The eastern enlargement of the European Union in 2004 : the pattern of the trade relationships between the Netherlands and the acceded countries

Bachelor Thesis Economics

Supervisor: Hans van Ophem

The eastern enlargement of the European Union in 2004:

The pattern of the trade relationships between the Netherlands and the

acceded countries

Name: Floortje Merten

Student number: 10379975

Faculty: Economics and Business

2

Table of Contents

Introduction

Methodology page 3

Theoretical Framework

The Europe Agreements page 4

Differentials between the CEECs and the EU15 page 5

Factors explaining bilateral trade page 6

Static effects of EU enlargement for the Netherlands page 7 Differences in the extent of bilateral trade as explaining factor page 8

Methodological section

Variables and dataset page 11

Dependent variables page 11

Explanatory variables page 11

Model specification page 12

Overall regressions page 12

Controlling for heteroskedasticity and serial correlation page 14

Country-specific regressions page 15

Results

Overall regressions page 17

Country-specific regressions page 19

Analysis of Dutch export relations page 22

Analysis of Dutch import relations page 23

Conclusion

Literature

Appendix 1: Descriptive Statistics

Appendix 2: Output of the fixed effects regressions as specified in

equation (3)

Appendix 3: Output of the fixed effects regressions as specified in

equation (4)

Appendix 4: Output of the fixed effects regressions as specified in

equation (5)

Appendix 5: Output of the fixed effects regressions as specified in

equation (6)

Appendix 6: Output of the fixed effects regressions as specified in

3

Introduction

‘’Greater economic integration may lead to a race to the top’’ concluded Richard E. Baldwin and Paul Krugman (2004) in their review about agglomeration, integration and tax harmonization. This statement seems very straightforward, though the past decades have been characterized by a heavy discussion about the economic benefits and drawbacks of increased trade liberalization. The

European integration developed quickly in the 1950s, 1960s and 1970s and started to reach higher levels (Baldwin & Krugman, 2004). After several historical landmarks in the 1980s, attention to the establishment of trading arrangements increased in order to stimulate the international economy (Hanson, 1998). The establishment of the European Union (EU) is a very notable development in this process of economic integration. This endeavor to create world’s largest single market (Hanson, 1998) has become a topic of big interest for politicians and economists and is even a common topic of living room discussion. Does the increased trade liberalization by the establishment of the European Union really benefit the member countries as much as initially thought?

One of the most recent expansionary developments in the composition of the EU is its fifth enlargement in May 2004. Cyprus, Czech Republic, Estonia, Hungary, Latvia, Poland, Lithuania, Slovenia, Malta and Slovakia entered the EU. This implies that these ten Central and Eastern European Countries (CEECs) could from then benefit from the elimination of tariffs when trading with other member countries (Turrión & Velázquez, 2004). Most of these countries have a lower per capita income than the original member countries (Turrión & Velázquez, 2004), which partly explains the doubts concerning the benefits of EU enlargement for the eastern member countries. On the other hand, international economic integration is a familiar concept with its roots stemming from the late 1940s and has also been proved to be welfare increasing (Baldwin & Krugman, 2004).

The theory that the disappearance of barriers to trade will boost international trade is well known. One way to look at the benefits of increased trade liberalization is to analyze the effects it has on the magnitude of a trade relationship. In this paper, the development of different trade relationships will be analyzed in order to find out if the trade relationships gradually increased in its extent over time or if the increased integration triggered a leap in the trade relationships. By a leap we mean a sudden increase in the percentage change of the extent of imports and exports. Research will be carried out on the trade relationships between the Netherlands, a country that has been a member of the EU since its establishment, and the CEECs to answer the following research question:

The eastern enlargement of the European Union in 2004: Does a leap in the trade relationship between the Netherlands and the acceded countries exist?

Methodology

This paper starts with a theoretical framework to give some background knowledge about the implications of becoming a member country. We will elaborate the possible consequences of increased trade liberalization. This theoretical framework will provide us with information about which factors are important to include in our model in order to make useful predictions about the extent of the import and export relations. At the end of the theoretical framework, the

developments of the ten examined trade relationships will be presented.

The empirical research will cover a time period of fifteen years in total. To investigate if the percentage change in import and export is similar across the CEECs, we will estimate linear

regression models based on the explanatory variables which appeared to be important from economic literature. Research will be carried out on the entire dataset consisting of the ten CEECs during 1995 to 2009 and on each CEEC separately. By comparing the developments of the extent of the import and export relationships, it will become clear if a leap in the trade relationships between the Netherlands and the acceded countries exists.

4

Theoretical framework

The Europe Agreements

The EU enlargement in May 2004 can be seen as a peak in the ongoing process of economic integration. The Europe Agreements (EA) underlie this enlargement and constituted a legal framework of the accession process. In the 1990s, these agreements had been proclaimed as an initiative for a higher level of integration of the goods markets between the member countries of the EU and the 10 potential new member countries (Egger & Larch, 2011). The EA consist of agreements which ensured the elimination of trade tariffs among the 15 original member countries of the EU and the CEECs (Egger & Larch, 2011). Because during the establishment of these agreements many of the CEECs faced different degrees of problems, the agreements are made specific for the individual countries (Ramsey, 1995). The agreements consist of the following economic principles (Ramsey, 1995):

• A common legal basis and form

• Political dialogue and institution of association • Free movement of trade

• Free movement of workers • The right of establishment • The provision of services

• Competition provisions of legislation • Economic co-operation

• Cultural co-operation • Financial co-operation

Because of the transformation the new member countries had to go through, the accession process started many years before the actual accession in order to fully prepare the new member country. For example, all tariffs expect for agricultural tariffs and tariffs on sensitive products have already been eliminated by the EU since 1997. In 2002, the CEECs removed all their tariffs to the EU (Breuss, 2002). This process of preparation can take more than ten years. In most cases, the

transformation implied moving from a situation where trade is centrally planned to a situation where trade is liberalized (Mardas, 2010). On average, the overall openness of the CEECs to world markets measured as external trade as a percentage of GDP increased from 56% in 1993 to 80% in 2001 (José & Aurora, 2006).

The degree of liberalization the CEECs will experience when entering the EU is characterized by a Common Market. This implies that all tariffs between the member countries are removed, that the countries adopt a common external trade policy for nonmember countries and that there is free movement of capital and labor between member countries. Member countries give up sovereignty in capital flows and labor movements (Appleyard, Field & Cobb, 2009, p. 394).

Much research has been carried out on the effects of the EA on international trade. From the empirical model of bilateral trade by Egger and Larch (2011) it appeared that the EA reduced trade within the CEECs while it stimulated trade between the CEECs and the EU15. This trade redirection effect was much larger for the CEECs than for the fifteen original member countries of the EU (EU15). On average, a CEEC signing the EA with the EU15 experienced an increase in corresponding bilateral trade of 30% (Egger & Larch, 2011). The effects of the EA have had a much larger impact on the GDP of the CEECs than on the GDP of the EU15, which was only slight. On average, annual GDP of the involved CEECs increased by 5% between 1994 and 1999 (Egger & Larch, 2011). After all, Egger and Larch conclude that the cumulative welfare effects of the EA between 1994 and 1999 will be around 3% of the GDP of the ten CEECs and will be around 2% of the GDP of the EU15 (Egger & Larch, 2011), so the overall effect was beneficial for both parties involved. Looking at the actual enlargement instead of the accession process, the gains for the CEECs will be

5 on average ten times as large as the gains for the EU15 (Breuss, 2002). The overall welfare effect will again be positive (Breuss, 2002).

Differentials between the CEECs and the EU15

Before the CEECs joined the EU, large economic differences between the CEECs and the EU15 existed. One of them is income per capita: this measure was much lower in the CEECs than in the EU15. Income per capita in the CEECs in purchasing power parity-terms lies at 40 percent of income per capita in the EU15. Theory asserts that a lower income per capita is due to lower labor

productivity (Baldwin, Francois & Portes, 1997). Increased economic integration could quickly change this. Opening up borders implies an exchange of knowledge and new production

technologies. Because income per capita and thus labor productivity is higher in the EU than in the CEECs, there will be a larger flow of knowledge and production technologies from the EU to the CEECs than the other way around. According to Egger and Larch (2004), this phenomenon partly explains why the CEECs benefited more from the EU enlargement than the EU15. Another reason why income per capita in the CEECs is lower than in the EU15 is because of the worse state of capital stocks in the CEECs (Baldwin et all, 1997). Increased liberalization could also change this in a rapid pace. Installing new machines when capital becomes available would be a relatively simple task because of the high educational level in the CEECs (Baldwin et all, 1997).

A coherent distinction between the CEECs and the EU15 is the difference in economic size: the economic size of the EU15, calculated by Baldwin, François and Portes as differences in income levels multiplied by differences in population, is relatively large compared to the CEECs (Baldwin et all, 1997). Opening up trade barriers expands producer and consumer possibilities, which increases efficiency. This results in higher outputs and an increase in income levels (Baldwin et all, 1997). Because of the relatively moderate size of the CEECs compared to the EU15, opening up trade barriers would imply a relatively much larger increase in the availability of consumer and producer possibilities in the CEECs than in the EU15.This is another phenomenon which explains why the CEECs have benefited more of the EA than the EU15 did.

Looking at the nature of trade between the EU15 and the CEECs, we might also explain the differences in welfare gains. From the Hecksher-Ohlin theorem it follows that a country should export the good whose production is intensive in the factor it is abundant in. A country should import the good whose production is intensive in the factor of which the country has a shortage (Krugman, Obstfeld & Melitz, 2014, p. 127). The EU15 indeed turns out to be a net exporter of machines and equipment because it is abundant in capital. However, because of a certain degree of protection before the accession, the Heckscher-Ohlin theorem doesn’t describe the trade pattern between the EU15 and the CEECs well. Before the enlargement, trade between the EU15 and the CEECs mainly consisted of two-way trade in similar goods (Baldwin et all, 1997). As a consequence of this trade pattern, the countries involved do not take full advantage of economies of scale. Because of free trade, countries generate a comparative advantage and start exporting the goods whose production is intensive in the factors the country disposes of the most. If one country focuses on the production of only these goods, it can produce those goods on a larger scale and become more efficient in this production process because of knowledge spillovers (Krugman et all, 2014, pp. 179-180). If all involved countries adopt this system of specialization, all of the involved countries benefit from economies of scale (Krugman et all, 2014, pp. 179-180). Moreover, specialized knowledge also stimulates innovation (Krugman et all, 2014, p. 182). The increase in efficiency drives down

production costs and increases the extent of international trade: for the same amount of exports, more goods can be imported. Although all of the involved countries in international trade gain from the creation of economies of scale, the effects will be more significant for the CEECs than for the EU15 since the EU15 already disposed over more advanced production technologies than the CEECs in the years before the accession (Baldwin et all, 1997).

6 Factors explaining bilateral trade

Turrión and Velazques (2004) also state that, besides the benefits the accession of the CEECs will have for the enlarged EU as a whole, mainly the CEECs will gain. They emphasize the fact that those gains need not necessarily have to be the same across the CEECs. Although intra-trade will

experience an increase because of closer economic relations, this development in the magnitude of trade might not be similar across the acceded countries (Turrión & Velazques, 2004).

Differences between CEECs may cause differences in the trade relationship between the Netherlands and the CEECs. This analysis points us at which factors are useful to consider when predicting the extent of bilateral trade. As stated before, economic integration is positively related to labor productivity if the candidate country opens up to countries with a higher marginal product of labor (Baldwin, Francois & Portes, 1997). Hence it follows that there is more to gain for a

candidate country with a relatively low marginal product of labor before the accession compared to the EU15 than for a candidate country whose marginal product of labor is already at a comparable level to the EU15 before it becomes a member country. A relatively low marginal product of labor of a candidate country before the accession may cause a larger leap in the trade relationship between the candidate country and the Netherlands regarding the accession.

The differences in economic size between the CEECs and the EU15 partly explain why the EU enlargement creates relatively more opportunities for the CEECs than for the EU15 (Baldwin et all, 1997), though this widening of the range of opportunities is not similar for the ten CEECs involved. Economic size is another indicator for differences across the CEECs regarding the accumulation of import and export levels as a consequence of the eastern enlargement. This indicator combined with the distance between two countries involved in a bilateral trade relationship constitute the Gravity model, a model which is being used very often for analyzing trade patterns. The Gravity model asserts that the extent of bilateral trade is positively related to economic size and negatively related to distance (Bussière, Fidrmuc & Schnatz, 2008). Since the distance between the Netherlands and the candidate country the closest to the Netherlands compared to the candidate country located the furthest from the Netherlands differs by around 600 kilometers, the indicator distance is expected to have an impact on the quantity of products transported between the Netherlands and the CEECs.

Four other factors are often considered to be important to include in the Gravity model. The first factor considers language. Two countries that share the same language are expected to trade more with each other than countries that don’t. The second factor considers territories. Two countries appear to be more closely connected if they were part of a common dominion in the past, which stimulates trade. The third factor considers if countries share a common border. Two

countries that share a common border experience lower transaction costs than two countries that do not, which has a positive effect on bilateral trade. The fourth factor considers trade restrictions: a free trade arrangement will have a positive effect on bilateral trade (Bussière et all, 2008). Of course, the higher the degree of protectionism before the EU accession, the larger the increase in the extent of bilateral trade between the Netherlands and the considering country we will observe in our analysis of the trade relationships.

Within the CEECs, economic uncertainty and instability had a negative effect on

international trade and are caused by unexpected variation in inflation rates and exchange rate volatility (Baldwin et all, 1997). The prospect of becoming a member of the EU decreases economic uncertainty and instability by making investors aware of the economic direction in which the transition country is heading. By becoming a member country of the EU a constraint is placed on arbitrary trade and on indirect tax policy changes which constitutes a downward pressure on inflation rate volatility (Baldwin et all, 1997). The CEECs will also arrive on the path of eventual Economic and Monetary Union (EMU) membership which will decrease economic uncertainty and instability even further (Baldwin et all, 1997). Being a member of the EMU has all the characteristics of a common market like the EU does, but also implies unification of the monetary policy. This decreases the likelihood of asymmetric shocks, which in turn decreases economic uncertainty and instability and increases international trade (Appleyard, Field & Cobb, 2009).

7 It appears that price stability is a prerequisite for rapid economic growth (Gylfason, 1999). High inflation causes a decrease in export volume in proportion to GDP (Gylfason, 1999). We expect the EU accession of the CEECs to decrease inflation rate volatility. The more monetary uncertainty a country experienced before EU accession, the more a country could gain from the accession.

Hacker and Hatemi-J (2004) carried out research on the effect of exchange rate changes on a country’s trade balance and observed a J-curve effect. The J-curve summarizes the effects of

depreciation in the short run and in the long run. Because import and export only slowly react to changes in exchange rates, first a small drop in the export-to-import ratio is observed after depreciation has taken place. It takes some time until the export-to-import ratio and the trade balance will increase. Overall, the long run effect will exceed the short run effect (Hacker & Hatemi-J, 2004).

Most of the research carried out on exchange rate volatility before 1980 points at a significant negative relationship between exchange rate volatility and trade flows. The primary explanation for this finding is the theory of choice under uncertainty: the quantity of exports and imports decreases when an increase in exchange rate uncertainty takes place (De Grauwe & Skudelny, 2000). However, since around 1980 the conclusions of research carried out on the effect of exchange rate volatility on international trade varied widely regarding the significance of the findings, perhaps because during this time period more statistics about volatile exchange rates were available (De Grauwe & Skudelny, 2000). De Grauwe and Skudelny (2000) tested the effect of exchange rate volatility on intra-EU trade flows between 1962 and 1995 by adding a variable of the real exchange rate volatility to the Gravity model. The implemented regressions resulted in a significant negative coefficient for the exchange rate volatility variable. De Grauwe and Skudelny subsequently conclude that complete elimination of exchange rate variability creates new trade flows within the EU.

Static effects of EU enlargement for the Netherlands

Because in this paper the trade relationship of the CEECs with the Netherlands is examined, it is also interesting to consider the effects of the EU enlargement on the Netherlands specifically.

Economic integration implies a different treatment for member countries as for nonmember countries. The EU enlargement therefore leads to a shift in trade patterns between members of the EU and nonmembers. Two effects that directly emerge from the establishment of a group of member countries with reciprocally preferential trade conditions compared to nonmember countries are the trade creation effect and the trade diversion effect (Appleyard et all, 2009, pp. 394-395). The first effect, the trade creation effect, implies that the disappearance of trade barriers increases commercial relations between member countries (Turrión & Velazques, 2004). Trade creation implies a shift from a producer with higher resource costs to a producer in the member country with lower resource costs. This creates beneficial welfare effects (Turrión & Velazques, 2004). For the Netherlands, the trade creation effect could take place if the domestic production of a good could be replaced by cheaper production in one of the CEECs. This would be observable in our analysis of the trade relationships between the Netherlands and the CEECs as an above expected increase in the extent of bilateral trade.

The second effect is called the trade diversion effect. This implies a shift in the origin of a product from a producer in a nonmember country with lower resource costs to a producer in a member country with higher resource costs (Appleyard et all, 2009, pp. 394-395). Lower prices because of the removal of trade tariffs could cause a replacement of exports of an original member country by exports of a recently acceded country (Turrión & Velazques, 2004). The trade diversion effect would be harmful for Dutch GDP levels. Although the trade diversion effect would imply an increase in bilateral trade between the Netherlands and the CEECs, trade diversion has a negative effect on overall welfare and would decrease Dutch GDP. On the long term, this would exercise a downward pressure on the extent of trade between the Netherlands and the CEECs.

8 0 .005 .0 1 .015 0 .005 .0 1 .015 0 .005 .0 1 .015 1995 2000 2005 2010 1995 2000 2005 2010 1995 2000 2005 2010 1995 2000 2005 2010

Czech Republic Hungary Slovakia Estonia

Slovenia Cyprus Malta Poland

Latvia Lithuania

Panel overview

Dutch exports as % of total Dutch exports

Differences in the extent of bilateral trade as explaining factor

The graphs below display that the extent of bilateral trade is different across the investigated trade relationships. In answering our research question, we analyze trade patterns instead of the

magnitude of trade. However, the initial extent of bilateral trade, by which we mean the extent of trade before the CEECs entered the transition process, might also be an important factor in explaining trade patterns.

Figure 1: Dutch exports as percentages of total Dutch exports (Panel overview)1

Table 1: Dutch exports to CEECs in thousands (USD) and Dutch exports to CEECs as percentages of total Dutch exports (Panel overview)

1 _{The numbers on the y-axis multiplied by hundred are the percentages.}

Dutch exports to CEECs in thousands (USD) Dutch exports to CEECs as % of total Dutch exports

1995 2009 Percentage change 1995 2009 Percentage change

Czech Republic 555.041 6.078.214 995.09 % 0.2880 % 0.9912 % 244.13 % Hungary 459.128 2.697.513 487.53 % 0.2383 % 0.5419 % 127.46 % Slovakia 151.988 1.146.544 654.36 % 0.0789 % 0.2303 % 192.05 % Estonia 65.966 321.382 387.19 % 0.0342 % 0.0646 % 88.61 % Slovenia 206.011 679.030 229.61 % 0.1069 % 0.1364 % 27.61 % Cyprus 74.832 393.758 426.19 % 0.0388 % 0.0791 % 103.71 % Malta 78.909 298.433 278.20 % 0.0409 % 0.0560 % 46.42 % Poland 1.259.994 7.519.788 496.81 % 0.6538 % 1.5108 % 131.96 % Latvia 141.065 281.532 99.58 % 0.0732 % 0.0566 % -22.74 % Lithuania 92.966 526.254 466.07 % 0.0482 % 0.1057 % 119.15 %

9 0 .005 .0 1 0 .005 .0 1 0 .005 .0 1 1995 2000 2005 2010 1995 2000 2005 2010 1995 2000 2005 2010 1995 2000 2005 2010

Czech Republic Hungary Slovakia Estonia

Slovenia Cyprus Malta Poland

Latvia Lithuania

Panel overview

Dutch imports as % of total Dutch imports

Dutch exports to the CEECs significantly increased between 1995 and 2009. Apparently, the values of Dutch exports in 1995 to Poland, the Czech Republic and Hungary are the largest compared to the values of Dutch exports in 1995 to other CEECs. Between 1995 and 2009, Dutch exports to the Czech Republic almost multiplied by eleven. Dutch exports to Poland and Hungary almost multiplied by six. We indeed observe a large percentage change in Dutch exports to Poland, the Czech Republic and Hungary as a percentage of total Dutch exports when we compare the observation in 1995 with the observation in 2009. However, while Dutch exports to Hungary as a percentage of total Dutch exports more than doubled, Dutch exports to Slovakia as a percentage of total Dutch exports almost tripled. Dutch exports to Slovakia multiplied by more than seven between 1995 and 2009. This implies that not only in the case of the relatively large trade partners of the Netherlands we

experience a large increase in Dutch exports as a percentage of total Dutch exports. Dutch exports to Lithuania and Cyprus multiplied by more than four. Dutch exports to Lithuania and Cyprus as a percentage of total Dutch exports more than doubled. The percentage change in Dutch exports to Estonia, Slovenia, Malta and Latvia as a percentage of total Dutch exports was relatively modest. The case of Latvia is the only one for which the Dutch exports as a percentage of total Dutch exports decreased. Although it seems that the initial extent of a trade relationship is related to the

development of the trade relationship regarding increased trade liberalization, we should be aware of the fact that also smaller trade relationships can be influenced notably by the EU enlargement.

Figure 2: Dutch imports as percentages of total Dutch imports (Panel overview)

Looking at the Dutch imports from the CEECs we reach a somewhat similar conclusion. Dutch imports from Poland, Hungary, the Czech Republic and Latvia are relatively of the largest extent compared to Dutch imports from other CEECs in 1995. Again, also the percentage change in Dutch imports from Slovakia is large compared to the percentage change in Dutch imports from other CEECs. Dutch imports from the Czech Republic, Hungary, Poland, Slovakia and Cyprus as a

10 Table 2: Dutch imports from CEECs in thousands (USD) and Dutch imports from CEECs as

percentages of total Dutch imports (Panel overview)

change in Dutch imports from Poland as a percentage of total Dutch imports is relatively modest. Looking at Dutch import relations, we observe more cases where Dutch trade with a CEEC as a percentage of total Dutch trade decreased compared to Dutch export relations: Dutch imports from Slovenia, Latvia and Lithuania as a percentage of total Dutch imports decreased when we compare the observation in 1995 with the observation in 2009. Especially in the case of Latvia this is a notable development, because the Netherlands imported to a relatively large extent from Latvia in 1995. Although Latvia seemed to be a large trade partner of the Netherlands looking at import relations, increased trade liberalization did not contribute to an increase in the amount of import in this case. The percentage change in Dutch imports from Estonia and Malta as a percentage of total Dutch imports is relatively modest.

Dutch exports to and imports from the Czech Republic as a percentage of total Dutch exports and imports increased relatively the most between 1995 and 2009. Also in the case of Poland and the Czech Republic we observe a relatively large percentage change in bilateral trade with the Netherlands when we look at export relations as well as import relations. Dutch exports to and imports from Slovakia and Cyprus as a percentage of total Dutch exports and imports also increased to a large extent, though the trade relations between the Netherlands and Slovakia and Cyprus are relatively small compared to the trade relations between the Netherlands and the Czech Republic, Hungary and Poland.

The Czech Republic, Hungary and Poland are indeed the countries that gained the most from the EU enlargement. In a ten year time period, the Czech Republic increased its real GDP by around 5.5%. Hungary and Poland increased their real GDP by around 8.5% (Breuss, 2002). This is what we call trade creation: intra-trade between the new EU member countries and the EU15 increased because trade costs decreased due to the elimination of import tariffs.

Dutch imports from CEECs in thousands (USD) Dutch imports from CEECs as % of total Dutch imports

1995 2009 Percentage change 1995 2009 Percentage change

Czech Republic 358.121 4.880.745 1262.88 % 0.2055 % 1.1007 % 435.54 % Hungary 402.182 2.369.724 489.22 % 0.2308 % 0.5344 % 131.53 % Slovakia 143.356 1.426.499 895.07 % 0.0823 % 0.3217 % 291.01 % Estonia 161.667 443.075 174.07 % 0.0928 % 0.0999 % 7.69 % Slovenia 122.118 232.778 90.62 % 0.0701 % 0.0525 % -25.1 % Cyprus 17.831 94.335 429.05 % 0.0102 % 0.0213 % 107.89 % Malta 13.460 48.016 256.73 % 0.0077 % 0.0108 % 40.18 % Poland 943.784 4.244.203 349.70 % 0.5417 % 0.9572 % 76.71 % Latvia 507.654 169.290 -66.65 % 0.2914 % 0.0382 % -86.90 % Lithuania 247.711 333.976 34.82 % 0.1422 % 0.0753 % -47.02 %

11

Methodological section

In the empirical section we examine the relationship between the annual Dutch imports and exports as percentages of the total annual Dutch imports and exports respectively and the explanatory variables which appeared relevant from economic literature. First we carry out an overall regression of the import and export percentages on the explanatory variables based on the complete panel dataset to discover if we can conclude that a leap in the trade relationships exists when ignoring country specific differences. Subsequently, we carry out a regression of the annual Dutch imports and exports as percentages of the total annual Dutch imports and exports per CEEC on the

explanatory variables measured per CEEC to become able to compare the similarities and differences across the ten examined trade relationships.

Variables and dataset

The dataset is self-composed for the use of writing this bachelor’s thesis. The data covers a time period from 1995 to 2009 and contains cross-sectional observations across ten countries (Czech Republic, Hungary, Slovakia, Estonia, Slovenia, Cyprus, Malta, Poland, Latvia, and Lithuania). The dataset is a balanced panel which thus contains a total of 150 observations.

Dependent variables

The dependent variables are annual Dutch exports of goods to each CEEC as a percentage of the total annual Dutch exports of goods and the annual Dutch imports of goods from each CEEC as a percentage of the total annual Dutch imports of goods. We extract data concerning annual Dutch exports to and imports from each CEEC during 1995 to 2009 and data concerning total annual Dutch exports and imports during 1995 to 2009 from the iLibrary of the Organization for Economic Co-operation and Development (OECD). We use these numbers to compute the annual Dutch imports and exports as percentages of the total annual Dutch imports and exports, which we use as dependent variables.

Explanatory variables

The first independent variable we include in our model is Gross Domestic Product (GDP) per capita in current US dollars. The data is extracted from The World Bank DataBank. GDP per capita is measured as GDP divided by mid-year population. The values are computed without making adjustments for depreciation of assets or depletion of natural resources.

The second independent variable is total population per country. The data is extracted from The World Bank DataBank. The annual numbers are mid-year estimates and include all residents except refugees who are not permanently settled in the concerning country.

The third independent variable is the exchange rate between the Dutch currency and the currencies of the CEECs. The leap in the exchange rate which we observe if we do take into account the introduction of the Euro causes a discontinuity in the development of the exchange rate which we cannot explain based on our model. Because the developments in the exchange rates are more valuable than the exchange rates itself when we try to explain patterns in bilateral trade, we do not take into account the introduction of the Euro. We extract data on the exchange rate between the Dutch Guilder and the currencies of the CEECs from OANDA. The historical exchange rates measure the value of the foreign currency against the Dutch Guilder as the base currency. An increase in the historical exchange rate thus implies an increase in the value of the Dutch Guilder compared to the value of the foreign currency.

The fourth independent variable is the inflation rate in the CEECs as a percentage of the inflation rate in the Netherlands per year. We compute this data as inflation measured in consumer price index per CEEC divided by the inflation measured in consumer price index in the Netherlands in the same year. We extract the annual inflation rates for the CEECs and the Netherlands from The World Bank DataBank. The inflation rates are measured by The World Bank as the annual percentage change in the cost to the consumer of acquiring a basket of goods or services.

12 For the fifth, sixth and seventh independent variable we extract data from The World Bank DataBank on the added values as percentage of the GDP per country in the agricultural sector, industrial sector and manufacturing sector respectively. It concerns the net output per sector after subtracting intermediate inputs from the total of all outputs. No deductions for deprecation of assets or depletion of natural resources are made.

In a separate regression model we demonstrate that it is possible to include dummy

variables for every year included in the panel dataset. These dummy variables control for the effects that are constant across the panels but that vary over time. By extending our model like this, the fixed effects regression picks up those effects including for example movements of the world economy.

Model specification Overall regressions

The multiple regression model we will use for the implementation of the overall regression of the annual Dutch exports to the CEECs as a percentage of total annual Dutch exports on GDP per capita, total population per country, exchange rate, inflation rate in the CEEC relative to the inflation rate in the Netherlands and added value as a percentage of GDP per CEEC in the agricultural, industrial and manufacturing sector is as follows:

𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽1 𝐺𝐺𝐺𝐺𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽2 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽3 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽4 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖

+ 𝛽𝛽5 𝐸𝐸𝐺𝐺𝐸𝐸𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽6 𝐼𝐼𝐼𝐼𝐺𝐺𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽7 𝑀𝑀𝐸𝐸𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖+ 𝑢𝑢𝑖𝑖𝑖𝑖

(1) We use the model specification based on the same explanatory variables when we estimate the annual Dutch imports as a percentage of total annual Dutch imports per CEEC:

𝐼𝐼𝑀𝑀𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽1 𝐺𝐺𝐺𝐺𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽2 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽3 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽4 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽5 𝐸𝐸𝐺𝐺𝐸𝐸𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽6 𝐼𝐼𝐼𝐼𝐺𝐺𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽7 𝑀𝑀𝐸𝐸𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖+ 𝑢𝑢𝑖𝑖𝑖𝑖

(2) In regression model (1), 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 is the annual Dutch exports as a percentage of total annual Dutch exports, 𝐺𝐺𝐺𝐺𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 is the GDP per capita measured in US dollars, 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 is the

exchange rate between the Dutch Guilder and the foreign currency, 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 is the inflation rate in the CEEC divided by the inflation rate in the Netherlands, 𝐸𝐸𝐺𝐺𝐸𝐸𝐼𝐼𝑖𝑖𝑖𝑖 is the added value as a

percentage of GDP in the agricultural sector, 𝐼𝐼𝐼𝐼𝐺𝐺𝐼𝐼𝑖𝑖𝑖𝑖 is the added value as a percentage of GDP in the industrial sector and 𝑀𝑀𝐸𝐸𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖 is the added value as a percentage of GDP in the manufacturing sector. Our estimates are based on a dataset including observations across ten countries during fifteen years, so 𝑖𝑖 can take the value one to ten and 𝑡𝑡 can take the value one to fifteen. In regression model (2) the dependent variable is 𝐼𝐼𝑀𝑀𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖, which stands for the annual Dutch imports as a percentage of total annual Dutch imports. The specification of the explanatory variables in

regression model (2) is the same as the specification of the explanatory variables in regression model (1).

Because we deal with a panel dataset, we first carry out a Hausman test in order to decide if the fixed effects model or the random effects model should be used for an overall regression, meaning a regression including the time-series data as well as the cross-sectional data. No time-invariant though relevant explanatory variables are yet included because these drop out in the fixed effects regression. We compare the estimated coefficients by the fixed effects model with the same

estimated coefficients by the random effects model. If we get a significant P-value, we reject the null hypothesis that the estimated coefficients are the same in both regressions.

13 Table 3: Hausman test comparing the coefficients of the fixed effects regression with the coefficients of the random effects regression with Dutch exports to each CEEC as a percentage of

the total Dutch exports as dependent variable

Based on the above findings with annual Dutch exports to each CEEC as a percentage of the total annual Dutch exports as dependent variable in the regression equations, we reject the null hypothesis. The country specific error terms are correlated with the regressors, implying endogeneity. Therefore we use a fixed effects regression. The fixed effects model controls for omitted variables which effects differ across the CEECs but are constant over time. Conducting a fixed effects regression is comparable to creating dummy variables for each country which absorbs these constant, country-specific explanatory variables (Stock & Watson, 2012, p. 396). This is very useful since these time-invariant variables appear to be of interest from economic literature while explaining bilateral trade. For example cultural differences, religious differences and differences in distance between the Netherlands and the CEECs are now controlled for. However, using the fixed effects model does not completely eliminate the threat of omitted variable bias.

Besides our statistical proof that using the fixed effects model in our research is the best option, it seems logic to use this model specification from economic reasoning on the purpose of this methodological section. We are interested in the time-varying variables on the basis of which we try to explain patterns in bilateral trade. The fixed effects model analyzes the effect of these time-varying explanatory variables on the dependent variable. Variables which remain constant during the investigated time period will not explain discontinuities in the trade relationships between the Netherlands and the CEECs due to EU accession.

If we conduct the Hausman test with annual Dutch imports to each CEEC as a percentage of the total annual Dutch imports as dependent variable in the regression equations our conclusion might be different. In the table below it becomes clear that in this case we get an insignificant P-value as result of the Hausman test, which implies that the random effects model would be save to use. To be able to make valuable comparisons between the developments in import and export relations, we prefer to use the same model specification for the annual Dutch exports to each CEEC as a percentage of the total annual Dutch exports and the annual Dutch imports to each CEEC as a percentage of the total annual Dutch imports. Therefore, we will adhere to the fixed effects model for the overall regression.

Fixed effects Random effects Difference GDPCEEC 1.54E-07 2.19E-07 3.46E-08 POPCEEC -4.54E-09 2.54E-10 -4.80E-09 EXRATE -9.39E-06 -.0000157 6.27E-06 INFLRATE -.0001139 -.0001146 -7.01E-07 AGRI .0002736 -.0000181 .0002917 INDU -.0002293 -.0002361 6.79E-06 MANU .0005743 .0005559 .0000184 Chi2(6)=45.05 Prob>Chi2=0.0000

14 Table 4: Hausman test comparing the coëfficients of the fixed effects regression with the coefficients of the random effects regression with Dutch imports to each CEEC as a percentage of

the total Dutch imports as dependent variable

The dependent variable, the annual Dutch imports and exports as percentages of the total annual Dutch imports and exports respectively, will be influenced directly by phenomena that vary over time, for example by fluctuations in the Dutch economy or fluctuations in the world economy. If we create year dummies to capture those effects that vary over time but that have the same effect on each CEEC, we estimate the following regression models2_:

𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽1 𝐺𝐺𝐺𝐺𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽2 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽3 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽4 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽5 𝐸𝐸𝐺𝐺𝐸𝐸𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽6 𝐼𝐼𝐼𝐼𝐺𝐺𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽7 𝑀𝑀𝐸𝐸𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖+ 𝜆𝜆𝑖𝑖+ 𝑢𝑢𝑖𝑖𝑖𝑖 (fixed effects) (3) 𝐼𝐼𝑀𝑀𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽1 𝐺𝐺𝐺𝐺𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽2 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽3 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽4 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽5 𝐸𝐸𝐺𝐺𝐸𝐸𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽6 𝐼𝐼𝐼𝐼𝐺𝐺𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽7 𝑀𝑀𝐸𝐸𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖+ 𝜆𝜆𝑖𝑖+ 𝑢𝑢𝑖𝑖𝑖𝑖 (fixed effects) (4) When we include the time variable and add an accession effect for the years after the accession including the year of accession itself and an interaction effect to equation (3) and equation (4) we become able to determine if the EU accession has significantly influenced our dependent variables. By extending our regression models this way, we estimate the following regression models3_:

𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽1 𝐺𝐺𝐺𝐺𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽2 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽3 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽4 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽5 𝐸𝐸𝐺𝐺𝐸𝐸𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽6 𝐼𝐼𝐼𝐼𝐺𝐺𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽7 𝑀𝑀𝐸𝐸𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖+ 𝜆𝜆 𝐸𝐸𝐼𝐼𝑀𝑀𝐸𝐸 + 𝛾𝛾 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐴𝐴𝐴𝐴𝐼𝐼𝐸𝐸𝐼𝐼 + 𝛿𝛿 𝐸𝐸𝐼𝐼𝑀𝑀𝐸𝐸 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐴𝐴𝐴𝐴𝐼𝐼𝐸𝐸𝐼𝐼 + 𝑢𝑢𝑖𝑖𝑖𝑖 (fixed effects) (5) 𝐼𝐼𝑀𝑀𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽1 𝐺𝐺𝐺𝐺𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽2 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽3 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽4 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽5 𝐸𝐸𝐺𝐺𝐸𝐸𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽6 𝐼𝐼𝐼𝐼𝐺𝐺𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽7 𝑀𝑀𝐸𝐸𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖+ 𝜆𝜆 𝐸𝐸𝐼𝐼𝑀𝑀𝐸𝐸 + 𝛾𝛾 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐴𝐴𝐴𝐴𝐼𝐼𝐸𝐸𝐼𝐼 + 𝛿𝛿 𝐸𝐸𝐼𝐼𝑀𝑀𝐸𝐸 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐴𝐴𝐴𝐴𝐼𝐼𝐸𝐸𝐼𝐼 + 𝑢𝑢𝑖𝑖𝑖𝑖 (fixed effects) (6)

2 _{The results of the fixed effects regression models as specified in equation (3) and equation (4) are observable}

in Appendix 2 and Appendix 3.

3 _{The results of the fixed effects regression models as specified in equation (5) and equation (6) are observable}

in Appendix 4 and Appendix 5.

Fixed effects Random effects Difference GDPCEEC 1.82E-07 1.65E-07 1.73E-08 POPCEEC 2.20E-10 1.72E-10 4.84E-11 EXRATE .0000147 7.69E-06 7.01E-06 INFLRATE -.0001156 -.0001133 -2.31E-06 AGRI .0003167 .0002953 .0000214 INDU -.0001021 -.0000841 -.000018 MANU .0003832 .000326 .0000572 Chi(2)5=3.11 Prob>Chi2=0.6824

15 Controlling for heteroskedasticity and serial-correlation

It seems reasonable that the variance of the error term of the annual Dutch imports and exports as percentages of the total annual Dutch imports and exports per CEEC is not constant during our invested time period. For example, by comparing a relatively large acceded country like Poland with a relatively small acceded country like Malta, we don’t expect the variation in bilateral trade

between the Netherlands and Poland to be similar to the variation in bilateral trade between the Netherlands and Malta. When we carry out the modified Wald test on groupwise heteroskedasticity, we indeed get a significant result4 _{whereby we reject our null hypothesis of constant error terms.}

Therefore, we should add the robust-option to our fixed effects regression to control for heteroskedasticity. Because we deal with a relatively small panel dataset with a relatively small panel as well as a relatively limited time period, the robust-option unfortunately doesn’t work well. We do not include the robust-option in our fixed effects regression because this will negatively influence the usability of our model5_.

Country-specific regressions

In order to explore the differences and similarities between the ten separate trade relationships, the regressions are also carried out for each separate trade relationship. We use the same variables as used in the overall regression as specified in equation (1) and equation (2). Again, we do not take into account routines to control for heteroskedasticity because our dataset per country is very limited.

The multiple regression model we use for the implementation of the regressions per CEEC of the annual Dutch exports as a percentage of total annual Dutch exports per CEEC on GDP per capita, total population per country, exchange rate, inflation rate in the CEEC relative to the inflation rate in the Netherlands and added value as a percentage of GDP per CEEC in the agricultural, industrial and manufacturing sector is as follows:

𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽1 𝐺𝐺𝐺𝐺𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽2 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽3 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽4 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽5 𝐸𝐸𝐺𝐺𝐸𝐸𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽6 𝐼𝐼𝐼𝐼𝐺𝐺𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽7 𝑀𝑀𝐸𝐸𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖+ 𝑢𝑢𝑖𝑖𝑖𝑖

(Time series) (7) The multiple regression model with the annual Dutch imports as a percentage of total annual Dutch imports per CEEC as dependent variable is as follows:

𝐼𝐼𝑀𝑀𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽1 𝐺𝐺𝐺𝐺𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽2 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽3 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽4 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽5 𝐸𝐸𝐺𝐺𝐸𝐸𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽6 𝐼𝐼𝐼𝐼𝐺𝐺𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽7 𝑀𝑀𝐸𝐸𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖+ 𝑢𝑢𝑖𝑖𝑖𝑖

(Time series) (8) To analyze the effect of the transition process and the accession, we again include a time variable, an accession effect and an interaction effect. To carry out this regression properly, we add the time variable, the accession effect and the interaction effect after eliminating insignificant coefficients6_.

4 _{Outcome modified Wald test: Prob>F = 0.0000}

5 _{We should be aware of the fact that the estimates of the fixed effects regression without controlling for}

heteroskedasticity might be incorrect.

6 _{To determine which coefficients should be eliminated from equation (7) and (8), we carry out fixed effects}

regressions on the entire panel dataset based on the equations as specified in (7) and (8). The results of these fixed effects regressions are observable in Appendix 5. When we determine which coefficients should be eliminated from equation (7) and (8) we assume a significance level of 5%.

16 The equations become specified as follows7_:

𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽1𝐺𝐺𝐺𝐺𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽2 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽3 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽4 𝐸𝐸𝐺𝐺𝐸𝐸𝐼𝐼𝑖𝑖𝑖𝑖 + 𝛽𝛽5 𝐼𝐼𝐼𝐼𝐺𝐺𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽6 𝑀𝑀𝐸𝐸𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖 + 𝜆𝜆 𝐸𝐸𝐼𝐼𝑀𝑀𝐸𝐸 + 𝛾𝛾 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐴𝐴𝐴𝐴𝐼𝐼𝐸𝐸𝐼𝐼 + 𝛿𝛿 𝐸𝐸𝐼𝐼𝑀𝑀𝐸𝐸 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐴𝐴𝐴𝐴𝐼𝐼𝐸𝐸𝐼𝐼 + 𝑢𝑢𝑖𝑖𝑖𝑖 (Time series) (9) 𝐼𝐼𝑀𝑀𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽1𝐺𝐺𝐺𝐺𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 + 𝛽𝛽2 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖+ 𝛽𝛽3 𝐸𝐸𝐺𝐺𝐸𝐸𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽5 𝑀𝑀𝐸𝐸𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖+ 𝜆𝜆 𝐸𝐸𝐼𝐼𝑀𝑀𝐸𝐸 + 𝛾𝛾 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐴𝐴𝐴𝐴𝐼𝐼𝐸𝐸𝐼𝐼 + 𝛿𝛿 𝐸𝐸𝐼𝐼𝑀𝑀𝐸𝐸 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐴𝐴𝐴𝐴𝐼𝐼𝐸𝐸𝐼𝐼 + 𝑢𝑢𝑖𝑖𝑖𝑖 (Time series) (10) We also carry out regressions of our dependent variables solely on the time variable, the accession effect and the interaction effect8_.

𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝜆𝜆 𝐸𝐸𝐼𝐼𝑀𝑀𝐸𝐸 + 𝛾𝛾 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐴𝐴𝐴𝐴𝐼𝐼𝐸𝐸𝐼𝐼 + 𝛿𝛿 𝐸𝐸𝐼𝐼𝑀𝑀𝐸𝐸 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐴𝐴𝐴𝐴𝐼𝐼𝐸𝐸𝐼𝐼 + 𝑢𝑢𝑖𝑖𝑖𝑖

(Time series) (11) 𝐼𝐼𝑀𝑀𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝜆𝜆 𝐸𝐸𝐼𝐼𝑀𝑀𝐸𝐸 + 𝛾𝛾 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐴𝐴𝐴𝐴𝐼𝐼𝐸𝐸𝐼𝐼 + 𝛿𝛿 𝐸𝐸𝐼𝐼𝑀𝑀𝐸𝐸 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐴𝐴𝐴𝐴𝐼𝐼𝐸𝐸𝐼𝐼 + 𝑢𝑢𝑖𝑖𝑖𝑖

(Time series) (12)

7 _{Our dataset is too small to get useful results out of the regressions as specified in equation (9) and equation}

(10). Hence, we do not further pay attention to the outcomes of these regressions.

17

Results

Overall regressions

If we carry out the regressions on the entire panel dataset, a regression over the years 1995 to 2004 and a regression over the years 2004 to 2009 and combine these fitted values in one graph, we are able to observe a discontinuity in the developments of the trade relationships. The fitted values of the regressions as specified in equations (3) and (4) are observable in figure 3 and figure 4. The output of these regressions is observable in Appendix 2 and Appendix 3.

From the output of the regression as specified in equation (3) with Dutch exports to the CEECs as a percentage of total Dutch exports as dependent variable, it follows that the coefficient on the population, the added value in the agricultural sector and the added value in the manufacturing sector are significant9_{. The coefficient on the added value in the industrial sector is insignificant.}

From the economic literature it follows that large income differences between the CEECs and the EU15 exist, which could be useful in explaining bilateral trade. Though, the coefficient on the GDP of the CEECs appears to be insignificant10_{. Also the coefficients on the exchange rate and the inflation}

rate are insignificant, although we expected these measures to be important in explaining bilateral trade based on the economic literature.

Figure 1 graphically shows that a leap exists for the annual Dutch exports as a percentage of total annual Dutch exports per CEEC and that annual Dutch exports to the CEECs as a percentage of total Dutch exports strongly increased between 1995 and 2009. However, if we carry out the regression as specified in equation (5) and perform a test on the significance of the accession effect and the interaction effect11_{, the accession effect and the interaction effect appear to be insignificant.}

Based on this finding, we conclude that EU accession has not significantly influenced the Dutch exports to the CEECs as a percentage of total Dutch exports12_.

Figure 3

9 _{We use a significance level of 5%.}

10 _{Our limited panel dataset might be the cause of insignificant results when the explanatory variable is}

actually of significant importance in the regression.

11 _{We test if the accession effect and the interaction effect are significant at the same time.}

12 _{Note that this conclusion is based on an overall regression. The accession effect and the interaction effect}

18 Although we can conclude that increased trade liberalization gave a boost to the Dutch export relations with the CEECs, the accession didn’t significantly influence Dutch exports as a percentage of total Dutch exports. We indeed observe a decrease in the growth rate of Dutch exports as a percentage of total Dutch exports after the accession in 2004.

Figure 4

From the output of the regression as specified in equation (4) with Dutch imports to the CEECs as a percentage of total Dutch imports as dependent variable, it follows that only the coefficients on the added value in the agricultural sector and the added value in the manufacturing sector are

significant. Again, for some explanatory variables which appeared to be of importance when we try to explain bilateral trade from economic literature, we get insignificant results13_.

In Figure 2 we observe a pattern that looks like a short downfall. From the graph it appears that the Dutch imports per CEEC as percentages of total Dutch imports increased over the total time period of the regression, but it experienced a discontinuity at the moment of accession. We should bear in mind that our dependent variable concerns values of imports and exports per CEEC as percentages of the values of total Dutch imports and exports. This implies that a downfall of the value of our dependent variable might imply that the increase in the value of the total annual Dutch imports of goods exceeded the increase in the annual Dutch imports of goods from the CEECs on average.

Although the pattern of the discontinuity is not as expected from economic literature, we observe an increase in the growth rate of Dutch imports from the CEECs as a percentage of total Dutch imports. When we carry out the regression as specified in equation (6) and perform a test on the significance of the accession effect and the interaction effect, the accession effect and the interaction effect appear to be significant14_{. Based on this finding, we conclude that EU accession}

has significantly influenced Dutch imports from the CEECs as a percentage of total Dutch imports.

13 _{Our limited panel dataset might again be the cause of insignificant results when the explanatory variable is}

actually of significant importance in the regression.

14 _{Again we carried out a test on the significance of the accession effect and the interaction effect at the same}

19 .00 03 .00 04 .00 05 .00 06 .00 07 .00 08 e x po rt s as pe rc e nt ag es 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch exports to Cyprus as % of total Dutch exports

0 .00 01 .00 02 .00 03 .00 04 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch imports from Cyprus as % of total Dutch imports

0 .00 5 .01 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch imports from Czech Republic as % of total Dutch imports

.00 2 .00 4 .00 6 .00 8 .01 .01 2 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch exports to Czech Republic as % of total Dutch exports

When we compare the fitted value line ‘1995-2009’ of figure 1 with the same fitted value line of figure 2, we can also conclude that increased trade liberalization induced a larger increase in the percentage change of Dutch exports to the CEEC than in the percentage change of Dutch imports per CEEC. Over a time period of fifteen years, the Dutch exports to the CEECs as a percentage of total Dutch exports grew almost four times as large. The Dutch imports from the CEECs as a percentage of total Dutch imports did not even double in the same time period.

Country-specific regressions

Figure 5 Figure 6

20 0 .00 1 .00 2 .00 3 .00 4 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch imports from Latvia as % of total Dutch imports

.00 05 .00 06 .00 07 .00 08 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch exports to Latvia as % of total Dutch exports

.00 04 .00 06 .00 08 .00 1 .00 12 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch exports to Estonia as % of total Dutch exports

.00 05 .00 1 .00 15 .00 2 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch imports from Estonia as % of total Dutch imports

.00 2 .00 3 .00 4 .00 5 .00 6 .00 7 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch exports to Hungary as % of total Dutch exports

.00 2 .00 3 .00 4 .00 5 .00 6 po ts as pe c e tag es 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch imports from Hungary as % of total Dutch imports

Figure 9 Figure 10

Figure 11 Figure 12

21 .00 04 .00 05 .00 06 .00 07 im po rt s as pe rc e nt ag es 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch imports from Slovenia as % of total Dutch imports

.00 6 .00 8 .01 .01 2 .01 4 .01 6 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch exports to Poland as % of total Dutch exports

.00 5 .00 6 .00 7 .00 8 .00 9 .01 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch imports from Poland as % of total Dutch imports

.00 04 .00 06 .00 08 .00 1 .00 12 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch exports to Lithuania as % of total Dutch exports

.00 06 .00 08 .00 1 .00 12 .00 14 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch imports from Lithuania as % of total Dutch imports

.00 08 .00 1 .00 12 .00 14 .00 16 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch exports to Slovenia as % of total Dutch exports

Figure 15 Figure 16

Figure 17 Figure 18

22 .00 03 .00 04 .00 05 .00 06 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch exports to Malta as % of total Dutch exports

.00 00 2 .00 00 4 .00 00 6 .00 00 8 .00 01 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch imports from Malta as % of total Dutch imports

.00 05 .00 1 .00 15 .00 2 .00 25 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch exports to Slovakia as % of total Dutch exports

0 .00 1 .00 2 .00 3 p p g 1995 2000 2005 2010 year 1995-2009 2004-2009 1995-2004 Observations

Dutch imports from Slovakia as % of total Dutch imports

Figure 21 Figure 22

Figure 23 Figure 24

Analysis of Dutch export relations15

Looking at the fitted value lines ‘1995-2009’ of each separate trade relationship, we conclude that Dutch exports to the CEECs as a percentage of total Dutch exports increased between 1995 and 2009. In our introduction we stated that the relatively largest trade partners of the Netherlands, namely the Czech Republic, Hungary and Poland, gained the most from the EU enlargement. We indeed observe a leap in the trade relationship between the Netherlands and the Czech Republic. The combined effect of the accession effect and the interaction effect is significant16_{. The Dutch}

export relationship with the Czech Republic was significantly influenced by increased trade liberalization and the accession in 2004 caused a leap in Dutch exports to the Czech Republic as a percentage of total Dutch exports. The interaction effect is insignificant, which implies that the growth rate of Dutch exports to the Czech Republic as a percentage of total Dutch exports was not

15 _{We base the analysis on the outcome of the regressions as specified in equation (11). In this regression, we}

only include the time effect, the accession effect and the interaction effect as explanatory variables.

23 significantly influenced by the EU enlargement. In figure 7, we even observe a lower slope coefficient of the fitted value line ‘2004-2009’ compared to slope coefficient of the fitted value line ‘1995-2004’

In the case of Hungary, we observe significance for the accession effect and the interaction effect. The accession had a significant impact on this export relation, although from figure 11 it appears that the development after the accession is negative. The accession effect and the interaction effect are also significant in the case of Poland. As immediately observable from figure 15, Dutch exports to Poland as a percentage of total Dutch exports increased to a larger extent in the years after the accession than in the years before the accession. We cannot observe a leap in the trade relationship in figure 15, though from the regression it follows that the accession significantly influenced Dutch exports to Poland as a percentage of total Dutch exports.

Although mainly the three largest trade partners of the Netherlands gained of EU accession compared to the other CEECs, some relatively small trade relationships are also significantly

influenced by EU accession. As stated in the introduction, bilateral trade between the Netherlands and Slovakia increased relatively to a large extent between 1995 and 2009. We indeed get significant results for the accession in the case of Slovakia. In figure 24 we observe a leap in the trade

relationship at the time of the accession. For Dutch imports from Malta as a percentage of total Dutch imports we also get significant results for the accession effect and the interaction effect, but from figure 21 it follows that this is the case because of an outlier in 2009. If we do not include the observation in 2009 in our dataset, we get insignificant coefficients for the Dutch export relation with Malta. We don’t get significant results regarding the accession for Dutch exports to Estonia, Slovenia, Cyprus, Latvia and Lithuania as a percentage of total Dutch exports.

EU enlargement has increased opportunities for the Netherlands regarding an enlargement of sale opportunities. Since the CEECs entered the transition process in 1995, a process of tariff reduction took place in a rapid pace. For the four relatively largest export partners of the

Netherlands in 199517_{, we indeed observe a significant effect of the accession on Dutch exports as a}

percentage of total Dutch exports. For the relatively small trade relationships we do not conclude that EU accession had a significant effect on Dutch exports as a percentage of total Dutch exports.

Analysis of Dutch import relations18

From our introduction it appeared that Dutch imports from the Czech Republic, Slovakia, Hungary, Cyprus and Poland increased relatively to the largest extent. For the Czech Republic and Poland we find a significant accession effect and interaction effect. In figure 8 and figure 16 the positive influence of increased trade liberalization is clearly observable: the slope of the fitted value line increased after EU accession and Dutch imports from the Czech Republic and Poland as a percentage of total Dutch imports increased to a large extent. For Dutch imports from Hungary as a percentage of total Dutch imports, we do not find significant results. From figure 12 it appears that the EU accession didn’t induce an increase in Dutch imports from Hungary as a percentage of total Dutch imports. Dutch imports from Slovakia as a percentage of total Dutch imports are significantly influenced by EU accession. Figure 24 displays the increase in the slope of the fitted value line and the significant increase in Dutch imports from Slovakia as a percentage of total Dutch imports between 1995 and 2009. For Cyprus, we also find significant results regarding the accession. From figure 6 is appears that after the accession Dutch imports from Cyprus as a percentage of total Dutch imports increased, though this development reversed in the last years of the investigated time period.

For Estonia, Slovenia, Latvia, Lithuania, and Malta we observe a decreasing fitted value line for ‘1995-2009’. However, in the case of Latvia and Lithuania, EU accession has smoothed this downward development. In the case of Estonia, Slovenia and Latvia we get significant results

17 _{Poland, the Czech Republic, Hungary and Slovakia.}

18 _{We base the analysis on the outcome of the regressions as specified in equation (10). In this regression, we}

24 regarding the accession, but from figure 10, figure 14 and figure 20 it appears that Dutch imports from Slovenia and Latvia as a percentage of total Dutch imports was decreasing most of the time between 1995 and 2009. For Dutch imports from Lithuania and Malta as a percentage of total Dutch imports, we get insignificant results.

The reduction of tariff rates in the EU15 made importing from the CEECs cheaper. Therefore, Dutch imports from the CEECs as a percentage of total Dutch imports increases in general. This is the trade creation effect: because importing from the CEECs became cheaper, some goods initially imported from countries with higher prices will after the reduction or elimination of tariff rates be imported from the CEECs. However, six out of ten of the fitted value lines for ‘1995-2009’ are decreasing: increased trade liberalization has not caused the Dutch imports from Estonia, Latvia, Lithuania, Slovenia and Malta to increase by an equal growth rate as the total Dutch imports. For Dutch imports from the four relatively largest import partners of the Netherlands in 199519 _{as a}

percentage of total Dutch imports, we find a significant effect of the EU accession, though for Hungary and Latvia the developments are ambiguous. Only in the case of Poland, the Czech Republic and Slovakia we can conclude that increased trade liberalization has unambiguously and positively influenced Dutch imports as a percentage of total Dutch imports.