EVALUATION OF THE EFFECTIVENESS OF ECUADOR’S
TRADE AGREEMENTS
Master in Economics
International Economics and Globalization Track
Prepared by: María Alejandra Ruano Casañas, 11088176 Supervisor: Dr. D.J.M. Veestraeten
Second reader: Prof. Dr. F.J.G.M. Klaassen Submission date: July 5th 2016
2 Statement of Originality
This document is written by María Alejandra Ruano Casañas 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 The author wishes to thank the Secretariat of Higher Education, Science, Technology and
4
ABSTRACT
This study uses the gravity model of international trade to analyze the effect that trade agreements have had on Ecuadorean exports in the period 1991-2014. Special emphasis is given to the Andean Community and to the European Union Generalized System of Preferences (EU GSP). The sample of this study comprises 44 countries, which are Ecuador’s main trade partners. First, a brief overview of the economic development of Ecuador and its trade is given. The different trade agreements are explained, as well. Second, the variables used in the gravity model and its application are presented. Third, using country-pair fixed effects, Ecuadorean exports are regressed against Ecuador’s GDP and its partner’s GDP and one dummy variable is added for each trade agreement. The EU GSP is disaggregated in 4 dummies according to its different stages. The data was considered averaged every four years to account for short business cycles and also on a year-by-year basis. The findings show that both with data averaged and non-averaged, the Andean Community has increased Ecuadorean exports. On the other hand, the coefficient of the dummy variables of the EU GSPs were not significant. This can indicate lack of effectiveness in the EU GSP for Ecuadorean exports. Key words: Gravity model, exports, Ecuador, trade agreements, European Union Generalized System of Preferences, Andean Community
5
TABLE OF CONTENTS
1. INTRODUCTION ... 6
2. RECENT ECONOMIC DEVELOPMENT OF ECUADOR AND ITS TRADE ... 9
2.1 BRIEF OVERVIEW OF ECUADOR’S ECONOMIC DEVELOPMENT... 9
2.2 ECUADOR EXPORT STRUCTURE AND TRADING PARTNERS ... 13
2.3 ANDEAN COMMUNITY AND EUROPEAN UNION GSP ... 17
3. THE GRAVITY MODEL OF INTERNATIONAL TRADE ... 21
3.1 DEFYING GRAVITY: UNTANGLING THE MODEL ... 21
3.2 IMPACT OF TRADE AGREEMENTS ON INTERNATIONAL TRADE ... 22
3.3 TO LAG OR NOT TO LAG? MANAGING ENDOGENEITY AND HETEROGENEITY ... 27
4. EMPIRICAL ANALYSIS ... 30
4.1 MODEL AND DATA ... 30
4.2 ESTIMATION RESULTS ... 35
5. CONCLUSIONS ... 42
REFERENCES ... 45
6
1. INTRODUCTION
Ecuadorean exports have increased throughout time. From 1991 to 2000 exports increased by 69.1% and from 2000 to 2010 they increased by 262.7%. According to data from the World Bank, in 2014, Ecuadorean exports of goods and services represented 28.6% of GDP. This highlights the importance of understanding what is driving exports. Furthermore, from 1991 to 2000, Ecuador has also increased its participation in different trade agreements, which may have positively affected the growth of exports. The evolution of Ecuadorean exports from 1991 to 2014 is presented in Figure 1a. Ecuador is a founder country of the Andean Community (CAN, in Spanish Comunidad Andina de Naciones), which began in 1969, but it was not until 1993 that it became a Free Trade Area (FTA). Moreover, Ecuador became an associated member of Mercosur in 2004.
Figure 1a Evolution of Ecuadorean exports 1991-2014
Source: Graph constructed by author based on UN Comtrade data.
Ecuador also has bilateral trade agreements with Cuba signed in 1995 and the ACE29 with Mexico, which began in 1987 and was renegotiated in 1993 and 1994. The ACE32/65 was signed with Chile from 1995 and 2008, respectively. Additionally, Ecuador is a beneficiary of the Generalized System of Preferences (GSP) from the European Union. The GSP provides unilateral
0 5000 10000 15000 20000 25000 30000 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 USD (m ill io n s/n o min al terms ) Exports
7 preferential tariffs for developing countries. The EU GSP began in 1969 but has had several modifications since then. The USA GSP began in 1976 with the Andean Trade Preference Act (ATPA) but changed in 2002 to the Andean Trade Promotion and Drug Eradication Act (ATPDA), which further increased tariff liberalization. The Turkey GSP started in 2002 and more recently the Belarus, Kazakhstan and Russia GSP in 2010. Figure 2 shows the amount of Ecuadorean exports divided by its main trading partners from 1991 to 2014.
Figure 2a Ecuadorean exports classified by its main trading partners 1991-2014
Source: Graph constructed by author based on UN Comtrade data.
There are also ongoing trade negotiations with South Korea, the European Free Trade Association, El Salvador, Nicaragua and Honduras and with the European Union. (IMF 2015).
Ecuador’s international trade participation is relatively small in comparison to other South American countries. Despite the existence of studies on the effectiveness of Latin-American trade agreements as a whole, (Jacobo 2010) there is not much information regarding the effectiveness of the trade agreements specifically for Ecuador. The most important trade agreement for Ecuador in the South American region is the Andean Community. The trade agreement with the European Union is expected to conclude this year after 6 years of negotiations (IMF 2015) which highlights the importance to understand the impact that the EU GSP has had thus far.
0 2000 4000 6000 8000 10000 12000 19 91 19 92 19 93 19 94 19 95 19 96 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 20 14 U SD (m ill ion s/n o m in al term s) CHILE CAN EU MERCOSUR MEX USA
8 Hence, the thesis will answer the question: “To what extent have the Andean Community, the EU GSP and other trade agreements affected Ecuadorean exports in goods1 during the period 1991-2014?”
The empirical model that will be employed is an augmented gravity model of international trade which in its basic form explains how the economic size and the distance between countries affect their amount of trade. The data for this analysis will be extracted from the World Bank, the UN Comtrade and the Ecuadorean Central Bank. The GDP and export values are deflated by the US Consumer Price Index (CPI), which is obtained from the Bureau of Labor Statistics of the US Department of Labor. The information regarding the preferential trade agreements is from the Ecuadorean Ministry of Trade.
Particularly, panel data with country-pair fixed effects will be used and a regression model based on the papers of Cheng & Wall (2005) and Correia (2008), for the period 1991-2014 (the time period was chosen based on the availability of information) including 44 countries which are Ecuador’s main trade partners (see Annex I). Real exports from Ecuador to its partner countries will be regressed in their log linear form on partner country real GDPs, populations and dummy variables for the different trade agreements and GSPs. The GSP from the European Union will be disaggregated into 4 dummies according to the different stages of liberalization. Fixed effects estimation will be used in order to mitigate heterogeneity due to cultural, historical, political factors and other variables that differ between countries but remain the same over time such as distance and language, amongst others.
The thesis is structured in 5 chapters. Chapter 2 presents a brief overview of Ecuador’s economic development. This chapter describes Ecuador’s export structure and introduces Ecuador’s trade agreements. Previous studies will be evaluated in Chapter 3 regarding the use of the gravity model for international trade and its effectiveness among trade blocs and for particular countries. Chapter 4 covers the data and the empirical model created to analyze the effects of the trade agreements on Ecuador exports as well as a discussion of the results. Finally, in Chapter 5 the conclusions are presented.
9
2. RECENT ECONOMIC DEVELOPMENT OF ECUADOR AND ITS
TRADE
This chapter will analyze the development of Ecuador’s economy and its two main crises, which may have affected the level of exports, its main trade partners with an emphasis on the Andean Community and the European Union as well as Ecuador’s main export products.
2.1 BRIEF OVERVIEW OF ECUADOR’S ECONOMIC DEVELOPMENT
Ecuador’s economic development is characterized by several cyclical booms throughout the years. These cyclical booms were the result of the high exports of raw materials such as cacao (1866-1925), banana (1946-1968), and oil (1972-until now) (BCE 2010). A new stage in the economic development of Ecuador began in the decade of the 70s with the discovery of oil in the Amazon rainforest which led Ecuador to become part of the Organization of the Petroleum Exporting Countries (OPEC). According to data from the World Bank, oil rents as a percentage of GDP went from 0.05% in the early 70s to 16.6% at the end of the same decade. This was also the decade of international trade and trade liberalization not only for Ecuador but also for the entire Andean region (Bolivia, Colombia, Ecuador and Peru). It began with the formalization of the Andean Pact in 1969, which later became the Andean Community (see section 2.3). In 1973, annual GDP growth rate reached its peak, 13.9%, but one decade later fell to -0.33%2. Lucero (2001) mentions the oil export oriented economy with the addition of declining oil prices as reasons. This crisis proved to be one of the first “hints” that relying largely on oil exports was detrimental for the economy.
During the 90s, Ecuador was hit by several crises. First, in 1995, Ecuador and Peru went to war, which lasted for one month, over a land dispute that had begun in the 19th century with the independence from the Spanish crown and ended in 1999 with a Peace treaty with Peru (BCE 2010). Then, between 1997 and 1998, Ecuador was gravely affected by a natural disaster known as “El Niño”. This natural phenomenon is the result of the temperature changes in the Pacific Ocean accompanied by extremely heavy rainfall and occurs irregularly every 7 years with a maximum delay of 15 years. Due to the intense floods and landslides, at least 286 people died and 30,000 went homeless. Crops were lost and infrastructure was severely damaged. (Vos et al. 1999). To make
2 Source: World Bank Data.
10 matters worse, from 1996 to 1998, oil prices fell drastically, around 31%3 on average. The “sucre” (Ecuador’s legal tender from 1884 to 2000) devalued 132% against the U.S dollar in the period 1995-1998 and in 1999 the sucre devalued 194% against the U.S dollar within one year4. Several banks went into bankruptcy, around 16 banks closed. Some banks were bailed out by the government absorbing around 23% of GNP. Thus diverting resources from social spending (Lucero 2001). By 1999, GDP growth had the biggest drop in Ecuadorean history thus far, reaching -4.74%. (See figure 2a) According to the Ecuadorean Institute of Statistics (INEC), during this decade, around 5.2% of the total population emigrated mainly to USA and Spain.
Figure 2a. Evolution of Ecuador’s GDP Growth
Source: Graph constructed by author based on World Bank Data
The year 2000 combined all the problems from the 90s and inflation reached its highest peak, 96.1% (see figure 2b). The sucre devalued even further, 1 dollar now became 25,000 sucres. According to Lucero (2001), the then President, Jamil Mahuad, considered that the only solution to foster investors’ trust and save the economy was to turn to dollarization. In March 2000, Ecuador embraced the US dollar as its new legal tender. Thus reducing the availability of monetary policy
3 Source: Federal Reserve Bank of St. Louis and US. Energy Information Administration. 4 Source: Banco Central del Ecuador. In 1999, one dollar was equivalent to 19,858 sucres.
-4.74% -6 -4 -2 0 2 4 6 8 10 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 An n u al %
11 instruments to boost the economy and counteract economic crises. Because since then, the Ecuadorean Central Bank can no longer control the supply of money and is dependent of the decisions of the U.S. Federal Reserve.
Figure 2b. Evolution of Ecuador’s inflation, consumer prices
Source: Graph constructed by author based on World Bank Data
Ecuador’s economical strains developed into social and political ones. From 2000 to 2007 Ecuador had five presidents. Two democratically elected presidents were removed by “coup d’état”, one was removed by the National Congress of Ecuador on the basis of “mental incapacity” and the other two were the ones who replaced the previous presidents and finished their mandates. Political stability began in 2007 when Rafael Correa was democratically elected. With the oil prices on the rise, (see figure 2c) the economy seemed to be blooming. GDP growth went from 2.2% in 2007 to 6.4% in 2008. However, prosperity was cut short, since a new crisis was in the horizon.
In 2009, the shadow of the international financial crisis, reached Ecuador. This crisis began in the United States in 2007 due to the subprime mortgage crisis and affected financial institutions in many developed countries. Dullien et al. (2010) state that developing economies were affected in an indirect way, through the trade channel and through the workers’ falling remittances, and highlights that “the economic consequences of these indirect effects were as severe as the direct
effects were on developed countries”. (Dullien et al. 2010, p. 17)
96.1 0 20 40 60 80 100 120 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 Ann u al %
12 Acosta (2009) also mentions the workers’ falling remittances, from those who had emigrated in the 90s, as a channel of how the international crisis affected Ecuador, and adds the low oil prices as another reason. As stated by Acosta, the remittances fell by 9.4% from 2007 to 2008, which represented a reduction of 1.6% of GDP and in 2009 the remittances fell by 27.04%. Similarly, the price of the crude oil barrel slipped 37.8% from 2008 to 2009. Acosta explains: “each
dollar less in the price of the Ecuadorean crude oil barrel equals, approximately, a net decrease in government revenue of 57.8 million of dollars” (Acosta 2009, p. 2) and in 2009, the price of the crude
oil barrel was on average 61.95 US dollars (see figure 2c). Oil prices fell since investors expected that the international crisis would diminish the demand for oil, thus selling the oil they had previously purchased not as raw material, but as financial instruments, pushing the prices downwards (BCE 2010). Moreover, the 2009 crisis was also strengthened by the appreciation of the dollar, which reduced the competitiveness of the Ecuadorean exports (Acosta 2009).
Figure 2c. Evolution of average annual crude oil prices
Source: Graph constructed by author based on data from Federal Reserve Bank of St. Louis
There is no doubt that one of the biggest weaknesses Ecuador faces is its dependency on oil exports (Acosta 2009), which is why during president Correa’s terms (he was chosen for a second term in 2009 and a third one in 2013), he made an emphasis in steering the Ecuadorean economy away from the dependency of oil. His economic development plan known as the “Productive Matrix”
99.67 61.95 0.00 20.00 40.00 60.00 80.00 100.00 120.00 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 U S Dolla rs p er b ar re l
13 is based on four main pillars according to the Ministry of Production, Labor and Productivity: developing new productive sectors such as mariculture and biofuels, industrialize primary traditional products, motivate Ecuadorean industries to produce products that are imported, and improve the quality of the products that are exported. Nevertheless, these structural changes to promote exports are not beneficial by themselves without the right market. Which is why the International Monetary Fund (IMF) in its 2015 Article IV Consultation, highlights the importance to promote trade integration and finalize the negotiations for a Free Trade Area with the European Union.
2.2 ECUADOR EXPORT STRUCTURE AND TRADING PARTNERS
Ecuador’s export structure is heavily dependent on agricultural products and raw materials. It can be divided in two categories: traditional and nontraditional products. Traditional products are bananas and plantain, shrimps, cocoa and its derivatives, tuna and fish and coffee and its derivatives. Figure 2d shows the evolution of Ecuadorean exports of traditional products in the period 2010-2014.
Figure 2d. Ecuadorean exports of traditional products
Source: Graph constructed by author based on data from Ecuador Central Bank and ProEcuador (Institute of Promotion of Exports and investment).
$0 $500,000 $1,000,000 $1,500,000 $2,000,000 $2,500,000 $3,000,000 Bananas and plantain
Shrimp Cocoa and derivatives
Tuna and fish Coffee and derivatives U SD 2010 2011 2012 2013 2014
14 Nontraditional products include natural flowers, wood, mining products, fruits, tobacco (plant). Figure 2d shows the evolution of Ecuadorean exports of nontraditional products in the period 2010-2014.
Figure 2e. Ecuadorean exports of nontraditional products
Source: Graph constructed by author based on data from Ecuador Central Bank and ProEcuador.
Ecuadorean exports have benefited from the Generalized System of Preferences of the European Union, United States, Canada, Switzerland, Japan and Norway since the early 70s. These preferences are given unilaterally to developing countries and allow them to pay less or no duties on their exports. The USA GSP began in 1976 with the Andean Trade Preference Act (ATPA) but changed it in 2002 into the Andean Trade Promotion and Drug Eradication Act (ATPDA), which further increased tariff liberalization. More recently, Ecuador has also been able to benefit from the GSP of Turkey and Russia since 2002 and 2010, respectively.
Likewise, there are bilateral trade agreements signed with Mexico in 1987 with additional renegotiations signed in 1993 and 1994. Other with Cuba in 1995 and there have been two bilateral
$0 $200,000 $400,000 $600,000 $800,000 $1,000,000 $1,200,000 Natural flowers Wood Mining products Fruits Tobaco (plant) U SD 2010 2011 2012 2013 2014
15 trade agreements with Chile, the first one known as the ACE325 signed in 1995 and the second one known as ACE65 signed in 2008 which increased trade liberalization with Chile. Among the main products that benefit from total tariff reduction or total tariff elimination in these bilateral trade agreements are cocoa and its derivatives, tobacco, tuna and fish and textile products.
Ecuador’s main trading partners are the USA, Chile, Peru, Colombia, Venezuela, Vietnam and several countries from the European Union (PROECUADOR 2015). Ecuador’s main trade agreements, according to the Institute of Promotion of Exports and Investment (PROECUADOR), included in the analysis of this paper are summarized in table 2a.
Table 2a Ecuador’s trade agreements
Trade Partner Type of Program Since Until European Union GSP 1969 Present USA GSP 1976 2012 Andean Community (CAN) Andean Pact FTA 1969 1993 1993 Present Mexico ACE 29 1987, 1993, 1994 Present Chile ACE 32/ACE65 1995, 2008 Present
Cuba ACE 1995 Present
Turkey GSP 2002 Present
MERCOSUR Associate member 2004 Present Belarus,
Kazahstan and Russia
GSP 2010 Present
Source: Table constructed by author based on Ecuador ministry of commerce and PROECUADOR
Ecuador became an associate member of Mercosur in 2004, which allows the country to benefit from certain tariff reductions when trading with Mercosur’s member countries: Argentina, Brazil, Uruguay, Paraguay and Venezuela. Each Mercosur country has a particular list of products
5 In spanish ACE: Acuerdo de Complementación Económica/Economic Complementation Agreement: is the terminology used by Latin American countries for bilateral trade agreements.
16 which specifies the percentage of tariff reduction given to and given by Ecuador as well as exceptions.6
6 The list of the products that benefit from the unilateral and bilateral trade agreements, as well as the exceptions, mentioned in Chapter 2, Section 2.2 can be accessed online via http://www.comercioexterior.gob.ec/acuerdos-comerciales-3/
17
2.3 ANDEAN COMMUNITY AND EUROPEAN UNION GSP
Ecuador’s main regional trade agreement is the Andean Community, known in its beginnings as the Andean Pact. It began in 1969 with 5 countries: Bolivia, Chile, Ecuador, Colombia and Peru. Its objective was the integration and the economic and social cooperation between its participants through a Free Trade Area (FTA). However, the FTA did not start immediately. The elimination of internal tariffs was a gradual process, which ended in 1993 when the FTA was able to function to its full capacity for Bolivia, Colombia, Ecuador and Venezuela. Chile left in 1976 and returned as an associated country in 2006 and Venezuela joined in 1976 to leave in 2006 as a political protest due to the FTA negotiations by Colombia and Peru with the United States. At the present time, the Andean Community is formed by Bolivia, Colombia, Peru and Ecuador. Figure 2f shows the evolution of Ecuadorean exports to the countries from the Andean Community in the period 1991-2014.
The Andean Community FTA embraces total tariff abolition without any exception. Apart from the absence of tariffs among member countries, since 2003 there is also free flow of people.
Figure 2f. Evolution of Ecuadorean exports from the Andean Community
Source: Graph constructed by author based on UN Comtrade data. 0 500 1000 1500 2000 2500 1991 1992 1993 1994 1995 1996 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 U SD (m ill ion s/n o m in al term s)
18 The European Union is the third most important market for Ecuadorean exports after the USA and the Andean Community (see graph 1a). From the total exports from the European Union in 2014, exports from Ecuador to Germany represent 18%, followed by Spain and Netherlands with 18% and 17% respectively.
Figure 2g. Ecuador’s main export markets from the European Union Countries share of EU imports
2014
Source: Graph constructed by author based on UN Comtrade data.
The European Union GSP went through several regulation modifications throughout the years regarding its conditions about quotas, product coverage, ceilings and depth of tariff cuts, granting either duty-free access or a tariff reduction. The first GSP began in 1969 and lasted for a decade. This program benefited agricultural, textile and industrialized products from 178 countries, Ecuador among them. Later, it was renewed for 10 more years. The second phase of the GSP began in 1991 with a special arrangement to aid Andean and Central American countries combat drug production and trafficking, thus motivating the replacement of drug cultivation with alternative products, as well as complying with labor and environmental standards. (UN 2008)
Belgium 7% France 11% Germany 18% Italy 14% Netherlands 17% Poland 2% Spain 18% United Kingdom 6% Other EU countries 7%
19 From 1995 to 2001 a new cycle of the EU GSP began, this new period was an extension to the previous one, including now special incentives to promote environment protection and social rights for all the beneficiary countries. The biggest departure from the 1991 GSP was the elimination of quantitative limitations of GSP imports which were replaced by four categories of duty-free treatment according to product sensitivity: 15% preferential margin for very sensitive products, 30% preferential margin for sensitive products, 65% preferential margin for semi-sensitive products and duty-free entry for non-sensitive products (United Nations 2008). Another addition was the introduction of the country-sector graduation mechanism by which products from the beneficiary countries which fell under certain criteria7 could no longer benefit from the preferential treatment.
The next phase lasted two years, from 2002 to 2004. In this period the tariff preferences were deepened to aid the fight against drug production and trafficking. Furthermore, the four categories of duty-free treatment were reduced to only two: non-sensitive and sensitive products. Non-sensitive products enjoyed duty free access, while sensitive products benefited from a tariff reduction of 3.5 percentage points in the full ad valorem rate of customs duty payable. (United Nations 2008). According to the European Commission around 3250 products are classified as non-sensitive, while 3750 as sensitive.
In 2005, India objected to the special arrangements given to Andean and Central American countries to combat drug production taking the dispute to the World Trade Organization (WTO). India considered that this special preference was not in accordance to the Most Preferred Nation principle (MFN) under which countries are forbidden to discriminate between trading partners. Since the European Union could not prove that such discrimination did not exist, a new regime began in late 2005 named the GSP plus which ran until 2008. The GSP plus included special arrangements for least developed countries, granting more favorable tariff treatment such as total suspension of duty for eligible products. From 2009 to 2013 the same conditions from the previous arrangements were maintained. (De Benedictis et al. 2011)
7 “The sector/country graduation combines a development criterion, expressed as a development index reflecting a country's per capita income and the level of its manufactured exports as compared with those of the Community, with a measurement of relative industrial specialization expressed as a specialization index based on the ratio of the beneficiary country's share of total Community imports in general to its share of total Community imports in a given sector; whereas combined application of these two criteria should make it possible to adjust the crude results of the specialization index, in terms of the sectors to be excluded, in line with the level of development.” Regulation (EC) No. 3281/94, page 2.
20 The new EU GSP, which began in 2014, features an annual review mechanism by which countries that were ranked by the World Bank as a “high” or “upper-middle-income” economy for three consecutive years must be removed from the list of beneficiaries of the GSP. According to the World Bank, Ecuador, for the third year in a row, has been considered as an upper middle income economy, meaning its Gross National Income (GNI) per capita has been in the range of $4,126 to $12,735. Therefore, in 2015 Ecuador was removed from the list of GSP beneficiaries.
In 2007, the Andean Community, as a bloc, began negotiations with the European Union for a trade agreement. However, in July 2009, Ecuador and Bolivia, who followed a more protectionist stance, left negotiations to signal disagreement with the high import tariffs on certain products. Specifically for Ecuador, the problem laid on the import tariff on bananas.
In December 2009, the Geneva Banana agreement was reached between the EU, USA and Latin American countries. This agreement reduces banana import tariffs from 176 euros per ton to 114 euros per ton8 within 8 years until 2019, at the latest. In 2012, the WTO announced that the EU and 10 Latin American countries (Ecuador among them) had signed the 2009 deal. Due to these improvements, Ecuador renewed formal negotiations in 2014.
Contrary to Ecuador, Peru and Colombia never ceased negotiations with the EU and in 2012 each signed their agreements. These trade agreements give these countries further benefits. The import reduction, for Peru and Colombia, is meant to gradually reach 75 euro per ton in 2020. This left Ecuador in disadvantage as Colombia and Costa Rica, who also signed a trade agreement with the EU, are Ecuador’s main competitors in the banana sector.
Since the European Union is such an important market for Ecuador (see figure 1b), it is adamant to finalize the negotiations with the European Union for the Free Trade Area (IMF 2015). In the regression analysis in this study, the different stages of the European Union GSP will be summarized according to the similarity of each scheme. Four stages have been considered: 1991-1994 Drug eradication program, 1995-2001 Duty free treatment according to product sensitivity, 2002-2004 deeper tariff liberalization and 2005-2014 GSP plus.
21
3. THE GRAVITY MODEL OF INTERNATIONAL TRADE
This chapter begins by presenting the gravity model of international trade and the different explanatory variables that have been included in the model throughout the years. An emphasis is made on its application as a tool to analyze the effect of unilateral, bilateral and regional trade agreements on trade.
3.1 DEFYING GRAVITY: UNTANGLING THE MODEL
Tinbergen (1962) found inspiration in Isaac Newton’s law of universal gravity and applied it to the study of international trade. This law states that the force by which two bodies (countries) are attracted (trade) to each other is positively related to their mass (economic size measured by GDP) and negatively to the distance between them. He intended to find out the variables that determine trade between countries and concluded that the main ones were the size of the countries and their geographical distance, think of transportation costs.
Since then, many authors have modified the Tinbergen model by adding more independent variables in their empirical studies resulting in the augmented gravity model. Linnemann (1966) added the population of the importing and exporting countries since he wanted to determine whether the size of the countries’ population is related to the amount of trade. Even though the population variable has been proven to be significant (Mátyás 1997; Cheng & Wall 2005), there is no agreement on whether the sign of the variable should be positive or negative. For instance, a positive coefficient on the population variable of the importing country would show that larger countries have greater capacity to absorb imports thus increasing trade. Meanwhile, a negative sign of the same variable, could show that larger countries count with more people, thus more resources, diminishing their need to import products; consider economies of scale. According to Jacobo (2010) the sign of the population coefficient could be negative or positive, only the interpretation changes, depending if the country exports less when it has a larger population (absorption effect) or a country with a larger population exports more than a country with a smaller population (economies of scale).
Mátyás (1997) highlighted the importance of adding country-pair fixed effects and included dummy variables that should measure the impact of participation in regional trade agreements. Mátyás considered a positive sign as trade creation and a negative sign as trade diversion.
22 Bergstrand (1985) was the first to introduce exchange rates in the gravity model. Rose (2000) and many others included common language as a dummy variable to account for cultural proximity since cultural differences such as language could increase transaction costs and reduce trade. Rose tries to control for omitted variables including the exchange rate and dummy variables related to culture and history which try to explain whether the countries shared a common border, were previously colonies or had the same colonizer. Additionally, he analyzed the influence of currency unions on international trade flows covering in his study about 186 countries during 1970 to 1990 and he concluded “two countries that share the same currency trade three times as much as they would
with different currencies”. (Rose 2000, p.1)
3.2 IMPACT OF TRADE AGREEMENTS ON INTERNATIONAL TRADE
Sanso et al. (1993) highlighted the alleged empirical success of the gravity equation as a tool to understand the determinants of bilateral trade. They used an augmented model with per capita income in a cross section of annual observations on trade in 16 OECD countries from 1964 to 1987. They agree with the conclusion that trade is inversely proportional to the distance between the countries. Rose (2003, 2004) as others before him, used panel data and analyzed 175 countries to examine whether World Trade Organization (WTO) membership promotes trade and found no significant evidence that being a WTO member increases trade, although he calls this result a mystery since “common sense and conventional wisdom accord an important role to the GATT/WTO
in creating trade”. (Rose 2003, p.23)
As many others, Rose (2002) and Carrillo & Li (2002) analyzed the effects of the trade agreements on international trade using the gravity model. Carrillo & Li reviewed the effect of the Andean Community and Mercosur on intra-regional and intra-industrial trade in the period 1980-1997 while Rose covered several regional trade agreements in the period 1948-1999 such as the North American Free Trade Agreement (NAFTA), the Association of Southeast Asian Nations (ASEAN), the European Union, the Caribbean Community and Common Market (CARICOM), Mercosur, among others. Rose’s research was focused on estimating protectionism of the countries (trade barriers) and how protectionism affects trade.
The authors use a panel data with time fixed effects. For the dependent variable, Rose considered the logarithm of the average value of real bilateral trade between one country and the other while Carrillo & Li used the value of one country’s imports from its partner.
23 There are some similarities in the independent variables. Both included distance and the logarithm of real GDP; although Rose considered the real GDP as the logarithm of the multiplication of the GDP per capita of both countries. The trade agreements are incorporated in both studies as dummies. Additionally, Carrillo & Li added the bilateral exchange rate defined as the log of the ratio of the countries’ real exchange rate with respect to the US dollar and deflated by the countries’ CPI. They add a dummy to account for the reopening of the international credit market in Latin America after 1990. Meanwhile, Rose adds common language and three dummy variables that represent whether the country is a colony at time t, if the country i ever colonized country j and vice versa, or if both countries remained part of the same nation throughout the sample. For example United Kingdom and Bermuda.
Both authors reach similar conclusions. The results for the main variables of the gravity model are consistent with the findings of previous studies. Countries further apart trade less, countries with the same language trade more as countries with higher GDP or GDP per capita. Thus concluding that the economic size of the countries is a determinant of trade, as countries with larger GDPs, “richer countries” are more likely to trade. However, Carrillo & Li found that Mercosur and the Andean Community only have had a small positive impact on trade. They considered that the impact was small because the volume of intra-regional trade had increased due to unilateral trade liberalization reforms implemented in the 1990s as well as the fact that the countries that belong only to the Andean Community or only to Mercosur have common borders and common transportation infrastructures that ease trade. In comparison, Rose also found that there is less trade in Latin America than what can be due to the existence of high trade barriers.
Empirical studies for South America have found significant evidence of trade creation such as the works of Martínez-Zarzoso & Nowak-Lehmann (2003) and Jacobo (2010). The authors analyzed trade creation and the determinants of bilateral trade flows for the European Union and MERCOSUR for 20 and 16 countries respectively. In the regressions, they include the logarithm of exports as the dependent variable and the populations and GDPs of the importer and exporter in logs as independent variables. However, Jacobo’s GDPs specification is different since he adds the GDPs as the multiplication of both variables. They all add dummies for the trading blocs, which take the value of 1 if the country belongs to the European Union and 0 otherwise. Furthermore, Martínez-Zarzoso & Nowak-Lehmann add the real exchange rate and an infrastructure variable for both the importer and exporter. The infrastructure index is constructed by the authors and
24 considers data on roads, paved roads, railroads and number of telephone lines. After doing the Hausman test, which is used to discriminate between fixed and random effects testing the null hypothesis that the explanatory variables and the individual effects are uncorrelated, they conclude that a panel data with fixed effects is preferred over random effects since the null hypothesis is rejected under random effects. Thus, the fixed effects specification provides consistent results.
Their results are similar. Jacobo and Martínez-Zarzoso & Nowak-Lehmann show that the GDP of the exporter and the importer have a positive effect on bilateral trade flows, meaning that an increase in the countries’ GDP increases exports. Thus the stronger a country´s economy is, the more the country will trade. Geographical distance, on the contrary, decreases exports, since the further the countries are apart, transaction costs increase as well as transportation costs. The dummy variables for trade blocs are statistically significant and show a positive sign, therefore membership of a trade agreement also increases exports. Additionally, Martínez-Zarzoso & Nowak-Lehmann found that the coefficient for exporter infrastructure was positive and significant while the importer infrastructure was not significant, meaning that investing in the infrastructure of the trade partner did not have spillover benefits for the investor. Meanwhile, the coefficient of the exchange rate variable was found to be positive and statistically significant which according to the authors indicates that price competitiveness is imperative. They find that a 10% depreciation of the exporter currency rises exports by 2.8%.
Martínez-Zarzoso & Nowak-Lehmann concluded that the population of the exporter country has a large and negative effect in exports meanwhile importer population has a large and positive one which they explained as bigger countries importing more than smaller ones due to the larger population. Jacobo also concludes that a country with a bigger population exports more but he states the economies of scale as a reason. (See Linnemann 1966)
There are specific studies for countries in Latin America such as the one from López & Muñoz (2008), who used the gravity model in its basic logarithmic form to explain the determinants of trade flows in Mexico and Chile in the years 1994-2012. They used panel data where the dependent variable is imports plus exports from the host country and the independent variables are: distance, dummy variables in order to measure the impact of the trade agreements, as well as the multiplication of GDPs from host and partner country. They find, in the case of Mexico, no evidence that its trade has increased as a result of their trade agreements, whereas for Chile, the dummy variables that measure the impact of trade agreements are positive and significant.
25 Two authors, Correia (2008) and Mejía (2011), analyzed the effectiveness of the European Union GSP on promoting Colombian exports. Both authors used panel data where the dependent variable is the logarithm of real exports from Colombia to its partner country. Among the similar independent variables used are the logarithm of Colombian real GDP and its partner’s, distance from Colombia to the capital city of the other country, as well as dummy variables for geographical and cultural characteristics such as common language, common border and whether the importer country is an island. The last variable was included considering that an island does not have neighboring countries therefore hindering trade (expected negative sign). Correia added a variable to account for a crisis that Colombia went through in 1999 as he considered that because the crisis had a negative impact on the economy; it would also have a negative impact on exports.
Another difference is how the authors account for the EU GSP. Mejía uses only one dummy variable for the EU GSP since he only considers the GSP Plus, which ran from 2005 to 2008, despite the fact that his period of analysis is 1990-2008. He considers the regional trade agreements (RTAs) using only one dummy which is 1 if Colombia and its trade partner are part of the same regional trade agreement or if Colombia receives any other preferences from the trade partner and 0 otherwise.
On the other hand, Correia disaggregated the EU GSP into 3 categories in order to analyze whether the consecutive changes in the GSP had benefitted Colombian exports and to what degree, expecting a positive coefficient on each variable. The first category was the EU GSP from 1991-1994 which is recognized for the drug eradication program, the second one from 1995-2001 known for the elimination of quantitative restrictions and the last one from 2002-2005 for further product coverage. Additionally, Correia adds independent dummy variables for each additional trade agreement in order to assess the effect that each one have had in Colombian exports. He considers therefore the USA GSP, the bilateral trade agreements with Mexico and Chile and Colombian regional trade agreements such as the Andean Community (Colombia is one of the founder countries) and the Mercosur as an associated country.
Mejía added the bilateral real exchange rate in his regression, determined as the product of the nominal exchange rate and the ratio of prices between the two countries and found that the coefficient is positive and statistically significant. The appreciation of the currency of the partner country by 1% raises Colombian exports by 0.6%. On the other hand, Correia did not include the exchange rate stating that most authors found it insignificant. Martínez-Zarzoso & Nowak-Lehmann
26 (2003) also included the exchange rate as an independent variable and found that it was significant. According to their findings a 10% change in the real exchange rate increases exports by 2.6%.
Correia analyzed three regression models estimated by Ordinary Least Squares (OLS), model 1 had no time effects plus country fixed effects, model 2 had time effects but no country fixed effects and model 3, time and country fixed effects. He concluded that the best one is the time and country fixed effects model since it provides the best fit. However, when applying this model, the variables that remain the same throughout time such as distance and language are dropped since they are included in the time fixed effects, therefore they disappear from the regression equation. To avoid dropping these variables, which are considered as the “standard features of the gravity model”, Martínez-Zarzoso & Nowak-Lehmann, who also applied the fixed effects model, estimated them running a second regression. In this additional regression, the dependent variable is the individual effects (the fixed effects that resulted from two previous regressions, one using the exchange rate as a variable and the other without) and the independent variables are distance, a dummy for common language and a dummy, which is 1 if the countries share a border and zero otherwise. Even though they found distance to be significant, the language and adjacency dummies are not, which is contrary to the findings of other authors and they propose that these results should be further investigated. Nevertheless, Cheng & Wall (2005) consider the elimination of these traditional variables as an advantage of the time fixed effects model. They highlight that the distance from one capital city of one country to the other is not an appropriate proxy for transportation costs. The capital of the country may not necessarily be the economic center, especially for bigger countries which have several cities as center of trade, and even more fundamentally the transaction costs such as shipping containers may decrease or increase over time.
The results found by Correia and Mejía contradict each other. The first author finds that the EU GSP had a positive effect in model 1 and model 2, concluding that the EU GSP had increased exports in Colombia while a negative effect in model 3 (considering both time and fixed effects) concluding the opposite. Correia considers that model 3 provides the right estimation since is the most theoretically consistent because it includes all time and country specific factors not considered otherwise. Hence, he attaches more emphasis to the results of the last model. According to model 3, the EU GSP had no success in promoting exports since the 3 coefficients of the variables that control for the different stages of the EU GSP are negative, increasing in magnitude and are significant. However, he explains that this is not the result of the lack of efficiency of the EU GSP but
27 of the improvement of the trade agreements with other countries such as the USA GSP and the FTA with the Andean Community since both are positive and significant.
Meanwhile, Mejía found that the variable for EU GSP was positive and significant. The results may differ since Correia disaggregated the EU GSP into different categories, in this way he could control better for the progressive nature of the GSP throughout the years. This progressive nature refers to the different stages that the EU GSP has gone through as stated in Chapter 2.
3.3 TO LAG OR NOT TO LAG? MANAGING ENDOGENEITY AND
HETEROGENEITY
In all the previous papers, endogeneity was not accounted for. In the gravity model one way that endogenity could be present is due to the impact that history has on trade. For instance, countries that have traded for many years are more likely to continue doing so even if there is no trade agreement between them. After years of negotiating in a foreign country, leaving the market could be more costly that continuing running the business. Before exiting, exporters will consider sunk costs such as operational and marketing costs, and other factors such as having, by then, a proper supply chain, customer loyalty, etc. (Eichengreen & Irwin 1998).
Eichengreen & Irwin (1998) were the first ones to consider endogeneity by adding dynamic effects to the gravity model. They used annual lags in a cross section model taking into account the evolution of trade in the period 1949-1964. In their regression, the dependent variable is the logarithm of the value of bilateral trade between countries and the independent variables are: the logarithm of the product of the countries’ national incomes, the logarithm of the product of the countries’ per capita income, a dummy for each different trade agreement and a dummy to account whether the countries are contiguous. They run two regressions; the first one is the simple general specification without lags and the second one considering lagged trade, lagged incomes and lagged incomes per capita of both countries. They find that the variables income and distance are still significant and concluded that “lagged trade exercises an important effect even after controlling for
the arguments of the standard gravity model”. (Eichengreen & Irwin 1998, p. 56)
Similarly, Bun & Klaassen (2002) highlighted the importance of lags in the gravity model, however, contrary to Eichengreen & Irwing (1998), they used a panel data with country pair and time fixed effects for the period 1950-1997 considering annual lagged trade as well as lagged income terms. They find that the lagged variables are strongly significant and that “transitory shocks to trade
28 persist for a long time”. Baier et al. (2014) agrees that trade is dynamic but argued that 5 year lags are superior to annual lags since trade cannot adjust within one year. To back this claim they give two reasons, the first one is that five-year lags increase estimation efficiency due to the size of the panel data and second that differencing for more than one year can avoid a spurious regression problem since it is possible that trade flow data and real GDP data are nearly unit root processes.
Cyrus (2002) examined endogeneity in the gravity model, considering that a potential problem for the model, when estimated using OLS, is the causality between income and trade, since the regression states that high income causes high trade, but in reality, the relation could be the opposite. According to the author’s hypothesis, OLS will overstate the importance of income. She used instrumental variables (based on the augmented Solow model of Mankiw, Romer & Weil, 1992) that are correlated with the Gross National Product (GNP) but uncorrelated with the error term to determine the true effect of income on bilateral trade. She included in the traditional gravity model the logs of factor accumulation variables (instrumental variables IV): physical capital accumulation, human capital accumulation and population growth. The results show, using IV, that a 1% increase in either country’s GNP causes a 0.52% increase in trade, while using OLS the result is 0.51%. Moreover, the coefficients on distance and adjacency are virtually unchanged. She concludes that the effect of income on bilateral trade is highly significant but using instrumental variables only slightly alters the results compared to OLS.
One type of endogeneity arises due to omitted variable bias. In order to find the best way to account for this, Cheng & Wall (2005) assessed ways that heterogeneity had been allowed for in an empirical analysis. They compare the results of 5 regressions that differ in the estimation: pooled cross-section, unrestricted country-pair fixed effects, symmetric country-pair fixed effects, first-difference model and Mátyás (1997) model which is a special case of the fixed effects model since it includes two fixed effects, one as an exporter and one as an importer. In all 5 estimations, the dependent variable is the logarithm of the real exports from origin country to destination. The explanatory variables in the last 4 estimations are: logarithm of real GDP for both origin and destination, origin and destination population and one dummy variable for each trade agreement (European Union, NAFTA, MERCOSUR, Australia-New Zealand, and the Israel-United States Free Trade Agreement). Only standard cross-section includes additionally the geographical distance between the countries, a dummy variable, which is 1 if the countries are contiguous, and another dummy for common language. Cheng & Wall use a balanced panel data of 29 countries for the
29 period 1982-1997. They leave 5 years between the observations as in the fixed effects estimation, as they consider the dependent and independent variables cannot adjust entirely in 1 year. They concluded that unrestricted country-pair fixed effects with time effects is statistically to be preferred to other specifications. This model explains 95% of the variations in exports.
Ecuador was included in the analysis of previous studies as part of the Andean Community, but not considered on its own. There is a lack of literature analyzing the effects of trade agreements in Ecuador using the gravity model, the only one seems to be Barragán (not dated). He examined the trade intensity between Ecuador and the countries in Asia-Pacific and used the gravity equation to explain the determinants of exports, imports and trade flows finding a negative relation with the distance, different religion and trade tariffs, and a positive one with population, real GDP of the countries in Asia-Pacific and common language.
30
4. EMPIRICAL ANALYSIS
This chapter presents in Section 4.1 the regression model based on the gravity model designed to assess the effect on Ecuadorean exports that the European Union GSP, the Andean Community as well as other trade agreements have had. Section 4.2 discusses the results of the empirical research.
4.1 MODEL AND DATA
The empirical model that will be employed is an augmented gravity model of international trade, which in its basic form explains how the economic size and the distance between countries affect trade flows between them. This model will be used to investigate whether the European Union GSP and the Andean Community have affected the amount of Ecuadorean exports in the period 1991-2014. The effect that other trade agreements and determinants may have had on Ecuadorean exports is also analyzed.
The sample includes 44 of Ecuador’s trade partners, countries that have trade agreements with Ecuador (See Annex 1).
The data for this study will be extracted from the World Bank, the United Nations Commodity Trade Statistics Database (UN Comtrade) and the Ecuadorean Central Bank. The export values and GDPs from Ecuador and its trade partner are deflated by the United States Consumer Price Index (CPI-U-RS) obtained from the Bureau of Labor Statistics of the US Department of Labor. The information regarding the preferential trade agreements is from the Ecuadorean Ministry of Trade. The period 1991-2014 is used due to the availability of data. Data for the dependent variable, exports from Ecuador to its trade partners, was only available from 1991 onwards.
A pooled panel data regression will be used following the econometric specifications of Cheng & Wall (2005). The observations are averaged every 4 years to account for the criticism mentioned by Cheng & Wall that both the dependent and independent variables cannot fully adjust in a year’s time. Thus, 6 periods will be analyzed, 1991-1994, 1995-1998, 1999-2002, 2003-2006, 2007-2010 and 2011-2014.
The regression will include time fixed effects and will be estimated using least squares with a dummy variable for each of the country pairs (LSDV) to account for heterogeneity.
31 Thus, the empirical model that will be applied to examine the impact of the trade agreements on Ecuador’s exports in the period 1991-2014 is the following:
ln 𝑋𝑖𝑗𝑡 = 𝛿𝑜+ 𝛿𝑖𝑗+ 𝛿𝑡+ 𝛽1ln 𝑌𝑖𝑡+ 𝛽2ln 𝑌𝑗𝑡+ 𝛽3ln 𝑁𝑖𝑡
+ 𝛽4ln 𝑁𝑗𝑡+ 𝛽5𝐶𝐴𝑁𝑖𝑗𝑡+ 𝛽6𝐺𝑆𝑃1𝐸𝑈𝑖𝑗𝑡+ 𝛽7𝐺𝑆𝑃2𝐸𝑈𝑖𝑗𝑡+ 𝛽8𝐺𝑆𝑃3𝐸𝑈𝑖𝑗𝑡 + 𝛽9𝐺𝑆𝑃4𝐸𝑈𝑖𝑗𝑡+ 𝛽10𝑀𝑒𝑟𝑐𝑜𝑠𝑢𝑟𝑖𝑗𝑡+ 𝛽11𝐺𝑆𝑃1𝑈𝑆𝑖𝑗𝑡+ 𝛽12𝐺𝑆𝑃2𝑈𝑆𝑖𝑗𝑡
+ 𝛽13𝐴𝐶𝐸32𝑖𝑗𝑡+ 𝛽14𝐴𝐶𝐸65𝑖𝑗𝑡+ 𝛽15𝑀𝑒𝑥𝑖𝑗𝑡+ 𝛽16𝐶𝑢𝑏𝑎𝑖𝑗𝑡+ 𝛽17𝐺𝑆𝑃𝑇𝑢𝑟𝑘𝑒𝑦𝑖𝑗𝑡 + 𝛽18𝐺𝑆𝑃𝐵𝐾𝑅𝑖𝑗𝑡+ 𝜀𝑖𝑗𝑡
The dependent variable is 𝑋𝑖𝑗𝑡 which denotes the exports, in real terms, from Ecuador (i) to its trade partner (j) in year t. The independent variables could be analyzed in four groups:
First, the intercept of the regression is split in three terms, 𝛿𝑜 is the intercept common to all years and country pairs, 𝛿𝑖𝑗 is the fixed effect for country-pairs added in order to account for heterogeneity and 𝛿𝑡is the year specific effect common to all trading pairs.
Second, the standard gravity variables: 𝑌𝑖𝑡 is Ecuador’s real GDP measured in millions of U.S. dollars in year t, and 𝑌𝑗𝑡 is the trade partner’s real GDP measured in millions of U.S. dollars in year t. Data was initially in nominal terms and was deflated using CPI-U-RS from the Bureau of Labor Statistics.
The use of real terms over nominal terms in both exports and GDP is ingrained in all the gravity model literature (Rose 2000, Bun & Klaassen 2002, Cheng & Wall 2005, Correia 2008). Baldwin and Taglioni (2006) mention that using the US CPI as a deflator is a measurement error that could create biases via spurious correlations due to global trends in the inflation rates. However, they also highlight that including time dummies offsets this error since time dummies can capture the effect of “globalization” as well as other time trends such as variations in the transportation costs.9
𝑁𝑖𝑡 and 𝑁𝑗𝑡 are Ecuador’s population and the partner’s population respectively in year t. Other traditional variables included in gravity models are distance, language and religion, however
9 Results were not significantly sensitive to the definition of GDP and exports. For instance, the model explains about 86.5% of Ecuadorean exports when the real GDP and real exports were used, while with nominal GDP and nominal exports the model explains about 86.6%. Additionally, the significance and signs of the explanatory variables remain the same. To maintain coherence with results presented by authors in Chapter 3 and to compare results, real variables are reported.
32 since these variables do not change over time, they are accounted here via the country-pair fixed effects.
Third, the dummy variables 𝐺𝑆𝑃1𝐸𝑈𝑖𝑗, 𝐺𝑆𝑃2𝐸𝑈𝑖𝑗,𝐺𝑆𝑃3𝐸𝑈𝑖𝑗and 𝐺𝑆𝑃4𝐸𝑈𝑖𝑗 represent the European Union GSP disaggregated in 4 stages, one dummy variable for each stage: (1) Drug combat, (2) Elimination of quantitative restrictions, (3) Further product coverage, (4) Further tariff liberalization: GSP plus. These variables assume the value of 1 if the partner country granted Ecuador, in year t, a certain EU GSP program and 0 otherwise. 𝐶𝐴𝑁𝑖𝑗𝑡 is the variable for the Andean Community which is 1 if Ecuador and the trade partner were part of the Andean Community in year
t.
𝑀𝑒𝑟𝑐𝑜𝑠𝑢𝑟𝑖𝑗, 𝐺𝑆𝑃1𝑈𝑆𝑖𝑗, 𝐺𝑆𝑃2𝑈𝑆𝑖𝑗, 𝐴𝐶𝐸32𝑖𝑗, 𝐴𝐶𝐸65𝑖𝑗, 𝑀𝑒𝑥𝑖𝑗, 𝐶𝑢𝑏𝑎𝑖𝑗, 𝐺𝑆𝑃𝑇𝑢𝑟𝑘𝑒𝑦𝑖𝑗 and 𝐺𝑆𝑃𝐵𝐾𝑅𝑖𝑗 represent the dummies for the other trade agreements: Mercosur, USA GSP disaggregated in 2 stages: ATPA, ATPDEA, Chile (ACE 32, ACE 65), Mexico, Cuba, GSP Turkey, GSP Belarus, Kazakhstan and Russia.
Table 3a presents detailed information on each variable, its definition and source. Table3a. List of variables
Variable Definition Source
𝑋𝑖𝑗 Real exports from Ecuador (i)
to its trade partner (j).
UN Comtrade, measured in millions of U.S. dollars at current prices delated using CPI-U-RS from the Bureau of Labor Statistics.
𝑌𝑖 Real GDP of Ecuador at market prices.
World Bank Database in millions of U.S. dollars at current prices deflated using CPI-U-RS from the Bureau of Labor Statistics.
𝑌𝑗 Real GDP of trade partner at
market prices.
World Bank Database in millions of U.S. dollars at current prices deflated using CPI-U-RS from the Bureau of Labor Statistics.
𝑁𝑖 Population of Ecuador. World Bank Database in millions of inhabitants.
𝑁𝑗 Population of trade partner. World Bank Database in millions of inhabitants.
𝐶𝐴𝑁𝑖𝑗 Dummy variable, takes the value of 1 if Ecuador and its
Ecuadorean Ministry of trade website.
33 trade partner are part of the
Andean Community (from 1993 to 2014 inclusive).
𝐺𝑆𝑃1𝐸𝑈
𝑖𝑗 Dummy variable, takes the value of 1 if the Ecuadorean exports to the trade partner have benefited by the EU GSP (Drug erradication program) during the period 1991-1994 inclusive.
United Nation. (2008). Generalized System of Preferences Handbook on the scheme of the European Community.
𝐺𝑆𝑃2𝐸𝑈
𝑖𝑗 Dummy variable, takes the value of 1 if the Ecuadorean exports to the trade partner have benefited by the EU GSP (Duty free treatment-product sensitivity) during the period 1995-2001 inclusive.10
United Nation. (2008). Generalized System of Preferences Handbook on the scheme of the European Community.
𝐺𝑆𝑃3𝐸𝑈
𝑖𝑗 Dummy variable, takes the value of 1 if the Ecuadorean exports to the trade partner have benefited by the EU GSP (Deeper tariff liberalization) during the period 2002-2004 inclusive. 11
United Nation. (2008). Generalized System of Preferences Handbook on the scheme of the European Community.
𝐺𝑆𝑃4𝐸𝑈
𝑖𝑗 Dummy variable, takes the value of 1 if the Ecuadorean exports to the trade partner have benefited by the EU GSP (GSP Plus) during the period 2005-2014 inclusive.12
United Nation. (2008). Generalized System of Preferences Handbook on the scheme of the European Community.
𝑀𝑒𝑟𝑐𝑜𝑠𝑢𝑟𝑖𝑗 Dummy variable, takes the
value of 1 if Ecuadorean exports benefited from Ecuador becoming an associated country of Mercosur during the period 2004-2014 inclusive.
Ecuadorean Ministry of trade website.
𝐺𝑆𝑃1𝑈𝑆
𝑖𝑗 Dummy variable, takes the value of 1 if the Ecuadorean exports to the trade partner have benefited by the USA GSP (ATPA) during the period 1991-2002 inclusive.
Ecuadorean Ministry of trade website.
When the data is averaged every 4 years, the periods considered for the stages of the EU GSP change: 10 Period for data averaged is 1995-2002 inclusive.
11 Period for data averaged is 2003-2006 inclusive. 12 Period for data averaged is 2007-2014 inclusive.
34
𝐺𝑆𝑃2𝑈𝑆
𝑖𝑗 Dummy variable, takes the value of 1 if the Ecuadorean exports to the trade partner have benefited by the USA GSP (ATPDA) during the period 2003-2012 inclusive.
Ecuadorean Ministry of trade website.
𝐴𝐶𝐸32𝑖𝑗 Dummy variable, takes the
value of 1 if Ecuador had a bilateral trade agreement with Chile in the period 1991-1994 inclusive.
Ecuadorean Ministry of trade website.
𝐴𝐶𝐸65𝑖𝑗 Dummy variable, takes the
value of 1 if Ecuador had a bilateral trade agreement with Chile in the period 1995-2014 inclusive.
Ecuadorean Ministry of trade website.
𝑀𝑒𝑥𝑖𝑗 Dummy variable, takes the value of 1 if Ecuador had a bilateral trade agreement with Mexico in the period 1994-2014 inclusive.
Ecuadorean Ministry of trade website.
𝐶𝑢𝑏𝑎𝑖𝑗 Dummy variable, takes the
value of 1 if Ecuador had a bilateral trade agreement with Cuba in the period 1995-2014 inclusive.
Ecuadorean Ministry of trade website.
𝐺𝑆𝑃𝑇𝑢𝑟𝑘𝑒𝑦𝑖𝑗 Dummy variable, takes the value of 1 if the Ecuadorean exports to the trade partner have benefited by the Turkey GSP during the period 2002-2014.
Ecuadorean Ministry of trade website.
𝐺𝑆𝑃𝐵𝐾𝑅𝑖𝑗 Dummy variable, takes the
value of 1 if the Ecuadorean exports to the trade partner have benefited by the Belarus Kazakhstan Russia GSP during the period 2010-2014.
Ecuadorean Ministry of trade website.
35
4.2 ESTIMATION RESULTS
Table 4a. Regression results with Integration Dummies, Pooled data Dependent variable= Logarithm of Real Exports of Ecuador
Independent MODEL 1 Variables 4 YEAR AVERAGES lngdpEcuador 1.767*** (0.348) lngdpj -0.149 (0.374) CAN 0.541** (0.212) EUGSP2 0.0179 (0.264) EUGSP3 -0.450 (0.329) EUGSP4 -0.286 (0.299) MERCOSUR -0.389 (0.350) USGSP2 0.0449 (0.177) ACE65 0.0960 (0.222) MEXICO -0.690*** (0.249) CUBA 0.387 (0.298) GSPTURKEY 0.729** (0.350) GSPBKR 1.671*** (0.628) Constant -19.68*** (4.612) Observations 253 R-squared 0.908
Country Pair FE YES
Time FE NO
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
36 Table 4a presents the results of the analysis of the determinants of Ecuadorean exports, with data averaged over four years in order to avoid short business cycle fluctuations. The dependent variables is the logarithm of real exports, while the independent variables are the log of real GDP of Ecuador and its country partner, and dummy variables for the various trade agreements. Following the specification of Cheng & Wall (2005), the method used is the LSDV method, since, as mentioned before the country-pair fixed effects control for heterogeneity in the model. However, unlike their specification, population of Ecuador and its trade partner’s as well as time effects were not considered in the model due to collinearity between the time effects and the GDP of Ecuador. This collinearity results since both the GDP of Ecuador and its population are variables that change across time but not across units (partner countries) thus their coefficients cannot be estimated accurately if time effects are added and vice versa. Moreover, in order to maintain consistency with the gravity model specification the time effects and populations were dropped. The dummy variables 𝐺𝑆𝑃1𝐸𝑈𝑖𝑗, 𝐺𝑆𝑃1𝑈𝑆𝑖𝑗 and 𝐴𝐶𝐸32𝑖𝑗 were excluded to avoid the dummy variable trap.
The results of the regression show that the coefficient of the GDP of Ecuador is positive and statistically significant at the 1% level of significance, meaning that taking into account the GDP of Ecuador’s trade partners as well as its trade agreements, a 1% increase in the Ecuadorean GDP is associated with an increase in the Ecuadorean exports by 1.76%. The more than proportional increase on exports highlights the importance of moderately constant output growth within the country as a way to signal the partner countries the reliability of trading with Ecuador.
Surprisingly, the coefficient of the GDP of the partner country is negative and not significant. This result opposes to the findings obtained by the different authors such as Correia (2008) and Mejía (2011), mentioned in Chapter 3. The authors describe that the higher the GDP of the partner country, the more likely is the partner country to import, as the GDP shows the economic health of the country. Given that the coefficient is not significant, further analysis is needed to determine the validity of this contradictory result.
The coefficient of the dummy for the Andean Community is positive and significant at the 5% level. Ecuadorean exports have increased 71.7% (𝑒0.541− 1) due to the Andean Community. This is in line with the results found by Carrillo & Li (2002) and Correia (2005) mentioned in chapter 3. They found that the coefficient of the dummy for the Andean Community had a positive and
