The effect of the European Single Market on trade activity : an analysis of the Dutch economy

(1)

Faculty of Economics and Business

Bachelor’s thesis

The effect of the European Single Market on trade

activity: An analysis of the Dutch economy

By Jerrel King

BSc Economie & Bedrijfskunde Specialisation: Economics Academic year: 2015-2016 Student number: 10528466 Supervisor: Oana Furtuna

(2)

ABSTRACT

This thesis investigates the effect of the European Single Market on Dutch trade activity. This Single Market is an advanced form of a preferential trading agreement that originated in the 1950s. It has evolved into a complex interconnected framework over the years. Most preferential trading agreements are being formed to ensure growth in countries’ trade activities. However, it is unclear what the actual effects of preferential trading agreements on trade are. The underlying model of this research question is the (augmented) gravity model of international trade. The effect of the European Single Market on Dutch trade activity is determined by means of applying fixed effects regression. Bilateral panel data for the years 1988 through 2014 is used in the dataset. The findings show that European Single Market-membership has a positive effect on overall Dutch trade activity and Dutch exports. Dutch imports, however, are unaffected.

(3)

Statement of Originality

This document is written by Student Jerrel King who declares to take full responsibility for the contents of this document.

I declare that the text and 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 is responsible solely for the supervision of completion of the work, not for the contents.

(4)

interconnectedness is the increase in trade over the years. Figure 1 shows how trade in real terms has evolved over a 16-year time span, up to the financial crisis of 2009. At the beginning of this century, the total value of goods traded in the world amounted $939,7 constant billion dollars. In 2008, before the crisis, this amount had more than doubled to $1892,3 constant billion dollars.

Figure 1 - Exponential trade growth adjusted for seasonality, 1998-2009 (constant billions of dollars, base year=2009)

Source: OECD insights; with data converted to constant US$

Numerous world events in general have led countries to consider integration of their economies and/or policies. The aftermath of the Second World War in particular gave six countries1 the incentive to unite and preserve order, welfare, and economic stability. These countries agreed to the Treaty of Paris (1951) and the Treaty of Rome (1957), establishing the European Coal and Steel Community (ECSC) and European Economic Community (EEC) respectively. These

1_{Belgium, France, Germany, Italy, Luxembourg, the Netherlands}

500 650 800 950 1 100 1 250 1 400 1 550 1 700 1 850 2 000

(6)

communities abandoned previous trade restrictions between their members, and thus promoted free trade characterized by zero import tariffs. They have been acknowledged as the first pillars of what is currently known as the European Union.

The European Union was formally established in 1993 with the signing of the Maastricht Treaty. Furthermore, the Maastricht Treaty introduced criteria for the formation of an official Economic Monetary Union with a common currency. This Economic Monetary Union was fully established in 1999 and harmonized the union members’ monetary and economic policies. Consequently, the euro was introduced as a common currency. An important characteristic of the European Union is that it functions as a Single Market. The European Single Market is an

advanced stage of economic integration where its members have abandoned their internal trade restrictions, exert a common external tariff and support the free movement of goods, services, capital and people.

However, the European Union is not the only trading block in the world (albeit in an advanced stage). Multiple countries have established some form of preferential trade agreement with each other, most of them being either free trade agreements or customs unions2_{. The most} notable ones are the NAFTA and MERCOSUR. The former is a free trade agreement within North America, the latter a customs union in Latin America. Furthermore, there are countries currently negotiating terms of potential prospective trade agreements. One of those being the Transatlantic Trade Investment Partnership (TTIP), which concerns a free trade agreement between the current two largest economies in the world: The United States of America and Europe. Given the fact that this potential comprehensive agreement concerns the two largest economies in the world, this could affect other countries and the world as a whole.

So why should countries engage in preferential trade agreements or rather economic integration in the first place? It is stated that free trade can stimulate economic growth. Samuelson (1962) argued that free trade could generate positive welfare due to the increased consumption possibilities and comparative advantages within countries. Proponents of

globalization therefore often use this statement as their main argument when pleading for further economic integration. But are trade levels significantly affected by economic integration? In this thesis this question will be researched and applied to one of the original founders of the ECSC

2_{Within the framework of economic integration custom unions are more advanced than free trade agreements, but less advanced} than common markets. They exert no internal tariffs between members and common external tariffs to non-members.

(7)

and EEC: the Netherlands. This thesis will address the issue whether or not the European Single Market positively affected Dutch trade activity.

The research question will be answered using the gravity model of trade. Chapter II provides a discussion of the previous literature and research regarding the effects of (European) economic integration on trade levels. Chapter III elaborately discusses the methods used in this thesis. The gravity model equation –which lays the foundation of this research– and its use within this thesis is also explained. Consequently, the application of the model with regard to panel data regression is discussed. In chapter IV the data(sets) are described. Subsequently, the results gained from this data will be explained in chapter V. Finally, chapter VI provides the overall conclusion of this thesis.

II. Literature review

As noted by Brada and Mendez (1985), the gravity equation has generally been used in two ways. The first includes adding dummy variables to the gravity model equation to measure the effect of integration. Using bilateral cross-section data the equation can subsequently be estimated for several years. The second entails estimating the gravity equation for a pre-integration period, and predicting future trade levels using these estimations. Then, the actual observations can be compared to these estimations. This chapter discusses the various

approaches and results of the gravity model –regarding (European) integration– used by researchers thus far. Furthermore, these approaches will be compared and contrasted with the approach followed in this thesis.

Aitken (1973) has been one of the first researchers to use the gravity model to determine the effects of European integration on trade. Aitken conducted research to determine the effect of the European Free Trade Association (EFTA) and the European Economic Community (EEC) on trade levels within theses communities. The aim of this research was to establish whether the formation of these free trading communities lead to trade creation3, using the following augmentation of the gravity equation:

(8)

log 𝑋_!" = log 𝑏_! + 𝑏_!log 𝐷_!"+ 𝑏_!log 𝑌_! + 𝑏_!log 𝑌_!+ 𝑏_!log 𝑁_! + 𝑏_!log 𝑁_!+ 𝑏_!log 𝐴_!" + 𝑏_!log 𝑃_!"!!"_{+ 𝑏}

!log 𝑃!"!"#$+ log 𝑒!" (1)

, where 𝑋!" is the dollar value of country i's exports to country j, 𝑌! (𝑌!) is the GDP value of country i (j), 𝐷_!" is the distance between the economic centres of i and j, 𝑁_! (𝑁_!) is the

population of country i (j), 𝐴_!" is a dummy variable accounting for neighbouring countries, and

𝑃_!"!!"_{, 𝑃}

!"!"#$ are dummy variables for trade between partners of the EEC and EFTA respectively.

Aitken gathered bilateral cross-section data for the years 1951 through 1967, for 12 countries. This dataset contained data before integration in 1959, as well as after integration. Aitken estimated the effects of the EEC and EFTA on trade for each year separately. His results showed moderate positive effects of both trading communities on trade. Specifically, the effect of the EEC on trade showed insignificance at a .05-significance level until 1961. The effect of the EFTA follows a similar trend and becomes significant at a .05-significance level in 1964. Furthermore, all other coefficients of the gravity equation were significant and consistent at .05-significance for every year.

Brada and Mendez (1985) applied a similar approach. However, their research was different since they assumed heterogeneity amongst countries. Brada and Mendez maintained a sceptic view about the extent of the positive effect of the EEC and EFTA on trade, and aimed to solve for additional (country-specific) factors that could attribute to trade creation. First, they extended their research by also determining the effects of integration for developing countries, entailing a more heterogeneous sample. This interest came from the view that the effect of

integration could depend on the extent of development of the integrating country. For this reason, they researched the effect of four4 additional preferential trading agreements, other than the EEC and EFTA. Second, Brada and Mendez emphasized the environmental influences of preferential trading agreements that could also affect trade. Specifically, they argued that countries close to each other should experience a greater post-integration stimulus to mutual trade than integrating countries far from each other. In order to control for these factors, they included two interaction

4_{Council for Mutual Economic Assistance (CMEA); Central American Common Market (CACM); Latin America Free Trade} Agreement (LAFTA); Andean Pact

(9)

variables in their revised gravity model5: 𝑃_!"log !!

!!

!! and 𝑃!"log 𝐷!", where 𝑃!" is a dummy

representing trade between partners of one of the six preferential trading agreements considered,

!!

!! and

!!

!! are the GDPs per capita for countries i and j which measure development, and 𝐷!"

is distance between countries i and j. Brada and Mendez took the same econometric approach as Aitken. They estimated the gravity equation for every year separately using bilateral cross-section data. Their results showed significance for the preferential trading agreements at a .10-significance level for the three years6 reported. These years were after the integration period. This was in line with the overall trade theory and the findings of Aitken. Furthermore, they did find significant effects for environmental factors. Inter-member distance resulted in being the most important environmental factor with significance at a .01-significance level for every year reported. The effect of development on trading agreements declined over time and became insignificant at a .10-significance level after the final year reported.

However, these positive effects of integration on trade are not the norm. Frankel, Stein and Wu (1995) also conducted research to determine the effects of several preferential trading agreements. Similar to the approach of Aitken, Brada and Mendez, they gathered bilateral cross-section data. In contrast to the research discussed above, they extended their research by

including a larger sample size. Using the gravity model, they determined that the effect of the EFTA on trade never showed significant results for the years 1965 through 1990. These results are in stark contrast with the findings of Aitken, Brada and Mendez. Despite these findings, Frankel et al. did find a significant result for the European Community (former EEC), arguing that trade within this community was three times higher than would have been without its existence.

As can be referred from above, there seems to be mixed results and no clear answer as to whether the certain trading communities are trade creating or not. The methods described above all apply bilateral cross-section data to the gravity model. However, a more recent study by Baier et. al (2007) takes a different approach by applying bilateral panel data. By taking panel data (for

5_{The revised equation used by Brada and Mendez: log 𝑋}

!!= A + 𝛼!log 𝑌!+ 𝛼!log 𝑌!+ 𝛼!log 𝑁!+ 𝛼!log 𝑁!+ 𝛼!log 𝐷!"+ 𝛽log 𝑄!"+ 𝛾!𝑃!"log (!_!!

!)( !!

!!) + 𝛾!𝑃!"log 𝐷!"+ log 𝑒!"

(10)

every five years), they added a time dimension to the gravity model. Trade values in the dataset would be measured with regard to differences over the years, instead of for each year separately. In addition, Baier et. al extended the view of Brada and Mendez regarding heterogeneity. They noted that there was unobserved heterogeneity. These are the unobserved components of the effect of preferential trading agreements on trade, which are not included in the model. It is for this reason that they applied panel data regression, in particular fixed effects regression. Fixed effects regressions can solve for unobserved heterogeneity given that this heterogeneity is constant over time. An elaboration on this technique and its assumptions follow in chapter III!_. Baier et. al also noted that the implementation of panel data –in contrast to cross-section data– had a large impact on their results. Specifically, when controlling for endogeneity bias using fixed effects regression, the effect of a free trade agreement yielded seven times the effect that was estimated by OLS.

Regarding the effect of European integration on trade levels, this thesis tries to expand research by focussing on a more advanced and more recent stage of European integration than those discussed above: the European Single Market. Unlike the literature discussed, this thesis focuses on one reporting country and its trading partner instead of multiple reporting countries. This approach will be used because it encompasses a more narrow view of the subject at hand. Moreover, this thesis will pursue the more recently applied approach by Baier et. al. to assess the effect of the European Single Market on Dutch trade activity. This involves the use of bilateral panel data instead of cross-section data to estimate the effects on trade. In particular, fixed effects regressions can be used which implicitly control for constant (country-specific)

unobserved heterogeneity while cross-section regressions do not. However, instead of estimating the gravity equation for each year separately, this thesis will use a single estimation for the years 1988 through 2014. This period represents years before and after integration. Based on theory, this thesis will provide an augmentation of the gravity model that involves other variables than those discussed above. These variables will be discussed in chapter III!_.

(11)

𝐈𝐈𝐈𝐚_{. Methodology}

As mentioned in the first chapter, this chapter provides a detailed discussion of the methods used underlying this thesis. The (augmented) gravity model of trade will be used to approach the research question. The gravity model –which has its foundation in physics due to Newton’s law of gravity– explains gravitational pull between economic entities. As can be referred from chapter II, numerous researches have used this model to determine the effect of various forms of economic integration on trade due to its explanatory power. This is the main reason why this model will be used as foundation of this thesis. However, the most common problem when applying a regression to this model is to isolate the effect of integration on trade. This chapter provides a discussion of the method used to mitigate this problem.

Tinbergen was the first researcher to implement the economic approach to the gravity model in 1962. Subsequently, Bergstrand (1985) provided the first theoretical economic foundation for the model. He noted that bilateral trade flows are generally explained by the following specification:

𝑃𝑋!" = 𝛽! 𝑌! !! 𝑌! !! 𝐷!" !! 𝐴!" !!𝑢!" (2)

Here, 𝑃𝑋_!" is the value of bilateral trade flow from country i to country j, 𝑌_! (𝑌_!) is the GDP value of country i (j), 𝐷_!" is the distance between the economic centres of i and j, 𝐴_!" is any other factor(s) affecting trade between i and j, and 𝑢!" is a log-normally distributed error term with [𝐸 𝑙𝑛 𝑢_!" = 0]7_{. For econometric purposes, this specification of the gravity model of trade can} be transformed to a linear form by means of applying natural logarithms to both sides of the equation. This linear form will be used to estimate the effect of the European Single Market on Dutch trade activity. In addition, all variables will be observed over time. Taking all of the above into account, the gravity model will be specified as follows:

ln 𝑃𝑋!"# = 𝛽!+ 𝛽!ln 𝑌!" + 𝛽!ln 𝑌!" + 𝛽!ln 𝐷!" + 𝛽!ln 𝐴!"# + ln 𝑢!"# (3)

7_{This result is necessary to ensure internal validity when applying a regression: implying that there are no sources of bias in the} model. This result is also one of the four conditions necessary to be able to assume an asymptotically normal distribution (convenient upon testing coefficients)

(12)

This equation is a modified version of equation 2, since this equation –in contrast to equation 2– accounts for time.

According to the gravity model theory it is expected that countries’ GDPs have a positive effect on trade activity, whereas distance has a negative effect. Furthermore, instead of only focussing on total trade volumes, this thesis will also provide an analysis on imports and exports separately. Several regression models will be constructed using different variants of equation 3. Panel data must be used in order to model the variables measured over time. This entails data consisting of multiple countries measured over multiple time periods. This equation can then be estimated by a regression with panel data. In this thesis fixed effects regression will be used.

𝐈𝐈𝐈𝐛_{. Fixed effects regression}

In this thesis fixed effects regression is used to estimate the effect of the European Single Market on Dutch trade levels. A mathematical elaboration on this method is provided in Appendix 1. Fixed effects regression is a method used for panel data. It controls for any variables that are omitted from the equation that vary between entities, but remain constant over time. These time-invariant variables do not cause any changes to the dependent variable over time, since we assumed them to remain constant. It is important to note that fixed effects regression implicitly controls for these variables. By applying fixed effects regression to the gravity model, the fixed effects estimator can be determined after which it is possible to assess the effect of the European Single Market on Dutch trade. An elaboration on the assessment of variables follows in chapter III!_.

There is a vast amount of unobserved variables imaginable that vary across countries, determine trade levels but remain constant over time. For instance, countries’ view towards international trade –established through historical events– could determine the overall trade activity. Or, perhaps there are country-specific regimes or laws that have an influence on trade. These are merely a few examples of variables that are hard to quantify, presumed to remain constant over time, but cannot easily be accounted for within a regression model. For this reason, this thesis applies fixed effects regression. This also coincides with the findings of Baier and Bergstrand (2007). Baier and Bergstrand researched the effects of several free trade agreements

(13)

on trade volumes using the gravity equation. They argue that fixed effects should be applied to the gravity equation rather than random effects for two reasons. First, they assume that there are unobserved time-invariant bilateral variables, influencing both the presence of a free trade agreement and trade volumes. They argue that these variables are controlled best using fixed effects because it allows for arbitrary correlations between these unobserved variables and the independent variable. On the other hand, with random effects it is assumed that there is no correlation between the unobserved variables and the independent variables. However, this latter result is less likely than the former. Second, they argue that previous evaluations of the gravity equation by Egger (2000) –using the Hausman Test8– have resulted in favour of fixed effects rather than random effects.

𝐈𝐈𝐈𝐜_{. Model assumptions}

As mentioned before, the gravity equation will be modelled by means of constructing a regression model, which subsequently will be estimated with fixed effects. However, when applying any regression model, there are some (specific) assumptions that need to be satisfied in order for the estimates to be reliable i.e. consistent. For the fixed effects estimators to be

consistent, the regression has four assumptions that need to be satisfied. Besides, if these assumptions hold the estimator is asymptotically normally distributed. This result is convenient upon testing. The following assumptions need to hold:

i. 𝑢_!"# has a conditional mean of zero

ii. (X!,!", X!,!", … , X!,!", 𝑢!,!"#, 𝑢!,!"#, … . , 𝑢!,!"#) are independent and identically distributed (i.i.d.) draws from their joint distribution

iii. Large outliers are unlikely

iv. There is no perfect multicollinearity

8_{Statistical test used to evaluate fixed effects estimation versus random effects estimation. It also tests the consistency of an} estimator relative to a less efficient –but consistent– estimator

(14)

𝐈𝐈𝐈𝐝_{. Model specification}

As mentioned before, the effect of the European Single Market on Dutch trade levels needs to be estimated. This effect will be estimated by exploiting the possibility of adding additional

variables to the linear gravity model, which was specified in the introduction of this chapter. These additional variables were defined as 𝐴_!"#. The primary variable of interest in this case is countries their membership in the European Single Market. To include this characteristic in the regression model, the variable PTA is added to the model. This is a dummy variable with the following specification:

𝑃𝑇𝐴_!" = 1, if countries i and j are both part of the European Single Market at time t

0 otherwise

Since this thesis has its emphasis on Dutch trade, country i is specified as the Netherlands. The regression model needs to be augmented with additional variables to correct for omitted variable bias: In this current form, the regression model will yield a biased estimator for PTA. There are more variables that could explain trade volumes, which are also correlated with countries partaking in the European Single Market. As mentioned by Stock and Watson (2015), the estimator will yield either an upward or downward bias, depending on the correlation between PTA and the variable(s) omitted. They specify this omitted variable bias with the following equation:

𝑏_! ! 𝑏_!+ 𝜌_!"𝜎!

𝜎_!

, where u represents the variable(s) omitted, 𝑏_! is the estimator of independent variable X, 𝜌_!" = Corr(𝑋_!, 𝑢_!), and 𝜎_! (𝜎_!) is the standard deviation of u (x). In principal, this equation states that the larger the sample, the closer 𝑏_! is to 𝑏_!+ 𝜌_!"!!

!! with high probability. Thus, leading to

biased and inconsistent results. However, it should be noted that the variables added should change over time, since fixed effects regression already implicitly controls for time-invariant variables (as mentioned in chapter III!_).

In addition, the regression model will be augmented with a variable accounting for the real effective exchange rate over time. This exchange rate measures the progress of the value of a

(15)

country’s currency –adjusted for inflation– against a basket of the trading partners of the country. The real effective exchange rate is used in many economic studies for a wide variety of purposes. One of those is assessing the drivers of trade flows. It can be argued that changes in real effective exchange rates could encourage/discourage trade activity. This is due to the fact that an increase (decrease) in the real effective exchange rate of a country means that its goods become relatively more expensive (cheaper). Thus, an increase (decrease) in net imports is expected. For this reason this variable will be included in the model. In the regression model this variable will be specified as 𝑅𝐸𝐸𝑅!"

Furthermore, a variable accounting for the total foreign direct investments over the years will be added to the model. This variable will be specified as 𝐹𝐷𝐼_!"9_{. The interest in this variable} comes from the view that foreign direct investments could be highly correlated with regional trade agreements –specifically the European Single Market– and could affect trade volumes. It can be argued that due to increased market size and integration, investing in the European Single Market is less risky relative to investing in an individual country that is not part of a union. Evidently, when not taking account of this correlation and not controlling for foreign direct investments, this could lead to large omitted variable bias (see equation above).

Moreover, the regression model will also include a dummy variable accounting for the Netherlands and its partners (j) having the same currency:

𝐶𝑈𝑅𝑅_! = 1, if country i and j have the same currency at time t 0 otherwise

A common currency reduces transaction costs, as there is no need to incur administrative costs when converting currencies. This could be an incentive for trade volumes to change, and thus a common currency can be considered a determinant of trade. This view is also in line with the findings of Frankel and Rose (2002). They conducted research, using data for over 200 countries, to determine the effects of a common currency on trade and income. They found that common currencies increase bilateral trade and increase overall trade openness. Also trade within a currency union is found to have tripled. Furthermore, Frankel and Rose conclude that the trade

9_{This variable reflects bilateral total FDI investments over the years. These total investments consist of the sum of inward and} outward FDI flows

(16)

created among members of a common currency union does not come at an expense of trade diversion from non-members.

Taking all of the variables described above into account, the conformed gravity equation model that is used in this thesis for testing will be displayed in the following way:

ln 𝑃𝑋_!" = 𝛽_!ln 𝑌_! + 𝛽_!ln 𝑌_!" + 𝛽_!ln 𝐷_! + 𝛽_!𝑃𝑇𝐴_!"+ 𝛽_!𝑅𝐸𝐸𝑅_!"+ 𝛽_!𝐶𝑈𝑅𝑅_!"

+ 𝛽!ln 𝐹𝐷𝐼!"(!"!#$)+ 𝐹𝐸!+ ln 𝑢!" (4)

, where 𝐹𝐸! represents the fixed effects that vary across trading partners (j).

The main variable of interest in this thesis is PTA, rather the European Single Market. As mentioned in section III!_{, after determining this fixed effects estimator it can be assessed}

whether or not this variable has a significant effect on trade. In terms of equation 3, the following hypothesis needs to be tested:

H_! ∶ 𝛽_! = 0 H_! ∶ 𝛽_! ≠ 0

Assuming that the assumptions hold (see section III!_{), it is presumed that the data is}

asymptotically normally distributed. The assessment of the significance of 𝑃𝑇𝐴_!"# will be done with statistical testing. This thesis will use the p-value approach. This approach is used to test hypotheses and entails setting significance level(s) – 𝛼 – after which the p-value can be

computed from the observations. The p-value is defined as the probability of a result equal to or more extreme than observed. If p-value ≤ 𝛼, H! can be rejected. In this thesis this method will be used to test the hypothesis multiple times for various forms of the gravity equation.

(17)

IV. Dataset

Having discussed the variables and corresponding method used, this section discusses the dataset in more detail. The data sample used in this thesis involves a total of 195 countries, consisting of the Netherlands and 194 of their trading partners. This data sample involves a strongly balanced 27-year time period from 1988 through 2014. This represents both before and after integration-periods, which is specified here as 1993.

The data for the bilateral trade values, exports and imports (all in euros) comes from the Eurostat Database for International Trade: International Trade in Goods Statistics (ITGS). Figure 2 represents the total Dutch trade activity data used in this thesis, where total trade values are comprised of the sum of Dutch exports and Dutch imports

Figure 2 - Total trade values (exports and imports) 1988-2014 in euros

Source: Eurostat

The trade values used in the sample involve all 194 Dutch trading partners. A list of these countries is available in Appendix 3. Upon estimation, these values are transformed to (natural) logarithms. The trade values concern the values of goods traded. In this database, ‘goods’ are defined as all moveable property including electricity.

0 5. 00 0e +1 0 1. 00 0e +1 1 1. 50 0e +1 1 2. 00 0e +1 1 To ta l t ra de va lu e 1990 1995 2000 2005 2010 2015 year

(18)

To account for the Gross Domestic Product variables, the World Bank database on GDP data is used. The GDP data used from this database includes all countries specified in Appendix 3. Moreover, within this database GDP is specified as the sum of gross value added by all resident producers in the economy plus any product taxes, minus any subsidies not included in the value of the products. These GDP values in current US dollars have been converted to euros according to the following method: For the years after the euro-introduction in 1999, the GDP values in dollars have been converted to euros using the yearly USD/EUR exchange rates. For the years considered prior to the euro-introduction (1988-1998), the GDP values in dollars have first been converted to the Dutch Guilder using the yearly USD/NLG exchange rates.

Subsequently, the GDP values –in Dutch Guilders– have been converted to euros using the Guilder/EUR parity (1NLG=0,453780216EUR). Similar to the trade data, this data will also be transformed to logarithms upon estimation. Figures 3 and 4 represent the GDP values of the Netherlands and their trading partners j –respectively– over time.

Figure 3 - Dutch GDP 1988-2014 (converted to euros)

Source: World Bank world indicators

2. 00 0e +1 13 .0 00 e+1 14 .0 00 e+1 15 .0 00 e+1 16 .0 00 e+1 17 .0 00 e+1 1 G D P N et he rla nd s 1990 1995 2000 2005 2010 2015 year

(19)

Figure 4 – GDP values of Dutch trading partners 1988-2014 (converted to euros)

Source: World Bank world indicators

With the use of Geodist database provided by CEPII, the variable representing distance within the gravity equation can be accounted for. This database contains numerous variables accounting for bilateral country-specific geographic characteristics. There are several alternatives to account for distance. In particular, the database distinguishes two kinds of distance measures: simple distances and weighted distances. The former requires only one city to calculate

international distances; the latter requires the use of multiple principle cities in each country. This thesis uses simple distances, since these involve only the most important city of each country.

Data on the real effective exchange rates is collected from Bruegel datasets, which is a relatively new database established in 2012. This dataset includes nominal and real effective exchange rates for 178 countries with a maximum of 172 trading partners. An alternative within this dataset is the real effective exchange rates computed with 67 trading partners. Furthermore, these exchange rates are calculated with base year = 2007. This thesis uses the real effective exchange rates calculated with 172 trading partners. These variables will be used based on the view that the 194 Dutch trading partners considered in this thesis are represented well by these

0 5. 00 0e +1 2 1. 00 0e +1 3 1. 50 0e +1 3 G D P Tra di ng Pa rtn er 1990 1995 2000 2005 2010 2015 year

(20)

172 partners used in the calculations of the Bruegel dataset. Moreover, Darvas (2012) states that the real effective exchange rates used are specified as follows:

REER_! = !""#! ! !"#!

!"#_!(!"#$%&') ,

, where REER! is the real effective exchange rate of the Netherlands against a basket of 172 of their trading partners’ currencies over time, and NEER_! = _!"#𝑆(𝑗)_!!(!)

!!! is the nominal effective exchange rate of the Netherlands. This is the geometrically weighted average of S(j)_!, the nominal bilateral exchange rate between the Netherlands and its trading partner j. CPI_!(!"#$%&')=

𝐶𝑃𝐼(𝑗)_!!(!) !"#

!!! is the geometrically weighted average of CPI indices of 172 trading partners, where CPI(j)_! is the consumer price index of trading partner j at time t, and 𝑤(!)_{is the weight} attributed to trading partner j ( !"#𝑤_!(!)

!!! = 1). Furthermore, it is argued that the motivation for the use of geometric weight averages comes from previous literature.

At last, the OECD database is used to account for the total Foreign Direct Investments (FDIs). Total FDIs are specified as the sum of inward and outward FDIs. Inward FDIs are the investments in the reporting country under study (the Netherlands), and outward FDIs are Dutch investments in their partner countries j. In this thesis the focus will be on total bilateral FDI investments and thus the variable in equation 3 accounting for FDI investments equals:

𝐹𝐷𝐼_!"(!"!#$) = 𝐹𝐷𝐼_!"(!"#$%&)+ 𝐹𝐷𝐼_!"(!"#$%&'). Upon estimation this variable will be expressed as

a logarithm.

V. Empirical results

This section presents the results of this research. Subsequently, these results are explained in detail. Tables 1 through 3 show the effects of the fixed effect estimators specified in equation 3. Total trade values, exports and imports have been regressed on multiple variations of these estimators. Tables 1 through 3 all distinguish six regressions for each variant. The effects will be assessed on a .01-signifance level. However, other significance levels will also be considered. In addition, these estimations have been repeated using an alternative estimation method: Pooled OLS. Because fixed effects regression is the main method of interest, the results on Pooled OLS are available in Appendix 2, tables 4 through 6.

(21)

As can be referred from table 1, the first regression involves a fixed effects regression of total trade on the original variables10 of the gravity equation. The GDP of the Netherlands and their 194 trading partners both have positive effects on total trade at a .01-significance level. Because the variables considered are expressed as logarithms, the estimates represent their elasticity. Specifically, a 1%-point increase in Dutch GDP increases total Dutch trade values with 1.148%-points. Similarly, a 1%-point increase in the GDP of Dutch trading partners increases total Dutch trade values with .350%-point.

Table 1 – Fixed effects regression, Total Trade

Independent variables (1) (2) (3) (4) (5) (6) Constant -20.361*** (2.088) -19.207*** (2.188) -26.813*** (2.651) -26.988*** (2.756) -18.133*** (1.716) -18.006*** (1.896) Y_! 1.148*** (.107) .1.100*** (.110) 1.455*** (.134) 1.463*** (.138) .786*** (.120) .779*** (.126) Y!" .350*** (.062) .354*** (.062) .277*** (.066) .276*** (.066) .711*** (.102) .713*** (.102) PTA_!" .334*** (.086) .311*** (.096) .329*** (.095) .265*** (.072) .263*** (.071) REER!" .0001*** (.00003) .0001*** (.00003) (.001) -.002 (.001) -.002 CURR!" -.064 (.070) .011 (.053) FDI!" .007 (.010) (.010) .007 Dependent variable: Total trade

Estimation method: (Country) Fixed effects

Notes: Robust standard errors are given in parentheses.

All variables are expressed in natural logarithms, with the exception of PTA, REER and CURR *: Significance at 𝛼 = .10

**: Significance at 𝛼 = .05 ***: Significance at 𝛼 = .01

Furthermore, the Pooled OLS regression on the same equation also yields positive effects. These Pooled OLS estimates are shown in Appendix 2, table 4. Distance shows to have a negative effect at .01-significance on overall trade values: A 1%-point increase in distance between the

10_{Note that 𝐷}_!_{is omitted from this table because it is a time-invariant variable and tables 1-3 involve fixed effects regression (see} Appendix 1). The estimates on distance are available in Appendix 2

(22)

Netherlands and their trading partners decreases total Dutch trade values with .680%-point. These findings are consistent with the overall expectations of the gravity model.

The second regression in table 1 includes the main variable of interest –PTA– in the standard gravity equation. This variable is significant at a .01-significance level. Moreover, it shows a positive effect implying that European Single Market-membership increases Dutch trade activity among their members. Specifically, if the Netherlands and their trading partner both are members of the Single Market, overall trade is 33.4%-points higher relative to non-membership. The standard variables of the gravity equation variables also show similar significance.

When the gravity equation is augmented with the variable accounting for the real effective exchange rate, the fixed effects estimator of the European Single Market decreases relative to the second regression. Regression 3 of table 1 shows that controlling for real effective exchange rates decreases the effect on total trade with 2.3%-points relative to the second

regression. This entails an increase in total Dutch trade activity of 31.1%-points when the Netherlands and their trading partner are both part of the Single Market relative to non-membership. This effect is significant at a .01-significance level. Furthermore, a 1%-point increase in the real effective exchange rate increases total Dutch trade activity with .01%-point. This effect of the real effective exchange rate on Dutch trade activity is also significant at a .01-significance level. These results imply that the second regression overestimated the effect of the Single Market on Dutch trade activity due to lack of other explanatory variables, and that the real effective exchange rates should be included.

Regression 4 adds the dummy variable accounting for the Netherlands and their trading partners both having the same currency. When adding this dummy, the fixed effects estimator of the European Single Market increases relative to regression 3. Specifically, membership in the Single Market increases total trade value with 32.9%-points relative to non-membership. This is an increase of 1.8%-points relative to the third regression. The effect remains positive and

significant at a .01-significance level. Despite this result, the effect of having a common currency shows no significance at 90% confidence. Implying that Dutch trade activity is not affected by the benefits that are encountered when the Netherlands and their trading partner have the same currency.

Regression 5 is a modification of the third regression equation. In this case, the variable accounting for a common currency has been omitted (since there seemed to be no significance at

(23)

a 90%). Instead, the variable accounting for FDI flows has been added. In this specific model, the standard gravity equation variables all show significant effects at 99% confidence. However, the variables accounting for FDI flows and the real effective exchange rate show no significance at a .01-significance level. Nevertheless, the effect of the European Single Market on total Dutch trade activity remains significant at a .01-significance level. Here, the effect of the Single Market on total Dutch trade activity is 26.5%-points, which is a decrease of 4.6%-points relative to the third regression.

Finally, regression 6 includes all the variables that are of interest in the fixed effects regression. In this regression, both the variables accounting for common currencies and FDI flows are added to the model. The results show –similar to the previous regression– that the standard gravity equation variables are all significant at 99% confidence. The main variable of interest –again– generates a positive effect for total trade. Specifically, membership of the European Single Market increases total Dutch trade with 26.3%-points relative to

non-membership. This is the lowest effect observed for this effect in these regressions. Nonetheless, this effect is still significant at a .01-signifiance-level. This result is consistent with the results generated from the regression performed under Pooled OLS. The variables accounting for countries having a common currency, FDIs flows, and the real effective exchange rates all are insignificant at a .10-signficance level. Thus, implying that none of these variables can determine total Dutch trade activity.

Overall these results show that the European Single Market has a positive effect on total Dutch trade values. Because total Dutch trade is comprised of imports and exports, it can be argued that this effect is mostly captured by Dutch exports. This could be explained through the fact that Dutch exports have always been strongly EU oriented whereas Dutch imports are not. To confirm this statement, separate regressions for imports and exports follow.

As mentioned before, this thesis tries to expand the research question at hand by also examining the different components of total trade, exports and imports. Similar to table 1, table 2 presents variants of the gravity equation model performed under fixed effects regression.

(24)

Table 2 – Fixed effects regression, Exports Independent variables (1) (2) (3) (4) (5) (6) Constant -20.068*** (1.945) -18.348*** (1.991) -25.914*** (2.208) -25.917*** (2.294) -20.769*** (1.656) -20.213*** (1.791) Y_! 1.165*** (.097) 1.094*** (.099) 1.457*** (.107) 1.457*** (.111) .815*** (.101) .786*** (.105) Y!" .286*** (.052) .292*** (.051) .205*** (.048) .205*** (.049) .746*** (.085) .755*** (.086) PTA_!" .498*** (.081) .490*** (.081) .490*** (.079) .332*** (.052) .325*** (.051) REER!" .0001** (.00004) .0001** (.00004) -.0002 (.0009) -.0002 (.001) CURR_!" -.001 (.062) .050 (.050) FDI!" .008 (.009) .007 (.009) Dependent variable: Exports

The standard gravity equation under fixed effects is represented by the first regression in table 2. The results reflect the expected effects, with all variables being significant at a .01-significance level. Similar results are found in table 5, where the equation has been estimated under Pooled OLS.

The second regression includes the main variable of interest in the standard gravity equation. The Single Market affects Dutch exports positively at a .01-significance level. Specifically, Dutch exports increase with 49.8%-points when the Netherlands and their trading partner are both members of the European Single Market relative to non-membership. Moreover, the standard gravity equation variables show similar significance.

In the third regression the variable accounting for the real effective exchange rate has been added. When controlling for this variable, the effect of the main variable of interest

(25)

decreases 0.8%-point relative to regression 2. Here, Dutch exports increase by 49%-points when the Netherlands and their trading partner are both part of the European Single Market, as

opposed to non-membership. Despite this decrease this effect still remains significant at a .01-significance level. However, the variable accounting for the real effective exchange rates is significant at 95% confidence but not at 99%. This result is in contrast with the results found under the regression performed on total Dutch trade activity. This means that the real effective exchange rates are better at explaining total Dutch trade activity than Dutch exports. Thus, implying that real effective exchange rates should predict Dutch imports better than Dutch exports (in a regression concerning the same variables).

The fourth regression adds the dummy variable accounting for the Netherlands and their trading partners both adopting the same currency. When estimating the equation in this form, the fixed effects estimator of PTA is significant at .01. The effect of PTA on Dutch exports is the same as in the third regression: Dutch Exports increase with 49%-points when the Netherlands and their trading partner are both part of the European Single Market relative to

non-membership. Moreover, the effect of having a common currency is insignificant and cannot seem to determine Dutch exports. This effect is similar to the regressions on total Dutch trade.

The fifth regression is a modification of regression 3. This regression includes the variable accounting for FDI flows while the dummy variable accounting for common currencies has been omitted. The main variable of interest –again– shows significance at 99%. When the Netherlands and their trading partner are both part of the European Single Market, Dutch exports increase with 33.2%-points. However, the effect of FDI flows on Dutch exports is statistically insignificant at 99% confidence. This is similar to the regressions on total Dutch trade, where FDI flows and the real effective exchange rates also showed insignificant effects.

In the final regression both the variable accounting for total FDI flows and the variable accounting for common currencies are added. Again, the standard gravity equation variables show significant effects at 99% confidence. The effect of the Single Market remains significant at a .01-significance level. Specifically, membership of the Single Market increases Dutch exports with 32.5%-points relative to non-membership. This entails a decrease of 16.5%-points relative to regression 4. This effect is also consistent with its Pooled OLS counterpart, which shows a positive significant effect at a .01-significance level. The variables accounting for the

(26)

real effective exchange rates, FDI flows, and countries having the same currency cannot seem to determine Dutch exports. This is similar to the results under the regressions on total Dutch trade.

These results show that the European Single Market has an overall positive effect on Dutch Exports. This conforms to the statement made above. However, a regression on Dutch imports needs to be assessed to confirm that the effect of the Single Market on total Dutch trade is dominantly generated through Dutch exports.

Finally the effects of the variables of interest on Dutch imports are presented in table 3. Similar to the estimations given in table 1 and table 2, the standard gravity equation variables under fixed effects are presented in the first regression of table 3. The results reflect the theory with all variables being significant at a .01-significance level. Similar results (with exception of Dutch GDP) are found in table 6, where the equations have been estimated under Pooled OLS

Table 3 – Fixed effects regression, Imports

Independent variables (1) (2) (3) (4) (5) (6) Constant -20.660*** (3.250) -20.061*** (3.473) -26.950*** (3.950) -27.264*** (4.107) -14.064*** (2.701) -14.487*** (2.956) Y! 1.017*** (.158) .992*** (.166) 1.318*** (.194) 1.332*** (.200) .522*** (.199) 544*** (.208) Y_!" .485*** (.089) .461*** (.090) .385*** (.100) .383*** (.100) .798*** (.156) .791*** (.157) PTA!" .171 (.114) .152 (.136) .184 (.131) .184 (.116) .189* (.114) REER_!" .0002*** (.00003) .0002*** (.00003) -.002 (.002) -.002 (.002) CURR!" -.115 (.097) -.038 (069) FDI_!" .015 (.019) .016 (.018) Dependent variable: Imports

(27)

The second regression includes the main variable of interest in the standard gravity equation. In contrast to the regressions in table 1 and table 2, the effect of the European Single Market is insignificant with 90% confidence. Thus, implying that there is no difference in trade regarding Dutch imports between members of the European Single Market and non-members. Apparently, the Single Market does not incentivize the Netherlands to import more from fellow Single Market-members. However, the standard gravity equation variables do show significance at a .01-significance level, similar to the regressions on Dutch exports.

The third regression adds the variable accounting for the real effective exchange rate to the model. When controlling for this variable, the effect of the European Single Market on Dutch imports remains insignificant at a .01-significance level. This is also in contrast with the

regressions on exports and total trade. However, the variable accounting for the real effective exchange rates is significant at a significance level of .01, contrary to the regression on exports. This confirms the statement that the real effective exchange rates are better at explaining Dutch imports as opposed to Dutch exports. It can be argued that the real effective exchange rates cannot explain Dutch exports well due to Dutch exports being EU oriented. This is because most EU members have the same currency as the Netherlands and thus are not prone to exchange rate changes. Furthermore, the effect of the real effective exchange rates in consistent with theory described in section III!_.

The fourth regression adds the dummy variable accounting for the Netherlands and their trading partners adopting the same currency. In this form, the fixed effects estimator of the Single Market remains insignificant at .01. This effect is –again– in contrast to the regression on Dutch exports. Moreover, the effect of having a common currency is insignificant at a .10-significance level.

The fifth regression adds the variable accounting for FDI flows, but omits the variable accounting for common currencies between the Netherlands and their trading partners. In this specific model, the standard gravity equation variables remain significant at 99% confidence. The effect of the European Single Market on Dutch imports still shows no significance at 99% confidence. Neither do the variables accounting for FDI flows and real effective exchange rates. This is similar to the results under the regressions on Dutch exports.

The final regression includes all variables of interest, as specified in equation 3. Similar to the regression on exports and total trade, the standard gravity equation variables are significant

(28)

at 99% confidence. Also, the results show insignificance for the dummy variable accounting for a common currency and the variable accounting for bilateral FDI flows at 90% confidence. The main variable of interest becomes significant at 90%. However, there is still no significant effect noticeable at either 95% or 99% confidence. This effect is inconsistent with the regression model performed under Pooled OLS. Under Pooled OLS, the effect of the European Single Market is positively affects Dutch imports at a .01-significance level. This could be explained by the fact that Pooled OLS does not control for the assumed constant unobserved heterogeneity across countries. Also, the effect of FDIs on Dutch imports seems insignificant at .01.

Corresponding to the statements made above, it can be concluded that the European Single Market has no significant effect on Dutch Imports. Furthermore, the variables used to augment to gravity equation (with exception of PTA) all cannot seem to explain any form of Dutch trade activity (at least in the considered combinations).

VI. Conclusion

This thesis aimed at determining the effect of the European Single Market on total Dutch trade activity using the gravity model. Taking into account country fixed effects, the effects of the European Single market were assessed on Dutch total trade, exports and imports. The results showed that the European Single Market has a positive effect on total Dutch trade activity. Specifically, total Dutch trade activity increases 26.3%-points when the Netherlands and their trading partner are both part of the European Single Market. This is the result generated when all the variables discussed in this thesis are accounted for. Furthermore, this effect was consistent with the effect estimated under a regression performed with Pooled OLS. Because total Dutch trade activity consists of the sum of Dutch exports and imports, it was argued that this positive effect was generated through the exports more so than the imports. This argument was based on the view that Dutch exports –as opposed to Dutch imports– have always been EU oriented. Enforcing this argument, the analysis on Dutch exports showed that the European Single Market has a positive effect on these exports. Furthermore, these results were overall higher than the results found under the analysis of total Dutch trade activity. Specifically, Dutch Exports increases 31.6%-points when the Netherlands and their trading partner are both part of the European Single Market. This result was also consistent with the results generated under Pooled

(29)

OLS. Finally, the analysis on Dutch imports confirmed the statement by showing insignificant results for the effect of the European Single Market on Dutch imports. Thus, the establishment of the European Single Market did not incentivize the Netherlands to import more from fellow European Single Market members. However, this result was inconsistent with the results found under Pooled OLS.

Further research on this topic could explore the effect of a more advanced European integration stage on trade levels. This could provide answers as to how far integration should go. The results in this thesis showed that Dutch trade activities were not stimulated by the benefits of having a common currency. This result implies that the Netherlands should not have engaged in integration beyond having a common currency had the justification of this integration been based on trade creation. To confirm this statement, further research should apply different

combinations of variables to assess whether there are any differences in results. Furthermore, a larger panel could be used to enlarge the sample, which would result in more precise estimates.

(30)

APPENDIX 1: (APPLIED) FIXED EFFECTS REGRESSION

As mentioned by Stock and Watson (2015), fixed effects regression entails estimating so called entity-demeaned variables. To find these entity-demeaned variables, the country-specific average is subtracted from each variable included in the model.

In terms of equation 3 (without taking account of logarithms), the country-specific average equation is:

𝑃𝑋_! = 𝛽_!+ 𝛽_! 𝑌 + 𝛽_! 𝑌_! + 𝛽_! 𝐷_! + 𝛽_!𝑃𝑇𝐴_!+ 𝛽_!𝑅𝐸𝐸𝑅_! + 𝛽_!𝐶𝑈𝑅𝑅_! + 𝛽_! 𝐹𝐷𝐼_!(!"!#$) + 𝐹𝐸_!+ ln 𝑢_!

, where 𝑃𝑋_! = (1/27) !"_!!!𝑃𝑋_!" (all other averages are calculated similarly).

Thus, in terms of equation 3 (without taking account of logarithms):

(𝑃𝑋_!"− 𝑃𝑋_!) = 𝛽_! 𝑌_!− 𝑌 + 𝛽_! 𝑌_!"− 𝑌_! + 𝛽_! 𝑃𝑇𝐴_!"− 𝑃𝑇𝐴_! + 𝛽_! 𝑅𝐸𝐸𝑅_!"− 𝑅𝐸𝐸𝑅_!

+ 𝛽_! 𝐶𝑈𝑅𝑅_!"− 𝐶𝑈𝑅𝑅_! + 𝛽_! 𝐹𝐷𝐼_!"(!"!#$)− 𝐹𝐷𝐼_!(!"!#$) + (𝑢_!"− 𝑢_!)

Note that the variables accounting for the constant unobserved effects and distance have been eliminated because they are constant: 𝐷! = 𝐷! and 𝐹𝐸! = 𝐹𝐸! => 𝐷!− 𝐷! = 0 and 𝐹𝐸!− 𝐹𝐸_! = 0

Define the entity-demeaned variables: (𝑃𝑋!"− 𝑃𝑋!) = 𝑷𝑿!", 𝑌!− 𝑌 = 𝒀!, 𝑌!"− 𝑌! = 𝒀!" … etc.

Subsequently, the Fixed Effects estimator(s) 𝛽_!" can be found by OLS:

(31)

APPENDIX 2: POOLED OLS RESULTS AND CORRELATIONS TABLE

Table 4 – Pooled OLS, Total Trade

Independent variables (1) (2) (3) (4) (5) (6) Constant -6.378*** (1.312) -5.981*** (1.312) -9.079*** (1.675) -9.004*** (1.686) -12.487*** (2.426) -12.225*** (2.492) Y! .288*** (.051) .261*** (.051) .378*** (.066) .375*** (.066) .561*** (.089) .550*** (.091) Y_!" D! .998*** (.009) -.680*** (.021) .985*** (.009) -.635*** (.022) .967*** (.010) -.594*** (.026) .967*** (.010) -.594*** (.025) .843*** (.018) -.403*** (.025) .844*** (.018) -.402*** (.025) PTA!" .252*** (.051) .330*** (.054) .307*** (.064) .415*** (.044) .397*** (.046) REER!" .0004*** (.00008) .0004*** (.00007) (.0008) -.001 (.0008) -.001 CURR!" .050 (.076) .045 (.050) FDI!" .070*** (.015) .069*** (.015) Dependent variable: Trade

Estimation method: Pooled OLS

(32)

Table 5 – Pooled OLS, Exports Independent variables (1) (2) (3) (4) (5) (6) Constant -3.426*** (1.301) -2.789*** (1.297) -6.203*** (1.625) -6.022*** (1.636) -11.763*** (2.292) -11.534*** (2.363) Y_! ..333*** (.051) .290*** (.050) .430*** (.063) .423*** (.063) .662*** (.085) .652*** (.087) Y!" D_! .879*** (.009) -.958*** (.019) .874*** (.009) -.887*** (.021) .855*** (.009) -.877*** (.024) .855*** (.009) -.875*** (.023) .741*** (.017) -.620*** (.021) .742*** (.017) -.619*** (.021) PTA!" .407*** (.049) .415*** (.050) .359*** (.059) .540*** (.037) .524*** (.040) REER!" .0004*** (.00008) .0004*** (.00008) -.003*** (.0009) -.003*** (.0009) CURR!" .120** (.069) (.049) .040 FDI!" .098*** (.014) .097*** (.014) Dependent variable: Exports

(33)

Table 6 – Pooled OLS, Imports Independent variables (1) (2) (3) (4) (5) (6) Constant -8.777*** (2.112) -8.093*** (2.133) -7.178*** (2.771) -7.305*** (2.791) -11.427*** (3.724) -11.805*** (3.793) Y! .021 (.082) -.023 (.083) -.088 (.108) -.083 (.108) .295** (.135) .311** (.138) Y_!" D_! 1.241*** (.013) -.400*** (.030) 1.235*** (.013) -.327*** (.035) 1.240*** (.016) -.249*** (.039) 1.240 *** (.016) -.250*** (.039) .986*** (.031) -.223*** (.042) .985*** (.031) -.224*** (.042) PTA_!" .418*** (.085) .534*** (.090) .573*** (.102) .365*** (.087) .391*** (.087) REER!" .0005*** (.0001) .0005*** (.0001) -9.86e-06 (.001) -6.02e-06 (.001) CURR_!" -.084 (.118) (.090) -.065 FDI!" .071*** (.024) .073*** (.024) Dependent variable: Imports

(34)

Ln(gdpnld) Ln(gdppart) Ln(dist) pta reer curr Ln(fdi) Ln(gdpnld) 1.0000 Ln(gdppart) -0.0544 1.0000 Ln(dist) 0.0788 -0.2417 1.0000 pta 0.0453 0.3029 -0.6919 1.0000 reer -.0005 -0.0266 0.1213 -0.0844 1.0000 curr 0.1734 0.2248 -0.4690 0.6343 -0.0613 1.0000 Ln(fdi) -.0200 0.7615 -0.3882 0.4238 -0.0748 0.3781 1.0000

(35)

APPENDIX 3: DUTCH TRADING PARTNERS INCLUDED IN RESEARCH AFGHANISTAN ALBANIA ALGERIA ANDORRA ANGOLA ANTIGUA & BARBDUDA ARGENTINA ARMENIA ARUBA AUSTRALIA AUSTRIA* AZERBAIJAN BAHAMAS BAHRAIN BANGLADESH BARBADOS BELARUS BELGIUM* BELIZE BENIN BERMUDA BHUTAN BOLIVIA BOSNIA HERZEGOVINA BOTSWANA BRAZIL BRUNEI BULGARIA* BURKINA FASO BURUNDI CAMBODIA CAMEROON CANADA CAPE VERDE CENTRAL AFRICAN REPUBLIC CHAD CHILE CHINA COLOMBIA COMOROS CONGO COSTA RICA COTE D’IVOIRE CROATIA* CUBA CYPRUS* CZECH REPUBLIC* DENMARK* DJIBOUTI DOMINICA DOMINICAN REPLUBLIC ECUADOR EGYPT EL SALVADOR EQUATORIAL GUINEA ERITREA ESTIONA* ETHIOPIA FAROE ISLANDS FINLAND* FRANCE* FRENCH POLYNESIA FIJI GABON GAMBIA GEORGIA GERMANY* GHANA GREECE* GREENLAND GRENADA GUATEMALA GUINEA GUINEA-BISSEAU GUYANA HAITI HONDURAS HONG KONG HUNGARY* ICELAND INDIA INDONESIA IRAN IRAQ IRELAND* ISRAEL ITALY* JAMAICA JAPAN JORDAN KAZAKHSTAN KENYA KIRIBATI KOSOVO KUWAIT KYRGYZ LAOS LATVIA* LEBANON LESOTHO LIBERIA LIBYA LIECHTENSTEIN LITHUANIA* LUXEMBOURG* MACAO MADAGASCAR MALAWI MALAYSIA MALDIVES MALI MALTA* MARSHALL ISLANDS MAURITHANIA MAURITIUS MEXICO MICRONESIA MOLDOVA MONGOLIA MONTENEGRO MOROCCO MOZAMBIQUE NAMIBIA NEPAL NEW CALEDONIA NEW ZEALAND NICARAGUA NIGER NIGERIA NORWAY OMAN PAKISTAN PALAU PANAMA PAPUA NEW GUINEA PARAGUAY PERU PHILIPPINES POLAND* PORTUGAL* QATAR ROMANIA* RUSSIA RWANDA SAMOA SAN MARINO SAO TOME & PRINCIPE SAUDI ARABIA SENEGAL SERBIA SEYCHELLES SIERRA LEONE SINGAPORE SLOVAKIA* SLOVENIA* SOLOMON ISLANDS SOMALIA SOUTH AFRICA SOUTH KOREA SPAIN* SRI LANKA

ST. KITTS & NEVIS ST. LUCIA ST. VINCENT & GRENADINES SUDAN SURINAME SWEDEN* SWITZERLAND SYRIA TAIWAN TAJIKISTAN TANZANIA THAILAND TOGO TONGA TRINIDAD & TOBEGO TUNISIA TURKEY TURKMENISTAN TUVALU UGANDA UKRAINE UNITED ARAB EMIRATES UNITED KINGDOM* UNITED STATES URUGUAY UZBEKISTAN VANUATU VENEZUELA VIETNAM YEMEN YUGOSLAVIA ZAMBIA ZIMBABWE

(36)

REFERENCES

Aitken, Norman D., “The effect of the EEC and EFTA on European Trade: A Temporal Cross-Section Analysis.” American Economic Review, Vol. 63, No. 5, 1973, pp. 881-892 Baier, S.L., Bergstrand, J.H., “Do free trade agreements actually increase members’

international trade?” Journal of International Economic, Vol. 71, Issue 1, 2007, pp. 72-95

Bergstrand, J.H., “The Gravity Equation in International Trade: Some Microeconomic

Foundations and Empirical Evidence.” The Review of Economics and Statistics, Vol. 67, No. 3, 1985, pp.474-481

Brada, J.C., Mendez, J.A., “Economic Integration among Developed, Developing and Centrally Planned Economies: A Comparative Analysis.” Review of Economics and

Statistics 67, No.4, 1985, pp. 549-556

Darvas, Z., “Real Effective Exchange Rates for 178 Countries: A New Database. Bruegel

Working Paper, 2012/06.

Egger, Peter. “A Note on the Proper Econometric Specification of the Gravity Equation.”

Economic Letters, 66, 2000, pp. 25-31

Frankel, J., Rose, A.K., “An estimate of the effect of common currencies on trade and income.” The Quarterly Journal of Economics, 117, 2002, pp. 437-466

Frankel, J., Stein, E., Wei, S., “Trading blocs and the Americas: The natural, the unnatural, and the super-natural.” Journal of Development Economics, Vol. 47, Issue 1, 1995, pp. 61-95

Samuelson, P.A., “The Gains from International Trade Once Again”, The Economic Journal, Vol. 72, No. 288, 1962, pp. 820-829

Stock, J.H., Watson, M.W. (2015). “Introduction to Econometrics”. Harlow, Essex: Pearson Education Limited

The effect of the European Single Market on trade activity : an analysis of the Dutch economy

Faculty of Economics and Business

Bachelor’s thesis

The effect of the European Single Market on trade

activity: An analysis of the Dutch economy

TABLE OF CONTENTS