The euro effect – the impact of EU bilateral real exchange rates on German net FDI : evidence from Germany and seven EU-countries

(1)

The euro effect – the impact

of EU bilateral real exchange

rates on German net FDI:

MASTER THESIS WITHIN: Economics NUMBER OF CREDITS: 30 ECTS

PROGRAMME OF STUDY: Economic Analysis AUTHOR: Aenne Ohainski

JÖNKÖPING May 2019

(2)

Master Thesis Economics

Title: The euro effect – the impact of EU bilateral real exchange rates on German net

FDI:

evidence from Germany and seven EU-countries

Author: Aenne Ohainski

Tutors: Tina Wallin and Toni Duras

Date: 2019-05-20

Key terms: net FDI, real exchange rate, euro, horizontal FDI, vertical FDI, market-oriented firm, cost-oriented firm, euro-effect, Germany, European Union

___________________________________________________________________________

Abstract

In literature it has been stated that in times of low capital barriers policies can impact real exchange rates (RERs) and, it has been shown that RERs influence foreign direct investment (FDI). As inward FDI is a growth stimulating factor for the German economy and as more than a third of inward FDI stems from countries in the European Union (EU), this study investigates the RER-FDI link between Germany and seven EU countries. The impact of bilateral RERs between Germany and seven EU countries on German net FDI inflows is examined for the period 1974-2018. Further, it is investigated how the euro introduction in 1999 affected the RER-FDI links. Using Ordinary Least Squares models it is found that in the pre-euro period a real German currency appreciation led to decreases in net FDI from most economies in scope. This negative RER-FDI link endures for the non-euro countries Sweden, Denmark, and the United Kingdom after the euro introduction. France, Italy, and Spain, euro countries, are subject to the euro-effect: the negative RER-FDI link changes to a positive link with the euro introduction. This phenomenon indicates an altering investment behavior. The results are strengthened by a panel estimation as robustness check. As the euro-effect was not discovered in previous studies nor is a theory established explaining the altering investment behavior of euro firms, this thesis suggests an alternative explanation.

(3)

Tables

Table 1 OLS models - sample 1974Q1-1998Q4 ... 18

Table 4 Fixed effects panel estimation ... 25

Table 5 Descriptive statistics ... ii

Table 6 Linear correlation coefficients ... iii

Table 7 Augmented Dickey-Fuller unit root test ... iii

Table 8 Maddala and Wu unit root test... iv

Table 9 OLS models with bilateral exports - sample 1999Q1-2018Q2 ... viii

Figures

Figure 1 Nominal German net inward FDI flows 2011Q1-2018Q2 ... 1

Figure 2 Theory overview ... 7

Figure 3 Horizontal FDI versus export decision in the HFT model ... 9

Figure 4 Vertical FDI versus export decision in the HFT model ... 9

Figure 5 FDI transition in the 90s ... 16

Figure 6 The 90s-transition captured by dummies ... 16

Figure 7 Supply under horizontal and vertical FDI in the pre-euro period ... 27

Figure 8 Supply under horizontal and vertical FDI in the euro period ... 27

Figure 9 Net FDI flows to Germany in the time period 1974Q1-2018Q2 ... v

Figure 10 Real exchange rate standard deviation in the pre-euro and euro period ... vi

Appendix

Appendix A: Definitions ... i

Appendix B: Tables ... ii

Appendix C: Figures ... v

(5)

Abbreviations

ADF Augmented Dickey-Fuller

BDI Bundesverband der Deutschen Industrie

CPI Consumer price index

EU European Union

FDI Foreign direct investment

GDP Gross domestic product

HAC Heteroscedasticity and autocorrelation consistent

HFT model Heterogenous trade model

MW Maddala and Wu

OECD Organisation for Economic Co-operation and Development

OLS Ordinary least squares

PPP Purchasing power parity

RER Real exchange rate

WTO World Trade Organization

Country codes

AT Austria DE Germany DK Denmark ES Spain FI Finland FR France GR Greece IT Italy NL the Netherlands SE Sweden

(6)

1

1 Introduction

Foreign direct investment (FDI) is an important factor for Germany, contributing to economic growth, increasing competitiveness, and intensifying integrity with other countries. For these reasons it is essential to understand the determinants of FDI flows. Besides factors such as gross domestic product (GDP) and interest rates, a vast amount of empirical research has shown that also real exchange rates (RERs) influence FDI flows through the foreign exchange effect on profits.1_{Since European Union (EU) countries are the leading investors in Germany, in 2016 74%} of German inward FDI stemmed from EU-countries (Deutsche Bundesbank, 2018a), the investment relationship between EU countries and Germany is of primary interest. However, a limited amount of empirical research has examined the connection between bilateral FDI flows across EU countries and bilateral real exchange rates since the euro introduction.

To fill this gap, this thesis aims to examine the impact of bilateral real exchange rates on German net inward FDI2_{from EU countries. There are two key reasons to investigate German net} inward FDI: firstly, FDI inflows contribute to sustainable growth in Germany, in 2013 2.6 million jobs based on German inward FDI; however, net inflows have been unstable and partial negative from 2011 until 2018 (see Figure 1). Moreover, the current shrinking share of the working population and the bare presence of raw materials make productive capital necessary to foster economic wealth (BDI, 2013).

Figure 1 Nominal German net inward FDI flows 2011Q1-2018Q2

Source: Own depiction using data from Deutsche Bundesbank.

Secondly, as Barrell and Pain (1998) argue, through the removal of capital barriers in the EU, policies can have a significant impact on RERs and, therefore, can influence FDI flows.

However, researchers do not agree whether a RER appreciation in the FDI host country positively or negatively affects FDI inflows to that country.3_{Several empirical studies argue for a}

1_{See, e.g., Froot and Stein (1991), Klein and Rosengren (1994), Kosteletou and Liargovas (2000) and Chen, et}

al. (2006).

2_{For an explanation how net FDI is composed, see Appendix A.}

3_{Empirics and theories predominantly model FDI flows and not net FDI flows. However, following the reasoning of}

Culem (1988) that net FDI flows are a good proxy for gross FDI flows (see Section 5 for a further discussion), this thesis also examines literature that uses (gross) FDI flows.

-10,000 10,000 30,000 20 11-Q1 20 12-Q1 20 13-Q1 20 14-Q1 20 15-Q1 20 16-Q1 20 17 -Q1 20 18-Q1 € mi lli on world EU-countries

(7)

2

negative relationship, i.e. that FDI flows to a foreign country decrease when this country’s currency appreciates (see, e.g., Cushman, 1985; Froot & Stein, 1991).4 _{Other researchers argue for a positive} link, i.e. that FDI flows to a foreign country increase when the foreign country’s currency appreciates (see, e.g., Campa, 1993; Lee, 2015).5

Chen et al. (2006) investigate the RER-FDI link on a firm level instead of on a country level. The economists predict that the RER effect on FDI depends on the firms’ structure. They distinguish between two types of firms: market-oriented and cost-oriented. A market-oriented firm creates a positive RER-FDI link, while a cost-oriented firm creates a negative RER-FDI link.

Since no firm level data is available for EU countries, the framework by Chen et al. (2006) is applied on country level in this thesis. As, for example, found in a study on country level by Kosteletou and Liargovas (2000), the effect of a RER appreciation leads to different FDI reactions also across EU economies. However, the RER-FDI link has not been examined between euro countries, possibly due to that a constant nominal exchange rate prevails. But as suggested by several theories, it is the real exchange rate between countries affecting FDI (see, e.g., Cushman, 1985). Nevertheless, it was found that the euro introduction decreased real exchange rate risk between the euro-countries and fostered FDI considerably (see, e.g., Schiavo, 2007). Moreover, it was found that the euro has stimulated the activity of cost-oriented firms while others claim activity of market-oriented firms increased (see, Petroulas, 2007; Coeurdacier et al., 2009).

This raises the question as to how the euro influenced the motives of companies to structure, i.e. to become either market-oriented or cost-oriented. The relative change between market-oriented and cost-oriented firms in turn affects the RER-FDI link. However, there is a scant amount of research examining the motives of firms to structure and no connection is drawn to the exchange rate regime.

As a consequence of the lack of empirics and theories about German net inward FDI from EU countries and the effect of the euro on heterogenous FDI flows reflected through the RER-FDI link, this thesis seeks to answer two questions: Firstly, how have bilateral real exchange rates between Germany and EU-countries affected German net inward FDI from these countries in the period 1974-2018 and secondly, how has the euro introduction influenced this linkage.

Accordingly, this study elaborates implications particularly interesting for EU policymakers. In times where capital is increasingly mobile, political decisions can have great effects on real exchange rates or directly on FDI (Barrell & Pain, 1998). Moreover, Dunning (2009) emphasizes that macroeconomic policies have taken increasing importance on the risk assessment of FDI by

4_{The situation in which a foreign RER appreciation decreases (net) FDI inflows to that country is, henceforth also}

referred to as a negative RER-FDI link.

5_{The situation in which a foreign RER appreciation increases (net) FDI inflows to that country is, henceforth also}

(8)

3

multinational enterprises and their locational decisions. Therefore, examining the relationship between the real exchange rate and German net FDI inflows creates valuable information for policymakers to foster, foresee, or steer heterogenous net inward FDI.

Moreover, the investigations in this study are important for German and European investors. When investors know how FDI flows are influenced by real exchange rates, they can develop investment strategies according to the locational choice of firms.

To find answers, the relationship between bilateral RERs and German net inward FDI from euro countries and non-euro countries are estimated with Ordinary Least Squares (OLS) models. Due to limited data availability and stationarity problems, seven EU countries are in scope of the study, namely Austria, France, Italy, Spain, Denmark, Sweden, and the United Kingdom.

This thesis finds that there existed a significant negative RER-FDI link for the majority of economies from 1974 until the euro introduction in 1999. The negative RER-FDI link endures for the non-euro countries, Denmark, Sweden, and the United Kingdom in the euro era. However, France, Italy, and Spain, euro countries, are subject to a positive RER-FDI link from the euro introduction onwards, suggesting an altering investment behavior evoked by the euro. This phenomenon is called euro-effect in this study. This empirical finding has not been discovered in previous studies nor is a theory established covering the results. Therefore, this thesis suggests an alternative explanation for the euro-effect.

The results, however, are limited due to data restrictions, a vaster sample of countries might yield stronger results, for or against the euro-effect. Moreover, the euro-effect was only found for German net FDI, investigating several EU economies’ net FDI can discover whether this effect holds for more EU countries.

The remaining thesis is structured as follows: Section 2 gives an overview of relevant literature conducted in the field of RER-FDI linkages and reviews the effect of the euro on FDI. Section 3 presents the theoretical frameworks, while Section 4 describes the methodology applied to answer the research questions. In Section 5 the data is described which will be used in Section 6 to estimate the econometric models. Moreover, Section 7 evaluates the results and presents an alternative explanation of the euro-effect. Finally, Section 8 concludes.

(9)

4

2 Literature review

This section reviews literature and empirics which are linked to the research questions in this thesis. Section 2.1 sheds light on the first research question, presenting literature which studied how real exchange rates affect FDI. The second research question has not been examined in literature to date. Therefore, Section 2.2 summarizes literature that touches upon the effect of the euro on the RER-FDI link.

2.1 The real exchange rate and FDI

The relationship between FDI and RERs has been studied frequently in the past decades. Different empirical approaches have yielded ambiguous relationships (see, e.g., Kosteletou & Liargovas, 2000) resulting in the establishment of several theories since the 1970s (Ahmadi-Esfahani & Phillips, 2008). The literature investigating the RER effect on FDI can be separated into four groups of theories and empirical findings:

Firstly, literature claiming that a foreign bilateral RER appreciation has a negative impact on FDI flows to the foreign country. This effect occurs as changes in the RER evoke relative wealth or cost differences. With a real appreciation of the FDI host country’s currency costs relatively increase for the FDI source country and, therefore, FDI flows to the FDI host country decrease (see, e.g., Cushman, 1985; Kohlhagen, 1977). Froot and Stein (1991) examined FDI flows of several economies, including Germany, for the period 1972–1987 and indeed find, in a linear regression analysis, that a German RER appreciation significantly decreases capital inflows. Likewise, Pain and Van Welsum (2003) find, by using a non-linear regression, that an appreciation of the RER had a negative short-run effect on German FDI inflows in the period 1972–2000.

Secondly, literature arguing that a foreign bilateral RER appreciation has a positive impact on FDI flows to the foreign country. An appreciating currency indicates a well-performing economy and can evoke investors’ expectations of increases in future profits and, thus, leads to an increase in FDI flows to that economy (see, e.g., Campa, 1993; Lee, 2015). This reasoning for the German economy is supported by, for example, the empirical study by Kosteletou and Liargovas (2000) that finds, using an OLS model, that a weaker German currency decreased German inward FDI between 1960 and 1997.

Thirdly, literature which rejects that the bilateral RER is affecting FDI flows. Economists argue that although a foreign RER appreciation can increase prices for FDI in that country, likewise the returns in that country are increasing. Increases in return and in the price offset each other and make the RER irrelevant in the FDI location decision (Blonigen, 1997). Empirical support for the independence of the RER and FDI is found by, for example, De Vita and Abbott (2007). Examining the United Kingdom’s inward FDI, including flows from Germany, for the period

(10)

5

1975–2001, with a generalized method of moments model, the economists show that the RER does not affect FDI.

Lastly, literature claiming that a foreign bilateral RER appreciation can have a positive as well as a negative impact on FDI flows to the foreign country, depending on the firm’s structure and investment strategy (see, e.g., Cushman, 1985). In theory there are two types of firms: market-oriented firms and cost-market-oriented firms.

Market-oriented firms produce in the foreign market and also serve it; this type of investment is termed horizontal FDI. Often it is assumed that market-oriented firms produce the same good in the home market and foreign market (see, e.g., Baldwin et al., 2008). The motives for the firm to invest horizontally are either expanding the market or avoiding export costs if the foreign market is already served (see, European Central Bank, Ad Hoc Task Force, 2005; Baldwin et al., 2008). It follows that when a foreign currency appreciates, horizontal FDI flows to that country increase since the firm remits profits earned in the foreign market to the domestic economy with a favorable exchange rate.

Cost-oriented firms produce abroad and reimport the output to the home market, an investment referred to as vertical FDI. The primary motive behind investing vertically is to fragment the production process strategically to save costs (European Central Bank, Ad Hoc Task Force, 2005). Firms conducting vertical FDI negatively react to foreign currency appreciations since the production abroad becomes relatively more expensive.

Chen et al. (2006) showed that for Taiwan the extent of FDI flows to China depends on the firms’ motives. With the use of micro data, the researchers identify market-oriented and cost-oriented firms in the economy. The first group’s FDI flows to China significantly increased when the Chinese currency appreciated relative to the Taiwanese. Cost-oriented firms’ FDI behavior showed the reverse reaction during the period 1991-2002. The mechanisms behind this are explained in detail in the theoretical part, in Section 3.1.

2.2 The euro and FDI

Many studies and empirics support that a monetary union fosters FDI flows between the participating countries (see, e.g., Pain & Van Welsum, 2003; Schiavo, 2007). Also, the impact of the euro on FDI flows has been studied frequently in the past, numerous empirics agree that the euro has increased bilateral FDI flows (see, e.g., Abbott & De Vita, 2011; Choi & Park, 2012). Economists argue that the euro has led to a stronger financial integration and the elimination of exchange rate uncertainty (decreasing the risk level of investment opportunities), increased stability of inflation rates, and a fall in capital costs (see, e.g., Schiavo, 2007; Coeurdacier et al., 2009). However, since the euro introduction in 1999, studies analyzing FDI flows to and from EU

(11)

6

countries mainly focused on investment relations to economies outside Europe (see, e.g., Abbott & De Vita, 2011). The impact of currency unions on OECD countries’ FDI flows was, for example, examined by Schiavo (2007). The researcher finds that a monetary union increases FDI flows by approximately 160% up to 320%. This increase is gained through, for instance, the elimination of exchange rate uncertainty. Contrarily, there are also studies emphasizing that it is not the euro, but the Single Market and the EU that have been stimulating FDI flows since the 90s (see, e.g., Flam & Nordström, 2008; Dinga & Dingová, 2011). This thesis, however, seeks to examine the impact of the euro on the RER-FDI link but not to investigate whether the euro has fostered FDI flows. Most studies do not consider heterogeneous firms i.e., cost- and market-oriented firms, and in turn neglect whether the euro affects the firm decision to either invest vertically or horizontally.

Nevertheless, to the author’s knowledge, there have been two studies investigating the effect of the euro on horizontal and vertical FDI. The study by Coeurdacier et al. (2009) considers heterogeneity among firms when examining cross border mergers of the manufacturing sector. The researchers find that the euro had a significant effect on firms that invest abroad and sell their product abroad, i.e., market-oriented firms conducting horizontal investments. FDI flows by market-oriented firms have been increasing through the euro by 200% within the euro area and by 70% between non-euro and euro-countries. Petroulas (2007) finds indications that the euro has been stimulating vertical FDI within the euro area. Thus, empirics have found absolute increases in horizontal and vertical FDI evoked by the euro but to examine the effect of the euro on the RER-FDI link requires empirics about the relative change between the two FDI types.

(12)

7

3 Theory

In order to find answers for the two research questions: firstly, how the real exchange rate has influenced net inward FDI and secondly, how this relationship has been affected by the euro introduction, it requires a theoretical framework covering both problems. Figure 2 shows the required theoretical framework: an explanation of how a currency union determines or alters firms’ decisions to become market- or cost-oriented and, thus, to invest horizontally or vertically (answering research question two), and how the resulting FDI flows by the two different types of firms are affected by the real exchange rate (answering research question one). However, a theory modelling the first mentioned relationship is lacking in the literature. Therefore, the model by Baldwin et al. (2008) is used as an auxiliary framework. It explains how the euro affected horizontal and vertical FDI flows in the euro area. However, horizontal and vertical FDI are modelled in two different frameworks: first, considering horizontal FDI and exports as a trade-off and second, considering vertical FDI and exports as a trade-off. Consequently, the trade-off between either conducting horizontal FDI or vertical FDI is missing.

The other part of the required theoretical framework, how the firm structure is related to the firm’s reaction to an appreciating RER, is modelled by Chen et al. (2006).

In the following, the theory by Chen et al. (2006) is explained and in the subsequent subsection, the theories by Baldwin et al. (2008) are presented.

3.1 The impact of the real exchange rate on horizontal and vertical FDI - Chen et al., 2006 Chen et al. (2006) extend the real options model by Dixit (1989) to explain the influence of the real exchange rate on FDI flows. The researchers assume that there are two types of firms:

(13)

8

oriented firms, conducting horizontal FDI, and cost-oriented firms, conducting vertical FDI. Both types of firms aim to maximize their profits in home currency terms and depending on the firm’s type, i.e., market-oriented or cost-oriented, the real exchange rate affects the profits differently.

A cost-oriented firm which is producing abroad to reimport the output to the home market suffers from a foreign real currency appreciation. The production costs in the foreign market increase in home currency terms and, thus, profits decrease. Therefore, an appreciating foreign currency reduces the likelihood that the cost-oriented firm enters the foreign market. This implies that there exists a negative relationship between FDI and the real exchange rate, i.e., a negative RER-FDI link.6

A market-oriented firm which is producing and selling its goods in the foreign market reacts reversely. A foreign real currency appreciation increases profits in home currency terms since earnings gained in the foreign market are remitted with a favorable exchange rate. Accordingly, with a foreign real currency appreciation the likelihood that the market-oriented firm enters the foreign market increases; this implies that there exists a positive relationship between the real exchange rate and FDI, i.e., a positive RER-FDI link.

Consequently, inferred from the theory by Chen et al. (2006), a German RER appreciation is expected to have a positive impact on German inward FDI if the FDI-sending economy is dominated by market-oriented firms. On the other hand, a German RER appreciation is expected to have a negative impact on German inward FDI if the FDI-sending economy is dominated by cost-oriented firms (see Figure 2).

3.2 The impact of the euro on horizontal and vertical FDI - Baldwin et al., 2008

The second research question focuses on how a firm’s decision of investing horizontally or vertically is influenced by the exchange rate regime, i.e. floating in the pre-euro period and fixed for the euro countries in the euro-period. Since there is no theoretical framework incorporating the exchange rate regime as a factor in the endogenous decision-making7_{, the concept by Baldwin et} al. (2008) serves as an auxiliary theory.

The researchers build their model on the Heterogenous Trade (HFT) model by Melitz (2003) in which firms have different levels of competitiveness. In the framework it is assumed that horizontal FDI is a substitute for exports and that vertical FDI is a complement to

6_{A negative impact of the RER on FDI has also been modelled in earlier theories (see, e.g., Kohlhagen, 1977); however,}

the reasoning partly differs across the theories as, for example, that bilateral RER appreciations or depreciations evoke relative wealth differences (see Froot & Stein, 1991).

7_{The knowledge-capital theory models the decision of investing either a horizontally or vertically endogenously (see,}

(14)

9

exports. Firms aim at maximizing their profits which negatively depend on the costs of exports and the costs of FDI and positively depend on the firm’s level of competitiveness.

Figure 3 depicts the profits of exporting firms, 𝜋𝑋, and the profits generated by horizontal FDI, 𝜋𝐻, in dependence of the firm competitiveness. Horizontal FDI has higher fixed costs, 𝐹𝐻, than exporting, 𝐹_𝑋, but the marginal costs are lower for horizontal FDI as no trade costs arise, therefore, the slope of 𝜋𝐻 is steeper than of 𝜋𝑋. The threshold when it becomes profitable to export instead of only serving the home market is shown by 𝑎𝑥. When the level of competitiveness is less than 𝑎𝑥 only the home market is served. Correspondingly, when the level of competitiveness is greater than at the intersection, 𝑎𝐻, the firm invests and sells abroad, i.e. conducts horizontal FDI. Therefore, this model predicts that the most uncompetitive firms only serve the home market, medium competitive firms export goods, and the most competitive invest horizontally abroad.

home export horizontal FDI home export vertical FDI

€ home export horizontal FDI € home export vertical FDI

Source: Own depiction following Baldwin et al. (2008). Source: Own depiction following Baldwin et al. (2008).

In a next step in Figure 3 the introduction of the euro is modelled. A common currency reduces marginal trade costs as, for example, administrative costs, hedging or the effort of dealing with two currencies. Consequently, the export profit curve, 𝜋𝑋, turns counterclockwise, resulting in a new curve 𝜋𝑋€. The reduction in marginal costs lowers the threshold of exporting; the new threshold is 𝑎𝑋€. Moreover, the threshold of conducting horizontal FDI is affected, the intersection

Figure 3 Horizontal FDI versus export

(15)

10

of export profits and FDI profits moves to the right to 𝑎𝐻€. Hence, as exports are becoming less costly, firms substitute horizontal FDI with exports and only the most competitive firms conduct horizontal FDI. Therefore, this model predicts that the euro introduction has increased the threshold to conduct horizontal FDI for euro firms, implying a reduction in horizontal activity with the euro introduction.

Similarly as in the case for horizontal FDI, Figure 4 shows the vertical FDI profit, 𝜋𝑉, and export profit, 𝜋𝑋, curves in dependence of the home market’s level of firm competitiveness. The thresholds for vertical FDI and exports are depicted by 𝑎𝑉 and 𝑎𝑋, respectively. 𝐹𝑉 and 𝐹𝑋 represent the fixed costs for investing vertically and for exports. Both profit curves are dependent on transport costs, thus, the euro introduction rotates both curves, 𝜋_𝑉 and 𝜋_𝑋, counterclockwise. However, as Baldwin et al. (2008) argue, FDI profits are affected more than export profits since vertical FDI requires additional transports (as e.g., transports of intermediate goods) than only exporting the final good. Consequently, the thresholds for exports as well as vertical FDI are lowered, moving from aX to a€X and from aV to aV€. This implies that exports and vertical FDI activity increase with the euro.

Applying the outcome of the theory by Baldwin et al. (2008) on the second research question, the euro introduction is expected to have led to a decrease in horizontal FDI activity but increased the vertical FDI activity (see Figure 2). Further, if market-oriented firms dominated the market in a euro country in the pre-euro period and if the positive effect of the euro on vertical FDI has been strong enough to reverse the horizontal-vertical FDI ratio then it is expected that the positive RER-FDI link changes to a negative RER-FDI link. However, if cost-oriented firms dominated the market in a euro country in the pre-euro period then it is expected that the euro introduction has strengthened the negative RER-FDI link.

(16)

11

4 Methodology

This section provides a description of the models used in this study to examine how German net FDI inflows are affected by bilateral real exchange rates and how the euro influenced this link. To determine the latter, the sample needs to be split into two so that the model is estimated for the period 1974Q1-1998Q4 (i.e., the pre-euro period) and the period 1999Q1-2018Q2 (i.e., the euro period). The reason for not using dummies to examine the effect of the euro introduction is that a euro-dummy can only handle shifts in levels of FDI but is not able to capture a possible alteration of the sign of the RER coefficient.

In the first step, OLS regressions are used to investigate FDI flows to Germany by estimating separate models for each country before and after the euro introduction. As a robustness check, two panel models are estimated, bundling the countries with a positive RER-FDI link and the countries exhibiting a negative RER-FDI link. Section 4.1 outlines the OLS model, while Section 4.2 discusses the panel model.

4.1 OLS model

To investigate the impact of the real exchange rate on net FDI to Germany this study is following a similar approach as Kosteletou and Liargovas (2000) by using an OLS model for each country in scope. The OLS model is chosen as its results are simple to interpret while it is also a powerful model to use when determining linear long-run relationships (see, e.g., Gujarati & Porter, 2009).

This study estimates the following OLS model for each EU country in scope: 𝐹𝐷𝐼_𝑡 = 𝑐 + 𝛽₁𝑙𝑜𝑔(𝑅_𝑡) + 𝑩𝑇𝑾_𝑡+ 𝜀_𝑡,

where 𝑐 is the intercept term. 𝑅𝑡 is the bilateral real exchange rate between the examined country and Germany in quarter 𝑡. The real exchange rate is estimated in log-form to simplify the interpretation of the coefficient 𝛽1.8 𝐹𝐷𝐼𝑡 is German net inward FDI in percentage of the investing country’s GDP. The 𝑛 × 1 vector 𝑾𝑡 is a vector of control variables where 𝑛 is the number of variables (see Section 5) and the 𝑛 × 1 vector 𝑩 is the vector of coefficients. Lastly, 𝜀𝑡 is the error term.

However, to be able to determine long-run relationships the model relies on the assumption that the variables used are stationary. It is well-known in the literature that applying an OLS technique on non-stationary variables may lead to spurious results.9_{The current study, therefore,}

8_{A one percent increase of the real exchange rate leads to a 𝛽}

1 increase of 𝐹𝐷𝐼𝑡, ceteris paribus.

9_{The problem of spurious regressions was first defined by Granger and Newbold (1974). If the variables in the model}

are non-stationary, the OLS estimation experiences endogeneity and the errors are serially correlated (Hansen & Philips, 1990).

(17)

12

uses the Augmented Dickey-Fuller (ADF) unit root test (see Dickey & Fuller, 1979) to determine for which countries it is possible to estimate the model in level form.

In the EU, gross FDI flows have increased over time for reasons such as reduced trade and investment barriers (see, e.g., Barrell & Pain, 1998), similarly net FDI flows increased since 1974 in absolute terms (see Figure 9 in Appendix C). It is, therefore, likely that the absolute expansion of the dependent variable over time influences the residuals to likewise increase over time in absolute terms. Consequently, the residuals might suffer from heteroscedasticity. In the presence of heteroscedasticity, the OLS estimator is still linear unbiased as well as consistent; however, it is no longer efficient (Gujarati & Porter, 2009). This thesis handles this problem by using the Newey-West (1987) method, which is possible to use to obtain OLS estimators corrected for heteroscedasticity and autocorrelation when it is present. The resulting standard errors are known as HAC (heteroscedasticity- and autocorrelation-consistent) standard errors.

4.2 Panel model

As a robustness check two fixed effects panel OLS models are estimated, bundling the countries with a positive RER-FDI link and the countries with a negative RER-FDI link.

As the countries exhibit different characteristics, the panel estimations allow for different intercepts, 𝑐𝑖, for each cross section by using cross-sectional fixed effects. Accordingly, the panel estimation takes the form:

𝐹𝐷𝐼𝑖𝑡 = 𝑐𝑖+ 𝛽1𝑙𝑜𝑔(𝑅𝑖𝑡) + 𝑩𝑇𝑾𝑖𝑡+ 𝜀𝑖𝑡,

where 𝑖 refers to the country and the remaining terms have the same interpretation as in Section 4.1. As with the OLS model, it is required that the panel series is stationary. As the power of panel-based unit root tests is higher compared to conducting separate tests for each individual series (Levin et al., 2002), panel stationarity tests are conducted. Since heterogeneity across countries is assumed over the investigated period and because of an unbalanced panel this thesis uses the Maddala and Wu (1999) (MW) unit root test.10

(18)

13

5 Data

This section describes the variables used to estimate the RER-FDI link. Real net FDI flows serve as the dependent variables. Culem (1988) argues that factors influencing investments and disinvestments are symmetric, suggesting that net FDI is similar to gross FDI but with a lower intercept. The independent variables of main interest used are real bilateral exchange rates.11_The construction of the FDI and RER variables are described in the following subsection, 5.1, followed by a presentation of the control variables and dummy variables in subsections 5.2 and 5.2.1, respectively.

All data, if not stated otherwise, is obtained from the OECD. Moreover, all data used in this study are in real terms, but for convenience, the prefix real in the subsequent sections is neglected. Furthermore, descriptive statistics and a correlation matrix of the main variables can be found in Table 5 and Table 6 in Appendix B.

5.1 Main variables

The data used in this study reach from 1974Q1 until 2018Q2 and is collected on a quarterly basis. As the dependent variable German net inward FDI flows12_{from EU-countries, measured in euros,} are used. The data are collected from the Deutsche Bundesbank and outliers have been removed. Nominal net FDI flows from country 𝑖, 𝑓𝑑𝑖𝑖, are deflated by the German consumer price index (CPI), 𝐶𝑃𝐼𝐷𝐸, using the base period 1994Q4. Thus, real net FDI flows to Germany from country 𝑖 are calculated as follows:

𝐹𝐷𝐼_𝑖𝑟𝑒𝑎𝑙 = 𝑓𝑑𝑖𝑖

𝐶𝑃𝐼𝐷𝐸.

Following the approach of Culem (1988) each real FDI series is divided by the investing country’s real GDP, 𝐺𝐷𝑃𝑖.13 This approach is used as it is assumed that FDI flows are linked to the size of the economy and, thus, increase in absolute terms over time. Consequently, the dependent variable, real net inward FDI as a percentage of GDP, is calculated as follows:

𝐹𝐷𝐼_𝑖 =𝐹𝐷𝐼𝑖𝑟𝑒𝑎𝑙

𝐺𝐷𝑃𝑖 ∙ 100.

11_{A justification for either considering the nominal or real exchange rate is the theory of purchasing power parity}

(PPP). Purchasing power parity states that the nominal exchange rate and the relative price level of the respective country move proportionally and that the real exchange rate reverts to a constant level in the long run (Christidou & Panagiotidis, 2010). Consequently, if the theory holds the two variables can be used interchangeably in the model. However, validity testing of the PPP hypothesis within Europe has yielded ambiguous results (see, e.g., Lopez & Papell, 2007; Berka et al., 2012). Therefore, and as argued by Cushman (1985), price levels play an essential role in long-term investment, this study uses real exchange rates instead of nominal.

12_{The Deutsche Bundesbank’s definition and measurement method of FDI can be found in Appendix A.}

13_{The quarterly nominal GDP series are in a first step deflated by the CPIs and afterwards converted to euro using}

(19)

14

For simplicity the variable 𝐹𝐷𝐼𝑖 is referred to as net FDI (instead of real net FDI as a percentage of GDP) in the remaining sections.

The real exchange rate is constructed using the bilateral nominal exchange rate, 𝑆(𝐷𝐸:𝑖)14, between country 𝑖 and Germany by deflating it with the CPI ratio between Germany and country 𝑖:

𝑅(𝐷𝐸:𝑖)= 𝑆(𝐷𝐸:𝑖)·

𝐶𝑃𝐼𝐷𝐸

𝐶𝑃𝐼𝑖 .

Accordingly, an increase in 𝑅(𝐷𝐸:𝑖) represents a real appreciation of the German currency. 5.2 Control variables

Several researchers argue that a vital factor attracting FDI is economic growth in the FDI host country (see, e.g. Culem, 1988). Investors can, for example, exploit a higher GDP growth rate in Germany compared to the GDP growth rate in the home market (Culem, 1988). Therefore,

𝐺𝐷𝑃_{𝑑𝑖𝑓𝑓,𝑖}, capturing the difference between German real GDP growth, 𝐺𝐷𝑃̂_𝐷𝐸_{, and the FDI}

origin’s country real GDP growth, 𝐺𝐷𝑃̂ , is included in the model: 𝑖

𝐺𝐷𝑃_{𝑑𝑖𝑓𝑓,𝑖} = 𝐺𝐷𝑃̂_𝑖 − 𝐺𝐷𝑃̂_𝐷𝐸.

In the following, the GDP growth differential is referred to as GDP differential. Culem (1988) finds, in line with theoretical predictions (see, e.g. Hymer & Rowthorn, 1970), that the GDP differential has a negative impact on FDI.

Boateng et al. (2015) argue that trade openness creates a business-friendly environment and investment freedom, thus, affecting FDI positively. Following this study, trade openness, 𝑇𝑂_𝑖, is included in the model and calculated as the total value of trade, comprising exports, 𝐸𝑋𝑖, and imports, 𝐼𝑀𝑖, divided by 𝐺𝐷𝑃𝑖 of country 𝑖:

𝑇𝑂𝑖 =𝐸𝑋𝑖+𝐼𝑀𝑖

𝐺𝐷𝑃_𝑖 .

14_{Where the German currency is the quoted currency and the foreign currency is the currency in which the price is}

(20)

15

Furthermore, differences in interest rates across countries play a key role in investors’ decisions (see, e.g., De Vita & Abbott, 2007; Culem, 1988), therefore, short-term interest rate differentials are incorporated in the model. Real interest rates are constructed by subtracting the quarterly inflation rates from the nominal short-term interest rates for each of the examined countries. To construct differentials, the German expected real interest rate for the point in time of investment, 𝑖𝑒

𝐷𝐸, is subtracted from the lagged real interest rate of country 𝑖, 𝑖𝑖,𝑡−1. Data of the German expected real interest rate is provided by the Deutsche Bundesbank and the differentials are calculated as follows:

𝑖𝑑𝑖𝑓𝑓,𝑖 = 𝑖𝑖,𝑡−1− 𝑖𝐷𝐸𝑒 .

Assuming that FDI is financed through sources from the home country (see De Vita & Abbott, 2007), 𝑖𝑑𝑖𝑓𝑓,𝑖 is expected to have a negative impact on German net inward FDI.

Moreover, it has been argued that wage differences across countries matter as firms prefer locating where wages are lower if not accompanied by lower productivity. Following the approach of Culem (1988) unit labor costs, wages adjusted by productivity, are used to construct differentials:

𝑈𝐿𝐶𝑑𝑖𝑓𝑓,𝑖 = 𝑈𝐿𝐶𝑖 − 𝑈𝐿𝐶𝐷𝐸.

It is expected that 𝑈𝐿𝐶𝑑𝑖𝑓𝑓,𝑖 has a positive effect on German net inward FDI.

Due to data constraints, the interest rate differentials and unit labor costs differentials can only be used for the euro period, i.e. 1999-2018. The GDP differentials and trade openness variables, however, are available for the entire estimation period, i.e. 1974-2018.

5.2.1 Dummy variables

In the models two dummy series are incorporated. Hájek and Horváth (2016) point out that the 1990s were subject to a transition process, which led to greater volatility in FDI, as also observed for the data in this study (see Figure 9 in Appendix B). Events and processes as the gradual east German integration since 199015_{, the establishment of the European Single Market in 1993, the} foundation of the World Trade Organization (WTO) in 1995 and the entry of Austria, Finland, and Sweden in the EU in the same year led to increases in FDI.16_{As these events induced shifts in}

15_{The German reunification in 1990 did not lead to a significant structural break in FDI, as foreign direct investment}

mainly has been taking place in West Germany. In 2003, for example, only 2.7% of the German inward FDI stock was accounted to East Germany (Buch & Toubal, 2009).

16_{The positive impact of the WTO on FDI is, for example, discussed by Hanh (2011). The positive impact of EU}

(21)

16

net FDI, 17_{dummies are used to capture these exogenous shocks. Dummies are incorporated from} 1995Q1 onwards in the model, one capturing the positive shift (95_𝑝𝑜𝑠𝑖) and one capturing the negative shift (95_𝑛𝑒𝑔𝑖).18 Figure 5 shows the increase of net FDI in the 90s for the seven EU countries in scope of this study. Evaluating all positive net FDI flows to Germany the median of 𝐹𝐷𝐼_𝑖 is 0.09 from 1974Q1-1994Q4, in the period 1995Q1-1998Q4 the median rises to 0.54. The median is -0.04 from 1974Q1-1994Q4 when examining all negative net FDI flows and drops to -0.2 after 1995. Figure 6 depicts how the dummies capture the absolute expansion of net FDI. The dummy 95_𝑝𝑜𝑠𝑖 takes from 1995Q1 onwards the value 1 when 𝐹𝐷𝐼𝑖 is positive and the value 0 when 𝐹𝐷𝐼𝑖 is negative for each country 𝑖. The reverse holds for 95_𝑛𝑒𝑔𝑖.

Figure 5 FDI transition in the 90s Figure 6 The 90s-transition

captured by dummies

Note: FDI on the y-axis refers to the variable 𝐹𝐷𝐼𝑖 as described in Section 5.1 and represents real net FDI flows to

Germany from country 𝑖 measured in percentage of real GDP of country 𝑖. For visualization purposes the range of the y-axis is limited to values between -3 and 3%, complete data is available upon request. The median is calculated as the median for the seven countries in scope for i) all positive FDI flows from 1974Q1-1994Q4; ii) all negative FDI flows from 1974Q1-1994Q4; iii) all positive FDI flows from 1995Q1-1998Q4; iv) all negative FDI flows from 1995Q1-1998Q4.

17_{FDI during the 1990s rose by more than 500%, while world production only increased by around 27% (Waldkirch,}

2009). As the FDI variable in this study is normalized by GDP (𝐹𝐷𝐼𝑖 = 𝐹𝐷𝐼_𝑖𝑟𝑒𝑎𝑙

𝐺𝐷𝑃_𝑖 ∙ 100) the magnitude of 𝐹𝐷𝐼𝑖 is

affected.

18_{Although FDI increased continuously during the 90s, it was decided to include dummies from 1995Q1 onwards}

since two countries in scope joined the EU and the WTO was established leading to higher liberalization. However, the results in this study are robust and yield the same outcome when including a dummy from 1997Q1 onwards (results are available upon request).

-3 -2 -1 0 1 2 3 FDI -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 FDI

median for negative and positive net FDI for 1974Q1-1994Q4 and 1995Q1-1998Q4

95_𝑝𝑜𝑠

(22)

17

6 Results and analysis

In this section, the models presented in Section 4 are estimated to answer the research questions and, thus, determine how bilateral RERs have influenced German inward FDI and how this relationship was affected by the introduction of the euro. Stationarity tests are performed in Section 6.1, while Section 6.2, 6.3, and 6.4 estimate the OLS models over different time periods. Finally, Section 6.5 provides a robustness check by estimating two separate panels: one comprising the economies with a positive RER-FDI link and the other containing the countries with a negative RER-FDI link.

6.1 Stationarity tests

To be able to capture long-run effects of the bilateral RERs on FDI the variables in the OLS models are required to be stationary. Therefore, the property of stationarity and completeness of FDI and RER data sets served as selection criteria for the countries in scope of this thesis. Complete data sets for the period 1974-2018 are available for ten EU-countries: Austria, Denmark, Spain, Finland, France, Greece, Italy, the Netherlands, Sweden, and the United Kingdom. Stationarity tests for all variables have been conducted with the ADF test, the results for net inward FDI flows and bilateral RERs can be found in Table 7 in Appendix B.

The ADF test results reveal that at a 1%-level, the hypothesis of a unit root for all FDI series is rejected. Further the tests of the bilateral RERs reveal that all countries’ RERs are stationary except the ones of Finland, Greece, and the Netherlands. Therefore, the latter three countries are excluded from the scope of this thesis. As the RERs in the regressions are incorporated in logarithmic forms, the ADF test is also conducted for the logarithmic real exchange rates of Austria, France, Italy, Spain, Denmark, Sweden, and the United Kingdom. It appears that none of the series has a unit root in log-form, besides that of Spain; therefore, the RER-FDI link between Germany and Spain is examined using the RER in level form.19

The control variables of the seven selected countries are likewise tested for stationarity. The GDP differential (𝐺𝐷𝑃𝑑𝑖𝑓𝑓,𝑖) and unit labor costs differential (𝑈𝐿𝐶𝑑𝑖𝑓𝑓,𝑖) series are stationary for all countries. The interest rate-differential series (𝑖𝑑𝑖𝑓𝑓,𝑖) for Italy and the United Kingdom have unit roots. If these two variables are differenced once they achieve stationarity, thus, making it possible to include them in the equations. The last variable in scope is trade openness (𝑇𝑂𝑖). Since there are severe stationarity problems for several countries, this variable is included in first difference as in this form the null-hypothesis of a unit root can be rejected for every series.

19_{As the selection criteria of stationarity for the RER series of Finland, Greece, and the Netherlands was breached,}

(23)

18

6.2 Pre-euro period - estimation of OLS models

To determine how RERs between the seven EU countries in scope and Germany influenced German net FDI in the pre-euro period OLS models, as explained in Section 4.1, are estimated. The estimation results with a stepwise inclusion of the control variables are presented in Table 1. Table 1 OLS models - sample 1974Q1-1998Q4

Dependent variable: 𝑭𝑫𝑰𝒊

Variable Euro countries

Austria France Italy a _Spaina, b

𝐶 0.181 *** _0.187*** _0.136*** _0.122** _0.010* _0.007 _0.091* _0.094* (0.035) (0.035) (0.046) (0.047) (0.006) (0.006) (0.05) (0.055) 𝑙𝑜𝑔(𝑅(𝐷𝐸:𝑖)) -0.830 * -0.820 * -0.035 -0.297 -0.178 ** -0.194 ** -0.090 * -0.094 * (0.472) (0.469) (0.825) (0.862) (0.072) (0.074) (0.05) (0.056) 95_𝑝𝑜𝑠𝑖 0.484 *** 0.480 *** 0.323 *** 0.331 *** 0.114 *** 0.113 *** 0.144 *** 0.143 *** (0.091) (0.091) (0.113) (0.114) (0.031) (0.032) (0.031) (0.032) 95_𝑛𝑒𝑔𝑖 -1.431 *** -1.417 *** -0.766 *** -0.697 *** -0.104 * -0.100 * -0.045 ** -0.045 *** (0.18) (0.179) (0.179) (0.188) (0.053) (0.051) (0.018) (0.014) 𝐺𝐷𝑃𝑑𝑖𝑓𝑓,𝑖 0.001 _-0.012 0.003 0.002 (0.025) (0.039) (0.007) (0.003) 𝐷(𝑇𝑂𝐷𝐸) -0.016 -0.039 0.002 0.000 (0.025) (0.033) (0.004) (0.003) 𝐷(𝑇𝑂𝑖) -0.008 0.063 0.005 0.004 (0.02) (0.042) (0.005) (0.003) observations 92 92 100 100 94 94 96 96 R-squared 0.543 0.567 0.241 0.260 0.278 0.297 0.575 _0.585 Non-euro countries

Denmark a _Swedena _{United Kingdom}

𝐶 5.069 *** _5.321*** _3.302* _3.745** _{0.024 0.016} (1.543) (1.735) (1.72) (1.585) (0.09) (0.093) 𝑙𝑜𝑔(𝑅(𝐷𝐸:𝑖)) -2.458 *** -2.582 *** -1.570 * -1.789 ** -0.268 -0.298 (0.766) (0.859) (0.853) (0.785) (0.241) (0.249) 95_𝑝𝑜𝑠𝑖 0.557 * 0.538 * 0.838 *** 0.839 *** 0.742 *** 0.726 *** (0.287) (0.275) (0.205) (0.201) (0.125) (0.124) 95_𝑛𝑒𝑔𝑖 -0.481 ** -0.469 ** -0.585 ** -0.536 ** -0.184 -0.213 (0.23) (0.217) (0.258) (0.244) (0.184) (0.183) 𝐺𝐷𝑃𝑑𝑖𝑓𝑓,𝑖 0.022 0.024 0.059 ** (0.032) (0.022) (0.027) 𝐷(𝑇𝑂𝐷𝐸) -0.003 0.063 *** 0.002 (0.036) (0.016) (0.021) 𝐷(𝑇𝑂𝑖) 0.007 -0.020 * 0.001 (0.023) (0.011) (0.029) observations 96 96 92 92 94 94 R-squared 0.271 0.277 0.451 0.500 0.295 _0.333

Note: Standard errors in parentheses. ***, **, * denote significance at the 1%, 5%, and 10% level, respectively.

a_{HAC-standard errors}

(24)

19

Table 1 shows that the RERs have a negative and significant impact on net FDI flows from all countries except from France and the United Kingdom.20_{This implies that during the period} 1974Q1-1998Q4 an appreciating German currency reduced net FDI from Austria, Italy, Spain, Denmark, and Sweden. For example, a one percent increase in the Austrian-German bilateral RER decreased net FDI to Germany as a percentage of Austrian GDP by 0.83 percentage points. This finding is in line with the empirics of, for instance, Froot and Stein (1991). Moreover, the negative RER coefficients suggest, inferring from the theoretical framework by Chen et al. (2006), that the majority of firms investing in Germany were cost-oriented, exploiting cost differences in production between Germany and the home market.

The dummy variables 95_𝑝𝑜𝑠𝑖 and 95_𝑛𝑒𝑔𝑖, capturing the transition in the 1990s, show the expected signs and are significant for all economies besides for the United Kingdom. Hence, predominantly confirming the expected significant absolute increase in net FDI flows in the mid-90s.

Table 1, moreover, presents the outcome of a stepwise inclusion of the control variables. It can be observed that including more factors in the model does not alter the sign of the RER variables. However, the majority of control variables are insignificant.

A finding that is not in line with the empirics by Culem (1988) is the positive GDP differential of the United Kingdom, 𝐺𝐷𝑃𝑑𝑖𝑓𝑓,𝑈𝐾. It is significant at the 5%-level and shows that a higher quarterly real growth rate in the United Kingdom than in Germany increases net FDI flows to Germany. Possibly a higher growth rate in the United Kingdom is linked to a higher level of competitiveness compared to Germany or the United Kingdom exploits more moderate wage increases in Germany. If a higher GDP growth rate of the home country than of the destination country can have divergent effects (positive as found in this study and negative as proposed by Culem [1988]) then it is likely that this tends to cause a diverse overall effect, resulting in insignificant coefficients for 𝐺𝐷𝑃_{𝑑𝑖𝑓𝑓,𝑖}, as found in Table 1 for the majority of countries.

Furthermore, the first difference of trade openness, 𝐷(𝑇𝑂𝑖), is only significant for the Swedish economy. An increase in German trade openness, 𝐷(𝑇𝑂𝐷𝐸), has a positive effect on net FDI from Sweden. This result is in line with the predictions of Boateng et al. (2015), indicating that high exports and imports relative to GDP are linked to a business-friendly environment, fostering FDI. However, an increase in Swedish trade openness, 𝐷(𝑇𝑂𝑆𝐸), has a negative impact on net FDI, possibly a substitution effect between FDI and trade.

The models’ R-squares suggest that the regressors explain between 24.1% up to 58.5% of the variation in net FDI. Moreover, note that some estimated models suffered from heteroscedastic

(25)

20

and autocorrelated residuals and, therefore, the affected models are estimated with HAC estimators as explained in Section 4.1.21_{Furthermore, some models’ residuals are not normally distributed, a} finding which is not surprising when it comes to modelling financial time series (see, e.g., Mills, 1999). However, severe effects are not expected as the estimators are still unbiased (Gujarati & Porter, 2009).

6.3 Euro period - estimation of OLS models

OLS models are estimated for the period 1999Q1-2018Q2 to investigate the RER-FDI link in the euro-period. To examine the potential effect of the euro on this link the results are compared to the findings in the previous section. As explained in Section 4, it is not possible to capture the euro introduction by dummies. The OLS models’ results are presented in Table 2.

The pattern found in the previous section alters for the euro countries. A German RER appreciation has a positive impact on net FDI flows from France, Italy, and Spain during the euro period (the coefficients are significant on at least the 5%-level). Deriving from the theory by Chen et al. (2006) net FDI flows to Germany are primarily horizontal in the euro period. These results are in line with the empirics of Coeurdacier et al. (2009) that found an increase in horizontal FDI in the euro period evoked by the euro. The outcomes in this and the previous section suggest that the countries’ investment behavior shifted from vertical, in the pre-euro period, to horizontal FDI in the euro period. This contradicts the theory by Baldwin et al. (2008) that predicts a decline of horizontal FDI evoked by the euro.

Turning to the non-euro countries, it appears that the negative sign of the RER coefficients has not changed compared to the pre-euro period. A German RER appreciation decreases net FDI flows from Denmark, Sweden, and the United Kingdom. A difference to the pre-euro period, however, is that the magnitude of the coefficients has increased in absolute terms, suggesting that vertical activity relatively increased or that net FDI flows react more sensitive to changes in the RER.

Table 2, like Table 1, shows the stepwise inclusion of the control variables 𝐺𝐷𝑃𝑑𝑖𝑓𝑓,𝑖 and 𝐷(𝑇𝑂_𝑖) and variables available since 1999, the interest rate differentials, 𝑖𝑑𝑖𝑓𝑓,𝑖, and the unit labor costs differentials, 𝑈𝐿𝐶𝑑𝑖𝑓𝑓,𝑖. The control variables do not alter the sign of the RER coefficients, except the one from Austria which turns positive, but remains insignificant.

(26)

21

Table 2 OLS models - sample 1999Q1-2018Q2

Variable _{Euro countries}

Austria France Italy c _Spainb

𝐶 0.441 ** _0.508* _0.200*** _0.167*** _0.290** _0.509*** _{-1.248 -1.744}* (0.2) (0.279) (0.056) (0.062) (0.121) (0.151) (0.793) (1.011) 𝑙𝑜𝑔(𝑅(𝐷𝐸:𝑖)) -0.213 0.590 9.599 * 15.228 ** 1.725 * 3.617 *** 1.592 * 2.289 * (3.616) (3.969) (5.773) (6.901) (0.938) (1.226) (0.943) (1.22) 𝐺𝐷𝑃𝑑𝑖𝑓𝑓,𝑖 -0.219 0.138 -0.103 * -0.027 (0.233) (0.111) (0.054) (0.054) 𝑖𝑑𝑖𝑓𝑓,𝑖 0.005 -0.068 0.153 *** 0.060 ** (0.112) (0.062) (0.051) (0.026) 𝐷(𝑇𝑂𝐷𝐸) 0.066 -0.032 -0.005 0.032 (0.092) (0.044) (0.026) (0.033) 𝐷(𝑇𝑂𝑖) -0.021 0.067 0.017 -0.017 (0.067) (0.057) (0.038) (0.046) 𝑈𝐿𝐶𝑑𝑖𝑓𝑓,𝑖 -0.159 0.123 -0.016 0.022 (0.226) (0.097) (0.027) (0.035) observations 76 76 78 78 75 75 76 76 R-squared 0.000 0.028 0.035 0.136 0.044 0.215 0.037 0.139 Non-euro countries

Denmark a _Sweden _{United Kingdom}c

𝐶 15.111 *** _15.140** _6.455** _6.104** _-0.686* _-0.627 (5.513) (6.348) (2.902) (2.93) (0.376) (0.409) 𝑙𝑜𝑔(𝑅(𝐷𝐸:𝑖))) -7.730 *** -7.803 ** -2.770 ** -2.676 ** -2.665 ** -2.447 ** (2.858) (3.295) (1.278) (1.292) (1.052) (1.212) 𝐺𝐷𝑃𝑑𝑖𝑓𝑓,𝑖 -0.031 0.194 -0.142 (0.107) (0.126) (0.189) 𝑖𝑑𝑖𝑓𝑓,𝑖 -0.109 * -0.118 * 0.152 (0.057) (0.064) (0.188) 𝐷(𝑇𝑂𝐷𝐸) 0.051 0.033 -0.036 (0.057) (0.054) (0.059) 𝐷(𝑇𝑂𝑖) -0.033 0.004 -0.007 (0.05) (0.039) (0.091) 𝑈𝐿𝐶𝑑𝑖𝑓𝑓,𝑖 _-0.056 0.026 0.144 (0.079) (0.106) (0.103) observations 77 77 75 75 75 75 R-squared 0.064 0.125 0.059 0.194 0.081 0.147

b_{Real exchange rate estimated in level form}

c_{Interest rate differential estimated in first difference}

For most of the countries the variable 𝐺𝐷𝑃𝑑𝑖𝑓𝑓,𝑖 is insignificant, it only impacts net FDI flows from Italy to Germany. The negative coefficient is in line with the results in Culem’s (1988)

(27)

22

study and implies that when German GDP is growing faster than Italian GDP, net investment flows to Germany increase, ceteris paribus.

The interest rate differential, 𝑖𝑑𝑖𝑓𝑓,𝑖, is significant for four countries, showing ambiguous signs: the negative sign for the non-euro countries, Denmark and Sweden, implies that when the lagged real interest rate is higher in the respective country than the expected real interest rate in Germany at the point in time of investment, net FDI flows to Germany decrease. For the euro countries, Spain and Italy22_{, the positive interest rate differential signalizes the reverse effect. The} negative impact of the Danish and Swedish interest rate differential on net FDI is in line with theory and empirics (see, e.g., De Vita & Abbott, 2007). The negative sign suggests that the funding for FDI takes place in the home country (Denmark and Sweden) and therefore, FDI increases when the interest rate is lower in the home country. Contrarily, the positive sign of the interest rate differential for Spain indicates that FDI is funded in the destination country (Germany). This particularly is likely since it prevails high financial integration in the euro-area, low transaction costs, and no nominal exchange rate risk (see, e.g., Baldwin et al., 2008).

Both trade openness variables, 𝐷(𝑇𝑂𝐷𝐸) and 𝐷(𝑇𝑂𝑖), are insignificant for all economies. This finding might not surprise since the EU countries throughout the past decades have been increasingly integrating politically, financially, and trade wise, possibly resulting in that trade openness in a vastly interconnected union is not decisive anymore. Likewise, the unit labor costs differentials, 𝑈𝐿𝐶𝑑𝑖𝑓𝑓,𝑖, are not significant for any of the countries. Similar results have been found by Klein and Rosengren (1994), proposing that relative wages are not influential in the FDI decision.

In the regressions with significant RER variables, the R-squares range from 0.035 to 0.215. Hence, the goodness of fit measures are relatively low which might be a result of the great volatility of net inward FDI over time and potentially omitted variables. However, the aim of this study is not to determine all factors that affect net FDI flows to Germany but instead to determine whether the RER, a variable that appears to be relatively stable when including control variables in the models, affects net FDI.

In sum, the combination of the results of the euro countries in this and the previous section contradicts the theory by Baldwin et al. (2008) that proposes a decline in horizontal FDI evoked by the euro. Contrarily, the findings indicate that through the euro a change in the euro-firms’ investment behavior occurred, a shift from predominantly vertical FDI activity to predominantly horizontal FDI activity, a phenomenon henceforth in this thesis termed as the euro-effect. However, the RER-coefficients for France and the United Kingdom are insignificant in the

(28)

23

euro period and for Austria in the euro period, therefore, the indication of the euro-effect needs to be examined further in the next sections.

6.4 Entire period - estimation of OLS models

To test whether the results from the previous sections are stable, this section estimates OLS models over the entire period for countries which showed the same RER-FDI link in the pre-euro and

euro-period. The economies Denmark, Sweden, and the United Kingdom,23_{non-euro countries,}

showed a negative RER-FDI link in both investigated periods and are therefore subject to estimations from 1974 until 2018. The euro countries’ RER coefficients, however, altered in sign from the pre-euro period to the euro period, an effect possibly stemming from the euro introduction. To further test the euro-effect hypothesis, this section investigates the RER-FDI link for the non-euro countries which is not expected to be affected by the euro introduction and, thus, should show a negative RER-FDI link over the period 1974Q1-2018Q2. The results are presented in Table 3.

Table 3 OLS models - sample 1974Q1-2018Q2

Variable Denmark a _Sweden _{United Kingdom}

𝐶 3.598 *** _4.080*** _4.084*** _4.043*** _{-0.202 -0.198} (1.372) (1.392) (0.914) (0.903) (0.14) (0.143) 𝑙𝑜𝑔(𝑅(𝐷𝐸:𝑖)) -1.728 ** -1.965 *** -1.953 *** -1.934 *** -0.922 ** -0.921 ** (0.68) (0.69) (0.448) (0.442) (0.365) (0.371) 95_𝑝𝑜𝑠𝑖 0.360 *** 0.337 *** 0.873 *** 0.841 *** 0.525 *** 0.523 *** (0.103) (0.104) (0.12) (0.118) (0.1) (0.101) 95_𝑛𝑒𝑔𝑖 -0.592 *** -0.618 *** -0.186 -0.182 -0.464 *** -0.485 *** (0.075) (0.085) (0.13) (0.127) (0.12) (0.123) 𝐺𝐷𝑃𝑑𝑖𝑓𝑓,𝑖 0.038 0.023 0.056 (0.029) (0.026) (0.044) 𝐷(𝑇𝑂𝐷𝐸) 0.005 0.059 *** -0.005 (0.027) (0.021) (0.027) 𝐷(𝑇𝑂𝑖) -0.004 -0.019 0.004 (0.02) (0.014) (0.039) observations 173 173 167 167 169 169 R-squared 0.460 0.468 0.505 0.541 0.289 0.296

The estimation results found in the previous sections are verified when examining Table 3, the RER-FDI link is negative throughout the analyzed period for all countries (the RER

(29)

24

coefficients are significant at least at the 5%-level). Including the control variables in the models does not alter the results. Only one control variable is significant, namely 𝐷(𝑇𝑂𝐷𝐸), an increase in German trade openness has a positive impact on net inward FDI from Sweden. This is in line with the finding in Section 6.2 and the predictions by Boateng et al. (2015). The R-squares imply a relatively good fit of the models ranging from 0.289 to 0.541.

6.5 Robustness check - estimation of panel models

As a robustness check two panel OLS models are estimated, bundling the countries with the same sign of the RER coefficients. The previous sections showed that the RER-FDI link across the euro countries and across the non-euro countries was similar. Hence, two panels are estimated: one panel containing the non-euro countries while the other consists of the euro countries. Furthermore, the models are estimated for the pre-euro period and the euro-period to examine whether the indication of a euro-effect found in previous sections holds when bundling the countries. Lastly, both panels are estimated over the entire period and it is expected that if the euro-effect exists the RER coefficient is insignificant for the euro country panel.

Before estimating the models, the MW unit root test is conducted, the results can be found in Table 8 in Appendix B. The unit root tests show that it is possible to examine long-run relationships as the FDI, RER and control variables are stationary.

Table 4 presents the panel estimation results and reveals that the panel outcomes are in line with the findings in previous sections. The bilateral RERs for the euro-countries and non-euro countries are significant on at least the 5%-level for both periods, 1974Q1-1998Q4 and 1999Q1-2018Q2. The sign of the RER coefficients supports the findings in sections 6.2, 6.3, and 6.4. The sign is negative for the non-euro countries during both periods while the sign is negative for the euro countries in the pre-euro period and positive in the euro period, strengthening the hypothesis of a euro-effect.

Table 4 also depicts the output of the panel regressions for the entire investigated period, 1974Q1-2018Q2. The RER-FDI link for the non-euro countries is in line with results in Section 6.4, the RER coefficient is negative and significant at the 1%-level. However, the RER variable in the euro-countries’ panel estimation over the entire sample is insignificant, signalizing that the investment behavior in dependence of the RER was not constant over time. This strengthens the usage of a split sample to investigate the impact of the euro on the RER-FDI link.24

24_{Using, for example, dummies as approach would not capture the shift in sign for the RER as it can solely capture}

(30)

25

Table 4 Fixed effects panel estimation

Variable 1974Q1-1998Q4 pre-euro period 1999Q1-2018Q2 euro period 1974Q1-2018Q2 entire sample

euro non-euro euro d _non-euro _euro _non-euro

𝐶 0.065 *** _1.900*** _0.217*** _2.967** _0.064*** _0.571*** (0.009) (0.385) (0.025) (1.194) (0.01) (0.215) 𝑙𝑜𝑔(𝑅(𝐷𝐸:𝑖)) -0.210 ** -1.365 *** 0.721 *** -2.141 ** -0.135 -0.347 ** (0.087) (0.293) (0.212) (0.908) (0.085) (0.161) 95_𝑝𝑜𝑠𝑖 0.230 *** 0.563 *** 0.177 *** 0.397 *** (0.025) (0.069) (0.014) (0.038) 95_𝑛𝑒𝑔𝑖 -0.107 *** -0.512 *** -0.167 *** -0.529 *** (0.039) (0.093) (0.02) (0.044) 𝐺𝐷𝑃𝑑𝑖𝑓𝑓,𝑖 -0.003 0.025 * 0.000 0.073 -0.006 0.032 ** (0.007) (0.014) (0.024) (0.064) (0.007) (0.014) 𝐷(𝑇𝑂𝐷𝐸) -0.003 0.029 ** 0.012 0.020 0.001 0.022 ** (0.003) (0.013) (0.008) (0.025) (0.003) (0.01) 𝐷(𝑇𝑂𝑖) -0.001 -0.012 -0.008 0.006 -0.003 -0.009 (0.004) (0.01) (0.009) (0.022) (0.004) (0.009) 𝐷(𝑖𝑑𝑖𝑓𝑓,𝑖) 0.055 * -0.026 (0.028) (0.072) 𝑈𝐿𝐶𝑑𝑖𝑓𝑓,𝑖 0.009 -0.008 (0.018) (0.044) observations 377 281 251 219 637 505 R-squared 0.331 0.343 0.066 0.061 0.411 0.453 Note: Standard errors in parentheses. ***, **, * denote significance at the 1%, 5%, and 10% level, respectively.

d_{As it is expected that the euro-introduction enhanced homogeneity among the euro countries the model is estimated}

in a pooled regression.

The GDP differential, 𝐺𝐷𝑃𝑑𝑖𝑓𝑓,𝑛𝑜𝑛−𝑒𝑢𝑟𝑜, and German trade openness, 𝐷(𝑇𝑂𝐷𝐸), are positive and significant on at least the 10% level for the non-euro countries for the pre-euro period and the whole period. The positive GDP differential effect on net FDI was also found for the United Kingdom in Section 6.2. Moreover, it appeared in Sections 6.2 and 6.4 that an increase in German trade openness likewise positively impacts net FDI flows from Sweden.

The first difference of the interest rate differential, 𝐷(𝑖𝑑𝑖𝑓𝑓,𝑒𝑢𝑟𝑜), is positive and significant at the 10% level for the euro-countries in the euro period, consistent with the findings for Italy and Spain in Section 6.3. Further control variables are insignificant, and the R-squares range from 0.061 to 0.453, indicating a medium goodness of fit.

In summary, the panel estimation strengthens the findings of previous sections: that the RER has a negative impact on net FDI from the non-euro countries over the entire investigated period, and for the euro countries a negative impact in the pre-euro period and a positive impact in the euro period. Thus, supporting the existence of a euro-effect.

The euro effect – the impact of EU bilateral real exchange rates on German net FDI : evidence from Germany and seven EU-countries