Location Choice Network Patterns of Japanese Multinational Companies in Europe

This research investigates network patterns of location choice of multinational companies by using multinomial logit method. It empirically analyses regional economic factors, which were significant for attracting investments of Japanese companies during the last decade, by using the most detailed regional data possible. In addition to previous studies, this paper particularly addresses factors, which follower Japanese companies considered important in their investment decisions. For Japanese multinational company to locate near to other already established company from the same country there could be such reasons as: they tend to follow their business customers or because of existing intra-firm linkages already established in Japan, which they carry on in their investment decisions. The aim of the paper is threefold. Firstly, it analyzes significant regional economic factors, which follower Japanese companies consider important in choosing regions with already established Japanese firms and, secondly, it analyzes those regional economic factors, which are significant for those companies, which choose to locate near to hubs of other Japanese companies. Thirdly, by using distances between regional centers, this paper tries to establish significance of physical distance in establishing hub of Japanese companies. Paper hypotheses that Japanese companies disregard geographical distance in their investment decisions as they create networks of Japanese companies.


Introduction
Nowadays firms are free to choose locations, which best suit their interests.Although some companies in industries such as mining and retail are limited to certain locations where natural resources or customers are, most manufacturing and wholesale industries are relatively free to choose where to be located.Moreover, transportation costs have been decreasing and modernization of telecommunications have facilitated and fastened flow of information.For instance, European regional integration has greatly reduced trade costs and barriers to serve customers in a different country or region.Some researchers use term "footloose" which refers to idea that company is not tied to a specific location, but is free to choose any location according to the factors they find important (Clark, 1969).
On the other hand, companies are exposed to greater competition in such an integrated area, because number of potential competitors increases as well, compared to number of competitors they face in a country protected by entry barriers.As a result, one of solutions to stay competitive in such wide and integrated markets is to select the location, which best suits their strategic, operational and financial interests.
In previous studies, several factors emerge according to which multinational companies choose investment location.The first one is a market-companies want to serve markets where demand is high (market potential factor and companies can enjoy economies of scale), the second one is costs-companies want to locate where operational costs are smaller (this includes labor costs, expenses of renting/buying property, various taxes etc.), the third one-companies tend to locate in the same region as companies with similar attributes are located-this includes similar industries and similar country of origin (so called agglomeration effect) and the forth one-availability of resources as companies don't want to face shortages of labor force, government services, natural resources etc.This study analyzes location choice of 1,023 Japanese investments in 236 regional locations belonging to 18 European countries over the time period from 1995 to 2005.Following results derived from McFadden's (1984) paper, logit model is chosen.This study analyzes company's location choice from the perspective of European regions-the ability of regions to attract Japanese investment, attract followers and attract companies, which might serve neighboring regions.
In earlier papers, the most common model used was conditional logit and there were number of papers, which use nested logit and mixed logit models.This research uses multinomial logit, which allows classification of regions to reflect their experience of attracting Japanese companies and their position in the network of Japanese companies.
By using nested logit method, Head (2004) analyzes location choices of Japanese investment in nine EU countries during the period from 1984 till 1995.This study explores various market potential indices, notably the index developed by Harris (1954) and Krugman (1992).They also include other variables to characterize production costs-wage rate, corporate and social tax.It has been found that market potential is a significant and positive factor for investors' location choice.Social tax and corporate tax rate are found to be negative and significant.For nested logit model analysis, regions are grouped into their respective non-overlapping country groups.
Earlier paper by Head (1995) uses conditional logit to analyze location choice of 751 Japanese manufacturing plants in US in the 1980s.They found strong agglomeration effects on location choice, especially for companies within a similar industry or belonging to keiretsu.Similarly, Basile (2003) analyzes location choice of EU and US multinational companies in Europe.Contrary to the study by Head (2004), they try to different groups of countries e.g., North-South, Anglo-Continent-South etc.It has been found that country boundaries don't matter in the case of EU and US multinational companies.Contrary to agglomeration forces-peripheral regions also attracted significant investment and regions receiving assistance from structural funds were particularly attractive to European and US multinationals.
By using conditional logit model, Alegira (2006) analyzes location decisions of European firms across Europe with large sample size of 4,803 foreign investment projects in 246 regions.This study considerably extends geographical areas of previous location choice papers.Similarly to previous studies, market potential has been found to have positive and significant influence.For other included variables, income per capita is insignificant at country and regional levels.This study measures agglomeration effects by the number of foreign projects located in the region or country one year before the location decision is made.Study of agglomeration effects reveal that they tend to dominate on regional level, but at the country level economic activity tends to concentrate on peripheral countries.
Heterogeneity of investors has been analyzed with the mixed logit method, which has been used in Rasciute (2007) paper on the location choice of foreign investors in 13 Central and Eastern European Countries.This study reports high heterogeneity of investment location decisions.Market effect also has been observed, suggesting that larger host country will be more likely to be selected and this effect tends to be stronger for larger investing firms.In addition, less profitable firms are likely to invest in central locations, whereas more profitable firms will choose peripheral countries.
There are also studies comparing investment location choices in different areas of Europe.Disdier (2004) analyzes location choice of French firms in Eastern and Western Europe by using both conditional and nested logit methods.Results suggest that French firms will choose countries with already established French firms of the same industry, which confirms with results of previous papers.Other factors such as GDP and unemployment have positive and significant influence.Negative influences have been observed for such variables as GDP per capita, distance from France and wage rate.Additionally, this paper also introduces exchange rate volatility, which has negative and significant effect.
Several papers include unemployment rate, but results are mixed for this factor.As Disier (2004) points out that high unemployment rate might suggest labor market imperfections, but on the other hand from investor's view point might signal availability of large labor supply.For example in case US, positive influence has been reported by Coughlin (1991).This empirical research contributes to previous studies in such a way as it extends geographical spectrum of analysis and lowers statistical level of regional analysis.This paper also introduces several new factors in our analysis to characterize development level of infrastructure, distance to other Japanese companies and their www.ccseninvestmen The rest of 3 describe

Theore
Discrete ch is conditio can be w profits-Ja possible al Firm's loc firm will c where V j is Linear exp vector of e A firm cho below give Further, M estimated u Dependent particular r to the regi Japanese category(3 borders oth Figure 1).data in the Figure 2.     year.r and lue 3, if one or several Japanese companies are located in the region, which numbers more than two regions with present Japanese investor, thus forming hub of Japanese companies.If a region receives no new investment from Japanese investors, the value of the variable is 0. Presence of Japanese company in the region is determined by its registered address provided in the database.
On the right hand side, independent variables are region specific with a few country specific variables.For a summary of the independent variables, see Table 1.Size of the region is characterized with number of population, which can also characterize market potential.Labor costs and labor availability in the region is calculated by an average wage in the region and unemployment rate, which is country specific variable.Development level of infrastructure is captured by density of roads in the region, which is year specific, and by dummy variable, which characterizes that region has or is located in directly neighboring region of a large airport (20 million passenger movements per year).Distance captures physical distance between region centers of the selected region and region, which has the largest number of present Japanese companies.Wage rate might characterize labor cost as well as income level.High income levels in the region might be attractive for Japanese investors, but in the same time high labor costs have negative influence.In addition, higher labor costs might also indicate high labor quality, which is also attractive.Positive, insignificant coefficient is reported in Alegria (2006).Another labor characteristic is unemployment rate, which signals availability of labor force in the region.Disdier (2003) uses both GDP and GDP per capita, which could be used interchangeably with wage rate.Similarly, Head (2004) and Alegria ( 2006) uses several market potential calculation methods, which also include neighboring economies' GDP weighted by distance.The expected sign suggested by previous studies is positive.
Infrastructure in a region is measured by road density, which is calculated as a ratio of length of road network to the area of the region.Positive value is expected for this variable as investors will prefer to invest in the region with already established infrastructure.Another measurement of infrastructure development is availability of a large international airport in the region and surrounding regions.Distance is measured by geographical distance to the assumed center of the region (source MCRIT database, 2006).
Multinomial logit has been used in several papers related to analysis of location choice.Wei (2005) analyzes entry modes of FDI in China by using categorical variables of entry mode.In case of Europe, Louri (2000) uses multinomial logit to determine outward FDI activity of Greek firms, where the categories are firms' decisions to export, engage in FDI or not in engage at all in export or FDI activities.
Japanese investors are more likely to select region where is already established companies, because of positive spillover of information and possible cost savings in procurement.It is also possible that branches from same parent company are established in the same region for purpose of cost savings and facilitation of information exchange between daughter companies.

Empirical Results
In the first part of analysis, results of multinomial logit are reported in Table 2, which represents likelihood of regions to attract Japanese investors.There are totally 236 regions in the study, which roughly corresponds to number of regions at NUTS2 level.
Three categories, which represents manner of investment, are distributed rather evenly over the eleven years,with 212 or 8.2% entry regions of any given time have been chosen once, 353 or 13.6% regions during observation period have been chosen more than once, 601or 23.2% regions in the similar time period has been located next to more than two regions with Japanese companies.The rest is 1430 or 55.08% of regions, which has not been chosen at all.
Results are reported in two specifications, where in the second specification variables measuring infrastructure development are not observed.Log likelihood doesn't improve significantly.Most variables for category (1) are insignificant at observed significance levels, but for categories (2) and (3)-largely significant.Scope of this research concentrates on follower companies, which are represented by those categories (2) and (3).
It has been observed that size of region measured by the number of inhabitants has been negative for the first two categories, and positive, significant for companies located in hubs-category (3).On the other hand, other labor characterizing factor average wage rate in the region are positive and significant for the category (2).There is no variation between categories, but for category (3) this variable is insignificant.Companies, which are followers and did not invest in the regions close to regional hubs, found such factors significant as average wage rate, unemployment rate, infrastructure development level and distance.Insignificant factors were presence of a large airport and population size.
Those companies located close to the hubs of Japanese companies considered such factors significant as: population size of the region and infrastructure development level.
Figure ional attractivene icated inside circl nese companies' p of companies in th his region have at esence of establish hich has 5 compa boring region, dist companies investing close to hubs.

Table 1 .
Data summary

Table 2 .
Multinomial logit regression results, all regionsUnemployment rate, which could be explained as an indicator of availability of labor force, has negative sign for the first category, although insignificant.Variable turns significant for category (2) in the first model, suggesting Japanese investors' preference for regions with higher unemployment rate.Earlier studies show various signs for unemployment rate influence on the investment decisions.For example in