• No results found

The impact of the EU–South Korea FTA on trade with environmental goods : A gravity model approach

N/A
N/A
Protected

Academic year: 2021

Share "The impact of the EU–South Korea FTA on trade with environmental goods : A gravity model approach"

Copied!
37
0
0

Loading.... (view fulltext now)

Full text

(1)

The impact of the EU–South Korea FTA

on trade with environmental goods

– A gravity model approach

Author: Nils Norell

Spring semester 2020 Master Thesis, 30 credits Economics

¨

Orebro University School of Business Supervisor: Dan Johansson

(2)

Abstract

Quantitative research on the effect FTAs have on trade with environmental goods is a rather uncharted study field. Partly because environmental concerns have only played a limited roll in trade policy, and partly because there is no multi-laterally agreed upon definition of environmental goods. Applying the ’Combined List of Environmental Goods’ compiled by the OECD, this study empirically as-sesses whether the EU–South Korea Free Trade Agreement has led to a statistically significant increase in trade with environmental goods. Given the parties’ stated objective to explicitly promote environmental goods, the study investigates whether the agreement with its provisions has led to a statistically significant higher rate of growth in trade with environmental goods compared to other goods. The study utilises the gravity model and PPML estimations with high-dimensional fixed effect covering the years 2007 to 2018. The results show a statistically robust increase in trade with environmental goods of 13.5 percent on average. The analysis also indicates phasing-in effects of the FTA on trade with environmental goods. Seven years after the agreement was provisionally applied, the total effect is estimated to be approximately 20 percent. Further, the results indicate a higher growth rate of trade with environmental goods compared to non-environmental goods. The trade with non-environmental goods appears to be unaffected by the implementation of the FTA.

(3)

Contents

1 Introduction and Motivation 1

2 Theoretical background 5

2.1 The Gravity model . . . 5

2.2 Methodological literature review . . . 6

3 Data 10 3.1 Description of environmental goods . . . 10

3.2 Variables . . . 12

3.2.1 The dependent variable . . . 12

3.2.2 Standard gravity variables . . . 13

3.2.3 Regional Trade Agreement . . . 13

3.3 Descriptives . . . 13 4 Econometric specification 15 4.1 Model structure . . . 15 5 Results 19 6 Discussion 22 7 Conclusion 27 8 Appendix 32

(4)

1

Introduction and Motivation

Until the beginning of the century, environmental issues only played a minor roll in trade policy. Through the years, the World Trade Organization (WTO) has unsuccessfully tried to gain impact concerning environmental aspects of trade. The General Agreement on Tariffs and Trade (GATT), which defines the rules of global trade in goods, has barely dealt with concerns of natural degradation. It is only in Article XX that a single exception clause is established to safeguard the environment (Berger et al., 2017). As a consequence, with a rising number of ecological problems and an increasing awareness of those, environmental concerns within trade policy have found other paths. The number of Regional Trade Agreements (RTAs), which according to the definition of the OECD (2018), cover bilateral and plurilateral trade agreements, free trade agreements (FTAs) and economic partnerships, has grown steadily since the 1990s. Unlike their predecessors, these new types of agreements do not only aim to lower and eliminate trade restrictions such as tariffs and quotas. Instead, they address other policy areas such as environmental norms by explicitly formulating environmental provisions (George & Yamaguchi, 2018; Zurek, 2018). The inclusion of such provisions can be motivated with the potential for more environmentally friendly production and thus consumption. Additionally, they are desired to foster technological change and create new jobs (UNEP 2018).

Despite the potential for a ’greener’ progression, these RTAs with environmental provi-sions are not immune to criticism. While trade optimists point out the potential benefits mentioned above, critical voices describe environmental provisions as distractions from the controversy around the concepts of free trade and globalisation. Another angle of crit-icism, especially concerning developing countries, build upon the argumentation of ’green protectionism’. Developed countries are accused of discriminating cheaper products from their markets. (Berger et al., 2017)

Nevertheless, it seems evident that the majority of policy makers, in contrast to the past, are convinced that trade policy can be used as a tool to achieve inclusive growth which can contribute to a sustainable development (Zurek, 2018). For instance, in its trade and investment strategy report ’Trade for All’, the European Union (2015) states to continue its path towards free trade by negotiating deep and comprehensive FTAs with ambitious Trade and Sustainable Development (TSD) chapters.

(5)

The FTA negotiated between the EU and South Korea 20091 is one of the most com-prehensive agreements (Eurostat, 2018). Within the TSD chapter in this so called ’new generation’ FTA, the parties commit themselves to address the global environmental and climate related challenges. Additionally, the parties state that they ’shall strive to facilitate and promote trade and foreign direct investment in environmental goods and services, including environmental technologies, sustainable renewable energy, energy effi-cient products and services and eco-labelled goods’ (emphasis added, EU 2011, p. 63). Several years after its implementation, at the request of the Commission, Braml et al. (2018) drafted an evaluation report of the EU–South Korea FTA. The highly comprehen-sive review indicates among other things that the trade in environmental goods between the parties has risen since the agreement was provisionally in place. However, as their study is rather descriptive and only covers a limited number of environmental goods, it does not provide a full picture. Therefore, it is not possible to conclude by how much trade with environmental goods grew and whether the increase is a causal effect of the agreement and its provisions.

These aspects considered, it is of interest to investigate whether the FTA has led to an increase in trade with environmental goods, and whether the stated objective of promot-ing trade in environmental goods has caused the parties to trade those goods to a higher extent. Thus, the purpose of this study is to empirically assess whether the EU-South Korea Free Trade Agreement has led to a statistically significant increase in trade with environmental goods. Additionally, given the stated objective to explicitly promote en-vironmental goods, the study will investigate if the agreement with its provisions has led to a statistically significant higher rate of growth in trade with environmental goods compared to other goods.

Thus far, the scope of literature on empirical analysis of environmental provisions and environmental goods within trade policy is rather limited. No study has – to the best of my knowledge – assessed the impact of FTAs on the trade with environmental goods. However, OECD scholars have studied environmental aspects within RTAs. Sauvage (2014) proposes that the stringency of environmental policies impacts the demand for en-vironmental goods. When including enen-vironmental provisions, Mart´ınez-Zarzoso (2018)

1The EU-South Korea Free Trade Agreement was provisionally applied in July 2011 and formally

(6)

find a positive relationship between trade agreements and improved environmental qual-ity. Further, George and Yamaguchi (2018) conduct a qualitative study on environmental provisions and conclude that the U.S. and the EU implement their provisions to a greater extent than other economies.

Considering quantitative research, the majority of the most frequently cited econometric studies on RTAs have focused on measuring the average effect on trade flows on an aggregate level. Originating from Tinbergen’s (1962) gravity equation and by conducting state-of-the-art econometric analysis, these papers find mainly positive effects (Baier & Bergstrand, 2007; Cipollina & Salvatici, 2010; Kohl, 2014; Magee, 2008; Yotov et al., 2016). Yet, some studies have narrowed their data sets down to a more disaggregate level.

Grant and Lambert (2008) investigate a rather comparable hypothesis to the one pro-posed in this study. Similar to this work, the authors examine a subset of goods, however, focusing on agricultural goods. They pose the question whether agricultural goods are af-fected differently than other goods when joining RTAs. The authors find that the impact of RTAs on agricultural trade is higher compared to non-agricultural trade (72 vs. 27 per-cent, respectively). Further, Soete and Van Hove (2017) state that FTAs can either lead to an increase in product differentiation while in other cases they decrease established trade flows. Similarly, Jayasinghe and Sarker (2008) analyse trade creation and diversion effects of North American Free Trade Agreement (NAFTA), however, concerning agri-food products. Their findings point at positive trade effects within the agreement but displaced trade with trading partners outside of the agreement. Furthermore, on an even more disaggregate level, Zugravu-Soilita (2018) examines the causal effect trade has on environmental goods and the resulting impact on pollution. In a broader context related to the objectives of this study, Lakatos and Nilsson (2017) analyse the FTA between the EU and South Korea. In addition to positive trade effects, the authors find that this development emerges already in the negotiation phase of the FTA. Similarly the National Board of Trade (2019) identifies a positive effect of 36 percent on trade of the EU-South Korea FTA.

Considering all aspects mentioned, especially the fact that adjacent studies covering en-vironmental goods are mostly descriptive and qualitative, there is a clear need for a more comprehensive empirical investigation of environmental goods in relation to FTAs.

(7)

Fur-thermore, this study is conducted in the light of the stagnated Environmental Goods Agreement (EGA) negotiations. Thus, the result of this study can give an indication on how trade flows of certain environmental goods are affected by promotion and trade liberalisation between developed economies. However, it is important to keep in mind that this study cannot give any indication on the impact on environmental quality both during production, consumption and disposal.

The econometric model of this study is based on the theoretical findings from Tinbergen’s (1962) gravity model and Yotov et al.’s (2016) guidelines for advanced trade policy anal-ysis. The definition of environmental goods is based on the OECD’s ’Combined List of Environmental Goods’ (CLEG). The underlying data is obtained from UN Comtrade, the World Bank, the Centred’Etudes Prospectives et d’Informations Internationales (CEPII) and the WTO. The time period covered by this study is ranging from 2007 to 2018, comprising yearly observations on 236 countries. The results show a statistically ro-bust increase of 13.5 to 20 percent in trade with environmental goods. Additionally, the results indicate a higher growth rate in trade with environmental goods compared to non-environmental goods.

The rest of the paper is structured as follows: The second chapter introduces the theoret-ical foundations of the gravity model, followed by a methodologtheoret-ical literature review of trade policy analysis. The third chapter presents the definition of environmental goods as well as data sources and characteristics. The econometric specification for the model is presented in chapter four. The results of the econometric analysis can be found in chapter five. Those are discussed subsequently in chapter six. The paper closes with a conclusion and suggestions for future research.

(8)

2

Theoretical background

The number and depth of RTAs has grown steadily in the last decades (WTO 2020). Consequently, so has the literature investigating the impacts of these RTAs. Below follows an introduction of the gravity model and a methodological review on how the model has developed throughout the years.

2.1

The Gravity model

Often named the workhorse of international economics, the gravity model is by far the most common and precise econometric tool measuring trade flows. According to Yotov et al. (2016), the perks of the model are numerous. First of all, the intuition behind the equation is relatively evident. Tinbergen (1962), who introduced the equation, created an analogy to Newton’s Law of Universal Gravitation. Similar to the reference model in physics, the product of the mass of two objects, in this case economies, often represented by their GDP, is directly proportional to the gravitational force or in other words, their trade volume. The distance between the countries is related inversely proportional to the size of trade flows. In its basic form, the gravity model can be expressed as follows:

Xij = G ·

Yi· Yj

dij

.

(Cf. Tinbergen 1962) • Xij: Bilateral trade flow between trading partners i and j

• G is a constant

• Yi/j: The economic size of trading partners i and j

• dij: Bilateral geographical distance between trading partners i and j. Additionally,

trade frictions are captured in this term and, thus, the parameter should not be taken too literal as merely distance. It also includes tariffs, transport cost and trade costs

Additional factors that are proven to affect trade flows are geographical and country-specific factors such as adjacency, contiguity, access to ports, colonial and cultural ties as

(9)

well as common languages (Head & Mayer, 2014).

Secondly, the gravity model is a structural model with solid theoretical foundations. This means that the framework is suitable when conducting counterfactual analysis, e.g. mea-suring effects of trade policy. According to Head and Mayer (2014), even the first gravity estimations’ main purpose was to measure the efficiency of trade promoting policies. For example, Tinbergen (1962) himself found increases on four to five percent in bilateral trade within the Commonwealth and the Benelux customs union.

Furthermore, the model can be used for a broad range of tasks covering multiple countries and sectors. Hence, it can show that countries and sectors are interlinked and that policy changes in one place can trigger effects in others (Larch & Yotov, 2016). Lastly, the gravity model has the intrinsic property of high explanatory power. Empirical gravity studies on aggregate as well as disaggregate levels systematically yield values for the goodness of fit between 60 and 90 percent (Baier & Bergstrand, 2007; Larch & Yotov, 2016).

2.2

Methodological literature review

In spite of these properties, the gravity model was for a long time described as a ’theoret-ical orphan’ (Soete & Van Hove, 2017). However, in the beginning of the century, several scholars, unsatisfied with the theoretical foundation, addressed this issue. Originating from the Ricardian model of trade, Eaton and Kortum (2002) found that it yields results consistent with gravity models. This is not surprising given the findings of Arkolakis et al. (2012) which show that gains from trade do no vary dependent on the micro-theoretical base of the different models. In other words, independently whether the underlying model is constructed under consideration of monopolistic competition (Anderson, 1979; Ander-son & Van Wincoop, 2003), the Heckscher-Ohlin framework (Bergstrand, 1985; Deardorff, 1998) or heterogeneous firms (Chaney, 2008; Helpman et al., 2008; Melitz & Ottaviano, 2008), structural gravity will result. For a comprehensive review of the development and a more precise definition of the structural gravity equation, see Head and Mayer (2014) and Yotov et al. (2016).

Originally, OLS models were chosen to estimate the trade effects. However, due to recent development of econometric techniques, the inclusion of high dimensional fixed effects and other estimators such as Poisson Pseudo-Maximum Likelihood (PPML) are favoured

(10)

over standard OLS estimations in order to address several sources of biases. Anderson and Van Wincoop (2003) argue that earlier gravity studies do not sufficiently account for multilateral resistance and, thus, suffer from severe omitted variable bias. According to the authors, the majority of papers has neglected the fact that trading goods involves costs which causes reluctance among countries to export. Based on Krugman’s (1995) thought experiment, Anderson and Van Wincoop (2003) state that bilateral trade is not simply dependent on total cost, but rather on relative trade costs. To illustrate, two countries, say Czech Republic and Slovakia, surrounded by Italy, Germany and Poland, will most likely trade to a smaller extent with each other compared to if they were remote from the rest of the world, such as New Zealand and Australia.

With the consideration of multilateral resistance, the gravity model gained an impor-tant position within empirical trade research (Head & Mayer, 2014). Several subsequent papers (Feenstra, 2015; Redding & Venables, 2004) address this issue by introducing im-porter and exim-porter fixed effects, which have become a frequently applied computational method in gravity studies (see for example Baier and Bergstrand (2007); Magee (2008); Grant and Lambert (2008); Anderson and Yotov (2010); Soete and Van Hove (2017)). Further, on a theoretical note, the main effect of the implementation of a trade agreement is a reduction in bilateral cost of trade. Therefore, as a second order effect, a decrease in multilateral resistance follows. However, apart from the direct impacts between the parties, the formation will additionally affect excluded countries. Intuitively, this means that integrated parties, ceteris paribus, become relatively more remote from all other countries (Yotov et al., 2016).

According to Santos Silva and Tenreyro (2006), most papers have paid insufficient at-tention to zero-trade flows and heteroskedasticity. Therefore, the authors introduce the Poisson Pseudo-Maximum Likelihood (PPML) estimator as an alternative to OLS. In the majority of studies, the variables of interest are found in logarithmic form. As a consequence, the results can be interpreted as elasticises. In their seminal paper, San-tos Silva and Tenreyro (2006) criticise this practice due to the fact that zero values are not considered as they cannot be logarithmised. Disregarding this mathematical feature will cause the estimator to drop the zero-observations resulting in a systematic selection bias (Head & Mayer, 2014). According to Santos Silva and Tenreyro (2006), there are three possible explanations for zeros occurring in trade related data sets. First of all, it is rather plausible that some trading partner simply do not trade with each other during

(11)

some periods of time. As Santos Silva and Tenreyro (2006, p. 642) put it: ’it would not be surprising to find that Tajikistan and Togo did not trade in a certain year.’ However, the absence of trade is not the only possible explanation for zeros in the data set. Due to a lack of liability of the reported data, zeros can also represent missing values. In most cases, it is impossible to distinguish between ’true’ zero-trade flows and missing values. Lastly, a comparably minor problem the authors mentioned is that zeros might occur due to rounding errors of small values.

Yet, another problem concerning the logarithm of the independent variables, addressed by Santos Silva and Tenreyro (2006) is Jensen’s inequality. The formal proof shows that the expected value of the logarithm is unequal to the logarithm of the expected value. Based on this inequality, the authors state that the interpretation of elasticities in log-linear OLS models in the presence of heteroskedasticity can be severely biased. As the expected value of the logarithm is influenced by higher-order moments, the transformed errors correlate with the covariates when error terms are heteroskedastic. Thus, the derived estimator is not reliable. As trade data typically is heteroskedastic, the coefficients of OLS models ought to be interpreted with caution (Santos Silva & Tenreyro, 2006). On the other hand, although yielding robust results, Head and Mayer (2014) argue that rather than choosing Poisson Pseudo-Maximum Likelihood (PPML) as the single gravity estimator, it should be used as a complement to OLS and alternative estimators in a robustness process. When using the gravity equation it is not to be excluded that the error term and the indi-cator for a FTA can be correlated. Scholars are unanimous that the resulting endogenity causes incorrect results, however, their arguments are diverging regarding the direction of the bias (Baier & Bergstrand, 2007; Head & Mayer, 2014; Magee, 2008; Roy, 2010; Soete & Van Hove, 2017). Baier and Bergstrand (2007) argue that endogenity is causing a standard OLS estimator to underestimate the true effect of a FTA. The authors argue that countries primarily enter FTAs because they are different from each other. Further, they state that there are non-observable trade restrictions such as domestic regulations and institutional settings which prohibit trade. As a result, the dominating part in the error term becomes negative and, thus, correlates negatively with FTA. Magee (2008) on the other hand, presents a contrary argumentation suggesting that a standard OLS model overestimates the true effect of a FTA. The author argues that the error term is positively correlated with the implementation of a FTA as countries have inherent characteristics such as historical and colonial ties. These features are dominating explanatory factors as

(12)

to why countries trade above average with each other. Thus, they face larger incentives to enter a FTA. In accordance with Magee (2008), Roy (2010) concludes, what he calls, ’positive selection’ and argues that countries negotiate FTAs because they are similar. Another desirable estimation approach with possibly high explanatory power is the in-clusion of intra-trade flows, defined as gross production minus exports (Yotov, 2012). As consumers in an increasingly integrated world choose between consumption of foreign and domestic goods, including national trade flows can identify border effects and capture the trade diversion effect a FTA may cause. In other words, considering international trade flows exclusively might bias the estimator downwards and underestimate the true trade effect (Dai et al., 2014).

(13)

3

Data

The data are given in a panel structure, covering the years 2007 to 2018 and include 236 countries. To be able to estimate and distinguish the possible effect a FTA may have on the trade with environmental goods, there is a need to define what is meant by environmental goods. After defining those within the context of this study, all other data sources are discussed and a descriptive analysis is presented.

3.1

Description of environmental goods

To this point, there is no universally accepted definition of which goods can be listed as environmental goods. On a multilateral level, both the Doha Ministerial Declaration (launched in 2001) and the Environmental Goods Agreement (EGA) (launched in 2014) have failed to reach a consensus. The international community’s inability to define en-vironmental goods is therefore up to this day an unsolved problem. The lists of goods resulting from negotiations of the above mentioned parties, as well as the one proposed by the OECD, are provisionally used in different contexts.

An objection concerning the existing lists is that the goods included are too favourable for developed countries while neglecting the perspective of developing countries

(Zugravu-Soilita, 2018). Balineau and de Melo (2013) examine the objectives of the involved

parties by calculating an index over ’revealed comparative advantage’. The study points at a propensity towards mercantilistic behaviour as countries tend to promote goods for which they hold comparative advantages and exclude goods for which they face high tariffs. Additionally, there is some ambiguity in the characterisation of environmental goods. Steenblik (2005) describes several classification issues regarding the Harmonised System (HS) on the 6-digit level as it is insufficiently specific and argues that goods with no environmental use risk to swamp the volume of trade in the specific HS subheadings. Further, the possible multiple use (’dual use’) of certain products, which does not neces-sarily need to be environmentally friendly, is not accounted for. The author also stresses the concern that goods are defined by their relative environmental performance, which becomes complicated as technology evolves. For a more detailed exposition of the main challenges of defining environmental goods, see Steenblik (2005).

(14)

However, several intergovernmental organisations such as the OECD and Eurostat have outlined some guidelines for analytical purposes. According to the OECD’s ’Combined List of Environmental Goods’ (CLEG), products are considered as ’environmentally-related goods’ if they comprise ’activities which produce goods and services to measure, prevent, limit, minimise or correct environmental damage to water, air and soil, as well as problems related to waste, noise and eco-systems. This includes cleaner technologies, products and services that reduce environmental risk and minimise pollution and resource use’ (OECD/Eurostat, 1999, p. 9). Overall, the CLEG contains 248 products and is divided into eleven subcategories:

• Air pollution control

• Cleaner or more resource efficient technologies and products • Heat and energy management

• Environmental monitoring, analysis and assessment equipment • Natural resources protection

• Noise and vibration abatement • Renewable energy plant

• Management of solid and hazardous waste and recycling systems • Clean up or remediation of soil and water

• Waste water management and potable water treatment

• Environmentally preferable products (EEP) based on end use or disposal

Beside the CLEG, the United Nations Environment Programme (2018) has compiled a list of environmental goods. It contains in total less goods than the OECD’s list. Out of the 147 and 248 goods, respectively, only approximately 39 percent overlap. This may be a possible explanation why it has not been part of any multilateral or bilateral negotiations. Additionally, the UNEP’s list contains a larger share of environmentally preferable products which implies that the list is more suitable for developing countries as their share in the international trade concerning this group of goods is comparably large. It even holds potential for developing countries to become net exporters (Steenblik, 2005; Zugravu-Soilita, 2018). Taking this into consideration and keeping in mind the stated

(15)

objective of this study — estimating the trade with environmental goods between two developed economies — this study chooses the compiled list of the OECD for defining environmental goods.

3.2

Variables

Yearly data is obtained from different sources and can be categorised into dependent trade flow variables, standard gravity variables and Regional/Free Trade Agreement variables.

3.2.1 The dependent variable

In order to address the research questions, the dependent variable is obtained differenti-ated between environmental goods and non-environmental goods. Thus, it is possible to compare the two estimates.

The data for trade flows of environmental goods are available at UN Comtrade and were accessed via World Integrated Trade Solution (WITS) coded in the 2007 Harmonized Sys-tem (HS)2version. This version is chosen since Sauvage (2014) presents the CLEG in this version. However, the system is updated every fifth year covering modifications, merges and divisions of existing codes and introductions of new codes. This means, despite data being available for more years (e.g. HS 2002/1996/1992), it requires transformations which go beyond the scope of this study. Hence, the resulting time range of this study is 2007 to 2018. For further readings on HS-conversion, see UNSD (2017).

The data for non-environmental goods (i.e. all other goods making up the global trade excluding the 248 products on the CLEG) are also obtained in the HS 2007 version and cover all trade among 236 countries from 2007 to 2018. After extracting all individual goods denoted in HS 2007 codes, the data is summed up into one single data point per year per country pair. This is applied to both environmental and non-environmental goods yielding two observations per year per country pair.3

In line with Yotov et al. (2016), the study draws upon so called ’mirror data’. This entails analysing reported import data of goods which is considered more reliable as countries are

2Harmonized Commodity Description and Coding System is a system used by more than 200 countries

in the world for customs purposes and for international trade statistics (WCO 2020).

3Note: As some countries do not trade environmental goods among each other, there might only be

(16)

expected to report goods flowing into the country more accurately due to their tariff and tax base (Bacchetta et al., 2012). Moreover, following the suggested practice of Shepherd (2013) and Yotov et al. (2016), all trade flow data are obtained in nominal prices.

3.2.2 Standard gravity variables

The study includes the standard gravity model variables referred to in section 2.1. The data for GDP are extracted from The World Bank (2020) and given in USD in nom-inal prices. To capture bilateral trade costs, a distance variable, defining the distance between each country pair in kilometers, is obtained from the Centred’Etudes Prospec-tives et d’Informations Internationales (CEPII). The model additionally includes Mayer and Zignago’s (2011) standard gravity variables, specifying whether a country pair has a common border, speaks the same language, or shares a colonial link. For a more com-prehensive description of each standard gravity variable, see Mayer and Zignago (2011).

3.2.3 Regional Trade Agreement

The variable for the FTA indicates whether both importer and exporter country are either a member of the European Union or South Korea, respectively. As the FTA was provisionally applied in July 2011, the study chooses this year as the starting point for the agreement. To control for other trade effects outside of the trade agreement between the EU and South Korea, an additional variable is included. The dummy indicates whether any kind of RTA exists between a country pair. For this purpose a data set containing information on all existing RTAs from the WTO RTA database is utilised. Further, a dummy variable indicating the membership in the European Union is introduced.

3.3

Descriptives

When merging the trade data with the two data sets comprising GDP and trade costs and comprising information on standard gravity variables, a certain amount of country pairs is lost. This is a trade off between the maximal amount of observations and the additional information provided by other sources. Additionally, it cannot be excluded that some observations were lost due to coding errors in the ISO3 country codes. Most of the observations lost entail relatively small countries which do not account for a large share in global trade, and, thus, should not influence the result significantly. Given the

(17)

panel structure of the data, the standard summary statistics provide a comparably low degree of information. They can be found in the Appendix (see Appendix, Table A1).

Figure 1: Share of trade in environmental goods

2007 2009 2011 2013 2015 2017 0 5 10 15 20 25 30 P ercen ta ge share (%) Global trade EU-South Korea Source: WITS

The above graph depicts the development of the share environmental goods account for in global trade. The graphs differentiate between all global trade (dashed) and trade between the European Union and South Korea. While environmental goods make up a larger share of the trade for the EU and South Korea, the share is lower looking at a global level. An increase in trade can be observed from 2010 and this positive trend is continuous over the year when the agreement was provisionally in place (2011). It reaches a peak in 2014 and later decline until 2017. However, the shares of trade with environmental goods are consistently higher after 2011 and remain larger compared to global trade. Econometric methods are needed to confirm the descriptive trend and control for other drivers.

(18)

4

Econometric specification

The empirical model specification is based upon the guidelines proposed by Yotov et al. (2016), which represent the state-of-the-art econometric technique for trade policy analysis. This includes high-dimensional fixed effects addressing the previously mentioned methodological issues, such as multilateral resistance (Anderson & Van Wincoop, 2003) and endogeneity (Baier & Bergstrand, 2007). Additionally, in accordance with Yotov et al. (2016), Baier and Bergstrand (2007) and National Board of Trade (2019), the model specification include phasing-in effects in order to address that the FTA might have a non-linear effect and/or an effect changing over time. Due to a limited access to the necessary data, this study will, unlike Yotov et al.’s (2016) and Braml et al.’s (2016) recommendation, not include intra-trade flows.4

In order to address the previously mentioned issues concerning zero trade flows and heteroskedasticity, the empirical models will take into account Santos Silva and Tenreyro (2006)’s proposition to use the Poisson Pseudo Maximum Likelihood estimator (PPML). However, following Anderson and Yotov (2010), Head and Mayer (2014), Yotov et al. (2016) recommendation, OLS estimates will also be obtained in a robustness process.

4.1

Model structure

The first model for analysing the research question is a standard OLS estimation con-sidering the aforementioned traditional gravity covariates. To obtain the effect of the EU–South Korea FTA on trade volumes, a dummy variable is introduced. Further, to be able to capture the FTA’s effect on the trade with environmental goods explicitly, an interaction term is included. These two coefficients are then compared in order to conclude whether trade with environmental goods has grown at a higher rate than trade with all other goods.

In addition, the model comprises two dummy variables considering a EU membership and the existence of any RTA between a country pair. Controlling for the impact the EU as an economic union has, enables the main FTA variable to only capture the trade effect between EU countries and South Korea. If this had not been controlled for, the

4In addition to the general data availability issues, Yotov et al. (2016) point out several difficulties

when determining an accurate representation of intra-trade flows. As it in most cases requires extrapo-lation and linear interpoextrapo-lation, it depends heavily on strong assumptions.

(19)

effect would include the trade between EU countries and, thus, cause a bias. The second control variable is included in order to account for trade effects of all other RTAs in place during the years covered.

lnXij,g,t= β0+ β1lnGDPi,t+ β2lnGDPj,t+ β3lnDISTij + β4CN T Gij+

β5LAN Gij + β6CLN Yij + β7EUij,t+ β8RT Aij,t+ β9F T Aij,t+

β10Envirg+ β11 F T Aij,t#Envirg + εij,t

(1)

lnXij,g,t in equation (1) denotes the natural logarithm of the traded value between

ex-porter i and imex-porter j for a given type of good g at time t. β0 is a constant. lnGDPi,t

and lnGDPj,t are the natural logarithm of the nominal GDP for exporter i and for

im-porter j at time t, respectively. lnDISTij refers to the natural logarithm of bilateral

distance in kilometers between exporter i and importer j. CN T Gij, LAN Gij, CLN Yij

are dummy variables denoting whether the exporter and the importer share a common border, speak the same language or whether they ever had a colonial tie. EU is a dummy variable indicating whether the exporter and the importer are both members of the Eu-ropean Union at time t. RT Aij,tis a dummy variable denoting whether a country pair is

part of the same Regional Trade Agreement at time t. F T Aij,t is equal to one for trade

between the EU countries and Korea for the years after 2010 and forward, and equal to zero otherwise. Envirg is an indicator variable denoting whether environmental goods

are traded. It varies only over the type of good but not over country-pairs or time. As the aggregated trade volume of environmental goods is considerably smaller compared to the aggregate of all other goods (see 3.3 Descriptives), it is important to account for this. F T Aij,t#Envirg, the main variable of interest, is an interaction term which indicates

whether a country pair is part of the EU–South Korea FTA and whether trade with environmental goods is conducted. Lastly, εij,t refers to an error term presumed to be

(20)

In equation (2), a PPML estimator is introduced and, thus, the dependent variable is no longer in logarithmic form.

Xij,g,t= exp

h

β1lnGDPi,t + β2lnGDPj,t+ β3lnDISTij + β4CN T Gij+

β5LAN Gij + β6CLN Yij + β7EUij,t+ β8RT Aij,t+ β9F T Aij,t+

β10Envirg+ β11 F T Aij,t#Envirg

i × εij,t

(2)

In equation (3), time-varying exporter and importer fixed effects, πi,tand χj,t, are included

in order to control for multilateral resistance (MTR) and other unobservable as well as observable factors changing over time concerning exporters and importers. By including exporter and importer-time fixed effects, the estimator absorbs time-variant factors and therefore estimates for GDPi and GDPj can be excluded. For further reading of this

feature, see Yotov et al. (2016). When applying fixed effect, no constant is included in the equation.

Xij,g,t = exp

h

πi,t+ χj,t+ β1lnDISTij + β2CN T Gij + β3LAN Gij+

β4CLN Yij + β5EUij,t+ β6RT Aij,t+ β7F T Aij,t+ β8Envirg+

β9 F T Aij,t#Envirg

i × εij,t

(3)

To account for the aforementioned issue concerning endogeneity (ENDG), equation (4) includes country pair fixed effects, denoted by µij. This enables the model to account for

all time-invariant bilateral trade costs which might differ between countries. Similar to exporter and importer-time fixed effects, country pair fixed effects absorb information. In this case the time-invariant variables such as distance, common language, contiguity and colonial ties (Yotov et al., 2016).

Xij,g,t= exp

h

πi,t+ χj,t+ µij + β1EUij,t+ β2RT Aij,t+ β3F T Aij,t+ β4Envirg+

β5 F T Aij,t#Envirg

i × εij,t

(21)

In equation (5), two lagged variables, t + 3 and t + 7, are introduced in addition to the variables described in model (4) to allow for phasing-in effects (PHSNG). The years are chosen in order to ensure reasonable intervals and amount of lagged dummies.

Xij,g,t= exp

h

πi,t+ χj,t+ µij + β1EUij,t+ β2RT Aij,t+ β3F T Aij,t+

β4F T Aij,t+3+ β5F T Aij,t+7+ β6Envirg+ β7 F T Aij,t#Envirg+

β8 F T Aij,t+3#Envirg + β9 F T Aij,t+7#Envirg

i × εij,t

(5)

Finally, in accordance with Santos Silva and Tenreyro (2006), a heteroskedasticity-robust RESET test (Ramsey, 1969) is conducted. The test is able to assess whether the general specification in a linear model is appropriate. The underlying null hypothesis states that the models are not determined by a misspecification of error terms. The specific functional misspecification cannot be derived from a positive result, however, it can be concluded that the rejected null hypothesis points at an appropriate model specification.

(22)

5

Results

Table 1: Estimating the trade effects of the EU–South Korea FTA

OLS (1) PPML (2) MTR (3) ENDG (4) PHSNG (5) Log GDPi 1.310∗∗∗ 0.823∗∗∗ (0.006) (0.023) Log GDPj 0.887∗∗∗ 0.802∗∗∗ (0.006) (0.026) Log Distance −1.137∗∗∗ −0.549∗∗∗ −0.619∗∗∗ (0.020) (0.047) (0.032) Contiguity 0.857∗∗∗ 0.539∗∗∗ 0.393∗∗∗ (0.102) (0.155) (0.087) Language 0.946∗∗∗ 0.194 0.116 (0.041) (0.122) (0.080) Colony 0.758∗∗∗ −0.065 0.281∗∗∗ (0.758) (0.065) (0.096) EU 0.768∗∗∗ −0.049 0.572∗∗∗ 0.516∗∗∗ 0.514∗∗∗ (0.061) (0.100) (0.097) (0.141) (0.139) RT A 0.634∗∗∗ 0.131∗∗∗ 0.125∗∗∗ −0.006 −0.003 (0.028) (0.042) (0.023) (0.004) (0.004) F T A 0.158∗∗∗ −0.001 −0.007 −0.020 −0.022 (0.049) (0.018) (0.061) (0.027) (0.023) F T A(t+3) 0.003 (0.018) F T A(t+7) 0.001 (0.015) Envir −3.429∗∗∗ −2.507∗∗∗ −2.515∗∗∗ −2.522∗∗∗ −2.522∗∗∗ (0.018) (0.038) (0.038) (0.038) (0.035) F T A#Envir 0.667∗∗∗ 0.103∗ 0.111∗∗ 0.126∗∗ 0.096∗ (0.038) (0.057) (0.057) (0.056) (0.075) F T A#Envir(t+3) 0.040∗∗∗ (0.014) F T A#Envir(t+7) 0.045∗∗∗ (0.015) Observations 373,744 373,744 397,150 407,067 407,067 R2/P seudo R2 0.635 0.550 0.861 0.987 0.987 T otal F T A#Envir 0.181

RESET test (p-value) 0.0000 0.0004 0.0014 0.7197 0.7003

Note: All estimates are obtained with data from 2007–2018. Column (1) applies a standard OLS estimator, (2) a PPML estimator, (3) a PPML estimator with importer and exporter time fixed effect, (4) adds country-pair fixed effects, and (5) introduces lagged variables in order to allow for phase-in effects of the FTA. Standard errors are clustered by country pair and are reported in parentheses. Constants are placed in Table A2. The p-values read as follows: ∗p < 0.1;∗∗p < 0.05;∗∗∗p < 0.01.

(23)

Table 1 displays the outcomes of the models aiming to answer the question whether the FTA has led to an increase in trade with environmental goods. The increase in trade with environmental goods within the EU–South Korea FTA is given by an interaction term between the FTA and the environmental goods indicator variable. Considering the interaction structure of the models, the main effect of the FTA dummy variable has to be interpreted given that the good traded is a non-environmental good. Different outcomes are obtained throughout the model specifications. While the standard OLS model indicates a rather moderate positive effect of [exp(0.158)-1] x 100 = 17.12 percent, which is statistically significant at the one percent level, all other estimates are small and insignificant. The interaction term F T A#Envir yields consistently positive coefficients throughout model (1-4), which are [exp(0.667)-1] x 100 = 94.84 percent, [exp(0.103)-1] x 100 = 10.85 percent, [exp(0.111)-1] x 100 = 11.74 percent and [exp(0.126)-1] x 100 = 13.43 percent. Model (5), which estimates the cumulative effect of FTA over time, will be discussed further down.

To obtain the effect of the FTA given that an environmental good is traded, main and interaction effect are summed up. Thus, for model (1) the increase in trade with

envi-ronmental goods within the EU–South Korea FTA is 128.19 percent.5 However, in the

remaining models, the F T A yields economically and statistically insignificant estimates which are therefore treated as zero. The resulting average effect sizes for the increase in trade with environmental goods are 10.85, 11.74 and 13.43 percent.

The control variable for EU membership points at a highly positive effect on trade volumes which is significant at the one percent level. However, this result cannot be observed in the PPML model in column (2) which indicates an economically and statistically insignificant negative effect. When controlling for existing RTAs, the coefficients obtained from OLS (1), PPML (2) and PPML with importer and exporter time fixed effects (3) indicate a at the one percent level significant, positive relation, all else equal. Only the specification with higher level fixed effects (4) suggests a negative trade effect which is, however, not significant.

In column (5), two lags for the effect of the FTA are introduced in order to capture phasing-in effects. The estimate for F T A yields, similar to the previous PPML models (2-4), economically and statistically insignificant estimates. The three coefficients for

(24)

F T A#Envir, F T A#Envir(t+3) and F T A#Envir(t+7) suggest statistically significant

estimates ranging from ten to one percent. This implies a cumulative overall statistically significant effect of the FTA on environmental goods of exp(0.181)-1] x 100 = 19.84 percent.

The standard gravity coefficients behave in accordance to the theory described earlier and are highly significant at the one percent level in 13 out of 16 cases. Additionally, the indicator variable for environmental goods, Envir, is, as expected, largely negative and significant at the one percent level throughout all specifications. According to the estimation, the global trade volumes is 342.9 (1) to 252.2 (4) percent smaller when con-sidering environmental goods compared to all other goods. In the bottom of Table 1, the p-values of the Ramsey RESET test can be found. The null hypothesis is rejected in all models besides in model (4) and (5).

(25)

6

Discussion

Both hypotheses concerning the trade with environmental goods between the EU and South Korea can be confirmed over all model specifications. The outcomes show that the initiation of the FTA goes along with an increase in trade with environmental goods. The results of the PPML models are relatively consistent ranging between 11 to 20 percent. The OLS estimate constitutes an outlier suggesting a 128 percent increase.

Considering the coefficients of the standard OLS model (1), the intuitive interpretation suggests economically and statistically significant estimates for the trade in environmental goods. Additionally, all coefficients of the control variables are statistically significant at the one percent level. The standard gravity covariates in the OLS specification are all in line with the meta-analysis presented in Head and Mayer (2014). For example, the GDP effect of exporter country is stronger than the GDP effect for the importer country (1.3 vs. 0.89 percent). The distance coefficient indicates a strongly negative, elastic effect of 1.14 percent given a one percent increase in distance, which is comparably larger than the inelastic 0.89 percent in the meta-analysis. The three dummy variables for contiguity, language and colonial links also yield stronger effects than Head and Mayer (2014). However, as the standard OLS estimation does not take into account the panel structure of the data set, it evidently suffers from a severe bias. The standard OLS estimation treats all observations as homogeneous, which is a rather strong assumption due to the coverage of both different time periods and countries. It is intuitive that observations within the same country or within the same year are correlated to each other. Given that the OLS model does not allow for autocorrelation, the error term is most likely wrongly specified. Therefore, it is rather improbable that the coefficients yield the true average effect for trade with environmental goods after implementing the FTA. Due to the econometric flaws of the standard OLS estimation, three different PPML models were specified.

When introducing the PPML estimator in model (2), most coefficients appear to be smaller or become insignificant. The coefficient for the FTA, given a non-environmental good is traded, is nearly zero and insignificant. The estimator for trade with environ-mental goods is only significant at the ten percent level. The decrease of effect sizes is in accordance with other studies. For example, the National Board of Trade (2019) finds that average effects of indicator variables such as RTA, EU and FTA (and the

(26)

in-teraction) on trade are lower when choosing a PPML over a standard OLS estimation.6 Further, in line with Santos Silva and Tenreyro (2006), the standard gravity covariates decrease substantially when choosing PPML over OLS. Interestingly, the distance vari-able yields results that are equivalent to what Head and Mayer (2014) points out. It is therefore likely, as the authors state, that the effect of distance is underestimated when using PPML estimations.

In model (3), when importer and exporter time fixed effect are included in order to control for multilateral resistance and other unobservable as well as observable factors changing over time, there is no considerable change concerning the effects of interest. However, the control variable for EU membership seems to be largely affected by the inclusion of the aforementioned fixed effect. Thus, it is plausible that the model succeeds to account for factors that influence trade with goods between EU countries.

Model (4) utilises higher level fixed effects to account for all time-invariant bilateral trade costs and to deal with endogeneity. While the FTA coefficient remains insignificant, the interaction term is slightly bigger compared to model (3). Examining the direction and size of the F T A coefficient, it seems like the previous models (1-3) marginally under-estimate the FTA effect on trade. Conversely, the direction and the effect of all other RTAs points at the opposite, stating that the previous models slightly overestimates the trade effect. However, as neither of these coefficients are economically and statistically significant in model (4), it is not possible to draw any meaningful conclusion concerning endogenity and the direction of the bias.

Considering the other control variables, both similarities and differences compared to other studies occur. The trade effect of EU membership in model (4) of [exp(0.516)-1] x 100 = 67.53 percent is larger but yet in line with previous studies. When specifying a nearly identical model (PPML estimation without intra-trade flows), the National Board of Trade (2019) suggests a 40 percent trade effect. The coefficients for RTA are rather small and in some cases economically and statistically insignificant. This is not in line with the majority of previous gravity studies which often suggest average trade effect of 50 to 120 percent (Baier & Bergstrand, 2007; Kohl, 2014; Soete & Van Hove, 2017; Yotov et al., 2016).

6However, when including intra-trade flows, the National Board of Trade (2019) finds larger trade

(27)

In model (5), phase-in effects of the FTA are introduced to account for a non-linear development of the impact of the FTA. The economic interpretation of the coefficients suggests a strong trade effect on environmental goods of ten percent in the first three years of the agreement. The increase flattens later to roughly four to five percent between 2014 and 2018. Overall, after seven years of its provisional implementation, the FTA has still a significantly positive effect on trade with environmental goods of 4.5 percent. Con-sidering the immediate as well as the phasing-in effects, the FTA has led to a statistically significant increase in trade with environmental goods by approximately 20 percent. The phasing-in effect is somewhat in line with previous studies. Baier and Bergstrand (2007), Yotov et al. (2016) and the National Board of Trade (2019) find substantial trade effects ten to twelve years after implementations of a FTA/RTA. However, these studies support the finding only in general using aggregated data and not explicitly testing for specific FTAs or certain groups of goods like in this study.

In addition to the previous discussion, the p-values of the Ramsey RESET test give a good indication on the validity of the models. The null hypothesis, stating that there is no sign of functional misspecification, cannot be rejected when the p-value is larger than the significance value. In the standard OLS model (1), the null hypothesis clearly can be rejected which indicates that the logarithmic specification is, in addition to the aforementioned limitations, not appropriate. Similarly, the PPML models (2-3) without the most sophisticated fixed effects are not able to pass the RESET test. However, in the case of model (4) and (5), when importer and exporter-time fixed effects and country-par fixed effect are included, the test yields no indication of any functional misspecification. Thus, the most sophisticated models, (4) and (5), are considered trustworthy even though the estimates are relatively comparable to the other PPML models.

Concerning the second part of the research question, growth rates can be evaluated when comparing the trade effect of the FTA given that a non-environmental good is traded vs. given an environmental good is traded. This is done by comparing the coefficient for the FTA with the sum of the coefficient for interaction term. The latter not only suggests a significant increase in trade in environmental goods, it also points at a higher rate of growth. While the model indicates no growth in trade with non-environmental goods, average effects for environmental goods range between 11 and 20 percent. As the coefficients for the FTA in the PPML models (2-5) are insignificant, they are treated as zeros. This seems plausible, as the coefficients as well as the standard errors are small.

(28)

It can therefore be assumed that the FTA’s effect on non-environmental goods is either zero or extremely small and imprecisely measured.

This outcome is only partly in line with earlier research for the EU–South Korea FTA. Given that the largest share of goods traded comprises non-environmental goods, the difference between aggregated analysis should not be too large. For example, Braml et al. (2018), the National Board of Trade (2019) and Lakatos and Nilsson (2017) find significantly positive trade effects within the EU–South Korea FTA. However, it is likely that different data sets and model specifications impact the estimates. Both Braml et al. (2018) and the National Board of Trade (2019) include intra-trade flows which evidently tend to double or even triple FTA estimates. Furthermore, a limitation of this study is the fact that it only covers twelve years which possibly limits the between variation and, thus, the explanatory power of the data.

However, a possible explanation for the different significance levels of the growth rates for non-environmental and environmental goods concerns the heterogeneity within the goods traded. When comparing two different groups, the most desirable case would be a high degree of between variance and a low degree of within variance. If there is too much variation within the groups, the results can become insignificant. In the case of environmental goods, quite homogeneous features can be assumed as the products belong to the same class and partly serve a similar purpose. Non-environmental goods on the other hand comprise products from different sectors, from automobiles to textiles. The degree of heterogeneity in trade within this group can thus be assumed to be rather high. This can explain the significant impact of the FTA on environmental goods and the insignificant result concerning non-environmental goods. However, given that both coefficients and standard errors are comparably small, the results rather indicate a non-existent impact on trade volumes.

Another potential cause for differences concerning the rates of growth of trade between different groups of goods concerns tariffs. As environmental goods are industrial goods, it is plausible that they benefit from faster tariff reductions compared to non-industrial goods. As the latter are included as an aggregate, no differentiation was possible. Further, to match the tariff easing to the respective product groups, an extensive study of the technicalities of the FTA and the HS codes is necessary. This aspect was beyond the scope of this study but opens an interesting angle for future research.

(29)

Comparing this study’s main findings to Grant and Lambert (2008), who stated a compa-rable research question addressing agricultural goods, a similar conclusion is drawn. Both this study and Grant and Lambert (2008) suggest that a successful removal of trade bar-riers through trade liberalisation can stimulate exports and imports of the types of goods emphasised. Additionally, their study support the finding that growth rates between types of goods can differ significantly.

A rather obvious improvement of this study would be an inclusion of the earlier mentioned intra-trade flows in order to capture possible trade creation effects. Further, as the definition of environmental goods is imprecise, other lists could be included in order to test the robustness of this study’s findings. Additionally, a conversion of HS codes can enable longer times series which are likely to improve the explanatory power. Additionally, future studies should compile more information on similar environmental provisions in other RTAs in order to compare and assess their impact. A limitation of this study is that it does not contain any information on environmental provisions in other RTAs. Therefore, it can be concluded that the EU–Korea FTA has had a larger impact on environmental goods than other goods. However, it cannot be obtained with certainty that this effect is due to the TSD chapter and the environmental provisions. It is likely that the growth in trade with environmental goods is affected by other factors, such as a higher demand for green products, altered national and international policies and new technologies.

(30)

7

Conclusion

One of the European Union’s main objectives in order to contribute to a sustainable development is to negotiate deep and comprehensive FTAs with ambitious Trade and Sustainable Development (TSD) chapters. In this study the EU–South Korea FTA is empirical analysed in order to assess whether the stated objectives to promote and fa-cilitate trade in environmental goods is achieved. The results show a statistically robust increase in trade with environmental goods of 13.5 percent on average. The analysis also indicates phasing-in effects of the FTA for trade with environmental goods. Seven years after the agreement was provisionally in place, the total effect amounts to approximately 20 percent. Further, the results indicate a higher rate of growth in trade with environ-mental goods compared to non-environenviron-mental goods. The trade with non-environenviron-mental goods appears to be unaffected by the implementation of the FTA.

More research on trade effects of RTAs at a disaggregate level is necessary. This study gives an indication that agreements can have different effects depending on the type of good. A possible explanation of diverging estimates can be certain emphases in the agreements. Researchers should not only deal with RTAs on an econometric level striving to develop more sophisticated models but also engage with the agreements in terms of content. Countries and economic unions can put a special focus on certain types of goods and services. This can not only encourage the trade with those but can also lead to a distortion in trade with other goods. The aggregate trade might contain too much variance to provide significant results.

Finally, whether the increase in trade with environmental goods should be seen as a success, opens up potential for deeper research. In order to draw meaningful conclusions on ecological impacts, many other factors must be considered. Questions of production, consumption and disposal of goods need to be raised as well as the carbon footprint of global trade. A mere quantitative study cannot assess the impact on sustainability but is a first step. The increase in trade with so called environmental goods should motivate researchers and public authorities to further examine their impacts.

(31)

References

Anderson, J. E. (1979). A theoretical foundation for the gravity equation. The American economic review, 69 (1), 106–116.

Anderson, J. E., & Van Wincoop, E. (2003). Gravity with gravitas: A solution to the border puzzle. American economic review, 93 (1), 170–192.

Anderson, J. E., & Yotov, Y. V. (2010). Specialization: Pro-and anti-globalizing, 1990-2002 (tech. rep.). National Bureau of Economic Research.

Arkolakis, C., Costinot, A., & Rodrıguez-Clare, A. (2012). New trade models, same old gains? American Economic Review, 102 (1), 94–130.

Bacchetta, M., Beverelli, C., Cadot, O., Fugazza, M., Grether, J.-M., Helble, M., Nicita, A., & Piermartini, R. (2012). A practical guide to trade policy analysis. Geneva, CH: World Trade Organization.

Baier, S. L., & Bergstrand, J. H. (2007). Do free trade agreements actually increase members’ international trade? Journal of international Economics, 71 (1), 72–95. Balineau, G., & de Melo, J. (2013). Removing barriers to trade on environmental goods:

An appraisal. World Trade Review, 12 (4), 693–718.

Berger, A., Brandi, C., & Bruhn, D. (2017). Environmental provisions in trade agree-ments: Promises at the trade and environment interface (tech. rep.). Briefing Pa-per.

Bergstrand, J. H. (1985). The gravity equation in international trade: Some microeco-nomic foundations and empirical evidence. The review of ecomicroeco-nomics and statistics, 474–481.

Braml, M., Gabriel, F., Larch, M., Woon Nam, C., Steininger, M., & Yotov, Y. V. (2018). Evaluation of the Implementation of the Free Trade Agreement between the EU and its member states and the Republic of Korea. Final report: Main Report (tech. rep.). Civic Consulting and the Ifo Institute.

Chaney, T. (2008). Distorted gravity: The intensive and extensive margins of international trade. American Economic Review, 98 (4), 1707–21.

Cipollina, M., & Salvatici, L. (2010). Reciprocal trade agreements in gravity models: A meta-analysis. Review of International Economics, 18 (1), 63–80.

Dai, M., Yotov, Y. V., & Zylkin, T. (2014). On the trade-diversion effects of free trade agreements. Economics Letters, 122 (2), 321–325.

(32)

Deardorff, A. (1998). Determinants of bilateral trade: Does gravity work in a neoclassical world?, In The regionalization of the world economy. University of Chicago Press. Eaton, J., & Kortum, S. (2002). Technology, geography, and trade. Econometrica, 70 (5),

1741–1779.

European Union. (2011). Official Journal of the European Union, l 127, 14 may 2011 [Article 13.6 § 2]. https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri= OJ:L:2011:127:FULL&from=EN

European Union. (2015). Trade for all: Towards a more responsible trade and investment policy (tech. rep.). European Commission. Publications Office of the European Union, Luxembourg.

Eurostat. (2018). Archive: South Korea-EU - trade in goods - data 2015 [2020-06-02]. https://ec.europa.eu/eurostat/statistics- explained/index.php/Archive:South Korea-EU - trade in goods - data 2015

Feenstra, R. C. (2015). Advanced international trade: Theory and evidence. Princeton university Press.

George, C., & Yamaguchi, S. (2018). Assessing implementation of environmental provi-sions in regional trade agreements. OECD.

Grant, J. H., & Lambert, D. M. (2008). Do regional trade agreements increase members’ agricultural trade? American journal of agricultural economics, 90 (3), 765–782. Head, K., & Mayer, T. (2014). Gravity equations: Workhorse, toolkit, and cookbook, In

Handbook of international economics. Elsevier.

Helpman, E., Melitz, M., & Rubinstein, Y. (2008). Estimating trade flows: Trading part-ners and trading volumes. The quarterly journal of economics, 123 (2), 441–487. Jayasinghe, S., & Sarker, R. (2008). Effects of regional trade agreements on trade in

agri-food products: Evidence from gravity modeling using disaggregated data. Review of agricultural economics, 30 (1), 61–81.

Kohl, T. (2014). Do we really know that trade agreements increase trade? Review of World Economics, 150 (3), 443–469.

Krugman, P. (1995). Increasing returns, imperfect competition and the positive theory of international trade. Handbook of international economics, 3, 1243–1277.

Lakatos, C., & Nilsson, L. (2017). The EU-Korea FTA: Anticipation, trade policy uncer-tainty and impact. Review of World Economics, 153 (1), 179–198.

(33)

Larch, M., & Yotov, Y. V. (2016). General equilibrium trade policy analysis with struc-tural gravity.

Magee, C. S. (2008). New measures of trade creation and trade diversion. Journal of International Economics, 75 (2), 349–362.

Mart´ınez-Zarzoso, I. (2018). Assessing the effectiveness of environmental provisions in regional trade agreements: An empirical analysis. OECD.

Mayer, T., & Zignago, S. (2011). Notes on cepii’s distances measures: The geodist database. Melitz, M. J., & Ottaviano, G. I. (2008). Market size, trade, and productivity. The review

of economic studies, 75 (1), 295–316.

National Board of Trade. (2019). The trade effects of EU regional trade agreements – Evidence and strategic choices.

OECD. (2018). Environment and regional trade agreements [Obtained 2020-02-15]. http: //www.oecd.org/environment/environment-and-regional-trade-agreements.htm OECD/Eurostat. (1999). The environmental goods and services industry: Manual for data

collection and analysis. OECD Publishing.

Ramsey, J. B. (1969). Tests for specification errors in classical linear least-squares regres-sion analysis. Journal of the Royal Statistical Society: Series B (Methodological), 31 (2), 350–371.

Redding, S., & Venables, A. J. (2004). Economic geography and international inequality. Journal of international Economics, 62 (1), 53–82.

Roy, J. (2010). Do customs union members engage in more bilateral trade than free-trade agreement members? Review of International Economics, 18 (4), 663–681.

Santos Silva, J., & Tenreyro, S. (2006). The log of gravity. The Review of Economics and statistics, 88 (4), 641–658.

Sauvage, J. (2014). The stringency of environmental regulations and trade in environ-mental goods. OECD.

Shepherd, B. (2013). Gravity model of international trade: A user guide. ARTNeT, United Nations ESCAP.

Soete, S., & Van Hove, J. (2017). Dissecting the trade effects of europe’s economic inte-gration agreements. Journal of Economic Inteinte-gration, 193–243.

Steenblik, R. (2005). Liberalising trade in ’Environmental goods’: Some practical consid-erations (tech. rep.). OECD Publishing.

(34)

The World Bank. (2020). World bank open data [obtained: 2020-03-28]. https://data. worldbank.org/

Tinbergen, J. (1962). Shaping the world economy; suggestions for an international eco-nomic policy.

United Nations Environment Programme. (2018). Trade in environmentally sounds tech-nologies: Implications for Developing Countries.

United Nations Statistics Division. (2017). Correlation and conversion tables used in un comtrade.

World Customs Organization. (2020). What is the harmonized system (HS)? [obtained: 2020-03-26]. http://www.wcoomd.org/en/topics/nomenclature/overview/what-is-the-harmonized-system.aspx

World Trade Organization. (2020). Regional trade agreement data base: https://rtais. wto.org/UI/charts.aspx

Yotov, Y. V. (2012). A simple solution to the distance puzzle in international trade. Economics Letters, 117 (3), 794–798.

Yotov, Y. V., Piermartini, R., Monteiro, J.-A., & Larch, M. (2016). An advanced guide to trade policy analysis: The structural gravity model. World Trade Organization Geneva.

Zugravu-Soilita, N. (2018). The impact of trade in environmental goods on pollution: What are we learning from the transition economies’ experience? Environmental Economics and Policy Studies, 20 (4), 785–827.

Zurek, K. (2018). Fr˚an handel och h˚allbarhet till handel f¨or h˚allbarhet i EU:s frihandel-savtal (A. M. Bakardjieva Engelbrekt Antonina & L. Oxelheim, Eds.). In A. M. Bakardjieva Engelbrekt Antonina & L. Oxelheim (Eds.), EU i en v¨arldsordning under omvandling. Europaperspektiv 2018. Sant´erus Academic Press.

(35)

8

Appendix

Table A1: Summary statistics of all the variables used in the regressions

Variables Observations Mean Std. Dev Min Max

T rade value 427,248 403 4,480 0 504,000 GDPi 422,729 751,000 2,320,000 132 20,500,000 GDPj 400,458 506,000 1,510,000 20,4 20,500,000 Distance 397,151 7378.444 4528.728 59.61723 19904.45 Contiguity 397,151 0.024 0.152 0 1 Language 397,151 0.149 0.356 0 1 Colony 397,151 0.020 0.140 0 1 EU 427,248 0.041 0.198 0 1 RT A 408,770 0.105 0.306 0 1 F T A 427,248 0.030 0.172 0 1 Envir 427,248 0.403 0.490 0 1

Note: T rade value (dependent), GDPi andj are denoted in million USD. Distance in

kilo-meters. Remaning variables are binary variables. Data sources for each variable can be found in 3.2.

Table A2: Constants for model specifications in Table 1

OLS (1) PPML (2) MTR (3) ENDG (4) PHSNG (5)

Constants −30.572∗∗∗ −18.403∗∗∗ 27.942∗∗∗ 23.352∗∗∗ 23.352∗∗∗

(36)

Table A3: List of the 236 countries potentially used in the regressions

Afghanistan Albania Algeria American Samoa

Andorra Angola Anguilla Antarctica

Argentina Armenia Aruba Australia

Austria Azerbaijan Bahamas Bahrain

Bangladesh Barbados Belarus Belgium

Belize Benin Bermuda Bhutan

Bolivia Bonaire Bosnia Herzeg. Botswana

Bouvet Island Br. Indian Ocean Terr. Br. Virgin Isds Brazil

Brunei Darussalam Bulgaria Burkina Faso Burundi

Cabo Verde Cambodia Cameroon Canada

Cayman Isds Central African Rep. Chad Chile

China China, Hong Kong China, Macao Christmas Isds

Cocos Isds Colombia Comoros Congo

Cook Isds Costa Rica Croatia Cuba

Cura¸cao Cyprus Czechia Cˆote d’Ivoire

Dem. Rep. of Korea Dem. Rep. of Congo Denmark Djibouti

Dominica Dominican Rep. Ecuador Egypt

El Salvador Equatorial Guinea Eritrea Estonia

Falkland Isds (Malvinas) Fiji Finland Fmr Sudan

Fr. South Antarctic T. France French Polynesia Gabon

Gambia Georgia Germany Ghana

Gibraltar Greece Greenland Grenada

Guam Guatemala Guinea Guinea-Bissau

Guyana Haiti Vatican City State Honduras

Hungary Iceland India Indonesia

Iran Iraq Ireland Israel

Italy Jamaica Japan Jordan

Kazakhstan Kenya Kiribati Kuwait

Kyrgyzstan Laos Latvia Lebanon

Lesotho Liberia Lithuania Luxembourg

Madagascar Malawi Malaysia Maldives

(37)

Mauritius Mayotte Mexico Mongolia

Montenegro Montserrat Morocco Mozambique

Myanmar N. Mariana Isds Namibia Nauru

Nepal Neth. Antilles Netherlands New Caledonia

New Zealand Nicaragua Niger Nigeria

Niue Norfolk Isds North Macedonia Norway

Oman Pakistan Palau Panama

Papua New Guinea Paraguay Peru Philippines

Pitcairn Poland Portugal Qatar

Rep. of Korea Rep. of Moldova Romania Russia

Rwanda Saint Barth´elemy Saint Helena St. Kitts and Nevis

Saint Lucia St. Maarten St. Pierre and M. St. Vincent and Gren.

Samoa San Marino Sao Tome and Pr. Saudi Arabia

Senegal Serbia Seychelles Sierra Leone

Singapore Slovakia Slovenia Solomon Isds

Somalia South Africa South Georgia South Sudan

Spain Sri Lanka State of Palestine Sudan

Suriname Swaziland Sweden Switzerland

Syria Tajikistan Tanzania Thailand

Timor-Leste Togo Tokelau Tonga

Trinidad and Tobago Tunisia Turkey Turkmenistan

Turks and Caicos Isds Tuvalu USA Uganda

Ukraine United Arab Emirates United Kingdom United States M.O.I

Uruguay Uzbekistan Vanuatu Venezuela

Vietnam Wallis and F. Isds Western Sahara Yemen

References

Related documents

The FTA between Uruguay and Mexico includes standstill and ratchet, which are in force and tightly restrict countries’ policy space, even though they do not apply to existing

Elevated levels of circulating endotrophin are associated with the presence of NASH, but not with fibrosis, in patients with NAFLD. Our results suggest that endotrophin could be a

The sanctions dummy results resemble previous results however with a slightly smaller negative value of -2.159, indicating that an influence of increase in sanctions will

This thesis engages in the debate on trade governance as means for marginalized groups to strengthen their ability to keep control over the development

A Free Trade Agreement (FTA) between the European Union (EU) and the Republic of Korea (hereafter referred as South Korea) has been under negotiation for several years. Discussions

A case study is presented on how the company Krusell Thailand can import semi- finished plastic cases from suppliers in China to a free trade zone warehouse in Thailand, where it will

In the case of the EUMED, 6 it does not include trade liberalization in agricultural goods and they found a negative significant influence on export trade flows for MENA..

220 Labour provisions in trade agreements may entail other positive effects like increasing labour related capacity building, 221 but there is a growing consensus in the