The effect of Euro adoption on FDI flows
from the US
Evidence from Germany
Erkin Ilkay Özdemir
10583092
June 2016
Bachelor of Science Thesis
Thesis coordinator:
Dr. D.F. Damsma
Thesis supervisor:
Ms. E. Jakucionyte
University of Amsterdam
Faculty of Economics and Business
Specialization: Economics
Table of contents
1. Introduction 5
2. Literature review 7
2.1 FDI flows and macro-economic variables 7
2.2 Determinants of FDI flows 8
2.3 Currency union effects on FDI 10
2.4 FDI flows into Germany: trend and developments 11
3. Model specification 15
3.1 Main explanatory variables 15
4. Empirical analysis 18
5. Results 25
5.1 OLS regression results 25
5.2 IV regression 27
5.3 IV results 28
6. Conclusion 29
7. Reference list 30
Statement of originality
This document is written by Erkin Ilkay Özdemir, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business of the University of Amsterdam is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
In this paper, a time-series analysis is used in order to examine the effect of the Euro adoption on FDI flows into Germany. Using annual data from 1970-2013, an OLS regression is done. Multiple IV regressions have been applied in order to control for possible endogeneity. This gives no evident and significant results. The estimated results of the latter approach provide evidence that the euro adoption has a positive and significant effect on US FDI flows into Germany.
1. Introduction
Did the euro adoption in Germany have an effect on Foreign Direct Investment inflows from overseas? This paper answers the question by examining what happened to the US FDI flows into Germany. The data used in this research includes a time period from 1970-2013. Foreign direct investment is a category of cross-border investment made by a resident in one economy with the objective of establishing a lasting interest in an enterprise that is resident in an economy other than that of the direct investor (OECD, 2008).
Germany is an interesting country to investigate because Germany has the largest domestic market in Europe and has attracted an enormous amount of competitive and export-oriented multinational enterprises since the 1960s. The results should be of interest for both academicians who are investigating the impact of euro adoption on FDI flows, but also policy makers within Europe who are engaged in debates on the benefits and costs of adopting a common currency. Such debates can become more intense in the future, when European countries that planned to join the EU have to decide whether or not to adopt the euro as their national currency. For example Turkey, an EU-candidate, has to choose
between the Turkish Lira and the Euro.
Normally, an implementation of a common currency should have a positive effect on FDI flows. Main explanation for this is that a currency union increases intra EU-trade which can increase the demand for products produced by subsidiaries of US corporations. This can give those corporations an incentive to increase their direct investment in existing
subsidiaries or even establish new subsidiaries. Moreover, a common currency increases transparency of cost of capital. This means, that investors can make a better comparison between companies of different countries on how they are financed. The increased
transparency could increase the direct investment activities of those investors. Furthermore, a common currency could stabilize the exchange rate risk, which should give investors and multinational enterprises an incentive to invest more in a currency union. Aristotelous already performed such a study in 2005. He investigates the effect of European Monetary Union on US FDI flows into the EU by using panel data for the period in 1966-2003. His results shows that the adoption of the euro had positive and significant effect on US FDI flows into the countries that adopted the euro while there was no increase in US FDI flows into the countries which did not adopt the euro. The most important remark on the results of the study is the fact that the sample contains a period of five years in which the EMU was
in effect. This implies that the results cannot be considered as conclusive when making statements about the long-term impact of the euro adoption on US FDI flows. Therefore, the biggest contribution of this study is to test whether the results of Aristotelous (2005) also applies in the long-term because the period in which the euro has been adopted by Germany is 15 years. That is more than twice as long as the time dimension in Aristotelous (2005).
First, a review of the relevant literature on this subject will be presented. Thereafter, an econometric analysis will be done by performing an OLS regression in order to answer the research question. The dependent variable will be real FDI flows from US to Germany and the main explanatory variable will be a dummy variable which has a value of zero if Germany did not adopt the euro and one of otherwise. Time-series data may have a stochastic trend which can give misleading values regarding to hypothesis tests, confidence intervals and forecasts. In order to get rid of a stochastic trend, non-stationary data will be differenced. Therefore, the variables will be tested for non-stationarity by practising an Augmented Dickey-Fuller test in order to determine which of them have to be differenced. Thereafter, will the OLS regression be tested for heteroscedasticity and autocorrelation in order to check if the OLS estimates are BLUE. Also an IV regression will be done in order to tackle the
potential simultaneous causality between GDP growth of Germany and FDI flows. Finally, the findings based on the empirical analysis will be presented and followed by a conclusion. The results show that the euro adoption did have a significant positive effect on US FDI flows in Germany.
2. Literature Review
2.1 FDI flows and macro-economic variables
This section consists of an overview of existing literature about the relationship between FDI flows and economic growth. The purpose of this section is to emphasize the importance of studying the determinants of FDI.The largest part of the existing literature about the relationship between FDI and economic growth are based on cross-country analyses. Nevertheless, there have been done particular studies based on country-specific analyses. Most of the researches find a positive relationship between FDI and economic growth.
Country-specific studies such as Kokko (1994) find a positive effect of FDI on
economic growth in Mexico by using an OLS regression analysis. According to these studies, advanced competition between multinational companies and technology spill-overs caused this effect. Girma (2005) investigates the direct employment effect of foreign acquisitions in the U.K. Her results suggested that foreign acquisitions reduced labour-use inefficiency which was accompanied by negative as well as positive employment effects suggesting that foreign acquisitions can both create and destroy jobs in the short and medium run. However, despite the fact that foreign acquisition promotes efficiency, it also preserves jobs by helping enhance the long-run survival prospects of domestic establishments (Girma, 2005). The contention that FDI has a positive direct employment effect is further supported by a study of Becker and Muendler (2008) in which they conclude that multinational enterprises that expand abroad retain more domestic jobs than competitors without foreign expansion. They use a propensity-score matching approach in order to make expanding MNEs comparable to non-expanding firms. This effect between FDI flows and job retention is higher among highly educated workers.
Also some cross-country studies on the effect of FDI flows on economic growth exist. Borensztein et al. (1998) tests the effect of FDI on economic growth in a cross-country framework, utilizing data on FDI flows from industrial countries to 69 developing countries over the last two decades. Their results suggest that FDI works as an important driver for the transfer of technology, contributing relatively more to growth than domestic investment. But FDI contributes only to economic growth when sufficient capability of advanced
came to the conclusion that in developing countries in which domestic investment is stimulated, the effect of FDI on output growth is larger. Finally, in a meta-analysis of different studies, Ozturk (2007) came to the conclusion that consensus has been reached among academia and practitioners that FDI tends to have significant effect on economic growth through multiple channels such as capital formation, technology transfer and spill over, human capital enhancement, etcetera. As a result, there are several reasons to stimulate FDI flows. First, it works as a vehicle for technology transfers which stimulates economic growth. Secondly, FDI flows reduce labour-use inefficiency and retain jobs.
2.2 Determinants of FDI flows
This section provides insights on the determinants of FDI. Again both country-specific as well as cross-country researches have been done in order to examine the determinants of FDI. Ang (2007) investigates possible determinants with the help of annual time series data for the period 1960-2005 in Malaysia. He finds a positive relationship between real GDP and FDI inflows. Besides real GDP, there’s a small but positive impact of GDP growth on inward FDI. According to Ang (2007) FDI flows gets stimulated when trade openness, financial and infrastructure development improves. But higher corporate taxes and appreciation of the real exchange rate have a negative effect on FDI flows. Finally, he comes to the conclusion that FDI inflows increase with higher macroeconomic uncertainty. This seems paradoxal because normally, investors should be discouraged when the returns are more risky. A possible explanation may be the fact that foreign investors perceive more uncertainty as greater potential investment return which can lead to a shift towards more speculative type of foreign investment. A research in Turkey shows similar results with previous literature: host country’s market size, openness to foreign trade and the infrastructure have a positive effect on FDI inflows but there is a negative relationship between exchange rate volatility of the domestic currency and FDI inflows (Erdal et al, 2002). On the contrary, Wakelin et al. (2002) investigate the effect of the level of the exchange rate volatility in the exchange rate and exchange rate expectations on both outward and inward FDI flows in 12 developed countries and to USA between 1983 1995. They find no empirical evidence for an effect of exchange rate volatility on either US outward investment or inward investment. Chen et al. (2006) also study the impact of exchange rate movements on FDI flows. By practicing a real
options model on industry panel data on Taiwan’s outward FDI into China, the empirical results shows that exchange rate volatility has a significant negative effect on Taiwan’s outward FDI flows into China. The results of Chen et al. (2006) contradict with the results of Cushman (1985). Cushman examines the effect of real exchange rate risk by pooling
estimation results on FDI flows of US to five industrialized countries. According to his results, exchange risk has a positive effect on FDI flows. Conclusively, the effect of exchange rate volatility on FDI is currently ambiguous according the developed theories. Shamshudding (1994) uses a single-equation econometric model for 36 less developed countries for the year 1983 in order to examine the economic determinants of FDI. According to his results, the size of the market of the host country measured by GDP per capita is the most important factor in attracting FDI. Besides market size, cost factor such as wage cost and the
investment climate in the host country are important variables, which influence FDI. Furthermore, a panel data study on the determinants of FDI flows in the Gulf Cooperation Council (GCC) countries shows that oil potential and oil utilization have a negative effect on FDI flows (Mina, 2007). But, a positive relationship between the relative amount of oil utilization, measured by oil production relative to oil reserves, and FDI inflows exists. An econometric analysis of US outward FDI flows based on multinational enterprises for the period 1971-1989 shows that the most important factors in the investment decision are market size, factor costs, labour and capital (Barrell et al, 1996). Egger et al. (2001) performed a research in order to analyse distance as determinant of exports and FDI flows within a three factors New Trade Theory model assuming that distance affects both trade costs and plant set-up costs. The results of a Hausman-Taylor SUR model shows that distance affects outward FDI flows negatively. Finally, Janicki et al. (2004) examined the determinants of bilateral FDI flows between the EU members and eight central and east European candidate countries. The results from the cross-section data in 1997 showed similarities with other studies. The key determinants are again the economy size, country risk, labour costs and openness to trade of the host country.
Different Regional Integration Agreements could also affect the FDI flows from US to Germany in the period 1970-2013. A paper investigates the effects of economic integration on the size and character of FDI flows (Balasubramanyam et al, 2002). The impact of RIA’s on FDI flows and their welfare effects were investigated by using 381 bilateral FDI flows in 1995. The results imply that RIAs do not determine the size and direction of FDI flows but rather
the economic characteristics (including income, population, market friendliness and
distance) of both the source and host country. Worth (1998) performs a research existing of two parts. First, he examines the determinants of FDI for manufacturing industries in general and specifically for agricultural industries. The second part examines how RTA’s affect FDI determinants. He comes to the conclusion that FDI goes to countries with a higher GDP growth rate. According to his results, the greatest effect of RTA’s on FDI is through their effect on GDP and market size (Worth, 1998). Finally, Sethi et al. (2003) examine possible changing trends in the location of FDI flows and the determinants of FDI resulting from a macro-economic and firm strategies view. An OLS regression model is used to develop a statistical analysis of abroad investments by US multinational enterprises. The results show that US MNE’s increase their investment in Asian countries in order to profit from the lower wage levels and entry new emerging markets. The shift of FDI flows from developed markets to emerging markets could have a deflating effect on US FDI flows to Germany.
2.3 Currency union effects on FDI
No many studies have been done on the effects of common currency on FDI. General consensus from the existing literature is that currency unions have a positive effect on FDI flows. Before mentioning studies, which examined this effect, an explanation why currency unions affect FDI flows should be provided. A positive relationship is expected because currency unions expand trade between member countries, which also increases demands for products of subsidiaries of MNEs. Therefore, US corporations will most likely increase their direct investment in their subsidiaries in order to meet the increase in demand for products and services (Aristotelous, 2005). The studies of Rose (2001) and Glick and Rose (2002) find a positive and statistically significant effect of currency unions on trade. This result is
accompanied by a study of Micco et al (2003) in which they examined the impact of EMU on bilateral intra-EMU trade and find a positive and significant impact. It can be concluded that there is enough evidence to expect a positive relationship between currency unions and FDI flows.
Petroulas (2004) examines the effect of the European Monetary Union on inward FDI flows by using a difference-in-difference approach and fixed effects with common time controls. The results show that both inward FDI flows into the Euro-area and FDI flows to and from non-member countries become larger due to the introduction of the Euro. Aristotelous et al. (1996) practices a similar method by using a cross-section, time-series
approach to study the determinants of FDI from the US and Japan in the EU. By using annual data from 1980 to early 1990s he finds strong evidence on an increase in FDI flows in a barrier free European market. According to Aristotelous et al. (1996) FDI flows in the EU also depend on market size and the real exchange rate. The study of Schiavo (2007) agrees roughly with the findings of Aristotelous. Schiavo uses a gravity model on a sample of OECD countries in order to investigate the impact of EMU on FDI flows and concludes from his results the hypothesis that besides the fact that a common currency eliminates exchange rate volatility, currency unions have a positive impact on FDI flows. The most important study for this research is the study of Aristotelous (2005) in which he investigates the effect of EMU on US FDI flows into the European Union by using panel data from fifteen EU countries for the period 1966-2003. The empirical findings show that the euro adoption has a positive effect on US FDI flows into countries that adopted the euro. The contribution of this study is to examine the long-term effect of euro adoption on US FDI flows into Germany.
2.4 FDI flows into Germany: trend and developments
This section gives an overall overview on trends and developments on FDI flows into Germany and is obtained from a country profile study on inward FDI flows in Germany and its policy context (Jost, 2010).
The main reason for the increase of FDI flows to Germany in the 1960s is the successful reintegration of Germany into the world economy after World War II. The European unification process also contributed to this increase. Many large multinational enterprises worldwide had established affiliates in Germany. After the German reunification in 1990, the total FDI stock amounted to 111 billion US dollars and eventually has risen six-fold since then. The most attractive determinants for foreign MNEs were the size of the German market, the competitiveness of the German corporate sector translated into efficient suppliers, high quality infrastructure, skilled labour force, trade openness and low financing costs on German capital markets. But there was also criticism on the investment environment. For example, high marginal taxes, high wages and relative inflexible and the overregulated labour market were regarded as harmful to investing in Germany.
Furthermore, low foreign investment in East Germany after reunification was criticized because it led to more inequality in terms of inward FDI flows compared with West
the East German wages adjusted rapidly to the West German level after reunification (Jost, 2010) despite low labour productivity. Secondly, the deindustrialization process gave MNEs the opportunity to supply the East German economy via their West German affiliates.
Around 65% of total inward FDI during the past decade is concentrated in the service sector. Important reasons for this development are privatization and liberalization in the telecommunication sector and in the electricity, gas and water supply sector. Roughly one third of FDI in Germany was accounted by the manufacturing sector. Many big MNEs (most from the US) invested in production facilities, distribution and service centers in Germany and therefore contributed to the rebuilding and reintegration of Germany into the world economy by transmitting capital and technology after World War II. In recent years, MNEs still continued to expand their activity in Germany but, now via undertaking cross-border mergers and acquisitions.
The global economic and financial crisis had several effects on FDI flows. First of all, FDI flows to Germany declined in 2008 by 68%. Net equity capital investments halved, reinvested earnings became negative and the amount of net lending of foreign MNEs to their subsidiaries in Germany declined to 1.5 billion US dollars. However, compared to many other developed economies, FDI flows to Germany started to increase much earlier, namely in 2009 despite the fact that in 2009 the recession of the Germany was at its worst level. The reason for this increase was the general improvement of the business environment starting in the second quarter of 2009.
Figure 1
*FDI flows from OECD countries to Germany in billions of US dollars at constant prices *Data obtained from OECD database
As illustrated in the graph, there is a strong decline in FDI flows to Germany in 2008 from OECD countries confirming the study of Jost (2010). Moreover, the total FDI flows started to increase strongly in 2009 showing that despite the financial crisis, Germany still was an attractive location for foreign direct investment. However, FDI flows declined again in 2012. This decline could be caused by increased economic uncertainty which led transnational corporations in developed countries to postpone their new investments or even divest in foreign assets (Unctad Report, 2013).
-5E+10 0 5E+10 1E+11 1.5E+11 2E+11 2.5E+11 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 FD I fl ow s year
Figure 2
*Real FDI flows in millions of US dollars * Data obtained from U.S. Bureau of Economic Analysis database
It can be concluded that overall FDI flows to Germany decreased in 2008 and the graph in figure 2 confirms the same decline. There is a strong decrease in 2011, which is striking, because overall FDI flows from OECD partner countries decreased in 2012.
-1E+10 -5E+09 0 5E+09 1E+10 1.5E+10 2E+10 2.5E+10 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 FD I fl ow s year
3. Model specification
3.1 Dependent variable and main explanatory variables
There have been developed many theories drawn from different areas of economics and business in order to explain the determinants of FDI. These theories emphasized both demand-related and supply-related determinants of FDI. Consequently, an econometric model which incorporates both supply-related and demand-related determinants of FDI has been used. The demand-related determinants in this model are: GDP and GDP growth. The supply-related factors are: relative unit labour costs, real exchange rate and exchange rate volatility. This study examines the effect of euro implementation on FDI flows by estimating
the following model:
𝐹𝐹𝐹𝐹𝐹𝐹𝑡𝑡 = ẞ0 + ẞ1𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐹𝐹𝐺𝐺𝑡𝑡 + ẞ2∆𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐹𝐹𝐺𝐺𝑡𝑡 + ẞ3𝑈𝑈𝑈𝑈𝐺𝐺𝐹𝐹𝐺𝐺𝑡𝑡 + ẞ4∆𝑈𝑈𝑈𝑈𝐺𝐺𝐹𝐹𝐺𝐺𝑡𝑡 + ẞ5𝐺𝐺𝑈𝑈𝑅𝑅𝑅𝑅𝑡𝑡 +
ẞ6𝐺𝐺𝐺𝐺𝐺𝐺𝑡𝑡 + ẞ7 𝑉𝑉𝑉𝑉𝑅𝑅𝑡𝑡 + ẞ8𝑈𝑈𝑈𝑈𝐺𝐺𝑅𝑅𝑉𝑉𝐺𝐺𝐹𝐹𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝑡𝑡 + ẞ9𝐺𝐺𝑈𝑈𝐺𝐺𝑡𝑡 + Ɛ
T refers to the period from 1970-2013 and Ɛ is the error term. The dependent variable,𝐹𝐹𝐹𝐹𝐹𝐹𝑡𝑡,
is the real FDI flows from US to Germany measured in US dollars. The real FDI flows are calculated by dividing nominal FDI flows by the US GDP deflator. The first explanatory variable in the equation, 𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐹𝐹𝐺𝐺𝑡𝑡 , is the real GDP of Germany measured in dollars at
constant 2005 prices and captures the effect of the market size (total income and
production) on the investment decision of US companies. The sign is expected to be positive because the larger the host country’s total income and its potential for development the more FDI get stimulated. Moreover, a large market is important for efficient utilization of resources and exploitation of economies of scale, which improves the investment climate for transnational companies. The second variable, ∆𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐹𝐹𝐺𝐺𝑡𝑡, is the real GDP growth of
Germany and measures the effect of the change in aggregate demand in Germany on its FDI inflows. An increase in demand should lead to an increase in FDI flows according to the acceleration principle, which implies that an increase in demand for consumer goods will cause an increase with a larger magnitude in the demand for machines and other investment necessary to make these goods (Aristotelous, 2005). Therefore, a positive sign is expected. US real GDP is the third variable and the same sign as the German real GDP coefficient is expected. The intuition behind this is that a higher income level should give investors more opportunities in terms of amount of money to invest abroad. The study of Aristotelous et al. (1996) confirms a positive relationship between FDI flows and GDP, which measures market
size. This also holds for the fourth variable, real GDP growth of the US. So both the level and the growth of GDP are included because many existing literature confirm a positive
relationship between real GDP and FDI flows. The motivation to include the growth of GDP as a separate variable is based on the study of Ang (2007) in which he finds a small positive effect of GDP growth on FDI flows. GDP growth is a measure and signal of increasing market demand, which should attract FDI. The fifth variable,𝐺𝐺𝑈𝑈𝑅𝑅𝑅𝑅𝑡𝑡, captures the effect of the
difference in relative unit labour costs between the US and Germany. In this study, the variable is formulated in the following way: the relative unit labour costs of the US divided by the relative unit labour costs of Germany. According to empirical studies, higher unit labour costs in the US relative to Germany should have a positive effect on the FDI flows from the US to Germany. Lower labour costs in Germany makes it more attractive for foreign companies to set up a subsidiary in Germany. The sixth variable,𝐺𝐺𝐺𝐺𝐺𝐺𝑡𝑡, captures the effect of
the real exchange rate on FDI flows. The real exchange rate is defined in the following way: the amount of German marks/euros needed in order to buy one US dollar. In this case, the real exchange rate is calculated by the nominal exchange rate multiplied with the ratio of the GDP deflators of both the US and Germany. So an increase in the exchange rate means that more German marks/euros are needed in order to buy one US dollar, which implies a depreciation of the German Mark (since 2002, the euro). The depreciation of the German Mark/euro increases the relative wealth of US companies and therefore results in a lower cost of capital and thus an increase in FDI flows to Germany. Therefore the sign of the coefficient is expected to be positive. The seventh variable, 𝑉𝑉𝑉𝑉𝑅𝑅𝑡𝑡, captures the exchange
rate volatility between the US dollar and mark/euro and is measured by the moving standard deviation of the growth of the exchange rate.
𝑉𝑉𝑉𝑉𝑅𝑅𝑖𝑖𝑡𝑡 = [(𝑚𝑚) �(ln 𝑄𝑄1 𝑡𝑡+1−1− ln 𝑄𝑄𝑡𝑡+1−2)2]1/2 𝑚𝑚
𝑖𝑖=1
Q is the monthly nominal exchange rate and m, the order of the moving average, is set equal to 12. Exchange rate volatility generates uncertainty as the variance of expected profits increases. This could cause investors to hesitate on investing abroad. On the other hand, risk seeking investors could perceive more uncertainty as greater potential return and therefore increase their investment. As mentioned before in the literature review, there is no
eighth variable of the equation, 𝐺𝐺𝑅𝑅𝑉𝑉𝐺𝐺𝐹𝐹𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝑡𝑡, is the KOF Index of Globalization and
measures the three dimensions of globalization, namely economic, social and political aspects. The motivation to use the general globalization index as a variable instead of just economic globalization index is based on the fact that one of the components are trade and FDI flows which could imply simultaneous causality since the dependent variable is FDI flows. The Globalization Index captures the effect of globalisation on US FDI flows into Germany. The sign is expected to be positive because globalization leads to more activities abroad of US corporations which translates into an increase in FDI flows. The final variable, 𝐺𝐺𝑈𝑈𝐺𝐺𝑡𝑡 is a dummy variable which has a value of one since Germany adopted the euro, and a
value of zero since the beginning of the dataset till the euro adoption. Table 1
List of variables
𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐹𝐹𝐺𝐺𝑡𝑡 Real GDP of Germany at time t
𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐹𝐹𝐺𝐺𝐺𝐺𝐺𝐺𝑡𝑡 GDP growth of Germany at time t
𝑈𝑈𝑈𝑈𝐺𝐺𝐹𝐹𝐺𝐺𝑡𝑡 Real GDP of US at time t
𝑈𝑈𝑈𝑈𝐺𝐺𝐹𝐹𝐺𝐺𝐺𝐺𝐺𝐺𝑡𝑡 GDP growth of US at time t
𝐺𝐺𝑈𝑈𝑅𝑅𝑅𝑅𝑡𝑡 Relative unit labour costs at time t
𝐺𝐺𝐺𝐺𝐺𝐺𝑡𝑡 Real exchange rate at time t
𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝑅𝑅𝑉𝑉𝐺𝐺𝐹𝐹𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝑡𝑡 Globalization index for Germany at time t
𝑉𝑉𝑉𝑉𝑅𝑅𝑡𝑡 Exchange rate volatility at time t
𝑈𝑈𝑈𝑈𝐺𝐺𝑅𝑅𝑉𝑉𝐺𝐺𝐹𝐹𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝑡𝑡 Globalization index for US at time t
𝐺𝐺𝑈𝑈𝐺𝐺𝑡𝑡 Dummy variable: has a value of 1, since the euro is adopted at time t
There are is a possible threat in obtaining BLUE OLS estimators. Several studies mentioned in the literature review find a significant positive effect of FDI flows on GDP growth. This would lead to simultaneous causality, because the dependent variable “FDI” could affect the explanatory variables GDP and GDP growth of Germany. This causality would result in biased coefficients.
4. Empirical analysis
The following section will give insights into data analysis. First, a correlation matrix will be presented. The second part exists of stationarity check for the variables. The stationarity check determines which variables have to be adjusted if they are non-stationary. After the stationarity check, the model will be tested for autocorrelation and heteroscedasticity. All of this will be done with Stata 13.
Table 2
Correlation matrix
GERGDP ∆GERGDP USGDP ∆USGDP RULC RER USGlob GERGLOB Vol FDI
GERGDP 1 ∆GERPGDP -0.2539 1 USGDP 0.9868 -0.2748 1 ∆USGDP -0.2368 0.5156 -0.2342 1 RULC -0.1685 0.0211 -0.0955 0.1350 1 RER 0.9038 -0.2289 0.8842 -0.3324 -0.4346 1 USGLOB 0.9512 -0.2378 0.9201 -0.1499 -0.1774 0.8456 1 GERGLOB 0.9708 -0.3269 0.9628 -0.2214 -0.1109 0.8215 0.9484 1 VOL -0.1941 0.0423 -0.1912 0.2269 0.0241 -0.1608 -0.1283 -0.2071 1 FDI 0.1752 -0.0324 0.1887 --0.0247 -0.0360 0.1531 0.1895 0.2004 0.0918 1
The correlation between USGlob and GERGlob is very high (0.9484). If both variables are added to the regression it can lead to multicollinearity. Therefore, the variable which lead to higher adjusted 𝐺𝐺2 will be kept in the regression. The regression results in Table 5
show that the inclusion of GERGlob results in a higher𝐺𝐺2, therefore GERGlob will be kept.
Furthermore, there is a high correlation between US GDP and Germany GDP.
A variable is stationary if the probability distribution of the variable does not change over time. This implies that the mean and variance of those variables stays constant over time (Stock & Watson, 2012). It seems that often, time series on macroeconomic data includes a stochastic trend because movements in macroeconomic time series data are in the most cases the consequences of economic forces, which have a large unpredictable or random component. A stochastic trend implies a random trend that varies over time and
indicates nonstationary variables (Stock & Watson, 2012). OLS tests, if run on nonstationary variables, can be misleading regarding to hypothesis testings, confidence intervals and forecasts. A way to deal with stochastic trends or so-called unit root is to difference a nonstationary variable. In that case the variable can become stationary. Thereafter, the OLS regression can be performed. Before doing an ADF test, the number of lags must be
determined in order to allow the ADF formulation for higher order auto-regressive
processes. If too few lags are included, then potentially valuable information will be omitted. On the other hand, too many lags could introduce additional estimation errors into the forecasts (Stock & Watson, 2012). The two most commonly used methods for choosing the optimal amount of p orders of an auto regression are the Bayesian information criterion (BIC) and the Akaike information criterion (AIC). Generally, the BIC method underestimates the number of lags whereas AIC overestimates the number of lags. Both information criteria helps determining precisely how large the increase in 𝐺𝐺2 must be in order to justify the
additional lag. Finally, the optimal lag is indicated by the lowest value of the BIC or AIC. Table 3
Optimal Lag
Variable AIC AIC opt lag BIC BIC opt lag
FDI 48.1335 1 48.2317 1 GERGDP 52.3574 3 52.4652 1 ∆GERGDP 4.35782 0 4.40452 0 USGDP 54.6158 2 54.7546 2 ∆USGDP 3.99359 1 4.0861 1 RULC -2.25611 2 -2.11733 2 RER -2.00767 12 -1.8115 1 USGLOB 2.95762 1 3.05014 1 GERGLOB 3.90244 1 3.99496 1 VOL -5.0645 0 -5.01824 0
The optimal lag length from both the AIC and BIC is reported in Table 3. The listed optimal lags correspond to the minimum value of the BIC and AIC. BIC and AIC does not give always
the same amount of optimal lags. For example, GERGDP has an optimal leg length of 3 according to AIC but 1 according to BIC. In case of differences, the optimal leg length of AIC will be followed because many studies conclude that the ADF test works better with too many legs than too few. As mentioned earlier, AIC overestimates the amount of lags and BIC underestimates it.
As stated before, the variables must be tested for stationarity in order to get reliable results from OLS tests. The Augmented Dickey-Fuller test (ADF) will be used in order to test the variables for stationarity. The null hypothesis of the ADF test states that the variable in concern has a stochastic trend which implies a non-stationary variable. The alternative is that the variable does not have a stochastic trend, so that it is stationary. In Table 3, both the level and the difference of the variables are included for a stationarity test because it gives the opportunity to check whether a variable, which is non-stationary at level, becomes stationary if the difference is used. If the difference is significant implying a stationary variable, then the variable in differences can still be used for OLS estimates.
Table 4
ADF test results
Variable Level Difference
FDI -5.129*** -7.726*** GERGDP -0.488 -5.982*** ∆GERGDP -5.222*** Na USGDP 0.558 -4.306*** ∆USGDP -4.464*** Na RULC -3.421** Na RER -1.428 -5.036*** USGLOB -1.738 -6.029*** GERGLOB -1.342 -6.184*** VOL -6.894*** Na
• The lags of AIC are used with the ADF test because studies shows that ADF tests works better with too many lags than too few.
• *. Statistically significant at 10% level.
• **. Statistically significant at 5% level.
• ***. Statistically significant at 1% level.
Table 4 shows that most of the variables are non-stationary, so they have a stochastic trend. Stationary variables are: FDI, Germany and US GDP growth, RULC and VOL. It is obvious that the GDP growth for both countries is stationary because a growth variable already takes a difference between current value and the value one year earlier (∆GERGDP = 𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐹𝐹𝐺𝐺𝑡𝑡−
𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐹𝐹𝐺𝐺𝑡𝑡−1). The results also show that all the nonstationary variables can become
stationary when differenced because there is no insurance that the difference of a variable has no stochastic trend. The stationarity test is not done for the variable “euroad” which has the value of 1 since Germany has adopted the euro, and 0 if otherwise. The results from Table 4 imply that a modification of the main OLS regression is necessary in order to get more reliable hypothesis tests. The variables that are stationary remain unchanged. The remaining variables should be first-differenced which removes the possible stochastic trends from those variables. Also a lagged value of the change in FDI will be included in order to control for the persistence of a stochastic trend in the dependent variable. The definition of the variables changes slightly. A ∆ sign will be added to the variables which are
first-differenced. When GDP of the US and Germany gets differenced, they almost measure the same time-series as GDP growth. Therefore, in the remainder of the analysis, the GDP growth variables of both US and Germany will be excluded in order to avoid
multicollinearity.
3) 𝐹𝐹𝐹𝐹𝐹𝐹𝑡𝑡 = ẞ0+ ß1∆𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐹𝐹𝐺𝐺𝑡𝑡 + ß2∆𝑈𝑈𝑈𝑈𝐺𝐺𝐹𝐹𝐺𝐺𝑡𝑡 + ß5𝐺𝐺𝑈𝑈𝑅𝑅𝑅𝑅𝑡𝑡 + ß6∆𝐺𝐺𝐺𝐺𝐺𝐺𝑡𝑡 +
Table 5
OLS Regression
(1) (2) (3) (4)
FDI FDI FDI FDI
∆GERGDP -0.0190 -0.0224 -0.0224 -0.0225
(-1.25) (-1.49) (-1.51) (-1.51)
∆USGDP 0.0112* 0.0109* 0.0109* 0.0108*
(2.47) (2.61) (2.65) (2.61)
RULC -2.94617e+09 -2.86601e+09 -2.89494e+09 (-0.58) (-0.59) (-0.61)
∆RER 2.29415e+10* 2.57093e+10* 2.55446e+10** 2.67733e+10**
(2.40) (2.68) (2.85) (2.87)
∆USGlobindex 244740233.7 (0.28)
VOL -7.61318e+09 -2.26178e+09 -4.94917e+09
(-0.18) (-0.05) (-0.12)
euroad 3.20517e+09 3.73030e+09* 3.74973e+09* 3.52663e+09*
(1.81) (2.22) (2.31) (2.16)
LagFDI -0.367* -0.380** -0.379** -0.368**
(-2.69) (-2.85) (-2.89) (-2.82)
∆GERGlobindex 618238725.6 621261887.7 608181099.3
(1.30) (1.34) (1.30)
constant 4.93895e+09 4.37129e+09 4.34001e+09 1.53261e+09
(0.94) (0.86) (0.87) (0.88) N 43 43 43 43 𝐺𝐺2 0.4413 0.4666 0.4666 0.4611 Adjusted 𝐺𝐺2 0.3098 0.3411 0.3599 0.3533 F-statistic 3.36 3.72 4.37 4.28 t statistics in parentheses * p<0.05 ** p<0.01 *** p<0.001
Table 5 presents 3 OLS regressions. First, FDI is regressed against main explanatory variables except ∆GERGlobindex. In the second regression, ∆USGlobindex is replaced by
∆GERGlobindex as an explanatory variable in order to check whether it explains more of the variance of FDI. Both 𝐺𝐺2 and adjusted 𝐺𝐺2 increased meaning that equation 2 fits the data set
better. In the third regression, the variable “VOL” is omitted due to its non-significance. The variable “RULC” is omitted in the fourth regression. Omitting “VOL” and “RULC” leads to a
small increase in the adjusted𝐺𝐺2, but this increase is negligible in both cases. Therefore,
further empirical analysis will focus on the second OLS regression. Several tests will be performed for regression 2 with FDI flows as the dependent variable. After the test for stationarity, there are also other important tests before doing OLS regression. First, a test on homoscedasticity must be done in order to check whether the variance of the error is
related to the size of the explanatory variables. The most commonly used test on
homoscedasticity is the White test. The null hypothesis states that the variance of the error term is constant against the alternative which states a non-constant variance of the error term.
𝐻𝐻0: 𝑣𝑣𝐿𝐿𝑣𝑣 𝑢𝑢𝑖𝑖 = 1 and 𝐻𝐻1: 𝑣𝑣𝐿𝐿𝑣𝑣 𝑢𝑢𝑖𝑖 ≠ 1
The next step is to perform a White test. The results can be seen in the Appendix. With a p value of 0.1787 will the null hypothesis not be rejected. So there is not enough evidence to state that the error terms are heteroscedastic. The purpose of a test on autocorrelation is to check whether the residuals are related to residuals from the past. There also exist many tests for autocorrelation such as Durbin-Watson (D-W) and Breusch-Godfrey (B-G) test. For this research, first the Breusch-Godfrey test will be used and the results can be found in the Appendix. With a p value of 0.2632 is the test statistic not significant at 5% significance level. Also the Durbin-Watson test will be included in order to check if it aligns with the result of the B-G test. The null hypothesis states that there is no autocorrelation. The dU value is the upper critical value implying that if the d-statistic is greater than dU, the null hypothesis will not be rejected. With a d-statistic of 2.153897, the value is higher than the dU value (1.704), so the null hypothesis will not be rejected. This means that the residuals are not related to residuals from the past. Therefore, in the following analysis, it will be assumed that there is no autocorrelation. The conclusion from the White test, and both Durbin-Watson and Breusch-Godfrey test imply that the Gauss-Markov theorem holds for this particular OLS regression. Gauss-Markov theorem exists of four conditions namely:
1. The error term 𝑢𝑢𝑡𝑡 has a conditional mean of zero.
2. The variables are independent and identically distributed. 3. Large outliers are unlikely.
In other words, the OLS estimators are the most efficient linear conditionally unbiased estimators (BLUE) (Stock & Watson, 2012). Due to the fact that the OLS estimators are BLUE, it can be assumed that the OLS estimators are unbiased consistent and efficient.
5. Results
5.1 OLS regression results
Some regressions have been done in order to compare different set of explanatory variables. These results can be found in table 5. The explanatory variables GDP growth of both countries has been dropped from the regression. The main reason for this exclusion is that when the variable “GDP” becomes differenced, it measures almost the same effect as GDP growth. Including both variables would cause multicollinearity which leads to large variances of those variables. Also, a lagged value of the dependent variable “FDI” has been included in order to control for a possible stochastic trend. The 𝐺𝐺2 for this model is 0.4666
and this means that 46.66% of the variance of the sample variance of real FDI flows is
explained by the model. The drawback of using 𝐺𝐺2 is that 𝐺𝐺2 almost always increases when a
variable is added but adding a new variable does not always have to improve the fit of the model (Stock & Watson, 2012). Therefore it is more useful to look at the adjusted 𝐺𝐺2 that
does not necessarily increase when a new variable is added. The F (8, 34) statistic is 3.72 with (p< 0.001) test the joint hypothesis that all the slope coefficients are zero. The variable USGlob has been dropped from the regression due to high correlation with GERGlob. This can be seen in the correlation matrix. The motivation behind the choice to keep GERGlob instead of USGlob is that it explains more the variance of the dependent variable “FDI” measured by adjusted 𝐺𝐺2 as can be seen in Table 5. Now, the estimated coefficients will be
discussed. First of all, the results show that GDP growth in Germany has a negative non-significant effect on FDI flows. Regarding to earlier studies, this result seems weird because other studies find a positive relationship between market size and FDI flows. The GDP growth of US has a positive and significant impact on FDI flows. This positive relationship agrees with the results of earlier studies such as the research of Ang (2007). The next
variable, “RULC”, has a negative but non-significant impact on FDI flows. Its non-significance and negative sign does not align with earlier studies. According to the conclusions of
Shamshudding (1994) and Barrel et al. (1996) there is a significant relationship between FDI flows and labour costs. The real exchange rate has positive and significant impact on FDI flows. So when the exchange rate increases, the German Mark/euro depreciates. This depreciation gives US investors and MNAs an incentive to increase their investment in Germany. The Germany Globalization index has a positive but non-significant effect on FDI flows suggesting that in this model, it is not an important determinant of US FDI inflows into Germany. The effect of exchange rate volatility is negative but not significant. As mentioned earlier, there is no consensus about the effect of exchange rate volatility on FDI flows and this result shows that in this model, exchange rate volatility is not an important determinant. The coefficient estimate for the Euro dummy variable is the primary focus of this research. The coefficient is found to be positive as expected and statistically significant. This result does match with the conclusions of earlier studies. As mentioned before, all the studies done on the effect of EMU on FDI flows reach the same conclusions regarding to the effect of euro adoption on FDI flows: this effect is found to be positive and significant.
Limitations
Despite the fact that non-stationary variables have been changed into stationary variables by performing an ADF test, there are still some limitations, which have to be taken into account when interpreting the results. First of all, the model could be subject to
endogeneity. This means that FDI flows could also affect GDP growth, implying simultaneous causality. This causal relationship has been proved in the study of Kokko (1994).
Simultaneous causality could lead to a bias in the coefficients. One way to solve this problem is to do IV or TSLS regression. Furthermore, several studies such as Ang (2007) and Erdal et al. (2002) included more significant determinants on FDI flows such as the host country’s infrastructure, trade openness and taxation, which are not included in this research. This also may cause omitted variable bias. All of these mentioned points could lead to model misspecification. Finally, in contrast to the study of Aristotelous (2005), this research only used evidence from Germany. So it is a premature to generalize the conclusions of this empirical analysis to other countries, which adopted the euro.
5.2 IV Regression
The model could be subject to endogeneity. This means that FDI might affect
∆GERGDP implying simultaneous causality. A possible explanation behind this relationship is that FDI flows could stimulate positive spill-overs such as the transfer of technology, job retention and the introduction of capital goods which should have a positive effect on GDP. This could lead to a bias in the coefficients. An IV regression could solve this problem. In order to perform an IV regression, an appropriate instrument has to be found. This
instrument should meet two conditions. First of all, an instrument must be exogenous. That means that the instrument must not correlate with the error term of the original regression (Stock and Watson, 2012). Secondly, an instrument must be relevant implying that it has to correlate with the endogenous variable. If more instrumental variables than endogenous variables are used, than the IV regression transforms into TSLS regression. A combination of multiple instruments makes the estimation results more efficient and precise. Therefore more instruments will be included in this research.
In this analysis, the instrument variable should correlate with ∆GERGDP and should not correlate with FDI. Furthermore, there is no reason to assume that the correlation between USGDP, USGDPGR, RULC, RER, VOL, euroad and the error term is nonzero. Therefore, these variables are assumed to be exogenous and are included in the model because they are correlated with the endogenous variable “∆GERGDP”. In order to do IV regression, additional instrumental variables are needed. These variables must be correlated with ∆GERGDP but uncorrelated with the error term. Thus, these instrumental variables must not affect FDI directly because if they do, they should be included in the original regression equation. The first instrumental variable is trade of Germany. “Trade” is defined as the sum of exports and imports of goods and services measured as a share of GDP. According to Frankel et al. (1999), there is a positive relationship between trade and GDP growth. An attractive regulatory environment for trade should stimulate economies of scale and productivity within the German economy which is export-orientated. Furthermore, life expectancy and education spending are used as instrumental variables. If education
spending increases, this could lead to more high educated workers and increase the chance that a high educated person is getting hired and earns more. Increase in life expectation suggests that people live longer and spend more money which should has a positive effect on GDP.
5.3 IV results
Three IV regressions have been done in Table 9. Two tests are conducted in order to check if the instrumental variables are appropriate. These tests can also be found in Table 9. First of all, the relevance of the instruments is tested. The weakness of the instruments can be measured by the F-statistic in the first stage IV regression. If the F-statistic has a value larger than 10, the instrumental variables are assumed to be relevant. With a F-statistic of 1.83019 which is lower than 10, “trade” is a weak instrument in the first regression. There is no sufficient correlation with GDP growth. This can also be seen from the partial 𝐺𝐺2 which
has a value of 0.0511. The second test measures instrument exogeneity (over identification). The null hypothesis states that the instrument set is valid and the model is correctly
specified. The results are based on the Sargan-Hansen J- test. With a p value of 0.1912 the null hypothesis cannot be rejected. Due to weak instruments, it can be concluded that the coefficients in regression 1 are not consistent (Stock & Watson, 2012). The same
complications apply for the remaining regressions in which life expectation (regression 2) and education spending (regression 3) are added as extra instrumental variables. All the instrumental variables are assumed to be exogenous but weak according to the F-statistics and J-tests. The weakness of the instruments could translate into biased an inconsistent coefficients. For example, in both regression 2 and 3 are the variables “Lag∆FDI”, “∆RER” and “∆USGDP” not significant, which is in contrast with the results obtained from OLS regression. Moreover, ∆USGDP has a negative sign in regression 3 but a positive sign in the OLS
regression. It can be concluded that the instruments are weak and do not correlate
sufficiently with ∆GERGDP despite the fact that the null hypothesis on exogeneity cannot be rejected. The weakness of the instruments shows that the coefficients are inconsistent and probably biased what results in unreliable confidence intervals and hypothesis tests.
6. Conclusion
Several studies show that there is enough evidence to state that the adoption of a common currency is an important determinant of FDI flows. Main reason for this statement is that a common currency increases trade between currency union countries, which
increases demand for products of subsidiaries of multinational enterprises (MNEs). This increased demand gives MNEs more incentive to invest more in their existing subsidiaries or even build new subsidiaries. In this research, the relationship between euro adoption in Germany and the US FDI flows to Germany has been investigated by using data from the period 1970-2013. A literature review of existing studies has been done in order to provide background information and determine which variables have to be included in the empirical analysis. An OLS regression is used to perform the empirical analysis. Several tests have been used in order to check if the OLS estimators meet the Gauss-Markov assumptions. The outcome states that the euro adoption in Germany has a positive significant effect on US FDI flows into Germany. As a result the euro adoption can be seen as an important determinant. However, when interpreting the results, several limitations should be taken into account. First of all, there may be simultaneous causality between the variables “FDI flows” and “GDP growth” which can lead to biased coefficients. An IV regression is applied to control for this possible simultaneous causality but no significant and clear results were obtained.
Furthermore, according to earlier studies, there are more possible determinants of FDI flows such as infrastructure and taxes, which are not included in this model. This could lead to OVB. Finally, the results of this empirical analysis cannot be generalized to other countries, which adopted the euro. Despite the fact that this empirical research cannot be seen as a perfect study, it provides several basic insights on the relationship between FDI flows, GDP growth, relative unit labour costs and real exchange rate. Further research can take away the limitations of this empirical analysis. For example, an IV regression with more appropriate instrumental variables can be performed in order to tackle the endogeneity problem. Also, extra explanatory variables, which measures trade openness, infrastructure and taxation of the host country could be added to avoid omitted variable bias. Finally, a panel data analysis can be done for countries, which adopted the euro in order to check whether the
7. Reference list
Ang, J. B. (2008). Determinants of foreign direct investment in Malaysia. Journal of policy
modelling, 30(1), 185-189.
Aristotelous, K. (2005). Do currency unions affect foreign direct investment? Evidence from US FDI flows into the European Union. Economic Issues, 10(2), 1-10.
Aristotelous, K., & Fountas, S. (1996). An Empirical Analysis of Inward Foreign Direct Investment Flows in the EU with Emphasis on the Market Enlargement Hypothesis*. JCMS: Journal of Common Market Studies, 34(4), 571-583. Balasubramanyam, V. N., Sapsford, D., & Griffiths, D. (2002). Regional integration
agreements and foreign direct investment: Theory and preliminary evidence. The
Manchester School, 70(3), 460-482.
Barrell, R., & Pain, N. (1996). An econometric analysis of US foreign direct investment. The
review of economics and statistics, 200-207.
Becker, S. O., & Muendler, M. A. (2008). The effect of FDI on job security. The BE Journal of
Economic Analysis & Policy, 8(1).
Blonigen, B. A. (2005). A review of the empirical literature on FDI determinants. Atlantic
Economic Journal, 33(4), 383-403.
Borensztein, E., De Gregorio, J., & Lee, J. W. (1998). How does foreign direct investment affect economic growth? Journal of international Economics, 45(1), 115-135. Chen, K. M., Rau, H. H., & Lin, C. C. (2006). The impact of exchange rate movements on
foreign direct investment: market-oriented versus cost-oriented. The Developing
Economies, 44(3), 269-287.
Cushman, D. O. (1985). Real exchange rate risk, expectations, and the level of direct investment. The Review of Economics and Statistics, 297-308.
De Sousa, J., & Lochard, J. (2011). Does the Single Currency Affect Foreign Direct Investment?*. The Scandinavian Journal of Economics, 113(3), 553-578.
De Mello Jr, L. R. (1997). Foreign direct investment in developing countries and growth: A selective survey. The Journal of Development Studies, 34(1), 1-34.
Egger, P., & Pfaffermayr, M. (2004). Distance, trade and FDI: a Hausman–Taylor SUR approach. Journal of Applied Econometrics, 19(2), 227-246.
Eller, M., Haiss, P., & Steiner, K. (2006). Foreign direct investment in the financial sector and economic growth in Central and Eastern Europe: The crucial role of the efficiency channel. Emerging Markets review, 7(4), 300-319.
Erdal, F., & Tatoglu, E. (2002). Locational Determinants of Foreign Direct Investment in an Emerging Market Economy: Evidence from Tukey. Multinational business review, 10, 21-27.
Flam, H., & Nordström, H. (2007). The euro and Single Market impact on trade and
FDI. Manuscript, Institute for International Economic Studies, Stockholm University. Frankel, J. A., & Romer, D. (1999). Does trade cause growth? American economic review,
379-399.
Görg, H., & Wakelin, K. (2002). The impact of exchange rate volatility on US direct investment. The Manchester School, 70(3), 380-397.
Girma, S. (2005). Safeguarding Jobs? Acquisition FDI and Employment Dynamics in U.K. Manufacturing. Review of World Economics, 141(1), 165-178.
Glick, R., & Rose, A. K. (2002). Does a currency union affect trade? The time-series evidence. European Economic Review, 46(6), 1125-1151.
Janicki, H. P., & Wunnava, P. V. (2004). Determinants of foreign direct investment: empirical evidence from EU accession candidates. Applied Economics, 36(5), 505-509.
Jost, T. (2010). Inward FDI in Germany and its policy context.
Kokko, A. (1994). Technology, market characteristics, and spill overs. Journal of
development economics, 43(2), 279-293.
Micco, A., Stein, E., & Ordoñez, G. (2003). The currency union effect on trade: early evidence from EMU. Economic policy, 18(37), 315-356.
Mina, W. (2007). The location determinants of FDI in the GCC countries. Journal of
Ozturk, I. (2007). Foreign direct investment-growth nexus: a review of the recent literature. International Journal of Applied Econometrics and Quantitative
Studies, 4(2).
Petroulas, P. (2007). The effect of the euro on foreign direct investment. European Economic
Review, 51(6), 1468-1491.
Schiavo, S. (2007). Common currencies and FDI flows. Oxford Economic Papers, 59(3), 536-560.
Sethi, D., Guisinger, S. E., Phelan, S. E., & Berg, D. M. (2003). Trends in foreign direct investment flows: A theoretical and empirical analysis. Journal of international
business studies, 34(4), 315-326.
Shamsuddin, A. F. (1994). Economic determinants of foreign direct investment in less developed countries. The Pakistan Development Review, 41-51.
Xu, B., & Wang, J. (2000). Trade, FDI, and international technology diffusion. Journal of
Economic Integration, 585-601.
Ramb, F., & Weichenrieder, A. J. (2005). Taxes and the financial structure of German inward FDI. Review of World Economics, 141(4), 670-692.
Rose, A. K., & Van Wincoop, E. (2001). National money as a barrier to international trade: The real case for currency union. The American Economic Review, 91(2), 386-390. Unctad (2016). Press Release annual report2013.
http://unctad.org/en/pages/PressRelease.aspx?OriginalVersionID=143 U.S. Bureau of Economic Analysis. Retrieved 20 May 2016 from:
http://www.bea.gov/
Worth, T. (1998). Regional trade agreements and foreign direct investment. Regional Trade
Agreements and US Agriculture, (771).
World Bank (2016). World Bank Indicators. Retrieved 20 May 2016 from: http://data.worldbank.org/indicator?display=default
8. Appendix
Table 6
Descriptive statistics
Variable Obs Mean Std. Dev. Min Max
GERGDP 44 2310000000000.00 571000000000.00 1370000000000.00 3180000000000.00 ∆GERGDP 43 2.002329 2.055044 -5.61886 5.255006 USGDP 44 9010000000000.00 3280000000000.00 4340000000000.00 14500000000000.00 ∆USGDP 44 2.864046 2.073872 -2.77553 7.259087 RULC 44 1.036916 0.1696943 0.761506 1.553473 RER 44 0.9015765 0.3111238 0.3462609 1.472552 USGLOB 44 70.24591 6.548392 59.05 78.17 GERGLOB 44 66.98136 11.57379 46 81.92 FDI 44 3810000000 5520000000 -7320000000 23100000000 VOL 44 0.0228988 0.0184057 0 0.0876223 Table 7 Test on heteroscedasticity: White’s test
𝑅𝑅ℎ𝑖𝑖2 𝑣𝑣𝐿𝐿𝑣𝑣𝑢𝑢𝐺𝐺 1.81 (0.1787)
• The errors are homoscedastic under the null hypothesis and heteroscedastic under the alternative hypothesis.
• P-values are in parenthesis
• ** statistically significant at 5% level
Table 8
Test on autocorrelation: Breusch-Godfrey test and Durbin Watson test
𝑅𝑅ℎ𝑖𝑖2 𝑣𝑣𝐿𝐿𝑣𝑣𝑢𝑢𝐺𝐺 1.252 (0.2632)
d-statistic 2.153897
• There is no auto-correlation under the null hypothesis and auto-correlation under the alternative hypothesis.
• P-values are in parenthesis
Table 9
TSLS regression
(1) (2) (3)
FDI FDI FDI
∆GERGDP -0.0718 -0.271 0.298
(-1.06) (-0.36) (0.27)
∆USGDP 0.0131* 0.0218 -0.00301
(2.54) (0.62) (-0.06)
RULC -2.71139e+09 -2.08613e+09 -3.86839e+09
(-0.55) (-0.16) (-0.23)
∆RER 3.26742e+10* 6.08390e+10 -1.94422e+10
(2.42) (0.55) (-0.12)
VOL -1.28356e+10 -5.55938e+10 6.62848e+10
(-0.29) (-0.28) (0.24)
euroad 3.45721e+09* 2.35290e+09 5.50065e+09
(1.97) (0.38) (0.66)
LagFDI -0.419** -0.577 -0.126
(-2.88) (-0.82) (-0.13)
∆GERGlobindex 942519657.9 2.25384e+09 -1.48397e+09
(1.45) (0.44) (-0.20)
constant 5.87908e+09 1.19763e+10 -5.40322e+09
(1.06) (0.44) (-0.14) N 43 43 43 partial 𝐺𝐺2 0.0511 0.0028 0.0021 F-statistic 1.83019 0.095138 0.073034 J-test 9.95428 5.31978 4.43231 p value 0.1912 0.621 0.7289 t statistics in parentheses * p<0.05 ** p < 0.01 • T- ratios are in parenthesis
• The instrument variables are relevant under the null hypothesis if the F statistic is larger than 10 • * statistically significant at 5% level
• The instrument set is valid and the model is correctly specified under the null hypothesis. The reported value is the J-statistic with the p-value in parenthesis
