Economics, pp. 197-218.

 

Head, K., J. Ries, and K. Ruckman, 1997, “Industry Agglomeration and the Location of Foreign Affiliates,” in A.M. Rugman and J.-L. Mucchielli (eds.), Research in Global Strategic Management, London: JAI Press.

 

Heitger, B., and J. Stehn, 1990, “Japanese Direct Investment in the E.C.: Response to Internal Market 1993?,” Journal of Common Market Studies, pp. 1-15.

 

Hill, S. and M. Munday, 1992, “The UK Regional Distribution of Foreign Direct Investment: Analysis and Determinants”, Regional Studies, pp. 535-544.

 

Jacquemin, A., and P. Buigues, 1991, “Foreign Direct Investments and Exports in the Common Market: Theoretical, Empirical, and Policy Issues,” Center for European Policy Studies, Working Documents.

 

Kogut, B., and S.-J. Chang, 1991, “Technological Capabilities and Japanese Foreign Direct Investment in the United States,” Review of Economics and Statistics, pp. 401-413.

 

Krugman, P., 1991, Geography and Trade, Cambridge, MA: MIT Press.

 

McFadden, D., 1973, “Conditional Logit Analysis of Qualitative Choice Behavior,” in P. Zarembka (ed.), Frontiers in Econometrics, New York: Academic Press.

 

Mayer, T., and J.-L. Mucchielli, 1997, “Agglomeration Effects, State Policies, and Competition in the Location of Japanese FDI in Europe,” in A.M. Rugman and J.-L. Mucchielli (eds.), Research in Global Strategic Management, London: JAI Press.

 

Morsink, R.L.A. and W.T.M. Molle, 1991, “Direct Investment and European Integration”, paper presented at the conference “Direct Investment in Europe”, Universite catholique de Louvain, March, 1991.

 

Neven, D., and G. Siotis, 1996, “Technology Sourcing and FDI in the EC: An Empirical Evaluation,” International Journal of Industrial Organization, pp. 543-560.

 

Smith, D.F. and R. Florida, 1994, “Agglomeration and Industrial Location: An Econometric Analysis of Japanese-Affiliated Manufacturing Establishment in Automotive-Related Industries”, Journal of Urban Economics, pp. 23-41.

 

Swedenborg, B., 1979, The Multinational Operations of Swedish Firms: An Analysis of Determinants and Effects, Stockholm: The Industrial Institute for Economic and Social Research.

 

Thiran, J.-M. and H. Yamawaki, 1995, “Regional and Country Determinants of Location Decisions: Japanese Multinationals in European Manufacturing,” IRES, Universite catholique de Louvain, Discussion Paper Nr. 9517.  

 

Vernon, R., 1971, Sovereignty At Bay The Multinational Spread of U.S. Enterprises, New York: Basic Books.

 

Veugelers, R., 1991, “Locational Determinants and Ranking of Host Countries: An Empirical Assessment”, Kyklos, pp. 363-382.

 

Woodward, D.P., 1992, “Locational Determinants of Japanese Manufacturing Start-ups in the United States,” Southern Journal of Economics, pp. 690-708.

 

Yamawaki, H., 1993, “Location Decisions of Japanese Multinational Firms in European Manufacturing Industries”, in K. Hughes (ed.), European Competitiveness, Cambridge: Cambridge University Press.

 

Yamawaki, H., 1994, “Entry Pattern of s of Japanese Multinationals in US and European Manufacturing,” in M. Mason and D. Encarnation (eds.), Does Ownership Matter? Japanese Multinationals in Europe, Oxford: Oxford University Press.


 

Table 1. Numbers of US and Japanese Affiliates by European Regions, 1993.

 

REGIONS

NUTS 1 & 2

Sample for Statistical Analysis

Full Sample

 

 

 

 Japan

 

US

 

 Japan

 

US

 

BELGIUM

B

 

 

30

417

COMMUNAUTE FLAMANDE

B1

13

196

15

229

REGION WALLONNE

B2

 7

55

17

91

BRUXELLES

B3

8

68

 8

97

DENMARK

DK

 

 

2

NA

GERMANY

ALL

 

 

70

320

SCHLESWIG-HOLSTEIN

ALL15B

 

 

 0

1

HAMBURG

ALL6

 3

14

 3

21

NIEDERSACHSEN

ALL9

7

15

17

18

NORDRHEIN-WESTFALEN

ALL10B

17

65

 20

79

HESSEN

ALL7

56

 10

71

BAYERN

ALL2

15

54

 16

63

BADEN-WURTEMBERG

ALL1

 

 

 9

46

BERLIN

ALL3

 

 

 0

6

BREMEN

ALL5

 

 

 5

13

RHEINLAND-PFALZ

ALL11B

 

 

 0

0

SACHSEN

ALL13B

 

 

 0

2

GREECE

GR

 

 

2

NA

SPAIN

ESP

 

 

34

NA

NOROESTE

ESP1

 

 

1

 

NORESTE

ESP2

 5

 

5

 

MADRID

ESP3

 5

 

6

 

CENTRO

ESP4

 1

 

1

 

ESTE

ESP5

17

 

20

 

SUR

ESP6

1

 

 1

 

FRANCE

F

 

 

57

244

ILE DE FRANCE

F1

29

110

 30

129

BASSIN PARISIEN

F2

 9

31

 9

41

NORD-PAS-DE-CALAIS

F3

 

 

 0

7

EST

F4

 4

24

 5

27

OUEST

F5

 5

 

 5

1

SUD-OUEST

F6

 1

6

 1

7

CENTRE-EST

F7

 6

17

6

22

MEDITERRANEE

F8

 1

 

 1

10

IRLAND

IRL

 17

351

17

397

ITALY

I

 

 

29

336

NORD-OVEST

I1

6

50

 6

57

LOMBARDIA

I2

12

131

 15

151

NORD-EST

I3

 1

23

 1

23

EMILIA-ROMAGNA

I4

 

 

 0

23

CENTRO

I5

 3

11

 4

19

LAZIO

I6

 1

35

 1

38

CAMPANIA

I7

 

 

 0

8

ABRUZZI-MOLISE

I8

 

 

 0

11

SUD

I9

 1

4

 1

4

SICILIA

I10B

1

2

 1

 2

LUXEMBURG

LUX

 

 

3

NA

NEDERLAND

NL

 

 

29

514

NOORD-NEDERLAND

NL1

 

 

 1

27

OOST-NEDERLAND

NL2

 

 

 5

77

WEST-NEDERLAND

NL3

 

 

12

267

ZUID-NEDERLAND

NL4

 

 

11

143

PORTUGAL

P

 

 

9

NA

UNITED-KINGDOM

GB

 

 

150

1300

NORTH

GB1

 12

30

13

34

YORKSHIRE AND HUMBERSIDE

GB2

 1

42

 1

52

EAST MIDLANDS

GB3

 5

55

5

77

EAST ANGLIA

GB4

 

 

 0

39

SOUTH-EAST

GB5

 44

476

50

620

SOUTH-WEST

GB6

 7

70

9

88

WEST MIDLANDS

GB7

 22

63

24

78

NORTH-WEST

GB8

 12

160

 12

185

WALES

GB9

19

18

19

24

SCOTLAND

GB10B

 14

77

 15

100

NORTHERN IRELAND

GB11B

 1

3

 2

3

TOTAL REGION

 

340

2312

432

3528

 

 

 

 

 

 

 

Notes: NA indicates that data were unavailable at the time of data collection.  Total number of European regions in this table is 58.  The regional classification is based on the European Union’s NUTS-1 and –2 classification codes.  The criteria used to select the samples for statistical analysis are described in Section 3 of the text.

Table 2. Conditional Logit Results for Location Choice Model: All Manufacturing

 

 

Japanese Affiliates

 

 

US Affiliates

 

(1)

 

(2)

 

WAGES

 

11.807

(0.397)

 

 

-21.603

(1.701)c

 

UNION

-0.013

(1.997)b

 

0.033

(9.312)a

UNEMP

-0.108

(4.019)a

 

-0.143

(12.875)a

EDUC

-0.766

(0.590)

 

-3.270

(5.316)a

TECH

0.518

(1.358)

 

0.398

(2.417)b

TRANS

0.163

(3.191)a

 

0.195

(10.210)a

GDP

0.0012

(0.285)

 

0.018

(9.980)a

TAX

-6.542

(7.436)a

 

-11.754

(27.854)a

 

Log Likelihood

 

-1184.65

 

 

-7066.56

 

No of regions

38

 

31

No of observations

340

 

 

2312

 

Notes: Dependent variable is Prob(firm i chooses region k).  The levels of significance at a two-tailed test are: a = 1%; b = 5%; and c = 10%.

 

 

 

 

 


 

Table 3. Conditional Logit Results for Location Choice Model: Selected Industries

 

 

Japanese Affiliates

 

 

US Affiliates

 

 

Machinery

 

 

Machinery

 

Chemicals

 

(1)

 

(2)

(3)

(4)

(5)

(6)

 

WAGES

 

-14.499

(0.386)

 

 

8.607

(0.227)

 

-46.200

(2.669)a

 

 

-16.551

(0.961)

 

65.060

(2.340)b

 

28.150

(1.060)

 

UNION

-0.021

(2.733)a

 

-0.033

(3.908)a

0.030

(6.172)a

0.024

(4.738)a

0.010

(1.542)

0.019

(2.745)a

UNEMP

-0.150

(4.590)a

 

-0.107

(3.056)a

-0.184

(12.176)a

-0.132

(7.885)a

-0.130

(5.027)a

-0.170

(6.140)a

EDUC

-0.484

(0.280)

 

-2.021

(1.145)

-3.328

(3.834)a

-3.766

(4.306)a

-4.230

(3.326)a

-3.587

(2.722)a

TECH

3.667

(3.217)a

 

1.124

(0.852)

2.039

(3.804)a

-0.346

(0.536)

-14.775

(7.498)a

-14.086

(6.943)a

AGGLO

 

 

 

1.133

(3.813)a

 

0.914

(6.527)a

 

0.814

(6.735)a

TRANS

0.212

(3.552)a

 

0.246

(3.947)a

0.229

(8.992)a

0.233

(9.228)a

0.258

(6.574)a

0.181

(4.608)a

GDP

-0.0064

(1.202)

 

-0.015

(2.614)a

0.018

(7.555)a

0.0091

(3.307)a

0.015

(3.808)a

0.013

(3.013)a

TAX

-8.516

(7.210)a

 

-7.322

(6.106)a

-14.395

(25.631)a

-12.495

(21.446)a

-9.681

(10.918)a

-9.934

(11.535)a

 

Log Likelihood

 

-841.90

 

 

-834.47

 

-4061.23

 

 

-4040.42

 

-1510.18

 

-1487.83

 

No of regions

38

 

38

31

31

31

31

No of observations

246

 

 

246

1368

1368

489

489

 

Notes: Dependent variable is Prob(firm i chooses region k).  The machinery sample includes observations in non-electrical machinery (NACE 32), office equipment (NACE 33), electrical machinery (NACE 34) and transportation equipment (NACE 35).  The chemicals sample consists of observations in chemicals (NACE 25) including pharmaceuticals (NACE 257).  The levels of significance at a two-The levels of significance at a two-tailed test are: a = 1%; b = 5%; c = 10%.

 

 

 

 


Appendix

 

Sources and Definitions of the Variables

 

 

WAGES is constructed by dividing the regional value (in ECU) of wages and salaries in manufacturing in 1989 by the number of wage earners in that region in manufacturing in 1989. This variable is defined at the regional level and constructed from EUROSTAT, 1993, Structure et activite de l’industrie, donnees regionales 1988-1989, Theme 4, Serie C, Luxemburg.

 

UNEMP is a four-year average of the unemployment rates for the region in 1987-1991. It is defined at the regional level and constructed form EUROSTAT, 1990 and 1993, Regions Statistical Yearbook, Theme 1, Series A, Luxemburg.

 

UNION is the unionization rate in 1988 at the national level.  It is taken from OECD, 1991, Perspectives de l’emploi, Paris.


EDUC
is the regional number of students in high school on 1990, divided by the population aged between 15 and 24 years in the same region in 1990.  It is defined at the regional level and obtained from EUROSTAT, 1993, Portrait des regions, tomes 1-3, Luxembourg. 

 

TECH is the cumulated number of patents granted in the United States to the firms domiciled in an EC country in manufacturing over the 1963-1986 period in a specific industry, divided by the total cumulated number of patents for all EC countries in that industry.  This variable is defined at the national level for a specific industry and constructed from The US Patent and Trademark Office, 1987, Patenting Trends in the US: 1963-1986, Washington, D.C.: US Patent and Trademark Office.  This variable is aggregated to total manufacturing in the specification of Table 1.

 

AGGLO is constructed as the ratio of two variables.  The numerator is the number of wage earners in the region in the specific industry, divided by the number of wage earners in that region in total manufacturing.  The denominator is the number of wage earners in the EU in the specific industry, divided by the number of wage earners in the EU in total manufacturing.  These variables are constructed for 1989 and obtained from EUROSTAT, 1993, Structure et activite de l’industrie, donnees regionales 1988-1989, Theme 4, Serie C, Luxemburg.  This variable is defined at the regional level for the machinery sector (NACE 32 and 34) and the chemical sector (NACE 25) for the statistical analysis of Table 2.

 

TRANS measures the density of transport network in the region. TRANS is constructed from principal component analysis (factor analysis) by utilizing the correlations within the three observed variables, SEA, AIR, and ROAD.  Assuming each variable is represented as a vector, we derived a factor that represents “density” of transportation network with string positive loadings on SEA, AIR, and ROAD.    AIR is the quantity of goods (in 1000 tons) unloaded in the region in 1990, divided by the population of the same region in 1990.  SEA is the quantity of goods (in 1000 tons) loaded in the region in 1990, divided by the population of this region in 1990.  ROAD is the length of highways in the region measured in km in 1990, divided by the area of this region in km square.  The three variables are all obtained from the same statistical source, EUROSTAT, 1990 and 1993, Regions Statistical Yearbook, Theme 1, Series A, Luxemburg.  TRANS is defined at the regional level

 

GDP is an index of regional Gross Domestic Product per capita in 1990 (EC total=100) and obtained from EUROSTAT, 1990 and 1993, Regions Statistical Yearbook, Theme 1, Series A, Luxemburg.  NGDP is an index of Gross Domestic Product per capita in 1990 at the national level.

 

TAX is the nominal corporate tax rate in 1990 for the EC member states.  This variable is defined at the national level and obtained from Commission des Communautes europeennes, 1992, Inventaire des impot percus dans les Etats membres des Communautes europeennes, Luxembourg.



[1] There exists a literature that examines the location choice of foreign direct investment flows in the European Common Market by using aggregate time-series data.  Among others, Culem (1988) found that aggregate economic growth is a determinant of locations for US FDI flows in the European Common Market.  Morsink and Molle (1991) found evidence that exchange rate variability influenced the destination of FDI flows.      

[2] McFadden (1973).

[3] In the previous studies of MNEs’ choices of countries, Swedenborg’s (1979) showed that Swedish direct investment was drawn to countries where labor costs were relatively high, reflecting in fact the high skill content of labor force.  Dunning’s (1980) study of US MNEs also found the share of local production in foreign markets was high where wages were relatively high.  Yamawaki (1993) finds a significant negative effect of labor cost on the presence of Japanese MNEs in the European Community in the late 1980s.

[4] Carlton (1983) and Coughlin et al. (1991) for some evidence.

[5] GDP per head may not necessarily measure actual market size for a country like Ireland where the presence of MNEs is large.  This is true for Ireland in the late 1990s through the 2000s.  This issue, however, is less significant for the period examined in this paper.  Ireland’s GNP per capita was larger than or close to GDP per capita in the late 1980s through the early 1990s.

[6] Effective tax rate may be more appropriate to measure the effect of corporate taxation because it considers not only the statutory tax rate but also tax incentives offered in the host country.  Estimates of effective tax rate for OECD countries are available for a recent period (e.g. Buijink, Janssen, and Schols, 2000).   They are, however, unavailable for the period examined in this paper.

[7] Yamawaki (1994) provides more detailed accounts on the Toyo Keizai data and describes the pattern of Japanese investment in Europe by using the 1991 data.

[8] The NUTS (Nomenclature of Territorial Units for Statistics) classification is the official regional classification defined and used by the Eurostat.

[9] The US Chamber of Commerce data are used in this paper because it provides us with company-specific information on location.  While US Department of Commerce data on foreign direct investment provides quantitative data on direct investment for certain industries and countries in the EU, it is not easily matched up with the Chamber of Commerce data because the Department of Commerce data do not provide company-specific location data.

[10] Subsidiaries classified in Scheswig-Holstein, Baden-Wurtemberg, Berlin, Bremen, Rheinland-Pfalz, Sachsen in Germany, Nord-Pas-de-Calais in France, Emilia-Romagna, Campania, Abruzzi-Molise in Italy, and East Anglia in the United Kingdom are thus eliminated and not in the sample.  All the seven regions in Spain are not included in US sample due to the unavailability of the data, while the Japanese sample includes all Spanish regions except Noroeste.  Finally, subsidiaries in the Netherlands are deleted from the both US and Japanese samples because of the unavailability of industry-and region-specific wages at the time of data collection. 

[11] In an attempt to examine if the MNE’s location choice is nationally focused rather than regionally focused, GDP per capita at the national level (NGDP) was added to the specification of equations (1) and (2).  For the Japanese sample, the coefficient for GDP remains insignificant, while the coefficient for NGDP is significant with a negative sign, reinforcing our earlier results of pan-European focus of Japanese MNEs.  On the other hand, for the US sample, both the coefficients for GDP and NGDP are significant with negative signs, suggesting a joint influence of regional and national markets.  Inference on the causal effects of these variables is not unequivocally established because of a strong correlation between NGDP and TAX, particularly in the US sample.   The coefficient for TAX is strongly affected when NGDP is included in the model specification, and its sign changes from negative to positive for the US sample.  Simple correlation coefficients between TAX and NGDP in the Japanese (n=340) and US (n=2312) samples are 0.775 and 0.964, respectively.  The simple correlation coefficients for NGDP and GDP in the Japanese and US samples are 0.683 and 0.774, respectively.

[12] The ranking of European countries in Table 1 shows that US MESs choose the Netherlands, Belgium, and Ireland after the United Kingdom.  It is well known that the usage of English is very common in the Netherlands and the Flemish and Brussels regions of Belgium.