Determinants of Foreign Direct Investment to the Netherlands:
An analysis of home country characteristics
Master Thesis International Economics and Business (IE&B) Groningen, March 2009
Author: Supervisor:
Jesper de Wit prof. dr. J.H. Garretsen
Student number: 1350676 Faculty of Economics and Business jesperdewit@gmail.com j.h.garretsen@rug.nl
Co-assessor:
prof. dr. S. Brakman
Abstract
Given the large size and rapid growth of foreign direct investment in the Netherlands, this study analyses the home country determinants of this phenomenon. Taking the gravity model as a benchmark, we select and test six home country determinants from the literature. This includes a new direction in FDI theory, namely the influence of third countries, measured as the home market proximity effect, for which we include a distance decay parameter. Our sample covers the period 1987-2006, includes 19 OECD home countries and over 85% of Dutch inward FDI during these years. To control for unobserved country-specific differences we employ a fixed effects panel methodology. In addition, we create several subsamples and perform robustness checks regarding the period, geographic outliers and the sector. Our results show that home country market size (positive) and home country productivity (negative) are the most important determinants. Besides, the home country corporate income tax rate (negative) and home market proximity (positive) are significant in most samples.
Table of contents
LIST OF ABBREVIATIONS AND ACRONYMS 4
1. INTRODUCTION 6
2. THE NETHERLANDS AND INWARD FDI 7
2.1 The Netherlands as a receiver of FDI: a comparison 8
2.2 Geographic trends 9
2.3 Sectoral trends 10
2.4 Conclusion 11
3. THEORETICAL BACKGROUNDS AND HYPOTHESES 11
3.1 Home country market size 13
3.2 Home country trade costs 13
3.3 Bilateral trade 14
3.4 Home country productivity 15
3.5 Home country corporate tax rate 16
3.6 Home market proximity 17
3.7 Hypotheses 17
4. DATA DESCRIPTION AND EMPIRICAL METHODOLOGY 18
4.1 Data description 18
4.1.1 Explanatory variables 21
4.2 Empirical methodology 23
4.2.1 Empirical model 25
5. ESTIMATION RESULTS 27
5.1 Full sample results 27
5.2 Period and European subsample results 29
5.3 Sectoral results 33
5.4 Discussion 35
6. FINAL REMARKS 37
6.1 Conclusion 37
6.2 Limitations and future research 38
REFERENCES 39
APPENDIX A: Structure of the Dutch economy: main characteristics 42 APPENDIX B: The investment position of the Netherlands 44
APPENDIX C: Empirical evidence 46
List of abbreviations and acronyms
AUS Australia AUT Austria
BEL Belgium
CAN Canada
CBS Centraal Bureau voor de Statistiek
CEPII Centre d’Etudes Prospectives et d’Informations Internationales DNB De Nederlandsche Bank (Central Bank of the Netherlands)
DNK Denmark
DTAX Taxdummy
EU European Union
EXPO Export Performance FDI Foreign Direct Investment
FIN Finland
FRA France
GER Germany
GDP Gross Domestic Product
GGDC Groningen Growth and Development Centre HMPH Home Market Proximity High
HMPL Home Market Proximity Low HMPS Home Market Proximity Standard IMF International Monetary Fund
IRL Ireland
ITA Italy
JPN Japan
LUX Luxembourg
NOR Norway
OECD Organisation for Economic Co-Operation and Development OLS Ordinary Least Squares
SPA Spain SWI Switzerland
SWE Sweden
TC Trade Cost
UNCTAD United Nations Conference on Trade and Development
UK United Kingdom
US United States of America
1. Introduction
During the last two decades, the inflow of FDI has been growing rapidly in the Netherlands, signifying the importance of foreign firms for the Dutch economy (DNB, 2008; Hogenbirk, 2002). This coincides with a global upsurge in FDI by multinational firms, whereby growth of world FDI inflow outpaced growth of world income and trade (UNCTAD, 2008; Barba Navaretti and Venables, 2004). The enormous rise of international production by multinational firms has been enhanced by the process of economic globalisation, which we define as the integration of global economies into the world economy. EU-membership and the adoption of the euro shows how the Netherlands actively deals with globalisation. Besides, the global increasing irrelevance of national borders due to the liberalization of regulations and other economic barriers means that product and service markets become contestable for both domestic and foreign firms. It is generally believed that the impact of multinational firms contributes to the economic well-being of both home and host economies, as it boosts output, employment and productivity (Barba Navaretti and Venables, 2004; Dunning and Lundan, 2008).
The Netherlands increasingly benefits from these developments, since the Dutch share in world inward FDI stock increased from 3.5% in 1990 to 4.4% in 2007 (UNCTAD, 2008). Around 5,000 foreign firms have operations in the Netherlands, employing more than half a million people (Ministry of Economic Affairs 2006; UNCTAD, 2008). Moreover, in 2007 the Netherlands was the sixth largest receiver of FDI. These are remarkable achievements for a small economy with less than 17 million people.
Two streams of FDI theory may explain this phenomenon: the industrial organization/microeconomic and macroeconomic theory (Grosse and Trevino, 1996; Dunning and Lundan, 2008). The former analyses the behaviour of individual firms. The latter, which is the focus of this study, stresses why investment between pairs or groups of countries flows in certain patterns. We define FDI as an investment by home country multinational firms in host country firms with the aim of obtaining a degree of management control. This involves either creating new establishments or by means of mergers and acquisitions. FDI flow also includes all other financial transactions between affiliated firms (e.g. loans, reinvested earnings and equity capital), as well as purchases and sales of real estate. FDI stock is an accumulation of annual FDI flows by a country. In contrast, this does not include portfolio flows, which are mainly short term and financial.
2002; Blonigen et al., 2005; Kimino et al., 2007). Given the rapid growth of FDI in the Netherlands and its overall size, it is important to understand the home country determinants of Dutch inward FDI as well. Therefore, our purpose is to answer the following research question:
“What are the home country determinants of inward FDI to the Netherlands?”
An important feature of this study is that the we solely focus on home country determinants. This approach is chosen since the host country determinants of the Netherlands are equal for all home countries. Therefore, we automatically control for host country effects. Taking the gravity model as a benchmark we select and test six home country determinants from the literature. Our sample covers the period 1987-2006, includes 19 OECD home countries and over 85% of Dutch inward FDI during these years. To control for unobserved country-specific differences we employ a fixed effects panel methodology.
This study is organised as follows. The next section analyses the inward FDI position of the Netherlands. Section 3 reviews the empirical literature on the home country determinants of FDI and formulates six hypotheses that are subsequently tested in our empirical model. Section 4 describes the data employed and specifies the empirical methodology. The forthcoming estimation results are analysed and discussed in section 5. In section 6 we give final remarks. Finally, the four Appendices provide additional information and data.
2. The Netherlands and inward FDI
The Netherlands is one of the largest receivers of FDI in the world, both in absolute and in relative terms. In absolute terms the Netherlands was in 2007 the sixth largest receiver of FDI with a stock of € 491,935 million (UNCTAD, 2008). In relative terms, the ratio of inward FDI stock as a percentage of GDP was 87.9%, this ratio was never so high for the Netherlands. These are remarkable achievements for a small country with a population of only 16.6 million people.
Hogenbirk and Narula (2001) summarize the most important locational factors influencing the choice of multinational firms for the Netherlands from several studies on Dutch inward FDI. These are favourable location in Europe, good infrastructure (harbour of Rotterdam and airport of Amsterdam), favourable fiscal conditions and the highly educated labour force.
The EU makes free flow of goods and services among member states a reality. Multinational firms do not need to have activities in each individual EU state anymore but can choose to serve all EU markets from one location (Tulder, 1999). Because of its favourable location the Netherlands can act as a gateway to Europe as within a radius of 500 kilometre from Amsterdam, a firm is able to reach almost 50% of the European population (Hogenbirk, 2002). Thus, EU integration resulted in an upgrading of the locational advantages of the Netherlands, making it more attractive for subsidiaries of multinational firms.
Appendix A describes and shows the main characteristics of the Dutch economy which contributed to the inward FDI pattern. The Netherlands proves to be a small, developed economy with high dependence on international trade, in which re-exports form an important share. The rest of this section illustrates the rather unique position of the Netherlands as a host country of FDI. First, the general investment position of the Netherlands is discussed in comparison with other major receivers of FDI. Subsequently, the relative importance of the largest home countries over time is analysed. Third, sectoral shifts in FDI that have occurred over time are examined. This section ends with a conclusion.
2.1 The Netherlands as a receiver of FDI: a comparison
This section compares the general investment position of the Netherlands and seven other major, comparable receivers of FDI. Table B1 (Appendix B) shows indicators from which important observations can be derived. First, the enormous growth of inward FDI stock from 1990 to 2007 of all countries signifies the rise of activities by multinational firms. In fact, global inward FDI stock went up from US$ 1,941 billion in 1990 to $ 15,210 billion in 2007 (UNCTAD, 2008). Concerning flows, we observe the same trend, global FDI inflows in 2007 even past the 2000 record, reaching a historical record of $ 1,833 billion. Since FDI flows are more volatile than FDI stocks, also for the Netherlands, the focus below is on FDI stocks (See also section 4.1, figure 2 and 3, for a comparison).
(including Hong Kong), the UK, France and Belgium the sixth largest receiver of FDI (UNCTAD 2008). Both the Netherlands and Belgium received more FDI than their much larger neighbour Germany.1
Taking into account the size of the countries the position of the Netherlands is even more striking. This is against common sense since it is expected that the limited market size and lack of resources in small economies makes them relatively unattractive as a location for FDI (Dunning and Lundan, 2008). In addition, also with a relative indicator small economies are popular hosts since together with Belgium, Ireland and Switzerland, the Netherlands has the highest ratio of inward FDI stock to GDP in the EU (UNCTAD, 2008).
2.2 Geographic trends
In 2007 inward FDI stock is dominated by foreign investments from the EU (66.3%) and from the US (18.3%), as shown in Table B2 (Appendix B). During the last two decades inward FDI stock increased with 12.6% per year. EU and eurozone investments increased above this average at cost of investments from the US, which increased below the average. The importance of the two other non-EU countries in the top-10, Switzerland and Japan, also declined. Therefore, the relative importance of home countries shifted towards EU countries, especially countries which adopted the euro.
The US, the UK, Belgium, Luxembourg, Germany and France are by far the largest investors with a FDI stock of at least € 40 billion in 2007. Except for the US, all these countries are located close to the Netherlands. The US is still the largest investor due to the size of its economy and the dominant position it kept in the past. Considering the size of Luxembourg, with only half a million people, its share in Dutch inward FDI stock is impressive. This is partly explained by the merger between Mittal Steel and Arcelor in 2006, the first was listed in Luxembourg and the latter in the Netherlands. So, large multinationals can play a large role in inward FDI stock.
1
2.3 Sectoral Trends
Figure 1 separates inward FDI stock between industry and services, which stand for the secondary and the tertiary sector, respectively.2 Both sectors have grown substantially during the last two decades. Services has been the dominant sector since 1988. The last decade the percentages of both sectors more or less stabilized. Inward FDI stock consists of around 60% services and 40% industry.
Figure 1: Sectoral division between industry and services of inward FDI stock, 1984-2007 (millions of euros and percentage share of total)
0 50,000 100,000 150,000 200,000 250,000 300,000 350,000 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 Industry Services % Industry % Services
Sources: DNB, own calculations 3
The dominance of services FDI makes sense when considering the large sectoral contribution of the tertiary sector in GDP and employment, as provided in Appendix A. More important, a major distinction between services and industrial products is that the latter can in principle always be sold trough exports. This is not possible in case of services (Stibora and Vaal, 1999). Services are intangible, non-storable and production and consumption occur simultaneously. Therefore, output of services occurs in a physical presence in the local market. For multinational firms this is basically not possible by trade, but only by FDI.
2 The primary sector is ignored since its size is neglectable. 3
The most important industrial subsectors with high foreign presence are Minerals, oils and chemicals, Metal and electronics and Food, beverages and tobacco (DNB, 2008). Multinational firms like BASF (Germany), IBM (US) and Nestlé (Switzerland) are examples of firms which invested in these three subsectors. In case of services a distinction can be made between Wholesale and retail trade, Transportation and storage and Banking and insurance. Daimler-Benz (Germany), Ericsson (Sweden) and BNP Paribas (France) are examples of multinational firms which invested in these three subsectors. Most foreign companies enter the Dutch market with merger or acquisitions instead of greenfields (Hogenbirk 2002).
2.4 Conclusion
Although small in terms of size and population the Netherlands is a very attractive location for FDI. This is explained by specifically Dutch factors like geographic location, openness and EU-membership, but also by high infrastructural and educational standards and a favourable fiscal climate. Due to this factors the Netherlands increasingly attracts FDI which fosters development of the economy.
The last two decades the relative importance of home countries has shifted towards European countries, especially countries which adopted the euro. However, the US and Japan still belong to the largest investors in the Netherlands. Besides market-seeking, they use the Netherlands as an export platform to serve the European market. Finally, the service sector has been by far the largest receiver of FDI during the last decade.
The next issue, and central in this study, involves the home country characteristics of Dutch inward FDI. To deal with this issue section 3 describes the home country determinants of inward FDI from the literature. Afterwards, these determinants are tested for FDI inflow to the Netherlands.
3. Theoretical backgrounds and hypotheses
analysis and our focus is on the (home) country as the unit of analysis, therefore we take the country as the level of analysis.
Barba Navaretti and Venables (2004) give an overview of the home and host country determinants of FDI from the literature. The gravity model is taken as the starting point since it approximates quite well the cross-country pattern of FDI. Moreover, Ekholm (1998) and Shatz (2003) find that approximately 60% of the cross-country variation in FDI can be explained by the gravity framework. Basically, the gravity model argues that (bilateral) FDI is determined by the GDP of each country, the distance between them and possibly also other common factors like regional trade agreements. Though the gravity model provides a useful basis, it is important to move beyond, since this relationship is also valid for many other spatial economic interactions, for example trade flows.
When analysing FDI it is useful to distinguish between horizontal and vertical FDI (Barba Navaretti and Venables, 2004). The first one involves market-seeking FDI implicating the duplication of part of a firm’s activities in a host country with the aim of having better and cheaper market access to that host country. The latter involves efficiency-seeking FDI since part of a firm’s production is carried out in a host country to profit from low-cost inputs. The largest share of FDI among developed economies involves by far horizontal FDI. As shown in Table A1 and Table B2 and described in section 2, the inward FDI stock in the Netherlands is almost only originating from developed countries, with similar welfare standards. Therefore, inward FDI in the Netherlands is assumed to be horizontal. According to theory, and further explained below, horizontal FDI is positively influenced by market size, trade costs, firm-level economies of scale and similarity of factor endowments.
The selection of home country determinants is based on previous studies that have used a gravity model and/or a multidimensional model to determine home country characteristics of FDI in a high-income OECD country (Grosse and Trevino, 1996; Hogenbirk, 2002; Blonigen et al., 2005; Kimino et al., 2007) and on theoretical models and empirical evidence of FDI in general (Barba Navaretti and Venables, 2004; Blonigen, 2005).4 Appendix C (Table C1) gives an overview with additional information about the four most important studies used to derive home country determinants for this study.
Finally, the chosen determinants are: home country market size, home country trade costs, bilateral trade, home country productivity, home country corporate tax rate and home market proximity. The exact influence of these location-specific factors has generated conflicting results
4 According to the World Bank (2008) a country is a high-income OECD when 2007 gross national income (GNI) per
in previous research, due to various theoretical models, a lack of internationally comparable data and/or the econometric methodology employed. Based on an integration of the theoretical backgrounds described below and the inward investment position of the Netherlands analysed in section 2, we state a hypothesis at the end of each section. Section 3.7 provides an overview of the six formulated hypotheses. Section 4.1 describes how we measure these hypotheses.
3.1 Home country market size
Larger countries in terms of GDP are expected to sustain more large firms that are positioned to expand internationally and should hence make larger investments in the international capital market. Larger economies suggest greater availability of capital resources and intangible assets such as technical knowledge and marketing expertise that can be used to start foreign operations. In addition, Table B2 gives evidence for this relationship for the Netherlands, as it shows the dominant position of the US, the largest economy of the world.
Nevertheless, Kimino et al. (2007) emphasize that past evidence regarding home country market size is mixed due to the econometric methodology employed. This mainly involved OLS on pooled data. The shortcoming of this method is that it neglects country-specific effects, leading to unreliable results since country-specific differences are contributing to variation in FDI stock. Using country fixed effects Kimono et al. (2007) find that home country GDP does not influence FDI to Japan. In contrast, using pooled OLS Grosse and Trevino (1996) and Hogenbirk (2002) conclude that home country GDP is an important indicator for FDI to the US and to the Netherlands, respectively. Finally, both Hogenbirk (2002) for the Netherlands and Blonigen et al. (2005) for the US use population as a proxy for home country market size and find that it negatively effects FDI. Apparently, a large customer base in the home country makes it less necessary to expand to another market to achieve economies of scale.
Even though the evidence is mixed we expect a positive relationship between home country market size and FDI to the Netherlands based on the theoretical arguments described in the first paragraph of this section.
3.2 Home country trade costs
coordination costs, and could also be a barrier to FDI. Hogenbirk (2002) finds that cultural distance between the home country and the Netherlands negatively influences FDI. The geographic distance is not significant. Whereas Grosse and Trevino (1996) conclude that both geographic and cultural distance are negatively related with inward FDI to the US. In addition, Blonigen et al. (2005) show that closed countries, measured as the inverse of trade openness, are less likely to start operations in the US. Kimino et al. (2007) argue that geographic and cultural distance are at best only borderline significant for inward FDI to Japan.
The role of trade costs is easier to analyse when making a distinction between horizontal and vertical FDI. Theory predicts that trade costs can effect FDI either positively or negatively (Barba Navaretti and Venables, 2004). Assuming that a firm already exported to a host country the effect of trade costs depends on the kind of investment considered. Horizontal FDI is positively influenced by rising trade costs. In this case the goal of the multinational firm is to gain market access in a host country, and with rising trade costs starting a subsidiary is becoming cheaper relative to exporting. In contrast, vertical FDI is negatively influenced since rising trade costs increase the cost of trading components produced in the host country. Because Dutch inward FDI is assumed to be horizontal we expect a positive relationship between trade costs and FDI.
3.3 Bilateral trade
Both trade and FDI indicate market penetration of the home country in the host country. Whether trade and FDI are substitutes or complements is a topic of severe debate in the literature (Blonigen, 2005). The trade effects of FDI are related to the motivations underlying FDI behaviour.
the sense that they can be used (in a non-rival way) firm wide, across multiple plants, causing firm-level increasing returns to scale. Thus, it is more attractive for a firm to locate production in a variety of markets rather than have one production plant that exports to several markets. Firms or sectors with firm-level economies of scale dominating plant-level economies of scale are more likely to serve foreign markets through subsidiaries than through exports (Ekholm, 1998).
Against this are two main theories of complementarity between trade and FDI, namely demand complementarities and vertical relationships (Blonigen, 2001). The first theory suggests that a firm’s presence in a foreign market, by exports, increases total demand for its products trough several channels, for example, the provision of sales and after-sales services or more efficient and quicker deliveries and distribution, through which inward FDI complements home country exports. Grosse and Trevino (1996) and Hogenbirk (2002) show that this complementary relationship exists for both the US and the Netherlands, respectively. Secondly, vertical relationships cause complementarity in the sense that a multinational firm conducts part of the production in a foreign market which increases host country imports of intermediate products, which actually involve exports within the multinational firm. In addition, this may also increase host country exports, either back to the home country or to a third country. In case of the latter, the host country functions as an export-platform, as we described for the Netherlands in section 2.
Blonigen (2001) examines disaggregated product-level data for Japanese automobile parts in the US market and concludes that substitution and complementary effects can occur simultaneously, depending on the product-level. A substitution effect occurs in case of increased automobile parts production by Japanese firms in the US and a complementarity effect occurs from increased automobile production by Japanese firms in the US. The latter is caused by the fact Japanese producers still import automobile parts from Japan. Although both theory and empirical evidence are mixed we expect trade and FDI to be complementary for the Netherlands. This is based on the results of demand complementarity effects that Hogenbirk (2002) finds for the Netherlands and since the Netherlands functions as a distribution centre for Europe, which involves vertical relationships.
3.4 Home country productivity
higher productivity, particularly in service sectors, in which the quality of labour is important. Hogenbirk (2002) finds that Dutch FDI inflow is positively influenced by a higher wage rate in the Netherlands relative to the home country and argues that this is an indicator of the high productivity and high labour quality in the Netherlands. This is also shown in Table A1, which shows that productivity in the Netherlands is higher than in most other countries. So, ultimately productivity matters more than wage.
As shown before, the inward FDI stock in the Netherlands is almost only originating from developed countries, with similar welfare and productivity standards as the Netherlands. The theory of horizontal FDI argues that FDI is promoted by similarity of factor endowments, which we call productivity, between home and host countries (Markusen and Maskus, 2002; Barba Navaretti and Venables, 2004). Looking at the origin of inward FDI, this is obviously the case for the Netherlands. Furthermore, although the difference in productivity levels between the home countries is relatively small, there remains a difference. Firms from home countries with a relatively high productivity level are expected to keep a relatively larger share of their production at home, i.e. they prefer to export, to profit from this productivity. On the other hand, firms from countries with a relatively low productivity level will prefer to conduct a relatively larger share of their production abroad. Acknowledging that FDI takes place in a horizontal context, we expect a negative relationship between home country productivity and FDI.
3.5 Home country corporate tax rate
3.6 Home market proximity
A new direction in the determinants of FDI is the inclusion of the influence of third countries. Several papers find evidence of the influence of additional, potential host countries on the outward FDI decision, also called host market potential or export-platform FDI (Blonigen et al., 2007; Garretsen and Peeters, 2008). Meaning that a home country invests in a particular host country with the intention of serving third markets with exports of final goods from the subsidiary in the chosen host country. The host market potential variable is measured using FDI from a single home country to several host countries. So, every host country has its own market potential, dependent on its location. However, we analyse home country determinants, taking FDI to a single host country from several home countries. Therefore, every home country has its own market potential, which depends on home country location with respect to third countries. Consequently, following Blonigen et al. (2005), we recognize that these third country effects may be an important home country determinant as well.
Blonigen et al. (2005) argue that the proximity of the home country to surrounding, third markets alters the margin of whether to service the host market with exports from the home country or through FDI. When the home country has low-cost access to these surrounding, large markets, this increases the opportunity cost of exporting to the host country. Since the home country is used to export relatively cheap to its surrounding, large countries, export to the host country is relatively more expensive, leading to higher subsidiary production in the host country. For example, Ireland (home country) is surrounded by the UK (large third country), which increases the opportunity cost of exporting to the Netherlands (host country), resulting in more FDI inflow from Ireland to the Netherlands. Blonigen et al. (2005) call this the home (parent) market proximity effect. They find that the US receives more FDI from OECD home countries that are surrounded by large, third countries. We expect this to be the same for the Netherlands. Next we formulate the six hypotheses.
3.7 Hypotheses
literature is consistent about the possible home country determinants. Based on an integration of the theoretical backgrounds and the investment position of the Netherlands analysed in section 2, we formulate six hypotheses:
H1: There is a positive relationship between market size of home countries and FDI inflows to the Netherlands
H2: There is a positive relationship between home country trade costs and FDI inflows to the Netherlands
H3: There is a positive relationship between home country exports and FDI inflows to the Netherlands
H4: There is a negative relationship between home country skill and FDI inflows to the Netherlands
H5: There is a positive relationship between home country corporate tax rate and FDI inflows to the Netherlands
H6: There is a positive relationship between home market proximity and FDI inflows to the Netherlands
4. Data description and empirical methodology
This section explains the framework used to test the six hypotheses. Section 4.1 starts with a short introduction about the analysed data and then describes the dependent and explanatory variable(s). Section 4.2 specifies the empirical methodology, including diagnostic checks and the empirical model.
4.1 Data description
Like Blonigen et al. (2005) we take the annual real inward FDI stock (FDI) per home country as the dependent variable to measure FDI inflows to the Netherlands. We calculate this by converting nominal annual FDI stock with a price index of gross fixed capital formation, as indicated in the upper row of Table D2.5 Since our sample consists of 19 home countries and the sample period covers 20 years there are in total 380 observations. Instead of annual inward FDI stock we could also take annual inward FDI flow as a measurement of FDI inflows. Figure 2 (stock) and figure 3 (flow), on the next page, show both measures. We aggregate total FDI from all home countries in both figures. The difference between stock and flow is that the latter develops more erratic over time. When analysing individual home countries the same difference in pattern is observed (DNB, 2008). Inward FDI stock is preferred to annual FDI inflow as a dependent variable since it develops more stable over time. This fits better with economic theory since stock is more reliable and long-term oriented against the more volatile and short-term oriented flow. For instance, hypothesis 1 states that home country market size positively effects FDI inflows to the Netherlands. Since FDI stock develops more stable, like GDP (and the other explanatory variables), this improves the quality of our estimation results as FDI flows may suddenly change from year to year.
Figure 2: Total real inward FDI stock in the Netherlands, 1984-2007 (in millions of euros and as % of real GDP in the Netherlands)
0 100000 200000 300000 400000 500000 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 0.0 20.0 40.0 60.0 80.0 100.0 120.0 in million euro in % GDP
Sources: DNB, CBS, GGDC, own calculations
5
Figure 3: Total real inward FDI inflow in the Netherlands, 1984-2007 (in millions of euros)
0 10000 20000 30000 40000 50000 60000 70000 80000 90000 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 in million euro
Sources: DNB, CBS, own calculations
One econometric disadvantage of taking FDI stock is the danger of spurious regression as both FDI stock and our six explanatory variables either increase or decrease over time. We perform an ADF Fisher unit root test on all variables which all prove to be non-stationary, as the null hypothesis of a unit root is not rejected for any variable.6 Due to unavailability of data it is not possible to increase the number of observations for the home countries, for instance with quarterly data or for a longer sample period. Therefore the reliability of the ADF Fisher unit root is smaller and this diagnostic issue becomes less important. Nevertheless, our estimation results cannot exclude spurious regression. Section 4.2, regarding the econometric methodology, extends further about non-stationarity and the other diagnostic checks.
Figure 2 plots the development of the Dutch inward FDI stock for the period 1984-2007, both as a percentage of GDP and in absolute terms. Both measures show the rise of Dutch inward FDI stock. Figure 2 also shows that there is a rapid increase in inward FDI stock around the middle of the 1990s, starting in 1997 to be precise. To verify whether this is induced by different selected home country determinants we will perform a robustness check. In section 5.2 we divide the regression period into two decades, 1987-1996 and 1997-2006. In addition, using pooled OLS we also perform a Chow breakpoint test to test for a structural break. For the decade 1997-2006 a structural break is significant at 5%. Moreover, for the period 1997-2001 (the period with most rapid growth) a structural break is significant at 1%.7
6
For all variables we test for unit root in level and include individual intercept and trend. ADF Fisher unit root test estimation output is available on request.
7
To examine the effect of geographic outliers, we perform additional robustness checks in section 5.2. First, a sample of only European countries and second a sample of countries who adopted the euro. Furthermore, in the last part of our estimations, analysed in section 5.3, will use data on sectoral FDI, which means that we divide Dutch inward FDI stock into industry and services FDI stock. Regrettably, for reasons of confidentiality by the DNB these data are unavailable for a profounder sectoral division. In addition, sectoral data are only available for 14 of the 19 home countries. These countries are provided in Table D1.
4.1.1 Explanatory variables
We aim to explain the Dutch inward FDI stock with the six home country variables described below. Every variable is based on one hypothesis. Table D2 shows the definitions, abbreviations and sources of these variables.
First, home country market size is measured by real home country GDP in millions of euros. Estimation output shows a high correlation among GDP and population, consequently we only use GDP as a proxy for market size.8 For theoretical reasons we would prefer to include them both due to their opposite effect on inward FDI, as shown in Table C1, for instance by Hogenbirk (2002), who observes no correlation between GDP and population. We choose GDP since theory prefers GDP as an indicator of market size (Barba Navaretti and Venables, 2004). Besides, it better indicates the economic size and welfare of a home country than population.
Second, since the focus is on home country determinants, we use, like Blonigen et al. (2005), the inverse of trade openness as an indicator of trade cost for home countries. Trade openness is defined as imports plus exports divided by GDP. We use this measurement since relatively closed countries (which means a higher inverse of trade openness) are assumed to have relatively higher trade costs since trade is less important for the concerning home country economy. In addition, since we prefer a fixed effects model, as explained in the next section, time invariant variables like geographic or cultural distance cannot be regressed.
Third, following Kimino et al. (2007) export performance (EXPO) of the home country measures whether FDI and trade are substitutes or complements. Export performance is defined as the ratio of home country export to the Netherlands against home country import from the Netherlands. A positive sign indicates complementarity between export and FDI since exports to
8
the Netherlands increase relatively to imports with the Netherlands. A negative sign indicates the opposite, substitutional effect.9
Fourth, home country productivity is called home country skill and is measured as the real labour productivity per hour worked in euros.
Fifth, the home country corporate income tax rate is measured as the statutory corporate
income tax rate. Like Garretsen and Peeters (2008), we use a dummy variable to capture the role
of tax rates in neighbouring home countries. This is useful since a home country is expected to invest more abroad, including the Netherlands, when its tax rate is above the average rate of comparable countries. The tax dummy equals 1 if the statutory corporate income tax rate is larger than the average statutory tax rates in other countries in the sample (including the Netherlands) and 0 otherwise.
Sixth, the home market proximity variable indicates proximity of the home country to other surrounding markets and measures therefore the ease with which firms can export from their home location. Based on Blonigen et al. (2005) this variable is calculated as the distance-weighted sum of the market sizes (GDPs) of proximate countries: HMPi = Σj GDPj/(Dij)φ, where i
is the home country, j is another non-Netherlands country from the sample, Dij is the distance (in
kilometres) between i and j and φ is the distance decay parameter. The distance decay parameter is derived from Garretsen and Peeters (2007), who use it to measure host market potential. Blonigen et al. (2005) do not use this parameter for home market proximity. Therefore, to our knowledge, we are the first one to use it concerning HMP.
A shortcoming of our home market proximity variable is that we limit ourselves to countries in our sample for reasons of consistency and availability of data. For the non-European countries, fortunately only four out of nineteen home countries, it would be interesting to take their surrounding countries into account, for instance the role of China for Japan. One of our robustness checks is taking a subsample of only European countries in section 5.2. Consequently, we will observe the effect on the home market proximity effect
The distance decay parameter φ can take three values: 0.5, 1 or 2, leading to three measures of home market proximity. By varying φ we can change the punishment countries face for being located relatively isolated. So, when φ=0.5 relatively remote countries are punished less for being located isolated. This is called home market proximity high (HMPH). Alternatively, when φ=2 relatively remote countries are punished more for being located isolated, which is
9 Since home country exports and imports are highly correlated with home country GDP we take export performance to
called home market proximity low (HMPL). Home market proximity standard (HMPS) is the regular distance decay parameter, this is the measure used by Blonigen et al. (2005).
Table D3 shows the ranking of the sample countries based on market size and on two market proximity measures (with φ=0.5 and φ=2) for 1987 and 2006. The rankings are in ascending order, meaning that the largest or most proximate countries are ranked first. Several conclusions can be drawn. First, distance, and thus location, matters. Though the US and Japan are ranked highest based on market size they score, as expected by our choice of home countries, very low on market proximity. Therefore, Italy and Spain are better examples. For Belgium and, in particular, Luxembourg this is the other way around. Second, distance decay parameter, φ, is a relevant parameter. Regarding for example the UK and Switzerland, the first becomes more proximate with a more important role for distance and the latter becomes less proximate. In addition, the high ranking of Canada is explained by its bordering with the US. Finally, as also shown by Garretsen and Peeters (2007), although they analyse host countries’ market potential, the ranking of home countries’ market proximity is relatively stable over time.
The summary statistics of the variables are provided in Table D4, giving an indication of the magnitudes of the seven variables we use. For instance, when looking at the minimum and maximum value for home country skill and the standard deviations with respect to the mean, we observe, as expected, that the country differences are relatively small when compared to GDP. Table D5 gives the correlation matrix, showing that correlation is not above 0.7.10 The next section specifies the empirical methodology (panel data and fixed effects) used in estimating the variables, here we also discuss the diagnostic issues of unobserved country heterogeneity, model misspecification, heteroskedasticity, stationarity and autocorrelation.
4.2 Empirical methodology
The sample just involves 20 annual observations for each home country, therefore, the utilization of panel techniques will generate more efficient parameter estimates than separate equations per country. The latter is also not the aim of this study. From a methodological point of view, applying panel data gives several advantages over conventional cross-sections or time-series data. It provides, for instance, more informative data, allows us to study individual dynamics and gives information on the time-ordering of events (Hsiao, 2003; Brüderl, 2005; Baltagi, 2008; Wit, 2009
10
(Methodology Paper)). As a result, the reliability of the estimated regression parameters is better with a panel specification.
When using fixed effects, panel estimation also allows us to control for unobserved home country heterogeneity and model misspecification, which is not possible in case of pooled OLS. As already put forward in section 3.1, pooled OLS may cause inappropriate parameter estimates since it omits unobserved country-specific effects. Blonigen et al. (2005) even admit that this is a major shortcoming of their paper. Garretsen and Peeters (2008), though they focus on host country determinants of FDI, also find that pooled OLS has to be rejected in favour of the fixed effects model. Kimino et al. (2007) compare pooled OLS with both fixed and random effects methodologies and conclude that pooled OLS is the least reliable specification. Pooled OLS does not control for heterogeneity and leads to an unacceptable degree of aggregation bias and to statistically meaningless results. Furthermore, using fixed effects is expected to minimize the problem of heteroskedasticity, to which pooled OLS suffers.11 Finally, by performing the Likelihood ratio test we find that cross-section (country) effects are significant.12
The fixed effects model is favoured over the random effects model for both economic and econometric reasons. Common sense argues that countries are complex economic entities for which it is impossible to aggregate all possible determinants of FDI into a single model. Fixed effects will compensate for this. Moreover, in contrast to the random effects model, the fixed effects model does not require the assumption of no correlation between country-specific errors and the errors of the variables (Brüderl, 2005). In addition, from an economic point of view, there is little justification for treating the individual country effects as uncorrelated with the errors of the variables when almost no macroeconomic variables can be argued to be truly exogenous (Kimino et al., 2007). Furthermore, the random effects model is useful if the individual countries in the sample are randomly chosen and represent a larger population of countries (Hill et al., 2001). This is absolutely not the case for us because, as described in section 4.1, the inward FDI stock of the 19 chosen home countries forms 88,4% of total inward FDI stock over the sample period. Finally, the Hausman test shows that the fixed effects model is favoured over the random effects model.13
Two related diagnostic checks are stationarity and autocorrelation. Considering stationarity, as described in section 3, all our variables turn out to be non-stationary. However, the shortness of our time period does not allow us to do a reliable and valid ADF Fisher unit root test
11
Heteroskedasticity is detected by the White test. Estimation output is available on request. Therefore, when performing a pooled OLS estimation, in section 5.1, we use White heteroskedasticity-consistent standard errors & covariance.
12 Likelihood ratio test estimation output is available on request. 13
to check for stationarity. The number of observations cannot be increased due to unavailability of data. So, we need to accept the danger of spurious regression. In addition, since the series are non-stationary there is autocorrelation. Our estimation results show that the Durbin-Watson statistic is not always near 2.14 Meaning that the errors are correlated over time.
When performing the ADF Fisher unit root test and test for unit root in first difference instead of level all series are stationary.15 So, when using differencing of the variables in equation (1) (next page) non-stationarity is solved and the Durbin-Watson statistic improves. Neither taking dlog of all variables nor taking dlog of only FDI in (1) gives significant results. Nevertheless, whether significant or not, we have to accept these limitations since changing our variable specifications would lead to estimation results which have economically no value. Regarding the first experiment, taking dlog of all variables, a change in GDP, for instance, is according to economic theory unrelated to FDI inflow, measured by either stock or flow, since it is total FDI that matters. With regard to the second experiment, we basically change inward FDI stock into inward FDI flow. However, as explained in section 3, we prefer stock due to its stability and long-term orientation. Concluding, we prefer valid, economical results above completely sound econometrics and accept the danger of spurious regression. Moreover, despite autocorrelation, our results show no high standard errors or low t-statistics.
4.2.1 Empirical model
As in the theoretical framework described in section 3, we use the gravity model as the benchmark model. There are two major changes compared with the basic gravity model. First, we neglect the host country variables. Since the Netherlands is always the host country, these variables only vary over time and not in a cross-sectional sense. To control for the effects of time we use period (annual) fixed effects. Also these turn out to be significant using the Likelihood ratio test.16 Second, by including the home market proximity effect location or geography is allowed to play a role. Table 1 (next page) shows all six variables and the their expected sign in the regression, as argued in section 3.
14
The estimation output for the 1997-2006 subsample (section 5.2) shows that there’s no autocorrelation, as the Durbin-Watson statistic is 1.93. This output is available on request.
15
For all variables we test for unit root in first difference and include individual intercept and trend. ADF Fisher unit root test estimation output is available on request.
16 Moreover, we discuss the period differences of inward FDI stock in section 4.1. The likelihood ratio test estimation
Table 1: Estimation variables and their expected sign
Hypothesis Variable Expected sign
1 Home market size (GDP) +
2 Home trade cost (TC) +
3 Bilateral trade (EXPO) +
4 Home productivity (SKILL) -
5 Home corporate income tax rate (DTAX) +
6 Home market proximity (HMP*) +
So, to define the home country determinants of FDI inflow to the Netherlands we will test the following panel model, with both country and annual fixed effects17:
ln FDIit = ln αit+ β1ln GDPit+ β2ln TCit+ β3 ln EXPOit
+ β4ln SKILLit + β5 DTAXit+ β6 ln HMP*it+ εit (1)
(i = 1,2,3, …, N; t = 1,2,3, …, T)
FDI is the annual real FDI stock from home countries in the Netherlands, and subscripts i and t, respectively, index cross-section units of a specific home country varying from 1 to 19, and time starting from year 1987 till 2006. GDP is the market size of the home countries, TC captures trade costs, and EXPO is the export performance of the home country. The home country productivity level is represented by SKILL and DTAX is a the taxdummy. Finally, home market proximity, HMP*, is captured by either HMPS, HMPH or HMPL. The regressions are executed by using EViews 6.0. The next section shows and analyses the estimation results.
17
5. Estimation results
The first three sections provide and analyse the estimation results of equation (1) for the full sample, the period and European subsample and the sectoral subsample, respectively. Section 5.4 discusses our results with the previous literature.
5.1 Full sample results
Table 2 (next page) gives the estimation results for the full sample (1987-2006, 19 home countries).18 Column (1) shows the results for equation (1) using pooled OLS. As explained above, we prefer the fixed effects model above pooled OLS. 19 ,20 Column (2) shows the results for equation (1), with both country and annual fixed effects but without the home market proximity variable. Column (3) shows the results when including the standard (φ = 1) home market proximity variable. Columns (4) and (5) show the results when we vary distance decay parameter φ. We observe that the inclusion of HMP* does not change the results of the other explanatory variables.
Turning to the first hypothesis, home country GDP is, as expected, an important determinant of inward FDI in the Netherlands. This supports the theory that larger countries are able to sustain more large firms that are positioned to expand internationally and make larger investments in the international capital market and in the Netherlands.
Both hypothesis 2 concerning trade costs and hypothesis 3 concerning bilateral trade are not significant in the full sample. Regarding the first one, most FDI is increasingly originating from European countries (Table B2), and due to the continuing European integration trade costs are declining. Therefore, trade cost only plays a minor role. With respect to export and FDI, there is no evidence for neither a substitutional nor a complementary effect. Consequently they are likely to exist both, depending on the period, home country, sector and/or product-level.
Hypothesis 4 regarding home country skill is confirmed for the full sample. So, in a context of horizontal FDI, there is negative relation between the home country productivity level and inward FDI to the Netherlands. The Dutch economy has a relatively high productivity
18 Estimation outputs of all regressions from EViews 6.0 are available on request. 19
We also added geographic distance in the pooled OLS regression, which is also significant at 1% (and negative, as expected). Geographic distance is measured as the distance in kilometres between the Netherlands and the concerning home country (Source: CEPII (2008)). Moreover, HMPS is also significant at 1% when added (and positive, as expected). In both cases the sign and significance of the other explanatory variables remains the same.
20 Not mentioned before, also adjusted R-squared improves when using the fixed effects approach. Indicating that the
standard, therefore it is worth to start operations in the Netherlands, especially if the home country standard is relatively low.
Table 2: Regression results full sample
Independent variable Dependent variable: FDI
(1) OLS1 (2) Basis (3) HMPS (4) HMPH (5) HMPL
Home GDP 1.501 4.950 5.001 5.223 4.950
0.112*** 0.673*** 0.675*** 0.687*** 0.674***
Home Trade Cost -3.074 -0.054 -0.032 0.080 -0.054
0.319*** 0.575 0.576 0.578 0.577
Home Export Performance 1.014 -0.329 -0.359 -0.398 -0.329
0.155*** 0.242 0.244 0.244* 0.242
Home Skill 2.190 -6.371 -6.355 -6.445 -6.370
0.510*** 1.276*** 1.276*** 1.272*** 1.278***
Taxdummy -0.827 -0.226 -0.222 -0.204 -0.226
0.192*** 0.123* 0.123* 0.123* 0.123*
Home Market Proximity Standard 3.054
(φ = 1) 3.494
Home Market Proximity High 10.608
(φ = 0.5) 5.858*
Home Market Proximity Low 0.034
(φ = 2) 2.208
Constant -29.038 -34.517 -62.458 -171.825 -34.585
2.201*** 7.858*** 32.922* 76.227** 9.027***
Country dummy No Yes Yes Yes Yes
Period dummy No Yes Yes Yes Yes
Observations 380 380 380 380 380
R-squared 0.513 0.874 0.87 0.875 0.874
Adjusted R-squared 0.506 0.858 0.86 0.859 0.858
Standard errors in italics
* significant at 10%, ** significant at 5%, *** significant at 1%
1
OLS: White heteroskedasticity-consistent standard errors & covariance
We reject hypothesis 5 since the home country corporate income tax rate negatively effects FDI at the 10% significance level. This suggests that when a home country’s corporate tax rate exceeds the average rate of our sample countries, it will have a lower FDI stock in the Netherlands. It should be noted that the corporate tax rate is just one of a variety of tax parameters (e.g. method of double taxation) which influences the investment decision of multinational firms. In the estimation results of section 5.2 we extend deeper about taxation.
market size is relatively more important than distance. So, a larger market surrounding the home country increases the opportunity cost of exporting to the Netherlands, which increases FDI to the Netherlands. Nevertheless, home country market size (GDP) remains a more important indicator than surrounding market sizes. Since HMPH is the only significant proximity effect parameter we show this one in the other two sections as well (except for service FDI in section 5.3). In addition, the sign and significance of the other explanatory variables does not change when the distance decay parameter varies.
5.2 Period and European subsample results
This section checks the robustness of the previous results for both period and geographic effects. Table 3 (next page) shows the estimation results for these effects. As shown in figure 2 (section 4) and tested with the Chow breakpoint test, the growth of FDI stock accelerated during the second decade. Column (1) and (2) show that different determinants explain growth for the two decades. During the decade 1997-2006, and largely equal with the full sample results, GDP, skill and corporate taxation determine inward FDI. During the decade 1987-1996, GDP, export performance and home market proximity are the major determinants. Several conclusions about the effects of the two periods can be drawn.
First, like in the full sample results, GDP is for both decades positively significant at 1%. Moreover, it is the only variable which is significant for both decades. This shows the importance of home country market size as a determinant of inward FDI.
Second, the results for the period 1997-2006 largely match with the full sample results. This makes sense as the largest part of the Dutch inward FDI accumulated during this decade, as shown in figure 2, in section 4.1. Besides GDP, skill is also again significant at 1%. Trade cost and bilateral trade are again no major determinants of FDI.
trade-off between taxes and public services. Taxation is just one of a variety of home country governmental policies that influence the investment decision, other determinants are, for instance, infrastructure and education (Barba Navaretti and Venables, 2004). Therefore, home countries with relatively high corporate tax rates compensate for this with additional public benefits from which multinational firms from these home countries are expected to profit. Although the results are contrary to our expectations, the complexity of taxation and the trade-off largely explain the negative relationship between the home country corporate tax rate and Dutch inward FDI.
Table 3: Regression results - 4 subsamples
Independent variable Dependent variable: FDI
(1) 1987-1996 (2) 1997-2006 (3) Europe (4) Eurozone
Home GDP 4.141 6.717 5.623 4.102
1.186*** 2.229*** 0.779*** 1.099***
Home Trade Cost -0.849 -0.772 0.290 2.069
0.765 1.275 0.661 1.064*
Home Export Performance -1.015 0.215 -0.372 0.576
0.335*** 0.427 0.302 0.588
Home Skill -1.315 -10.293 -7.628 -7.984
1.945 3.237*** 1.395*** 1.883***
Taxdummy 0.128 -0.862 -0.170 -0.461
0.125 0.271*** 0.140 0.203**
Home Market Proximity High 53.0501 -16.490 23.0392 5.340
(φ = 0.5) 13.232*** 20.181 11.764* 12.586
Constant -710.537 165.749 -319.620 -73.128
169.203*** 267.950 147.611** 158.371
Country dummy Yes Yes Yes Yes
Period dummy Yes Yes Yes Yes
Observations 190 190 300 200
R-squared 0.941 0.831 0.864 0.859
Adjusted R-squared 0.929 0.795 0.844 0.830
Standard errors in italics
* significant at 10%, ** significant at 5%, *** significant at 1%
1
HMPS and HMPL are also significant for this sample, though with a lower β
2
HMPS is also significant for this sample, though with a lower β
Third, considering the first decade we observe the importance of export performance and the home market proximity effect. These determinants replace skill and the corporate tax rate, trade cost is again not significant. Market size is again positively significant.
multinational firm gains sufficient knowledge and experience in serving the Dutch market, FDI is preferred to exporting. Acknowledging that the European integration process was less advanced during the first than during the second decade, it makes sense that substitution dominates complementarity. Multinational firms were more focused on serving just the Dutch market than on using the Netherlands as an export-platform, which suggests complementarity as explained in section 3.3. Therefore, during the first decade, FDI and trade are seen as alternatives. Second, as shown in figure 1 (section 2) the share of industry in total inward FDI stock was larger during the first than during the second decade. Since plant-level economies of scale are assumed to play a larger role in industry than in services, the (industrial) multinational firms see FDI and exporting as alternatives. In section 5.3, when analysing industry FDI, we find evidence for this assumption. The third and last significant variable for the first decade is the home market proximity effect. Corresponding with our hypothesis it positively effects FDI like in the full sample. Moreover, it is significant at 1%. This justifies the inclusion of the home market proximity effect as a determinant. Furthermore, the result is the same independent of the distance decay parameter we use. So, both distance to large countries and the size of these countries are important during the decade 1987-1996.
Finally, when comparing both decades we observe that productivity and corporate taxation over time replace the substitution and home market proximity effect. According to our view the process of European integration helps to explain these developments. During the first decade Europe was less integrated and the euro still had to come into existence. In the latter decade, 1997-2006, integration advanced, removing trade and FDI barriers and lowering the role of the market proximity effect, especially the role of distance. Nevertheless, differences in productivity and taxation remain and have become more dominant in determining Dutch inward FDI.
In contrast, trade cost, bilateral trade and corporate taxation are no major home country determinants.
In addition, as noted in section 4.1.1, our home market proximity variable only includes home countries from our sample and not countries which are located close to these home countries. For instance, in the full sample analysis (section 5.1), the home market proximity variable for Japan does not include China. This problem is less relevant for the European subsample since all these home countries are located close to each other. Therefore, the home market proximity variable for Europe basically might give a better reflection. Nevertheless, both in the full sample (Table 2, column 4) and in the European subsample HMPH is positively significant at 10%.
Column (4) in Table 3 shows the determinants of 10 home countries which have adopted the euro, Table D1 shows these eurozone countries. Although their average share in Dutch inward FDI is only 32.8% (DNB, 2008; own calculations), their share has been rising very quickly, contributing to about half of the Dutch inward FDI stock in 2007 (Table B2). The results in column (4) confirm again the positive effect of market size and the negative effect of productivity. Second, corporate taxation effects FDI again in a negative manner, suggesting a trade off among taxes and public benefits, as described above. This makes sense as the countries continue to determine their own tax policy, despite of a common currency.
Finally, trade cost is significant for the eurozone. The positive relationship among trade cost and FDI confirms our hypothesis regarding horizontal FDI among countries with similar welfare standards, which is shown in their common currency. Since trade cost is measured as the inverse of trade openness the positive relation shows that the euro stimulates market-seeking FDI, even if the eurozone countries are relatively closed, which involves higher trade costs.
Before moving to the sectoral division our estimation results in Table 2 and 3 lead us to conclude that:
Using country and annual fixed effects improves the quality of the estimation results. Home country market size is the most important (positive) home country determinant of
The home market proximity variable proves to be a relevant (positive) home country determinant, though evidence remains mixed. The influence of the surrounding market size is relatively more important than the distance to this market.
Bilateral trade and trade costs are of minor importance. 5.3 Sectoral subsample results
In this section we distinguish between industry and service inward FDI stock. Figure 1 (section 2.3) shows that Dutch inward FDI stock consists of around 60% services and 40% industry. Both sectors differ in characteristics like economies of scale, trade-costs and/or factor endowments. Therefore, different home country determinants might effect inward FDI in both sectors. Column (2) and (3) in Table 4 (next page) show that this is actually the case. Like in the pervious sections the role of GDP and skill is for both sectors in accordance with our hypotheses. In addition, as projected by hypothesis 6, home market proximity standard is also an important determinant in both sectors. Differences are that for industry, trade cost, export performance and home market proximity high are significant. For services, on the other hand, corporate taxation and home market proximity low are more important.
Column (1) shows the determinants of the 14 countries from which sectoral data is available. These countries are provided in Table D1. The results from column (1) are almost identical to the full sample results (Table 2, column 4). The only difference is that the taxdummy is now significant at 1% and both home market proximity high and standard are significant at 5%. So, compared with the full sample results the chosen home country determinants are just more significant .
trade costs of (intermediate) goods rise. Therefore, the negative relation indicates that higher trade cost deters industrial FDI since the Netherlands becomes less attractive as an export-platform.
Table 4: Regression results - Sectoral division
Independent variable Dependent variable: FDI
(1) Total (2) Industry (3) Services
Home GDP 5.008 8.243 3.483
0.684*** 1.047*** 0.758***
Home Trade Cost -0.692 -2.314 0.243
0.571 0.874*** 0.666
Home Export Performance -0.323 -1.002 -0.210
0.243 0.372*** 0.295
Home Skill -7.535 -13.094 -5.869
1.203*** 1.841*** 1.419***
Taxdummy -0.439 -0.207 -0.736
0.118*** 0.182 0.139***
Home Market Proximity High 15.7341 36.7131
(φ = 0.5) 6.171** 9.443***
Home Market Proximity Low 11.7922
(φ = 2) 5.031**
Constant -231.936 -525.888 -36.338
79.757*** 122.056*** 11.330***
Country dummy Yes Yes Yes
Period dummy Yes Yes Yes
Observations 280 280 280
R-squared 0.882 0.825 0.870
Adjusted R-squared 0.863 0.797 0.849
Standard errors in italics
* significant at 10%, ** significant at 5%, *** significant at 1%
1
HMPS is also significant for this sample, though with a lower β
2 HMPS is also significant for this sample, though with less significance
