An empirical analysis of the process of becoming an EU member based on the gravity model

An Empirical Analysis of the Process

of Becoming an EU Member based on

the Gravity Model

Bachelor Thesis

BSc. Economics and Business

Economics and Finance Track

Renée de Both

Student number: 10382569

Supervisor: Egle Jakucionyte

2

Abstract

This research aims to achieve more insights in the effects of future EU member states during the process of becoming an official EU member on international trade. The model that is used to estimate the effect is a modified version of the gravity model. The different stages in the process of becoming an official EU member are added as dummy variables to this model. Those stages include: attaining official candidate status, starting formal negotiations, signing the accession treaty and the final stage of becoming an official EU member. The econometric analysis is applied by running the random effects model and the Prais-Winston regression. The empirical results suggest that a country that proceeds in the process of becoming an EU member has an increase in exports. Except for the coefficient of the candidate status, the coefficients of the different stages are significantly different from zero. Therefore, it is concluded that during the process of becoming an EU member, exports increase significantly. The effect on exports is the highest in the final stage: becoming an official member state.

Statement of Originality

This document is written by Renée de Both who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

3

Table of Content

1

INTRODUCTION

p. 4-6

2

LITERATURE REVIEW

p. 7-14

2.1 EU Enlargement and Economic Integration p. 7

2.2 Gravity Model and Research Field p. 11

2.3 Process of Becoming EU Member p. 14

3

METHODOLOGY

p. 15-18

3.1 Model p. 15

3.2 Hypothesis p. 16

3.3 Data Sample and Collection p. 16

4

ANALYSIS

p. 19-23

5

CONCLUSION

p. 24-25

6

REFERENCES

p. 26-28

7

APPENDIX

p. 29

4

1. Introduction

As of July 2013 the European Union has a new member: Croatia, the second ex-Yugoslavian country after Slovenia which joins the EU. In addition, the EU is in the midst of enlarging the European Union with other Balkan countries. However, the duration of the process of becoming an official EU member differs across countries. Prior studies suggest that when a country officially joins the EU, its level of international trade increases. Since the process of becoming an EU member can last many years, the question arises whether there will be an impact on international trade in the meantime. This research aims to achieve more insight in the effect of future EU member states during the process of becoming an official EU member on international trade. Therefore, the research question is: what is the impact of the different stages in the process of becoming an EU member on the level of international trade?

After the fall of the Berlin Wall, the enlargement of the EU began to play an important role in European politics and became a priority of the European Commission (EC, 2015). At first, the EU responded by providing aid and offering preferential arrangements only in terms of market access. The process of accession to the EU started by signing the European Association Agreements in 1991, since then, thirteen new countries have joined the EU. In the following years, the progress of some countries to qualify for candidacy was fast, compared to other countries. Those countries included Czech Republic, Estonia, Hungary, Poland and Slovenia. Their accession negotiations started in March 1998, thereafter Bulgaria, Latvia, Lithuania, Romania and the Slovak Republic followed in December 1999. At the moment, 28 countries are official EU member states. In addition, six countries acquired the official candidate status, these include Albania, Iceland, the former Yugoslav Republic of Macedonia, Montenegro, Serbia and Turkey. Albania has not started formal negotiations yet and formal negotiations with Iceland have been put on hold (EC, 2015). This short summary of the past 25 years and the present situation demonstrates that the statuses of the countries are varying within this time period. Therefore, sufficient developments exist to examine the impact of the different stages in the process of becoming an EU member on the level of international trade.

A considerable amount of research investigates the effects of enlargement of the EU and the impact of trade agreements on trade flows. For example, The Dutch Bureau of Statistics (CBS) published an article about the index value of the Dutch exports in real terms to indicate the differences between trade with old, new and non-EU countries. In Figure 1, this index is presented. From here it can be concluded that Dutch exports to more recent EU members has grown much faster than exports to

5 other countries (CBS, 2013). Therefore, it is relevant to gain more insight into the effect on international trade of future EU member states.

Figure 1. Index of Value of Dutch Export (CBS, 2013)

Until now there is no research that focuses on the effects of different stages in the process of becoming an EU member on international trade. This process can take many years, for example, Turkey applied for a candidate status in 1987 and it was recognized as an official candidate in 1997. The formal negotiations started in 2005 and are not closed to date. Another example, Croatia applied for membership in 2003, negotiations started in 2005 and in 2011 the accession treaty was signed. This demonstrates the difference in time periods for every stage across countries.

Croatia was mentioned in a top ten list published by the CBS, as being one of the countries with the largest increase in imports of Dutch goods. In January 2014 the Dutch export to Croatia increased by 22 percent with respect to January 2013 (CBS, 2015). Is this only an effect of the accession of Croatia in July 2013 or did the candidate status of Croatia before play a certain role? It may be that a country is able to benefit from being a candidate during the process of becoming an EU member in such a way that their international trade increases in the meantime.

6 To provide a scientifically grounded answer to the research question, I expand the standard gravity model. I add dummy variables for different steps that are undergone to officially join the EU and estimate the effect of the stages in the process on exports. The explanatory power of these variables is tested and analyzed by running a random effects model and a Prais-Winston regression.

The thesis proceeds as follows. Section 2 discusses prior studies about the effects on international trade of EU enlargement. Subsequently, a discussion about the gravity model and its application follows and the process of becoming an EU member is explained. The methodology (Section 3) provides the model, the hypothesis, the data sample and collection method. The summary statistics, estimation methods and empirical findings are presented and discussed in the analysis (Section 4). Finally, in the conclusion (Section 5), the main results are wrapped up and suggestions for future research are outlined.

7

2. Literature Review

2.1 EU Enlargement and Economic Integration

In this part of the literature review the reasons for enlargement of the EU will be taken into consideration, for incumbents as well as for new entrants. The discussion whether solely international trade agreements could enhance economic integration follows. Subsequently the trade-offs of economic politics on European level are emphasized. Then, the main findings of economic integration are discussed and this part is closed by discussing the impact of economic integration on growth.

In 1993, during a meeting in Copenhagen, the European Council announced its idea of enlarging the EU eastwards. However, every EU incumbent has a veto right to reject a new member. Therefore, it is important to understand why a country wants another country to join the EU and thereby wishes to enlarge the EU. Baldwin (1995) refer to three different concerns that influence the decision to vote for a country to join the EU, namely ‘high’ politics, ‘low’ politics and economics. High politics refers to the role of European integration to maintain democracy and peace. Low politics refers to the way in which an effect of the enlargement on certain special interest groups can take on an importance that far exceeds the economic weight. The economic concern involves the standard aggregate welfare gain or loss, which is typically expressed in percent of GDP.

According to Baldwin (1995) high politics is the driving force for incumbents to enlarge the EU eastwards. Promising or denying an EU enlargement has influence on long run expectations. Denying the prospect of enlarging the EU by Central and Eastern European Countries (CEECs), for instance, would have dimmed the participation of CEECs in the wealth and security of a united Europe. By promising an EU membership, the countries could be able to integrate Western and Eastern Europe. Though, there are disagreements about enlarging the EU amongst incumbents, because of historical and geographic considerations. For example, it is more beneficial for a country that has more trade and shares a border with future EU member states to enlarge the EU. These differences in trade and border sharing between EU incumbents make an inclusion of a country in the EU more difficult (Baldwin, 1995).

While high politics is used as an argument for enlargement, low politics is a determining factor for how long this process will take (Baldwin, 1995). Economic integration could harm people in the society and there are special interest groups which try to prevent this to defend those people. If those special interest groups have influential power, Baldwin (1995) mentions three difficulties that have to be taken into account. Those difficulties include budget issues, veto issues and the ‘big step’

8 factor. When the EU decides to enlarge, there will be extra costs for the incumbents to extend the Structural Funds to new member countries which qualify to receive transfers out of that fund. The aim of the Structural Funds is to converge the per capita income levels of the EU member states. Financing this cost means a cut in EU spending or an increased tax for incumbents. This budget issue could be a consideration of special interest groups. The voting issue refers to the fact that including new countries into the EU will lead to changes in the political landscape, because new entrants receive votes in the Council of Ministers. The last issue concerns the ‘big step’ factor and considers the ability of the governments and businesses in new member countries to deal with the obligations of the EU membership. Western governments and businesses have decades of experience to deal with the EU regulation, but new entrants lack this experience (Baldwin, 1995).

According to Baldwin (1995), the economic interest in enlarging the EU is not as important as the high politics argument. But in a later article, Baldwin et al. (1997) estimate the long-run economic benefits and costs of enlarging the EU15 and conclude with positive thoughts with respect to EU enlargement. They argue that the eastern enlargement is a bargain. It is estimated that the net costs, transfers less benefits, are between zero and ECU8 billion. The estimated upper bound is 0.01 percent of the EU15’s GDP, which is low compared to the historical numbers of former enlargements in central Europe. Though, the eastern enlargement is not about minimizing net costs, but it is about ensuring peace and stability in Europe. ‘Economic integration is the means, not the end’ is a quote from Baldwin et al. (1997).

The question remains why the CEECs, or countries in general, want to join the EU. It is argued that economic gains for the CEECs of an EU membership are large (Baldwin, 1995). Although, the only large, additional economic gain that requires full EU membership is to be qualified for the transfers of the Structural Funds. This is because the Europe Agreements already lead to lowering EU trade barriers for CEEC industrial products and this enhances economic integration without being a full member of the EU. But even when transfers are not considered, the real income will rise substantially (Baldwin, 1997). A membership of the CEECs allows these countries to be a part of the advanced industrial economies in Europe. Thereby, the membership affects the domestic politics and facilitates further reforms.

Taken into account that economic integration is the means and could facilitate further reforms, here it is pointed out that formal economic agreements are not sufficient. Because even when under formal economic agreements trade barriers are reduced, international markets for goods, services and capital are not completely exploited as they could be under complete economic integration. The question is why so much trade is absent from the international markets. According to Rodrik (2000),

9 national borders appear to be boundaries of political and legal jurisdictions. Such boundaries lead to separate markets in actually the same way as transportation costs and border taxation do. International trade is subject to transportation costs arising from differences in political and legal systems. There are different reasons why such transportation costs arise, but Rodrik (2000) argues that contract enforcement is the most important one. International courts may be unable to interfere between residents of different countries. Therefore, international transactions lead to higher risk.

This risk issue has an impact on how far international economic integration will reach and it is argued that it is limited by national sovereignty (Rodrik, 2000). The question arises whether it is possible to exploit international markets while politics remain at the local level. To answer this question, Obstfeld et al. (2005) use the macroeconomic trilemma, or also called the Impossible Trinity, as a guiding policy framework. Countries face a confrontation of three objectives which are jointly unable to attain. The three objectives are 1) stabilizing the exchange rate, 2) free international capital mobility and 3) monetary autonomy oriented toward domestic goals. Policymakers are only able to attain two out of the three objectives, because all three objectives are not consistent internally. The consequence of the trilemma is that one objective should be given up. When monetary policy activism is chosen to set local interest rates differently than the world interest rate, the arbitrage opportunities in open capital markets is in conflict with the objective of a fixed exchange rate regime (Obstfeld et al., 2005).

According to Rodrik (2000), complete international economic integration will only be attainable if jurisdiction is expanded to a global level or if a shrinkage of local politics takes place. He argues that global federalism will be the most likely outcome because of three reasons. First, the technological progress will facilitate a global government, since the new technology facilitates communicating with each other all over the world. Second, a large part of the population does not want to give up the opportunities that an integrated market can deliver. Third, citizens are not likely to give up their citizenship rights of representation and self-government, which leads to pressure on politicians to keep accountability (Rodrik, 2000).

But if economic agreements are not sufficient to enhance complete economic integration, are there only positive consequences of complete economic integration attained through EU enlargement? No, there are trade-offs of higher level economic politics. Alesina et al. (1995) mention the trade-off between the benefits of large political jurisdictions and the requirement of keeping individuals with different interests, preferences, culture and history together. This fundamental trade-off plays a certain role in three different cases. The first case is analyzed by Alesina and Perotti (1998), they emphasize the trade-off in the case of public goods provisions. Fiscal integration in political

10 jurisdictions has benefits, it leads to economies of scale of public goods and the opportunity of mutual insurance. These benefits also come at a cost, because when a political jurisdiction increases, the political landscape becomes unsure. The reason for this is that the outcome of the policy then depends on the aggregation of preferences of a more diverse population. The more diverse the population, the larger is the uncertainty and the higher the costs (Alesina and Perotti, 1998).

The second case is the trade-off within a redistribution system. If countries form a fiscal union, they can decrease economic uncertainty by mutual insurance when one country is hit by an idiosyncratic shock. If there is no fiscal union, countries cannot insure each other by transfers. The cost is that it may cause dispersion in political preferences and this results in political uncertainty (Alesina et al., 1995).

The third case is the trade-off regarding institutional differences. When the social security system is applied to different political levels, centralized funding can result in inefficient outcome because of free-riding by national politics. If regions have lower administrative standards, they are able to free-ride on the tax revenues collected in other regions aiming for social security benefits. On the other hand, when it is decentralized, local governments are only able to rely on their local social security system (Alesina et al., 1995).

Discussing the reasons and consequences of European enlargement and economic integration brings us to the fundamentals of international trade: why do people actually want to enhance international trade? Krugman (1979) shows that trade is not only a result of international differences in technology or factor endowments, but trade is a way of extending the market and allowing exploitation of economies of scale. The effects of trade are similar to labor force growth and regional agglomeration. This evidence seems useful in understanding trade between industrial countries, because trade, and gains from trade, will take place even among countries with identical tastes, technology and factor endowments (Krugman, 1979).

This leads to the principle economic argument for enlargement of the EU. That is to expand the free trade area, creating a single market, where international trade can benefit from greater mobility of goods and labor. Hence, EU citizens can take advantage from the fact that a single market leads to increasing economies of scale. This has an impact on the unit costs of products which is reduced by an efficient market, leading to efficient productivity and product diversity (Dixit and Norman, 1980).

The importance of economic integration in theory has been outlined. Henrekson et al. (1997) provide practical empirical evidence of European integration on increased economic growth. It is found that an European Union or European Free Trade Association membership has a positive and significant

11 effect on economic growth and that there is no significant difference between an EU or EFTA country. According to the results, an EU or EFTA membership may increase growth rates by 0.6-0.8 percentage points. The study suggests that a further integration can be growth-enhancing in the long run (Henrekson et al., 1997).

2.2 Gravity Model and Research Field

After arguing the positive effects of economic integration on national economic growth, the question arises whether there is a positive impact on international trade. Anderson (1979) argues that the gravity model explains best the factors that lead to increased international trade. This model is applied in a lot of different circumstances and normally has a good fit and seems to be fruitful in formalizing the predicted international trade flows.

There is strong empirical relationship between the size of the country’s economy and the volume of both imports and exports (Krugman et al., 2012). On top of that, distance has a negative effect on international trade. These relationships are represented in the gravity model. The reason for the name is the analogy to the Newton’s law of gravity: the gravitational attraction between two objects is proportional to their masses and decreases with the distance between them (Krugman et al., 2012).

The purpose of the gravity model is to estimate the level of international trade of countries and to analyze the result of economic integration on international trade. The standard model used by researchers is written as:

Log Xij = α0 + α1logYi + α2logYj + α3logNi + α4logNj + α5logDij + α6Aij + loguij,

where Xij is the value of exports of country i to j; Yij is the GDP; Nij is the population; Dij is the distance

between the two countries and A is any other variable enhancing or resisting trade between the two countries. Some researchers add dummy variables to the gravity model to improve the statistical significance or estimate the impact of the variables, like border sharing, joining a trade agreement or having the same language (Greenaway and Milner, 2002).

The number of applications of the gravity model is extensive. This gravity modeling became popular especially during the early 1990s to investigate potential effects on trade (Greenaway and Milner, 2002). There are three different reasons for this upcoming interest during this period. The first reason is the comeback of negotiations of free trade agreements, which asks for potential economic consequences. The second reason is the development of the gravity model which gained a broad

12 acceptance as an empirical tool. Third, the use of the gravity framework ex post as well as ex ante was seen as useful (Greenaway and Milner, 2002).

However, the use of the gravity model has been controversial and modeling issues have appeared. Some argue that the model even lacked theoretical underpinnings (Greenaway and Milner, 2002). Intuitively, the model is plausible, as the early literature of Tinbergen (1962) derived intuitive explanations for the relationship of the variables included in the gravity framework. One can imagine the fact that larger countries located closer to each other are likely to trade more with each other. Though, there was no formal dependency on standard trade models. Anderson (1979) addressed this shortcoming by a model of homogeneous goods to derive the importance of transport costs and assumes that distance and transport costs are related, called Armington Preferences. Armington Preferences ensure that larger countries, which have more traded goods, trade more. If all goods are traded, national income is then the total value of traded goods. Bergstrand (1985, 1989) develops the analysis further by introducing a general equilibrium model from which a gravity equation could be derived. Helpman and Krugman (1985) address this by including the gravity equation in a model of monopolistic competition with increasing returns to scale, which also yields the reliability of predictions with respect to trade patterns. In sum, one can no longer argue that the gravity model exists in a theoretical vacuum, since it is shown to be derivable from trade models (Greenaway and Milner, 2002).

Econometric issues could arise when interpreting the coefficients of the gravity model. OLS cross-sectional or time averaged panel estimation can eliminate the time variation in international trade flows and lead to inconsistent estimates (Mátyás, 1997). A suitable estimation method and specification of the gravity model lowers the risk of interpreting parameter inconsistency.

Greenaway and Milner (2002) conclude in the paper that the prospects for extending the use of the gravity model by investigating regionalism are substantial. According to them, the importance of measuring and testing the model is likely to expand. There is already a more discriminating and careful application of the model, due to panel data and stronger theoretical underpinnings.

Greenaway and Milner (2002) list a number of studies of international trade flows affected by trade agreements, using a gravity model. A common finding from the literature is that there is a positive relationship between trade agreements and trade effects. When controlling for distance, countries that are members of the same trade agreement trade more with each other than would otherwise be expected. Even if trade agreements did not affect liberalization, there are positive trade effects found. Therefore, the conclusion is that trade agreement and, hence, economic integration lead to a higher level of international trade (Greenaway and Milner, 2002).

13 Not all the research done in this research field will be cited, but there is one study which is emphasized here. This relatively recent article, that goes more into depth with respect to the enlargement of the EU is Papazoglou et al. (2006). It researches the potential trade effects of the most significant enlargement of the EU in 2004. The purpose of the research is to quantify the gains from trade from the expansion of the EU single market. It is assumed that transition economies already behave more like market economies and that the EU is the dominant trading partner of those transition economies. The gravity model is used to establish an export equation and is written as:

Log Xij = α0 + α1logYi + α2logYj + α3logNi + α4logNj + α5logDij + α6ADJDMij + α7INTDM + loguij,

where Xij is the value of exports of country i to j; Yij is the GDP; Nij is the population; Dij is the distance

between the two countries, ADJDM is a dummy variable which equals unity if the countries share a geographic border and INTDM is a dummy variable which equals unity if the country is an EU member.

The model is estimated by capturing the trade patterns of the EU15 economies to forecast the trade patterns of each of the ten new EU countries in the accession scenario as well as the no-accession scenario. The analysis relies on a panel of trade data for 1992 – 2003 and the predicted value for the two scenarios refers to 2006. Then the difference between the trade patterns with accession and without accession predicted for 2006 is assigned as the impact of the EU enlargement. According to Papazoglou et al. (2006), four main conclusions can be drawn from the results of the research. The first one is that all the accession economies face a rise in their level of trade, therefore, these countries become more integrated with the trading system. The second conclusion is that the increased exports generally increase with EU15 members, while trade decreases with the rest of the world. The third conclusion is that all accession countries have an increase in imports higher than an increase in exports. The last conclusion is that, as a result, those countries need to stimulate their competitiveness to a higher degree.

The overall result is that the impact of the EU enlargement in 2004 is not only to generate a higher level of international trade, but also has an impact on the direction of trade flows. The article shows that trade patterns in the enlarged EU are path-dependent and it is the job of the EU to take this into account while planning further enlargements.

No academic literature on different stages of becoming a member of a trade organization is available. Therefore, in the next part, the steps that have to be taken by countries to become an EU member are discussed.

14 2.3 Process of Becoming EU Member

The EU enlargement happens not just in one step and a lot of considerations have to be taken into account. As Baldwin (1995) has argued, the enlargement of the EU could be delayed because of pressure of special interest groups. That is a reason why the EU develops different steps that have to be taken to proceed the integration between Western and Eastern Europe.

On the official website of the European Commission, the steps towards joining the EU are described (EC, 2015). Including the final stage, there are four stages in the process of becoming an EU member. These stages include:

1) Official candidate status for membership.

2) Formal membership negotiations, a process that aims to establish the EU law and to implement the judicial, administrative, economic and other reforms to meet the conditions for accession, known as the accession criteria.

3) Acceding country, when negotiations are completed to the satisfaction of both parties and the accession treaty is signed. The country is allowed and able to join the EU.

4) Official EU membership.

The formal membership negotiations cannot start until all EU countries unanimously agree on a mandate for negotiations with the candidate country. These negotiations take place between ministers and ambassadors of the EU governments and the candidate country, this is called an intergovernmental conference. The negotiations are based on two elements. The first element is screening, the EC investigates each policy field in detail, to determine how well the country is prepared. The results of the examination are presented to the EU member states in the screening report. The conclusion of this report is whether the EC recommends to open negotiations directly or recommends that specified conditions are to be met first; this is called opening benchmarks. The second element is the negotiating position, the candidate country has to submit its position and the EU has to adopt the common position. The pace of the negotiations depends on the tempo of reform and adjustment of the candidate country to the EU law.

Negotiations are not ended until all EU governments are satisfied with the progress of the candidate. The accession treaty is the document that confirms the EU membership of the country. It includes the detailed terms and conditions of membership and transitional arrangements and deadlines. This treaty will only be final and binding if it is supported by the EU council, EC and the European Parliament and when it is signed by the candidate country and the EU members. Once the treaty is signed, the candidate becomes an acceding country. The country is expected to be an official EU member on the date stated in the treaty (EC, 2015).

15

3. Methodology

In this section it is explained how empirical evidence is gained in order to answer the research question. In the literature review I argue that the gravity model is applied in a lot of different circumstances and it seems to be fruitful in formalizing the predicted international trade flows. For this reason, the gravity model is applied to pursue an econometric analysis. In the first part of this section, I clarify the application of the gravity model in this research, in the second part I formulate the hypothesis and finally, I describe for which sample the research requires data and from which sources the data is collected.

3.1 Model

The gravity model uses an empirical methodology to capture the trade pattern of countries. In this research the standard gravity model is expanded in order to answer the research question. The adapted model that will be estimated is written as:

LogXij = α0 + α1logYi + α2logYj + α3logNi + α4logNj + α5logDij + α6CANDi + α7NEGOi + α8ACCEi + α9EUi +

α10BORDij + α11LANGij + logui,

where Xij is the value of exports of country i to j; Yij is the GDP of country i and j; Nij is the population

of country i and j and Dij is the distance between the two countries. The rest of the variables in this

regression are dummy variables. The variables CAND, NEGO, ACCE and EU are gathered from the literature, that explains the process of becoming an EU member. These variables refer to the steps that a country has to take to officially join the EU. The dummy variables take a value of unity if the specific stage in the process applies to a country. Therefore, the dummies are set equal to unity in one of the following variables: if countries have an official candidate status (CAND), if formal negotiations have started (NEGO), if they are acceding countries (ACCE) or if they are official EU members (EU).

To improve the statistical accuracy of the estimated model, control variables are added to the model. The control variables in this research model are border sharing (BORD) and language (LANG). These dummy variables equal unity, if countries in question share a geographic border or if they have a common official language, and zero otherwise.

According to the literature review, each variable in the model has predictable effects on the international trade level between trading partners. The GDP variables are expected to have positive impact on the level of international trade (Henrekson et al. (1997). An increase of GDP in the exporting country leads to more availability of domestic production for exports and an increase in

16 GDP in the importing country leads to an increase in imports. Therefore, the prediction is that both α1 andα2 will be positive. On the other hand, it is expected that the population variables do not have

the same sign. A larger population creates larger domestic demand, which enhances imports and restrain exports. Thus, the prediction is that α3 will be negative and α4 will be positive. Regarding the

distance variable, α5 is expected to be negative. According to Rodrik (2000), distance leads to higher

transportation cost, which reduces the level of international trade.

This research controls for border sharing and common languages, because literature outlined a significantly positive effect of these variables on the level of international trade (Papazoglou et al., 2006). If two countries share a border or if they have a common official language, it means that they have lower transportation and communication costs and easier market access, therefore, it is expected that α10 and α11 are positive.

Papazoglou et al. (2006) suggest that an EU membership leads to higher levels of international trade, thus α9 is expected to be positive. The dummy variables of the different stages of

becoming an EU member are therefore also expected to be positive.

3.2 Hypothesis

The predictions of the effects of the added variables to the gravity model result in the hypothesis of this research: a country that proceeds further in the process of becoming an EU member sees an increase in exports to EU member states and countries which are in the process of becoming an EU member. An EU membership enhances international trade and it is expected that different steps that have to be taken to become an EU member also enhance international trade. This because when it becomes more apparent that a country will be an EU member in the nearby future, it is assumed that a country is able to act more like an EU member.

To test whether the hypothesis should be rejected or not, the described model is estimated by running a regression using the statistics package of STATA. According to these estimates, the variables and their significance are analyzed to answer the research question.

3.3 Data Sample and Collection

In this part of the methodology I describe for which sample the research requires data and from which database the data is collected. To estimate the model, 36 countries are included in the sample. The reason for this is that it is preferred to have countries in the sample which are, nowadays, in some way connected with an EU membership. Therefore, it is chosen to include the 28 countries that are official EU members today, the six candidate countries at this moment and two potential candidates. The countries included in the sample are presented in the table below and are categorized based on their status at this moment of writing. The statuses of the countries are varying

17 within the timeframe of the sample, which is incorporated in the estimation of the model. The sample contains annual panel data over fourteen years. The first year of the sample is 2000, the sample ends in 2013.

Table 1. Sample countries

EU member EU candidate Potential candidate

Austria Italy Albania Bosnia and Herzegovina

Belgium Latvia Macedonia, FYR Kosovo*

Bulgaria Lithuania Iceland

Croatia Luxembourg Montenegro

Cyprus Malta Serbia

Czech Republic Netherlands Turkey

Denmark Poland Estonia Portugal Finland Romania France Slovakia Germany Slovenia Greece Spain Hungary Sweden

Ireland United Kingdom

* This designation is without prejudice regarding the current situation of Kosovo and is in line with UNSCR 1244 and the ICJ Opinion on the Kosovo Declaration of Independence.

The data for the variables are obtained from different databases. The value of exports (X) from one country to another is obtained from the IMF Direction of Trade Statistics and is denoted in current US dollars. The nominal values of exports are deflated to obtain the real values of exports, since it is not desirable to include the relationship of price changes in the estimation. The method I use for the deflation technique is as follows. The World Bank database provides the export value index and the export volume index for each country, with 2000 as base year. These indices are used to calculate the unit value index, which is the appropriate index to deflate the nominal values. By dividing the export value index by the export volume index, the unit value index for each country and every year is obtained. The real value of exports is calculated by dividing the nominal value by the corresponding unit value index and multiplying this with hundred.

GDP data (Y) is taken from the World Bank database and is denoted in constant 2005 US dollars. The population data (N) is obtained from the database of Eurostat, where the population is indicated on January 1 of each year. To get the data for the variables distance (D), language (LANG) and border sharing (BORD), the GeoDist database of CEPII is inquired. The data for the variable of the distance between two countries is based on latitudes and longitudes of the most important cities or agglomerations of the country in terms of population, instead of the capital cities. The reason to use this calculation is because these important cities are the places where the transportation costs of

18 distances specifically do matter. The data for the language variable is based on the fact that two countries officially share a common language. Languages that are not spoken by at least twenty percent of the population are ruled out. The data obtained for the variable of border sharing is based on the fact whether two countries are geographically contiguous or not. The three countries Kosovo, Montenegro and Serbia were not included in the GeoDist database. Therefore, data for distance, language and border sharing are obtained using the website ‘World Atlas’, which is also inquired by Papazoglou et al. (2006).

To provide the data for different stages of the process of becoming an EU member, different accession treaties, monitoring and progress reports of the EU are used to define the status of each country for each year. In order to form the dataset the following information is gathered from the websites: the accession treaties provided by Eur-Lex (2015), enlargement information of the European Commission (2015) and the Strategy and Progress Reports provided by the European Commission (EC, 2015).

19

4. Analysis

This section aims to establish an analysis obtained from the results of the model and the dataset. The hypothesis is tested based on the empirical results. In the first part, the dataset is examined to indicate the missing values, the average duration of stages and the descriptive statistics. Subsequently, different estimation methods are discussed. The empirical results from these estimations are presented in the last part.

4.1 Summary Statistics

After assessing the missing values in the dataset, I observed that there is no possibility to correct for all missing values without significant distortion of the analysis. There is no data available of exports values for Kosovo, Montenegro and Serbia until 2005. Since Kosovo is independent since 2008 and Montenegro since 2006, not all the data is available. To improve the analysis of the research, Kosovo, Montenegro and Serbia are excluded from the dataset while running a regression. The consequence is that 33 countries are left. The dataset does not contain missing values for other countries and variables.

To show that it is of practical matter to analyze the impact of the different stages on the level of international trade, the average duration of the different stages is presented in Table 2. From here it can be concluded that the average durations are long enough to be relevant. The time length of a status of a country is only included in the calculation if it starts and ends in the specific timeframe of the research sample. The average duration of the official EU membership is not calculated, since no country ever left the EU after acceding.

Table 2. Average duration of stages

Stage Average time period in years

Official candidate status 2.00

Formal membership negotiations 4.21

Acceding country 1.23

Official EU membership -

To give an insight in the dataset, the descriptive statistics are provided in Table 3. The variables are presented after the log-transformation, since the regression analysis is executed with the transformed variables.

20 Table 3. Summary statistics

Variable Observations Mean Std. Dev. Minimum Maximum

Xij_log 14784 18.30459 3.501506 0 25.291 Yi_log 14784 25.47129 1.802136 22.38855 28.78221 Yj_log 14784 25.47115 1.802216 22.38855 28.78221 Ni_log 14784 15.73556 1.499078 12.53914 18.22875 Nj_log 14784 15.73556 1.499078 12.53914 18.22875 Dij_log 14784 7.11744 0.673245 4.087945 8.49333 CANDi 14784 0.0238095 0.1524605 0 1 NEGOi 14784 0.1363636 0.3431859 0 1 ACCEi 14784 0.034632 0.1828522 0 1 EUi 14784 0.7034632 0.4567459 0 1 BORDij 14784 0.0814394 0.273518 0 1 LANGij 14784 0.030303 0.1714256 0 1 4.2 Estimation Methods

Different estimation methods are examined to decide which estimator is the most appropriate method to estimate the specified model. In this case it is not an option to use the fixed effect estimator, since the model contains time-invariant variables, such as border sharing and common language. The fixed effect estimator removes the effect of those time-invariant variables to assess the net effect of the predictors on the outcome variable and is used to study the causes of changes within the country-pairs (Plümper and Troeger, 2007).

An option to estimate a model including time-invariant variables is to ignore the possibility of effects of country pairs and run a pooled OLS regression. However, the exclusion of exogenous variables that have an influence on the endogenous variable causes omitted variable bias. Therefore, the probability is low that unbiased coefficients can be obtained by pooled OLS, in particular, when the amount of observations is small (Plümper and Troeger, 2007).

The random effects estimator is an estimator that could perform better than pooled OLS, if the model includes time-invariant variables. This estimator assumes that the error term across entities is not correlated with the predictors and this allows time-invariant variables to play a role as explanatory variables. It is necessary to include variables that may influence the predictor variables. Otherwise, the consequence of not including all the influential variables in the model is omitted variable bias. On the other hand, random effects estimator allows to generalize the inferences beyond the sample used in the model (Plümper and Troeger, 2007).

The study of Papazoglou et al. (2006) is comparable to this research, thus, their estimation methods are also taken into consideration. The regression methods used are the random effects estimator and the Prais-Winsten estimator with panel-corrected standard errors. The two advantages of the Prais-Winsten method is that it does not assume that data for all country pairs

21 carry the same degree of autocorrelation and that it corrects for simultaneous error correlation across country pairs.

To decide which estimation method is appropriate to use, several tests are performed. The Hausman-test is applied to decide between the fixed and random effects estimator. It tests whether the errors terms are correlated with the regressors (Hausman, 1978). The test yields a p-value of 0.0000 that rejects the null hypothesis. Based on this result, the fixed effects estimator should be used. However, due to the inclusion of time-invariant variables, this estimation is not an option, despite the fact that the Hausman-test suggest to use the fixed effects estimator.

The Breusch-Pagan Lagrangian multiplier test is applied to decide between the pooled OLS estimator and the random effects estimator. The null hypothesis in this test is that variances across entities are zero and therefore, no panel effect exists (Breusch and Pagan, 1979). The test yields a p-value of 0.0000 that rejects the null hypothesis. Based on this test result, the random effects estimator should be used.

Plümper and Troeger (2007) argue that when time-invariant variables preclude the fixed effects estimation, the random effects estimation may perform as a viable second best option. On top of that, Papazoglou et al. (2006) decided to extend their random effects estimation by using the Prais-Winston method, because of the two advantages it has. After taking all these considerations into account, I estimate both the random effects panel data model and the Prais-Winston regression.

To test for heteroskedasticity, another version of the Breusch-Pagan-test is used, which is called the Breusch-Pagan / Cook-Weisberg test. The null hypothesis in this test is that variances are constant. The test yields a p-value of 0.0000 that rejects the null hypothesis. Based on this test result, it is indicated that heteroskedasticity is present in the model. To correct for this, the option of robust standard errors is used while running the regression. The results of the Hausman-test and the two different versions of the Breusch-Pagan-test are included in the appendix.

4.3 Empirical Results

The estimation results for the model with random effects and the Prais-Winston regression with panel-corrected standard errors are presented in Table 4. The R² indicates the proportion of the variance in the dependent variable that is attributable to independent variables in the regression models. The R² of the model estimated by random effects equals 72.28 % and the model estimated by the Prais-Winston method explains 77.99 % of the variation in exports. Both estimated models have a significant F-test or chi², since the value is lower than 0.05. This indicates that all the coefficients in the models are jointly significantly different from zero and the models have a good fit.

22 Table 4. Regression results

Random effects Prais-Winston

Yi_log 1.684309*** (0.1161255) 1.294835*** (0.0587199) Yj_log 1.205334*** (0.0699899) 0.9909354*** (0. .0298726) Ni_log - 0.6557726*** (0.1236012) - 0.2437059*** (0.0683375) Nj_log - 0.2887502*** (0.0809305) - 0.0447581 (0.0641022) Dij_log - 1.711637*** (0.105716) - 1.560499*** (0.0676486) CANDi 0.0221008 (0.1871423) 0.281796 (0.1775698) NEGOi 0.3361884*** (0.1241244) 0.7908547*** (0.1593504) ACCEi 0.502517*** (0.1415004) 0.9708904*** (0.1590101) EUi 0.5869204*** (0.1402146) 1.081689*** (0.15391) BORDij 0.493017* (0.2428818) 0.5759457*** (0.1321941) LANGij - 1.239636** (0.4998299) - 0.7857767*** (0.3066077) Constant - 28.73233*** (0.9594819) - 25.2217*** (1.169334) No. observations 14,784 14,784

No. country pairs 1,056 1,056

R² 0.7228 0.7799

Chi²-test 0.0000

F-test 0.0000

Notes: Dependent variable: exports Xij. Robust standard errors in parentheses. Significance levels: * significant at 5 percent; **significant at 2 percent; *** significant at 1 percent

The coefficients of the standard variables of the gravity model, GDP (Y) and population (N) for both countries and the distance (D) between those countries, correspond to the predictions of the signs, except for the coefficient for the population variable of the importing country. This coefficient is not significantly different from zero with the Prais-Winston estimation, but all the other coefficients of the standard variables are significant at a one percent level. One reason for the fact that the

23 coefficient for the population variable of the importing country does not correspond to the prediction could be that in eleven countries, the amount of population decreased during the time period of the sample. These countries account for one third of the sample and include Albania, Bulgaria, Croatia, Estonia, Germany, Greece, Hungary, Latvia, Lithuania, Poland and Romania. This may be caused by population aging, leading to a decreasing amount of population in the model, while nonetheless exports increase in the model.

Regarding the control variables, the coefficient for border sharing (BORD) corresponds to its prediction, while the coefficient for common language (LANG) does not. This coefficient was expected to be positive, but both regressions estimate it to be negatively related to export. One reason for the negative relationship may be that communication costs are decreasing these days and, therefore, it is not beneficial to export especially to a country which shares a common language with respect to other countries. Another reason could be that new technological progress enhances communication between countries across Europe, leading to higher exports between countries that do not share a common language (Rodrik, 2000). Finally, the sample does not contain a lot of countries which share a common language with each other. The coefficients for all the control variables are significant at a five percent level at least, but mostly at a one percent level.

Now referring back to the hypothesis of the research, the coefficients of the expanded gravity model are analyzed. In both regressions, the coefficients are significantly different from zero, except for the candidate status (CAND). In addition, all coefficients have a positive effect on exports, hence, leading to more international trade. There are differences in coefficients between the two estimated models, but there are also similarities. The positive effect on trade is the largest when a country becomes an official EU member, which is in line with the considerations of enlarging the EU. The second similarity is that the two effects on trade of an acceding country and an official EU member are not substantially different. From this observation it can be concluded that most of the positive effects on international trade are already exploited due to signing the official accession treaties. The third similarity is that the positive effects on trade are becoming larger when a country proceeds further in the process of attaining the official EU membership. This proves that when it becomes more apparent that a country will be an EU member in the nearby future, there are more exports. The final similarity is that the effect of being official candidate in the first stage is not significantly different from zero. This suggests that when a country is accepted by the EU as official candidate, its exports may not increase immediately in the first stage of the process. The rest of the coefficients for the different stages are significant at a one percent level. Overall, according to this research, it is concluded that a country that gets further in the process of becoming an EU member has an increase in exports. Therefore, it is proved that the hypothesis of this research should not be rejected.

24

5. Conclusion

In this thesis I attempted to evaluate what the impact is of the different stages in the process of becoming an EU member on the level of international trade. My hypothesis is that a country that proceeds further in the process of becoming an EU member has increased exports to EU member states and countries which are in the process of becoming an EU member. To give an answer to the research question, I estimated a modified gravity model. The standard variables of the gravity model are GDP of the exporting and importing country, population of both countries and the distance between the countries. The different stages that are undergone to officially join the EU are added as dummy variables to this model. Those stages include: attaining official candidate status, starting formal negotiations, signing the accession treaty and the final stage of being official EU member. The control variables which are added to the model are border sharing and common language. After taking the considerations of Plümper and Troeger (2007), Papazoglou (2006) and the results of the Hausman-test and Breusch-Pagan Lagrangian multiplier test into account, I decided to both run the random effects model and the Prais-Winston regression.

The empirical results suggest that a country that participates in the process of becoming an EU member has an increase in exports, since the coefficients of the four variables for the different stages affect exports positively. Except for the coefficient of the candidate status, the coefficients of the different stages are significantly different from zero. The second finding is that the positive effects on trade become larger when a country proceeds further in the process of attaining the official EU membership. Therefore, it is concluded that exports increase during the process of becoming an EU member. The effect on exports is the highest in the final stage: becoming an official member state.

Referring back to the case of Croatia, as the results suggest, the substantial increase in Dutch exports to Croatia may be explained by the different stages Croatia has undergone to become official EU member. According to this research, the effects on international trade do not differ substantially between the stage of an acceding country and the final stage of being official EU member. Therefore, the higher levels of Dutch exports to Croatia may already be achieved from the moment Croatia and the EU incumbents signed the accession treaty. In addition, increased exports could also be partially explained by the formal negotiations, since formal negotiations also have a positive effect on exports in my model.

The contribution of this research in practice is that for firms in EU member states it is beneficial to focus on future EU members in advance of the process of becoming an EU member.

25 Then, those firms are prepared to expand their trade market in countries that are in the process, due to higher levels of international trade. This study contributes to the existing research by providing a potential expansion of the gravity model.

A suggestion for future research is to examine the effect of the different stages in the process of becoming an EU member on the amount of foreign direct investment. There are two main reasons why it may be advantageous for a firm to invest in a foreign country (Hymer, 1976). The first reason is that it could be profitable to acquire firms in more than one country in order to remove competition. The second reason is that some firms have a competitive advantage in a particular industry and therefore are able to exploit these advantages by establishing foreign operations. Prior studies suggest that an EU membership positively affects the amount of foreign direct investments. This effect may also apply for countries which are in the process of becoming an EU member. Foreign direct investment depends for a large part on the accessibility of a country, which becomes larger when countries are members of the same trade agreements (Hymer, 1979). Further research can examine the potential increase in foreign direct investment when a country proceeds further in the process of becoming an EU member.

26

6. References

Alesina, A., Perotti, R., & Spalaore, E. (1995). Together or separately? Issues on the costs and benefits of political and fiscal unions. European Economic Review, Vol. 39, Issues 3–4 (April 1995), pp. 751–758

Alesina, A., & Perotti, R. (1998). Economic Risk and Political Risk in Fiscal Unions. The Economic Journal, Vol. 108, Issue 449 (July 1998) pp. 989–1008

Anderson, J. (1979). A Theoretical Foundation for the Gravity Equation. The American Economic Review, Vol. 69, No. 1 (March, 1979), pp. 106-116

Baldwin, R. (1995). The Eastern enlargement of the European Union. European Economic Review, Vol. 39, Issues 3–4 (April 1995), pp. 474–481

Baldwin, R., Francois, J., & Portes, R. (1997). EU enlargement costs and benefits. Economic Policy, Vol. 30, Issue 82 (April 1997), pp. 125-176

Bergstrand, J. (1985). The Gravity Equation in International Trade: Some Microeconomic Foundations and Empirical Evidence. The Review of Economics and Statistics, Vol. 67, No. 3 (August, 1985), pp. 474-481

Bergstrand, J. (1989). The Generalized Gravity Equation, Monopolistic Competition, and the Factor-Proportions Theory in International Trade. The Review of Economics and Statistics, Vol. 71, No. 1 (February, 1989), pp. 143-153

Breusch, T., & Pagan, R. (1979) A Simple Test for Heteroscedasticity and Random Coefficient Variation. Econometrica, Vol. 47, Issue 5 (September, 1979), pp. 1287-1294

Centraal Bureau voor de Statistiek. (2013, June 15). Goederenexport sterk gericht op EU. Retrieved on June 15, 2015, from

http://www.cbs.nl/nl-NL/menu/themas/internationale-handel/publicaties/artikelen/archief/2013/2013-06-25-infographic-i-h.htm

27 Centraal Bureau voor de Statistiek. Internationale Handel, top-tien stijgers en dalers. Retrieved on

June 15, 2015, from http://www.cbs.nl/nl-NL/menu/themas/internationale-handel/cijfers/extra/internationale-handel-top-tien.htm

Dixit, A., & Norman, V. (1980). Theory of International Trade. Cambridge, United Kingdom: Nisbet.

Eur-Lex. EU Law and Related Documents. Retrieved on June 1, 2015, from http://eur-lex.europa.eu/collection/eu-law/treaties-accession.html#new-2-46

European Commission. (2013, June 27). European Neighbourhood and Enlargement Negotiations. Retrieved on May 13, 2015, from http://ec.europa.eu/enlargement/policy/steps-towards-joining/index_en.htm

European Commission. (2014, Octobre 14). European Neighbourhood and Enlargement Negotiations. Retrieved on June 1, 2015, from http://ec.europa.eu/enlargement/policy/steps-towards-joining/index_en.htm

European Commission. (2015, February 2). Economic and Financial Affairs. Retrieved on June 1, 2015, from http://ec.europa.eu/economy_finance/international/enlargement/index_en.htm

Greenaway, D., & Milner, C. (2002) Regionalism and Gravity. Scottish Journal of Political Economy, Vol. 49, No. 5 (November 2002) pp. 574-585

Hausman, J. (1978). Specification Tests in Econometrics. Econometrica, Vol. 46, Issue 6 (November, 1978), pp. 1251-1271

Helpman, E., & Krugman, P. (1985). Market Structure and Foreign Trade: Increasing Returns, Imperfect Competition and the International Economy. Apollo, United States.

Hymer, S.H., (1976). The International Operations of Nationals Firms; A study of Direct Foreign Investment. Cambridge, United Kingdom: MIT Press.

Henrekson, M., Torstensson, J., & Torstensson, R. (1997). Growth effects of European integration. European Economic Review, Vol. 41, Issue 8 (August 1997), pp. 1537–1557

28 Krugman, P. (1979). Increasing returns, monopolistic competition, and international trade. Journal of

International Economics, Vol. 9, Issue 4 (November 1979), pp. 469–479

Krugman, P., Obstfeld, M., & Melitz, M. (2012). International Economics. Essex, United Kingdom: Pearson Education Limited.

Mátyás, L. (1997). Proper Econometric Specification of the Gravity Model. World Economy, Vol. 20, Issue 3 (March, 1997), pp. 363–36

Obstfeld, M., Shambaugh, J., & Taylor, A. (2005). The Trilemma in History: Tradeoffs Among Exchange Rates, Monetary Policies, and Capital Mobility. Review of Economics and Statistics, Vol. 87, No. 3 (August 2005), pp. 423-438

Papazoglou, C., Pentecost, E., & Marques, H. (2006). A Gravity Model Forecast of the Potential Trade Effects of EU Enlargements: Lessons from 2004 and Path-dependency in Integration. World Economy, Vol. 29, Issue 8 (August 2006), pp. 1077–1089

Plümper, T., & Troeger, V. (2007). The Estimation of Time-Invariant Variables in Panel Analyses with Unit Fixed Effects, SSRN Working Paper

Rodrik, D. (2000). How far will economic integration go? The Journal of Economic Perspectives, Vol. 14, No. 1 (January, 2000), pp. 177-186

29

7. Appendix

Table 1. Hausman-test

Coefficients

(b) (B) (b-B)

fixed random difference S.E.

Yi_log 2.978648 1.684309 1.29434 0.1191082 Yj_log 1.710291 1.205334 0.5049568 0.1032761 Ni_log - 4.14599 - 0.6557726 - 3.490217 0.2789625 Nj_log - 0.9593455 - 0.2887502 - 0.6705953 0.2683556 CANDi - 0.1549637 0.0221008 - 0.1770645 0.0093853 NEGOi - 0.0370061 0.3361884 - 0.3731946 0.025329 ACCEi - 0.0657466 0.502517 - 0.5682636 0.0315696 EUi - 0.2672926 0.5869204 - 0.854213 0.0453241

b = consistent under Ho and Ha; obtained from xtreg B = inconsistent under Ha, efficient under Ho; obtained from xtreg Test: Ho: difference in coefficients not systematic

Chi² (8) = 425.36 Prob > chi² = 0.0000

Table 2. Breusch-Pagan Lagrangian multiplier test for random effects

Estimated results:

Var sd = sqrt (Var)

X_log 12.26054 3.501506

e 1.137353 1.066468

u 1.941181 1.393263

Test: Var (u) = 0

Chibar2(01) = 36332.78 Prob > chibar2 = 0.0000

Table 3. Breusch-Pagan / Cook-Weisberg test for heteroskedasticity

Ho: Constant variance

Variables: Yi_log Yj_log Ni_log Nj_log Dij_log CANDi NEGOi ACCEi EUi BORDij LANGij Chi² (11) = 13453.72