The effects of the EU : South Korea Free Trade Agreement : a panel data analysis of changes in bilateral trade

(1)

The effects of the EU - South Korea Free Trade Agreement:

A panel data analysis of changes in bilateral trade

Bobby Odenhoven 10246398

Economics and Business, University of Amsterdam Specialization: Economics

January 2016

(2)

Statement of Originality

This document is written by Student Bobby Odenhoven 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)

Acknowledgements

With the finalisation of this thesis, I would like to thank my supervisor Rutger Teulings for his guidance and help. Thanks to his useful advice and comments I was able to perform a regression method (dynamic models) that was beyond the scope of the course Econometrics in my Bachelor. Furthermore, I would like to thank him for emphasizing the importance of meeting deadlines, something that helped me to stay on track at times when I lacked a clear overview.

Additionally, I want to thank my friends and family who supported me while writing this thesis and who provided me with useful feedback and advice.

(4)

Abstract

The Free Trade Agreement between the EU and South Korea was implemented in 2011, eliminating nearly all trade tariffs. This thesis investigates to what extent the FTA has an effect on bilateral trade and its two components; imports and exports. Quarterly data measured among 21 countries is used, ranging from 2000-2014. These 21 countries include the fifteen largest countries in the EU and to include third country effects six large OECD trade partners of the EU are added. Three dependent variables; bilateral trade, export and import are tested against two FTA dummies that state whether the FTA is signed or implemented provisionally. This is done to measure whether there is a different effect for each implementation stage. The dummies measure the effect of the FTA once by using a static model and again by using a dynamic model. The model used in this research is based on the gravity equation and other variables that are included are the GDP, the NEER and the exchange rate. The results show that the FTA caused an increase in bilateral trade and export in the long run, implying the FTA had a positive effect. Specifically, it is found by using the dynamic model, that exports and bilateral trade are increased by 22% and 25% respectively.

(5)

1. Introduction

The Free Trade Agreement (FTA) between the European Union and South Korea was provisionally entered into force in July 2011. This FTA was the second largest trade agreement ever implemented globally and the most inclusive FTA the EU has ever

negotiated. Hence, the EU called it the first of a new generation of FTAs (European Union, 2011). Cooper (2011) states that at the time the negotiations started, South Korea was a much smaller market in comparison to the EU, it consisted of just 48.6 million consumers. Consequently it ranked 12th

as an export market for the EU. On the contrary, the EU market

with 492 million consumers was the second largest marketfor South Korean exports, with a

13% share of total exports. Cooper (2011) states that the FTA between the EU and South Korea was a step taken to strengthen both parties’ economic ties beyond their home regions. Furthermore he found that as the trade flows between the two parties had grown over the past decade, the FTA would build upon this trend. According to the European Union (2011) trade between the two parties was expected to increase due to the implementation of the FTA. More specifically, the European Union (2011) estimated this increase to be at an amount of €19.1 billion for the EU and €12.8 billion for Korea in the long run.

When the provisional agreement entered into force in July 2011, the majority of

custom duties were already removed and within the first five years of the implementation all the duties would be removed. The FTA would thus work in a progressive manner, by

removing more barriers over time. In 2007, four years prior to the implementation, the first negotiations had been launched between South Korea and the EU. According to an article published in the Korea Times in 2009, the FTA would boost the bilateral trade by as much as 40% in the long run (“S. Korea Strives to Ink FTA”, 2009). It has been eight years since these first negotiations took place and four years since the FTA was provisionally entered into force. Research has been limited to the expectations on what the effects of the FTA will be in the long run, leaving the data on what has actually happened until now untouched. This paper will research the effects that the FTA has had in the four years since it has been implemented. In particular, this paper will examine whether the FTA has had a significant effect on the bilateral trade between Korea and the EU. Furthermore it will examine whether the FTA has affected the two components of bilateral trade, export and import, independently. Hence, the paper will measure to what extent the Free Trade

Agreement has had an impact on bilateral trade, defined as exports from EU to South Korea added together with the imports of South Korea into the EU.

(7)

The paper will begin with a literature review where relevant information on

international trade and related topics will be discussed. This will be followed by a section discussing the data and method used for the empirical research. This section also includes a multiple regression model on panel data that is used to look for a significant relation

between bilateral trade and the FTA. Then, the results will be presented and the section following to this will discuss these results. Lastly, a conclusion will be drawn and the limitations of the research will be stated.

(8)

2. Literature review

2.1 General introduction to trade agreements

Agreements on trade are viable for Europe, according to the European Commission (2015). The Commission expects that over the next ten to fifteen years, 90% of world demand will be generated outside of Europe. Correspondingly, they make it a priority for Europe to stimulate opportunities abroad. To stimulate such opportunities, Europe could negotiate agreements on trade with its partners.

Global agreements on trade have a long history, starting with the General Agreement

on Tariffs and Trade (GATT), which was negotiated during the United Nations Conference on Trade and Employment (WTO, 2015). The GATT was signed in 1974 by 23 nations. It existed until 1994, when 123 nations signed it and the World Trade Organization (WTO) was created which took over the purpose of the GATT (WTO, 2015). Both the EU and South Korea are members of the WTO, together with 159 other countries.

As stated by the WTO (2015), the main objective of the agreements that they oversee

is that ‘trade flows as smoothly, predictably and freely as possible’. As noted by Balassa (1961), this objective can be met in two ways, by economic integration and economic cooperation. As he states, economic cooperation between countries aims at eliminating discrimination, so that a country does not prioritize some goods and services originating from one country to another. An example of this is the first article of the GATT, the most-favoured-nation (MFN), where it states that countries cannot discriminate against another. This article states that when a country grants a member of the WTO a special favour, it must do so for all other members of the WTO. In this way cooperation between all members is stimulated and consequently, trade will run more smoothly and freely.

The second concept Balassa (1961) mentions in his paper is economic integration, where economic policies between different countries become unified to make the flow of imports and exports easier. As Balassa (1961) states, economic integration sometimes compromises measures taken to reduce forms of discrimination. An example of economic integration is a free-trade area, in which tariffs and other restrictions between the included countries are removed, but where each country keeps its individual tariffs against countries that are not member of the free-trade area. In this way economic integration makes trade between some countries smoother and freer, but it fails to include all other countries that are member of the WTO. The reason that economic integration is possible in spite of the MFN is because the MFN allows for some exception; the creation of free trade area’s and

(9)

agreements being one of them (Brownsell, 2012).

Balassa’s five stages of economic integration

Balassa (1961) classified five different stages of economic integration, where each stage varies in the extent of integration. As mentioned by Crowley (2001), the stages are, respectively, a free trade area, customs union, common market, economic union and total economic integration. Over time, these stages originally classified by Balassa have been elaborated and now differentiate between a monetary union, a political union and an economic union (Laffan, O’Donnell, Smith, 2000). The stages of integration that are most commonly used today are summarized in Figure 1. An example of a customs union (stage 3) is the European Union Customs Union, where the EU together with a number of

neighbouring countries abolished all tariffs between countries in the customs union and implemented a common tariff for outside countries (Viner & Oslington, 2014).

It has to be noted that the stages of economic integration do not necessarily have to

happen sequentially. For a country to economically integrate with another it does not have to start at stage 1, it can also instantly implement the conditions necessary for stage 3.

Furthermore, it should be noted that for a common market, it can occur that the participating countries either keep their own tariffs against countries not included, or that they implement a common tariff. Additionally, for stage 6, a monetary union, no other ways of integration have to be fulfilled. The only necessary condition is that participants have a similar currency and one central bank. Finally, the last stage, as Balassa defines, calls for a unification of monetary, fiscal and social policies and it requires the implementation of a supra-national body that makes the decisions for the participating countries.

(10)

Figure 1. The stages of economic integration.

The figure shows to stages of economic integration based on Balassa’s (1961) original five stages.

2.2 The FTA between the EU and South Korea

The EU - Korea FTA was the most inclusive FTA ever negotiated as import duties were to be totally eliminated on nearly all products. More specifically, under the FTA industrial, fishery and a gricultural products from the EU would have reduced or zero tariffs when imported into South Korea (European Union, 2011). As Krugman, Obstfeld and Melitz (2012) state, tariffs are the simplest form of trade policies with the mere reason to raise the cost of shipping goods to another country. The purpose of tariffs can be divided into two, with the first being the provision of revenue and the second one being the protection of specific domestic sectors.

The European Union found that as the FTA eliminates trade tariffs, both European

and South Korean importers would save money by the eliminated costs of import duties. It was found that the total amount of import duties saved by Korea could reach € 850 million (European Union, 2011). Due to the eliminated tariffs, prices would fall causing increased competitiveness for exporters. It was thus expected that the FTA would benefit both

1. Preferential trading area

- tariffs for participating countries are reduced, not removed 2. Free trade area

- tariffs between participating countries are fully removed - countries’ national tariffs are kept against outside countries 3. Customs union

- tariffs between participating countries are fully removed - common external tariff for outside countries

4. Common market

- tariffs between participating countries are fully removed - free movement of capital and labour

5. Economic union

- free movement of capital and labour 6. Monetary union

- unifying the currency between participating countries - a single central bank

7. Total economic integration

- free movement of capital and labour

- unifying the currency between participating countries - same or near same fiscal policy

- integration of politics

(11)

countries by increasing their trade.

In addition to the elimination of tariffs, the FTA called for alternative measures to

stimulate trade. Specifically, it called for cooperation between Korea and the EU on

reducing technical barriers to trade, including transparency and cooperation on standards and regulations (Free Trade Agreement, 2010). As mentioned by Finalyzson and Zacher (1981), an example of a technical barrier is the difference in technical standards for certain products between countries, making trade in these products more difficult than other products.

As mentioned by Krugman, Obstfeld and Melitz (2012), a technical barrier is a form

of a non-tariff barrier. Non-tariff barriers are certain restrictions that limit trade between countries, but not in the form of a tariff. An example of a non-tariff barrier is an import ban or a quota on certain imported or exported products. These non-tariff barriers have the same effect and purpose as a tariff but are different in how they operate. A tariff is a certain tax that is levied on a product while a non-tariff barrier can take multiple forms, one of them being a technical barrier.

The improved non-tariff barriers called for in the FTA were mostly relevant for trade

in four sectors. These sectors include consumer electronics, motor vehicles and parts, pharmaceutical products and chemicals. As can be seen from figure 2, which represents the main exported and imported products of the EU from South Korea, these sectors accounted for a large share of the total trade. Specifically, they accounted for 70% of the imports and 61% for the exports in 2011 (The European Commission, 2015). For these four sectors the FTA introduced stronger transparency rules and the altering of regulations so that most products would comply with international standards. With these new measures enforced, trading of these products was simplified and was therefore likely to increase with the implementation of the FTA.

Top 5 exported products

1. General industry machinery & equipment 2. Transport equipment

3. Chemicals & pharmaceuticals

4. Primary products (agricultural products, food, etc.)

5. Scientific and controlling instruments

Top 5 imported products

1. General industry machinery & equipment 2. Transport equipment

3. Office and telecommunication equipment 4. Iron and steel

5. Chemicals & pharmaceuticals

Figure 2. The top 5 exported products from the EU to South Korea and the top 5 imported products from South Korea in 2011 (European Union, 2011).

(12)

Balassa’s five stages applied

When looking at the situation discussed in this paper, we can distinguish several stages of integration that the EU and South Korea went through. In 1997, the two parties signed an ‘Agreement on Cooperation and Mutual Administrative Assistance in customs matters’. With this agreement, the two parties would cooperate in simplifying, harmonizing and computerizing customs procedures (European Community, 1997). By doing this, the parties were to exchange each other’s professional, scientific and technical data related to their custom regulations.

In 2001, the two parties signed another agreement that was called ‘the Framework

Agreement on Trade and Co-operation’ (Framework Agreement for Trade and Cooperation, 2001). This agreement aimed at increasing cooperation in and diversifying trade to benefit both countries. To do so, the two parties would work towards eliminating barriers to trade in the future, by improving transparency and gradually removing non-tariff barriers.

When looking at this last agreement, it can be said that by signing this agreement,

EU and South Korea entered stage 1 of economic integration, a preferential trading area (PTA). By signing the Framework Agreement on Trade and Co-operation the parties agreed on gradually removing non-tariff barriers in certain areas, which are characteristics of a PTA. The parties did not remove all tariffs nor did they implement a common external tariff, so they remained in stage 1.

In 2010 however, the two parties signed a bilateral free trade agreement, moving to

stage 2. By signing this FTA, as mentioned earlier, the two parties agreed to eliminate or reduce trade barriers in manufactured goods, agricultural products and services. The

agreement did not aim for the alteration of the tariffs for outside countries, nor did it aim for free movement of capital and labour. Because the two parties did not implement a common external tariff, the FTA developed a so-called ‘rules of origin’ (European Community, 2011). This means that only products originating in either South Korea or the EU can benefit from the preferences granted by the FTA.

It can therefore be concluded that by signing the FTA, South Korea and the EU

moved to the second stage of economic integration, a free trade area.

Trade diversion and trade creation

The first effect of a FTA is trade diversion. As described by Krugman, Obstfeld and Melitz (2012) diversion occurs when economic integration causes trade to be diverted away from a more efficient supplier to a less efficient supplier. This means that due to a FTA a country

(13)

stops trading with a more efficient country that is not included in the FTA, but instead diverts to a country that is included (Viner, 1950). To illustrate this, imagine the situation where a country (A) trades with two other countries (B and C). Without the implementation of any trade tariffs, the price of country C is lower than country B, meaning that country C is more efficient at supplying the good. When tariffs are introduced, both prices are increased by the same amount, resulting in a lower price of country C with tariffs. When a FTA is formed between country A and B, all trade tariffs on both countries’ products will be

eliminated. This could cause the price without tariffs of country B to be lower than the price of country C with tariffs, which would mean that country B would become the new trade partner. From this example it becomes clear that with a FTA, the price of good from a country included in the FTA can become lower than the price of another, more efficient, country that is still subject to trade tariffs.

The second effect that can take place is trade creation. As Viner (1950) describes,

trade creation occurs when economic integration creates trade that would not have taken place without this integration. With implementing a FTA, this could cause the included countries to increase their trade, which would not have happened without the FTA. Trade creation takes place when first the prices with tariffs of the other country are higher than domestic prices. In this case a country does not import a certain product but it rather produces it domestically. According to Suranovic (2007), under the FTA the tariffs are removed and prices of products in the other country decrease and become lower than domestic prices. As a result the country will start importing the products from the other country, instead of producing them domestically, hence more trade is created (Viner, 1950). This is different than trade diversion, where the country is engaged in importing already but where the country changes from a more efficient trade partner to a less efficient one.

Trade creation and diversion- effects on surpluses

With trade diversion, the FTA changes the price of an imported product as tariffs for the included countries are eliminated and the product imported from the included country becomes cheaper than the product from non-included countries (Suranovic, 2007). In reaction to the decreased price, domestic producers will lower the price of their product in order to compete with the imported product. The changes in domestic prices as well as the price of the imported product have effects on consumer surplus, producer surplus and on the government revenue. The consumer surplus will increase, as prices of imported goods as

(14)

well as domestic goods will decrease. The producer surplus and the government revenue will decrease, as all the revenue of the tariffs on the imported product will be lost due to the FTA (Suranovic, 2007).

As explained by Viner (1950), trade creation has effects on the consumer surplus and

the producer surplus. It does not affect the government revenue as the concerning product was originally not imported but produced domestically, so there was no tariff revenue for the government in the initial situation. Consumer surplus will increase, as the price of the

product is now lower than before. Producer surplus will decrease as the good is now imported instead of produced domestically. Also, in order to compete with the imported products, domestic producers have to lower the prices of their substituting products. Although this results in a loss for producers, it is a gain for consumers and therefore has an offsetting effect (Suranovic, 2007).

Expectations and hypotheses

By expanding trade agreements beyond the GATT, South Korea and the EU moved two stages forward in economic integration. Furthermore, with the implementation of the FTA, trade-diverting effects could possibly take place, causing trade between the EU and South Korea instead of more efficient countries. On the other hand trade creation effects are possible by the elimination of tariffs and non-tariff barriers to trade.

When looking at the existing literature concerning the effects of a FTA, it is found that FTAs mainly lead to an increase in trade. Firstly, Baier and Bergstrand (2007) found that on average a FTA doubles the members’ bilateral trade after 10 years. Moreover, Wong and Chan (2003) found that the FTA between China and ASEAN (Association of Southeast Asian Nations) had a significant positive effect on trade between the two parties. Following the main consensus of the literature, it is thus expected that the FTA between EU and South Korea will increase bilateral trade.

For this reason the hypothesis that will be tested is that the FTA resulted in an

increase in total bilateral trade. Furthermore, the hypotheses that the FTA increased exports and imports independently are tested as well. Additionally, it will be investigated whether there was a difference in impact between the signing and the provisional implementation of the FTA. From the information given on the website of the European Council, it can be recognized that there was notable time period between signing the agreement and the provisional implementation for each country. A table of these exact dates can be found in table A1 in the appendix.

(15)

2.3 Factors determining trade

To evaluate the effect of the FTA on bilateral trade it is also important to identify factors that determine the volume of trade between countries. In the section following, the relevance of adding these factors to the empirical model will be explained.

Firstly, as Tinbergen (1962) mentions, there is a strong relationship between the size

of a country’s economy (gross domestic product) and the volume of its exports and imports. The gravity model was originally proposed by Newton to define the attractive force between two objects (Head & Mayer, 2013). Later, Tinbergen (1962) found that this model could also be applied to international trade flows. More specifically, he found that when looking at the monetary flow of trade between two countries, this was dependent on both countries’ GDP and also on the distance between them. Large economies, economies with a high GDP, tend to spend large amounts on imports because they have large incomes. Moreover as these countries will probably produce a wide range of products, they are likely to attract other countries’ spending on exports.

However, as exports and imports are used to measure the GDP of a country, trade

influences the size of GDP too. Thus, it can be said that there is simultaneity between size of trade and the size of the GDP between the two parties. The higher the GDP, the higher the value of total trade will be and vice versa. As one country’s import is another country’s export, total trade between two countries can be calculated by adding two countries’ exports or imports. On the other hand, bilateral trade is calculated differently than total trade. In this paper, bilateral trade will be computed by adding the exports from the EU to South Korea to the imports from South Korea into the EU. It should be noted that since bilateral trade makes up only a fraction of total trade, this simultaneousness will play no significant role in this research.

Another factor that is considered to be important for the volume of trade is the

nominal effective exchange rate, the NEER (United Nations, 2008). The NEER depicts the relative price of a country’s currency compared to all major currencies traded, while the ‘normal’ exchange rate represents the price of a currency compared to only one other currency (United Nations, 2008). As stated by the Bank of Japan (2009), the NEER of a country is calculated by taking the geometric average of nominal exchange rates with its largest trading partners, taking into account the shares in total exports and imports that each of those trading partner has. In order words, each trading partner is given a ‘weight’

(16)

adjusted for this ‘weight’ (Bank of Japan, 2009). After this step, the average value is taken from these adjusted nominal exchange rates, which is the NEER.

As Bahmani-Oskooee (2010) found in his research, a decrease in the NEER reflects a

depreciation of the domestic currency compared to its trading partners in nominal terms, while an increase in the NEER reflects an appreciation. Several studies have been performed to measure the effect of depreciation of the NEER on the trade balance of a country. In these studies, trade balance is defined as the difference between countries’ exports and imports. As the study of Houthakker and Magnee (1969) show, a depreciation of the currency improves the trade balance for a country. This means that when the NEER depreciates, the domestic currency becomes relatively cheaper, resulting in exports of this country to become relatively cheap and imports relatively expensive. It would thus mean when the NEER depreciates, the country would export more but import less products, leading to an increase in the trade balance. As this paper measures bilateral trade and not the trade balance, trade is defined as EU’s export to South Korea together with imports from South Korea into the EU. Therefore a change in the NEER will have an ambiguous effect on the bilateral trade, as the change in NEER has contradictory effects on exports and imports.

The exchange rates between two countries’ currencies, in this case the exchange rate

between euro and the Korean won, also influence trade. As Pilbeam (2013) mentions, when we define the exchange rate as domestic currency units per unit of foreign currency,

depreciation of the domestic currency implicates that the domestic currency becomes cheaper than the foreign currency. This means that when there is depreciation, the country’s imports will decrease and the exports will increase. As these are opposite effects, the total effect on bilateral trade will be ambiguous.

In order to form an expectation on what will happen in case of depreciation the

Marshall-Lerner condition must be fulfilled. As Bahmani-Oskooee and Niroomand (1998) state, the Marshall-Lerner condition formulates that the sum of the import and export elasticities of demand is greater than 1. When this condition is fulfilled, it is found that a depreciation of the currency will have a positive effect on bilateral trade. This is true as in that case the positive effect on exports will be larger than the negative effect on imports. In the rest of this paper it is assumed that the Marshall-Lerner condition is fulfilled. It can therefore be summarized that in this paper, depreciation of the currency is expected to have a positive effect on bilateral trade.

(17)

3. Methodology

3.1 The static model

In order to test the hypotheses formed in section 2, the following basic regression model will be used:

log 𝑇𝑟𝑎𝑑𝑒 !,! = 𝛽!log (𝐺𝐷𝑃)!,!+ 𝛽!log (𝑁𝐸𝐸𝑅)!,!+ 𝛽!log (𝑆)!,!+ 𝛽!𝐹𝑇𝐴!"#$%&! + 𝛽!𝐹𝑇𝐴!"#$%&!+ 𝛼!+ 𝜃! + 𝜏!∗! + 𝜀!,!

The dependent variable in the multiple regression is formulated as bilateral trade, the sum of the value of exports from the EU to South Korea and the value of imports from South Korea into the EU. To measure the independent effects on exports and imports, two additional regressions will be performed where imports and exports will function as the dependent variable. In this basic regression model, bilateral trade is defined as ‘Trade’.

The independent variables that will be used in this research are: the gross domestic

product of the exporter represented by 𝐺𝐷𝑃!,! , the nominal effective exchange rate of the exporter (𝑁𝐸𝐸𝑅!,!), the exchange rate (𝑆!,!), the dummy variable for a signed (𝐹𝑇𝐴!"#$%&!) and the dummy variable for the provisionally applied FTA (𝐹𝑇𝐴!"#$%&!). For these dummy variables, a 1 depicts that the FTA has been signed or provisionally applied, while a 0 depicts the opposite.

To make later interpretations of the estimated variables easier the natural logarithmic

values are used, as this represents percentage changes. Using natural logarithmic values also takes out large fluctuations in the values of the real data by using percentages. Furthermore, as the variables in the data used take large values, their standard errors are likely to be big. Using logarithms solves this problem.

Furthermore, time, country and trend fixed effects are added to the model. The time

fixed effects are added to correct for any country-invariant effects that change over time. An example of such a time effect is whether or not the world economy is in a recession at a certain time period. Instead of time fixed effects, the model could also correct for relevant effects of South Korea. These South-Korea specific effects would include the GDP, exchange rate and NEER for South Korea. As these effects are the same for all the EU countries but change over time, they are a specific form of time effects. However, the general time fixed effects correct for a larger scale of time effects than the South Korea specific effects. If we only include the South Korean effects in the form of the GDP,

(18)

exchange rate and NEER, there will be an omitted variable bias. Hence, to counter this potential outcome the general time fixed effects are used, not the South Korea specific effects.

Furthermore, country fixed effects are added. Examples of such country specific

time-invariant effects are the characteristics of a political system of a country, which could have an influence on a country’s GDP and trade.

Additionally, trend fixed effects are added to correct for unobserved country specific

effects. An example of such an effect is that transportation costs for countries are declining over time. The rate of which these costs decline differs per country and therefore they cannot be depicted by country fixed effects. By adding the fixed effects to the model, the time, country and trend effects are removed so that the model only focuses on the effects that the independent variables varying over time have on bilateral trade.

In order to interpret the outcome of the regression, certain assumptions on the error

term must be made in order for the estimates of the coefficients to be unbiased, efficient and consistent. Firstly, the error term must be unrelated to the predictors (Stock & Watson, 2007). Secondly, as stated by Stock and Watson (2007), the error terms must be independent of each other; there is no autocorrelation allowed. Autocorrelation occurs when in panel data, the error term from one observation are correlated with the error term of another observation. Thirdly, the errors are assumed to be homoskedastic. When this assumption is met, the error term has a constant variance across chancing values of the independent

variables (Stock & Watson, 2007). As stated by Antonie, Cristescu and Cataniciu (2010), the heteroskedasticity assumption is restrictive when used in panel data. It is likely that the error term will have a different variance across the countries used in the dataset, and that there will be heteroskedasticity.

3.2 The additional model; the dynamic model

In addition to the first model, another model will be tested to include the dynamics of the first model. As the first model is a stationary model, the statistical properties of the estimators are assumed not to change over time (Nason, 2006). Stationarity in a model implicates that the mean and variance of the different variables estimated in the model do not depend on the time periods, and are thus constant (Pelletier, 2014). Trade is dynamic, which means that countries in the EU now decide if they want to trade with South Korea in the future. The decision made is therefore based on economic indicators of today. It is

(19)

therefore unlikely that the assumption mentioned above is fulfilled, and an additional non-static model needs to be tested.

The dynamic model adds time lags of the variables to the existing model, to correct

for the correlation of the current estimators with the estimators of previous time periods. For each variable the most relevant time lags are chosen and added to the regression model. That is, when the GDP of a certain year is likely to be correlated with the GDP of the year before, the lagged GDP of one year will be included in the model. For the dependent variable one lag will be added, to correct the effect on current trade for the effect of trade from previous years. Also for the GDP one lag of a year will be added.

The quarterly GDP is an indicator that is rather complex to estimate (Young, 1993).

This is why the estimated value of the GDP is usually prepared one month after the quarter has already ended (Hirsch, McCully, Parket et al., 1993). Therefore it is likely that the GDP of previous time periods is correlated with the GDP of the current period. The lagged GDP will probably have an influence on trade in the current period too. In the model a lag of one year will be included, as it is assumed that the GDP correlates with the GDP of one year before. For the NEER and exchange rate four lags will be added, to measure the effect of previous quarters on the current estimators.

The model including lagged effects of the variables is represented below:

log (𝑇𝑟𝑎𝑑𝑒)_!,! = 𝛽_!log (𝑇𝑟𝑎𝑑𝑒)_!,!!!+ 𝛽_!log (𝐺𝐷𝑃)_!,! + 𝛽_!"log (𝐺𝐷𝑃)_!,!!! + 𝛽_!log (𝑁𝐸𝐸𝑅)_!,!+ 𝛽_!"log (𝑁𝐸𝐸𝑅)_!,!!!+ 𝛽_!"log (𝑁𝐸𝐸𝑅)_!,!!! + 𝛽_!!log (𝑁𝐸𝐸𝑅)_!,!!!+ 𝛽_!"log (𝑁𝐸𝐸𝑅)_!,!!!+ 𝛽_!log (𝑆)_!,!

+ 𝛽_!"log (𝑆)_!,!!!+ 𝛽_!"log (𝑆)_!,!!!+ 𝛽_!"log (𝑆)_!,!!!+ 𝛽_!!log (𝑆)_!,!!! + 𝛽_!𝐹𝑇𝐴!"#$%&

!+ 𝛽!𝐹𝑇𝐴!"#$%&!+ 𝛼! + 𝜃! + 𝜏!∗!+ 𝜀!,!

To estimate the total effects of the independent variables, the delta method1

will be used. This method can be used to calculate the long run effect of the variables, which are comparable with its estimated counterparts in the static model. As pointed out by Sosa-Escudero (2009), the aggregate effect of the variables will be their long run effect. The standard errors of the estimates will be calculated with the delta method as well.

As mentioned by Nickell (1981), adding lagged variables in a fixed effect regression

1_{When a simple regression model is Y}

t= α + β0X1t + γYt−1 + 𝜀, the delta method calculates the long run effect

(20)

will cause the lagged variables and error term to be correlated. He found that the variables estimated with a dynamic fixed effects model would be biased with a factor 1/T, where T is the number of time periods used in the dataset. This leads to the conclusion that the

estimates of the dynamic model will be biased with approximately 1/60. As this is a relative small factor the bias will be ignored in this paper.

3.3 The regression method

Since the data used is in panel-data format either a random effects or a fixed effects model can be used to estimate the variables. In a random effects model, the variation across countries (country, time and trend effects) is assumed not to be correlated with the independent variables (Torres-Reyna, 2007). Therefore, in a random effects model, those variations are included in the error term. An example of this is that in a random effects model, the political system of a country (country effect) must not be correlated with the GDP of a country. In a fixed effects model, this assumption is relaxed. With using a fixed effects model, the characteristics of each country are allowed to be correlated with the independent variables in the model. Using the example from before, the political system of a country may influence the GDP in a fixed effects model, something that is likely to be true in this research. Therefore those variations are included as fixed effects in the model, instead of including them in the error term.

When looking at the model used in this research, the variations across countries are

likely to have an influence on the independent variables. The fixed effects model is therefore the appropriate model to use. A Hausman test2 is performed to check whether the

appropriate model to use is indeed a fixed effects model. As the outcome of this test rejected the null hypothesis3 the assumption to use a fixed effects model is confirmed.

2_{A Hausman test compares random effect and fixed effect estimates and can be used to decide which model is}

the appropriate one to use (Cameron & Trivedi, 2010).

(21)

4. Data

4.1 The data set

As this paper aims to measure the effect the FTA had in the last four years, the data used is gathered over a time period of fourteen years. By analysing these fourteen years, the period before the implementation of the FTA can be compared with the period after to see if any changes occurred. The variables are measured on a quarterly basis from the first quarter of 2000 to the last quarter of 2014, covering 60 quarters in total.

As the EU currently consists of 28 member states, the data also has to be measured

over different countries. The paper focuses on the following selection of countries: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg,

Netherlands, Portugal, Spain, Sweden and the United Kingdom. The variables are observed for these 15 countries over 60 quarters.

In the following paragraphs the incentives for the selected timeframe and the choice

of countries will be discussed. As data over different time periods and different countries is used, a panel data set is created. By performing a multiple regression on the gathered data, it is possible to measure the effect of the FTA on bilateral trade.

Time scope of the dataset

The data used in this paper limits its time scope to fourteen years and only includes fifteen EU countries. The data needs to be gathered over 60 quarters and to do so different sources are used. For some countries included in the research, data on the variables lacked for years earlier than 2000. For other countries, the data was incomplete. In order to create a reliable and balanced dataset, the time scope will be limited to the years 2000-2014.

Selection of countries in the dataset

In the FTA all the 28 member states of the EU are included. To obstruct from missing data for certain variables, the paper limits itself to the fifteen largest countries in the EU, reducing the risk of an incomplete dataset.

Figure 3 depicts the GDP in millions of US dollars of the countries included in this

research, the first fifteen countries, compared to the ones not included and the total GDP of all the member states. The figure shows the size of GDP for two years, 2000 and 2014. From this figure it can be seen that the first 15 countries have a much larger GDP than the other 13 countries. More specifically, in 2000 the first fifteen countries accounted for 90% of the total GDP of the 28 member states. In 2014 this number decreased to 86%. In figure 2A in the

(22)

appendix the graphs for all the relevant years can be found. For all the years used in this dataset, the share of the total GDP of the first 15 countries is bigger than the other 13 countries. It can therefore be stated that the fifteen countries used in this research are the ones with the highest GDP.

In addition to the fifteen European countries in the dataset, six supplementary

countries are added. The control countries included are Canada, Chile, Japan, Mexico, Switzerland and the United States. As all the countries in the EU signed the FTA

simultaneously, it is likely that the dummy variables of the EU will be perfectly correlated. Adding the six countries will solve the correlation problem of the dummy variable FTAsigned.

Furthermore, as discussed in the literature review, the introduction of the FTA can

have two effects; trade creation and trade diversion. As these two effects are dependent on another country that is not included in the FTA, we can call those effects third country effects (Chen & Joshi, 2010). As the extra countries represent large OECD countries it will be likely that they are trade partners of the EU. To include the third country effects in the model the six supplementary countries are added.

.

Figure 3. Shares of the total GDP

This figure shows the shares of the total GDP of the first 15 countries and the other 13 countries. The first fifteen countries include: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden and the United Kingdom. The other 13 include: Bulgaria, Croatia, Cyprus, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Malta, Poland, Romania, Slovakia and Slovenia (OECD.stat, 2015).

90% 10%

Share of total EU GDP in 2000

First 15 countries Other 13 countries

86% 14%

Share of total EU GDP in 2014

(23)

4.2 The variables Dependent variable

As mentioned earlier, the dependent variable bilateral trade is defined as total export from the EU to South Korea added together with the total imports of the EU originating from South Korea. Moreover, the two components of bilateral trade, exports and imports, will function as dependent variables. To compute the value of these variables, data from the International Monetary Fund is collected. The IMF-website contains a Direction of Trade Statistics (DOTS) database, where countries list the value of their exports and imports between them and their trading partners. In this research, the exports in US dollars from the 15 included EU countries to South Korea and their imports in US dollars originating from South Korea are used.

Independent variables

To estimate the effect of the FTA, several independent variables correlated with the

dependent variable are included in the regression. Firstly, the GDP of the fifteen countries is included. To gather the data on this variable, the quarterly nominal GDP data sourcing from the OECD Economic Outlook is collected. For countries in the Eurozone this data is

measured in millions of euro, for countries outside the Eurozone it is measured in the national currency. The GDP data is converted into millions of dollars using the quarterly exchange rate of the country’s currency.

Secondly, the NEER of the fifteen countries is included. Data on the NEER is

gathered from the OECD Economic Outlook (Datastream, 2015). The NEER is measured quarterly among the 15 different countries and it is computed as index numbers, with 2010 as the base year.

Thirdly, the exchange rate of the fifteen countries is included. The exchange rate of

the euro and the other currencies are gathered from the OECD-website. The exchange rates from this website are quoted as national currency per US dollar. The European national exchange rates against the US dollar and the Korean exchange rate against the dollar are transformed into the exchange rate between the national currency and the Korean won. The exchange rate used in the model is then quoted as Korean won per European currency.

Lastly, two dummies for the implementation dates of the FTA are included. The first

dummy variable depicts whether or not the country signed the FTA, whilst the other dummy depicts whether the FTA was provisionally implemented. The corresponding information on

(24)

these dates is found on the website of the European Council. The exact dates are then transformed into quarterly dates and added to the dataset.

Moreover, an extra dummy variable is added which states whether the FTA is

completely implemented or not. The dates for the complete implementation differ from the dates the FTA was provisionally implemented.

4.3 Descriptive statistics

Table 1 shows the most important statistical information on the logarithmic values of the variables.

Table 1. Summary of the statistics

Variable Observations Mean

Std.

Dev. Minimum Maximum

all log-values Bilateral trade 1260 2.915 0.731 0.529 4.470 Export 1260 2.47 0.82 0.315 4.231 Import 1260 2.615 0.803 - 0.264 4.264 GDP 1260 5.816 0.594 4.299 7.248 NEER 1260 1.995 0.043 1.845 2.187 Exchange rate 1260 2.824 0.767 0.213 3.333 FTA-signed 1260 0.296 0.457 0 1.000 FTA-provisionally 1260 0.237 0.426 0 1.000

This table summarizes the most important statistical information on the dependent variables; bilateral trade, export and import. It also states the information on the independent variables. All the variables are measured in logarithmic values with base 10.

In Table 3 the number of observations, the mean, standard deviation, minimum and maximum value of the variables used in the model can be found. It can be seen that the dataset is balanced. By taking the logarithmic values the standard deviations are not as high as they would have been using the untransformed numbers. It can be seen that the minimum value of Import is a negative number. As the value of imports and exports is measured in millions, a number lower than one million will be depicted as a number smaller than 0. By transforming the variables in logarithmic functions, values smaller than 0 become negative.

(25)

In Table 2 the correlations between the independent variables are depicted. It can be seen that there are no high correlations, except for the correlation between the dummy variable FTA-signed and FTA-provisionally. However, as these variables are dummy variables they are nominal and not ordinal. The Pearson correlation test, used in Table 2 measures whether there is a linear relation between two variables. As the dummy variables can take the values 1 and 0 it is therefore not relevant to measure the correlation between them and this high correlation can be ignored.

Table 2. The correlation between the independent variables

GDP NEER Exchange rate FTA-signed FTA-provisionally GDP 1 NEER 0.1908 1 Exchange rate -0.0002 0.1769 1 FTA-signed 0.0198 0.0933 -0.0414 1 FTA-provisionally -0.0434 0.1051 -0.0649 0.8602 1

This table shows the correlation between the independent variables in the model. It can be seen that there are no high correlations, the highest one being the correlation between the two FTA dummy variables.

(26)

5. Results

In this section the results of the empirical research will be discussed. First the results of the static regression will be analysed followed by the results of the dynamic regression.

5.1 Testing the residuals

Before running the regression, the assumptions on the error term, stated in the third paragraph, must be tested. After running a modified Wald test4 it is found that there is no homoskedasticity, the residuals are heteroskedastic. Also, a Lagram-Multiplier test5 is performed to check for autocorrelation. It is found that there is autocorrelation as the null hypothesis can be rejected. Lastly, a Parasan test6 is done to test whether the residuals are cross-sectional dependent. This would mean that the residuals are correlated across the countries used in the data set. It is found that the residuals are not correlated; they are thus cross sectional independent. To correct for heteroskedasticity and autocorrelation, the regression of the static model will use robust and clustered standard errors as suggested by Hoechle (2007). As in the dynamic model lags of the variables are added, this model will only be corrected for heteroskedasticity, so robust standard errors will be used.

5.2 Results of the regression on the static model

The estimates of the variables found in the static regressions are summarized in Table 3. The results from the model without the inclusion of FTAcomplete will be analysed. The dummy variable FTAsigned is insignificant for all the dependent variables. However, the dummy variable FTAprovisional is significant for total trade. As the dummy variables are not

logarithms, they need to be interpreted differently than the other variables. As the dependent variable is a natural logarithm, the effect of the FTA dummy can be calculated by taking 𝑒!.!"_{and transforming that into a percentage. After doing this, it is found that by applying} the FTA provisionally, total trade is increased by 22%.

When looking at the logarithm of GDP, it can be seen that this variable is not

significant for all three dependent variables. It does have a positive effect on total trade and import, something that is expected. The negative effect on export is not expected but as the result is not significant it will be ignored.

The effect of the log(NEER) is significant for only exports. As all variables are

4_{Modified Walt test for group wise heteroskedasticity, the null hypothesis is homoscedasticity (Torres-Reyna,}

2007)

5_{Lagram-Multiplier test for serial correlation, the null hypothesis is no serial correlation (Torres-Reyna, 2007).} 6_{Parasan CD test for cross-sectional dependence, the null hypothesis is that residuals are not cross-sectional}

(27)

logarithms, it means a 1% increase in the NEER will have a 1.95% increase in exports. An increase in the NEER is depreciation, and as stated before in the literature review, will cause exports to increase. This result is therefore in line with what was expected from the

literature.

As in this model the exchange rate is defined as the amount of Korean won per

European currency, an increase in the exchange rate would mean an appreciation of the European currency. This implicates that when the exchange rate increases with 1%, the Korean Won depreciates and it would be expected that exports from the EU decrease and imports from EU out of South Korea would increase. This is because for depreciation of the Korean won, the won becomes cheaper than European currency, stimulating EU countries to import but discouraging European exports to South Korea because it becomes too expensive. From Table 3 it can be seen that for total trade, import and export the logarithm of the

exchange rate is significant. This means that when the exchange rate increases by 1%, the total trade, export and import will increase by 0.17%, 0.06% and 0.17% respectively. The positive effect on exports are not expected, however it could imply that an increase in the exchange rate has a bigger effect on imports, leading to a positive result in total trade.

When the results of the two regressions, one with the FTA complete dummy added and

the other one without are compared, it can be seen that the results of the estimations do not differ significantly. When omitting the FTA complete dummy, the standard errors of the estimates also do not change significantly when compared to the model without the extra dummy. It can therefore be said that by leaving the FTA complete_{variable out of the}

regression, there was no case of omitted variable bias (Stock & Watson, 2003).

(28)

Table 3. Results of the static regression

Table 3 shows the results of the first regression, with either the log of total bilateral trade, log of export or log of import as dependent variable. The table shows the results using clustered robust standard errors, corrected

for heteroscedasticity, serial correlation and non-normality. The standard errors of the estimations are given between parentheses. The estimated effects of the independent variables are calculated with including FE ,θt and τit . The effects are estimated once with the addition of the dummy FTAcomplete_{and once without. The R}2 _is

added to show how much of the variance in the dependent variable is explained by this model. A * indicated significance at α=5% and **indicates significance at α=1%

5.3 Results of the regression on the dynamic model

The estimates of the variables found in the dynamic regressions are summarized in Table 4. The results from the model without the inclusion of FTAcomplete

will be analysed. For the FTA effects, it can be seen that the dummy FTA provisional

is the only effect that is significant for both total trade and exports. It is found that the provisional implementation increased total bilateral trade by 25% in the long run, and increased the exports by 22% in the long run. Both the effects are positive, which was in line with the expectations. Furthermore, the effects of the provisional implementation of the FTA are bigger in dynamic model than in the static model.

It can be seen that in the dynamic model, the estimates of the logarithm of GDP are insignificant for all dependent variables. This is the same for the static model and therefore it can be stated that there is not a significant effect of GDP on bilateral trade.

The effect of the NEER is only significant for the export, and it is found that an increase in the NEER of 1% leads to a long run increase in exports of 2.15%. This positive effect corresponds to the expectations from the literature review. The effects of the NEER on total trade and imports are not statistically significant so there is not enough evidence to state

Including

FTAcomplete_? NO YES

Dependent

variable Total Trade Export Import Total Trade Export Import log(GDPi,t) 0.39 (0.30) -0.90 (0.71) 0.60 (0.46) 0.38 (0.29) -0.90 (0.71) 0.60 (0.45) log(NEERi,t) 0.15 (0.40) 1.95* (0.79) -0.10 (0.53) 0.13 (0.39) 1.95**(0.80) -0.14 (0.52) log(Si,t) 0.17**(0.02) 0.06** (0.02) 0.17** (0.03) 0.17**(0.02) 0.06** (0.02) 0.17** (0.02) FTAsigned _{-0.08 (0.08)} _{-0.07 (0.08)} _{-0.13 (0.08)} _{-0.07 (0.07)} _{-0.07 (0.09)} _{-0.12 (0.07)} FTAprovisional _{0.20* (0.08)} _{0.18 (0.09)} _{0.13 (0.15)} _{0.13 (0.08)} _{0.16 (0.09)} _{0.02 (0.17)} FTAcomplete _- _- _- _{0.10 (0.07)} _{0.02 (0.09)} _{0.15 (0.11)} R2 _0.79 _0.81 _0.59 _0.79 _0.81 _0.59

(29)

the NEER had a long run effect on total trade and imports.

The logarithm of exchange rate is significant for the total trade and import as

dependent variables. Moreover, the results show that a 1% increase in the exchange rate leads to a 0.31% increase in total trade and a 0.22% increase in imports in the long run. As stated before, the positive effect on total trade and imports are in line with what is expected.

When the results of the two regressions, one with the FTA complete dummy added

and another one without, are compared, it can be seen that the results of the estimations do not differ significantly. Therefore by leaving out the variable FTAcomplete

there was no omitted variable bias (Stock & Watson, 2007).

Table 4. Results of the dynamic regression

5.4 Empirical results and discussion

In this paper it is found that the provisional implementation of the FTA between the EU and South Korea increased total trade and exports. This is in line with the prediction stated in the literature review. More specifically, by implementing the FTA the exports are increased by 22% in the long run. The bilateral trade is increased by 25% in the long run.

By implementing the FTA, the two parties moved forward in economic integration.

From the results it can be concluded that this increased trade between the two parties. This finding corresponds with what was expected by the European Union (2011).

Including

FTAcomplete_? _NO _YES

Dependent variable Total Trade Export Import Total Trade Export Import

Long run variables

log(GDPi,t) 0.04 (0.27) -1.17 (0.83) 0.36 (0.50) 0.05 (0.26) -1.17 (0.83) 0.36 (0.49) log(NEERi,t) 0.31 (0.44) 2.15* (0.88) -0.08 (0.62) 0.29 (0.44) 2.14* (0.89) -0.11 (0.61) log(Si,t) 0.27**(0.04) 0.17 (0.09) 0.22** (0.06) 0.28**(0.04) 0.17 (0.10) 0.22**(0.05) FTAsigned _{-0.09 (0.09)} _{-0.08 (0.08)} _{-0.15 (0.11)} _{-0.08 (0.09)} _{-0.08 (0.09)} _{-0.14 (0.10)} FTAprovisional _{0.22** (0.07)} _{0.20* (0.09)} _{0.14 (0.16)} _{0.16 (0.08)} _{0.20* (0.09)} _{0.03 (0.19)} FTAcomplete _- _- _- _{0.08 (0.08)} _{0.01 (0.09)} _{0.15 (0.13)}

Table 4 shows the results of the second regression, using a dynamic model, with either the log of total trade, log of export or log of import as dependent variable. The table shows the results using robust standard errors, corrected for heteroskedasticity.

The standard errors of the estimations are given between parentheses. The estimated effects of the independent variables are calculated with including FE θt and τit. The effects are estimated once with the addition of the dummy FTAcomplete _{and once}

without. A* indicated significance at α=5% and **indicates significance at α=1%

(30)

The research also analysed whether the different stages of the implementation of the FTA, the signing and provisional implementation, had different effects. As only the dummy

FTA provisional_{was found to be significant, there is no evidence to state that the different stages}

had an effect on bilateral trade.

Furthermore, in this paper the GDP is not found to have a significant effect on

bilateral trade. This finding is not confirmed by the results when the gravity model is used to analyse the effects on trade, as suggested by Krugman, Obstfeld and Melitz (2012).

The NEER is only found to have a significant effect on exports, not on total bilateral trade and imports. The positive effect of the NEER on exports supports the findings of other researches, as the one of Bahmani-Oskooee (2010).

The exchange rate is found to be significant on imports and bilateral trade in the

dynamic model. The positive effect of the exchange rate on bilateral trade and imports supports the suggestions by Pilbeam (2013).

(31)

6. Conclusion and limitations

In this paper a multiple regression with fixed effects is conducted to measure the long-run effects of the implementation of the FTA between the EU and South Korea. It is found that by implementing the FTA, in the long run exports are increased with 22% and bilateral trade is increased with 25%. There was not enough evidence found that the FTA resulted in a change in imports.

The findings imply that in the four years that the FTA has been implemented, there

has been a significant increase in trade between the two parties. Thus, the research question can be answered conformingly: the FTA increased bilateral trade until now by 25% and is expected to increase the bilateral trade even more in the future. This result could imply that trade creation has taken place, as the FTA could have caused trade that would otherwise have not taken place. However, the result could also imply trade diversion, as the EU might have increased trade with EU in spite of a more efficient country it could have traded with.

It was also investigated whether trade responded to the signing of the FTA. As the

FTAprovisional

is the only variable found to be significant, it can be concluded that this stage was the only stage with an effect on bilateral trade. The signing of the FTA did not lead to an increase or decrease in trade.

To increase the significance of the variables used, the size of the data set could be increased. To do so, future research could look at a larger time period and include more European countries.

This paper also made the crucial assumption the Marshall Lerner condition was

fulfilled, however this assumption was not tested. Future research could therefore look at the type of products that were imported and exported and if any reasons are found for those products to be elastic or not. For example, if the imports consist mostly of necessities and not luxurious goods, demand is likely to be price inelastic. This could be a reason for the imports not to be affected by the FTA.

Additionally, it would be interesting to measure the sizes of trade diversion and trade

creation that took place after the FTA. In order to do so, future research could look at the actual values of trade between the EU countries and the control countries in the periods before and after the FTA.

(32)

References

Agreement between the European Community and the Republic of Korea on cooperation and mutual administrative assistance in customs matters, European Community-Republic of Korea, May 13, 1997, Official Journal of the European Union L 121, pp. 0014 – 0018. Retrieved from:

http://trade.ec.europa.eu/doclib/docs/2008/september/tradoc_140298.pdf

Antonie, M. D., Cristescu, A., & Cataniciu, N. (2010, June). A panel data analysis of the connection between employee remuneration, productivity and minimum wage in Romania. In Proceedings of the 11th WSEAS Int. Conf. MCBE 2010 (pp. 134-139).

Bahmani-Oskooee, M. (2001). Nominal and real effective exchange rates of Middle Eastern countries and their trade performance, Applied Economics, 33(1), 103-111.

doi: 10.1080/00036840122490

Bahmani-Oskooee, M. & Niroomand, F. (1998). Long-run price elasticity’s and the Marshall-Lerner condition revisited. Economics Letters, 61(1), 101-109.

Baier, S. L. & Bergstrand, J. H. (2007). Do free trade agreements actually increase members’ international trade? Journal of International Economics, 71(1), 72-95.

Balassa, B. (1961). The Theory of Economic Integration. Abingdon: Routledge

Baldwin, R., di Nino, V., Fontagé, L., de Santis, R. A., & Taglioni, D. (2008). Study on the impact of the euro on trade and foreign direct investment. (European Economic and Monetary Union Working Paper No. 321). Brussels: Office for Infrastructures and Logistics

Bank of Japan Research and Statistics Department, (August 14, 2009). “Explanation of the Effective Exchange Rate (Nominal, Real)”. Retrieved from:

(33)

Brownsell, L. (2012). Bilateral and Regional Trade Agreements, Advocates for International Development Report 2012. Retrieved from:

http://www.a4id.org/sites/default/files/user/Bilateral%20and%20Regional%20Trade %20Agreements.pdf

Cameron, A.C., Trivedi P.K., (2010). Microeconomics using Stata (Rev. Ed.). College Station: Stata Press.

Chen, M., Joshi S. (2010). Third-Country Effects of the formation of Free Trade Agreements. Journal of International Economics, 82(2), 238-48.

doi:10.1016/j.jinteco.2010.06.003

Cooper, W. H., Jurenas, R., Platzer, M. & Manyin, M. (2011). The EU-South Korea Free

Trade Agreement and its Implications for the United States (CSR Report No.

RL33539). Retrieved from Congressional Research Service: http://www.fas.org/sgp/crs/intel/RL33539.pdf

Crowley, P. M.. (2006). Is there a Logical Integration Sequence After EMU? Journal of Economic Integration, 21(1), 1–20.

The European Commission, (October 20, 2015). “European Union, Trade in goods with South Korea”. Retrieved from:

http://trade.ec.europa.eu/doclib/docs/2006/september/tradoc_113448.pdf The European Commission, (October 30, 2010). “Phase 2: the European Monetary

System”. Retrieved from:

http://ec.europa.eu/economy_finance/euro/emu/road/ems_en.htm

The European Commission. (November 9, 2015). Trade, Policy, Agreements. The official website of the European Commission. Retrieved from:

http://ec.europa.eu/trade/policy/countries-and-regions/agreements/

European Union, (2011). The EU-Korea Free Trade Agreement in practice. Retrieved from: http://trade.ec.europa.eu/doclib/docs/2011/october/tradoc_148303.pdf

Fagerberg, J.. (1988). International Competitiveness. The Economic Journal, 98(391), 355– 374.

(34)

Framework Agreement for Trade and Cooperation, European Community- Republic of Korea, March 30, 2001, Official Journal of the European Communities, pp. 48-53. Retrieved from:

http://trade.ec.europa.eu/doclib/docs/2005/february/tradoc_111835.pdf

Free Trade Agreement, European Union- Republic of Korea, September 16, 2010, Official Journal of the European Union L 127/1, Retrieved from:

http://eur-lex.europa.eu/legal-content/EN/TXT/?uri=uriserv:OJ.L_.2011.127.01.0001.01.ENG

Finalyzson, J.A. & Zacher, M.W.(1981). The GATT and the regulation of trade barrier: regime dynamics and functions. International Organization, 35, 561-602.

doi:10.1017/S002081830003424X.

Harrison, A. (2013, January). FOB/CIF Issue in Merchandise Trade/Transport of Goods in BPM6 and the 2008 SNA. Paper presented at Twenty-Fifth Meeting of the IMF Committee on Balance of Payments Statistics, Washington, D.C.

Head, K., & Mayer, T. (2013). Gravity equations: Workhorse, toolkit, and cookbook. Handbook of international economics. In G. Gopinath (Ed.) Handbook of

International Economics (Vol 4, pp. 131-195). Amsterdam: Elsevier B.V.

Hoechle, D. (2007). Robust standard errors for panel regressions with cross-sectional dependence, The Stata Journal, 7(3), 281-312.

Houthakker, H. S., & Magee, S. P.. (1969). Income and Price Elasticities in World Trade. The Review of Economics and Statistics, 51(2), 111–125.

http://doi.org/10.2307/1926720

Laffan, B., Donnell, R. & Smith, M. (2000). Europe's experimental union: Rethinking

integration (pp. 102-110). London: Routledge

Nason, G.P. (2006) Stationary and non-stationary time series. In H. Mader, S.C. Coles (Eds.) Statistics in Volcanology. London: Geological Society of London.

Pelletier, D. (2014). Time Series- Autocorrelation [Powerpoint slides]. Retrieved from: http://www.stat.ncsu.edu/people/bloomfield/courses/st430-514/slides/MandS-ch10-sec05-11.pdf

The effects of the EU : South Korea Free Trade Agreement : a panel data analysis of changes in bilateral trade