Do Foreign Investors’ Location Determinants in Service Functions Differ According to Sectors? An Empirical Analysis of EU for 1997 to 2011

Abstract

European Union (EU) is very attractive for foreign direct investment (FDI) in services and policy makers should know the reasons explaining investment location choices of foreign firms in order to attract them. This article explores FDI location determinants in service functions in the EU-28. Studies dealing with such an issue stay generally focused on the location of service functions in the manufacturing sector. Our assumption is that the location determinants of service functions may differ according to sectors. So, we propose to study whether the location pattern of some service functions is sector independent, whereas it is sector dependent for other service functions. Using a database developed by Ernst and Young, we estimate mixed logit models on foreign firms’ location choices. Our contribution is to consider simultaneously three sectors and eight service functions for 271 European regions, during the period 1997 to 2011. Our fundamental findings are that service functions location choices are different according to sectors and that location determinants vary according to the service function considered. The only variables significant for all service functions are agglomeration variables. However, our contribution is to distinguish between different types of agglomeration (regional, sectorial, functional, and group agglomeration) and to show that some agglomeration variables act differently according to service functions.

Keywords

FDI determinants of services location choices mixed logit models European union

Introduction

In a recent study, echoing the work of Levitt (1972), Crozet and Milet (2014) asked, “Is everybody in services?” They clearly answered no. They formulated another question: “Is everybody more and more in services?” They clearly answered yes.

These two authors noted that although some firms do not sell services, the share of the “pure manufacturers” is declining over time. This observation made for French manufacturing firms between 1993 and 2007 is also true for a growing number of countries (Demmou 2010). By becoming more global, manufacturing sectors in developed economies are becoming less concerned with pure production and more concerned with design, research, marketing, or product conceptualization (Crozet and Milet 2014). They include more and more service activities and propose product services (Tukker and Tischner 2006).

As François and Hoekman (2010) underline, if both final and intermediate demand factors are important in explaining the growing share of services in the economy, there is a fundamental function that many services perform in the overall economic growth and economic development: “they are inputs into production.” Services facilitate transactions through space (transport and telecommunications) or time (financial services). Services are also determinants of the productivity of the factors of production (labor and capital) that generate knowledge, goods, and other services. François and Hoekman (2010) illustrate this giving the examples of education, research and development (R & D), and health services, which are inputs into the production of human capital.

In its recent report on international trade, European Commission (2013) also insists on the major role played by services in all modern economies by improving the performance of other industries, since they can provide key intermediate inputs, especially in a globalized world. Indeed, while the share of services in gross exports worldwide is about 20 percent, almost half (46 percent) of value added in exports is contributed by services-sector activities (United Nations Conference on Trade and Development [UNCTAD] 2013). This is partly explained by the fact that most manufacturing exports require services for their production (François and Woerz 2008).

Beyond their presence as input or output in production and trade flows, services also represent a great part of foreign direct investment (FDI), which may be seen as an alternative economic strategy to trade. Indeed, firms can seek to complement or substitute external trade, by producing (and selling) goods and services in countries other than where they were first established. At this stage, it is important to keep in mind that services have unique characteristics that affect their tradability. According to François, Pindyuk, and Woerz (2009), their two most obvious characteristics are intangibility and nonstorability. They note that services also require differentiation and joint production, with customers having to participate in the production process. It may explain why in practice FDI remains a major channel for foreign providers to supply services.

By 2005, services represented three-fifths of the global FDI stock up from a half in 1990 (UNCTAD 2007). This can be seen in part as an illustration of the difficulties of cross-border trade for services and in part of the need for a direct and ongoing presence to supply many services (Lejour and Smith 2008). Today, about 60 percent of global FDI stock is still in the service sector (François and Pindyuk 2013). FDI is expanding more quickly for services than for goods and is increasing at a more rapid pace than conventional trade in services.

At the end of 2010, services made by far the largest contribution to both outward (58.5 percent) and inward (57 percent) stocks of European Union (EU) FDI with the rest of the world (Eurostat 2013). If we focus on inward FDI stocks in EU-27, we note two main facts. Firstly, manufacturing represented only 20 percent of inward FDI and almost two-thirds of EU-27 inward stocks of service FDI were held in financial and insurance activities in 2010. Secondly, the United States accounted for 35 percent of EU-27 inward stocks of FDI from the rest of the world in 2011. This country consolidated its position as the major FDI stockholder in EU-27, having invested mostly in financial and insurance activities, followed by manufacturing. Other countries with significant share in EU-27 inward positions were Switzerland (12 percent), Japan (3.8 percent), Canada (3.6 percent), followed by Brazil (2 percent), Singapore (1.8 percent), and Russia (1.4 percent).

These figures show that EU is very attractive for FDI in services and for very different countries. However, they do not inform about the specificities of European countries. Indeed, EU is not a simple bloc: multinationals may prefer to invest in one country rather than another. There are different reasons explaining their choice that policy makers should know in order to choose the good policy if they want to attract FDI. Moreover, besides governments at a national level, regional authorities also wish to attract multinationals’ activities, as the growing importance of EU regional policy proves it. Thus, the aim of this article is to explore these reasons by analyzing the location determinants of service activities in the regions of EU-28.

As decreases in communication and coordination costs have facilitated the international fragmentation of the value chain, contributing to the internationalization of service functions that surround the production process (Py and Hatem 2009), the literature on location decisions related to these service functions has developed steadily (Cf. infra). Nevertheless, it has mainly focused on the study of service functions in the manufacturing sector (Holmes 2005; Defever and Mucchielli 2005; Defever 2006; Henderson and Ono 2008; Defever 2012), with only few papers addressing the specific case of service sectors (Kolstad and Villanger 2008; Tomlin 2008; Nefussi and Schwellnus 2010; Castellani, Meliciani, and Mirra 2014). The research examining jointly service functions and service sectors is scarce. Py and Hatem (2009) make a first step in this direction by looking at the location determinants of both the service sector and different service functions. However, they do not consider that the location determinants of service functions differ according to sectors and to our knowledge, there is no study of this type in the empirical literature.

Our assumption is precisely that the location determinants of service functions may differ according to sectors. More precisely, we propose to explore whether the location pattern of some service functions are sector independent, whereas it is sector dependent for some other service functions. Thus, using a database developed by Ernst and Young, called the European Investment Monitor (EIM) and estimating mixed logit models (MLMs) on multinationals’ location choice, our contribution is to consider simultaneously different sectors (3) and different service functions (8) for a great number of regions (271 Nomenclature des Unités territoriales statistiques [NUTS] 2 European regions), during a long period of time (the period 1997 to 2011). Our fundamental findings are that service functions location choices are different according to sectors and that traditional location determinants highlighted by the literature vary according to the service function considered. The only variables that are significant for all service functions are agglomeration variables. However, our contribution is to distinguish between different types of agglomeration (regional, sectorial, functional, and group agglomeration) and to show that some agglomeration variables act differently according to service functions.

This article is organized as follows. The second section reviews the literature. The third section describes the data. The fourth section introduces the econometric methodology and provides a description of the explanatory variables. The fifth section presents the results. The sixth section highlights our main conclusions.

Literature Review

As mentioned by some researchers dealing with the location determinants of service FDI (e.g., Py and Hatem 2009; Hahn and Bunyaratavej 2010; Nefussi and Schwellnus 2010; Castellani, Meliciani, and Mirra 2014, for the most recent works), the great bulk of research on the location determinants of FDI has focused so far on the manufacturing sector. However, since the mid-2000s, works dedicated to various aspects of the internationalization of services (trade, offshoring, outsourcing, etc.) have noticeably increased.

The part of the relatively recent literature interested in the location determinants of service FDI takes its roots in two distinct fields of economics: international economics and economic geography. With respect to international economics, authors dealing with FDI in services often refer to the field of research related to international division of the production process (Lassudrie-Duchêne 1982; Sanyal and Jones 1982; Dixit and Grossman 1982). More specifically, the theory of trade in specific tasks (Grossman and Rossi-Hansberg 2008) among other theoretical works related to offshoring appears as a recent development of this latter research perspective. On the other hand, the location of FDI in services can also be explained by the new economic geography (NEG) literature developed along the lines of the seminal work of Krugman (1991) as a trade-off between centrifugal forces linked to the labor force immobility and to the existence of transport costs and centripetal forces linked to increasing returns and pecuniary externalities in the location of economic activities.

Beyond drawing on two theoretical fields of economics, research on the location determinants of service FDI deals with two distinct issues: the location of service functions on the one hand and of service sectors on the other hand. Following Duranton and Puga (2005)¹ and Defever (2006), the terminology “service function” refers to a stage or a service activity upstream or downstream the production stage itself within the firm’s value chain, whatever the sector where the firm performs its activities. Considering service functions leads to bring together strictly identical service activities. A service sector gathers all firms whose main activity is of a service nature and is the same according to the nomenclature of economic activities under consideration. A service sector also encompasses a variety of service functions.

Among works on service sectors or functions, some deal with a particular or a small number of similar sectors—such as business services (Castellani, Meliciani, and Mirra 2014; Nefussi and Schwellnus 2010) or banks (Zhu et al. 2012)—whereas others focus on a particular or a small number of similar functions—such as headquarters (Henderson and Ono 2008; Davis and Henderson 2008; Strauss-Kahn and Vives 2009), call centers/shared service centers/information technology services centers (Hatem 2004; Bunyaratavej, Hahn, and Doh 2007; Doh, Bunyaratavej, and Hahn 2009; Hahn and Bunyaratavej 2010), logistics (Hatem 2005), sales offices (Holmes 2005), and R & D (Hatem 2007; Sachwald and Chassagneux 2007). Other authors propose a broader analysis of the location determinants of service sectors or functions: this is the case, for example, with Tomlin (2008) who examines fourteen service sectors; Defever (2006, 2012) takes into account five production and service functions in the same time (headquarters, R & D, production, logistics, and sales and marketing), whereas Py and Hatem (2009) develop an analysis encompassing seven functions (production, headquarters, R & D, logistics, sales, delivery of services, and call centers).

Authors dealing with the location determinants of service FDI generally agree on a relative similarity of the latter with those explaining FDI in the manufacturing sectors or in the production function (Nefussi and Schwellnus 2010; Py and Hatem 2009; Castellani, Meliciani, and Mirra 2014; Montout and Robin 2012). This leads researchers to support the idea that no specific theory is needed to account for the location determinants of service FDI. Markusen (2005, 24) asserted, “we can make good progress in understanding the offshoring of white-collar services at the theory level from our existing portfolio of models. Many important features of offshoring of white-collar services can be modeled from a recipe that mixes and matches elements from the existing inventory.” Hence many works draw on the model developed by Head and Mayer (2004), rooted in NEG, and so does the present contribution. Py and Hatem (2009, 78), for example, justify their position saying that “we do not assume that the model explaining service activities location choices radically differs from the one explaining manufacturing activities location choices but that the relative weight of each location determinant greatly varies according to the activity under consideration.”² Such a position is still supported by recent works (Yin, Ye, and Xu 2014). Then for manufacturing as well as service activities, the firm’s location choice is a function of the resources provided by the host country (which impact production costs), demand variables (among which market potential), and agglomeration variables. However, some determinants are more significant for service FDI. For example, research insists on the importance of the education level of the manpower and of a common language or more broadly cultural similarity (Hatem 2004; Bunyaratavej, Hahn, and Doh 2007; Sachwald and Chassagneux 2007; Py and Hatem 2009). This also seems to be the case with market size, which can take the form of downstream demand size for service activities provided to other businesses (Hatem 2005; Nefussi and Schwellnus 2010; Sachwald and Chassagneux 2007; Kolstad and Villanger 2008; Py and Hatem 2009; Castellani, Meliciani, and Mirra 2014). Yin, Ye, and Xu (2014, 29) even show that “FDI in services tends to be mainly motived by market-seeking and client following purposes which are more prominent for services FDI than manufacturing FDI.”

Beyond that, just as any economic activity, service activities tend to agglomerate in order to capture externalities. However, whereas manufacturing activities seek to benefit from sectorial externalities, functional specialization plays a key role in the location choice of service activities (Duranton and Puga 2005; Defever and Mucchielli 2005; Defever 2006), without completely excluding the search of sectorial effects for the latter (Py and Hatem 2009 ³). The role of both same function scale externalities linked to the presence of other units operating in the same service function and diversity scale externalities linked to the presence of other service functions is brought to light by Davis and Henderson (2008) in the case of headquarters location. Although functional agglomeration effects appear to be important for services location, co-location of headquarters’ activities with production plants or with prior service investments within the same parent company remains a major determinant of service FDI location (Henderson and Ono 2008). More generally, there seems to be a trade-off between functional agglomeration of service activities around greater cities and co-location with prior affiliates (Defever and Mucchielli 2005; Defever 2006, 2012; Henderson and Ono 2008). Nevertheless, some works point to the fact that services cannot be dealt with as a whole and that location determinants vary according to the service nature (Hatem 2004, 2005; Defever 2006; Doh, Bunyaratavej, and Hahn 2009; Sachwald and Chassagneux 2007; Tomlin 2008; Py and Hatem 2009; Montout and Robin 2012).

In these analyses, costs are not totally considered as irrelevant. They are often mentioned but do not appear as primary drivers of service location decisions, except in specific cases (Bunyaratavej, Hahn, and Doh 2007; Montout and Robin 2012), and they do not reduce to labor costs offered by host countries. Characteristics of the host country (wages, taxes, quality of infrastructures, and even institutional quality), as well as evolutions of the exchange rate, impact on the cost multinational firms can provide their services to customers. For instance, Tomlin (2008) mentions a deterrent effect on service FDI played by higher unit labor costs in the host country compared to the home country in her study of Japanese FDI into the United States. The author also studies the impact of real-exchange rate appreciations, which are found to lead to higher FDI entry rate. Holmes (2005) examines the location determinants of manufacturing firms’ sales offices and concludes that fixed costs and frictional costs and to a lesser extent matching costs⁴ play an important role in location decisions depending on firms’ size and characteristics. Sachwald and Chassagneux (2007) show that quality and costs of R & D impact on global development centers’ location but not on local development centers or global research laboratories.

To conclude, the approach to the location determinants of service FDI through the lens of functions has improved our understanding of location choices with respect to service activities. However, this approach often restricts itself to the analysis of service functions in the manufacturing sector and is not interested in the location determinants of service functions in service sectors. Py and Hatem (2009) have started to overcome this limitation by examining the location determinants of both the service sector (compared to the manufacturing sector) and different service functions (compared to the production function). However, they did it in a successive way, without crossing sectors and functions considering that “location principles of service functions can be relatively independent of the sector in which the company performs its activity”⁵ (Py and Hatem 2009, 68). Our purpose is to elaborate on this literature and to assess the relevance of such an assumption. As mentioned earlier, our own assumption is that the location determinants of service functions may differ according to sectors. More precisely, as this has not been considered before to our knowledge, we propose to explore whether the location pattern of some service functions are sector independent, whereas it is sector dependent for some other service functions. Moreover, we will not consider the manufacturing sector or the production function as the reference unit of analysis (it has extensively been done) but lead a relative analysis of the location of different service functions belonging to different service and manufacturing sectors.

Characteristics and Distribution of FDI in European Countries

This section presents the FDI data used in this article and provides an overview of their main features in the EU for the fifteen-year period studied.

Description of the EIM Database

Launched in 1997, the EIM database identifies foreign direct investment projects into Europe. The sources of information utilized are, among others, global, national and regional media, news sites, companies, and governments websites. This database has also been used by Defever (2006, 2012).

It has several major characteristics. First, there is no minimum size criteria defined for selecting investment announcements, although the number of investments where less than ten jobs are created turns out to be very low. The media announcements are likely to mainly concentrate on large projects of important multinational firms. Moreover, project details include information like global region and country origin of parent company; the city, region, and country receiving the investment project; the type of investment, if it is a new location, a new co-location, or an expansion of a prior investment; and at last the sector and activity (function) where the investment is made.

In our study, using the fact that the EIM database provides the exact location of each investment project, we associate precisely to each one the corresponding NUTS 2 unit.⁶ There are 271 NUTS 2 European regions that firms can choose for an investment but only a part of them are chosen. We distinguish two databases over the period 1997 to 2011, an exhaustive one and a restricted one. The complete database covers more than 40,000 location projects when one considers all sectors and all functions: we use it exclusively for the construction of agglomeration variables (Cf. infra). The limited database covers almost 25,000 location choices at the individual firm level in EU-28: we use it to explore our problematic, that is to say leading a relative analysis of the location of different service functions belonging to different service and manufacturing sectors. We distinguish three main sectors named industry (with subsectors like automatic assembly, electronics, food, paper, pharmaceuticals, textiles, etc.), business services (with three subsectors that are business services, scientific research and software), and other services (with subsectors like transport, financial intermediation, hotels and restaurants, real estate, telecommunication and post, etc.). We only retain private activities belonging to the secondary or the tertiary sectors where services can be an input or an output. Then, we remove agriculture, forestry, fishing, cultural activities, education, health and social work, and construction in the definition of our sectors.

With respect to functions, we remove education and training and manufacturing in order to keep the eight service functions: contact center, headquarters, internet data center, R & D, shared services center, testing and servicing, logistics, and sales and marketing. The first six correspond to business service functions. Logistics, which refers to all activities linked to the transport of goods, and sales and marketing, which includes business representative offices and trade and marketing offices or subsidiaries, can be considered as other service functions.

Otherwise, we make two choices that are not systematic in the literature. Firstly, unlike Head and Mayer (2004) or Defever (2006, 2012) but similarly to Basile, Castellani, and Zanfei (2008) and Py and Hatem (2009), we both consider investments from non-European multinational firms and investments from European multinational firms. This may be viewed as an asymmetry as the location of investments from European firms is likely partly determined by the distance between the home country of investor and the country of location of investment, while it is unlikely that investments by non-European firms, like American or Chinese ones, in EU-28, are determined by the distance. However, FDI projects from European firms in EU-28 are an important proportion of our database (cf. infra) and hence cannot be excluded. It would provide a rather bad representation of firms’ location choices in Europe.

Secondly, unlike Defever (2006, 2012), Py and Hatem (2009), or Montout and Robin (2012), we retain all project investments, creations (known as greenfield), and also extensions (known as brownfield). We think that this latter category may be linked to the location choice determinants because, in this case, that is really the decision of the firm to reinforce the initial investment.

Descriptive Statistics

Regarding the evolution of FDI location choices, Figure 1 confirms two facts. Firstly, there is a growing importance of FDI in service functions in EU-28. There were 1,084 investment projects in 1997 and more than the double in 2011 with 2,448 projects. Secondly, the trends regarding global FDI flows are also observed for the evolution of inward FDI in service functions in EU-28. As a matter of fact, Figure 1 shows two peaks in the 2000s, one in 2000 and a second in 2007. The same peaks are observed in the global economy for these years. In 2000, global FDI inflows grew faster than other economic aggregates like world production and trade, reaching a first record of US$1,411 billion and they still rose more in 2007 to reach US$1,979 billion, well above the previous level of 2000 (UNCTAD 2009). We can also note that when there was a significant slowdown for the global FDI flows like in 2003 and 2009 (UNCTAD 2004, 2010), this is also the case for FDI in service functions in EU-28. All these observations lead us to conclude that investment projects in service functions in EU-28 are a good reflect of what is ascertained for the global economy.

Figure 1.

Evolution of location choices in service functions in EU-28, 1997 to 2011 (number). Source: Our own database constructed from European Investment Monitor database.

Figure 2 maps the geographic distribution of those FDI projects across the NUTS 2 regions for all cumulated years and shows that each region has received some investments. However, their distribution is not homogenous: Great Britain, Ireland, Belgium, the Netherlands, and Germany attract a lot of them. Moreover, there is a region-capital effect in a great number of countries: for example, the region which receives the higher number of investments in Portugal, France, Poland, Hungary, or Sweden is, respectively, that of Lisbon, Paris, Warsaw, Budapest, and Stockholm. This fact is not very surprising as it concerns FDI in services.

Figure 2.

Cumulated distribution of foreign direct investment projects across Nomenclature des Unités territoriales statistiques 2 regions.

Descriptive statistics drawn from our database (not reported here) show an increasing trend in the number of investments in service functions in all sectors over the whole period. A majority of FDI in service functions is done by companies belonging to service sectors rather than to industry (61.5 percent and 38.5 percent, respectively), which justifies our choice to analyze service functions FDI originated not only from industrial companies but also from service companies. To go further into the analysis, Figure 3 gives the distribution of the FDI projects in service functions by sectors. The investments in industry and the investments in business services converge at the end of the period: in 1997, FDI in industry represented 48 percent of total FDI and FDI in business services 24 percent. In 2011, the proportion of FDI in industry amounted to 37 percent of total FDI and to 41 percent for business services FDI, whereas FDI in other services fluctuated within the 20 percent to 28 percent range over the whole period.

Figure 3.

Evolution of foreign direct investment location choices in service functions in EU-28 by sectors, 1997 to 2011 (percent). Source: Our own database constructed from European Investment Monitor database.

A more thorough analysis of service functions completes our descriptive analysis. Table 1 crosses the detailed functional and sectorial composition of inward FDI in service functions in UE-28 over the global period. It indicates that there are privileged functions: far ahead and whatever the sector, sales and marketing (54.8 percent of the total investment projects in service functions), followed by logistics (12 percent), R & D (11.5 percent), and headquarters (11.2 percent). The first two are what we named the “other services functions” while the last two pertain to “business services functions.” More precisely, R & D encompasses both fundamental scientific research and applied development and headquarters corresponds to administration, management, and accounting activities. Moreover, an additional analysis (not reported here) shows that investments in the sales and marketing function strengthen over time in comparison with investments in other service functions. Apart from the sales and marketing function, a further analysis shows however that the service functions, which dominate in the sector of industry do not systematically occupy the same rank in the sector of business services or in the sector named “other services.” For example, the second more important service function in the sector of industry is R & D but headquarters in the sector of business services and logistics in the sector of other services.

Table 1.

Number of Investment Projects by Service Functions and Sectors (1997–2011).

Sectors service functions	Industry	Business services	Other services	Percentage by function
Contact center	237	486	398	4.5
Headquarters	1,231	1,060	499	11.2
Internet data center	52	122	98	1.1
Research and development	1,879	851	116	11.5
Shared services center	182	119	114	1.7
Testing and Servicing	461	238	117	3.3
Logistics	1,358	45	1,574	12.0
Sales and marketing	4,173	6,285	3,150	54.8

Source: Our own database constructed from European Investment Monitor database.

Empirical Implementation

After the presentation of the reduced-form equation derived from the theoretical model by Head and Mayer (2004), we present the econometric model used for our empirical analysis and we describe the explanatory variables.

Reduced-form Equation

The equation we estimate is derived from the theoretical framework proposed by Head and Mayer (2004), anchored in the seminal NEG model (Krugman 1991). On the supply side, firms operate in monopolistic competition: each one produces a different variety of a good and is a monopolist on this variety. On the demand side, the utility of the consumers increases with the number of varieties offered on the market. The demand emanating from a representative consumer in region j for a firm located in region i is obtained by maximizing the consumer’s utility, modeled by a constant elasticity of substitution function. In particular, it depends on the price of the variety produced in region i and sold in region j. This price is the price after delivery and is defined as a combination of the mill price and iceberg-type transport cost. The optimum price set by a representative firm is obtained by applying a constant markup over marginal costs of production. Then, the net profit of this representative firm in a region i corresponds to the sum of profits realized on each market j diminished by the fixed cost associated with setting up a new plant in j. It is a decreasing function of production costs and an increasing function of the market potential, defined as the sum of the market capacities of region j, weighted by the transport costs between j and all other regions.

Considering the literature derived from Head and Mayer (2004), in which some additional explanatory variables have been added in order to provide a more complete description of the FDI location determinants, leads to the following reduced-form equation:

π_{i j} = α_{1} M P_{j} + α_{2} S_{j} + α_{3} d_{i j} + α_{4} Z_{i j} .

where π_ij denotes the profit of firm i in region j; MP_j is the market potential of region j, a part of the demand-side variables, the other part being per capita gross domestic product (GDP) that we add in our final specification (see below); S_j groups all the supply-side variables that includes production costs in j; d_ij is the distance (both geographical and cultural) between region i and j and Z_ij includes all the other control variables put forward in the literature. Z_ij can vary with i and j or only with j and includes policy variables and variables capturing agglomeration externalities.

Econometric Methodology

In order to model the firms’ location choices inside the regions of the EU, we assume that they take the same form as random utility models. Formally, if a firm has a choice between a set of j = 1, …, J possible regions, it will choose the region that yields the highest profit among the J region. In other words, it chooses j if $π_{i j} > π_{i l}, \forall l \neq j$ . Let us decompose the profit made by firm i in region j into a deterministic part V_ij and a stochastic part ∊ _ij as follows:

π_{i j} = V_{i j} + ∊_{i j},

where the stochastic part is independently and identically distributed (iid) according to a type I extreme value distribution. The deterministic part depends linearly on observable attributes of the region j and the firm i:

V_{i j} = β^{'} X_{i j} .

This form is sufficiently general to allow some observable characteristics to depend only on the region j as described in equation (1). Then, profit maximization implies that firm i chooses region j, if $V_{i j} - V_{i l} > ∊_{i l} - ∊_{i j}$ .

The most widely used model for analyzing firms’ location choices is the conditional logit model proposed by McFadden (1974). In our context, this model focuses on the characteristics of the regions that can be constant across investors or that can vary across firms. However, this model is affected by a serious drawback as it is based on the Independence from Irrelevant Alternatives (IIA) assumption, which implies that there is no correlation across alternative location choices. This assumption is clearly untenable since both unobserved characteristics of the firms and unobserved correlations across choices may cause of violation of this assumption. Consequently, a number of papers (Crozet, Mayer, and Mucchielli 2004; Head and Mayer 2004; Mayer, Mejean, and Nefussi 2010; Rasciute and Pentecost 2010; Spies 2010) are based on the estimation of nested logit models that allow relaxing the IIA assumption. In these models, an assumption is made on the sequence of decisions that a firm takes. For instance, in the EU case, a nested logit modeling will imply that a firm first chooses the country, then it chooses the region within the country. The IIA assumption is then supposed to apply within each group. An a priori choice has then to be made on the structure of the firms’ decisions. However, this strategy is not applicable in our case since no geographical structure will be relevant simultaneously for all services of our data set.

Recently, some papers (Defever and Mucchielli 2005; Defever 2006, 2012; Basile, Castellani, and Zanfei 2008) have adopted a MLM approach as suggested by Train (2003). MLM does not rely on the IIA assumption and it is not necessary to assume some hierarchy in the firms’ decisions. As the estimation of these models is numerically very intensive, their application to the location choices of firms is recent. In MLM, the error term is decomposed into two parts as follows (Train 2003):

∊_{i j} = v_{i j} + u_{i j} = μ_{i}^{'} Y_{j} + u_{i j},

where Y_j is a vector of variables for firm i and region j that are the variables in X_ij that only vary with the region; μ _i is the vector of randomly distributed parameters with density g(.) over all firms and u_ij is an iid error term with mean zero. The term μ _i ′ Y_j represents the correlation between the unobserved firm characteristics μ _i and the observable characteristics of the destination region Y_j. Variables directly linked to firms are considered as fixed coefficients. All the parameters of the model are estimated using simulated likelihood assuming a distribution for the random parameters. As we do not wish to make a priori assumptions as to the sequence of decisions of firms, we explore their location determinants using MLMs.

Explanatory Variables

As stated above, a firm will choose to locate in a particular European region if that region yields the highest profit among all the possible regions. More precisely, the probability that a firm locates in a particular region depends on how the characteristics of that region affects the firm profitability relative to the characteristics of other regions. Therefore, region-specific characteristics that can affect profits are considered as explanatory variables and have been selected according to the existing literature on location choices of multinationals (see the second section). However, for reasons of data availability, we sometimes had to use national data to characterize the regions. In that case, as many works in that field (as is, e.g., Head and Mayer 2004; Basile, Castellani, and Zanfei 2008), we implicitly make the assumption that the national data are a good approximation of the regional ones (see Table 2 for more details on the source and the availability of the data and Table 3 for their descriptive statistics). Moreover, it sometimes makes sense since some variables are identical or quite similar in regions belonging to the same country (such as the tax rate on corporate income or the governance performance).

Table 2.

Presentation of Explanatory Variables.

Variables	Definition	Availability	Source	Type
Dist.	Geographic distance between origin of the investor and destination region	1997–2011	CEPII-distance	Firm-region
Lang.	Dummy equal to one when the country of origin of the investor and the destination region share the same language	1997–2011	CEPII-distance	Firm-region
MP	Harris-type market potential	1997–2011	Constructed using Eurostat and distance—CEPII	Region
GDP/cap	GDP product at current market prices—purchasing power standard per capita	1997–2011	Eurostat	Region
Wage	Regional wages/regional GDP	1997–2010	Eurostat	Region
Unemp.	Unemployment rates (people over fifteen years)	1997–2011	Eurostat	Region
Dens.	Total population divided by the surface area	1997–2011	Eurostat	Region
Funds	EU structural funds—objective 1	1997–2011	Eurostat	Region
Law	The index “rule of law” ranges from −2.5 (weak) to 2.5 (strong) as a measure of governance performance	1998–2011	World bank and worldwide governance indicator	Country
Open	Total exports to GDP (percent)	1997–2011	UNECE	Country
Tax	Tax rate on corporate income	1997–2011	Eurostat	Country
Skill	Number of students in first and second stage of tertiary education (levels 5 and 6)	1998–2011	Eurostat	Country
Motor	Total length of motorways (kilometer) divided by the surface area	1997–2011	UNECE	Country
Rail	Total of length of railway (kilometer) divided by the surface area	1997–2011	UNECE	Country
Air	Air transport, registered carrier departures worldwide	1997–2011	World bank	Country
Internet	Fixed broadband Internet subscribers (per 100 people)	1998–2011	World bank	Country
Agglo.GL	Total number of investment projects in a region	1998–2011 or 2002–2011	Constructed using EIM database	Firm-region
Agglo.S1 and Agglo.S5	Count of the number of investment projects in the same sector, in the region of the host country but not in the same function and in the same group the year preceding investment or five years preceding investment	1998 to 2011 or 2002–2011	Constructed using EIM database	Firm-region
Agglo.F1 and Agglo.F5	Count of the number of investment projects in the same function, in the region of the host country but not in the same sector and in the same group the year preceding investment or five years preceding investment	1998–2011 or 2002–2011	Constructed using EIM database	Firm-region
Agglo.G1 and Agglo.G5	Count of the number of investment projects in the same group, in the region of the host country, the year preceding investment or five years preceding investment	1998–2011 or 2002–2011	Constructed using EIM database	Firm-region

Note: CEPII, Centre d’études prospectives et d’informations internationales; EIM, European Investment Monitor; GDP = gross domestic product; MP = market potential; UNECE, United Nations Economic Commission for Europe.

Table 3.

Descriptive Statistics.

Variables	Mean	Standard deviation
Dist.	2,630.90	2,154.37
Lang.	–	–
MP	6,435,356	5,180,115
GDP/cap	20,747.07	8,557.37
Wage	.00463	1,666.95
Unemp.	8.91	5.36
Dens.	372.53	897.56
Funds	–	–
Law	1.07	.64
Open	53.82	27.77
Tax	27.69	8.33
Skill	601,702.35	758,327
Motor	.1892	.8919
Rail	.054	.039
Air	192,164.34	271,414.10
Internet	10.19	10.98
Agglo.GL	62.97	77.59
Agglo.S1	7.48	16.29
Agglo.S5	29.65	66.73
Agglo.F1	26.67	51.35
Agglo.F5	99.98	197.61
Agglo.G1	.02	.13
Agglo.G5	.05	.29

Note: GDP = gross domestic product; MP = market potential.

Following the discussion in reduced-form equation subsection, our explanatory variables capture various dimensions of a region and can be classified into five categories: the demand and the supply side of the market, the policy dimension, the geographic and cultural distances, and finally the agglomeration economies. We have chosen to separate here the variables included in Z in Equation (1) into two categories instead of one to facilitate the presentation.

Demand-side variables

To assess the size of regional markets and thus market demand, we consider two traditional determinants which should be positively correlated with the firms’ location choices: the regional GDP per capita, which controls for the host region’s development level, and a market potential indicator, which captures the fact that the market horizon of any multinational is much larger than the region in which it has established its plants (Casi and Resmini 2010). Constructed following Harris (1954), the advantage of this second measure is to take into account the demand emanating from the local market, weighted by internal distance, and the demand emanating from neighboring regions, weighted by bilateral distance. This indicator is defined by,

M P_{j} = G D P_{j} + \sum_{l \neq j} G D P_{l} / D_{j l},

where MP_j is the market potential for region j, GDP_j is the GDP in region j, and D_jl is the distance between capitals of regions j and l. This measure of market potential has been used in a large number of empirical studies (e.g., among many others Pusterla and Resmini 2007; Basile, Castellani, and Zanfei 2008; Py and Hatem 2009; Montout and Robin 2012). Note that due to data limitations for our extended sample of European regions, we are not able to construct a market potential based on bilateral trade flows (as in Head and Mayer 2004).

Supply-side variables

To characterize the local input market, we select several variables. As for the labor market, we use measures of wages, unemployment rate, and skill. In most papers, as for instance in Pusterla and Resmini (2007) or more recently in Cieslik (2013), which both analyze eastern European states, it was expected that a high wage rate would be negatively associated with the number of foreign firms located in a region. However, as Basile, Castellani, and Zanfei (2008) explain it clearly: lower wages may attract firms searching for lower labor costs, but high wages may be the signal of skilled workers and then attract firms in higher value added activities. This point is very important in our case since service functions are varied and do not necessarily need identical skills. The same uncertainty appears for the expected sign of the unemployment variable: a high rate may be an indication of a large supply of labor, which would attract firms but it can be seen by others as a signal of a relatively rigid labor market which would discourage them.

Conversely, the impact of the skill variable is unambiguous. To the extent that foreign firms locate in a place abroad in order to increase their efficiency, the expected sign of the relation between the probability of a region of being chosen and the level of skills of its labor force is positive. More skilled workers are supposed to have a better productivity. In our study, the level of education of the labor force is measured by the number of students in the first and second stages of tertiary education. The data are available at the national level and not at the regional level. However, we assume that this is a better proxy than the secondary school enrolment ratio, available for regions but not sufficiently discriminating within the EU.

Finally, to complete the characterization of the local input market, as Basile, Castellani, and Zanfei (2008), we introduce population density. This variable may proxy the cost of land. The higher is the population density and the higher is the cost of land, which should reduce the region attractiveness. However, this variable may also capture the effect of the agglomeration of consumers, which should increase the region attractiveness.

Policy variables

Economic policies may affect foreign firms’ location choices through various incentives. Foreign investors may be sensitive to the country fiscal policy, to the legal enforcement and the business environment. To capture these effects, we introduce two variables, the tax rate on corporate income and a measure of governance performance. The impact of corporate tax is not univocal (Bénassy-Quéré, Gobalraja, and Trannoy 2007). By reducing profits, tax may discourage firms’ location choices. However, a high level of taxation may be the signal of an abundant supply of public goods, which should increase the attractiveness of a place (Gabe and Bell 2004; Dembour 2008). To capture governance performance, we use an index constructed by the World Bank and named “rule of law”: it reflects perceptions of the extent to which agents have confidence in and abide the rules of the society and in particular the quality of the contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence. We expect a positive sign for the coefficient associated with this variable.

We also introduce a variable for economic openness: as argued by Moosa and Cardak (2006), it is usually expected to find a positive relationship between FDI and a country’s openness (measured here by the total exports to GDP). However, FDI can be a strategy used by a foreign firm to supply a foreign market when a country is relatively closed to exchanges (with other ones) and the sign of the relationship will be negative in that case.

Moreover, as in Groh and Wich (2012), we add three infrastructure variables, which take into account the length of motorways and railways and the number of flights as for air transport. Although these data are also available only at the country level, they give an idea of the easiness of access to the market from a given location. We expect that regions with a good accessibility are more attractive for foreign firms and thus are likely to receive more investments than other regions. To better characterize these various ways of communication, a variable relative to Internet is added: it is expected that a high number of Internet subscribers exert a positive influence on foreign firms looking for a foreign location.

Finally, as our study concerns European countries, we introduce a variable with the aim of capturing the role of EU Cohesion policy in attracting foreign investors in Europe. This European common policy, which represents about one-third of the EU budget, aims at achieving social and economic cohesion and thus reducing regional imbalances. Basile, Castellani, and Zanfei (2008) control for its role introducing the amount of structural funds allocated to each region. We proceed more simply with a dummy variable taking value 1 if a region is eligible for objective 1, which is the most important one for the studied period. Named “convergence objective,” the objective 1 covers Europe’s poorest regions whose per capita GDP is less than 75 percent of the EU average.

Geographic and cultural variables

As Py and Hatem (2009), we introduce two distance variables, constructed by the Centre d’études prospectives et d’informations internationales, in order to measure transaction costs associated with the decision of a location abroad. Firstly, we use a variable designed to measure geographic distance: it relates to the physical distance between the country of origin of the investor and the host region of investment. According to Gorbunova, Infante, and Smirnova (2012), the greater the geographic distance the more FDI stock grows because given a greater distance, FDI tends to substitute trade. On the contrary, according to Py and Hatem (2009), the relationship between geographic distance and FDI should be negative because it is easier to invest in a country, which is near the country of origin of the investor.

Secondly, we use a dummy variable designed to measure cultural distance that is to say the potential existence of a common language between the country of origin of the investor and the host region of investment. The expected sign in this case is positive because it is easier to deal with persons who have the same language.

Agglomeration variables

Agglomeration externalities are another factor that may influence firms’ location choices. Even if the number of firms in a place increases competition and may reduce its attractiveness, positive externalities resulting from an easier access to input needed for production and technological developments associated with geographic proximity may counterbalance the drawbacks and lead to the concentration of activities. Identified by Marshall (1890), these externalities explain why agglomeration of activities arises and persists and they have been analyzed in a large number of empirical studies concerning Europe (e.g., among many others Disdier and Mayer 2004; Head and Mayer 2004; Devereux, Griffith, and Simpson 2007; Montout and Robin 2012). Many works in the economic geography literature show that a firm’s location decision is influenced by location choices of previous firms with similar attributes (the same industry or national origin, e.g.) (Devereux and Griffith 1998; K. Head and Mayer 2004). Moreover, foreign firms’ location behavior has been reported to be more significant than local incumbents’ (Head, Ries, and Swenson 1999; Basile 2004). However, most studies are concerned with firm agglomeration with respect to sector (sectorial agglomeration). A few works only deal with functional agglomeration (Duranton and Puga 2005; Defever and Mucchielli 2005; Defever 2006, 2012; Py and Hatem 2009) and few papers consider several agglomeration variables together (Head and Mayer 2004; Defever and Mucchielli 2005; Defever 2006, 2012; Py and Hatem 2009; Montout and Robin 2012). Our intention is to elaborate on agglomeration variables to provide a thorough analysis of agglomeration processes in FDI host countries.

In this article, we control for the assumption that firms may benefit from locating close to other firms with the aid of four variables. Firstly, we look at the impact of the global agglomeration that is to say the number of investment projects in a given region. Following a cumulative process, firms tend to locate where other foreign firms are already present. Secondly, we consider the effect of the sectorial agglomeration that is to say the number of investment projects in the same sector in the same region of the host country. Firms tend to locate in regions where a high number of firms belonging to their operating sector are already established because they probably used similar technologies, inputs, and types of workers. Thirdly, we take into account the functional agglomeration with the number of investment projects in the same function in the same region in order to see if there are similarities between the two kinds of agglomeration, sectorial, and functional. Finally, we look at the group agglomeration by considering the number of investment projects in the same group, in the same region of the host country.

Our approach allows differentiating between sectorial agglomeration and the functional one, for three sectors and a great number of functions during a long time. Moreover, it is more complete than the other studies because two types of computations are made. The first one corresponds to the horizon of one year for 1998 to 2011: the agglomeration variables include the number of firms in the same sector/function/group set up in the same region the year preceding the investment choice. The second one corresponds to the horizon of five years for 2002 to 2011: the agglomeration variables include the number of firms in the same sector/function/group set up in the same region the five years preceding the investment choice. With this distinction, our aim is to capture possible differences between short- and medium-term effects in the process of agglomeration.

The correlation matrices between national variables, regional variables, and origin–destination variables are reported, respectively, in Tables 4, 5, and 6. No correlation is higher than .46 in absolute values in the first two cases. In the third case, unsurprisingly, the correlations between agglomeration variables are higher, especially with the general agglomeration variable and up to .83 between the general agglomeration variable and the functional agglomeration variable. However, as we show in the next section, these variables remain individually highly significant, so that multicollinearity between them is not an issue.

Table 4.

Correlation Matrix between National Variables.

	Law	Open	Tax	skill	Motor	Rail	Air	Internet
Law	1	.2252	.3044	.0860	.1138	.1462	.3857	.3814
Open		1	−.1403	−.4657	.2068	.4480	−.3272	.1827
Tax			1	.2937	.1580	.2073	.3666	−.1667
Skill				1	−.1489	.1147	.7984	.1527
Motor					1	−.2549	−.1143	.0061
Rail						1	.1253	.1528
Air							1	.2077
Internet								1

Note: GDP = gross domestic product; MP = market potential.

Table 5.

Correlation Matrix between Regional Variables.

	MP	GDP/cap	Wage	Unemp.	Dens.	Funds
MP	1	.5023	−.1622	.3044	−.2042	.2421
GDP/cap		1	−.3776	.4328	−.2708	.5011
Wage			1	.0679	.3739	−.1948
Unemp.				1	−.0761	.1413
Dens.					1	−.1741
Funds						1

Note: GDP = gross domestic product; MP = market potential.

Table 6.

Correlation Matrix between Origin–Destination.

	Dist.	Agglo.GL	Agglo.S1	Agglo.F1	Agglo.G1
Dist.	1	.1441	.1194	.1083	−.0015
Agglo.GL		1	.7040	.8385	−.0081
Agglo.S1			1	.6965	−.0079
Agglo.F1				1	.0296
Agglo.G1					1

Results

In this section, we report the estimation results for the MLM models applied to all functions and sectors. With respect to estimation, we use 500 replications and we assume that the coefficients associated with the variables observed at the regional or at the national level are random coefficients affected by unobserved firm characteristics. We assume a normal distribution and for these variables, we will also report their mean and standard deviation. Conversely, the variables that are specific to each firm-region pair are already variable over firms by definition so that the associated coefficients are assumed to be fixed. The models were estimated using simulated maximum likelihood estimation (Train 2003). Because some variables are measured at the national level, we use clustered-robust standard errors.

Firstly, we present the estimation results concerning determinants of investment location choices for four service functions whatever the sector with agglomeration variables computed for a horizon of one year.⁷ The four selected functions are those that are privileged by foreign firms over the global period: they represent almost 90 percent of the total number of investment projects when they are classified by function as we have seen in Table 1.⁸ These estimations represent our benchmark. Secondly, we present the estimation results of determinants of investment location choices for each function in each sector.

Determinants of Location Choices for Service Functions Whatever the Sector: A Benchmark

Table 7 displays the estimation results of the MLMs estimated for four service functions. A majority of coefficients are significant in this table, but their signs vary according to functions. For instance, the greater the geographic distance between the origin of the investor and the host region is, the greater the probability of that region being chosen for an investment in sales and marketing and headquarters functions is. An interpretation is that they may be better implemented through a local presence rather than in a remote way. Conversely, the sign is negative for logistics and R & D, meaning that it is better to be near the origin country for these functions. For this latter function, this result is consistent with previous results on the location determinants of R & D activities (Kumar 1996; Ambos 2005). The signs also vary for the wage variable: firms which invest in sales and marketing prefer lower wages whereas they prefer high wages, signal of skilled workers, when they invest in logistics and in headquarters, which supports Basile, Castellani, and Zanfei (2008) findings. The coefficient of the unemployment variable is only significant in two cases, sales and marketing (at 10 percent) and headquarters (at 1 percent), with a negative sign. It seems that in these cases, a high unemployment rate is the signal of labor market rigidity, which is not attractive for foreign investors.

Table 7.

The Determinants of FDI Location Choices by Service Function—Mixed Logit Estimation Results.

Service functions	Sales and marketing	Logistics	Headquarters	R & D
Explanatory variables	Coefficients	Coefficients	Coefficients	Coefficients
Dist.	.141*** (.018)	−.150*** (.036)	.177*** (.049)	−.102** (.042)
Lang.	.736*** (.028)	.448*** (.061)	.860*** (.066)	.663*** (.060)
MP	.070** (.032)	−.136** (.068)	−.066 (.082)	−.157** (.070)
GDP/cap	.940*** (.050)	−.239** (.114)	.826*** (.126)	.671*** (.122)
Wage	−.487*** (.066)	.581*** (.123)	.906*** (.220)	.045 (.151)
Unemp.	−.053* (.027)	−.048 (.057)	−.324*** (.075)	.014 (.059)
Dens.	.372*** (.072)	1.692*** (.588)	.975*** (.256)	3.613*** (.881)
Dens. (squared)	−.023*** (.006)	−.154*** (.053)	−.006*** (.020)	−.347*** (.081)
Funds	−.023 (.046)	−.015 (.078)	−.406*** (.117)	−.028 (.080)
Law	.296*** (.055)	−.496*** (.092)	.391** (.178)	.086 (.109)
Open	−.050 (.057)	.705*** (.137)	.134 (.145)	−.120 (.129)
Tax	.343*** (.078)	.471** (.183)	−.366* (.216)	−.228 (.150)
Skill	.105*** (.031)	.174*** (.063)	.007 (.075)	.085 (.069)
Motor	−.001 (.134)	−.457 (1.937)	−.061 (.327)	−.316 (.297)
Rail	−.727 (.612)	−3.817*** (1.314)	−2.909* (1.626)	−.274 (1.283)
Air	−.148*** (.021)	−.064 (.055)	.078 (.055)	−.052 (.044)
Internet	.210*** (.043)	.272*** (.087)	.189* (.112)	.054 (.086)
Agglo.GL	.864*** (.021)	.792*** (.035)	.906*** (.044)	.792*** (.037)
Agglo.S1	.313*** (.014)	.192*** (.041)	.450*** (.032)	.275*** (.034)
Agglo.F1	.158*** (.018)	.321*** (.037)	.072** (.035)	.263*** (.036)
AGGLO.G1	.307 (.187)	2.390*** (.173)	2.650*** (.264)	3.092*** (.146)
Error component (SD)
MP	.001 (.255)	−.003 (.865)	−.029 (.495)	.016 (.615)
GDP/cap	−.002 (.450)	.007 (1.158)	.053 (.740)	−.012 (.930)
Wage	−.009 (.384)	.012 (.873)	−.327 (.645)	.054 (.653)
Unemp.	−.175 (.201)	−.222 (.328)	1.085*** (.126)	−.167 (.343)
Dens.	.001 (.065)	.394** (.173)	−.140 (.120)	.996*** (.209)
Dens. (squared)	.002 (.001)	.004 (.015)	−.002 (.010)	.016 (.012)
Law	−.892*** (.062)	−.015 (.398)	.845*** (.206)	−.125 (.325)
Open	.003 (.147)	.118 (.362)	−.077 (.289)	−.018 (.322)
Tax	.344 (.227)	.451 (.417)	1.004*** (.307)	.242 (.377)
Skill	−.047 (.055)	−.006 (.152)	.025 (.193)	.005 (.136)
Motor	.027 (.064)	.112 (4.254)	−.084 (1.014)	−.169 (.977)
Rail	5.078*** (1.346)	−1.383 (4.150)	8.397*** (2.198)	−1.122 (3.839)
Air	−.013 (.043)	.105 (.988)	−.062 (.102)	.001 (.102)
Internet	−.175 (.143)	−.334** (.161)	−.113 (.352)	.314 (.187)
Likelihood	−52,796	−13,443	−9,830.6	−12,264

Note: Standards errors in parentheses. FDI = foreign direct investment; GDP = gross domestic product; MP = market potential.

*Significant at 10 percent.

**Significant at 5 percent.

***Significant at 1 percent.

Coefficients for variables of language and GDP per capita are always significant and positive as it is expected, except for logistics as for the latter indicator. If we consider jointly the density variable, for which the coefficient is always significant and positive, with the squared density variable, for which the coefficient is systematically significant and negative, the results show a nonlinear effect as documented in NEG models. A concentration of people in a given place is positive in terms of demand, but it may be negative in terms of costs of land, pollution, or traffic congestion. The coefficient of the skill variable is only significant in two cases, sales and marketing and logistics, with a positive sign, as expected.

With respect to the policy variables side, there are also surprising results. For instance, structural funds act negatively in the case of an investment project in the headquarters function, firms preferring to locate in richer regions. At the same time however, the negative coefficient for the tax variable (significant at 10 percent) seems to indicate that a high level of taxation, by reducing profits, discourages location for the headquarters function. For sales and marketing and for logistics, the sign for the tax variable is positive, as we expect it since a high taxation is the signal of an abundant supply of public goods and a good attractiveness of a region for these types of functions. It is also worth noting that as expected, the sign of the coefficient obtained for the variable of governance performance is significant and positive for sales and marketing and headquarters functions, but it is surprisingly negative for logistics, suggesting that the worse the governance performance in a country, the higher the likelihood of transport firms location and more generally of the logistic function.

Mixed results are also obtained for communication means. As expected, the quality of Internet is significant with a positive sign for three functions, sales and marketing and logistics (at 1 percent) and headquarters (at 10 percent), but the length of motorways is never significant. The rail and air transport variables are only significant at 1 percent, respectively, for logistics and for sales and marketing with a negative sign. Although we must interpret the latter result with caution due to the way the air transport variable is measured (see Table 2), one possible interpretation for sales and marketing is that the commercial function can be performed in the seat whenever airlines’ connections are good. For logistics, the worse the rail network in a region, the more the foreign firm is incited to locate a subsidiary.

Finally, we must have a look at the market potential and agglomeration variables. Surprisingly, the coefficient of the first is only significant (at 5 percent) with a positive sign for sales and marketing and significant with a negative sign for logistics and R & D. However, other estimates show that this result only appears when we take into account agglomeration variables.⁹ When the latter are not included, the coefficient associated with potential market variable is significant and positive, as expected. In fact, it seems that agglomeration variables capture the potential market effect, weakening the impact of this variable.

Otherwise, we may note that agglomeration variables are almost all significant for the four functions and these results are also the same for a horizon of five years. Both sectorial and functional agglomeration variables play an important role for service activities, which contradicts the results of Davis and Henderson (2008). Moreover, previous location choices impact on further location projects (group agglomeration), except for sales and marketing activities, as already mentioned by Defever (2012). In general, externalities play a very important role in the decision of a firm to locate in a place or in another. A deeper analysis by function and by sector will help to discriminate between them.

Determinants of Location Choices for Service Functions: Differences between Sectors

As in the previous section, we only interpret the estimation results for the four previous functions. For the other functions, we report in the Appendix the estimation results for the functions for which there are enough observations: Contact Center and Testing and Servicing (Tables A2 and A3).

Our estimation results may be analyzed in two ways. Firstly, the variables that impact on the location of all service functions under study for each of the three sectors (industry, business services, and other services) can be studied. Secondly, an analysis for each service function according to sectors can be carried out. The first analysis aims at revealing whether there are some variables specific to each of the sectors for the location of service functions whatever their nature. Our results (Tables 8 –11) do not support such an assumption: location determinants in each sector vary according to the service function considered (even if some of the determinants can sometimes be observed for two of the service functions at stake) or are also observed in other sectors. Our second analysis yields more interesting results. For each service function, we propose to bring to light only specific location determinants, which differentiate sectors, considering that common variables (affecting all sectors) have already been specified in our benchmark analysis.

Table 8.

The Determinants of FDI Location Choices in Sales and Marketing by Sectors—Mixed Logit Estimation Results.

Sectors	Industry	Business services	Other services
Explanatory variables	Coefficients	Coefficients	Coefficients
Dist.	.070* (.038)	.231*** (.029)	.048 (.032)
Lang.	.533*** (.056)	.845*** (.041)	.798*** (.060)
MP	.136** (.057)	.087* (.051)	.023 (.068)
GDP/cap	.728*** (.093)	1.187*** (.078)	.879*** (.101)
Wage	−.257* (.136)	−.452*** (.110)	−.584*** (.123)
Unemp.	−.129*** (.050)	−.098** (.043)	.038 (.056)
Dens.	.544** (.211)	.323*** (.115)	.335* (.189)
Dens. (squared)	−.042** (.017)	−.016* (.001)	−.018 (.015)
Funds	−.020 (.089)	−.006 (.070)	−.001 (.088)
Law	.228** (.105)	.462*** (.091)	.173* (.096)
Open	−.089 (.127)	.086 (.096)	.007 (.115)
Tax	.299** (.145)	.510*** (.129)	.244* (.139)
Skill	.176*** (.066)	.212*** (.060)	.042 (.065)
Motor	−1.752 (1.559)	−.120 (.262)	−0,019 (.220)
Rail	1.419 (1.228)	−2.343** (.977)	−4.613*** (1.289)
Air	−.123*** (.044)	−.149*** (.033)	−.211*** (.041)
Internet	.469*** (.900)	.256*** (.069)	−.076 (.081)
Agglo.GL	.910*** (.036)	.914*** (.034)	.796*** (.044)
Agglo.S1	.287*** (.029)	.138*** (.029)	.313*** (.034)
Agglo.F1	.100*** (.029)	.231*** (.031)	.281*** (.037)
Agglo.G1	.580* (.349)	.206 (.290)	.296 (.351)
Error component (SD)
MP	.001 (.441)	−.006 (.345)	.002 (.489)
GDP/cap	−.017 (.743)	−.015 (.559)	−.027 (.801)
Wage	.184 (.441)	.051 (.459)	−.026 (.684)
Unemp.	−.010 (.612)	.637*** (.107)	−.054 (.636)
Dens.	.128 (.113)	−.033 (.089)	−.068 (.126)
Dens. (squared)	−.003 (.010)	−.002 (.007)	−.001 (.011)
Law	.976*** (.112)	.958*** (.099)	.694*** (.127)
Open	.026 (.270)	−.041 (.209)	−.041 (.264)
Tax	.063 (.484)	.638*** (.243)	.226 (.509)
Skill	.044 (.106)	−.183*** (.054)	−.149** (.058)
Motor	1.390 (1.032)	.136 (.781)	−.088 (.757)
Rail	3.947 (2.910)	−4.518*** (2.025)	9.397*** (1.770)
Air	−.089 (.065)	.002 (.067)	.010 (.075)
Internet	.626*** (.142)	−.198 (.195)	.011 (.290)
Likelihood	−17,371	−22,953	−12,204

Note: Standards errors in parentheses. FDI = foreign direct investment; GDP = gross domestic product; MP = market potential.

*Significant at 10 percent.

**Significant at 5 percent.

***Significant at 1 percent.

Table 9.

The Determinants of FDI Location Choices in Logistics by Sectors—Mixed Logit Estimation Results.

Sectors	Industry	Other services
Explanatory variables	Coefficients	Coefficients
Dist.	−.196*** (.061)	−.152*** (.045)
Lang.	.360*** (.099)	.593*** (.082)
MP	−.218** (.110)	−.067 (.091)
GDP/cap	−.515*** (.186)	−.027 (.149)
Wage	.574** (.226)	.534*** (.153)
Unemp.	−.164* (.093)	.038 (.078)
Dens.	1.257** (.636)	1.140** (.469)
Dens (squared)	−.115** (.057)	−.103** (.042)
Funds	.021 (.113)	−.103 (.111)
Law	−.335** (.165)	−.559*** (.112)
Open	1.022*** (.230)	.368** (.171)
Tax	.644** (.316)	.275 (.209)
Skill	.284*** (.104)	.114 (.085)
Motor	−.412 (3.682)	−.511 (1.136)
Rail	−2.110 (2.078)	−5.257*** (1.804)
Air	−.045 (.091)	−.106 (.072)
Internet	.337** (.131)	.252** (.115)
Agglo.GL	.813*** (.054)	.753*** (.046)
Agglo.S1	.305*** (.058)	.064 (.062)
Agglo.F1	.433*** (.056)	.288*** (.052)
Agglo.G1	3.342*** (.241)	1.180*** (.257)
Error component (SD)
MP	−.016 (1.129)	.001 (1.248)
GDP/cap	−.041 (1.455)	.001 (1.808)
Wage	−.047 (1.209)	.010 (1.120)
Unemp.	.513* (.269)	.167 (.570)
Dens.	.257 (.240)	.130 (.247)
Dens. (squared)	.004 (.021)	.010 (.020)
Law	−.021 (.634)	−.013 (.570)
Open	−.600 (.413)	−.043 (.541)
Tax	.751 (.512)	−.026 (.944)
Skill	−.017 (.231)	−.024 (.176)
Motor	.051 (10.24)	−.019 (9.721)
Rail	−.390 (6.359)	−5.190 (4.731)
Air	−.006 (.167)	.036 (.121)
Internet	.064 (.521)	−.412** (.216)
Likelihood	−5,803.6	−7,401.6

Note: Estimations for logistics projects in business services were removed because they related to a too low number of projects (45). FDI = foreign direct investment; GDP = gross domestic product; MP = market potential.

*Significant at 10 percent.

**Significant at 5 percent.

***Significant at 1 percent.

Table 10.

The Determinants of FDI Location Choices in Headquarters by Sectors—Mixed Logit Estimation Results.

Sectors	Industry	Business services	Other services
Explanatory variables	Coefficients	Coefficients	Coefficients
Dist.	.128* (.077)	.405*** (.136)	.139* (.080)
Lang.	.601*** (.101)	1.163*** (.138)	1.086*** (.164)
MP	.042 (.121)	−.233 (.151)	.086 (.197)
GDP/cap	.757*** (.183)	1.325*** (.253)	.595* (.326)
Wage	.773*** (.284)	.922** (.447)	.985* (.540)
Unemp.	−.280*** (.108)	−.374** (.150)	−.352* (.182)
Dens.	1.266*** (.457)	3.885*** (1.205)	.732* (.422)
Dens. (squared)	−.088** (.037)	−.282*** (.090)	−.038 (.033)
Funds	−.504*** (.189)	−.318 (.205)	−.277 (.299)
Law	.421* (.233)	.569 (.347)	.364 (.386)
Open	.011 (.248)	.144 (.338)	.166 (.292)
Tax	−.037 (.336)	−.456 (.422)	−.504 (.489)
Skill	.007 (.109)	.282* (.158)	−.019 (.211)
Motor	−.620 (3.191)	−.218 (1.086)	.108 (.264)
Rail	−2.177 (2.137)	−6.745** (3.154)	−5.422 (4.399)
Air	.060 (.073)	−.026 (.131)	.198 (.169)
Internet	.335* (.174)	.359 (.230)	−.109 (.335)
Agglo.GL	.885*** (.062)	1.127*** (.088)	.904*** (.115)
Agglo.S1	.344*** (.056)	.190** (.074)	.580*** (.089)
Agglo.F1	−.025 (.051)	.229*** (.070)	.120 (.083)
Agglo.G1	−	2.288*** (.570)	3.330*** (.586)
Error component (SD)
MP	.029 (.817)	.068 (.545)	.039 (.932)
GDP/cap	−.080 (1.221)	.271 (.856)	−.099 (1.592)
Wage	.145 (1.068)	−.397 (.899)	.702 (.900)
Unemp.	−1.040*** (.187)	1.347*** (.219)	.941*** (.333)
Dens.	.057 (.191)	−.940*** (.283)	.071 (.243)
Dens. (squared)	.022 (.013)	.010 (.014)	−.005 (.021)
Law	.581 (.410)	−.757* (.399)	1.485*** (.394)
Open	.403 (.390)	1.276*** (.372)	−.097 (.481)
Tax	−1.510*** (.399)	.898* (.510)	.368 (1.121)
Skill	−.026 (.165)	.063 (.176)	.162 (.136)
Motor	.199 (6.193)	−.590 (1.983)	.423 (1.759)
Rail	−4.670 (4.804)	3.047 (6.122)	14.79*** (4.692)
Air	.022 (.167)	.210 (.151)	.239 (.181)
Internet	−.626* (.294)	.144 (.541)	.748 (.555)
Likelihood	−4,910.9	−3,190.9	−1,646.8

Note: Standards errors in parentheses. FDI = foreign direct investment; GDP = gross domestic product; MP = market potential.

*Significant at 10 percent.

**Significant at 5 percent.

***Significant at 1 percent.

Table 11.

The Determinants of FDI Location Choices in R & D by Sectors—Mixed Logit Estimation Results.

Sectors	Industry	Business services	Other services
Explanatory variables	Coefficients	Coefficients	Coefficients
Dist.	−.154*** (.055)	.060 (.080)	−.058 (.216)
Lang.	.528*** (.075)	.897*** (.117)	1.610*** (.396)
MP	.025 (.087)	−.493*** (.140)	−.053 (.390)
GDP/cap	.606*** (.154)	.747*** (.231)	.944 (.796)
Wage	.093 (.189)	.170 (.285)	−1.205 (1.276)
Unemp.	−.063 (.074)	.055 (.119)	−.233 (.376)
Dens.	1.166*** (.449)	4.599** (1.860)	.292 (1.644)
Dens. (squared)	−.123*** (.041)	−.437** (.171)	−.045 (.146)
Funds	−.062 (.106)	.080 (.144)	−.902** (.043)
Law	.221* (.131)	−.104 (.218)	.262 (.927)
Open	−.192 (.163)	−.100 (.262)	.857 (.892)
Tax	−.163 (.210)	−.154 (.292)	−.538 (1.040)
Skill	.050 (.085)	.223 (.150)	.428 (.523)
Motor	−.427 (.527)	−.337 (.577)	−.414 (1.514)
Rail	1.054 (1.623)	−4.269 (2.894)	−.141* (.823)
Air	−.014 (.057)	−.156 (.097)	.011 (.346)
Internet	.024 (.104)	.092 (.186)	−.319 (.613)
Agglo.GL	.699*** (.045)	.963*** (.078)	.936*** (.193)
Agglo.S1	.450*** (.045)	.022 (.064)	.330 (.228)
Agglo.F1	.139*** (.045)	.502*** (.069)	.360* (.215)
Agglo.G1	3.211*** (.164)	2.385*** (.331)	3.603*** (1.306)
Error component (SD)
MP	−.011 (.957)	−.015 (.853)	−.007 (4.154)
GDP/cap	.010 (1.486)	.020 (.982)	−.182 (3.009)
Wage	.001 (1.277)	.331 (.604)	.003 (4.654)
Unemp.	.064 (.724)	.438 (.319)	.582 (1.039)
Dens.	.249 (.186)	−1.323*** (.446)	.181 (.844)
Dens. (squared)	−.023** (.011)	.016 (.018)	−.001 (.073)
Law	−.041 (.613)	.024 (.441)	.352 (1.807)
Open	.047 (.467)	.526 (.412)	.465 (1.858)
Tax	−.677 (.452)	.486 (.506)	.815 (1.859)
Skill	.013 (.177)	−.031 (.159)	−.376 (.580)
Motor	−.297 (1.308)	.325 (1.351)	−.503 (3.381)
Rail	.656 (5.449)	10.42*** (3.874)	.208 (22.62)
Air	.001 (.135)	−.196* (.100)	.094 (.499)
Internet	.410* (.219)	.640** (.270)	.687 (.896)
Likelihood	−8,125.9	−3,599.7	−441.96

Note: Estimates for other services may be not significant due to a low number of observations. Standards errors in parentheses. FDI = foreign direct investment; GDP = gross domestic product; MP = market potential; R & D = research and development.

*Significant at 10 percent.

**Significant at 5 percent.

***Significant at 1 percent.

Table 8 shows that the location of sales and marketing units in industry as well as in business services is negatively correlated to the unemployment rate and the squared density variable. This could reveal that firms in these sectors are more sensitive than their counterparts in the other service sectors to the rigidity of the labor market and to negative externalities due to the concentration of population in a place. What also appears as relevant factors for industry and business services firms in order to locate sales and marketing units are the education level of the workforce and the development of information and communication technologies. As regards firms belonging to the service sector, contrary to industrial firms, an important, determinant of the sales and marketing function is the rail variable: for them, the more rail infrastructures are developed, the less FDI is needed certainly because services can be provided from previously located subsidiaries or from the parent company (service exports).

Table 9 displays estimation results pertaining to the location choices of logistics activities according to sectors. In industry, the location choices of logistics units are surprisingly negatively correlated to demand-side variables—which suggests that location choices are mainly motivated by industrial organization considerations¹⁰ rather than market ones - but positively correlated to the tax variable, the workforce level of education and to the sectorial agglomeration variable. All these variables are not significant for firms belonging to service sectors. As for sectorial agglomeration, this means that logistics units of industrial firms tend to locate close to other logistics units belonging to the same field of industrial activity,¹¹ which is not the case in the other services sector. In other services (mainly the transport sector), location choices are more sensitive to developed raid infrastructures than their industrial counterparts. As already mentioned, it seems that the more the rail infrastructures in a country, the more transport companies can perform their activities from abroad.

Table 10 summarizes estimation results for the location choices of headquarters activities by sectors. We observe that the poorer EU regions (targeted by EU structural funds) attract less headquarters location projects of industrial firms than richer ones. Moreover, the level of development of rail connections is negatively correlated to location choices of headquarters in the business service sector, suggesting that if connections are underdeveloped, a local headquarter is required. Finally, functional agglomeration appears as an important determinant of the location of business services headquarters contrary to headquarters in industry and other services, meaning that for firms in business services, the presence of other headquarters in the same place is valued. Surprisingly for the location of headquarters activities, the Internet penetration rate does not seem to be a crucial determinant for any of the sectors under study.

Table 11 gives the main location determinants of R & D units according to sectors. For industrial firms, R & D location choices are negatively affected by the geographical distance with home country, which is not the case in service sectors. We may think that according to sectors, motives of R & D internationalization are different: when a company operating in business services sets up an R & D unit abroad, most of the time it follows its clients and such a unit is dedicated to support its clients’ activities rather than its own R & D: the distance does not play then, however it could explain why market potential is an important determinant. This is not the case for industrial companies that prefer nearest locations for their R & D activities, as mentioned earlier. Moreover, contrary to firms in service sectors, industrial firms search for other firms performing their activities in the same sector in order to choose a location for an R & D unit. However, both categories of firms look for other R & D activities to locate their own R & D units (functional agglomeration). Finally, whatever the sector, the skill variable is never significant, which is an odd result for R & D location choices. One possible explanation is that human resources do not matter much in R & D internationalization strategies of firms in EU since EU provides quite similar skills.

Conclusions

Today, services play a major role in all modern economies. They are present as input and as output in the production process and in trade flows, but they also represent a great proportion of foreign direct investments. Our original analysis of FDI location determinants in the EU-28 regions from the double point of view of service functions and sectors results in three key findings.

Firstly, service functions location choices are different according to the three sectors considered in this study, that is, industry, business services, and other services. At least, we might think that location patterns in business services could be similar either to industry (as those services are provided at least in part to industrial clients) or to the other services sector (as both of them belong to the broad category of services). However, location determinants of service functions in business services are often specific to this sector. This result contradicts Py and Hatem (2009) assumption according to which service functions location principles are sector independent.

Secondly, traditional location determinants highlighted by the literature, as for example, geographical distance between the country of origin of the investor and the host region of investment, wage, or corporate tax, vary according to the service function considered: they are not always significant and the correlation may be positive or negative. Apart from the language variable for which the coefficient is always positive and very significant, the only variables that are significant for all service functions in all sectors relate to agglomeration variables. After controlling for other variables and introducing a mid-term horizon for our regressions, agglomeration variables remain significant.

Thirdly, considering different types of agglomeration, that is to say distinguishing between regional, sectorial, functional, and group agglomeration, we bring to light that it is not always the same type of agglomeration variables that act according to service functions and to sectors. Moreover, both functional and sectorial agglomeration variables appear as important determinants of FDI location choices of service functions. Such an observation was already made by Py and Hatem (2009), but this challenges other research works (Defever and Mucchielli 2005; Defever 2006) that consider that manufacturing activities are attracted by other manufacturing activities in the same sector whereas the location of service functions (other than logistics) is rather explained by the search of functional externalities.

Using a different methodology from the other studies done in this area, our study leads to new questions. By highlighting the heterogeneity of the foreign investment location determinants in service functions, according to both sectors and service functions, our results show the difficulty to conduct a policy in this field. As a matter of fact, it should be difficult for a government to have a specific strategy and to find the right policy decisions that would attract investments in its country or in a given region. In particular, is it better to favor a sector or to favor a service function or even both to attract FDI in service functions? Is it better not to intervene or to support already existing local specificities relying upon the smart specialization through a bottom-up process?

The answer to these questions is likely to be linked to the specificities of a region relative to another one. It is why in a forthcoming study, we intend to go further into the analysis by identifying precisely the regions in which foreign investors choose to invest relative to others and which kind of service activities are located in each of them.

Footnotes

Appendix

Table A3.

The Determinants of FDI Location Choices in Testing and Servicing by Sectors—Mixed Logit Estimation Results.

Sectors	Industry	Business services
Explanatory variables	Coefficients	Coefficients
Dist.	−.123 (.116)	.042 (.169)
Lang.	.638*** (.164)	.993*** (.261)
MP	.116 (.173)	−.049 (.295)
GDP/cap	.200 (.302)	.730* (.434)
Wage	.555 (.557)	1.046 (.939)
Unemp.	−.308* (.170)	.466* (.280)
Dens.	3.068 (1.391)	2.567 (2.187)
Dens. (squared)	−.291 (.129)	−.244 (.198)
Funds	−.015 (.229)	−.059 (.344)
Law	−.137 (.308)	.169 (.561)
Open	.429 (.296)	−.806 (.974)
Tax	.096 (.487)	−.736 (.980)
Skill	.398 (.205)	.616** (.304)
Motor	−17.79 (7.504)	−3.631 (12.34)
Rail	3.103 (3.468)	2.270 (7.407)
Air	.035 (.150)	−.191 (.248)
Internet	.542 (.317)	.260 (.656)
Agglo.GL	.801*** (.088)	1.075*** (.162)
Agglo.S1	.163* (.098)	−.169 (.133)
Agglo.F1	.211 (.116)	.178 (.172)
Agglo.G1	3.460*** (.451)	1.927* (1.009)

Error component (SD)
MP	.105 (.718)	.056 (1.157)
GDP/cap	−.042 (1.137)	.220 (1.473)
Wage	1.875*** (.536)	1.180 (1.022)
Unemp.	−.866** (.311)	−1.122** (.445)
Dens.	−.844* (.349)	.503 (.376)
Dens. (squared)	−.006 (.024)	−.058 (.044)
Law	.039 (.587)	−.385 (.631)
Open	1.540** (.483)	−2.032 (.880)
Tax	.976 (.664)	1.345 (1.233)
Skill	−.053 (.320)	.195 (.466)
Motor	−5.914 (13.86)	3.089 (6.545)
Rail	3.527 (7.559)	4.554 (10.65)
Air	.009 (.181)	−.210 (.217)
Internet	−.800* (.443)	1.388** (.706)

Likelihood	−2,206.8	−1,080.5

Note: Standards errors in parentheses. MP = market potential; GDP = gross domestic product.

*Significant at 10 percent.

**Significant at 5 percent.

***Significant at 1 percent.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Université de Franche-Comté (France) and by UTBM (France).

Notes

References

Ambos

2005. “Foreign Direct Investment in Industrial Research and Development: A Study of German MNCs.” Research Policy 34:395–410.

Basile

2004. “Acquisition versus Greenfield Investment: The Location of Foreign Manufacturers in Italy.” Regional Science anf Urban Economics 34:3–25.

Basile

Castellani

Zanfei

. 2008. “Location Choices of Multinational Firms in Europe: The Role of EU Cohesion Policy.” Journal of International Economics 74:328–40.

Bénassy-Quéré

Gobalraja

Trannoy

. 2007. “Tax and Public Input Competition.” Economic Policy 22:386–430.

Bunyaratavej

Hahn

Doh

. 2007. “International Offshoring of Services: A Parity Study.” Journal of International Management 13:7–21.

Casi

Resmini

. 2010. “Evidence on the Determinants of Foreign Direct Investment: The Case of EU Regions.” Eastern Journal of European Studies 1:93–118.

Castellani

Meliciani

Mirra

. 2014. “The Determinants of Inward Foreign Direct Investment in Business Services across European Regions.” Regional Studies. doi:10.1080/00343404.2014.928677.

Cieslik

2013. “Determinants of the Location of Foreign Firms in Polish Regions: Does Firm Size Matter?” Tijdschrift voor Economische en Sociale Geografie 104:175–93.

Crozet

Mayer

Mucchielli

J.-L.

. 2004. “How Do Firms Agglomerate? A Study of FDI in France.” Regional Science and Urban Economics 34:27–54.

10.

Crozet

Milet

. 2014. “The Servitization of French Manufacturing Firms.” CEPII Working Paper, n°2014-10, Paris, France.

11.

Davis

J. C.

Henderson

J. V.

. 2008. “The Agglomeration of Headquarters.” Regional Science and Urban Economics 38:445–60.

12.

Defever

2006. “Functional Fragmentation and the Location of Multinational Firms in the Enlarged Europe.” Regional Science and Urban Economics 36:658–77.

13.

Defever

2012. “The Spatial Organization of Multinational Firms.” Canadian Journal of Economics 45:672–97.

14.

Defever

Mucchielli

J.-L.

. 2005. “Décomposition internationale de la chaîne de valeur—Une étude de la localisation des firmes multinationales dans l’Union européenne élargie.” Revue Economique 56:1185–205.

15.

Dembour

2008. “Competition for Business Location: A Survey.” Journal of Industry, Competition and Trade 8:89–111.

16.

Demmou

2010. “Le recul de l’emploi industriel en France entre 1980 et 2007. Ampleur et déterminants: un état des lieux.” Economie et Statistique 438–440:273–96.

17.

Devereux

M. P.

Griffith

. 1998. “Taxes and the Location of Production: Evidence from a Panel of US Multinationals.” Journal of Public Economics 68:335–67.

18.

Devereux

M. P.

Griffith

Simpson

. 2007. “Firm Location Decisions, Regional Grants and Agglomeration Externalities.” Journal of Public Economics 91:413–35.

19.

Disdier

A.-C.

Mayer

. 2004. “How Different Is Eastern Europe? Structure and Determinants of Location Choices by French Firms in Eastern and Western Europe.” Journal of Comparative Economics 32:280–96.

20.

Dixit

A. K.

Grossman

G. M.

. 1982. “Trade and Protection with Multistage Production.” Review of Economic Studies 49:583–94.

21.

Doh

J. P.

Bunyaratavej

Hahn

E. D.

. 2009. “Separable But Not Equal: The Location Determinants of Discrete Services Offshoring Activities.” Journal of International Business Studies 40:926–43.

22.

Duranton

Puga

. 2005. “From Sectoral to Functional Urban Specialization.” Journal of Urban Economics 57:343–70.

23.

European Commission. 2013. International Trade and Foreign Direct Investment. Luxembourg: Eurostat Pocketbooks, Publications Office of the European Union.

24.

Eurostat. 2013. “Europe in Figures—Eurostat Yearbook 2013.” Accessed June, 2015. http://epp.eurostat.ec.europa.eu/statistics_explained/index.php/Eurostat_yearbook

25.

François

Hoekman

. 2010. “Services Trade and Policy.” Journal of Economic Literature 48:642–92.

26.

François

Pindyuk

. 2013. “Consolidated Data on International in Services v8.7.” IIDE Discussion Paper, 201301-01, IIDE: Institute for International and Development Economics, Vienna, Austria.

27.

François

Pindyuk

Woerz

. 2009. “Trends in International Trade and FDI in Services.” IIDE Discussion Paper, 200908-02, Institute for International and Development Economics, Vienna, Austria.

28.

François

Woerz

. 2008. “Producer Services, Manufacturing Linkages and Trade.” Journal of Industry, Competition and Trade 8:199–229.

29.

Gabe

T. M.

Bell

K. P.

. 2004. “Tradeoffs between Local Taxes and Government Spending as Determinants of Business Location.” Journal of Regional Science 44:21–41.

30.

Gorbunova

Infante

Smirnova

. 2012. “New Evidence on FDI Determinants: An Appraisal over the Transition Period.” Prague Economic Papers 2:129–49.

31.

Groh

A. P.

Wich

. 2012. “Emerging Economies’ Attraction of Foreign Direct Investment.” Emerging Markets Review 13:210–29.

32.

Grossman

G. M.

Rossi-Hansberg

. 2008. “Trading Tasks: A Simple Theory of Offshoring.” American Economic Review 98:1978–97.

33.

Hahn

E. D.

Bunyaratavej

. 2010. “Services Cultural Alignment in Offshoring: The Impact of Cultural Dimensions on Offshoring Locations Choices.” Journal of Operations Management 28:186–93.

34.

Harris

1954. “The Market as a Factor in the Localization of Industry in the United States.” Annals of the Association of American Geographers 44:315–48.

35.

Hatem

, ed. 2004. “Call and Shared Service Centres in the Euro-mediterranean Region.” Anima Papers and Studies 10:48.

36.

Hatem

2005. “Les investissements internationaux en logistique en Europe.” Conference International Trade and logistics, corporate strategies and the global economy, Université du Havre, 28-29 septembre.

37.

Hatem

2007. “Les investissements internationaux dans les centres de R&D en Europe: une analyse à partir des bases de données projets de l’AFII.” Economies et Sociétés 17:699–722.

38.

Head

C. K.

Ries

J. C.

Swenson

D. L.

. 1999. “Attracting Foreign Manufacturing: Investment Promotion and Agglomeration.” Regional Science and Urban Economics 29:197–218.

39.

Head

Mayer

. 2004. “Market Potential and the Location of Japanese Investment in the European Union.” The Review of Economics and Statistics 86:959–72.

40.

Henderson

J. V.

Ono

. 2008. “Where Do Manufacturing Firms Locate Their Headquarters?” Journal of Urban Economics 63:431–50.

41.

Holmes

T. J.

2005. “The Location of Sales Offices and the Attraction of Cities.” Journal of Political Economy 113:551–81.

42.

Kolstad

Villanger

. 2008. “Determinants of Foreign Direct Investment in Services.” European Journal of Political Economy 24:518–33.

43.

Krugman

1991. “Increasing Returns and Economic Geography.” The Journal of Political Economy 99:483–99.

44.

Kumar

1996. “Intellectual Property Protection, Market Orientation and Location of Overseas R&D Activity by Multinational Enterprises.” World Development 24:673–88.

45.

Lassudrie-Duchêne

1982. “Décomposition internationale des processus productifs et autonomie nationale.” In Internationalisation et autonomie de décision, edited by Bourguinat

, 45–56. Paris, France: Economica.

46.

Lejour

A. M.

Smith

P. M.

. 2008. “International Trade in services—Editorial Introduction.” Journal of Industry, Competition and Trade 8:169–80.

47.

Levitt

1972. “Product Line Approach to Service.” Harvard Business Review, September-October, 41–52.

48.

Markusen

J. R.

2005. “Modelling the Offshoring of White-collar Services: From Comparative Advantage to the New Theories of Trade and FDI, the Brookings Forum (Offshoring White-collar Work: The Issues and Implications).” May 12–13, NBER Working Paper, 11827, Cambridge, MA.

49.

Marshall

1890. Principles of Economics. London, UK: Macmillan.

50.

Mayer

Mejean

Nefussi

. 2010. “The Location of Domestic and Foreign Production Affiliates by French Multinational Firms.” Journal of Urban Economics 68:115–28.

51.

McFadden

1974. “Conditional Logit Analysis of Qualitative Choice Behaviour.” In Frontiers in Econometrics, Chap . 4, edited by Zarembka

, 105–42. New York: Academic Press.

52.

Montout

Robin

. 2012. “La localisation des investissements des pays émergents en Europe.” Revue Française d’Economie 27:71–104.

53.

Moosa

I. A.

Cardak

B. A.

. 2006. “The Derminants of Foreign Direct Investment: An Extreme Bounds Analysis.” Journal of Multinational Financial Management 16:199–211.

54.

Nefussi

Schwellnus

. 2010. “Does FDI in Manufacturing Cause FDI in Business Services? Evidence from French Firm-level Data.” Canadian Journal of Economics 43:723–50.

55.

Pusterla

Resmini

. 2007. “Where Do Foreign Firms Locate in Transition Countries? An Empirical Investigation.” The Annals of Regional Science 41:835–56.

56.

Hatem

. 2009. “Internationalisation et localisation des services, une analyse sectorielle et fonctionnelle appliquée aux firmes multinationales en Europe.” Economie et Statistique 426:67–95.

57.

Rasciute

Pentecost

E. J.

. 2010. “A Nested Logit Approach to Modelling the Location of Foreign Direct Investment in the Central and Eastern European Countries.” Economic Modelling 27:32–39.

58.

Sachwald

Chassagneux

. 2007. “Les facteurs de localisation des centres de R&D à l’étranger: le cas de l’Europe.” Economies et Sociétés 17:723–50.

59.

Sanyal

K. K.

Jones

R. W.

. 1982. “The Theory of Trade in the Middle Products.” The American Economic Review 72:16–31.

60.

Spies

2010. “Network and Border Effects: Where Do Foreign Multinationals Locate in Germany?” Regional Science and Urban Economics 40:20–32.

61.

Strauss-Kahn

Vives

. 2009. “Why and Where Do Headquarters Move?” Regional Science and Urban Economics 39:168–86.

62.

Tomlin

K. M.

2008. “Japanese FDI into U.S. Service Industries: Exchange Rate Changes and Services Tradability.” Japan and the World Economy 20:521–41.

63.

Train

K. E.

2003. Discrete Choice Methods with Simulation. Cambridge, UK: Cambridge University Press.

64.

Tukker

Tischner

. 2006. “Product-services as a Research Field: Past, Present and Future, Reflections from a Decade of Research.” Journal of Cleaner Production 14:1552–56.

65.

UNCTAD (United Nations Conference on Trade and Development). 2004. World Investment Report—The Shift Towards Services. New York: United Nations.

66.

UNCTAD (United Nations Conference on Trade and Development). 2007. World Investment Report—Transnational Corporations, Extractive Industries and Development. New York: United Nations.

67.

UNCTAD (United Nations Conference on Trade and Development). 2008. World Investment Report—Transnational Corporations and the Infrastructure Challenge. New York: United Nations.

68.

UNCTAD (United Nations Conference on Trade and Development). 2009. World Investment Report—Transnational Corporations, Agricultural Production and Development. New York: United Nations.

69.

UNCTAD (United Nations Conference on Trade and Development). 2010. World Investment Report—Investing in a Low-carbon Economy. New York: United Nations.

70.

UNCTAD (United Nations Conference on Trade and Development). 2012. World Investment Report—Towards a New Generation of Investment Policies. New York: United Nations.

71.

UNCTAD (United Nations Conference on Trade and Development). 2013. World Investment Report—Global Value Chains: Investment and Trade for Development. New York: United Nations.

72.

Yin

. 2014. “Location Determinants of Foreign Direct Investment in Services: Evidence from Chinese Provincial-level Data.” LSE Asia Research Center Working Papers 64. London, UK.

73.

Zhu

Eden

Miller

S. R.

Thomas

D. E.

Fields

. 2012. “Host-country Location Decisions and Early Movers and Latecomers: The Role of Local Density and Experiential Learning.” International Business Review 21:145–55.