ESSAYS ON THE ECONOMIC INTEGRATION OF THE COOPERATION COUNCIL OF THE ARAB STATES OF THE GULF

Transcription

1 ESSAYS ON THE ECONOMIC INTEGRATION OF THE COOPERATION COUNCIL OF THE ARAB STATES OF THE GULF by Adham Al Said This thesis is presented for the fulfilment of the requirements of the Doctor of Philosophy degree at The University of Western Australia The University of Western Australia Business School Economics 2011

2 ABSTRACT Economic integration is a process through which a group of countries seek cooperation and standardisation in a wide range of policies. These policies affect both domestic and regional economies through the movement of goods, services, and factors of production. In some cases economic integration is of deeper form, such that supernational institutions are developed to achieve its objectives. Economic Integration takes a number of forms such as Free Trade Areas, Customs Union, Common Market, or a Monetary Union. Today the World Trade Organisation (WTO), originally The General Agreement on Tariff and Trade (GATT), dubs these agreements as Regional Trade Agreements (RTA). The post-war trading system has been marked with two conflicting forces, multilateralism and regionalism. The GATT, signed in 1947, formalised the Most Favoured Nation (MFN) approach in international trade according to which countries pass preferential trade policy equally to all trading partners. However, within the multilateral trade liberalisation framework, concessions were given for restricted MFN based agreements. Article XXIV of the original GATT Agreement allowed certain types of trade agreements within a subset of countries that may be discriminatory in nature. These agreements became known as Preferential Trade Agreements or PTAs. The 21 st century has experienced a substantial increase in the number of RTAs. A great number of these RTAs involve more than two countries. This phenomenon has become to be known as New Regionalism. The underlying motives for creating RTAs shifted substantially from the traditional view of trade liberalisation. The implications of these RTAs have become more significant as deeper integration is sought for purposes of development. The importance of RTAs is stressed further where developing countries are becoming more involved in the process. These agreements are part of the international trading system today and they are here to stay. This thesis analyses the rationale for the creation of RTAs and their potential effects on the countries involved. It is concerned with measuring the effectiveness of these RTAs using a barometer made up of three components, trade, incomes and growth, and prices. This is applied to a specific RTA, the Gulf Cooperation Council (GCC); a group of six major oil producing developing countries. Since the 1980s, the GCC countries, located on the Persian Gulf, have been and are undergoing a long-term economic integration process. This project involves creating ii

3 a Free Trade Area, Customs Union, Common Market, and a Monetary Union. After a long period of slow paced economic integration, the GCC has recently invigorated the process. Currently, its progress is at the Common Market stage of integration. The GCC aims to achieve a monetary union within the foreseeable future but faces a number of challenges to the achievement of this goal. The GCC provides a unique economic integration experience and there are many factors that make it a likely candidate for overall success. This thesis provides an economic analysis of the experience thus far. The thesis adopts three instruments to evaluate the GCC s economic integration experience. First, it uses the gravity model of trade. The findings of the trade analysis based on the gravity model indicate that the GCC s intra-regional trade patterns have remained largely unaffected by trade liberalisation between member countries. Intraregional trade patterns are explained sufficiently by economic and geographical variables. Second, the thesis uses neoclassical growth theory models of income convergence. It is found that the GCC countries exhibit some degree of reduced income dispersion. However, RTA effects are not clearly identifiable as the main catalyst. Finally, the Law of One Price approach is applied to investigate the behaviour of disaggregated prices within the region. This indicates that there still exist distortions that create a wedge between cross-country prices, which prevents their equalisation. Although the GCC has advanced through the several stages of economic integration, this thesis finds that it has yet to reach its full potential. While significant progress has been made in the past decade, challenges remain to develop the GCC in an effective RTA. iii

4 TABLE OF CONTENTS ABSTRACT TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES ACKNOWLEDGEMENTS ii iv viii ix xi CHAPTER 1 INTRODUCTION The Context Aims and Objectives Thesis Plan Contributions of the Thesis 8 CHAPTER 2 ECONOMIC INTEGRATION Introduction Regionalism and Trade Implications of RTAs The Gravity Equation Regionalism, Incomes, and Growth Growth Theories Overview Empirics of Growth and Income Convergence Economic Integration and Prices Purchasing Power Parity PPP Empirical Methodologies Concluding Remarks 42 CHAPTER 3 INCEPTION, ACHIEVEMENTS, AND CHALLENGES Historical Background The Formation of the GCC 49 iv

5 3.2.1 Achievements The Soci-Economic Characteristics of the GCC Demography Incomes and Growth Other Economic Dimensions Resource Endowments Monetary Indicators Trade Future Challenges The Country Level Regional Level 75 CHAPTER 4 ECONOMIC INTEGRATION AND INTRA-REGIONAL TRADE Introduction The Gravity Model Approach to Trade Application to World Trade and the GCC Step1: The Traditional Gravity Approach to Determining Total Trade Step 2: Trade within the GCC Countries Data and Empirical Results: The Traditional Model (Step 1) Data Empirical Results Comparison with Other Studies Trade Creation and Trade Diversion Trade within the GCC Countries (Step 2) Conclusions 114 Appendix A4.1 Sensitivity Analysis 117 Appendix A4.2 Data Sources and Description 121 Appendix A4.3 Economic Foundation of the Gravity Model 126 v

6 CHAPTER 5 HOW MUCH CONVERGENCE Introduction Economic Integration Stages of Economic Integration The Gulf Cooperation Council (GCC) Convergence Absolute Convergence Conditional Convergence Dispersion and Convergence The Case of the GCC Other Regional Integration Experiences Deviations from the Mean Approach Summary and Conclusions 158 Appendix A5.1 A Neoclassical Growth Model 160 Appendix A5.2 Sensitivity Analysis 164 Appendix A5.3 Data Descriptions and Sources 168 CHAPTER 6 PRICE CONVERGENCE Introduction The GCC Inflation Experience Inflation within the GCC region Inflation in Other Economic Regions Exchange Rates and Prices Versions of PPP Empirical Methods of testing PPP GCC verses the Group of Seven Price Differentials 184 vi

7 6.4.1 Deviations from the Mean Intra-GCC Price Differentials Prices of Broad Commodity Groups Micro Prices The Data Convergence Price Differentials The Case of Coke Concluding Remarks 206 Appendix A6.1 Data 211 CHAPTER 7 PERSPECTIVES Thesis Review and Major Findings Limitations of the Study Implications of the Findings 220 REFERENCES 223 vii

8 LIST OF TABLES Table 2.1 Gravity Equation Empirical Studies 23 Table 2.2 Selected Growth and Convergence Studies 33 Table 2.3 Selected PPP and Price Convergence Studies 44 Table 3.1 Major GCC Economic Integration Achievements Table 4.1 Bilateral Exports 87 Table 4.2 First Set Of Estimates of the Gravity Equation 89 Table 4.3 Second Set of Estimates of the Gravity Equation 95 Table 4.4 Prior Estimates of Elasticities from Gravity Equations 100 Table 4.5 GCC Trade Matrices, 1995 and Table 4.6 GCC Import Ratios of Total Trade 110 Table 4.7 Disaggregated Trade Estimates 112 Table 4.8 Estimate of Country-Product Dummy Variable Coefficients 113 Table A4.1.1 Estimates of the Gravity Equation 119 Table A4.2.1 Countries Included In Gravity Model Sample 122 Table A4.2.2 Regional Trade Agreements 123 Table A4.2.3 Data Sources 125 Table 5.1 Basic Economic Indicators of the GCC 132 Table 5.2 Absolute Convergence Estimates, Cross-sectional 134 Table 5.3 Absolute Convergence Estimates, Country Groups 135 Table 5.4 Cross-Sectional Conditional Convergence Estimates, Table 5.5 Output Shares of the GCC Countries 146 Table 5.6 Convergence the Deviations from the Mean Approach 157 Table A5.2.1 Convergence linear panel estimates, Table A5.2.2 Cross-Sectional Conditional Convergence Estimates Based On Oil Production 167 Table A5.3.1 Data Descriptions and Sources 168 Table A5.3.2 Countries Included in Regressions 170 Table 6.1 Inflation Differentials, GCC, Mean, Table 6.2 GCC Prices, Table 6.3 Convergence, Disaggregated Price Estimates 202 Table 6.4 Convergence Estimates of Coke Prices 205 Table 6.5 Coke Inter-City Price Convergence Estimates 207 Table 6.6 Summary of Price Convergence Analysis 210 Table A6.1.1 Price Data Details 211 viii

9 LIST OF FIGURES Figure 2.1 Regional Trade Agreements in Force Figure 2.2 World Income Distribution 25 Figure 2.3 Income Convergence, Figure 3.1 The GCC and The Middle East 54 Figure 3.2 Population in 1980 and Figure 3.3 Population Growth Rates 56 Figure 3.4 Fertility 57 Figure 3.5 Population Composition 1980 and Figure 3.6 Real Gross Domestic Product Figure 3.7 Real GDP Growth Rates 61 Figure 3.8 Real GDP Per Capita Figure 3.9 Oil Reserves Figure 3.10 Oil Production 64 Figure 3.11 Gas Reserves Figure 3.12 Gas Production 65 Figure 3.13 Official Nominal Exchange Rates 66 Figure 3.14 Deposit Interest Rates 67 Figure 3.15 Inflation Rates Figure 3.16 Money Supply Growth Figure 3.17 Trade as Percentage of GDP Figure 3.18 GCC Ratio of Internal to External Trade 70 Figure 3.19 Current Account Balance Figure 3.20 Total Foreign Reserves 72 Figure 3.21 Crude Oil Prices 72 Box 3.1 Article 4 GCC Charter 50 Figure 4.1 The Gravity Model Concept 81 Figure 4.2 RTA Dummies Compared 98 Figure 4.3 Transportation Cost Proxies Compared 99 Figure 4.4 Intra-Regional Imports as Proportion of GDP 103 Figure 4.5 Extra-Regional Trade as a Proportion of GDP 104 Figure 4.6 Intra/Extra Regional Imports Ratio 106 Figure 5.1 Absolute Growth Divergence, Figure 5.2 World s Distribution of Income 142 Figure 5.3 Dispersion and Convergence 143 ix

10 Figure 5.4 GCC Average Growth Dispersion 144 Figure 5.5 Relative Importance of GCC Members, 1980 and Figure 5.6 GCC Growth: Weighted Mean and Standard Deviation 147 Figure 5.7 GCC s Real GDP Growth 148 Figure 5.8 Weighted Standard Deviation of Growth 149 Figure 5.9 Growth in EU 15: Weighted Mean 150 Figure 5.10 Growth in Australian States: Weighted Mean 151 Figure 5.11 United States Weighted Mean of Growth 152 Figure 5.12 Weighted Mean of Real GDP Growth 153 Figure 5.13 Income Convergence, Figure 5.14 GCC Income Convergence, Figure 5.15 EU15 Income Convergence, Figure 6.1 GCC Weighted Inflation Figure 6.2 EU15 Weighted Mean of Inflation 174 Figure 6.3 Australian States and Territories Weighted Mean of Inflation 175 Figure 6.4 United States Mean of Inflation 176 Figure 6.5 GCC Exchange Rates 182 Figure 6.6 G7 Countries Exchange Rates 182 Figure 6.7 GCC Inflation Rates 183 Figure 6.8 G7 Inflation Rates 184 Figure 6.9 Inflation Convergence 186 Figure 6.10 Cross Country Inflation Differentials, Figure 6.11 Disaggregated Inflation Rates by Country, Figure 6.12 Disaggregated Inflation Rates by Group, Figure 6.13 Standard Deviation by Group, Figure 6.14 Distribution of Relative Price of Commodity Groups by Sub-Periods 197 Figure 6.15 Coke Price Distribution 204 x

11 ACKNOWLEDGEMENTS This thesis is the culmination of hard work and persistence. This would not have been possible without superb supervision and guidance by my supervisors Professor Kenneth W. Clements and Associate Professor Dr. Abu Siddique. Their insights and guidance were invaluable in getting me past the finish line. I appreciate their continuous support and encouragement throughout my candidature, which drove me to strive for more. I would also like to thank Professor Jeffery Sheen for his invaluable comments as a discussant of my paper presented at the 2007 PhD Conference in Economics and Business at UWA. His insights and comments were very helpful in developing my research. I would also like to thank the UWA Economics Seminar participants for continuous support and constructive feedback on my research. My thanks extend to Professor Ranjan Ray, Professor Srikanta Chatterjee, and Associate Professor Dilip Dutta for taking the time to evaluate and examine my thesis. Their insights, encouragement and suggestions were appreciated and improved the final version of the thesis. My research was facilitated by some generous scholarships and grants. My research would not have been possible in the first instance without the generous scholarship and support from Sultan Qaboos University. I am also grateful to the Australian Research Council, UWA Graduate Research School, and UWA Business School for their generosity and assistance. This thesis would not be as it is if not for their generous support. I would also like to thank my colleagues and friends Mr. Kenneth Yap, Ms. Janine Wong, Ms. Xing Gao, Ms. Mei-Hsiu Chen, and Mr. Tom Simpson. Their continuous encouragement and support have helped greatly during my research. Finally, I would like to acknowledge the incredible support provided by my wife Basma Al Said during the production of this thesis. Not only did she have to tolerate my academic temperament, she endured a postgraduate degree and two pregnancies. I am also thankful to having been blessed with three lovely kids Sara, Tariq, and Laith who made the journey all the more tolerable. xi

12 CHAPTER 1 INTRODUCTION 1.1 The Context The world trading system has undergone significant changes since the late 19 th century. Based on the most-favoured-nation (MFN) approach, countries extended equal treatment of trade concessions to all their trading partners. This approach, which fostered non-discriminatory trade policies that helped reduce tariffs, lost its footing after World War I, when the world trading system became segmented. The MFN approach was revived after World War II through the leadership of the United States. A formal agreement, the General Agreement on Tariffs and Trade (GATT), was signed between the major economies of the time with objective of supporting a multilateral approach to trade liberalisation. Critically, however, Article XXIV of the GATT agreement allowed for specific forms of non-mfn trade agreements, specifying three main conditions for their existence: first, reduction of a substantial portion of trade barriers between the countries involved; second, prevention of increased non-member discrimination; and third, a specified time frame for economic integration, generally accepted as ten years (Frankel 1997). Non-MFN agreements allowed under Article XXIV have generally been in the area of regional economic integration, a process often referred to as regionalism. Economic integration of this type is the central topic of this thesis. Economic integration has gained new vigour in the past two decades with the New Regionalism phenomenon. The international trade scene has experienced substantial increases in the number of bilateral and regional economic integration agreements. These agreements have taken many forms, including Free Trade Agreements (FTA) and Economic Integration Agreements. Although many such agreements, both bilateral and regional, have been concluded between developed and developing countries, a significant number of regional agreements have signed been between developing countries. This has led to a complex, interwoven international trading system dotted with economic integration agreements. The effects and outcomes of bilateral and regional agreements on international trade have been fiercely debated. This is especially relevant given their interactions with the World Trade Organization (WTO) as a multilateral trade liberalisation platform. While it has been suggested by Bhagwati (1991) that a fractured international trading system would put a stumbling block in the way of free trade, opposing views consider 1

13 the positive outcomes of intensified bilateral and regional trade liberalisation would instead become the building blocks of free trade (Ethier 1998). The nature of FTAs has changed dramatically since the late 20 th century. They often take the form of regional economic integration among geographically neighbouring countries. Such agreements seek to develop interdependent regions built around joint economic policies that are not restricted to trade alone; because of this, agreements beyond trade liberalisation are said to incorporate deeper integration. This is one of the main features of the New Regionalism. Economic integration in these instances involves developmental policies that aim for harmonisation of laws and regulations in areas such as financial and capital markets, labour markets, and infrastructure. These arrangements are not strictly restricted to regional agreements however; even bilateral FTAs include reforms to a wide range of policies and legislation. This is especially true where a developed and a developing country strike a partnership (Ethier 1998; Schiff and Winters 1998; Whalley 1998). Deep economic integration is a multifaceted process. The degree of integration varies from one agreement to another. Regional bloc agreements reflect the differences in their objectives and goals. Traditionally, four main stages of economic integration are identifiable: free trade areas, customs unions, common markets, and monetary unions; this final stage usually involves the adoption of a common currency. The effects of these different stages or degrees of integration are profound. At each stage the degree of integration increases, and coherent and joint policies become a necessity. Trade liberalisation, free movement of factors of production, and capital market harmonisation are a few of the steps required to achieve greater economic integration. These initiatives have important effects on the economies of countries involved in such processes. Analysing these effects allows an assessment of the degree of functionality of an economic integration agreement. 1.2 Aims and Objectives This thesis is concerned with the types of economic integraiton discussed above, and its economic effects on the countries involved. The thesis will consider the effects of economic integration among developing countries by analysing a select region: in this case, the Gulf Cooperation Council (GCC). The GCC is a regional bloc formed in 1981 by six Arab Gulf states: Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and the United Arab Emirates (UAE). The GCC is a relatively young economic integration region whose members have followed the economic integration path outlined above, 2

14 and where recent developments have attracted considerable attention. Over the past twenty-five years, the region has launched a free trade area, a customs union, and a common market. It is the aim of the thesis to analyse the effect of these developments on the region s trade, incomes, and prices. The thesis considers this region in its core analysis, but it includes comparisons with other economic regions with comparable experiences. Specifically, the thesis aims to address a number of issues with respect to economic integration by identifying specific effects of economic integration on macroeconomic variables. The objectives of the thesis are outlined in the following: 1. To measure the effects of regional trade agreements on intra-regional trade. FTAs and CUs are stages of economic integration that focus on trade liberalisation. The effectiveness of these forms of economic integration stages can be gauged by analysing trade patterns. Trade patterns within and without regional trade agreement can indicate the effects of integration. 2. To identify patterns of income convergence within economic integration areas. As deeper economic integration affects the movement of factors of production within a region, which in turn is expected to influence incomes within the region, income convergence, in terms of reduced regional differences, becomes an important subject for analysis. Income convergence within the integration area can be a measure of effectiveness of joint policies. 3. To analyse intra-regional price differences within economic integration areas. A consequence of economic integration is the reduction of barriers between member countries. Free movement of goods and persons, and the ability to trade freely within the region, are expected to affect prices across member countries. The conditions are favourable for reduced price differentials across borders as a result of economic integration. This thesis thus focuses on three key indicators of economic integration: trade; income and growth; and prices. Individual chapters are dedicated to investigating in depth these aspects of GCC members economies. 1.3 Thesis Plan This thesis is made up of a number of essays that address the objectives stated above. In Chapter 2 conceptual approaches to economic integration are discussed. The 3

15 chapter considers the development of regional economic integration in the world before moving on to discuss the three conceptual pillars of the thesis: trade, income convergence, and price convergence. The chapter outlines the broad framework of the thesis. The economic integration of the GCC is central to this thesis, and Chapter 3 presents a historical perspective prior to its formation. It offers insights into the motives underlying the creation of the Cooperation Council. The GCC s geographic, social, and economic characteristics are also discussed; as are the goals and achievements of the region s economic integration. Chapter 3 outlines the major goals the GCC hopes to achieve, and where progress had been made. Finally, the chapter discusses the major challenges that face the region at both national and regional levels. The three key aspects of economic integration described above, trade, income and growth convergence, and price convergence, are addressed in Chapters 4, 5, and 6. The first to be considered is trade, as economic integration generally starts at the trade level. This usually takes the form of liberalised trade through Free Trade Areas (FTA) or a Customs Union (CU). FTAs liberalise trade by reducing the tariffs between the countries involved, while each member state is free to set external tariffs vis-à-vis the rest of the world. The tariff reductions are subject to strict rules concerning the origin of products. A FTA is relatively easy to create since there is no obligation for members to adopt a uniform external tariff against the rest of the world; however, the rules ensuring the operation of the FTA make it administratively demanding. CUs, in contrast, are based on liberalised trade between member countries with no internal tariffs. Moreover, CUs unify members external tariffs so that non-members can treat the bloc as a single trading entity. Once a CU is established, trade becomes easy as there are no rules governing the origin of products. Each port of entry into the bloc is identical to the others in terms of customs treatment. Trade liberalisation takes place among a limited number of countries. Third parties may be discriminated against through existing or new barriers to trade introduced by the regional bloc, for that reason proponents of multilateral trade prefer to dub FTAs and CUs as Preferential Trade Agreements. Jacob Viner (1950) identified these distortions to trade as trade creation and diversion. Countries that form an economic integration region in the form of a FTA or CU are likely to shift their imports from non-members to members, since they are preferred trading partners. If imports are now sourced from a higher-cost producer as a result, trade diversion has taken place. 4

16 There are opposing opinions about the effectiveness of trade liberalisation through economic integration. Some consider that the removal of barriers to free trade benefits the countries involved. However, trade agreements are not without their externalities, captured in the concept of trade diversion and creation. The thesis investigates these effects for the GCC in Chapter 4. While the GCC experience is central to this analysis, the chapter also provides comparisons with other Regional Trade Agreements (RTA) experiences. The approach to analysing the effect of economic integration on trade utilises the Gravity Equation. Borrowed from physics, this equation was introduced into international economics almost fifty years ago by Tinbergen (1962). The concept of the gravity equation of trade is that country size matters. Larger countries are likely to trade more with each other than with smaller countries: thus, larger masses (countries) gravitate towards each other through trade. The gravity equation has been used extensively in the literature on international economics and trade and has been successfully linked to existing trade theories, making it a useful tool to analyse trade patterns between countries by considering natural trade flows, which are subject to a number of geographical, social, and economic limitations. Chapter 4 utilises the gravity approach to examine how RTAs as a form of economic integration have affected their members trade patterns. The model controls for natural factors that affect trade, and thus provides a means to isolate potential effects of economic integration. A by-product of this approach is the ability to identify potential trade creation and diversion from the model. Chapter 4 examines this issue specifically for the GCC by decomposing the region s trade into intra- and extraregional trade. The decomposition highlights the effects of the diversion or creation of trade within the region in terms of progress towards economic integration. The thesis is also concerned with the effects that economic integration may have on the income and growth of the countries involved. It is a multifaceted process, so the degree of harmonisation and standardisation in policy may manifest in favourable macroeconomic effects. Chapter 5 aims to analyse these effects within the context of neoclassical growth theory. The fundamental building block of the neoclassical approach is the Solow (1956) growth model. The model has developed over time, and other theories have been proposed to identify the factors that affect growth. Competing theories include endogenous growth models advocated by Lucas (1988) and Romer (1986). Chapter 5 uses the neoclassical models that have incorporated convergence effects, to focus on the experience of countries involved in economic integration based on their incomes. 5

17 Growth models often refer to convergence in terms of β and σ convergence. The former refers to convergence between countries income levels to some given steady state. The latter refers to reduction in the dispersion of the countries income. The Solow growth model is generally used to test absolute β-convergence, based on initial incomes. A regular empirical finding rejects the hypothesis that countries have converged, based on initial incomes. This has led theorists to turn to different approaches, including conditional convergence and endogenous growth theories. Chapter 5 adopts the conditional convergence approach when considering β- convergence. Unlike the basic Solow growth model, the neoclassical conditional convergence model captures growth effects by augmenting the number of characteristics specific to countries. The key characteristic that the chapter aims to evaluate is the regional economic integration in the form of FTAs and CUs. Chapter 5 also considers σ-convergence through the reversion-to-the-mean technique. Ben-David (1993; 1996) applied this technique to the European Union (EU) to assess convergence in a nonparametric methodology. The usefulness of this approach is its application to a small number of countries, as is the case with the GCC. It also allows for the estimation of half-lives of convergence. The third pillar of the analysis is price convergence. In Chapter 6 the effect of economic integration on price is examined. Chapter 6 tests the Law of One Price (LOP) given economic integration. The LOP states that the price of a good in two countries should be equal when converted into a common currency. This proposition has evolved into the concept of Purchasing Power Parity (PPP), in its absolute form, stipulates that the value of a basket of goods in two countries are equalised when compared in a common currency. Since the absolute version of the PPP does not hold, a weaker version allows for differences in price levels, so that foreign and domestic prices are proportional to each other. This weaker version is known as relative PPP. The LOP and PPP propositions have been tested extensively in the literature. There are mixed results where PPP and, by default, LOP do not hold. This is unsurprising since, in the real world, barriers to trade and transportation costs create a wedge between prices in different countries. The relevance of such results in the literature is generally related to exchange rate behaviour. Since the baskets of goods across countries are priced in different currencies, the exchange rate is an important part of the conversion. Implications of PPP not holding are reflected in unpredictable real exchange rates. The Economist magazine introduced a novel approach to PPP in 1986, adopting McDonald s Big Mac hamburger as the uniform good in more than twenty 6

18 countries across the world where the franchise operates using the US dollar price ratio of these burgers. The Economist s Big Mac Index measures currency deviations from PPP. While the Big Mac Index is not a perfect predicator of exchange rate movements, it does provide an insight in exchange rate deviations from their expected long-term values (Ong 2003). Cross border price differentials are affected by many factors, however two stand out with respect to international trade. The first involves trade barriers and the second, transportation costs. As transportation costs are generally correlated with distance between trading, there is little scope for areas of economic integration to affect them directly. Trade barriers, however, are a core target of economic integration. Trade liberalisation, standard harmonisation, and the free movement of factors of production are a few ways that allow for the reduction of price wedges: a FTA, CU, Common Market, or Monetary Union can reduce deviations from PPP and get close to upholding LOP. In theory, the greater the economic integration, the greater the likelihood of approaching LOP. A number of studies have applied this proposition to the United States, Canada, and Europe, measuring the effects of borders, reduced barriers, and even the adoption of a single currency. Such studies look at micro prices to test the proposition of LOP. Chapter 6 also examines this issue within the GCC countries. Has their progress towards economic integration been sufficient for prices to closely reflect those other member countries? Having established both an FTA and a CU, the GCC countries are likely to exhibit only small deviations from LOP. Trade barriers are expected to be lower, and movement of goods within the region to be easier. An added feature of the GCC countries is fixed exchange rates. Unlike the European Union s experience prior to the introduction to the Euro, nominal exchange rates of different GCC countries vis-àvis the dollar have been very stable over time. While not integrated monetarily, the GCC states enjoy an added benefit from fixed exchange rates. Chapter 6 uses these underlying factors of the GCC s status to examine the degree to which prices have converged. The analysis first focuses on changes in price levels to develop an aggregated picture. It then uses disaggregated prices at the micro level to distinguish trends within major spending categories within the GCC. The findings in Chapter 6 indicate that GCC prices converge. However, price equalisation does not take place. Finally, Chapter 7 summarises the findings of the thesis and their implications to the GCC s economic integration. 7

19 1.4 Contributions of the Thesis This thesis contributes to the literature in a number of ways. It takes a broad economic perspective of the GCC experience. Previous research on the GCC has concentrated on the political underpinnings of the region, of individual countries experiences within the region, the energy perspective, and most recently the proposed monetary union. Leading examples for this previous research include Ramazani (1988), Iqbal and Erbas (2004), Fasano and Iqbal(2002) Zaidi (1990), Fasano and Iqbal (2003). This thesis examines the effects of the GCC s achievements in economic integration on its members as a group. This is particularly relevant as the thesis analyses the effects empirically, in relation to the major milestones the region has achieved, such as the customs union and common market. This has not been done before in the depth presented here. Another contribution of this thesis is to offer a barometer of the functionality of economic integration within the GCC. Using three main areas of analysis, the thesis provides insights on how well the integration has worked. Through isolating trade, income, and price convergence effects, the outcome of the analysis provides policy implications. The results provided in this thesis can be used as an assessment tool for the progress of the GCC s economic integration. This tool can also be used to examine other economic integration experiences. 8

20 CHAPTER 2 ECONOMIC INTEGRATION 2.1 Introduction Regional economic integration has taken centre stage in international economies with the rise of Regional Trade Agreements (RTAs) in the second half of the 20 th century. The General Agreement on Trade and Tariff (GATT) in 1947 provided a clause to safeguard multilateral trade agreements from the effects of RTAs. However, although Article XXIV of GATT required RTAs to report their intent, it did not prohibit their creation. RTAs are not restricted to developed countries alone: a great number of them are between developing countries. RTAs between developing countries are dubbed South-South agreements, in contrast to North-North agreements which involve developed countries only. A third arrangement developed during the 20 th century was of North-South RTAs involving both developed and developing countries. This interaction within the framework of regional economic integration introduced a new way of thinking about RTAs and their uses as developmental tools. Regionalism can have many different interpretations, so it is important to distinguish what is meant by regionalism and economic integration in the context of this thesis. Arndt (1993) categorises four types of regionalism often referred to in the literature. These are Preferential Trade Agreements (PTAs), growth triangles, open regionalism, and sub-national regionalism. The first type of regionalism refers to arrangements such as Free Trade Agreements (FTAs), Customs Unions (CUs), or Economic Integration Areas (EIAs). These agreements discriminate against nonmembers in areas of trade preference or other competencies and require substantial governmental involvement to negotiate and bring them into effect. Usually, supernational bodies are responsible for enacting these agreements. Examples of these agreements include the North American Free Trade Agreement (NAFTA), the European Union (EU), and MERCOSUR in South America. The second type of regionalism, growth triangles, occurs due to economic, geographic and cultural ties between sub-national regions. There are no binding agreements or bureaucratic institutions to ensure the operation of this type of regionalism. Regions and states become highly integrated because of proximity. A common example is the growth triangle formed by Singapore, Johor (Malaysia), and Riau (Indonesia). 9

21 Open regionalism forms the third possible type of regionalism. This particular type borrows from PTA-type agreements in terms of having some institutional structure. However, unlike PTAs, open regionalism does not discriminate against non-members; nor does it bind member countries into particular forms of concessions. The core thrust of open regionalism is based on the Most Favoured Nation (MFN) approach to trade liberalisation, and trade barriers are reduced for both members and non-members. The Asia Pacific Economic Cooperation (APEC) is a prime example of such an arrangement (Arndt 1993). Finally, the fourth type of regionalism takes place at the sub-national level. This is often a product of economic planning from a mostly bygone era. However, Arndt (1993) points out a shift towards sub-national regionalism has been observed by researchers in regions such as Europe, South Asia, and North America. This type of regionalism is different from the previous three as it concentrates on a specific region or state that lobbies for greater autonomy or secession. Government involvement may be prominent here just as it is in the PTA case. Which of these four types are relevant to the context of regionalism used here? The first type of regionalism, the PTA, is relevant to this chapter: regionalism which involves government initiatives that aim economically and politically in some cases to integrate their economies. The term regionalism in this chapter is used synonymously with regional economic integration or simply economic integration to refer to a specific PTA between neighbouring countries. The concept of RTAs is used synonymously with PTAs. This chapter will also refer to agreements such as Free Trade Agreements (FTAs) and Customs Unions (CUs) as RTAs. The proliferation of RTAs in their different forms, bilateral and multilateral, has led researchers to question their effects on welfare and economic development. To address these issues, a number of approaches have been developed over the past few decades, including static welfare analysis, simulations, and empirical methodologies. Given the potential effects of RTAs, this chapter is interested in exploring regional economic integration and its theoretical underpinnings. This will be achieved by analysing three major areas where regional economic integration can be shown to have potential economic consequences: trade, growth and incomes, and price equalisation. This will provide the background to the analyses in Chapters 4, 5, and 6 on specific economic issues relevant to regional economic integration. This chapter will take the following structure: Section 2.2 discusses what is meant by regional economic integration and considers the forms it has taken. It 10

22 examines the current trends of RTAs, the welfare implications of the proliferation of regionalism, and its effect on trade. Section 2.3 explores the relationship between RTAs and growth, through the relevant empirical literature that links trade liberalisation generally, and regionalism specifically, to growth. In section 2.4, the price convergence effects of RTAs will be discussed. The PPP theory will be used as an analytical framework in this section. Finally, the chapter will conclude with a summary of the impact of RTAs based on the three areas explored. 2.2 Regionalism and Trade Regional Economic Integration in the form of RTAs refers to the building of cooperative ties between countries with the aim of improving their joint welfare. The nature of RTAs is different to those of Multilateral Trade Negotiations such as those initiated by the General Agreement on Trade and Tariffs (GATT), which was established in 1947 and renamed the World Trade Organisation (WTO) in RTAs are made legally possible by Article XXIV of the GATT agreement and the Enabling Clause of 1979 for developing countries. Customs Unions (CUs) and Free Trade Agreements (FTAs) were not prohibited outright, but were discouraged from pursuing discriminatory practices against non-members. Established and newly created CUs and FTAs were expected to notify GATT/WTO of their arrangements. Over the past fifty years a great number of RTAs have been initiated. The WTO has received more than 400 notifications of RTA formations since 1948; of these, 300 were reported post-1995 (WTO 2010). The WTO has developed a comprehensive RTA database detailing all notifications in various forms. Using this database it is possible to depict the growth in RTA numbers over time. The database distinguishes between agreements based on goods and those based on services, a distinction that needs to be taken into consideration when observing the numbers. Furthermore, the database includes not only trade blocs but also bilateral trade agreements between countries. The RTA forms included are FTAs, CUs, PTAs, Economic Integration Agreements (EIAs), and combinations of FTAs & EIAs, and CUs & EIAs. Figure 2.1 plots the number of RTAs from 1958 to The bars in the chart indicate the total number of RTAs notified to the WTO in a given year. Each bar is divided into the different type of RTA arrangements for that given year. All the RTAs included here are still in force. The black line shows the cumulative number of RTAs in force over time. 11

23 Number of Agreements (Bars) Cumulative Number of Agreemen (Line) Figure 2.1 illustrates an explosion in the number of RTAs created during the past two decades. During the 1960s, 70s and 80s the number of active RTAs was well below 40. This changed dramatically during the 1990s and 2000s, as is shown by the sharp climb of the black line. RTA numbers increased more than fivefold over this period. The composition of the rapid increase is worth noting. In the 1990s the majority of new RTAs were in the form of FTAs. Many of these would have been bilateral trade agreements between countries, or between countries and trade blocs. Economic Integration Areas (EIAs) in three forms standalone, with a FTA, or with a CU made up a small share of the increasing numbers. However, this stopped in the 2000s. While EIAs have not increased in numbers, FTA & EIAs have. 1 These types of RTAs are mostly bilateral country agreements that include both goods and services, but there are also some trade bloc and country agreements in this category. Figure 2.1 Regional Trade Agreements in Force FTA CU PTA FTA & EIA CU & EIA Source: WTO (2010) Figure 2.1 can be divided into two main sections, pre-and post The pre wave of regionalism is often referred to as first regionalism and the post-1990 as new regionalism, a term coined by Bhagwati (1990). Each era presents distinct characteristics. The pre-1990s period was dominated by a strong multilateral push towards trade liberalisation, led primarily by the United States (Bhagwati 2008). The emphasis on industrialisation and import substitution policies during the post WWII period meant that developing countries were less likely to engage in multilateral trade negotiations with vigour. The most influential event that ushered in new regionalism 1 This trend is also observed by Bhagwhati (2008). 12

24 was the shift in US trade policy, to champion multilateral trade liberalisation, marked by the signing of the US Canada Free Trade Agreement in the late 1980s (Frankel 1997). In addition, a stall in the Uruguay Round s negotiations eroded confidence in multilateral trade liberalisation efforts. The shift in sentiment of developing countries from anti-market policies to open trade liberalisation also contributed to the pattern observed in the post-1990 period (Krugman (1991), Ethier (1998), and Bhagwati (2008)). These RTAs, as indicated by Figure 2.1, take a number of forms which reflect the characteristics of the agreements. These differences are pointed out by Whalley (1998) in areas such as coverage of agreements, concessions provided, and depth of economic integration. Whalley (1998) identifies six major reasons why countries might participate in RTAs. First are the traditional gains from trade expected from concessions provided by trading partners. While trade theory supports the expectations of gains from trade, in the case of RTAs welfare implications can counter these, as Viner (1950) shows. Second, RTAs can be used to anchor reforms by ensuring they are binding. A common example is Mexico s accession into the North American Free Trade Agreement (NAFTA). Third, bargaining power in multilateral negotiations is increased. This is a direct response to the hegemony of larger economies in the GATT and now WTO negotiation rounds. Fourth, countries join RTAs to secure access to larger markets. This becomes insurance against market denial to countries not involved in RTAs (Bhagwati 2008). Fifth, RTAs create strategic linkages that go beyond trade: the European integration experience, for example, was based on securing peace in the region post WWII. Finally, RTAs allow interplay between regional and multilateral negotiations. Smaller countries may use RTAs as a vehicle to secure multilateral objectives with larger countries. Whalley s (1998) six points have been augmented by another, which has become evident recently: RTAs assist their members when competing for foreign direct investments (Ethier 1998). Given the variety of reasons why they might want to participate in RTAs, countries choose the types most suited to their needs. These choices reflect certain characteristics of the new wave of RTAs. Ethier (1998) identifies six major characteristics of new regionalism: 1. Small (often developing) countries join larger counterparts to form RTA agreements. Typical examples are Mexico in NAFTA and the smaller states in the European Union. 13

25 2. Smaller countries undergo reform prior to joining RTAs. These reforms are unilateral in nature. 3. RTAs under new regionalism exhibit a modest degree of trade liberalisation, indicating that they are not strictly trade based. In some cases such as the EU, smaller states enjoy market access and low tariffs prior to accession. 4. Asymmetric liberalisation is another characteristic of new regionalism, where smaller partners adhere to their larger counterparts requirements. Examples can be found in the EU s enlargement. 5. Deep integration within new regionalism RTAs is in many cases a dominant feature. Joint policy initiatives beyond trade liberalisation are a recurring theme. 6. Spatial emphasis is apparent in the number of agreements between neighbouring countries. The majority of RTAs reported to the WTO are Free Trade Areas (FTAs). FTAs are the most basic form of trade liberalisation. Country pairs or groups choose to reduce or completely eliminate tariffs between them. However, each country maintains its own tariffs against non-members. To ensure the smooth operation of an FTA, rigorous Rules of Origins (ROOs) must be put in place. These rules ensure non-members cannot take advantage of FTA concessions by using members as an indirect point of entry to target markets. A more intensive arrangement for a RTA is a Custom Union (CU). CUs are similar to FTAs in terms of eliminating trade barriers. However, CU members agree to a unified tariff schedule imposed on non-members. This feature is administratively useful as ROOs are no longer required. Goods entering any port of entry in a member country will be treated the same. Once within the CU, goods can move without further hindrance. These arrangements were the subject of the earlier welfare analysis of Viner (1950). RTAs have also taken other forms, including common markets and Economic Integration Areas (EIAs). These arrangements have increased in numbers, as indicated above in Figure 2.1. EIAs are generally less concerned with trade and more with harmonisation and standardisation across countries. Initiatives include common markets across a number of countries, such as the EU in the 1990s, and the Central American Common Market (CACM) and the Common Market for Eastern and Southern Africa (COMESA). The goal behind these arrangements is to create a single market through which factors of production can freely move. Efficiency gains and economies of scale may by-products of such arrangements. Works and capital can find more productive 14

26 uses if they can move within a greater geographical area. The European experience has been most prominent in this area. Finally, some EIAs take the form of monetary unions. Monetary unions are considered the next major stage after a common market in terms of economic integration. The Economic and Monetary Community of Central Africa and the West African Economic and Monetary Union are direct initiatives of monetary unions. The EU showed a gradual development towards the European Monetary Union (EMU), a movement that spanned more than 40 years. Monetary Unions are a significant development for any EIA. They involve a unified central bank that conducts a unified monetary policy. The Mundell (1961) criteria of adjustment through movement of capital and labour are essential; a synchronised business cycle is also an advantage Implications of RTAs RTAs have been shown to cause a number of distortions in trade patterns and welfare. The first approach to identify distortions was devised by Viner (1950), and adopted by Lipsey (1957) and Meade (1955); it is often referred to as the Vinerian analysis of trade creation and trade diversion. It asserts that RTAs such as customs unions are not necessarily welfare enhancing. Custom unions can create trade by switching import supply to efficient suppliers within the union, on the other hand, they can divert trade from more efficient suppliers outside the union. An alternative approach to RTAs is taken by Kemp and Wan (1976), who show that a customs union can find a Pareto-best outcome such that the welfare of non-members of a customs union is not hurt while the welfare of the members of a customs union is improved. A third approach is led by Cooper and Massell (1965) and Bhagwati (1968). In this approach, the role of economies of scale is considered as means for developing countries to take advantage of regionalism. Although the argument of economies of scale is not crucial for welfare improvements, by adopting Kemp and Wan s (1976) Pareto outcomes, developing countries may find positive impacts of regionalism. Finally, Brecher and Bhagwati (1981) present a mechanism by which welfare effects can be analysed given parametric and policy variations in customs unions. A proponent of regional economic integration, Krugman (1991) points to three major influences that RTAs can create: trade diversion, beggar-thy-neighbour effects, and trade warfare. The first by-product of RTA trade diversion is similar to that identified in the Vinerian approach, but Krugman (1991) states it need not lead to increased protectionism: instead, it may lead to incorrect specialisation based on the 15

27 formation of RTAs. The second condition of a beggar-thy-neighbour policy may take the form of negative effects on non-members. Trade warfare refers to the condition where RTAs form sufficient market power to develop a bullish approach to trade policy and consequently hurt overall welfare. While these conditions can be accepted as natural by-products of many RTAs, Krugman (1991) cites a number of inherent advantages too. He argues that the creation of an RTA reduces the distortion already in place due to existing external tariffs. Consumers incentives distortion will be reduced as a result of a shift towards trade with a member of the RTA. Prior to the creation of an RTA, external tariffs would have restricted consumers to domestic production. Krugman (1991) also points out that there are important size, productive efficiency, and competitive advantages to be gained. He cites the European Community s (EC) experience where the Treaty of Rome, signed in 1957, facilitated substantial increases in intra-industry trade that would not have been possible otherwise. He also points out that another advantage for an RTA region is the improvement of terms of trade against the rest of the world. This last point may be the obverse of the previously mentioned beggar-thy-neighbour effect, however. Krugman (1991) agrees that RTAs such as custom unions will divert trade from non-members; however, he turns to Kemp and Wan (1976) to argue that CUs that reduce external tariffs against non-members may be beneficial for both members and non-members. This however is subject to the RTA s intents and purposes. If self preservation and welfare are more important than the welfare of the rest of the world, RTAs may set their external tariffs above the unilateral levels of its members. However, he argues that RTAs leave individual countries better off than otherwise despite the distortions they cause, as they improve terms of trade for their members. Krugman (1991) presents a simple but interesting model regarding the optimal number of RTAs in the world. He finds the ideal number of RTAs that will reduce negative welfare effects is three. Kemp and Wan (1976) propose that it is possible to create a Customs Union (CU) that is not welfare reducing. Their model is described as a general, effective, simple approach to the issue of CUs welfare effects. They assume a competitive equilibrium of world trade where subsets of countries decide to create a CU. Prior to the creation of a CU there is no restriction on tariff setting or taxes on trade. They propose that a tariff vector and compensatory transfers within the union will result in competitive equilibrium without causing a decline in welfare for members or nonmembers. This is described as freezing trade with the rest of the world, and has the 16

28 effect of mirroring changes in the welfare of the world by observing what happens to members of the CU (Winters 1997). While the Kemp and Wan (1976) model has shed a positive light on CUs specifically and RTAs generally, Winters (1997) emphasises that it cannot be used as a criterion for welfare change. He interprets the model as one which aims to find the price vector which maintains trade composition between the union and the rest of the world. The tariff imposed by the CU is then represented as the difference between the CU price vector and the world. Winters (1997) stresses the importance of understanding which are the welfare relevant indicators, pointing to non-members imports and terms of trade as the measures to observe before and after a CU has been created. RTA agreements can be a product of natural factors. The argument for natural trading blocs is based on the assumption of geographical and economic ties. Countries that are neighbours are more likely to be trading partners than those that are far apart. The argument of natural trade partnership is based on the observation that proximate countries generally become trading partners prior to launching a RTA. Natural trading blocs are less likely to reduce welfare than their unnatural counterparts. Outsiders have the most to lose from these arrangements; but this may be outweighed by the increased benefits to the bloc from trade creation (Krugman 1991). The proposition of natural trade blocs was tested extensively by Frankel et al. (1995) using the gravity model of bilateral trade. Their findings support natural trading blocs in a number of examples. Controlling for a number of economic and geographical variables, they show that in regions such as Latin America natural trading patterns overlap those of actual trade blocs. The likely positive outcomes of RTAs discussed above, are not absolute. Bhagwati et al. (1998) emphasises their negative welfare effects. He primarily discredits the arguments that trade diversion, as described by Viner (1950), is negligible, citing a number of reasons why the argument is flawed. He refers to a number of studies such as those by Yeats (1996) and Wei and Frankel (1996) of trade diversion in MERCOSUR and the European Union respectively, and supports the argument that trade diversion is significant despite relatively low tariff levels. RTAs proponents suggest that lower tariffs levels prior to creating such unions imply lower trade diversion effects. Bhagwati et al. (1998) argues that until the Uruguay Round, barriers to trade remained significant and high. He points out specific barriers to trade, such as anti-dumping sanctions that countries may use selectively, which are influenced by RTA memberships. While emphasising RTAs diversionary effects, Bhagwati et al. (1998) discredits other major 17

29 supporting arguments, including the idea of Natural Trading Partners, and assertions regarding volume of trade and transport-cost. Although proponents of multilateralism point to the diversionary effects of RTAs and welfare, others point out that the welfare analysis applies to the old regionalism. The typical Vinarian analysis is less applicable to Bhagwati s (1991) new regionalism. This is specifically due to the different nature of RTAs in the 1960s and 70s, compared to those of later years. These changes stem from the shift in rationales behind current RTAs which, according to Ethier (1998), include: small countries wishing to create RTAs with larger ones, significant unilateral reforms taking place in small countries, making trade liberalisation a priority, which it may not be in current RTAs, one-sided agreements which help smaller countries liberalise, and finally a desire for deep integration. These characteristics require a different way of thinking about RTAs, and may negate some of the objections cited against them. The above discussion of regionalism and specific concepts of trade creation and trade diversion reveal that the Vinerian approach remains a static approach to regionalism while more dynamic approaches have been introduced to address the changing nature of RTAs and their goals. The following section will present some of the relevant empirical work used to model RTAs effects The Gravity Equation The gravity model of trade has been the traditional econometric tool used by economists to measure the effects of RTAs. First adopted by Tinbergen (1962), it estimated bilateral trade flows based on the distance between the two countries and their economic size, measured by their GDPs. The model was formalised by Anderson (1979) in line with economic theories of trade. His model was a general equilibrium model with differentiated products. Contributions by Helpman and Krugman (1985) produced a gravity model that addressed intra-industry trade where countries relative factor endowments and their labour productivities are similar (Baier et al. 2008). Bergstrand (1989) proved that the gravity equation could be derived from the Hecksher- Ohlin model, establishing its creditability. The gravity model in its simplest form is estimated as: (2.1) log X ij 1 2 log GDPi 3 log GDPj 4 log Distanceij ij. 18

30 On the left hand side X ij is the total exports of country i to country j. The countries size are measured by gross domestic products GDP and GDP. Transportation cost is proxied by using the distance between countries i and j. The s are parameters to be estimated. k i A number of issues have been raised concerning the estimation procedures of this model. Earlier estimates used OLS to estimate the gravity equation. However Egger (2000, 2002) points out that cross-section OLS estimates have the problem of biased estimates due to the omission of variables. He suggests the use of fixed effects in a panel setting. Other studies such as Anderson and van Wincoop s (2003) apply a nonlinear estimation to the gravity equation, incorporating prices of importers and exporters in the model. This helps them address the omitted variable problem. In an extensive use of the gravity model Frankel (1997) and Frankel et al. (1995) incorporate RTAs in the model by adding dummy variables to represent RTA membership. These studies are cross-sectional over several decades using a similar specification to equation (2.1). In the case of Frankel et al. (1995) there is some evidence to support the natural trade bloc argument for some countries in South America, but it does not hold in Europe. These results were subject to criticism, such as Baier et al. (2008), that argues using probit or OLS to detect RTA while using bilateral trade patterns can skew the results, as unobservable effects that are correlated with the trade flows are not represented. Earlier studies apply the gravity model to detect trade creation and trade diversion in the European Community (EEC) after it was first created. Balassa (1967) critiques the method of using average income elasticities of exports and imports derived from cross-section regressions. He points to the difference between developed and developing countries elasticities. The developed countries will have elasticities compared to their developing counterparts. The developed countries will experience greater trade shares of the national products due to specialisation. The effect is opposite in developing countries due to protectionism. Balassa (1967) suggests that this approach will lead to a greater tendency of trade creation findings of the EEC and argues that disaggregation of imports and exports would be more informative. He suggests an appropriate method is to compare income elasticities before and after regional integration, while controlling for income growth. He finds suggestive evidence of trade creation and diversion in a number of sectors, and also finds that trade with nonmembers is affected in various ways. In his assessment, the EEC has led to a basket of mixed results with regard to effects on regional trade. Economic conditions during the late 1950s led to a number of influencing factors in some sectors, while in other sectors 19 j

31 reduction in intra-regional trade was directly due to the EEC. The integration effect cannot be completely isolated, in Balassa s opinion. Likewise, Aitken (1973) uses a version of the gravity equation similar to Equation (2.1). His equation includes both income and population, measures for transportation cost, dummies for EEC, European Free Trade Area (EFTA), and common borders. Aitken uses a cross-sectional OLS for the period , where he expects RTA effects to be registered in the transportation cost measure and common border dummy. An added benefit is the ability to isolate annual effects and observe patterns over the time period. Aitken (1973) asserts this approach allows detection of the RTA effect in the first year. The role of dummy coefficients is central to the estimation of integration effects. The estimated trade flows are compared to actual trade flows to indicate any trade diversion based on integration. Aitken (1973) finds support for the trade effect of the EEC, but not with the EFTA. He goes further to estimate the projected trade for 1967 using 1958 as the transition year, comparing these to dummy variables estimates for trade effects. Using his projections and dummies, Aitken finds in favour of trade diversion towards the end of the period despite gross trade creation effects detected earlier in his sample. These findings of trade creation within the EEC are pointed out in the specific case of the United Kingdom. Sapir (1992) points out that in the years following the UK s accession into the EEC, it experienced a marked increased in intra-eec imports. Generally, Sapir (1992) finds support for an effect of EEC on intra-regional trade. However, he does not attribute it completely to the integration process, arguing that there are other forces at play, including external trade policy. Concentrating on intraindustry trade, Sapir (1992) points to evidence in favour of increasing patterns within the EEC. He attributes this increase in overall intra-industry trade as the influence of late comers like Greece, Portugal, Spain and Italy into the EEC. These countries experienced greater growth in intra-industry trade than the original members. Intraregional trade was also influenced by sectors. Studies that disaggregate the EEC trade shares find both trade creation and diversion. While the European experience is of great interest to researchers, the global scope of regional integration has not been not neglected. Studies of regional integration effects have been undertaken by Frankel (1997), Frankel et al. (1995), and Soloaga and Winters (1999). Frankel et al. (1995) use the gravity equation to determine how much regionalisation has taken place. In doing so they address the issue of natural trading blocs stressed by Krugman (1991). Intra-continental RTAs are less likely to reduce 20

32 welfare compared to inter-continental ones. The argument here is that proximity plays a role in determining the optimal location of RTAs, assuming high transportation costs that prohibit inter-continental trade. Frankel et al. (1995) test this proposition of natural trading blocs using the gravity equation. In their specification they use total trade as the dependent variable for 63 countries for the period 1965 to Their specification uses the product of trading pairs incomes and population in the estimation, instead of individual parameters for each country. They find their estimates are in line with theoretical assumptions about incomes and population and distance. They find richer countries trade more, but at a less than proportional growth rate over time. Another important result is the negative coefficient of distance. In line with theoretical intuition, they find that proximity plays a role in generating more trade. When disaggregating trade based on manufacturing, agriculture, and raw materials, Frankel et al. (1995) find support for the contention that physical cost is not the only concern with respect to distance With respect to RTA effects, the authors find trade bloc effects. These are pronounced in Latin America, in both MERCOSUR and the Andean Pact, compared to NAFTA. The RTA dummies used here become positive in the latter part of the period for these RTAs, a trend not matched in South East Asia s ASEAN or the Pacific s APEC. The core indicators of income, income per capita and distance explain trade patterns sufficiently. Using generic dummies for RTAs, such as FTAs and CUs, returns positive and significant results for the former but not the latter. This is especially true of manufacturing. Cultural and colonial indicators included by Frankel et al. (1995) are also significant. Finally, accounting for factor endowment by absolute difference of GDP per capita or country difference in capital to labour ratios proves less substantial than anticipated. The main conclusion is that countries with similar endowments trade more with each other than otherwise. Soloaga and Winters (1999) utilise the gravity equation for a broader assessment of RTAs. In their case they include nine RTAs, ranging from North-North agreements such as the EU and NAFTA to South-South agreements such as MERCOSUR and the Andean Pact. Soloaga and Winters use X in equation (2.1) as imports into country i ij from country j as their dependent variable. They introduce three distinct dummies to measure the effects of each RTA over time. The first of these is a dummy that captures whether two countries are members of a given RTA. This will measure the overall added trade due to the RTA. The second dummy captures the imports if country i is a member of a given RTA. Soloaga and Winters (1999) describe this as an openness measure of the RTA to imports by its members. The third dummy is assigned to an 21

33 exporting country j that is a member of a RTA. This is intended to represent the openness of the RTA to exports. Traditional trade diversion and creation in this framework are measured by income elasticities. These dummies offer a measure of RTA effect. Soloaga and Winters (1999) estimate their equations based on crosssections and pooled periods for years from 1980 to They control for typical gravity equation variables such as income, population, and distance, and include cultural and geographical variables such as language, common borders and island states. They include fixed effects to address the specification issues Matyas (1997) raised regarding gravity equation estimations, using both fixed effects and RTA dummies discussed above. This is to avoid treating all trading partners symmetrically as Matyas (1997) fixed effect approach would suggest. Soloaga and Winters (1999) find little evidence that RTAs in their new forms have increased intra-regional trade significantly. Elasticities did not change significantly before and after the creation of these RTAs, although not surprisingly they find that the EU and EFTA both produced trade diversion. In Latin America there were some positive signs that RTAs tend to import more from the rest of the world, but Soloaga and Winters (1999) argue that this effect is largely due to unilateral trade liberalisation prior to creation of the RTAs. NAFTA was found to be insignificant in terms of intra-regional trade effects. The authors suggest NAFTA places less emphasis on trade liberalisation than on other integration issues. These studies show that RTAs have considerable effect on intra-regional trade. However, trade diversion effects are also evident, and increased trade due to RTAs is not a feature of all trade blocs. In ASEAN, APEC, and NAFTA, for example, trade flows can be explained mainly by traditional variables of the gravity equation. Western Hemisphere RTAs such as MERCOSUR and the Andean Pact show strong RTA effect on intra-regional trade. However, a possible reason for this is unilateral trade liberalisation, as Soloaga and Winters (1999) explain. Table 2.1 indicates some of the major studies that utilise the gravity equation to test various RTA-related effects on trade. 2.3 Regionalism, Incomes and Growth In the previous section, the trade effects of economic integration in the form of RTAs was discussed. This section will focus on the effects of economic integration on growth and incomes. The aim is to identify if RTAs affect growth and income convergence within regions. As a starting point, the distribution of income across all countries during the period 1980 to 2008 is observed. Figure 2.2 illustrates the 22

34 Tinbergen (1962) Balassa (1967) Table 2.1 Gravity Equation Empirical Studies Study Comments Conclusions Basic gravity model based on incomes and distances between trading partners. Using the model proposed by Tinbergen (1962) in a disaggregated crosssection study of the EEC. Model explains trade flows very well as applied in the Finnish case. Disaggregating trade into a number of sectors shows both trade creation and diversion within the EEC. Sapir (1992) Investigates the intra-regional effects of the EEC. EEC has increased intra-industry trade within the region. Enlargements of the EEC played a role in increasing patterns of intra-industry trade. Consequently some degree of trade creation took place within the Community. Aitken (1973) Anderson (1979) Dummies are used in cross-sectional OLS representing two European RTAs. Formal general equilibrium model incorporating the gravity equation applying the Armington assumption. Confirmation of RTA effect especially in the case of EC. Trade diversion evident. Bergstrand (1985,1989) Derived the gravity equation from the Heckscher-Ohlin model of relative endowments. (continued on next page) 23

35 Table 2.1 (Continued) Gravity Equation Empirical Studies Study Comments Conclusions McCallum (1995) Border effects between the US and Canada using the gravity equation. Large border effects when trade is across borders compared to within each country. Frankel et al. (1995) Frankel (1997) Rose (2000) Used gravity equation to test for natural RTAs. Included dummies for regional and other social and demographic regressors. Using gravity to test for RTA effects while controlling for other variables. Gravity equation used in detecting effects of a currency union on trade. Natural RTAs are detected in South America and Europe. North America s NAFTA, East Asia s ASEAN do not display any natural RTAs due to proximity. Feenstra (2002) Tested for border effects using consistent methods including fixed effects. Border effects calculated in traditional way are larger for smaller countries. This argument is supported by use of simple border effect dummies to detect average effects. Fixed effect estimation returns consistent estimation in the gravity equation. Anderson and van Wincoop (2003) Baier and Bergstrand (2007) Specific treatment of prices in the gravity equation. Introduced multilateral resistance as a key regressor in their model. This was to address the suspected problem of omitted variables. Omitted variable bias from uncaptured effects of deeper integration in the gravity equation. Simultaneity bias due to endogenous variables such as GDP and more specifically the FTA measure. Fixed effects shown to be more efficient than Random Effects or IV. Phase-in effects are taken into consideration. Do FTAs increase members international trade? Yes! It doubles members trade over a period of 10 years as FTAs phase in. 24

36 ,097 1,408 1,808 2,322 2,981 3,828 4,915 6,311 8,103 10,405 13,360 17,154 22,026 28,283 36,316 46,630 59,874 76,880 98, , ,755 Qatar Number of countries Saudi Arabia, Oman Bahrain Kuwait, UAE ,097 1,408 1,808 2,322 2,981 3,828 4,915 6,311 8,103 10,405 13,360 17,154 22,026 28,283 36,316 46,630 59,874 76,880 98, , ,755 Kuwait UAE Qatar Number of countries Oman Bahrain Saudi Arabia distribution of the natural log of GDP per capita (PPP $US) obtained from the IMF s World Economic Outlook 2009 database. The horizontal axis is measured in PPP $US, converted from natural logs. Each bar refers to the upper limit of the range included in the count. The Gulf Cooperation Council (GCC) countries are featured in the figures to illustrate changes over time. Figure 2.2 World Income Distribution (PPP $US) A Income per capita 18 B Source: IMF (2009) Income per capita Panel A in Figure 2.2 shows the histogram of GDP per capita of The histogram includes 147 countries. Countries GDP per capita in PPP $US ranged from 163 in Myanmar to 46,500 in Qatar. The average GDP per capita is 4,136 for all countries. Within the 147 countries included, 101 countries fall below the average and 25

37 46 above the average. There are three distinct peaks within the distribution, and the GCC countries are spread out within the distribution although all GCC countries are above the average GDP per capita. Qatar, Kuwait, and the UAE are by far the richest countries in the world based on the PPP estimates used. Oman is relatively the poorest within the region. Bahrain and Saudi Arabia are well below the top three in terms of GDP per capita. The distribution depicted in Panel B shifts to the right in terms of the average GDP per capita of PPP $US 14,469. The distribution ranges from $US 330 in the Congo to $US 86,000 in Qatar, which is joined by Luxembourg at the top of the list. Although the spread of the distribution has increased dramatically over the 29 years, the number of countries below and above the average has changed very little. There are 98 countries below the average, and 49 countries above. Figure 2.2 also shows an interesting pattern within the GCC. While Qatar, Kuwait and the UAE s GDP per capita remain high, Bahrain, Saudi Arabia and Oman appear to have caught up with their GCC counterparts. This may suggest a degree of convergence between RTA members over time. While the histograms suggest that, overall, countries may not converge in terms of their per capita GDPs, the case of the GCC countries implies otherwise. Figure 2.3 depicts this relationship of growth against initial endowments of incomes in a scatter plot based on the neoclassical convergence concept. Based on the Solow (1956) and Swan (1956) models of growth, countries at different initial levels of income will converge to a given steady state of long run growth. Poorer countries will eventually catch up with their richer counterparts. The initial state is measured by a number of factors, including income as an indicator of economic development. The key catalyst of growth in this context is capital accumulation. Countries with low per capita income, and thus developing, experience greater returns on capital investment: they have much to gain from accumulating capital through saving and investment for production purposes. Their wealthier counterparts with greater capital stocks experience diminished returns. Consequently, the growth rates are expected to be greater for developing countries and relatively slower for developed ones. Based on the above, the relationship between initial conditions and growth is expected to be negative. The relationship is very sensitive to which countries are included. In the sample of 147 countries, Zimbabwe is the poorest based on PPP valuations of GDP per capita. When included, no convergence can be detected. However when removed, the relationship between growth rate and incomes in 1980 is weakly negative. Very little evidence of convergence can be detected from this simple 26

38 Growth (% p.a.) exercise. This is not surprising as there many reasons why poorer countries have not converged with richer ones. In a specific case, Figure 2.3 isolates six GCC countries. In a simple convergence exercise the six countries exhibit signs of the catch-up scenario suggested by neoclassical growth theory. 14 Figure 2.3 Income Convergence, Oman Bahrain y = x Saudi Arabia Kuwait UAE Qatar y = x Log of Per Capita GDP 1980 Source: IMF (2009) Growth Theories Overview Growth theory attempts to explain the reasons why on a global scale countries have not converged following the assumptions put forward in the neoclassical models led by Solow (1956) and Swan (1956). The premise of the Solow model of growth was built around capital accumulation and diminishing returns. This allows countries to reach a given steady state. Growth is the by-product of population growth and technological growth. Figure 2.3 depicts this relationship to some extent by controlling for initial incomes as proxy for country differentials. Poorer countries with lower capital stock will experience faster growth rates as a result of greater marginal productivity. This has clearly not been the case for the world in the past three decades, and this inadequacy has led researchers to find explanations of what drives growth. Abandoning the exogenous assumptions of the neoclassical theorists, Romer (1986) and Lucas (1988) have led work on new growth theories, dubbed Endogenous Growth Models. These models emphasise the role of human and knowledge capital. The significance of endogenous human and knowledge capital in growth models was their lack of diminishing returns. Unlike physical capital, which experiences diminishing returns, the 27

39 spillovers of human and knowledge capital allow long-run positive growth of incomes (Barro 2004) Empirics of Growth and Income Convergence The empirical literature on growth is large, and has taken a number of directions based on neoclassical and endogenous growth theories. Explanations of the determinants of growth have taken many forms and included many variables. Two main areas can be identified as the most important issues: winners and losers in long term growth, and the variability in growth rates over time for specific groups of countries (Rogers 2003). These issues boil down to a specific concept, convergence. Are countries approaching a common per capita income level? As indicated in the previous section through the distribution of countries across the world and the scatter plot, there is no evidence of such behaviour. A number of studies aim to address the issue of convergence by examining distributions and predicting outcomes. Quah s (1996) and Pritchett s (1997) are examples of such studies, which gave rise to the concept of a convergence club, where countries would converge in separate groups. These results are fragile, and sensitive to which countries are selected (Rogers 2003). In a different direction, empirical studies have concentrated on regional and cross-country growth to detect convergence. These studies include Barro et al (1991), Barro (1991), Barro and Sala-i-Martin (1992), and Sala-i-Martin (1996). These studies generally use crosssectional samples of countries and test convergence based on neoclassical assumptions. In the case of Barro et al. (1991), regional sets of US states, Japanese prefectures, and European regions are used as units of analysis. A number of outcomes of these regressions have surfaced as stylised facts. These include: there are no simple determinants of growth from the extensive list used by these studies; initial income still plays an important role with respect to growth; government size is less of an issue than quality governance; human capital remains weakly related to growth while health and wellbeing indicators are generally found to be robust; institutions play an important part in growth outcomes; and finally open economies generally grow at a faster rate than others (Sala-i-Martin 2002). Levine and Renelt (1992) have put many of the variables used in the cross-country studies to the test and find that initial incomes, under conditional convergence, have a robust relationship with growth. Other traditional explanatory variables such as political and economic variables are not robust, however. They also find that investment plays an important role in growth: investment shares of 28

40 GDP and trade shares of GDP have a positive and robust relationship. However, investment shares of GDP affect the robustness of trade policy variables. Above are some of the studies that have considered growth empirically and tested the relationship between growth and multitude of variables. The emphasis from this point onwards is the relationship between trade generally, and RTA specifically, on growth. Thus, posing the question, what effect does trade have on growth? Lewer and Van den Berg (2003) review the empirical literature that has tested the relationship between trade and economic growth. Their interest is in the studies that specifically use trade measured by export growth and income growth. They break down the studies into four major categories: cross-section studies, time-series studies, per capita income studies, and simultaneous equations studies. The first two categories are studies that regress growth of capital, labour stocks and other explanatory variables, and finally growth of trade. These regressions generally are linear, and have their theoretical backing from neoclassical growth models. Lewer and Van den Berg (2003) point to a remarkable consistency in the results of studies that can be classified as cross-section studies and time-series studies specifically, and other studies in general. In the case of cross-sectional studies, based on a survey of 196 studies they report an average coefficient value of This is interpreted as meaning that a 1% increase in trade, measured in many cases by growth in exports, leads to one fifth of a percent growth in output. This result is not dissimilar to time-series regressions. Lewer and Van den Berg (2003) report an average coefficient of This, however, includes all the studies that have adjusted for stationarity and those that have not. When these effects are taken into account the coefficients differ substantially: studies that do not adjust for the unit root problem report a coefficient of on average, while those that do report a coefficient of on average. The time-series trade effect is affected by whichever subset of studies is considered; those based on classifications of income, on openness, or on single and multiple regressions. In the case of single or multiple regressions based on time-series data, the results are very similar: coefficients on average are and respectively. A number of criticisms have been made against econometric studies of growth. The effect of the trade variable can be used to capture income effects that are not entirely related to that variable. Of these studies key Frankel and Romer s (1999) and Rodriguez and Rodrik s (2000) stand out (Noguer and Siscart 2005). The emphasis in this case is on non-trade effects due to geographical variables. The implications of the geographical characteristics of countries can be summarised as three main effects: 29

41 diseases and morbidity due to the environment, endowments of resources and agricultural productivity, and finally institutions (Noguer and Siscart 2005). The inclusion of robust instruments that measure the effect of non-trade factors that may have been captured by trade measures will yield precise estimates. Noguer and Siscart (2005) propose an instrument that they assert to be precise compared those used in previous studies. They point to the criticism made by Rodriguez and Rodrik (2000) that geographical instruments used in growth empirical studies may be correlated with other geographical variables which may affect income through non-trade channels. They Rodriguez and Rodrik point out that those regressions may be spurious. Based on this, Noguer and Siscart (2005) construct a geographical characteristic instrument using a gravity equation variant. They find inclusion of this improved instrument to reduce the effect of volume of trade on incomes. Using the Instrumental Variable estimation, they find a one-to-one relationship between trade share and income growth: thus, countries which experience an increase of trade share of GDP by 1% experience a 1% increase in income per capita. This result verifies the findings of Frankel and Romer (1999), that trade does affect income. Rodriguez and Rodrik (2000) present a case of trade policy implications on growth. They differentiate themselves by testing how trade policy affects growth. They critique a number of empirical studies that link openness to growth to trade policy implications. These studies include Dollar (1992), Sachs and Warner (1995), Edwards (1998), and Frankel and Romer (1999) as main contributions to the literature on trade and growth. Rodriguez and Rodrik (2000) argue that in the case of Dollar (1992) there are problems with the indices created to detect real exchange distortions and variability. The critique is based on the fact that each of these indices measures the trade orientation of countries used in the cross-sectional study. In the case of real exchange distortions, the index is formed of the prices (based on Summer and Heston s 1988 database) of each country relative to the United States. Rodriguez and Rodrik (2000) point out three conditions that must be ignored for Dollar s (1992) index to measure trade orientation correctly: export taxes or subsidies, the Law of One Price, and transportation costs and geographical factors. They show that ignoring these conditions produces ambiguous effects on the index and consequently on its perceived interpretation. The other critique of Dollar s approach to growth empirics is the second index of real exchange variability. This captures relative price variation over time. In this case Rodriguez and Rodrik (2000) argue that the measured effect is influenced by factors other than trade policies, and show that the distortion index is not robust, unlike the variability index which 30

42 withstands alternative specifications of the growth equation used by Dollar (1992). The insignificance of the distortion index is due to its correlation with other factors such as omitted geographical variables as Rodriquez and Rodrik (2000) suggest. Sachs and Warner (1995) develop an openness index using five criteria: tariffs exceed 40%, non-tariff barriers cover more than 40% of imports on average, a socialist economic system exists, major exports are controlled by a state monopoly, and the black market premium exceeds 20%. These criteria are used to classify countries as open or not. Rodriguez and Rodrik (2000) argue that this index of openness is misleading in terms of trade policy, and show that of the five criteria only two contribute significantly to growth. The trade-specific criteria do not do not affect growth when tested individually. Creating sub-categories of the Sachs and Warner (1995) openness measure, they find only the black market premium and state monopoly of exports are statistically significant with coefficients close to those of the overall index. They ask, since these two indicators of openness are significant, are they related directly or indirectly to trade policy? Sachs and Warner (1995) argue that in the case of a state based export monopoly, distortions are caused such that the overall trade of the country will decline. In creating this measure they source the World Bank s study on African countries in This choice restricts the index used for these criteria to African countries that underwent structural reform programs during the period of study. Rodriguez and Rodrik (2000) argue this causes two selection bias problems: first, the omission of non-african countries that also restricted trade through monopoly; and second, the omission of African countries with restrictive trade policies that did not fall under the structural programs. This leads to an upward bias of the monopoly criteria used by Sachs and Warner (1995). In effect Rodriguez and Rodrik (2000) argue that the monopoly measure correlates with that of the sub-saharan Africa dummy used by Sachs and Warner, and the results are influenced by non-trade policy related factors. The critique of the black market criterion follows similar concerns about its shortcomings as an indication of trade policy relevance. The argument here is that black market premiums need not reflect trade policy distortion alone. If foreign currency rationing has led to deviations between official and market exchange rates, importers may be led to black market transactions. The behaviour of importers and exporters will have consequences for resource allocation, as Rodriguez and Rodrik (2000) point out. If both importers and exporters use the black market to source their foreign currency, such 2 Rodriguez and Rodrik (2000) point to the World Bank study Adjustment in Africa: Reforms, Results, and the Road Ahead

43 consequences will be alleviated. The critical issue they point to in Sachs and Warner (1995) is that the black market premium points to policy distortions, corruption, or economic mismanagement, and not to trade policies. They confirm this by using the same black market premium as a continuous variable against growth, where they find the 20% criterion identified by Sachs and Warner (1995) weighs heavily on those countries that fall in that category. As the two significant components of the openness criteria of Sachs and Warner (1995) are shown to be less related to trade policies and more to imbalances and other factors, the effects of openness of 2.20 to 2.45 in Sachs and Warner s (1995) regressions may be due to other factors overall, and not trade policies as Rodriguez and Rodrik (2000) point out. In an extensive review of regional trade agreement and development Schiff and Winters (2003) present a number of links between RTAs and economic growth. The premise is that trade policy in the form of preferential treatment agreements may lead to some economic growth. The contrary argument is that multilateral trade liberalisation may produce economic growth. Three main factors identified in the literature influence this potential relationship: openness, Foreign Direct Investment (FDI), and knowledge transfer. These are the channels through which trade can influence economic growth. The choice of RTA partners plays an important role in how such effects take place; for instance, it is less likely that a South-South RTA will benefit from spillovers associated with knowledge and FDI, compared to a North-South RTA. In an RTA-specific study, Ben-David (1993) finds that trade liberalisation within the European Economic Community (EEC) and European Free Trade Area (EFTA) has convergence effects in terms of incomes within Europe. He finds a link between trade and income convergence when considering major trading partners. In this approach, Ben-David (1996) does not define RTAs specifically but focuses on country groupings based on trade patterns. These results imply that if RTAs sufficiently increase their member s trade, there is a likelihood of income convergence among them. Ben- David (1996) links the results to the Hecksher-Ohlin factor equalisation, with trade as the channel through which income convergence takes place. This sub-section has outlined some of the key studies that related trade to growth, and by proxy income convergence. These studies are summarised in Table Economic Integration and Prices The main aim in this section is to identify the effects of regional integration on prices. Trade liberalisation has a potential indirect effect on prices. Reduction of trade 32

44 Table 2.2 Selected Growth and Convergence Studies Study Comments Key Results Barro (1991) Barro et al. (1991) Examines economic growth across a cross-section of countries based on neoclassical growth models. Key variables include initial incomes, human capital measured by schooling, investment, and political stability. Examines income or β-convergence within countries by comparing incomes across US states in one instance and across European regions in another. Simple correlation between income and growth is weak. However, when human capital is accounted for, the link is negative and significant. Human capital is one of the key variables for poorer countries catching up with richer ones. Government spending and investments can be distorting. Political instability is negatively related to growth. Sub-Saharan Africa and Latin America s weak growth are left unexplained by this framework. Poorer states or regions grow faster than relatively richer ones. This is dubbed β-convergence and is found to be around 2% per annum. In Europe the cross-country regional comparison yields similar results to that of US states. There are no significant differences when comparisons are made within countries and across countries. (Continued on next page) 33

45 Table 2.2 (Continued) Selected Growth and Convergence Studies Study Comments Key Results Barro and Sala-i-Martin (1992) Ben-David (1993 and 1996) Tests for convergence using the neoclassical growth model. This is applied to US states and comparably to other countries. Relative steady states are captured by individual characteristics of each state or country. (1993) Links between timing of trade liberalisation and income convergence is examined. The European Economic Community (EEC) is used as a case study. This is done by observing the dispersion of the six original members and then the inclusion of new members. (1996) Examines income convergence within a group of countries. Trade s effect on income convergence is the main focus. 34 Convergence is supported in the form of faster growth towards steady state when farthest away. Growth rates are faster for poorer states in the US than for their richer counterparts, when only initial incomes are considered. When specific characteristics are included the rates are similar at 2% a year. Convergence between countries is evident in the conditional form. Openness, technological diffusion, and capital mobility also affect rates of convergence. (1993) Trade liberalisation is shown to have an effect on income convergence. This is observable when new members join the EEC. Income convergence patterns are not related to long term trends or previous divergences. (1996) Evidence of convergence with specific group of countries that trade substantially with each other. This result is emphasised when countries are grouped based on their trade policies. (Continued on next page)

46 Table 2.2 (Continued) Selected Growth and Convergence Studies Study Comments Key Results Quah (1996) Regularity of conventional empirical finding of cross-sectional convergence due to statistical regularity and not actual convergence. Paper studies dynamic distribution changes of country incomes. Sala-i-Martin (1996) Uses the classical approach to convergence to test the concepts of β-convergence and σ- convergence for a set of 110 countries. Study also includes country datasets such as US states, Japanese prefectures, and European regions. Paper indicates that conditional convergence makes the neoclassical approach plausible within the convergence debate. Support for convergence clubs found. Income distribution bimodal. In the classical approach absolute β-convergence and σ-convergence are not detected in the crosscountry dataset. There is support for conditional convergence, however. The rate of convergence is approximately 2% per annum. Absolute β-convergence and some σ-convergence are observed for OECD countries during subperiods. The rate of convergence is close to 2% per annum This result is also verified within country states and regions. Both σ- convergence and β-convergence are observed. The rate of convergence is also close to 2% per annum. (Continued on next page) 35

47 Table 2.2 (Continued) Selected Growth and Convergence Studies Study Comments Key Results Frankel and Romer (1999) Paper examines the relationship between trade and income by devising geographic instrumental variables. Justification of use of such instrumental variables is to avoid endogeneity of trade effects in incomes. Geography explains part of the trade and income relationship but not all. A positive effect of trade on income is found. A 1% change in trade to GDP leads to 0.5% change in per capita GDP. Trade within country was found to affect incomes as well. As countries get larger there is a small but positive effect on income. 36

48 barriers may reduce some of the distortions to cross-country prices of goods. However, cross-country prices are influenced by a number of factors besides barriers to trade. Transportation costs and exchange rates also play a significant role in cross-country price deviations. While RTAs cannot do much about the former, there is body of literature that suggests that through currency unions they may have an effect on trade and growth; one such is Rose (2000) Purchasing Power Parity An appropriate starting point for the discussion of regional integration and prices is purchasing power parity. The PPP theory asserts that prices across borders will be equalised barring any impediments to trade or transportation costs. Proposed formally by Cassel (1928), PPP theory came to be perceived as a powerful tool for exchange rate forecasting. The practicality of the theory was supported by proponents such as Keynes (1923), but also received criticism from opponents like Viner (1933). While not everyone agreed on the practical usefulness of PPP theory, many studies of the US and European countries were conducted during the 1920s and later revived post-world War II. The PPP doctrine received renewed impetus in the 1970s with the monetary approach led by Mundell (1968, 1971) which links the Quantity Theory of Money to exchange rates (Dornbusch 1988). 3 A major issue in the PPP doctrine is the deviations of exchange rates that seem to dispute the theory. While evidence has been found for and against PPP, explanations for the deviations are required. Balassa (1964) and Samuelson (1964) provide one explanation by pointing out that PPP deviations are subject to productivity bias. Dornbusch (1988) identifies the deviations from PPP as structural and transient. The structural deviations are caused by relative productivities between countries; the transient deviations are caused by market imperfections. PPP deviations are also influenced by so called border effects, such as barriers to trade, transportation costs, marketing costs, exchange rate variability, and local currency prices (Engel and Rogers 2001). These distortions are a key reason why the Law of One price or absolute PPP is based on the assertion that the price of a commodity is equalised when compared in the same currency. Therefore, (2.2) P SP i i 3 See Clements (1981) for exchange rate determination using the monetary approach. 37

49 where the left hand side is the domestic commodity price P i in the local currency, and the right hand side is the foreign commodity price P i, and S is the spot exchange rate. The same concept can be applied to price levels across countries. In absolute terms, the previous expression becomes P SP such that the price of a basket of goods in both countries is compared in the same currency. From this expression, the exchange rate is the ratio of the values of the two baskets. Therefore, (2.3) S P P. Taking the logs of (2.3) gives the expression s p p. Thus the log exchange rates are subject to the log difference of the value of the baskets. The assumption behind the expression is that the baskets of goods are comparable between the domestic and foreign countries. These consumer baskets are represented by their national Consumer Price Index (CPI). The use of CPIs implies that the consumer baskets are not likely to be identical in composition or weight. This is critical for absolute PPP, which applies to identical goods. A weaker version of PPP uses CPI changes instead of absolute levels. The relative PPP relationship is usually presented as a difference of logs of (2.3) to get (2.4) s p p. The expression here refers to price changes or inflation rates and exchange rate changes. The change in the real exchange rate is the excess of the nominal change over the inflation differential. Thus, the change in the real exchange rate is the deviation from parity, vis-à-vis, s p p, which we write as (2.5) r s p p. PPP theory has been used as a tool to determine exchange rates. These approaches are discussed in the following section PPP Empirical Methodologies The concept of PPP was empirically tested for both short and long periods. The stricter interpretations of PPP theory, that this relationship holds true all the time, has been tested numerous times; an often cited study is Frenkel (1978). This and other studies that followed post-bretton Woods are referred to as First Generation PPP studies (Froot and Rogoff 1995). The proposition tested by Frenkel (1978) was based on the inflation differential such that (2.6) s p p t t t t 38

50 where s is the exchange rate, t pt p t is the inflation differential between the home and foreign country, and are parameters to be estimated, and is a random error. t Frenkel (1978) tests for the condition where 1 to accept PPP. Thus inflation differentials affect the exchange rate directly, as suggested by equations (2.4) and (2.5). He finds support for PPP based on (2.6). His results were contested, however, because countries in his sample which experienced hyperinflation in were omitted; nor could the results be replicated for OECD countries. Furthermore, the causality of exchange rate effects on inflation differential was questioned by Isard (1977) amongst others (Froot and Rogoff 1995). Krugman (1978) shows that the endogeneity problem presented by Equation (2.6) can cause bias in OLS estimates of. He proposes using Instrumental Variables (IV) on Equation (2.6), a similar approach to that of Frankel (1981). They both reject PPP based on the IV approach. The main feature of this generation of PPP empirical tests was the short term horizon. The second generation of PPP tests emphasised longer horizons. These studies tested for stationarity of the real exchange rate. The real exchange rate was defined as (2.7) q s p p. t t t t Equation (2.7) equates the real exchange rate changes to changes in the nominal exchange rate and inflation differentials. Using Equation (2.7), studies in the so-called second generation tested the null of random walk. Testing for random walk utilised the developments in time-series produced by Dickey and Fuller (1979). What came to be known as the Augmented Dickey-Fuller test for stationarity is presented as q t q L q t t 1 t 1 t. (2.8) Equation (2.8) regresses the real exchange rate obtained from Equation (2.7) against its lag and combination of differenced lags. The regression also includes a trend when deemed necessary. The null hypothesis of 3 1 is tested to confirm the presence of a unit root. If the null is rejected then the real exchange rate is a mean reverting random walk process, and PPP holds. The results of the second generation were mixed. Long run tests have shown that it is more likely to accept the random walk hypothesis with respect to the real exchange rate. Criticisms of the power of the test, based on data availability, often cast doubt over the findings (Froot and Rogoff 1995). The methodology used in time-series evolved into error correction models and structural break tests which were used to distinguish 39

51 possible influences on the data. Cointegration tests introduced by Engle and Granger (1987) defining stationarity can be weakly proven by mean reversion of a linear combination of variables. In this case a linear combination of exchange rates and prices needs to be stationary (Froot and Rogoff 1995). The general empirical findings suggest that PPP generally and LOP specifically do not hold in the short run and are questionable in the long run (Taylor and Taylor 2004). Two major concerns of these tests generally, and of mean reversion specifically, are raised by Taylor (2003). The first is temporal aggregation, where studies have used annual, quarterly, or monthly data to test for mean reversion in a basic first-order autoregressive specification. The half-life estimates are often slow, as indicated above. This, Taylor (2003) points out, defies expectations of researchers of speedy recovery from deviations from the mean. The problem of aggregation rises from using low frequency data to infer adjustments in high frequency data. Taylor (2003) indicates that this causes bias towards longer half-life estimates. The second problem is attributable to linear specification of the mean reversion regressions. The assumption is that PPP deviations will die out in a linear fashion no matter how large they are. Taylor (2003) points out that there are conditions such as fixed and variable trading costs, and risk aversion which fall under a band of inaction where no arbitrage can occur. He refers to Heckscher s commodity points as the originating concept of this idea. In other words, Taylor (2003) points to the possibility of nonlinearities dissipation of deviations from the mean. Rogoff (1996) presents an important conundrum with respect to the empirical PPP literature: How can one reconcile the enormous short-term volatility of real exchange rates with the extremely slow rate at which shocks appear to damp out? (Rogoff 1996, p. 647). This he dubbed the purchasing power parity puzzle. The slow rate of convergence is indicated in the literature by half-life estimates of 3 5 years found using different empirical methods. Rogoff (1996) puts forth a number of arguments why this empirical predicament exists. He refers to the Balassa-Samuelson effect as an extension to PPP models that can help explain this pattern. Another plausible extension to existing PPP models is the connection between current accounts and the real exchange rate in the long run. Rogoff (1996) also suggests incorporating government spending linkages to real exchange rates as a means to understand the cause of long half-lives found in the literature. Finally, there is a multi-vector autoregression model approach. These models have given some promise of incorporating a number of economic explanatory variables that interact with the real exchange rate incorporating monetary shocks bounds. Despite all these possible approaches Rogoff (1996) sees 40

52 persistent and slow converging deviations of PPP as a direct result of international markets remaining segmented and trade being plagued with friction. These frictions include transportation costs and barriers to trade. While many studies have tested relative PPP, others have concentrated on the LOP or absolute PPP. Disaggregated or micro-prices used to test the LOP appeared as early as Isard (1977), and Giovannini (1988). Isard (1977) uses the disaggregated prices of manufacturing goods in three countries, the US, Japan, and Germany, and finds that price differences from the LOP persist for a long time. Giovannini (1988) finds similar results using disaggregated price comparisons. These studies also find that deviations are consistent with exchange rate movements (Froot and Rogoff 1995; Rogoff 1996). The most interesting and well known LOP specifically and the PPP in general is the Big Mac Index. Introduced in 1986 by The Economist magazine, the index used the prices of the McDonald s Big Mac burger as a representative of an identical basket of good across countries. The value of the Big Mac was compared across countries once converted in US dollars. The relative dollar prices of Big Macs indicated over or under valuation (Ong 2003). Parsley and Wei (2008) provide a useful comparison between the Big Mac Index verses the Consumer Price Index as a unit for LOP studies. Other studies chose to examine the LOP in the context of economic integration. These include Parsley and Wei (2008 and 1996), Bergin and Glick (2007), Rogers (2007), and Broda and Weinstein (2008). The unique aspect of the Parsley and Wei (1996) study is that they concentrate on the US only. In testing for the LOP the authors eliminate the common frictions referred to above by observing price differences in one country. Their data are divided into three categories: perishables, non-perishables, and services. They use prices relative to a chosen city within the US using the unit root test proposed by Dickey and Fuller (1979) and using distance as the measure of transportation cost. They find that in the case of non-tradable products, reversion to zero is present. They also show that services take longer to damp out differences in inter-city prices. Parsley and Wei (1996) find support for non-linearity of convergence especially in tradables. Rogers (2007) examines price convergence in Europe and the US after the launch of the monetary union. He makes use of the Economist Intelligence Unit to find the dispersion within the EU prior to the creation of the monetary union. The reduction in price dispersion after the launch of the union was minimal in comparison. Rogers (2007) reports the largest decline was experienced by the EU-11 countries that adopted the Euro. The US in comparison has shown lower price dispersion over time, especially 41

53 in non-tradables. The key findings of Rogers (2007) are explained by factors of economic integration. These include harmonisation of tax rates, convergence of incomes and labour costs, and trade and factor market liberalisation. These findings and other major studies are summarised in Table Concluding Remarks The phenomenon of the sharp increase in RTAs in the past two decades has placed emphasis on understanding the outcomes of these agreements. This chapter has presented the conceptual overview of economic integration by defining what is meant by economic integration, and has identified the main characteristics and underlying reasons for their formation. The chapter also discussed the three aspects of economic integration that act as a barometer of effectiveness, identified as trade convergence, income and growth convergence, and finally price convergence. The chapter discussed the effects of economic integration on each, citing evidence from the literature that links these aspects to economic integration. With respect to trade, reduction in trade barriers due to liberalisation of movement of goods within a given RTA has been cited as a catalyst of increased intra-regional trade. However, this effect is countered by potential diversion of trade from the rest of the world. The second aspect of economic integration effects considered here is income convergence. Using trade as a vehicle of growth, and considering the link between RTAs and increased trade, incomes within a region may be expected to converge due to economic integration. There are other potential channels by which trade, intensified by RTAs, may affect incomes, such as through FDIs and knowledge spillovers. This is expected to be an important channel when developing countries partner with developed countries to form RTAs. RTAs can also influence income convergence through free movement of factors of production with an economic integration area, bringing some balance to intra-regional differences. The third aspect of economic integration discussed in this chapter is price convergence. Linked to trade liberalisation and greater economic integration, price convergence is a significant measure of the effectiveness of RTAs. In the first instance of trade liberalisation, reduction in tariffs can be linked to reduction in price differences. Reduced non-tariff barriers, through harmonisation and standardisation of laws and regulations, also can play a role in bringing down price deviations between RTA members. Surprisingly, even with significant economic integration there exist price 42

54 wedges that may be attributable to market segmentation or non-tariff barriers. These findings come from studies that use highly disaggregated price data. The three aspects of economic integration presented in this chapter act as the measures which will be used in this thesis to determine the effectiveness of economic integration. 43

55 Table 2.3 Selected PPP and Price Convergence Studies Study Comments Key Results Isard (1977) Tests the Law of One Price using disaggregated data from manufacturing sectors in the US, Canada, Germany, and Japan. Rogoff (1996) High shor-term volatility of exchange rates is countered by slow dampening out of shocks. Persistent and slow converging deviations of PPP as a direct result of international markets remaining segmented and trade being plagued with friction. Parsley and Wei (1996) Examines the convergence of prices based on the Law of One Price given distortions such as barriers to trade and exchange rate volatility. This is done for 48 US states using highly disaggregated data. Tradable goods convergence to parity in 4 to 5 quarters. Services take three times as long. Tradable convergence speeds are faster than those of other cross-country studies. Study results support nonlinearity of convergence rates. Convergence is generally faster the larger the price deviations. (Continued on next page) 44

56 Table 2.3 (Continued) Selected PPP and Price Convergence Studies Study Comments Key Results Bergin and Glick (2007) Focuses on price dispersion variation over time. Price dispersion declines in the 1990s and increases in the following decade. This forms a U- shaped pattern. Distance plays a significant role in time varying dispersion. Resulting transportation costs play a significant role in dispersion behaviour. Parsley and Wei (2008) Rogers (2007) Examines the Euro effect on prices within the EU. This is done by using highly disaggregated data. The paper utilises the Big Mac Meal for 25 European countries. The prices are also compared across Euro and non-euro countries. Tests evidence of price convergence in Europe based on based on the effects of the monetary union. The paper compares Europe to the US with respect to price convergence. The paper uses disaggregated price data. This paper finds no significant effect of the Euro on price dispersion. Consequently, the Euro adoption had no significant effect on market integration. Paper finds convergence in prices in Europe. This result correlates with the completion of the Single Market within the EU. Traded goods dispersion declined such that it is comparable to the US The results are related to a number of factors including the EU s integration policies. (Continued on next page) 45

57 Table 2.3 (Continued) Selected PPP and Price Convergence Studies Study Comments Key Results Broda and Weinstein (2008) Highly disaggregated micro data on prices are used for the US and Canada. Using barcode data this paper finds that the Law of One Price holds in absolute form. This is true within a country and across borders. The distance coefficients are significantly smaller than those found in similar studies that use aggregate data. This is in the magnitude of 5 to 10. Market segmentation appears to be similar within and across borders. 46

58 CHAPTER 3 GCC INCEPTION, ACHIEVEMENTS AND CHALLENGES This chapter will preview the development of the Gulf Cooperation Council (GCC) and its major achievements. An understanding of why the GCC came into existence and the goals that member countries hoped to achieve will be outlined. The primary emphasis will be on the economic integration process, which has three stages: creating a customs union, a common market, and a monetary union. This chapter will also discuss the challenges that face the GCC region. 3.1 Historical Background In the recent history of the Gulf countries, the Pax Britannica is perhaps the most influential era that shaped and the region, shaping it through the 19 th and 20 th centuries. Earlier, British involvement was largely confined to securing shipping routes from India, through the Gulf to Britain. This role gradually changed as the shipping lanes were threatened by piracy in the Gulf. The continual harassment of British vessels in the region led to a number of expeditions to the Gulf to stop the perpetrators, and in 1820 an anti-piracy or General Treaty was signed between British India and Sheikdoms in the Gulf. The historical background material that follows is based on Onley (2009). The Sheikdoms of the time, which included several individual states known today as the United Arab Emirates (UAE), Bahrain, Qatar, and Kuwait, had all requested the British to provide formal protection at some time or another. Britain, however, was reluctant to become involved in the region beyond the protection of its interests at the ports of Muscat (Oman) and Basra, and maintaining a presence in southern Iran where administrative staff were stationed. Apart from Oman, which achieved an informal arrangement with Britain that provided protection to Muscat, the Gulf States could secure no agreements. The Maritime Truce of 1835 marked a shift in this policy. The treaty banned naval warfare between the Sheikdoms during the six months of the pearling season. The (now UAE) states, which came to be known as the Trucial States as a result of the treaty, lobbied the British India Resident (administrator) to make the agreement perpetual. The treaty was gradually extended to cover longer periods of times with the unanimous agreement of the rulers of the states, and was also extended to Bahrain in 1861 and Qatar in In effect this extended the British Empire s protection to the states; however the involvement was generally in the form of arbitration and mediation in 47

59 disputes. The sizable Gulf Squadron of naval ships used to maintain maritime security was the major deterrent used by the British during this period. British involvement was scaled up in the 20 th century with direct military intervention in conflicts and resolution of infractions by states or rebels by troops sent from British India. The discovery of oil was the catalyst for change in British policy. Realising the importance of the resource to its economy, Britain s involvement become more intense. A characteristic of British engagement was to isolate the region from other foreign powers such as France, Germany or Russia. This led to a great dependence of the Gulf States on Britain for development, especially once oil placed the region on the global map. Britain achieved a number of advantages by isolating the region from its competitors, of which access to oil exploration and production concessions were the most important. In return the Sheikdoms received protection, and numerous developmental projects and programs. The post-war era marked the deterioration of the Pax Britannica. Nationalist movements in Arab countries applied great pressure on Britain to leave the region. Simultaneously, communist influences from Russia and China were fuelling resentment against Imperial presence in the region (Onley 2009; Holden 1971). Changes in British policy were reflected in greater autonomy of Gulf Sheikdoms; Kuwait gained independence in 1961, followed by other Gulf States in the early 1970s. The British role in the region specifically, and in its colonial interests generally receded. In 1968, a declaration of withdrawal from the region was made by the British Government. This was to take place in Until that year, Britain stood by its protection commitments to the Gulf States; an example of this was the repulsion of the Iraqi invasion of Kuwait soon after its independence. Britain finally withdrew from the region completely by 1977, after assisting in the Dhofar war in southern Oman. The withdrawal of the British from the Gulf region and other colonies east of the Suez Canal was a cause of major security concern in the region. The newly independent and autonomous states faced a number of challenges. As disputes and differences were settled between states, attention turned to external threats. These were mainly revolutionary movements fuelled by nationalism and Marxist influences, leading to conflict in some countries. Other threats were represented by the competition for hegemony over the region between bigger countries like Iraq and Iran (Holden 1971). The Iraq Iran war was a direct result of this. The USSR invasion of Afghanistan was another alarm: internal and external threats from communist states or movements were viewed with concern. However, the most serious threat perceived by the Arab countries 48

60 of the Gulf was the Iranian revolution in the 1970s and the consequent change in the regime of that country. Although the Iraq Iran war and the Russian invasion of Afghanistan contributed to the process, the anticipated export of revolutionary ideology from Iran is now considered to have been the major driver behind the creation of the GCC in 1981(Ramazani 1988). During these events, the formation of the GCC was not the only initiative. In 1974 the Shah of Iran proposed a Gulf security organisation. This was rejected by the other Gulf countries, including Iraq. Saudi Arabia in 1977 proposed an Arab security network within the Gulf. This was also rejected by the Gulf States. However, Saudi Arabia did sign a number of bilateral security agreements with its Arab Gulf neighbours, with the exception of Kuwait. In the years preceding the GCC, Kuwait and Oman proposed two polar plans for the security of the region. Kuwait followed a self-reliance approach to security within the region, while Oman proposed a strategic alliance with the United States to form a joint maritime force to ensure security in the Gulf generally, and in the Strait of Hormuz specifically. Neither proposal gained approval. In 1980 Saudi Arabia introduced a proposal mid-point between those of Kuwait and Oman: to confine cooperation between the Arab Gulf states mainly to the area of security. This proposal became the stepping-stone to an agreement of cooperation in 1981 (Ramazani 1988; Bellamy 2004). 3.2 The Formation of the GCC In 1981 six countries, Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and the United Arab Emirates, created the Cooperation Council of the Arab States of the Gulf (GCC). The GCC countries agreed to cooperate and unify their policies in a number of key economic, political, and social areas. The original charter specified the overall cooperation initiative and the areas involved, identified in Box 3.1. The Economic Agreement of 1981 also emphasised the economic goals of the GCC. These fell in to four major areas: The first was to establish economic nationalisation among citizens of the GCC. The second was to achieve economic integration through a number of consecutive steps, creating a Free Trade Area, a Customs Union, a Common Market, and finally a Monetary and Economic Union. The third goal the agreement outlined was the standardisation of laws and regulations relevant to the proposed economic integration process. The fourth and final goal was the development of synergies between the infrastructures of the six countries through joint ventures. 49

61 Box 3.1 Article 4 GCC s Charter 1. To effect coordination, integration and inter-connection between Member States in all fields in order to achieve unity between them. 2. To deepen and strengthen relations, links and areas of cooperation now prevailing between their peoples in various fields. 3. To formulate similar regulations in various fields including the following: i. Economic and financial affairs ii. iii. Commerce, customs and communication Education and culture 4. To stimulate scientific and technological progress in the fields of industry, mining, agriculture, water and animal resources; to establish scientific research; to establish joint ventures and encourage cooperation by the private sector for the good of their peoples. Source: The Cooperation Council of the Arab States of the Gulf - Seretariat General, The Cooperation Council - Charter. Available from: [16 November 2009]. The Agreement was replaced by a new Economic Agreement in 2001, in which the emphasis shifted from cooperation to integration. The new Agreement emphasised a number of areas where integration needed to proceed. These included the Customs Union, the Common Market, and the Monetary and Economic Union. The agreement also included areas such as international economic relationships between the GCC and other groups, integration of development initiatives across the region, labour force and education development, scientific and technical progress, infrastructure integration, and development (The Cooperation Council of the Arab States of the Gulf Secretariat General 2010) Achievements The GCC has a achieved a number of goals worth highlighting. The major achievements are reported in Table 3.1. These milestones are linked to the Unified Economic Agreement (1981) and The Economic Agreement (2001). Since its inception the GCC has established a Free Trade Area (FTA), in External tariffs ranged between 4% and 20%, which allowed some degree of speculation by traders who imported into member countries with lower tariffs and sold them in those where tariffs are higher (Ramazani 1988; Legrenzi 2003). Within the Free Trade Area, duty free goods included those which had 40% added value within the region, produced by 50

62 Free Trade Area Table 3.1 Economic Integration, Achievements, Types of Integration Achievement Comments Article Three (Unified Economic Agreement 1982) Customs Union Article One: The Customs Union (The Economic Agreement 2001) Common Market Article Three (The Economic Agreement 2001) Launched 1983 Implemented as planned in 2003 Completed in 2008 Tariffs range 4% to 20%; 40% GCC national value added; 51% or more GCC ownership in production plant. Common external tariff of 5% on most goods. Monetary and Economic Union Article Four (The Economic Agreement 2001) International Economic Relations Article Two (The Economic Agreement 2001) Transportation, Communication, and Infrastructure Article Twenty-three: Infrastructure Integration (The Economic Agreement 2001) Not yet implemented Signed agreements with Lebanon, Singapore, EFTA. Currently negotiating with ASEAN, Australia, China, India, Japan, MERCOSUR, New Zealand, Pakistan, Syria, and Turkey. Electricity power grid linkage (2010) Two members, Oman and UAE, opted out of the Currency Union. The remaining members postponed the launch of the Currency Union from its original 2010 deadline. Negotiations with EU started in 1984 and were suspended in Recently Bahrain and Oman signed Free Trade Agreements with the United States. Other members are also in the process of negotiating agreements individually. Three phases 1. Bahrain, Kuwait, Qatar, and Saudi Arabia (2005) 2. Oman and the UAE (2009) 3. Completion (2010) 51

63 plants with at least 51% ownership by GCC nationals (Dar and Presley 2001; Legrenzi 2003). This condition, one of many non-tariff barriers that existed at that time despite the FTA, was enforced through the granting of certificates of nation of origin. Other barriers included favouring national products against other GCC states in government purchases. Exemptions were also retained on certain goods, including handicrafts and art (Legrenzi 2008). In 2003 the work on the Customs Union within the GCC was launched (The Cooperation Council of the Arab States of the Gulf - Seretariat General 2009). This included unification of laws and regulations relevant to trade within the region, and ensured the freedom of goods movement across national borders without the hindrance of tariffs or non-tariff barriers. In addition, a single entry point for duty levies was established across GCC ports. Finally, national treatment of GCC goods in each of the six countries was established. The Customs Union s launch underwent a transitional period during , in a lead-up to a fully functional system. During this period some goods were exempted from full Customs Union treatment based on specific members prohibitions (The Cooperation Council of the Arab States of the Gulf Secretariat General 2009). Despite its long delay the customs union within the GCC states marked an important step forward. However, this was a smooth progression. The decision to establish the custom union was made by the Supreme Council at its annual summit in December 2001, but the full implementation of the customs union s common external tariff was blocked prior to the 2003 launch deadline. This was largely due to the varying economic strategies of the GCC countries. Countries that sought to protect fledgling industrial industries negotiated higher tariffs, while those who wanted to maintain their role in re-exports and shipping services wished to maintain lower tariffs (Legrenzi 2008). Moreover, trade liberalisation talks within the region were sidetracked by individual GCC talks with the US. Contrary to the GCC s charter, most GCC countries were involved with the US in free trade talks. This detracted from the efforts to make the customs union functional, and created undesirable rifts between countries. However, a more pressing issue with the customs union was the revenue sharing scheme. Some consensus towards solving this issue was based on sharing revenue according to the destination of imports within the region. Until these issues were resolved the GCC s customs union could not reach its full potential (Middle East Monitor: The Gulf 2005). 52

64 The push for a common market required a number of intermediate steps that helped facilitate the objective of free movement of factors of production. These initiatives included equal cross-border employment opportunities for all GCC nationals in both private and public sectors, however, some reservations of member countries are still allowed. In 2005 the GCC agreed on a number of convergence criteria, not dissimilar to those of the European Monetary Union. These fell into two categories, monetary and financial. The former included inflation rates no higher than 2% above the GCC weighted average, interest rates no higher than 1.5% of the GCC average, and foreign cash reserves to cover a minimum of 4 months of imports. The latter dictated a government deficit of 3% of GDP or less and a public debt burden no higher than 60% of GDP (The Cooperation Council of the Arab States of the Gulf - Seretariat General 2010; in the Emerging Markets Monitor 2009). In 2008 the GCC Common Market was officially completed (The Cooperation Council of the Arab States of the Gulf - Seretariat General 2009). The common market ensured the free movement of GCC nationals, and in some cases non-nationals, across borders, and the free movement of capital across the region. A number of specific initiatives have taken place since the inception of the GCC in These include the ability of GCC nationals to conduct cross-country retail and wholesale activities, including production plants and distribution across GCC countries. Other initiatives have included the creation of a GCC Standardisation Organisation responsible for the harmonisation of goods and services standards, which established standards for goods produced within the region. Other initiatives have included economic citizenship of GCC nationals across the six countries, including equal treatment in stock ownership and incorporation of cross-border firms. In 2005 a Common Trade Policy was adopted to unify the GCC s External Trade Policy. 1 The GCC has also sought to integrate its economies through joint projects and coordinated policies in oil and non-oil sectors. In the oil sector the most evident achievement has been agreement on strategic plans in case of disruption to a member s ability to produce enough oil for domestic consumption, and oil lending to member states that find at least 30% of their expected exports disrupted due to damage to facilities (The Cooperation Council of the Arab States of the Gulf - Seretariat General 2009). The promise of major infrastructure development in this sector has not materialised, however. Initial plans to develop a GCC pipeline to terminals in Oman in 1 See The Cooperation Council of the Arab States of the Gulf - Seretariat General, Economic Cooperation. Available from: [May 2010] for more details. 53

65 the 1980s were never realised. Other projects, such as a GCC refinery, were shelved in the earlier years of integration (Ramazani 1988). In the non-oil sectors, aims have been better realised. In terms of joint projects, the establishment of the Gulf Investment Cooperation (GIC) was a positive step in cooperation: the investment entity lent funds to various industrial, fishery, and agricultural projects in the region and provided consulting expertise on investment projects (Legrenzi 2008; Ramazani 1988). Energy is another sector in which the GCC aimed to develop joint policies and projects. The most prominent of these was the GCC-wide electricity grid. In its first phase, proposed in the GCC Summit of 1997, Bahrain, Kuwait, Qatar and Saudi Arabia would link their grids. In 2001 the GCC established an Electric Interconnection Commission, responsible for overseeing the project. The first phase was implemented in 2005 by awarding several contracts to link the existing grids. In 2009 trials of the first phase were conducted. The second state was the linkage of Oman and the UAE s grid to the rest of the GCC members. Completion of the project is expected in 2010 (The Cooperation Council of the Arab States of the Gulf - Seretariat General 2009). 3.3 The Socio-Economic Characteristics of the GCC The GCC countries are located in the Arabian Peninsula: Figure 3.1 shows the region s central location in the Middle East. A number of observations can be made. The region s location is centred between South East Asia, South Asia, and Europe. Figure 3.1 The GCC and the Middle East Source: Perry-Castañeda Library Map Collection, 54

66 Millions The Middle East generally, and the Persian Gulf specifically, have been both strategic and volatile in contemporary history. The Gulf region holds large reserves of oil and gas, and this has placed a strategic importance on the area since the earliest discoveries. Saudi Arabia is the largest country of the six members of the GCC. The other members, Bahrain, Kuwait, Oman, Qatar, and the UAE, are considerably smaller. Although geographically different, the GCC countries share important similarities. First, they speak the same language. Second they share a great dependence on oil and gas for government revenues. Third, low levels of inflation have been enjoyed over the past few decades. However, the six countries differ in a number of vital socio-economic characteristics Demography Figure 3.2 illustrates the dramatic change in the GCC countries populations over the period The populations of the GCC countries are markedly small, with most countries below the 2 million mark in 1980: Kuwait, Oman, and the UAE had populations of 1.4, 1.2, and 1.0 million respectively. Bahrain and Qatar, the smallest countries in the region, had less than 300,000 inhabitants each in Saudi Arabia, the largest country by size, is also largest in terms of population, with 9.6 million inhabitants in Figure 3.2 Population in 1980 and 2008 (Millions) 2008, , 9.6 Bahrain Kuwait Oman Qatar Saudi Arabia UAE Source: World Bank (2010) Over three decades the picture alters dramatically. In Saudi Arabia the population more than doubled, to 24.6 million people in The increase was even more remarkable in the UAE, which experienced a fourfold increase in population from 55

67 to 4.5 million people. Qatar underwent a substantial population increase too; its inhabitants increased sixfold from 2.3 thousand to 1.3 million. Kuwait and Oman experienced a doubling of their populations to reach 2.8 million people in each country by Bahrain remains the smallest country in terms of population. Its population doubled from 350 to 775 thousand people during the same period. These changes in population can be traced over the time by observing the trends in Figure 3.3. The most interesting characteristic is the decline in growth rates from as high as 10% to as low as 3%. With the exception of Qatar, GCC countries converge to a relatively low annual growth rate. Two factors may have played a role in these significant increases in populations, especially in the cases of Saudi Arabia, Qatar, and the UAE. The first is natural growth or fertility, and the second is immigration. Figure 3.4 allows us to distinguish the first factor, where the fertility of the six countries, measured in births per woman, is plotted against time. 14 Figure 3.3 Population Growth Rates (% p.a.) Bahrain Kuwait Oman Qatar Saudi Arabia UAE Source: World Bank (2010) The decline in growth rates mentioned earlier is supported by a decline in fertility in the region. On average, women in the region have 3 to 4 children a marked decline from the 5 to 7 common earlier. However, the high growth rates experienced by Qatar and the UAE cannot be explained by fertility alone. Declining fertility and high population growth suggest immigration has played a role in the 1990s and 2000s both in these countries specifically and in the entire region. 56

68 Figure 3.4 Fertility (Births per woman) UAE Bahrain Kuwait Oman Qatar Saudi Arabia Source: World Bank (2010) The GCC populations also exhibit interesting compositional changes within their population. Figure 3.5 divides the population into three age categories: under 14, years, and over 65. The first and last categories represent dependents as a proportion of the total population. In the figure, two doughnut graphs are drawn for each country, representing 1980 and 2008 respectively. The former is indicated by the inner circle and the latter by the outer circle. In all the GCC countries, the percentage of under 14s is significantly high. Bahrain is the only exception, with less than 25% of the population falling into that category. Kuwait, Oman, and Saudi Arabia have the highest ratios of under 14s; more than 40% of the population. Qatar and the UAE also have more than 25% of their populations under 14 in The trend of growth, given the reduction in fertility and increase in migration, has changed the composition of these populations significantly. While the proportion of over 65s remains largely unchanged, the 0 14 and years have changed considerably with the exception of Bahrain, where an increase in the 0 14 years category has been experienced over the period and its years category shrank by 5%. Other GCC countries experienced reductions in the proportion of their population in the 0 14 years category by 10% or more. Kuwait and Qatar experienced the largest increase in their proportions of year-olds. In comparison, Saudi Arabia and the UAE experienced the smallest change of about 10%. The implications are significant. First, the composition changes reflect the earlier comments regarding the decline in fertility. They suggest a reduction in the dependency proportion of the population, especially since the over 65 years category 57

69 remained steady over the three decades. The second implication relates to the first but applies to the year category. Since this is considered the working age portion of the population, it indicates the measure of the potential labour force. The marked increase in this category implies more people are entering the job market. The increased demand for jobs places major emphasis on job creation within these economies. Figure 3.5 Population Composition 1980 and 2008 (Percent) Source: World Bank (2010) 58

70 Incomes and Growth In the previous section, a general picture of the GCC s demographic characteristics was presented. Here the macroeconomic characteristics of the six countries will be presented, to help develop a general understanding of the GCC countries and the trends affecting them during the period of economic integration The discussion will start with the Gross Domestic Product (GDP) for the six countries. Figure 3.6 depicts the GCC countries real GDP using the IMF s World Economic Outlook (WEO) October 2009 edition. The values reported in the WEO are in local currencies, but these have been converted into dollars using the official exchange rates reported in the World Bank s World Development Indicators 2010 to avoid the recurring problem of missing data with respect to GCC countries. Figure 3.6 Real Gross Domestic Product (Billions $US) UAE Saudi Arabia Qatar Oman Kuwait Bahrain Source: IMF (2009) World Bank (2010) The figure shows the cumulative real GDP of the six countries, where each bar is divided into the respective incomes. This gives a perspective of the size of these economies compared with each other. The GCC s real GDP has increased twofold over the period , from 250 to 500 billion US$. The contribution of each country to the region s real GDP is indicated by the different bands of each bar. Saudi Arabia dominated with more than 60 % of total GCC output in 1980; the second largest economies were Kuwait and the UAE at 15% each. Bahrain, Oman, and Qatar had shares of 1%, 3%, and 4% respectively. These shares changed over the period, but the 59

71 most noticeable adjustment was the UAE gaining a larger share of GCC total output, from 16% in 1980 to 24% in Oman and Qatar also experienced an increased share of total output, their contributions doubling over the period. In 2008 Oman contributed 6% of total GCC output, while Qatar s share rose to 8%. Bahrain and Kuwait s shares remained largely unchanged. The effect of changes in total GCC real GDP can be traced by the height of the country portions of each bar over time. The countries that gained larger shares over time show increased vertical height in the bands. Figure 3.6 indicates the persistent growth of real GDP in the GCC countries. These changes over time are depicted in Figure 3.7 for GCC countries compared to their arithmetic averages. This was done based on annual (Panel A) and five-year averaged growth rates (Panel B). In each panel the six countries are divided into two groups: on the left, Bahrain, Kuwait, and Oman; on the right, Qatar, Saudi Arabia, and the UAE. The GCC average is shown as a dark solid line in all graphs. Panel A shows that the growth rates of Bahrain and Oman have generally moved together, and closer to the GCC average. Kuwait on the other hand has experienced growth rates substantially different from the group average. This is true of its growth rates in the 1980s, but the largest fluctuations were in 1990/91. The main driver of this dramatic change was the devastation caused by the Iraqi invasion of Kuwait and the post-gulf War recovery. The right panel shows that the growth rates of Qatar and the UAE are volatile, compared to the GCC average. Saudi Arabia s growth rate is closely linked to the region s average. In Panel B the short term fluctuations are smoothed by taking five-year averages of growth rates. Deviations from the GCC average are evident in the panel, where Qatar and Saudi Arabia are obvious examples. Figure 3.7 illustrates that GCC growth rates are generally high; however, they are also volatile even when short term business cycles are smoothed. The Figure shows some degree of synchronicity in the region, although not complete convergence of growth rates. The GDP description above gives a useful overall picture of the GCC countries economic size. Examining the GDP per capita of the six countries as an indicator of well-being is also informative. Figure 3.8 gives the GDP per capita of the GCC countries over the period The figure indicates two levels of wealth: Kuwait, Qatar, and the UAE are relatively richer than their counterparts. These three countries also experienced greater volatility in their GDP per capita than their relatively poorer GCC neighbours. Qatar and the UAE remain richer than the rest of the region. Saudi Arabia experienced a decline in its GDP per capita from over 15,000 to less than 10,

72 Figure 3.7 Real GDP Growth Rates (% p.a.) A. Annual Bahrain Kuwait Oman Qatar Saudi Arabia UAE B. 5 Year Averages GCC Average 10 5 GCC Average Bahrain Kuwait Oman Qatar Saudi Arabia UAE Source: IMF (2009) 61

73 Thousands This is not surprising since it was noted earlier that the population increase in Saudi Arabia doubled, while its real GDP share fell. Bahrain and Oman have experienced a steady rise in their GDP per capita in the region, moving ahead of Saudi Arabia, but remain near the bottom of the table Figure 3.8 Real GDP Per Capita (Thousands $US) Bahrain Kuwait Oman Qatar Saudi Arabia UAE Source: IMF (2009) World Bank (2010) 3.4 Other Economic Dimensions In this section, additional dimensions of the economic characteristics of the GCC will be discussed. These include resource endowments, monetary indicators, and trade: three of the indicators that distinguish the GCC s economies and their performance over time Resource Endowments The GCC countries have been associated with their resource endowments, namely oil and gas. Kuwait, Qatar, Saudi Arabia, and the UAE are members of the Organisation of the Petroleum Exporting Countries (OPEC). The importance of these resources can be realised by observing their stock and production within the region compared to the rest of the world. The oil reserves are considered in Figure 3.9, which represents a pie chart of the world s proven oil reserves up to The GCC s oil reserves are about 40% of the world s proven reserves. The share of reserves within the 62

74 region is decomposed based on the countries. Bahrain is not included as its reserves are minimal. Saudi Arabia holds about one fifth of the world s proven reserves. Kuwait and the UAE are a distant second with 8% of proven reserves. Oman and Qatar have very small shares of world proven oil reserves. The oil production of the six GCC countries is shown in Figure 3.10, where yearly averages are displayed. GCC s total output of oil is captured by the vertical bars, which are divided into the production of each country. The production story follows that of the reserves: Saudi Arabia is by far the largest producer with 8.4 million barrels per day on average. The UAE and Kuwait produced 2.2 and1.8 million barrels per day on average during the same period and Oman and Qatar produced less than 700,000 barrels per day. Figure 3.10 also reveals two trends, the first a decline in overall output of the region, possibly due to excess supply during the late 1970s. The historically low prices undoubtedly spurred more production, as indicated by the figure. Production since the 1990s has steadily increased despite a few dips. Unlike its OPEC member neighbours, Oman shows signs of peaked production in the late 1990s. This is indicated by its shrinking band, and suggests Oman has only small reserves of oil. The gas story is quite different. Figure 3.11 shows the gas reserves within the GCC compared with the rest of the world. The GCC countries hold about 23% of the world s proven gas reserves, so they do not have the significant influence on the gas market that they hold in the oil market. The most significant player within the region is Qatar. Holding about 14% of the world s proven gas reserves, Qatar is one of the important producers of gas in the world. Saudi Arabia holds 4% of the total gas reserves, and the UAE 3.5%. Kuwait and Oman hold 1% and 0.5% respectively. The gas production picture is also different from oil, as shown in Figure Gas production in the region has climbed substantially from 1980 to 2008; increasing more than tenfold. In 2008, Qatar and Saudi Arabia were evenly matched at approximately 80 billion cubic meters, while the UAE produced 50 billion cubic meters. Unlike its oil production, Oman s gas output has shown consistent increases in the 2000s, reaching 25 billion cubic meters in Kuwait and Bahrain have maintained production at 13 billion cubic meters in

75 Figure 3.9 Oil Reserves 2008 (Trillion of barrels) Source: BP (2009) Figure 3.10 Oil Production (Millions of barrels per day) Source: BP (2009) 64

76 Figure 3.11 Gas Reserves 2008 (Trillions of cubic meters) Source: BP (2009) Figure 3.12 Gas Production (Billions of cubic meters) Source: BP (2009) 65

77 Monetary Indicators A common feature of GCC economies is exchange rate regimes. All six countries have maintained some form of fixed or managed exchange rate regimes. With the exception of Kuwait, they have fixed exchange rates vis-à-vis the US dollar. Kuwait instead has used a basket of currencies arrangement based on its trading partners. This changed briefly in 2003 when the GCC countries agreed to unify their exchange rate regimes as a transitional step towards monetary unification (Strum and Siegfried 2005). Figure 3.13 shows the official exchange rates of the six countries from 1980 to The figure is divided into panels grouping each set of three countries with similar US dollar values together. In the left panel Bahrain, Kuwait, and Oman are shown. Since a devaluation of the Omani Rial in the 1980s, the currency peg has been very stable over time. This is also true for the Bahraini Dinar. The Kuwait Dinar, however, shows minor fluctuations due to its exchange rate arrangement. This was briefly interrupted when all the GCC countries agreed to official peg their currencies to the US dollar in Kuwait reverted to its currency basket exchange rate in The right panel shows the exchange rates of Qatar, Saudi Arabia, and the UAE. Despite Saudi Arabia s currency devaluation, the exchange rates of the Qatari and Saudi Rials, and the UAE Dirham, have been very stable Figure 3.13 Official Nominal Exchange Rates ($US per local currency) Kuwait Bahrain Oman Source: World Bank (2010) The fixed exchange rate restricts the ability of the GCC countries to conduct effective monetary policy. Monetary policy within the region follows that of the US. Interest rates movements also mimic those of the US, despite differing business cycles. Moreover, the terms of trade of GCC countries are heavily influenced by US dollar 66

78 movements against other currencies, especially those of major trading partners such as the EU and South East Asian economies. However, a benefit of the US dollar peg is the stability of oil exports revenues priced in the same currency. Moreover, fixed exchange rates provide a nominal anchor, which helps keep inflation in check (Strum et al. 2008). Another outcome of the exchange rate arrangement is the behaviour of interest in the six countries. These are indicated in Figure 3.14, where the deposit interest rates across the GCC countries are compared. Five of the countries interest rates are available for most years, but UAE figures are incomplete. It is unsurprising that the GCC interest rates follow US interest rates. This is an expected outcome of the exchange rate regime. The implication is a small spread of interest rates across the GCC countries. This is very encouraging in terms of convergence criteria set by the GCC in 2005, indicated in Section 3.2.1, where interest rates should not deviate more than 1.5% from the group average Figure 3.14 Deposits Interest Rates (Percent) Bahrain Kuwait Oman Qatar Saudi Arabia UAE United States Source: World Bank (2009), Federal Reserve (2010) The effects of the exchange rate peg are evident in Figure 3.15, where the inflation rates of the six GCC countries over the period are shown. Overall the inflation rates within the region have been remarkably low. There are notable deviations, however, in the 1980s, early 1990s, and early 2000s. The fall in prices are likely to be related to declining oil prices in the 1980s and falling oil production. 67

79 Consequently, public expenditure, which drives economic activity within these countries, is expected to decline. The spike of 1990 is largely due to regional conflict. Bahrain exhibited sharp decline in 2000; this suggests the economic slowdown of that period may have affected Bahrain more than its neighbours. More importantly, the noticeable increase from 2002 onwards is uniform across all countries, with Oman, Qatar, and the UAE exercising above average inflation in A number of factors influence inflation rates in the region; they can be summarised as domestic demand rising and loose credit conditions, a number of booming sectors creating bottlenecks within the economies, a rise in the price of raw materials, and a rise in food prices. The US dollar peg had a role to play as well. The weaker US dollar reflected in the imports of these countries, which trade heavily with Europe (Strum et al. 2008). 16 Figure 3.15 Inflation Rates (% p.a.) Source: IMF (2009) Bahrain Kuwait Oman Qatar Saudi Arabia UAE The inflation picture is mimicked by the money supply growth of the GCC countries. These rates have been volatile over the past three decades, as shown in Figure A general decline in money growth rates from the 80s into the 90s is evident. However, in the 2000s money supply growth rates within GCC countries accelerated. This is compatible with the trend of growth the region has experienced. The monetary policy within the region is reactionary compared to that of the US, and the inflation patterns observed in Figure 3.16 are affected by some of the factors linked to the exchange rate arrangement. 68

80 Figure 3.16 Money Supply Growth (% p.a.) Source: World Bank (2010) Trade In this section an external characteristic of the GCC, namely trade, will be discussed. The aim is to identify regional effects, especially within the context of economic integration. The previous discussions about the Free Trade Area and the Customs Union are relevant here. Given these two initiatives, trade can be expected to intensify within the region. Trade from outside the region may be affected positively or negatively as a result of the GCC s trade policy. In Figure 3.17 the GCC s trade as a percentage of GDP is shown. It is immediately clear that the GCC countries are very open economies that trade heavily. Most GCC countries total trade proportions range around % of GDP. Bahrain s and the UAE s trade as percentages of GDP are larger than the others ; Saudi Arabia has the smallest proportion compared to its neighbours. This is in line with the general trend for comparably larger economies trade being a smaller percentage of their GDP compared with their smaller counterparts. In Figure 3.18 these changes are captured by the ratio of intra- to extra-gcc trade. The ratio depicts intra-regional trade against the rest of the world for each of the six countries. Some interesting patterns can be observed. The relatively poorer countries within the region, Bahrain and Oman, tend to trade significantly more within the region than the other four countries. 69

81 Figure 3.17 Trade as Percentage of GDP (Percent) Bahrain Kuwait Oman Qatar Saudi Arabia UAE Source: World Bank (2010) Figure 3.18 GCC Ratio of Internal to External Trade Source: IMF (2009) and author s calculations For 1980 Bahrain s trade within the region is the equivalent of more than three quarters of its total with the rest of the world. This declines to less than one fifth in Oman s ratio increases steadily in the 1980s. Its ratio peaks at 0.4 in 1990 before declining to levels below 0.2 in The remaining countries trade ratios remain largely below 0.1, with the exception of Qatar, whose ratio of above 0.1 for the 2000s confirms that its intra-regional trade has remained small and unchanged overtime. This 70

82 is unsurprising, since the GCC s economies are relatively homogenous because trade is dominated by exports of hydrocarbons. The trade patterns in the GCC can be further explored by examining the current account balances of the six countries. These are depicted in Figure 3.19 where the current account balance as a percentage of GDP is plotted for each of the GCC countries. Most of the six countries had surpluses during the 1980s. These declined over time. The 1990s saw most GCC countries experience currency account deficits. Kuwait was a particularly severe case because of the regional conflict of the first Gulf War. In the 2000s the trend was reversed: all GCC countries experienced current account surpluses again. Current account surpluses are related directly to revenues from oil exports, but also to the imports of consumer goods and services. In the 2000s these played a significant role in the patterns shown here (Strum et al. 2008). Figure 3.19 Current Account Balance (% of GDP) Source: IMF (2009) The role of hydrocarbons can be further emphasised with respect to foreign exchange reserve accumulation. Figures 3.20 and 3.21 depict this relationship by showing the total reserves and oil prices. In the 1980s reserves were modest for all the GCC countries except Saudi Arabia. Reserves started to increase in the 1990s and 2000s. This is shown in Figure 3.20, where in five out of the six countries foreign reserves increased dramatically, and most dramatically in the UAE and Saudi Arabia. This increase in reserves may be attributed to increases in the production of oil for export, and in the price of oil. Figure 3.21 plots two oil price series from 1980 to 2008: 71

83 the first is an average annual historical price series in 2008 US dollars and the second is an average annual spot price of Dubai Crude Oil. The figure shows the decline of prices in the 1980s and a stagnation in the mid-period before prices started to increase significantly in the late 90s. This increase in price and production (and by proxy exports of resources) is related directly to the increase in foreign exchange reserves within the region. Figure 3.20 Total Foreign Reserves ($US) Source: World Bank (2010) Figure 3.21 Crude Oil Prices ($US Per Barrel) Source: BP (2009) *Historical prices include mostly dated Brent Crude prices, which are converted (by source) into 2008 dollars. 72

84 3.5 Future Challenges The GCC countries face significant challenges at both country and regional level. At the country level, development challenges present themselves fairly uniformly in the form of resource depletion and the need to find new growth engines. The challenges are different at the regional level, and are related to the effective operations of the economic integration project the GCC is undergoing The Country Level In the previous section, a number of socio-economic characteristics were discussed and differences between countries were highlighted. One of the key economic factors is resource depletion. Oil contributes 30% or more to the GDP of the region, and about 75% of governments revenue (Fasano and Iqbal 2003). Countries like Bahrain and Oman must urgently develop alternative sources of income to hydrocarbons. Diversification is one of the key objectives the region is actively seeking to develop. Diversification so far has concentrated on areas of comparative advantage of each country. This includes petrochemicals and industries that are energy intensive, such as aluminium smelting. The Bahraini diversification experience targets financial services and regional tourism. Bahrain has also built on its comparative advantage by setting up aluminium production. The UAE has diversified into finance, international tourism, and trade related services. Saudi Arabia s diversification efforts concentrate on light industry among other strategies. Qatar has focused on hydrocarbons based on its large natural gas reserves, but is developing tourism in the form of conference and international event hosting (Strum et al. 2008; Jbili 2000). The region s demographic composition puts great emphasis on job creation, as mentioned above. A large expatriate working population forms three fourths of the total labour force in many cases (Fasano and Iqbal 2003). The number of foreign workers is mainly due to the small national populations but large demand for labour. GCC countries have access to labour at competitive wages from the Indian subcontinent, the Philippines and other Arab countries. This has seen a concentration of foreign workers in the private sector and national workers in the public sector of each country. To address this issue, most GCC countries choose to impose quotas on foreign labourers. Some have also embarked on nationalisation of the labour market as a short-term solution to the problem of high dependence on foreign workers. All six countries have 73

85 developed training and development initiatives to enable nationals to meet job requirements in the private sector (Fasano and Iqbal 2003; Strum et al. 2008). The policies that individual GCC countries have chosen to address depend largely on the individual member s conditions, but Fasano and Iqbal (2003) note a number of general areas where governments are working to address the issues above. Of these, stabilisation funds, privatisation, and foreign direct investment standout. A number of GCC countries have development stabilisation funds from oil revenues to help counter revenue volatility due to market price changes. Jbili (2000) makes the point that GCC countries need a formal arrangement of oil stabilisation funds that act as a cushion, as well as fiscal discipline to reduce government spending fluctuations reflecting oil prices. Another recurring theme for reforms is privatisation. GCC countries are encouraged to reduce public sector involvement in the production of commercial goods and services such as utilities and telecommunication. This approach expected to position the private sector to be the growth generating sector in the region as countries move away from oil (Jbili 2000). The GCC countries have pursued this objective in varying degrees; for example, Oman, Qatar, and the UAE have involved the private sector and foreign investments in infrastructure projects in the water and energy sectors. In comparison, Saudi Arabia has worked on privatising the telecommunication sector (Fasano and Iqbal 2003). Foreign direct investment (FDI) is a third area for GCC countries consider as part of addressing developmental challenges. FDI stock ratio to GDP for the GCC countries is comparatively low at 11%, as against the 23% global average. FDI flows in the region have been closely linked to oil prices and hence remain volatile (Strum et al. 2008). GCC countries are now trying to attract direct foreign investment by introducing reforms to their investment laws that allow for full foreign ownership of firms in nonhydrocarbon sectors. Some have reduced corporate income tax, and red tape, and allowed foreigners greater access to capital markets (Fasano and Iqbal 2003). Such measures taken by GCC countries to address their individual, but similar, challenges are at varying degrees of execution. There remain difficulties common to all countries, including increasing the non-oil sector s contribution to GDP, which remains small in most cases (Fasano and Iqbal 2003). 74

86 3.5.2 Regional Level The challenges the GCC faces as regional bloc stem from the members characteristic differences. The lack of progress in the 80s and 90s did not reflect positively on the region as far as economic integration project was concerned. The 2000s reinvigorated economic integration with introduction of the Customs Union (2003) and the Common Market (2008), but the Monetary and Economic Union remains a work in progress, as the six countries find they are unable to meet the original 2010 deadline. A number of reasons explain why the GCC has taken longer than expected to achieve some of its major milestones. Most point to members differing approaches to policy and implementation, a lack of harmonised standards and common regulations across the region, and bureaucracies that have hindered developments in several areas and specifically the trade integration process. This was the particular challenge when administering the FTA of 1983: the rules of origin were difficult to implement given the non-tariff barriers that existed at the time, including border procedures, varying product specifications across countries, and bureaucracy (Dar and Presley 2001). With non-tariff barriers still in place, the certificate of national origin was essential to establish the duty free status of products that met the rules of origin requirements; however, certain products were excluded from duty free status and remained so until the 2000s. An added distortion, furthermore is government purchases and contracts that usually give priority to domestic producers and service providers instead of other GCC members products (Legrenzi 2003). In an effort to reduce distortions to the free flow of goods, the Gulf Standardisation and Metrology Organisation (originally the Saudi Arabia Standards Organisation in the 1980s) was upgraded to an independent agency in 2001 (Strum and Siegfried 2005). This agency has played a key role in the lead-up to the customs union by promoting mutual recognition and harmonisation of national standards. The completion of the common market is another regional challenge the GCC countries have to address. A keystone of a common market is free movement of factors of production. Strum and Siegfried (2005) point to three areas where the GCC has to improve, to ensure the success of the common market: foreign ownership of equity and real estate, regulation of cross-country banking operations, and development of capital markets. GCC members are already involved in harmonising regulations with respect to investment, financial markets, and banking to facilitate the common market. Despite their efforts, the number of banks operating across borders remains small. The disparity 75

87 of banking regulations within the GCC is probably the largest hurdle facing integration in this area (Strum and Siegfried 2005). The monetary union is presently the most significant stage of economic integration that the GCC countries need to achieve. Article Four of the Economic Agreement of 2001 broadly outlines the goal of achieving a monetary and economic union and identifies the requirements. These include harmonisation of economic policies, banking legislation, and convergence criteria. The convergence criteria the GCC countries have set for the monetary union are analogous to the European Union s measures. These include monetary measures such as inflation, interest rates, foreign reserves, and money growth. Likewise, the GCC s convergence criteria include public debt and government deficits. There are a few concerns about how the GCC will meet its convergence criteria,. Inflation, for example, has diverged within the region so that countries have deviated from each other. Incomes remain divided into two groups, and trade remains significantly low. In terms of other convergence criteria such as interest rates, the region generally follows the US rates movements and thus is less likely to deviate greatly. Section showed GCC interest rates moving closer, especially in the last decade. With respect to other fiscal convergence criteria the story is somewhat different, and GCC budget balances vary widely. In the period , for example, Saudi Arabia s annual budget balance on average was -8.8% of GDP. Oman s annual budget balance was -0.1% on average for the same period (Strum and Siegfried 2005). The picture changes dramatically during the period , where all GCC countries experienced budget surpluses ranging from 5% to 30% of GDP. These values include the IMF s projections for 2007 and 2008 (Strum et al. 2008). Government debt varies considerably within the six GCC countries. Government debt as a percentage of GDP is highest in Saudi Arabia for the period , accounting for more than 80% of GDP. In contrast, the lowest government debt in the same time was in the UAE, where it was less than 10% of GDP on average (Fasano and Iqbal 2002). These debt levels declined significantly in the period and remain 20% of GDP or less for all countries. The difference between the GCC countries government debt has converged over this period. These values also include IMF s projections for 2007 and 2008 (Strum et al. 2008). In the cases of both government debt and budget balances, the role of hydrocarbons is significant. The increases in oil prices shown in Section emphasise the impact of oil on government spending and budget balances within the 76

88 region. Zaidi (1990) highlights the channel through which this effect takes place. He points out that in the case of the GCC countries, all highly open economies, government spending drives money growth and private spending. Increased oil receipts raise net foreign assets in the banking system. This is offset by government deposits. When governments spend oil exports receipts domestically, the money supply rises and consequent growth is expected. Zaidi (1990) suggests that liquidity excess is channelled through expenditure on imports of goods and services. The impact of this mechanism has been visible in the past decade where governments have spent on large scale infrastructure projects. This was the catalyst suggested by Zaidi (1990) behind the observed growth discussed in Section Besides meeting the convergence criteria, the GCC countries have other challenges to overcome prior to launching a common currency. The monetary union project faces the challenge of consensus. Oman announced its withdrawal from the single currency stage of economic integration in 2006 (Emerging Markets Monitor 2009). The reversion of Kuwait to its basket of currency regime in 2007 has also been a significant event in preparation for the monetary union. There is the question of choice of an exchange rate regime of the new currency. Should it be the US dollar pegged currency or a Kuwaiti Dinar basket exchange rate regime? The flexibility of a basket of currencies as an exchange rate regime can help combat imported inflation such as that observed in the 2000s (Emerging Markets Monitor 2009); however, as mentioned earlier, the dollar peg has ensured the stability of oil export revenues. Despite this, it is likely that the GCC s common currency may benefit from a basket arrangement if the GCC finalises a FTA with the EU (Fasano and Iqbal 2002). This is not currently a pressing issue, as GCC EU FTA talks have stalled, as indicated in Section Other necessary developments to ensure the success of the proposed monetary union are in the institutional realm. The GCC has generally been an inter-governmental forum for legislation and monitoring the integration process. Unlike the EU, the GCC Secretariat can only recommend policy options and nudge governments towards enforcing objectives and resolutions (Strum and Siegfried 2005). This may have worked in the earlier integration period, but in the case of a monetary union it is essential to establish supernational entities responsible for conducting monetary policy. It is also essential to develop standardised statistical indicators across the six countries. These would provide greater transparency (Fasano and Iqbal 2002; Fasano and Iqbal 2003). 77

89 It remains to be seen if the GCC will successfully launch functional monetary union despite some of the challenges mentioned above. Critically, the GCC needs to develop effective institutions that will enable it to achieve the broader objectives of economic integration. 78

90 CHAPTER 4 ECONOMIC INTEGRATION AND INTRA-REGIONAL TRADE 4.1 Introduction The regionalisation of world trade is not a new phenomenon; in fact, it has gained strength over the past few decades. The number of regional trade agreements (RTAs) reported to the World Trade Organization has increased rapidly in both developed and developing countries. Although not every RTA has been successful in improving the fortunes of its members, RTAs continue to exist and regenerate themselves. The reason behind this is that the underlying goals of typical RTAs are based not merely on trade perspectives but on a number of goals and objectives which include: i) gains from trade, ii) strengthening domestic policy reform, iii) increasing multilateral bargaining power, iv) guaranteeing access to markets and concessions, v) strategic linkages and alliances, and vi) influencing multilateral negotiations through regional interplay (Whalley 1998). This chapter is concerned with the first of these objectives, the traditional gains from trade, and will investigate the effects of RTAs on their members intra-regional trade. A by-product of this analysis is the determination of the trade creation and diversion effects, first introduced by Viner (1950). Arguing that customs unions may have negative welfare implications by diverting trade, Viner (1950) shows that not all trade agreements are necessarily welfare enhancing. Thus, RTAs can have profound effects on trade flows and welfare, for good or ill. This chapter has two main aims: to measure the effect of RTAs on international trade flows using empirical tools for the period 1995 to 2006; and to apply a disaggregated analysis to the Gulf Cooperation Council (GCC) region on a longer time period, from 1980 to The GCC consists of six developing countries Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and the UAE undergoing long-term economic integration. Significant developments in the integration process in recent years justify paying closer attention to this region. The paper proceeds as follows: Section 4.2 provides a brief background of an empirical model commonly used to examine bilateral trade flows, which is the gravity model. Section 4.3 will present a two-step methodological framework of the traditional gravity model and GCC s intra-regional trade. Section 4.4 discusses the data used to estimate the traditional gravity model and its empirical results. It also compares those 79

91 results with other studies in the literature. Section 4.5 discusses the concepts of trade creation and trade diversion with respect to the RTAs included in the sample. Section 4.6 discusses the disaggregated trade patterns within the GCC. The chapter concludes with a summary of findings and recommendations. 4.2 The Gravity Approach to Trade International trade flows are traditionally modelled using the gravity model. This model is originated from physics, in particular from Newton s work on gravity that led to the Law of Universal Gravitation. The law explains that the attractive force between two objects is the product of a constant, their masses, and distance squared. This concept has been applied to studies of migration, tourism, and commodity shipping (Bergstrand 1985). Gravity models are extensively used in economic analyses to predict trade flows (Anderson 1979, Bergstrand 1985, Feenstra 2004). Earlier applications of the model to trade flows were tested by Tingbergen (1962), Poyhonen (1963), and Linnemann (1966). The model proved to have a great degree of accuracy in determining trade flows between trading partners; however, it had no theoretical justification originally (Anderson 1979). Several attempts at linking the model to theory were made. Anderson (1979) used an expenditure approach to link the gravity model to an aggregate spending two-country model. Bergstrand (1985) approached the problem from a microeconomic perspective, using a general equilibrium model to achieve a generalised gravity equation. Deardorff (1998) proved that the gravity model can be derived from neo-classical trade theories and the Hecksher-Ohlin model. Deardorff (1998) applied both frictionless and impeded trade versions of the model in his derivation and also linked the gravity model to homogenous and differentiated goods. In doing so, he established a fundamental justification for the use of the gravity equation. The gravity model was extensively applied to RTA analysis. Frankel et al (1995) and Frankel and Wei (1998), for example, applied the gravity model to a cross-section of countries to determine trade flows and the welfare implications of regional trading blocs. Their focus was on the effect trade blocs had on trade flows, and included several dummy variables in an attempt to capture the effect. This chapter follows this literature to find the trade bloc s effects on its members. Other applications of the gravity model involved measuring border effects on trade. These studies include McCallum (1995), Feenstra (2002) and Anderson and van Wincoop (2003). In this literature, attempts were made to explain the border effect on 80

92 trade while correctly specifying the model. A number of these studies, such as McCallum (1995), concentrated on the border effect between Canada and the US. McCallum (1995) tested the border effects between Canadian provinces and some American states using a dummy for intra-provincial trade, and found substantial border effects between Canada and the US. Anderson and van Wincoop (2003) took the analysis a step further and developed what they called multilateral resistance to explain McCallum s (1995) results and their deviation from theory. Anderson and van Wincoop (2003) suggested that the McCallum results were exaggerated and biased because of variable omission and the specification of his model. Their model suggested a theoretically correct specification using multilateral resistance to explain the border effects. They find relatively smaller border effects between Canada and the US. Feenstra s (2002) approach differed from both studies above by using fixed effect methods. These isolate the effects of importers and exporters as fixed in the model. The gravity model has also been used to explain the effects of currency unions on trade. Such studies include Frankel and Rose (2002) and Rose (2000). Conceptually, Figure 4.1 illustrates the Gravity Model in international trade. It depicts three economies of different sizes, 1 being the largest and 3 the smallest. Figure 4.1 The Gravity Model Concept Medium Country 2 M 12 M 23 M 21 M 32 Large Country 1 M 13 3 Small Country M 31 M 12 +M 21 >M 23 +M 32 As each country is exactly the same distance from the other two, transport costs play no role in determining trade flows in this stylised version of the model. The notation M ij indicates the bilateral trade flow from country i to country j. As countries 1 81

93 and 2 are the largest, the gravity model predicts that trade between them is necessarily larger than trade between countries 2 and 3. Thus, M 12 + M 21 > M 23 + M 32. Based on the work of Deardorff (1998), Anderson and Wincoop (2003), and Feenstra (2004), the gravity model can be derived as follows. Consider country i with i=1, 2,,H countries in the world. Each country N i produces varieties of good k, where k=1,2,3,n. Country j s consumption of good k, imported from country i is denoted by C ijk. Utility of consumers in country j is given by (4.1) U 1 j cijk H N i i 1 k 1 where σ > 0. Equation (4.1) can be generalised for all goods. FOB prices paid by consumers in country j are assumed equal across all varieties N 1 imported from i. As buyers in country j face a transportation cost t ij when importing goods from i, they pay p i t ij for every unit imported. Let c ijk = w ijk Y j, where w ijk is the portion country j s income (Y j ) spent on imports of good k from country i. Let M j be the total imports of j so that corresponding budget share of its imports from i is w ijk = p i t ij c ijk M j. The above setup then implies that w ijk is a constant with respect to the k subscript, that is, the budget share of the N i varieties imported from country i into country j are the same for each variety. In other words, c ijk = c ij. Equation (4.1) can now be represented as (4.2) 1 H U N c j i ij i 1,, The budget constraint for country j is (4.3) Y H N p t c j i i ij ij i 1 Maximising the utility function (4.2) subject to (4.3) yields the demand function for good k imported by country j: (4.4) where P j is the CES price index, defined as H (4.5) Pj Ni pitij i 1 The value of total exports from country i to country j is defined as Xij Ni pitijcij. Using this identity and combining (4.4) and (4.5) yields (4.6) c ij p t i ij X P j pt 1 i ij ij NiYj P j Y j P j, 11 1,.. 82

94 Equation (4.6) can be rearranged to resemble a typical gravity equation by defining income of country i as the product of the quantity of goods produced N i and their price p i. Country i s income is thus Y i =N i p i. Using this property and (4.6) we can rewrite the gravity equation as Y H N p t c j i i ij ij i 1 Equation (4.6) can also be expressed in logs to obtain an additive log-linear form as (4.7) log X logy logy log p 1 log t 1 log P. ij i j i ij j This log-linear equation is the basis of estimation in this chapter. Incomes Y i and Y j are measured using GDP. Distance is used as a proxy for transportation cost t ij. For more details, see Appendix A Application to World Trade and the GCC This chapter s approach to estimating trade flows is based on two sequential steps. In the first step, we consider the determinants of bilateral exports of a sample of 145 countries. The objective here is to quantify the RTA effects on normal bilateral trade expected as a result of fundamental economic variables such as GDP and per capita GDP. In the second step, we consider commodity-specific bilateral exports among the GCC countries based on trade within the region. The second step permits more detailed examination of trade within the GCC. Details of the two steps follow Step1: The Traditional Gravity Approach to Determining Total Trade In the first step the bilateral exports between country pairs are taken to be a function of income, income per capita, distance, common borders, language, and RTA membership. We index countries by i = 1,,H, therefore, trade between countries i and j is determined as follows: (4.8) Xij f Yi Yj Ci C j ij ij ij ij where X ij represents bilateral exports from i to j; Y i is the income of the exporting country; Y j is the income of the importing country; C i is the per capita income of the exporting country; C j is the per capita income of the importing country; Distance ij is the spatial distance between the trading partners capitals; Adjacency ij is a dummy variable indicating a common border; Language ij is a dummy that captures the common languages shared by trading partners; RTA ij is a dummy that represents membership of a regional trade agreement by both trading partners. According to model (4.8), total 83..,,,Distance,Adjacency,Language,RTA,

95 trade between countries i and j is determined by the economic size of the two economies, their per capita affluence, geographic proximity, cultural differences (as measured by the language variable), and trading arrangements. Model (4.8) is taken to be log-linear: (4.9) log X logy logy log C log C log D ij 1 2 i 3 j 4 i 5 j 6 ij Adj Lang RTA, 7 ij 8 ij 9 ij ij where X ij is the value of exports from country i to country j; Y i is GDP of country i valued at nominal US dollars; Y j is GDP of country j valued at nominal US dollars; C i is per capita GDP of country i valued at nominal US dollars; C j is per capita GDP of country j valued nominal US dollars; D ij is the distance between the capital cities of country i and country j measured in kilometres; Adj ij is a dummy variable for adjacency or common borders, 1 for a common border, 0 for none; Lang ij is a dummy variable for common language that takes the value of 1 for common language and 0 for otherwise; RTA ij is a dummy variable for Regional Trading Arrangements that takes the value 1 when both i and j are members of the same agreement, 0 otherwise; and ε ij is a disturbance term. Equation (4.9) is directly related to the derived form in (4.7). Equation (4.9) includes size measures, GDP per capita, which are not included in the simplified model (4.7). The distance parameter 6 is reflective of the term (1- ) in (4.7). Since t ij, transportation cost, is not readily observable, distance is used as a proxy. Finally, equation (4.9) also includes a number of dummy variables that are used to explain qualitative variables of concern. The parameters can be interpreted as follows: 2 and 3 represent the income elasticities of the exporting country and importing country; these are expected to be positive as larger economies are expected to trade more with each other. The parameters 4 and 5 are elasticities relating to the wealth of countries as measured by GDP per capita. The use of GDP per capita instead of population is justified by Frankel (1997, pp 57 59). The distance parameter 6 represents the distance elasticity, which is expected to be negative. The parameters 7 and 8 reflect the effects of adjacency and common languages on bilateral trade. It is expected that if countries have common borders, more trade is facilitated, so the parameter 7 is expected to be positive. Countries with common languages may find it easier to trade with one another; thus the parameter 8 is 84

96 expected to be positive. Finally, the RTA parameter 9 represents the trade bloc effect on bilateral trade. This value may be positive or negative Step 2: Trade within the GCC Countries The second step deals with trade among members of the GCC, with trade disaggregated by product group. Total trade, as determined by Step 1, is split by product group; we then identify those members whose trade is systematically above or below expectation. The expected value of trade is estimated based on socio-geo-economic variables, similar to those used in Step 1. Suppose total trade is made up of n product groups, which we index by p=1,2,,n, so that X ij n X p 1 p ij, where p X ij is the exports of product group p from country i to country j. If we write S for the set of countries that are members of the GCC, total exports by country (4.10) i Sare then n j S p 1 X i j S X p X X i S. i ij ij. Thus Disaggregated trade within the GCC is determined by total trade of the two countries concerned, X i, X j, together with the economic/geographic/cultural variables of model p (4.11) Xij g Xi X j ij ij ij,, Adjacency, Distance, Country/Commodity Dummy. Model (4.11) is taken to be log-linear: (4.12) where log X log X log X log D p ij 1 2 i 2 j 3 ij 4 + Adj Country/Commodity Dummies, i, j S, ij k k ij k X is export value from country i to country j of product group p; X i total p ij exports of country i; X j total exports of country j; D ij is the distance between the capital cities of country i and country j measured in kilometres; Adj ij is a dummy variable for adjacency or common borders, 1 for a common border, 0 for none; γ k is Country/Commodity dummy takes a value of 1 if country i exports commodity p, 0 otherwise; and ij is a disturbance term. 4.4 Data and Empirical Results: The Traditional Model (Step 1) The first part of this section describes the data sources and data used in estimating the gravity model (4.9). The second sub-section will discuss the estimation results and RTA implications. 85

97 4.4.1 Data The data used were obtained for 145 countries. The list of countries is included in Appendix A4.2. Bilateral trade was measured in millions of dollars of exports from the exporting country i to the importing country j. The bilateral trade data were obtained from the IMF s Direction of Trade Statistics. The sample period is divided in five crosssections from 1995 to These are divided as follows; 1995, 1998, 2001, 2003 and The sample period aims to investigate the recent developments in RTA effects on international trade in the post-gatt era when the WTO became operational. This study differs from previous studies in the extent of coverage. Unavailable data points were considered zero trade, which may cause a downward bias on the estimates of elasticities. From a maximum 20,880 [= ] possible bilateral trade flows in every cross-section only 11,561 to 15,127 observations were useful for estimation. The variables GDP and GDP per capita were obtained from the IMF World Economic Outlook. GDP valued at nominal US dollars was used for all trading partners as well as GDP per capita also valued at nominal US dollars. Since transportation costs are not directly observable, distances in kilometres were used as a proxy. These were obtained from CEPII 1, where they are measured between capital cities. An alternative method to measuring transportation cost is to use the ratio of import c.i.f values to export f.o.b values for matched bilateral pairs of countries. The use of c.i.f/f.o.b values is justified in a number of studies, such as Geraci and Prewo (1977) and Hummels and Lugovskyy (2006). These values are calculated from the data available from the IMF s Direction of Trade Statistics database. Bilateral dummies are used to capture the situation where both trading partners belong to the same RTA. If trading partners did not belong to the same RTA, they received 0; if they did belong to the same RTA, they received a value of 1. Table 4.1 reports the means and standard deviations of bilateral exports across the five cross-sections in two cases. The first case includes zero trade observations, while the second case excludes them. Table 4.1 illustrates a number of noteworthy points: first, on average, bilateral exports are comparatively low in Case 1 compared to Case 2, as might be expected. Second, there is a substantial increase in trade on average over the sample period. In fact, bilateral exports have more approximately doubled on average over the past decade in case 1. 1 Centre D études Prospectives et D Informations Internationales, available at 86

98 Table 4.1 Bilateral Exports ($US Millions) Case 1: Including Zero Observations Mean Standard Deviation 2,387 2,714 3,161 3,516 5,069 Number of Observations 20,880 20,880 20,880 20,880 20,880 Case 2: Excluding Zero Observations Mean Standard Deviation 3,144 3,392 3,791 4,175 5,842 Number of Observations 11,561 12,855 13,977 14,254 15,127 Note: The means and standard deviations are expressed in millions of US dollars. During the last three years of the sample, the sample mean indicates an increase of 58% and 49% in Case 1 and Case 2 respectively. Third, there are large variations within the sample, more than doubling over the sample period. The standard deviation rises steadily during the sample period and mirrors the increase in the mean. This indicates a considerable variation in trading patterns towards the end of the period. It further indicates that not all countries included in the sample necessarily trade more, despite the implications of greater mean. Finally, excluding zero trade observation not only increases the mean of the sample but also increases the standard deviation. This may indicate that trade is actually more variable than expected Empirical Results Table 4.2 presents the OLS estimates of the gravity equation (4.9). The estimation was carried out over five cross-sections: 1995, 1998, 2001, 2003, and The coefficient of income logy i represents country i s elasticity of the exports with respect to income. During the period, this elasticity increased from 1.07 in 1995 to 1.19 in The coefficient of Y j, the importing country s income elasticity, increased marginally from 0.82 in 1995 and 0.88 in However, the importing country s income plays less of a role in determining the level of bilateral exports. Per capita income C i and C j measure the differences in size, where larger countries are expected to trade more with each other compared to their smaller counterparts. With respect to the per capita income of an exporting country, the coefficient of C i declined during the sample period from 0.08 to Similarly, the effect of the importer s GDP per capita, 87

99 C j, declined substantially from 0.10 in 1995 to 0.03 in As a result, the influence of per capita income on bilateral exports diminished towards the end of the period. Both income and per capita income are significant at the 1% level in most years. Table 4.2 indicates that throughout all years 1995 to 2006 the distance coefficient is of the correct sign (negative) and significant at the 1% level. In other words, the further the trading partners are from each other, the greater the cost of transportation and the lower their trade. The elasticity of bilateral trade with respect to distance fluctuates modestly over the period. The gravity equation (4.9) estimated in Table 4.2 includes a number of dummy variables. These are divided into two categories: first, the dummy variables of common borders or adjacency and common language; and second, RTA membership indicators. The adjacency dummy is significant at the 1% level in all years and shows an upward trend from 0.63 to 0.82, or 88% [=(e ) x100] to 127% effect above normal trade explained by economic factors. This suggests that countries on average traded more with their neighbours in the latter years of the sample period. Similarly, common language plays a statistically significant role, at the 1% level, in affecting bilateral exports. In fact, its coefficient increased during the sample period. The language dummy increased from 0.75 to 1.03, or 112% to 180% effect above normal trade. This confirms that language plays an increasingly important role in facilitating bilateral trade. These results suggest that cultural and geographical variables influence bilateral trade to a great extent and cannot be ignored. The second category of dummy variables is used to represent membership in RTAs described earlier. In the industrialised countries, the sample includes the European Union (EU), North American Free Trade Area (NAFTA), Closer Economic Relation (CER), and European Free Trade Area (EFTA). The EU, one of the oldest existing RTAs in the form of a customs union, shows negligible effect during the period. The EU dummy coefficient increased from in 1995 to in 2001 before declining to in This translates into a -13% to -8% effect of EU membership on the bilateral exports of the countries involved. The EU dummy is significant only in 2006, at the 10% level. This result is surprising since the EU is considered to have fostered greater trade amongst its members. The lack of statistical significance, however, does not allow such inference. 88

100 Variable (1) 1995 (2) Table 4.2 First Set of Estimates of the Gravity Equation Year 1998 (3) Income Y i (0.011) (0.011) (0.010) (0.010) (0.010) Y j (0.011) (0.010) (0.010) (0.010) (0.010) Per capita income C i (0.016) (0.015) (0.015) (0.014) (0.015) C j (0.015) (0.015) (0.014) (0.014) (0.014) Distance (0.024) (0.023) (0.023) (0.022) (0.023) Adjacency (0.121) (0.108) (0.112) (0.109) (0.115) Common Language (0.060) (0.057) (0.055) (0.055) (0.055) RTA EU (0.080) (0.080) (0.083) (0.082) (0.083) NAFTA (0.530) (0.509) (0.537) (0.550) (0.660) EFTA (0.518) (0.523) (0.551) (0.510) (0.499) CER (0.108) (0.119) (0.126) (0.126) (0.144) APEC (0.085) (0.096) (0.095) (0.095) (0.109) ASEAN (0.255) (0.289) (0.255) (0.239) (0.267) MERCOSUR (0.341) (0.313) (0.492) (0.505) (0.491) LAIA (0.120) (0.118) (0.136) (0.135) (0.134) CAN (0.223) (0.327) (0.372) (0.352) (0.293) CARICOM (0.232) (0.220) (0.205) (0.201) (0.204) COMESA (0.355) (0.325) (0.353) (0.338) (0.340) 2001 (4) 2003 (5) 2006 (6) Notes: Dependent variable: log of bilateral exports. White heteroskedasticity-consistent standard errors in parentheses. See Appendix for the full names of RTAs. (Continued on next page) 89

101 Variable (1) 1995 (2) Table 4.2 (Continued) First Set of Estimates of the Gravity Equation 1998 (3) CEMAC (0.174) (0.218) (0.212) (0.182) (0.188) CACM (0.748) (0.621) (0.556) (0.766) (0.774) WAEMU (0.249) (0.323) (0.353) (0.342) (0.364) GCC (0.215) (0.222) (0.226) (0.234) (0.257) PAN_ARAB (0.143) (0.140) (0.129) (0.121) (0.124) CIS (0.195) (0.214) (0.197) (0.181) (0.247) ECO (0.277) (0.328) (0.321) (0.364) (0.355) S.E. of regression R Number of Observations 11,399 12,695 13,722 13,989 14,857 Notes: Dependent variable: log of bilateral exports. White heteroskedasticity-consistent standard errors in parentheses. See Appendix for the full names of RTAs. Year 2001 (4) 2003 (5) 2006 (6) 90

102 Nevertheless, these values should be considered with caution: the sample does not include the enlargement of the EU to the 25 current members. Elsewhere in Europe, EFTA shows fluctuating effects during the period. In 1995, the EFTA RTA had a coefficient of 0.42, which translates to a 52% effect on bilateral trade within the region. Its effect peaks in 1998 at 0.84, or 131%. EFTA declines in effect in 2001 and 2003, to as low as 0.52 or 67%; however, in 2006 a coefficient of 0.81, translates to a 124% effect on bilateral trade. In all cross-section years, EFTA coefficients are insignificant. Little can be inferred about the true effect of the RTA. In North America, NAFTA s dummy shows consistent negative effects during the period, between and -1.93, or -72% and -86%. NAFTA s bilateral exports are not affected positively by the RTA, as suggested by the model. NAFTA s effect weakens over the sample period, especially in 2006 when it had an effect of -86% below normal trade. These results are significant at 5% in 1995 and NAFTA s coefficients are significant at the 1% level in all other years. On the other side of the globe, CER shows substantial positive effects on its members bilateral exports. This effect declines over the period, from 0.70 to 0.36, or 101% to 43%. During the sample period CER s effect peaks in 2001 at 0.74, which translates to 109%. These results are statistically significant in all years at the 1% level. Although these values are very large, they may be explained by CER s remoteness and the close proximity of its members, Australia and New Zealand. In South East Asia, the Association of South East Asian Nations (ASEAN) and Association of Pacific Economic Cooperation (APEC) represent the major RTAs. APEC includes Pacific Rim countries included in earlier-mentioned RTAs, such as the US and Canada from NAFTA. APEC is not an official RTA, but it is considered instrumental in fostering greater trade between its members. APEC exhibits consistent and statistically significant RTA effects in all years. Coefficients of 1.45 or 326% in 1995, and 1.32 or 274% in 2006, show strong APEC effects on its members normal bilateral exports. These results are significant in all years at the 1% level. The ASEAN dummy shows a decline in its members bilateral exports. In 1995 ASEAN s dummy coefficient was 1.09 or 196%. It declined to 0.84 or 132% in 1998 and continued to decline further through 2001, to a low of 0.73 or 108%. ASEAN s influence recovered marginally in 2003 but declined to a value of 0.57 or 77% in In all years the coefficients are significant at the 1% level. These results suggest a weakening effect of ASEAN on its members normal trade. A possible explanation for this trend in ASEAN s performance is the East Asia crisis that started in The regional and 91

103 global slow-down may very well have shadowed any positive effects the RTA exhibited during that period. In Latin America and the Caribbean the Andean Community (CAN), Central American Common Market (CACM), Southern Common Market (MERCOSUR), Latin American Integration Association (LAIA), and Caribbean Community and Common Market (CARICOM) represent the major operational RTAs. CAN s RTA dummy coefficients range between 0.77 and 0.18 or 116% and 20%. During the sample period in 1998, CAN s effect was 113%, declining to 75% in It recovered in 2003 to 95%, only to fall drastically to 20% in The results of CAN s RTA effects are significant at the 1% level in 1995, 5% level in 1998, and 10% level in 2003 and are insignificant in 2001 and Similarly, CACM s influence on its members bilateral trade flows degraded progressively over the sample period. Its coefficients range between 0.50 and -1.01, or 65% and -64%, above and below normal trade. In 1995 CACM s RTA dummy reported a value of 0.5 that declined to 0.17 in 1998: 65% compared 18%. CACM s effect became negative from 2001 onwards. Its coefficients declined from to and then to These results are insignificant all years. In the case of MERCOSUR, the dummy coefficients range between 0.11 and 0.99, or 11% and 170%. This RTA s effect declined from 0.15 to 0.11 in 1998 but recovered to 0.36 in In 2003 MERCOSUR s effect peaked at 0.99 or 170%. However, it declined to 0.70 or 101% in MERCOSUR s coefficients are only significant in 2003 at the 10% level. MERCOSUR s regional effect increased over the sample period according to the results reported here. LAIA s dummy also exhibited improved performance over the sample period. Its coefficient took the value of 0.35 or 42% in 1995, only to decline to 0.27 or 31% in However, LAIA s coefficient climbed to 40% in 2001 and to 68% in The dummy coefficient peaked in 2006 at 0.59 or 80% effect on bilateral trade. Unlike MERCOSUR, LAIA s coefficients are significant at the 1% level in most years and 5% in 1998 and In the Caribbean, CARICOM s dummy takes large values that range between 2.24 or 840% and 2.97 or 1851%. CARICOM s performance declines during the middle years of the sample period, where it falls to 2.31 or 907% in 1998 and 2.24 or 839% in 2001, before recovering in 2003 to reach 2.58 or 1220%. The significance of these results at the 1% level in all years suggests that CARICOM plays an important role in the bilateral trade of the region. 92

104 The sample includes three African RTAs: the Common Market for Eastern and Southern Africa (COMESA), Economic and Monetary Community of Central Africa (CEMAC), and West African Economic and Monetary Union (WAEMU). COMESA s dummy coefficients increased from 0.55 to 1.05, or 74% to 187% effect on its members intra-regional trade. During the sample period COMESA s dummy coefficient fell to 0.11 in 1998 and increased to 0.36 and 0.84 in 2001 and 2003, respectively. In most years COMESA s dummy was insignificant, except in 2003 and 2006 where it was significant at the 5% and 1% level, respectively. CEMAC s dummy coefficient ranged from 1.49 to 2.16 during the period. This translates into RTA effects from 345% to 768% above normal bilateral exports. CEMAC also experienced a decline in its performance, as indicated by the coefficients of the dummy variable. It declined in 1998 to 1.71 or 451%, and further in 2001 to 1.49 or 345%. In 2003 CEMAC s effect climbed to 1.83 or 524%. It peaked in 2006 at 768%. These results are statistically significant in all years at the 1% level. WAEMU exhibited strong effects of 1.86 or 542% in 1995, and 1.76 or 481% in During the sample period WAEMU s performance declined, much like its African counterparts. In 1998 and 2001, its effect was between 306% and 312%. In 2003 it climbed to 457%. WAEMU s dummy was significant at the 1% level in all years. WAEMU has a substantial effect on its members bilateral trade flows within the region. In the Middle East and North Africa, a number of RTAs are found. These include the Gulf Cooperation Council (GCC), and Pan-Arab Free Trade Agreement (PAN-ARAB). The GCC exhibited declining effects from 1995 to Its dummy coefficients reflected small RTA effects, with coefficients ranging between 0.12 and , or 13% and -60%. The GCC s performance declined progressively during the sample period to become negative at or -15 % in 1998, or -39% in 2001, and or 39% in In 2006 it reached the weakest effect of -60%. The GCC dummy was not significant in 1995 and However, in 2001 and 2003 the GCC dummy was significant at the 5% level, and at the 1% level in PAN-ARAB displayed an increasing effect on bilateral exports of 0.07 or 7% and 0.40 or 49%. However, PAN- ARAB fluctuated during the period under study, between 0.07 in 1995 and 0.29 in 1998 before falling to 0.07 in In 2003 PAN-ARAB s performance improved to 0.27 or 32% and in 2006 it peaked at 0.40 or 49%. These values are significant at the 5% level only in 1998, 2003, and PAN-ARAB s dummy was not significant in other years. In Central Asia and the former USSR, two main RTAs are included in this sample, the Commonwealth of Independent States (CIS) and Economic Cooperation 93

105 Organisation (ECO). CIS comprises the majority of former USSR states and exhibits substantial effects of the sample period. In 1995 CIS s coefficient took the value of 3.11 or 2,147%, but it declined to 2.29 or 887% in CIS s dummy coefficient rose to 2.50 and 2.70 in 2001 and 2003 respectively. In 2006 it declined to 2.19 or 792%. CIS was significant at the 1% level in all cross-section years. Finally, the ECO exhibited declining RTA effects on its members bilateral trade during the sample period. Its dummy coefficient declined from 1.73 in 1995 to 1.39 in Its coefficient continued to decrease in 2001 and However, it recovered to 0.84 in 2006, which translates to 131%. In 1995, 1998, and 2001 coefficients were significant at the 1% level; they were significant at 10% in 2003 and 5% in The results of the first step of this framework suggest that RTAs play an important role in determining bilateral trade flows between countries. The findings also suggest that RTAs in some developing countries are effective and may have a significant impact on regional trade flows. The main results of Table 4.2 can be contrasted with those presented Table A4.1.1 in Appendix A4.1. Table A4.1.1 uses PPP valued GDP and GDP per capita instead of nominal US dollar values. The results are discussed in the appendix. Alternatively, Table 4.3 reports OLS estimates of Equation (4.9) using the ratio of c.i.f/f.o.b imports and exports as a proxy for transportation cost. This ratio is an alternative measure to distance as a proxy of transportation cost. The values reported in Table 4.3 indicate marginal differences with respect to the elasticities of income of exporter and importer. Income elasticities of exporters in Table 4.3 range from 0.91 and 1.05, lower than the 1.07 and 1.19 reported in Table 4.2. The importing country s income elasticity, in contrast, is greater: Table 4.3 shows it ranging from 0.84 to 0.98, compared to 0.82 and 0.88 in the previous results. The results for per capita income are similar to those reported in Table 4.2. Table 4.3 reports significantly different results for transportation cost where c.i.f/f.o.b ratios are used instead of distance. This transaction cost proxy suggests a weaker effect than geographical distance on bilateral trade. Moreover, common borders or adjacency coefficients more than double compared to the results reported in Table 4.2. The common language dummy remains mainly unaffected by the change of proxy. Closer examination of the RTA dummies shows substantial differences in their coefficients compared to Table 4.2. In the case of the EU, the coefficients range between 1.59 and In Table 4.3 the EU dummy is significant in all years. Similarly, NAFTA s dummy coefficients change from and in Table 4.2 to and in Table 4.3. NAFTA s 94

106 Variable (1) 1995 (2) Table 4.3 Second Set of Estimates of the Gravity Equation (c.i.f/f.o.b Ratios) 1998 (3) Income Y i (0.012) (0.011) (0.011) (0.011) (0.011) Y j (0.011) (0.011) (0.011) (0.011) (0.011) Per capita income C i (0.015) (0.015) (0.015) (0.014) (0.014) C j (0.015) (0.015) (0.014) (0.014) (0.014) c.i.f/f.o.b (0.012) (0.010) (0.009) (0.009) (0.009) Adjacency (0.119) (0.113) (0.114) (0.120) (0.123) Common Language (0.064) (0.060) (0.059) (0.058) (0.057) RTA EU (0.078) (0.076) (0.079) (0.080) (0.082) NAFTA (0.320) (0.315) (0.408) (0.429) (0.570) EFTA (0.415) (0.422) (0.434) (0.422) (0.423) CER (0.117) (0.124) (0.119) (0.122) (0.129) APEC (0.098) (0.100) (0.100) (0.105) (0.119) ASEAN (0.266) (0.264) (0.256) (0.234) (0.261) MERCOSUR (0.385) (0.399) (0.541) (0.596) (0.573) LAIA (0.124) (0.131) (0.136) (0.132) (0.140) CAN (0.371) (0.424) (0.415) (0.406) (0.365) Year 2001 (4) 2003 (5) 2006 (6) Notes: Dependent variable: log of bilateral exports. White heteroskedasticity-consistent standard errors in parentheses. See Appendix for the full names of RTAs. (continued on next page) 95

107 Table 4.3 (Continued) Second Set of Estimates of the Gravity Equation (c.i.f/f.o.b Ratios) Year Variable (1) 1995 (2) 1998 (3) 2001 (4) 2003 (5) 2006 (6) CARICOM (0.251) (0.239) (0.223) (0.231) (0.224) COMESA (0.348) (0.349) (0.332) (0.360) (0.384) CEMAC (0.208) (0.260) (0.245) (0.260) (0.259) CACM (0.785) (0.446) (0.412) (0.363) (0.402) WAEMU (0.293) (0.275) (0.283) (0.323) (0.339) GCC (0.277) (0.260) (0.285) (0.306) (0.328) PAN_ARAB (0.151) (0.128) (0.132) (0.130) (0.134) CIS (0.203) (0.188) (0.190) (0.182) (0.177) ECO (0.275) (0.333) (0.303) (0.317) (0.268) S.E. of regression R Number of Observations 9,418 11,863 11,384 11,863 12,419 Notes: Dependent variable: log of bilateral exports. White heteroskedasticity-consistent standard errors in parentheses. See Appendix for the full names of RTAs. 96

108 coefficients are significant at the 5% level in all years except 1995, when they are not significant. EFTA also shows greater effect in Table 4.3, compared to previous results. In all cases the coefficients of EFTA more than doubled. In this case EFTA remained significant at the 1% level during all years. CER also shows an increase in range of coefficients, to from It is significant at the 1% level in all years. ASEAN s dummy increases in value when c.i.f/f.o.b ratios are used as a proxy for transportation cost. It increased from 1.09 to 2.48 in 1995, and likewise from 0.57 to 2.03 in ASEAN remained significant across all years at the 1% level. APEC s coefficients on the other hand declined in value from 1.45 and 1.32 to 1.19 and APEC s dummy was significant in all years at the 1% level. MERCOSUR, LAIA, CAN and CACM s coefficients all increased, in the results reported in Table 4.3. With the exception of MERCOSUR, Latin American RTAs were significant at the 1% or 5% level during the sample period. CARICOM also experienced a substantial increase in its values, from 2.62 and 2.97 to 4.10 and These results are statistically significant at the 1% level. In Africa, COMESA, CEMAC, and WAEMU experienced increases in their coefficients when c.i.f/f.o.b ratios were used. These results are significant at least at the 5% level in most cases. In the Middle East the GCC and PAN- ARAB s dummy coefficients increased significantly, compared to previous results. The GCC results were significant at least at the 5% level for all years, but insignificant in PAN-ARAB on the other hand was significant at the 1% level in all years. Finally, the Central Asian RTAs CIS and ECO experienced changes in their coefficients, compared to those reported in Table 4.2. These results were significant at the 1% level during all years. The comparison between the results reported in Table 4.2 using distance as a proxy for transportation cost compared unfavourably with those Table 4.3 based on the R-squared statistic alone. Using c.i.f/f.o.b explains bilateral trade flows about 3 to 4 percent better. Furthermore, the results in Table 4.3 are statistically significant in more cases than otherwise. Although the results are in some cases much greater compared to earlier results, they appear to be robust. In the case of geographical proximity, the common border or adjacency dummy plays an increasing role once distance is removed. Figure 4.2 compares the RTA dummy estimates of Table 4.2 and Table 4.3. In most cases using cif/fob places greater emphasis on RTA effects. 97

109 Table Figure 4.2 RTA Dummies Compared 5 4 CEMA CARICOM Y = 1.15X CIS 3 WAEMU 2 ASEAN EFTA CER ECO COMESA EU ARAB 1 LAIA APEC CAN GCC MERCOSUR 0 CACM NAFTA -1 Table Constructed from imports c.i.f values divided by exports f.o.b values, this ratio proxies transportation cost. Compared to physical distance between the capital cities of trading partners, this ratio suggests a significantly weaker effect of transportation cost. Using geographical distance as a transportation cost proxy suggests a one-to-one reduction in bilateral trade. The c.i.f/f.o.b ratio reports half as much reduction in bilateral trade based on transportation cost. Both measures are statistically significant and each tells a different story. When the transportation cost proxy is changed, the adjacency dummy reacts by increasing more than two-fold. Maintaining statistical significance, results in Table 4.3 suggest that sharing a common border is more influential when geographical distance is not considered. This effect does not apply to the common language dummy, however. It remains similar to the results reported in Table 4.2. Figure 4.3 illustrates the correlation between the two proxies used for distance in this paper. The figure shows that as distance increases so, generally, does the c.i.f/f.o.b ratio. The two proxies are loosely correlated, but positive nonetheless. Therefore, our transportation proxy responds positively to distance, indicating increasing transportation costs Comparison with Other Studies The main results in Section conform to trade theory with respect to economic parameters. It is worth noting similar studies that verify the results presented 98

110 log(cif/fob) above. This section will compare the results of the first step of the model with other relevant studies Figure 4.3 Transportation Cost Proxies Compared y = x Correlation Coefficent= log(distance) Table 4.4 presents some comparable studies that considered RTAs effects on bilateral trade. In the case of the economic variables, the model estimates in Table 4.2 and Table 4.3 comply with the correct signs expected in theory and those reported in Table 4.4. This study finds similar GDP effects on bilateral trade compared to Wei and Frankel (1997), Feenstra et al. (2001) and Krueger (1999). Income effects in Table 4.2 range between 1.07 and 1.19 compared to 0.91 and 1.05 in Table 4.3. These studies reported values of 0.96, 1.12 and 0.97, respectively. Even so, the difference is not substantial. In the case of the income of the importing country the results of this paper come close to a number of the studies listed in Table 4.4. Values between 0.82 and 0.88 reported in this paper are close to the values of 0.89 and 0.72 reported by Feenstra et al (2001), Krueger (1999), and Wei and Frankel (1997). The variation in per capita income is larger, however, in comparable specifications. Wei and Frankel (1997) report 0.21 and 0.06 for the exporter and importer countries respectively. This chapter finds values between and 0.09 for exporters, and 0.03 and 0.15 for importers. The values reported here are closer to Wei and Frankel (1997). The distance, adjacency, and common language dummies are similar in most cases to other studies. In the case of distance, this study reports values that range from to -1.14, marginally different from the reported by Rose (2000) or the 1.10 reported by Feenstra et al. (2001). However, this paper also makes use of the c.i.f/f.o.b ratios as a proxy for transportation cost: therefore Table 4.3 cannot be directly 99

111 Variable (1) Income Frankel et al (1995) (2) Table 4.4 Prior Estimates of Elasticities from Gravity Equations Wei and Frankel Rose (2000) Feenstra et al. (1997) (2001) (3) (4) (5) Krueger (1999) (6) Current Study (Averages over years) (7) Y i 0.96 (0.02) 1.12 (0.02) 0.97 (70.32) (0.01) 0.83 (0.01) Y j 0.89 (0.02) 0.72 (0.02) 0.89 (106.86) 0.85 Per capita income C i 0.21 (0.02) 0.41 (24.58) (0.01) 0.73 (0.02) C j 0.06 (0.03) 0.25 (21.77) 0.07 Distance (0.02) (0.05) (0.04) (0.04) (-49.05) Adjacency 0.72 (0.09) 0.42 (0.16) 0.63 (0.18) (0.16) 0.14 (2.56) 0.72 Language 0.47 (0.05) 0.59 (0.08) 0.50 (0.08) 0.69 (0.08) 0.73 (20.89) 0.86 Colony 1.75 (0.15) RTA/FTA 0.67 (0.14) 1.73 (0.11) EC/EU 0.24 (0.09) (0.16) 0.07 (1.08) NAFTA (0.63) (0.98) 0.11 (0.33) ASEAN 1.40 (0.29) 1.80 (0.33) 1.00 (5.52) 0.80 CER 0.50 (1.95) 0.60 MERCOSUR (0.46) 0.78 (0.42) (-0.85) 0.46 EFTA 0.04 (0.30) (0.32) 0.64 APEC 0.61 (0.21) 1.41 Notes: Standard errors in parentheses. NAFTA in column (3) refers to US-Canada FTA only. 100

112 compared to the studies in Table 4.4. Other studies report mixed results with respect to adjacency. However the values from Table 4.2 are comparable to the 0.72 reported by Frankel et al. (1995), and the 0.63 reported by Rose (2000). This, however, is not the case when c.i.f/f.o.b ratios are used. Adjacency takes much larger values, compared with other studies. In the case of common language, this paper finds stronger than usual effects compared with other studies. Table 4.2 reports values as high as 1.03 for language; other studies such as Krueger (1999) reported The estimates of RTA dummies coefficients presented here vary substantially from the literature. For the EU or European Community (EC), the coefficient ranges from to 0.24 in the studies referred to in Table 4.4. Table 4.2 reports values between and -0.01, these are similar to the findings of Wei and Frankel (1997). This study shows similar negative or weak RTA effects with respect to NAFTA, but reports coefficients for NAFTA smaller than other related studies as low as For CER, Krueger (1999) reported a 0.50 dummy coefficient. This study reports values between 0.36 and 0.74, close to the results reported in Table 4.4. For MERCOSUR, the literature reports different values, ranging from to Table 4.2 reports values between 0.11 and 0.99, well in line with these. The results reported for ASEAN are considerably different from those included in Table 4.2. Frankel et al. (1995), and Wei and Frankel (1997) found values of 1.4 and 1.8, higher than the 0.57 and 1.09 found in this study. In the case of EFTA, the results of this study also differ from the literature, with coefficients larger than other studies reported: values between 0.42 and 0.84 are larger than the 0.04 and reported by Frankel et al. (1995) and Wei and Frankel (1997). This study also reports larger coefficients for APEC, with values ranging from 1.32 to 1.45 greater than the 0.61 reported by Frankel et al. (1995). Although the results differ in a number of cases from other studies in the literature, there are commonalities that verify the results of this study. Other RTAs are less commonly tested. However, the interest of this paper is the GCC and the following section will discuss the region s trade patterns. On the whole, this study s findings largely verify previous studies results, implying that the gravity model (4.9) captures RTA effects reasonably well, and in line with the literature. 4.5 Trade Creation and Trade Diversion The results in Table 4.2 and Table 4.3 have broad implications with respect to the trade intensities within RTAs and their welfare. If a RTA increases trade by more than 101

113 100%, is such a trade bloc diverting trade from other countries or regions? To answer this question and contrast the empirical results what what is happening within these RTAs, it is useful to consider the import shares of GDP internally and externally. Consider Figure 4.4, where the intra-regional imports of each RTA are illustrated at the beginning and end of the sample period, 1995 and The figure presents intraregional imports as a percentage of the region s total GDP. An increase from the beginning to the end of the period indicates imports from within the RTA have risen. For most RTAs this indicates increased intra-regional imports. Panel B is a blow up of the cluster of RTAs closer to the origin. This is especially evident for ASEAN, where intra-regional imports jumped by about 20% over the past decade. The EU also experienced increased intra-regional trade, although by a smaller magnitude of less than 10%. The PAN-ARAB RTA exhibits increases, although small, as a share of its GDP. In contrast, NAFTA, CIS, and CARICOM all exhibit declining regional trade. The Caribbean RTA has experienced a decline of about 10% in intra-regional trade. Other RTAs such as CACM, LAIA, GCC, CER, EFTA, ECO, CAN, WAEMU and MERCOSUR show very small proportions of intra-regional trade; in fact, COMESA and CEMAC hardly register trade within the region as a proportion of their GDP. On the other hand, Figure 4.5 illustrates the comparable extra-regional trade of the RTAs included in the model. The values reported in Figure 4.5 show a significantly different picture to that in Figure 4.4. Most of the RTAs trade substantially more with non-members than members. CARICOM, NAFTA, PAN-ARAB, EFTA and GCC trade most with non-members. In the case of NAFTA trade with non-members was more in 2006 than in CACM, APEC, ECO, EU and MERCOSUR also trade more with non-members at the end of the period. COMESA, CAN and WAEMU trade marginally less with nonmembers towards the end of the period. CARICOM s decline in extra-regional imports is most noteworthy: It declines by about 10%. Figure 4.4 and Figure 4.5 give us an insight into which RTAs trade more within themselves and which with non-members, but they do not strictly tell us which RTAs are more biased towards intra-regional trade during the sample period. To determine this, it is useful to examine the ratio of intra-regional trade to extra-regional trade, an approach followed by Sologa and Winters (2001). Such a ratio will indicate whether RTAs are trading more among themselves, at the expense of non-members. 102

114 Figure 4.4 Intra-Regional Imports as Proportion of GDP A. All RTAs º NAFTA EU APEC 0.1 PAN-ARAB CARICOM 0.05 ASEAN CIS B. Sub-set RTAs PAN-ARAB 45º CARICOM ASEAN 0.03 CACM CIS GCC 0.02 MERCOSUR LAIA ECO CAN CER 0.01 EFTA WAEMU CEMAC 0 COMESA

115 Figure 4.5 Extra-Regional Imports as Proportion of GDP A. All RTAs 1 45º PAN-ARAB NAFTA CARICOM EFTA GCC ECO CER ASEAN B. Sub-set RTAs º EFTA 0.35 ASEAN 0.3 GCC ECO CACM CER LAIA CEMAC EU CAN CIS APEC MERCOSUR WAEMU 0 COMESA

116 Figure 4.6 plots these ratios for each of the RTAs included in this study. The ratio can be interpreted as follows: if the quotient of intra/extra regional imports increases, this indicates that a RTA is importing more from its members or less from the rest of the world, or both. In this case the RTA would be suspected of diverting trade. If the ratio declines; the RTA in question either imports less from its members or more from the rest of the world, or both. Consequently the RTA is trade-creating. For instance, take the developed countries RTAs, NAFTA and EU. Panel (i) in Figure 4.6 indicates that over the sample period NAFTA s ratio declined over time. As a result, NAFTA can be considered as a trade-creating RTA. In contrast, the EU s ratio, shown in panel (iv), climbed during the sample period. This increase was more pronounced in 1999 and 2000; in 2005 it fell marginally. Overall the EU appears to be trade-diverting, based on its intra/extra regional imports ratio. This is not the case for APEC, shown in panel (iv), which appears to be trade-creating. The intra/extra regional import ratio indicates a decline over the sample period, which suggests that this grouping was trading more with the rest of the world. In the case of CER there is no marked increase or decrease in the ratio when compared to other RTAs. The lack of significant change in its ratio suggests that CER was neither diverting nor creating trade during this sample period. In Asia the results are mixed. In Central Asia, CIS, illustrated in panel (i), exhibits significant fluctuations during the sample period. Although the overall ratio declined during the sample period, it increased and decreased repeatedly during the period. This indicates episodes of trade diversion, although CIS diverted less trade overall. Compared to CIS, ECO shows less volatility. Panel (viii) indicates that its ratio remained fairly steady during the sample period. However, towards the end of the period it edged upward, suggesting marginal trade diversion. This is similar to the case of the PAN-ARAB RTA, which appears to have diverted trade progressively over the past decade. Overall its intra/extra regional imports ratio climbed during the sample period. So did that of ASEAN, which appears to be leaning towards trade diversion. Its ratio increased steadily over the sample period. The GCC, on the other hand, exhibits trade creation as its ratio declines over time. It is noteworthy that not all ratios are comparable in magnitude. In many instances import ratios are very small. For example the ratios shown for ASEAN, ECO, PAN-ARAB and the GCC indicate that all trade very little within their regions; previous inferences of trade diversion are minimal in these cases. 105

117 Figure 4.6 Intra/Extra Regional Imports Ratio (i) (ii) (iii) MERCOSUR NAFTA LAIA CIS COMESA EFTA (iv) (v) (vi) PAN-ARAB 0.14 CACM 0.14 APEC EU CAN GCC (vii) (viii) (ix) CARICOM ASEAN WAEMU CER ECO CEMAC

118 The Caribbean and Latin America offer the most contrasting picture to those of Asia and North Africa, with their RTAs fluctuating between increased trade diversion and trade creation over time. The ratio for MERCOSUR, for example, shown in panel (ii), climbs at first, indicating trade diversion as its members deal more with each other compared to the rest of the world. As time progresses, MERCOSUR s trade diversion weakens. Towards the end of the period it stabilises. LAIA, in panel (ii), and CAN, in panel (v), exhibit opposite behaviours. Their ratios decline during the first few years of the sample period, indicating trade creation. However, during the last few years they become trade-diversionary. Panel (v) displays CACM s ratio, which in comparison to other Latin American RTAs declines during the first half of the sample period, suggesting trade creation. In the second half it spikes and then declines over time. Overall CACM shows mixed signs of trade diversion and trade creation. The Caribbean RTA CARICOM, illustrated in panel (vii), has an intra/extra regional import ratio that changes marginally over the sample period but towards the latter part of the sample period suggests temporary trade creation effects. Figure 4.6 also illustrates some of the RTA ratios for the African continent. WAEMU in panel (viii) shows intra/extra regional import ratios that suggest greater trade diversion during the sample period. WAEMU appears to concentrate trade within its members, a finding substantiated by the results reported earlier with respect to WAEMU. In contrast, CEMAC s ratio in panel (ix) suggests a trade diversion in the beginning of the sample period, but signs of trade creation after Trade creation weakens during the last few years of the period. COMESA does not exhibit any significant trade creation or trade diversion signals during the period. Panel (iii) shows that only in 2002 does COMESA appear to reduce trade with the rest of the world. 4.6 Trade within the GCC Countries (Step 2) The GCC, created in 1981, has been operational for the past three decades. Recently it has been undergoing significant economic integration progress that warrants attention. Trade liberalisation has gone hand in hand with efforts to enhance capital and labour mobility. What makes the GCC interesting is the political commitment towards economic integration. The GCC follows a standard approach to economic integration in its four main phases: Free Trade Area, Customs Union, Common Market, and Monetary and Economic Union. The GCC possesses many factors that should encourage success and that may not be present elsewhere, such as a common language, close historical and social ties, and a common religion. Moreover, the GCC economies are similar in their 107

119 dependency on oil and in the rapid economic progress they have made in the past three decades. The six member countries, Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and UAE, are considered middle-income or high-income countries based on their GDP per capita. The region is also strategically important, since a large portion of the world s proved oil reserves are located within its borders. In the light of these considerations, two questions need to be answered: First, is the GCC an effective RTA? Second, what effect does it have on trade patterns within the region? The answer to the first question lies in the empirical analysis that has already been presented. In Table 4.2 and Table 4.3, the RTA dummy of the GCC indicates the contribution the arrangement makes to intra-regional trade. The gravity equation implies effects between -60% and 180%, but how well does the gravity model fit the GCC trade flows specifically? The observed, fitted, and residuals matrices of the GCC s intra-regional trade are reported in Table 4.5 for the years 1995 and In 1995 the model understates the actual trade flow values in most cases. Take the case of Bahrain s exports to Saudi Arabia in Panel (i) indicates that Bahrain exported US$ 251 million to Saudi Arabia. The model predicts that Bahrain exported US$ 27 million indicated in panel (ii). The residual is US$ 225 million, the difference between the two values. The case is quite the opposite when the year 2006 is considered. Understimation is more pronounced, as indicated by the residuals. This is illustrated by a number of psotive values in panel (vi). The most severe case is Bahrain; its trade flows are underestimated in the model by US$ 4,975 million in Thus, the gravity model in step 1 does not predict trade flows well for the GCC. Since the interest of the paper is to understand the trade patterns within the GCC, further exploration of such trends is required. A discussion of intra-trade shares within the region will be helpful. A better picture of the GCC s intra-regional trade is given by considering imports as a ratio of total trade. Table 4.6 reports the 1995 and 2006 intra-gcc trade matrices expressed as percentages of the total. Table 4.6 gives a vivid picture of who trades more with their GCC counterparts. Bahrain and Oman trade the most within the GCC. Overall, the GCC region appears to trade less within itself over the sample period. What can explain the stagnation in trade flow within the region? Transportation costs 108

120 Importer Exporter Bahrain Kuwait Oman Qatar (i) Observed Table 4.5 GCC Trade Matrices, 1995 and 2006 (Millions $US) Saudi Arabia UAE Total Bahrain Kuwait Oman Qatar 109 (iv) Observed Bahrain ,594 Kuwait Oman ,478 1,935 Qatar ,348 Saudi Arabia 1, ,350 3,591 4,534 1, ,260 9,028 UAE , , ,703 5,464 Total 1, , ,269 5,016 1,901 3,079 1,592 2,992 5,470 (ii) Fitted (v) Fitted Bahrain Kuwait ,416 Oman Qatar ,152 Saudi Arabia , ,441 3,914 UAE ,155 2,469 Total , ,863 1,863 2,809 (iii) Residual (vi) Residual Bahrain Kuwait Oman , Qatar Saudi Arabia 1, ,197 3,163 4, ,477 UAE , , ,995 Total 1, , ,033 4, , ,195 3,013 Saudi Arabia UAE Total

121 cannot be expected to produce such low overall trade proportions. Distance within the region is not so prohibitive that it would reduce trade substantially; it barely exceeds one thousand kilometres between most capital cities. Exporter Importer Table 4.6 GCC Import Ratios of Total Trade (Percentages) Bahrain Kuwait Oman Qatar Saudi Arabia UAE 1995 Bahrain Kuwait Oman Qatar Saudi Arabia UAE TOTAL Bahrain Kuwait Oman Qatar Saudi Arabia UAE TOTAL Source: Author s calculations The answer may lie in the similarity of the economic structures of these economies. The common dependence on oil for developmental purposes and supporting large government sectors may lead to low intra-regional trade. These six countries are traditionally open economies, importing most of their consumer and capital needs from outside the region. Moreover, the non-complementary products produced within the region may be an influencing factor, eliminating the possibility of trade in different goods. Most countries within the region have a low level of diversification of exports products, so that intra-industrial trade is not particularly viable. This is true of most Arab countries, the GCC included (Al-Atrash and Yousif 2000). To understand the intra-regional trade patterns of the GCC region, disaggregated commodity trade analysis is useful to contrast with the patterns suggested by Table 4.6. This is the second step in the model s analysis; the first was estimating bilateral trade determinants using GDP and GDP per capita as proxies for size and economic 110

122 development. This second step involves estimating bilateral trade within the GCC, conditional on total intra-regional trade. To do this, the GCC s total trade is divided into ten 1-digit categories according to SITC revision 1 obtained from the UN Comtrade database. These categories are: (1) animal and vegetable oils and fats, (2) beverages and tobacco, (3) chemicals, (4) crude materials except fuels, (5) food and live animals, (6) machinery and transport equipment, (7) manufactured goods, (8) miscellaneous manufactured articles, (9) commodities and transactions not classified, and (10) mineral fuels and lubricants. The choice of this classification was based on the data available for the GCC countries between 1980 and Intra-regional bilateral trade exports were regressed on total bilateral trade, distance, adjacency and country/commodity combinations. Saudi Arabia is used as a base country, given its dominant size within the GCC, and mineral fuels were considered the base commodity. Mineral fuels, which include oil and petroleum extracts, form a substantial proportion of extra-gcc trade. Although there exists some regional trade within mineral fuels, emphasis here will be on other commodity categories. With this specification, there are 45 dummy variables representing Bahrain, Kuwait, Oman, Qatar and UAE across nine commodity categories. Table 4.7 presents the results of the second step of the model based on equation (4.12). The model suggests a significant role that total trade within the GCC plays in determining bilateral trade in the product groups. On average an increase of 1% in total trade of exports within the GCC is expected to increase exports of any specific product group by 0.6%. Similarly, an increase of 1% in total trade of imports within the GCC increases bilateral exports of specific product groups by 0.3%. The model also suggests that distance plays a lesser role in reducing trade within the GCC, compared to the overall model in Step 1; a -0.5 elasticity is reported in Table 4.7. Adjacency, however, exhibits strong effects within the GCC, promoting trade almost four times above the norm. Step 2 of the model included a number of country and commodity dummies. The coefficients of these dummies are reported in Table 4.8. Table 4.8 shows a number of interesting patterns in the GCC intra-regional trade. Firstly, although Saudi Arabia is the largest economy in the region, it trades less with its GCC neighbours in many of the selected commodity groups. Saudi Arabia trades more, compared to its neighbours within the region, in as many as half the commodities groups including animal and vegetable oils and fats, beverages and tobacco, crude materials, and commodities and 111

123 Variable Table 4.7 Disaggregated Trade Estimates P ' logx =β +β LogX +β LogX +β Adj +Σ γ (Country/Commoditydummies) +ε ij 1 2 i 3 j 4 ij k k ij k 112 Coefficient Total Trade of Exporter (0.027) Total Trade of Importer (0.026) Distance (0.062) Adjacency (0.075) R Number of Observations 4,455 Note: White standard errors in parentheses. other transactions. Bahrain, Kuwait, Oman, Qatar and the UAE trade more within the GCC in chemicals, food and live animals (except Qatar), machinery and transportation, manufactured goods, and miscellaneous manufactured goods, relative to Saudi Arabia, although they do not trade equally in each commodity category. In the case of animal and vegetable oils and fats, Oman trades most within the GCC compared to Bahrain, Kuwait, Qatar, and the UAE, but less than Saudi Arabia by 25% [=(e )x100]. Qatar trades the least within the region; 97% less than the base case. Bahrain, Kuwait, and the UAE s intra-gcc trade in this category falls below Saudi Arabia by 64% to 89%. These results suggest that trade within the GCC of animal and vegetable oils and fats is less than trade in mineral fuels. With the exception of Oman, other GCC countries trade less than base case Saudi Arabia with respect to beverages and tobacco. Oman trades approximately 27% more than Saudi Arabia in this category. Qatar trades the least in beverages and tobacco is, 95% less than the base case. Bahrain, Kuwait and UAE trade 62%, 41% and 18% less than the base case in this category. The GCC countries also trade less intraregionally compared to the base case in the crude materials category. Here, Kuwait trades most after Saudi Arabia, at 27% less; then, the UAE, Bahrain, Oman and finally Qatar. Other categories in which the other GCC countries trade relatively more than Saudi Arabia are chemicals, food and live animals, machinery and transportation, manufactured goods, and miscellaneous manufactured goods. In chemicals, Qatar trades most within the GCC relative to Saudi Arabia. The amount of trade within the GCC of Qatar is in the order of 837% more than the base case. Qatar is followed closely by Kuwait, then the UAE, Oman, and finally Bahrain. Saudi Arabia trades the least in chemicals within the region. Trade in food and live animals is primarily led by Oman,

124 Table 4.8 Estimate of Country-Product Dummy Variable Coefficients Importer Commodity Group Bahrain Kuwait Oman Qatar UAE 1. Animal Fats etc (0.235) (0.242) (0.213) (0.304) (0.334) 2. Bev. & Tobacco (0.203) (0.230) (0.193) (0.259) (0.317) 3. Chemicals (0.203) (0.214) (0.195) (0.233) (0.287) 4. Crude Materials (0.209) (0.227) (0.202) (0.237) (0.287) 5. Food & Animals (0.203) (0.213) (0.192) (0.233) (0.287) 6. Mach. & Tans (0.203) (0.214) (0.192) (0.233) (0.293) 7. Manufac. Goods (0.203) (0.213) (0.192) (0.233) (0.287) 8. Misc. Manufac (0.203) (0.213) (0.192) (0.233) (0.290) 9. Other (0.216) (0.242) (0.192) (0.254) (0.338) Notes: Dependent variable: log of bilateral disaggregated exports from i to j. White heteroskedasticity-consistent standard errors in parentheses. 113

125 followed closely by Kuwait, then UAE, Bahrain, Saudi Arabia, and Qatar. Oman trades 798%, or eight times more than the base case, while Qatar trades about 13% less than Saudi Arabia. In the category of machinery and transportation, Kuwait leads the GCC countries in intra-regional trade, while Oman and Bahrain come close after. Qatar and UAE fall within the bottom half, but still above the base case. Manufactured goods trade patterns shift in favour of Bahrain, trading the most in the region relative to the base case. Its trade flows are mirrored by Qatar and Kuwait to make up the top half of the GCC in this category. The UAE exceeds Oman relative to the base case. All five countries trade relatively more than the base case. In the miscellaneous manufactured goods category Kuwait trades most within the region; followed by Bahrain, the UAE, and then Oman. Qatar completes the top five countries. Again, Saudi Arabia trades the least within the GCC in this category. Finally, Saudi Arabia exceeds other GCC countries in intra-regional trade in the commodity and other transactions category. After Saudi Arabia, Oman trades most in this category, followed by the UAE. Bahrain, Kuwait, and Qatar trade similarly in these commodities relative to the base case. The results of this analysis indicate there is no clear pattern within the region where one country dominates intra-regional trade within the GCC. The second step of the model used here explains the trade patterns within the GCC, and shows that there are substantial differences between the largest economy, Saudi Arabia, and other countries in intraregional trade. Saudi Arabia trades more in animal and vegetable oils and fats, beverages and tobacco, crude materials, and commodity and other transactions, but less in chemicals, food and live animals, machinery and transportation, manufactured goods, and miscellaneous manufactured goods. In these particular categories there is no clear leader in intra-regional trade. This position switches between Bahrain, Kuwait, Oman, Qatar, and UAE. Overall, the results suggest that the above categories are traded more than mineral fuels within the region. 4.7 Conclusions Using a two-step framework, the gravity model has been utilised to analyse trade flow patterns between different countries in the world. The aim in the first step was to measure the effect of RTAs on bilateral trade flows. Comparing a significant portion of existing RTAs in the world, the model suggests that a number of RTAs in developing 114

126 countries explain a large portion of bilateral trade flow once economic factors are accounted for. RTAs play a significant role in the determination of trade flow between countries. Findings suggest that geographical proximity, shared borders, and language significantly influence bilateral trade flows. These findings are congruent with previous literature. It is in developing countries RTAs that this research has found large and significant effects on bilateral trade flows. Trade blocs in the Caribbean, Africa and Central Asia show RTAs strongly affect their trade patterns. Latin American, Middle Eastern, and North African RTAs remain less effective on bilateral trade flows. This chapter also investigated trade creation and trade diversion of the RTAs in question, with mixed results. A number of RTAs such as NAFTA, APEC, EFTA and CEMAC show relative signs of trade creation. On other hand, RTAs such as the EU, ASEAN, WAEMU, CAN, COMESA and PAN-ARAB show strong signs of trade diversion, although in a number of cases intra-regional imports are small enough that trade diversion is minimal. As seen from Step 1 in the model, the GCC appears to have no significant effect on its members bilateral trade. This is confirmed by examining the import ratios of these six countries, as illustrated in Table 4.6. GCC intra-regional import ratios are low for most of the countries at the beginning and the end of the period, 1995 and The second step of the framework used in this chapter identified the intra-regional trade patterns of the GCC region. In Step 2 the model quantifies commodity-specific interactions in GCC intraregional trade. The largest economy in the region, Saudi Arabia, dominates in less than half of the commodity groups. The other five countries, Bahrain, Kuwait, Oman, Qatar and the UAE, exceed Saudi Arabia s intra-regional trade in more than half of the cases. However, there is no clear trend of the other five countries exceeding the base case. Step 2 in the model also suggests that there is some degree of mineral fuel trade within the region despite the similarities of these oil dependent economies. Although the GCC countries have been undergoing economic integration for the past three decades, the RTA has not intensified intra-regional trade during the sample period. This is evident from the analysis from step 1 and the intra/extra-regional import ratios. New development in the region such as the launch of a customs union have yet to bear fruit. This may be attributable to the similar economic structures of all six economies, and their dependence on natural resources; the non-complementary products produced 115

127 within the region and the low level of diversification of export products may also be explanatory factors. Despite small trade volumes within the region, a number of countries such as Bahrain and Oman carry out a significant portion of their total trade within the region. Finally, the model suggests that a common border between these countries and other GCC countries plays an important role in intra-regional trade. 116

128 Appendix A4.1 Sensitivity Analysis Alternative results to the OLS regression of the model (4.9) are reported in Table A4.1.1 using PPP valued GDP and GDP per capita. Table A4.1.1 uses distance as a proxy for transportation cost. Accordingly if we contrast the results with those reported in Table 4.2 we note a number of differences. With respect to income the values are comparable to those reported in Table 4.2. Take column (6) in Table 4.2 as an example, the elasticity of the exporter s income in 2006 is 1.19; 1.17 was reported in Table A The elasticity of the importing country s income is 0.88 in 2006 as reported by Table 4.2; Table A4.1.1 reports an elasticity of The per capita income coefficients, however, are not comparable across the tables. In Table A4.1.1 the coefficients for per capita income are greater than those reported in Table 4.2 for all years. In 1995, for example, Table 4.2 reports per capita income elasticity for the exporter of 0.09, compared to 0.61 in Table A The same is true of the importer s per capita income elasticity: Table 4.2 reports a value of 0.10, compared to 0.50 in Table A With respect to cultural and geographical variables, Table 4.2 and Table A4.1.1 do not differ substantially. In the case of distance, Table 4.2 reports a value of for 2006, while Table A4.1.1 reports a value of The adjacency dummy is marginally greater in Table 4.2 compared to Table A This is more so in the years 2001, 2003, and 2006, where the coefficients of the dummy variable take the values of 0.71, 0.79, and 0.82 in Table 4.2. In contrast, Table A4.1.1 reports values of 0.66, 0.69, and 0.72 for the same years. The common language dummy s coefficients are not significantly different when Table 4.2 and Table A4.1.1 are compared. In 1995, Table 4.2 reports a coefficient of 0.75, a smaller value compared to the 0.85 in Table A For the rest of the years Table A4.1.1 s language coefficients remain greater than Table 4.2 s; but the difference is not significant. The main difference between Tables A4.1.1 and 4.2 is the RTA dummies coefficients. Take the EU for example: Table 4.2 reports a dummy coefficient of for 2006, which translates to -14% effect below normal trade. In contrast, Table A4.1.1 reports a coefficient of 0.46 that reflects a 59% RTA effect above normal trade. The difference is large and significant. Other RTAs also experience significant change in their effects when PPP based values are used. The GCC is another example where the use of PPP based value 117

129 changes its coefficients. In the year 1995, Table 4.2 reports a coefficient of 0.12, which translates to 13% RTA effect above normal trade. In Table A4.1.1 the corresponding value is -0.28, equivalent to a -24% effect below normal trade. 118

130 Variable (1) Income (PPP) Table A4.1.1 Estimates of The Gravity Equation (PPP Adjusted Income) Year 1995 (2) 1998 (3) Y i (0.011) (0.011) (0.011) (0.010) (0.010) Y j (0.011) (0.011) (0.011) (0.010) (0.011) Per capita income (PPP) C i (0.021) (0.020) (0.021) (0.020) (0.020) C j (0.020) (0.020) (0.019) (0.019) (0.019) Distance (0.024) (0.024) (0.024) (0.023) (0.024) Adjacency (0.119) (0.109) (0.111) (0.109) (0.114) Common Language (0.061) (0.059) (0.056) (0.056) (0.056) RTA EU (0.076) (0.077) (0.079) (0.080) (0.080) NAFTA (0.489) (0.504) (0.569) (0.578) (0.676) EFTA (0.445) (0.481) (0.496) (0.459) (0.460) CER (0.114) (0.134) (0.134) (0.136) (0.151) APEC (0.092) (0.100) (0.097) (0.096) (0.108) ASEAN (0.268) (0.308) (0.261) (0.245) (0.269) MERCOSUR (0.336) (0.303) (0.449) (0.486) (0.496) LAIA (0.133) (0.132) (0.136) (0.142) (0.147) CAN (0.246) (0.349) (0.367) (0.371) (0.269) CARICOM (0.232) (0.220) (0.202) (0.201) (0.199) COMESA (0.358) (0.340) (0.367) (0.349) (0.327) 2001 (4) 2003 (5) 2006 (6) (continued on next page) 119

131 Table A4.1.1 (Continued) Estimates of The Gravity Equation (PPP Adjusted Income) Year Variable (1) 1995 (2) 1998 (3) 2001 (4) 2003 (5) 2006 (6) CEMAC (0.173) (0.200) (0.206) (0.165) (0.163) CACM (0.741) (0.622) (0.562) (0.803) (0.802) WAEMU (0.259) (0.330) (0.356) (0.334) (0.362) GCC (0.211) (0.236) (0.227) (0.227) (0.228) PAN_ARAB (0.152) (0.172) (0.160) (0.153) (0.147) CIS (0.185) (0.203) (0.193) (0.182) (0.245) ECO (0.269) (0.314) (0.314) (0.372) (0.360) S.E. of regression R Number of Observations 11,399 12,751 13,781 14,051 14,920 Notes: Dependent variable: log of bilateral exports. White heteroskedasticity-consistent standard errors in parentheses. See Table A4.3for the full names of RTAs. 120

132 Appendix A4.2 Data Sources and Description This appendix describes the sample and data used in this chapter. The appendix indicates which countries are included in the sample in Table A Table A4.2.2 describes the RTAs included in the analysis. It lists their members and years of inception. The data variables, their units of measurement, and their sources are shown in Table A

133 Table A4.2.1 Countries Included in Gravity Model Sample Angola Equatorial Guinea Madagascar Slovakia Argentina Estonia Malawi Slovenia Armenia Ethiopia Malaysia Solomon Islands Australia Fiji Maldives South Africa Austria Finland Mali Spain Azerbaijan France Malta Sri Lanka Bahamas Gabon Mauritania St. Kitts And Nevis Bahrain Gambia Mauritius St. Lucia Bangladesh Georgia Mexico St. Vincent and the Grenadines Barbados Germany Moldova Sudan Belgium Ghana Mongolia Suriname Belize Greece Morocco Sweden Benin Grenada Mozambique Switzerland Bolivia Guatemala Myanmar Syria Brazil Guinea Nepal Tajikistan Brunei Darussalam Guinea-Bissau Netherlands Thailand Bulgaria Guyana New Zealand Togo Burkina Faso Haiti Nicaragua Tonga Cambodia Honduras Niger Trinidad and Tobago Cameroon Hong Kong Nigeria Tunisia Canada Hungary Norway Turkey Cape Verde Iceland Oman Turkmenistan Central African Republic India Pakistan Ukraine Chad Indonesia Panama United Arab Emirates Chile Iran Papua New Guinea United Kingdom China Ireland Paraguay United States Colombia Italy Peru Uruguay Comoros Jamaica Philippines Uzbekistan Costa Rica Japan Poland Venezuela Cote D Ivoire Jordan Portugal Yemen Croatia Kazakhstan Qatar Zambia Cyprus Kenya Russia Czech Republic Kiribati Rwanda Denmark Korea Samoa Djibouti Kuwait Sao Tome and Principe Dominica Laos Saudi Arabia Dominican Republic Latvia Senegal Ecuador Lebanon Seychelles 122

134 Table A4.2.2 Regional Trade Agreements Trading Blocs Created Members ASEAN Association of South East Asian Nations (AFTA) CER Closer Trade Relation Trade Agreement EU European Union GCC Gulf Cooperation Council NAFTA North American Free Trade Agreement MERCOSUR Southern Common Market APEC Asia Pacific Economic Cooperation EFTA European Free Trade Area CARICOM Caribbean Community and Common Market 1994 Indonesia, Malaysia, Philippines, and Thailand 1983 Australia and New Zealand 1957 (1992) Belgium (1957), Luxembourg (1957), France (1957), Germany (1957), Greece (1981), Italy (1957), Netherlands (1957), Denmark, Ireland, United Kingdom (1973), Greece (1981), Portugal, Spain (1986), Austria, Finland, Sweden (1995) 1981 Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and United Arab Emirates 1989 Canada, United States, and Mexico Argentina, Brazil, Paraguay, and Uruguay 1989 (1989)Australia, Brunei Darussalam, Canada, Indonesia, Japan, Malaysia, New Zealand, Philippines, Korea, Singapore, Thailand, United States (1991), China, Hong Kong (China), Taiwan (China) (1993), Mexico, Papua New Guinea, (1994) Chile, (1998)Peru, Russia, Vietnam Iceland, Norway, Switzerland 1973 (1973)Antigua and Barbuda, Barbados, Jamaica, St. Kitts and Nevis, Trinidad and Tobago, (1974) Belize, Dominica, Grenada, Montserrat, St. Lucia, St. Vincent and the Grenadines, (1983) The Bahamas (only part of the Caribbean Community, not the common market). (Continued on next page) 123

135 Table A4.2.2 (Continued) Regional Trade Agreements Trading Blocs Created Members COMESA Common Market for Eastern and Southern Africa CACM Central American Common Market CIS Commonwealth of Independent States ECO Economic Cooperation Organization 1993 Angola, Burundi, Comoros, Djibouti, Egypt, Ethiopia, Kenya, Lesotho, Malawi, Mauritius, Mozambique, Rwanda, Somalia, Sudan, Swaziland, Tanzania, Uganda, Zambia, Zimbabwe El Salvador, Guatemala, Honduras, Nicaragua, Costa Rica (1962) 1991 Azerbaijan, Armenia, Georgia, Moldova, Kazakhstan, Russian Federation Ukraine, Uzbekistan, Tajikistan 1985 Azerbaijan, Iran, Kazakhstan, Pakistan, Tajikistan, Turkey, Turkmenistan, Uzbekistan LAIA Latin American Integration Association Mexico, Argentina, Bolivia, Brazil, Chile, Ecuador, Paraguay, Peru, Uruguay, and Venezuela. PAN-ARAB 1997 Algeria, Jordan, Egypt, Lebanon, Morocco, Syria, Sudan, Tunisia WAEMU West African Economic and Monetary Union Benin, Burkina Faso, Cote D Ivoire, Mali, Mauritania, Niger, Senegal, Togo, and Guinea-Bissau (1997) 124

136 Data Series Bilateral Trade Data Description Annual bilateral exports valued in million US Dollars between each pair of trading partners Table A4.2.3 Data Sources Source Direction of Trade Statistics (DOTS) IMF (2007) GDP GDP per Capita Distance GDP valued at PPP nominal billion US Dollars GDP per capita valued at PPP nominal US Dollars Straight-line distance between capital cities in each country in the sample measured in kilometres. World Economic Outlook (2006) International Monetary Fund (IMF) World Economic Outlook (2006) International Monetary Fund (IMF) Centre D études Prospectives et D Informations Internationales < Adjacency Dummy variable 1 if countries share a common border Centre D études Prospectives et D Informations Internationales < Common Language Trade Bloc 0 if countries do not share a common border Dummy variable 1 if country pair speak the same language 0 if country pair do not speak the same language Dummy Variable 1 Country pair are both members of the same RTA 0 Country pair are not both members of the same RTA Centre D études Prospectives et D Informations Internationales < World Trade Organization < region_areagroup_e.htm> 125

137 Appendix A4.3 Economic Foundation of the Gravity Model This appendix, based on the work of Feenstra (2002), sets out the details of the derivation of the gravity model. Given that utility function is based on constant elasticity of substitution (A4.1) H (σ-1)/σ U j= N i(c ij), i=1 Where N i is the varieties imported from country i, c ij is the quantity consumed of the variety and σ is the constant elasticity of substitution. This utility can be maximised subject to the budget constraint or income (A4.2) Y H N p t c j i i ij ij i 1, where Y i is the income of country j, which is equivalent to its spending on other countries goods, p i is the price of the imported good, t ij is the transportation cost from country i to country j. In order to maximise the utility in equation (A4.1) the Lagrange multiplier method is utilised. Combining (A4.1) and (A4.2) to form the Langrangian function, we get: (A4.3) H H (σ-1)/σ i ij i i ij ij j i 1 i=1 L U N (c ) λ( N p t c -Y ). j Partially differentiating with respect to and c ij and equating to zero yields; -1 L (c ij) (A4.4) = =k, c p t ij i ij (A4.5) L H i=1 N p t c = Y i i ij ij where k 1. From (A4.4) we obtain an expression for c ij ; (A4.6) -σ c ij=(kpit ij). Substituting (A4.6) into (A4.5) gives us 126

138 -σ (A4.7) ( N p t (p t ) )k=y.. H i=1 i i ij i ij j The constant (A4.8) k is identified as Y -σ j k=k =. H 1-σ N i(pit ij) i=1 Given (A4.7) and (A4.8), an expression for demand can be derived in the form of (A4.9) c =k(p t ) ij = i ij -σ Y (p t ) - j i ij 1 i( pt i ij ) N (p t ) Y i ij j = x P - Pj j. The price index P j can be represented as (A4.10) H 1 1-σ 1-σ P j= N i(pit ij). i=1 It is thus easily verifiable that the denominator in (A4.9) is a simple manipulation of (A4.10). Since the demand for imported goods of country j is identified in (A4.9) it is possible to then derive an expression for export flows from country i to country j. Let the export flow from country i to country j be denoted as This identity simply states that the value of exports of country i to country j is equivalent to the product variety sold, its price, and the quantity consumed. Substituting (A4.9) into the identity allows us to derive an expression for flow of exports from country i to country j: (A4.11) X = N p t c ij i i ij ij (pi t ij ) Y j = Nipi 1- P j (p t ) = N Y 1- i ij i j 1- Pj pt 1- i ij = NiY j. P j X ij Ni pi tij c ij. 127

139 The expression (A4.11) in essence is the gravity equation of trade. Country i s income can be included in expression (A4.11) using Yi Ni pi y. Where y is the fixed firm s output which is produced based on the assumption of zero profits in a monopolistic competitive market (Krugman 1981). Substituting the expression into (A4.11) results in (A4.12) X ij Y i Y j p i y = Y iy j p i y p i t ij P j t ij P j 1 1. To estimate (A4.12) we take its logs: Note the this expression is identical to Equation (4.7) in section 4.2. For estimation purposes, however, transportation cost is not observed and thus is proxied for by distance or c.i.f/f.o.b ratios. 128

140 CHAPTER 5 HOW MUCH CONVERGENCE? 5.1 Introduction In 1981, five Gulf countries, Bahrain, Kuwait, Oman, Qatar and Saudi Arabia, together with the United Arab Emirates, agreed to coordinate their economic, political and social policies. To achieve this they established the Gulf Cooperation Council (GCC). Since then, the Council has initiated a major economic integration programme aimed at unifying the policies of these countries in order to assimilate their economies. The GCC countries have followed a traditional approach to economic integration, progressing from a free trade area to a customs union, followed by a common market and, finally, a monetary union. Convergence of these six economies is essential to facilitate successful economic integration. The central hypothesis of this chapter stipulates that the GCC economies demonstrate convergence with respect to growth and income. Convergence is a byproduct of the similarities between the GCC economies and their economic integration, and the validity of this claim will be tested using a number of distinct approaches to convergence. This chapter aims to answer two main questions: are the GCC economies converging with respect to economic growth; and does the economic integration of the GCC play a role in the convergence of the six economies incomes? This chapter analyses the convergence of the GCC economies based on economic growth and income. The analysis aims to determine the degree to which the GCC countries have converged, and the implications of this for their economic integration agendas. The chapter will progress as follows: section 5.2 presents an overview of the stages of economic integration and the GCC s experience to date. Section 5.3 presents basic macroeconomic indicators of the six member countries. Section 5.4 defines convergence in the context of a neoclassical framework. Section 5.5 presents an alternative view of convergence in terms of dispersion of growth, through which the GCC is compared to other regions. Section 5.6 analyses convergence using deviationsfrom-the-mean, it tests GCC-specific income convergence. Using this technique the GCC s experience is also compared with that of the EU. The chapter concludes with a summary of findings and a discussion of their implications. 129

141 5.2 Economic Integration Economic integration can be described as a process where restrictions to movement of goods, capital and labour are reduced or removed completely. The process involves harmonisation of laws, standards and regulations. In his Glossary of International Economics, Alan Deardorff (2000) defines it as follows: Economic integration refers to reducing barriers among countries to transactions and to movements of goods, capital and labour, including harmonisation of laws, regulations, and standards. Common forms include FTAs, customs unions, and common markets. Sometimes classified as shallow integration vs. deep integration. The process of economic integration, as suggested by the definition, is a complex and multifaceted process. It can take a number of forms. Deardorff s (2000) definitions of the two distinct approaches regarding integration illustrate this point: shallow integration leads to reduction or elimination of tariffs, quotas, and other barriers to trade in goods at the border, such as trade limiting customs procedures. In contrast, deep integration is economic integration that goes well beyond removal of formal barriers to trade and includes various ways of reducing the international burden of differing national regulations, such as mutual recognition and harmonization Stages of Economic Integration Whether shallow or deep, economic integration usually follows four sequential initiatives: Free Trade Areas (FTAs), customs unions, common markets and monetary union. Each stage increases the depth of economic integration between the countries involved. Each stage requires greater commitment from the participants, and further harmonisation and standardisation. In order to appreciate this progression a description of each stage is useful: 1. Free Trade Areas (FTAs) Within FTAs, barriers to goods movement are reduced but members maintain individual tariff schedules against goods imported from third countries. Operation is effective if FTAs requires certificates of origin to verify the source of imports. These documents distinguish goods produced within or outside the area. 2. Customs Unions In custom unions all tariffs are abolished and no restriction to trade exists, as with FTAs. In this case, however, customs unions members levy 130

142 identical tariffs on all third countries imports. A unified tariff schedule is adopted by all members. 3. Common Market Common markets are a progression from customs unions. Not only are goods allowed to move freely, but so are factors of production. Free movement of labour and capital across borders is facilitated, ensuring that equal treatment of factors of production applies across borders. 4. Monetary Union A monetary union is the culmination of the economic integration process. Once goods and factors of production move freely, members of a union seek to unify their national currencies. Members relinquish their national currencies and a common currency is adopted across political borders as legal tender. This stage requires national governments to surrender their sovereignty over monetary policy to a supranational central bank. These descriptions of the stages of economic integration provide a broad framework that requires considerable political will and resolve to put into effect. There are a number of examples of regions that fall within the stages discussed above. The most common example of a FTA is the North American Free Trade Area (NAFTA). The European Union (EU) has also made extensive progress along the outlined stages of integration. In 1999 the European Union launched a common currency the Euro although this has yet to be adopted by all member countries. Other examples exist of integration processes, such as the Association of Southeast Asian Nations (ASEAN), the Caribbean Community (CARICOM), the Economic and Monetary Community of Central Africa (CEMAC), the Mercado Comun del Sur (MERCOSUR) and the Gulf Cooperation Council (GCC). In each case there are varying degrees of integration along the four main stages. For example, CARICOM operates a common market, but MERCOSUR has yet to revive its FTA. The GCC recently launched a customs union (in 2003), and introduced a common market in Various developments within different economic regions suggest varying degrees of success in establishing operational and effective economic unions. 5.3 The Gulf Cooperation Council Officially known as the Cooperation Council for the Arab States of the Gulf (GCC), the countries making up the GCC share a number of common characteristics including language, religion, and cultural customs, and economic structures in which 131

143 resource exports are pivotal to growth. Most of the GCC countries are members of the Organisation of Petroleum Exporting Countries (OPEC). Given their mutual characteristics, they set out to coordinate their policies and generate integration with each other. Table 5.1 provides a brief summary of the main economic and demographic descriptive statistics of the GCC, comparing the years 1981 and The GCC s output has more than trebled in the past twenty-five years, and the mean GDP per capita has increased by 65% despite high population growth the population has more than doubled in the past two decades. The most populous country is Saudi Arabia where, in 2006, 23 million people lived. The least-populated country within the GCC is also the smallest: Bahrain, with a total of 750,000 people in However, the UAE has the fastest-growing population, more than tripling over the last twenty-five years. Table 5.1 Basic Economic Indicators of the GCC Measure of size GDP (PPP) Billion $US GDP per Capita (PPP) $US 14,215 23,548 Population (Million) Source: World Economic Outlook and author calculations. According to World Bank classification, countries with a Gross National Income (GNI) per capita between $3,596 and $11,115 are deemed upper middle-income. 1 Countries with a GNI per capita greater than $11,116 are classified high-income. On average, the GCC countries are considered to be upper middle to high-income economies. Based on GNI per capita in 2006, Oman falls in the upper middle-income category, while Bahrain, Kuwait, Qatar, Saudi Arabia and the UAE are considered as high-income countries. The progress of the GCC s economic integration has followed the four stages discussed above. In 1983 the GCC launched its FTA to help reduce trade restrictions between member countries. It did not progress to the next step until 2003, when it initiated a customs union whereby all remaining restrictions between member countries on movement of goods were removed and unified tariffs against non-members were established. In 2008 the GCC launched its common market, allowing greater mobility of capital and labour within the region. Finally, in 2010, the GCC plans to launch a common currency and create a GCC central bank. 1 World Bank (2008a). 132

144 Convergence is paramount if economic integration is to be successful. This is especially true in the last stage, where the GCC countries will adopt a common currency. The absence of convergence may prove problematic in maintaining a credible monetary union. This is particularly true if the economies involved are not behaving similarly, where macroeconomic shocks can affect each country differently. Convergence in a broader perspective can ensure greater uniformity with respect to unforeseen shocks and disturbances. 5.4 Convergence This section will apply the neoclassical approach to growth and convergence using the concepts of absolute and conditional convergence. The section will incorporate RTAs in the analysis to examine their effect on convergence within such regions Absolute Convergence Absolute convergence is a proposition of the Solow growth model and its variants to explain economic growth and its determinants. It is conceptualised as the catching-up effect, where poorer economies reach the levels of economic well-being their developed counterparts currently enjoy. This process in theory is related to the idea that growth and per capita income are inversely related, as emphasised in the neoclassical theory of growth (Barro & Sala-i-Martin, 1992). Neoclassical growth models assume diminishing returns to capital. This implies that poorer countries with less capital can experience greater returns on capital compared with their capitalendowed counterparts. As a consequence poorer countries will grow at a faster rate than richer countries. This definition of convergence can be dubbed absolute convergence in the sense that economies grow closer together with respect to their per capita incomes. Generally the concepts of absolute -convergence can be measured using the following regression: (5.1), 1 log yi t T log yit it. yit T Equation (5.1) has the left-hand side ratio of GDP per capita in country i in year t+t to that in an initial year t. The logarithm of this ratio, log (y i,t+t /y it ), is approximately the proportionate growth rate over the period [t, t+t] so that dividing by T gives the average annual growth rate. Equation (5.1) expresses this annual growth as a linear function of the log of GDP in the initial year, plus a random term it. On the right-hand side, y it is the initial GDP per capita at time t, and β is the parameter to be estimated. If β 133

145 < 0, convergence is present. Assuming country i has a lower GDP per capita than country k, the GDP growth rate of the former should exceed that of that latter. Eventually the GDP per capita levels will even out. The results of estimating Equation (5.1) are reported in Table 5.2. The data was sourced from the World Bank s World Development Indicators Using annualised real GDP per capita growth rates, Table 5.2 reports results for two ten-year and one seven-year sub-cross-sections for the period 1980 to The table also includes a long-run regression of growth rate from 1980 to 2006 in column (4). Based on the criteria of β<0 as an indicator of convergence, the estimates reported in the table do not indicate convergence in all sub-period cross-sections except for in column (3). The sub-period cross-section coefficients are only significant for at the 5% level. For the full period, the estimated coefficient of β is positive and significant at the 5% level. This suggests divergence instead of convergence, based on the neoclassical models of growth: thus it is sensible to infer no convergence takes place within the period from 1980 to Table 5.2 Absolute Convergence Estimates, Cross-Sectional, 1 log yi t T log yit it. yit T Period (1) (2) (3) (4) Constant (0.0116) (0.0109) (0.0067) (0.0091) Per capita GDP (0.0015) (0.0012) (0.0008) (0.0011) R-squared Adjusted Number of Observations Note: White heteroskedasticity consistent standard errors in parentheses The lack of international convergence in growth can be verified by plotting income growth over the sample period against the initial year s GDP per capita. The regression line in Figure 5.1 clearly indicates the lack of convergence within the sample. The regression line has a gradient of 0.002, which indicates lack of a negative relationship between the growth of GDP per capita and the initial GDP per capita. In absolute terms, the world exhibits no absolute convergence. However, it is possible to examine convergence for income-specific groups, where economies may be more homogeneous. 134

146 Grwoth (p.a.) 10% Figure 5.1 Absolute Growth Divergence, % 6% 4% 2% 0% -2% -4% -6% -8% y = x Ln GDP per capita 1980 Source: World Bank (2008b) Dividing countries into different income groups based on their GDP per capita may show a different picture. The results of convergence estimates based on income levels are reported in Table 5.3 for five different income categories. Table 5.3 Absolute Convergence Estimates, Country Groups, 1 log yi t T log yit it. yit T Low Income Lower Middle Income Country Group Upper Middle Income High Income (non-oecd) OECD (1) (2) (3) (4) (5) Constant (0.038) (0.032) (0.032) (0.056) (0.054 ) Per capita GDP (0.007) (0.005) (0.004) (0.006) (0.006 ) R-squared Adjusted Number of Observations Notes: White heteroskedasticity consistent standard errors in parentheses The income categories of Low Income, Lower Middle Income, Upper Middle Income, High Income (non-oecd) and OECD countries are based on the 2007 GNI per 135

147 capita World Bank (2008) figures. 2 Unlike the previous results, the estimate β is now negative in all cases. It is the largest for High Income (non-oecd) countries in column (4). While these results indicate significant convergence within the sub-income groups with the exception of the OECD, caution must be exercised. The number of useful observations is small in all cases and severely reduced for the High Income (non- OECD) group. However, for columns (1), (2), and (3) the results remain significant for a small sample. These estimates indicate that within income groups there is some degree of convergence Conditional Convergence The results of absolute convergence above do not support convergence across a large group of countries. This is not a surprising result since the concept of absolute convergence is crude at best. It relies on the assumption of common attributes for all countries included within the sample, and assumes that all countries are homogenous. Thus a Low Income country is expected to have the same attributes as an OECD country and to behave in a similar fashion when growth takes place. This assumption is not sensible, especially in light of the poor performance of the model. Therefore using the basic model of convergence, there is little evidence to support the hypothesis that poorer countries grow faster than their richer counterparts. Theoretically, the flaw of absolute or unconditional convergence is due to the fundamentals of the Solow model of growth, which implies that convergence will take place towards a common steady state of per capita income. This growth is negatively related to the initial levels of income. Thus when β<0, convergence is present; otherwise divergence has taken place (Durlauf, Johnson and Temple 2005). To address this issue of heterogeneity within the sample, modifications to the Solow models of growth were introduced. From these, conditional convergence and endogenous growth theories evolved. Conditional convergence uses the Solow growth model and augments it with country-specific attributes. In contrast, endogenous growth theories take the path of including knowledge and human capital within the modelling of growth. Romer (1986) and Lucas (1988) have led the research in this field. This chapter, however, follows the neoclassical approach to growth, utilising the work of Barro (1991), Barro and Sala-i-Martin (1992) and other cross-country studies. A detailed derivation of the neoclassical model based on their approach is presented in Appendix A World Bank (2008a). 136

148 Conditional convergence is based on accounting for differences between countries based on economic, political, social and demographic characteristics. The neoclassical model includes savings and population growth as control variables of growth towards the steady state; country differences are not accounted for. The model shown in Equation (5.1) above is expanded by the inclusion of a number of variables that address the shortcomings of the absolute or unconditional convergence results. This approach draws on the extensive work done by Barro and Sala-i-Martin in a number of studies on cross-country comparison. The conditioning variables are introduced to control for different steady states across a heterogeneous sample (Barro et al, 1991). The conditional convergence model is generally presented as follows:, 1 log y log it Xit it, yit T i t T (5.2) y where is a matrix of country specific characteristics discussed above. Equation (5.2) X it also includes initial GDP, as above. We shall use Equation (5.2) to estimate β and γ in a cross-section-based regression. A great number of conditioning variables are used to account for steady state differences and growth in per capita GDP, but not all have been shown to be significant. Levine and Renelt (1992) conducted a substantial sensitivity study of cross-country growth regressions using typical variables used in the literature. They found that investment shares of GDP correlated significantly and robustly with growth, and that investment and trade were related. However, trade policies and growth were not as robustly correlated when investment was included in the regressions. They also found support for conditional convergence, and that fiscal policy indicators robustly correlated with growth. Many of the other variables, including political indicators, were not robust, however. Some of the major control variables were also extensively tested by Barro and Sala-i-Martin (2004). They found support for conditional convergence such as those indicated by the neoclassical growth models. The following section reports the results of estimates of conditional convergence. One of the more significant conditioning indicators introduced to the model above is human capital. Since the neoclassical growth models concerned themselves only with physical capital, human capital was assumed a given and not included as a variable. Using education as a proxy for human capital, for example, data such as primary and secondary school enrolment quantify the stock available. Technology is another factor that the neoclassical model did not account for directly, since it remained 137

149 exogenous in the earlier growth models. Technological progress is measured in terms of the number of scientific and research publications. Other indicators include population growth and fertility to gauge demographic differences. Openness and terms of trade measure external policy and its implications on countries. Inflation is included to gauge monetary policy as central banks aim to maintain stable prices. Fiscal policy is proxied for by government expenditure. Domestic credit from commercial banks is used to represent the state of the financial conditions within countries. Political rights and civil liberties indices are used to proxy for institutions. Geographic features are limited to the lack of direct shipping lane access, captured by a landlocked dummy. Finally, the analysis includes dummy variables for some of the more notable RTAs such as the EU, NAFTA, ASEAN, CARICOM, and more importantly for our purposes, the GCC. The estimation of conditional convergence is subject to considerable debate. The main approach in the literature is to use cross-sectional linear estimation. Thus, the sign of coefficient of initial GDP,, indicates the presence (when < 0) or absence (β 0) of convergence. OLS cross-section regressions are considered suitable for the estimation by some, including Barro and Sala-i-Martin (2004), but not everyone agrees. In contrast to the cross-section approach, Islam (1995) suggests a panel estimation approach to take into account country effects on long-term economic growth. This approach requires the data to be handled in shorter time increments without compromising the long-term periods. Studies that choose panel estimation as an econometric method typically use five-year increments to divide the data over the whole sample period. This reduces the number of observations available but does not appear to affect the detection of long-term effects. Quah (1993, 1996) suggests a completely different approach to testing for convergence, arguing that the convergence trends detected in the literature are inherent in the data structure. Quah (1996) supports the argument of convergence clubs found in the literature, such as those of Baumol (1986) and Ben-David (1994). Quah s bi-modal method suggests polarisation of the distribution of countries, in that convergence is occurring not between all countries but within groups of countries. His results, however, have been found to be sensitive to country selection (Durlauf et al. 2005). For the purposes of this chapter, cross-sectional regressions are used to test for convergence. Using the conditional variables above, cross-sectional estimation of convergence can be carried out. Essentially, the estimation is concerned with two years, 1980 and The aim is to estimate the relationship between growth over the period 1980 to 2006 and the initial conditions of the variables listed above. There is an advantage in 138

150 using cross-sectional data in this case, in that it allows for a clear indication of convergence during the period without any contamination from business cycles and short-term changes: that is, the long-term perspective helps isolate clear trends. The increased depth in data may emphasise interactions between the conditioning variables and growth better, but the trade-off is the omission of interim changes during the period of interest. The results of the cross-sectional convergence are reported in Table 5.4. The table includes four main regressions that include different combinations of the variables listed above. Column (1) of the table includes estimates of the main variables discussed earlier. The coefficient of initial per capita GDP level is negative, This implies that initial levels of income are negatively-related to growth. Convergence is present when growth is conditioned based on economic, social and political variables. However, while most variables comply with economic intuition, such as the negative relationship between fertility and economic growth, most of the estimates are insignificant. In this case only the per capita GDP level and fertility are significant, at the 5% and 10% levels respectively. The remaining coefficients are not statistically significant. Furthermore, the sample size is severely reduced when the conditioning variables are introduced in fact about less than one quarter of the countries available can be included in such a regression. Alternatively, it is possible to omit some of the variables in column (1) and test for convergence without taxing the model. Omitting a number of the variables which were insignificant in the previous regression, column (2) of Table 5.4 presents results with the remaining indicators. The variables which severely reduce the number of observations include government spending as a percentage of GDP, life expectancy, terms of trade, secondary school enrolment, and scientific publications. The estimated coefficients for the level of per capita GDP remain negative and significant at the 1% level. Conditional convergence is present once initial conditions are accounted for. All the variables included in column (2) except openness are statistically significant. It should be noted here that the political rights coefficient is negative. This does not imply that political rights, a proxy for democratic rule, are negatively related to growth. The nature of the index, in which democratic countries are given a lower score than their non-democratic counterparts, explains such a result. When we reduce the 139

151 Table 5.4 Cross-Sectional Conditional Convergence Estimates, Model Variable (1) (2) (3) (4) Constant (0.0839) (0.0269) (0.0321) (0.0298) Per capita GDP (0.0033) (0.0018) (0.0020) (0.0022) Fertility (0.0108) (0.0054) (0.0061) (0.0062) Investment (0.0074) (0.0054) (0.0055) (0.0049) Government Spending (0.0040) Life Expectancy (0.0129) Political Freedom (0.0037) (0.0027) (0.0029) (0.0029) Inflation (0.0041) (0.0018) (0.0019) (0.0020) Openness (0.0065) (0.0024) (0.0024) (0.0025) Secondary School Enrolment* (0.0023) Scientific Publications* (0.0024) Terms of Trade (0.0054) Dummies Landlocked (0.0046) (0.0046) Custom Union (0.0036) Free Trade Area (0.0042) R-squared Adjusted Number of Observations Note: White heteroskedasticity standard errors in parentheses * Period average 140

152 number of variables to the essential statistically significant indicators, the explanatory power of the model remains relatively unchanged. The R-squared adjusted here is , compared with from the previous regression. This shows that the model is unaffected by the reduction in the number of variables. The model can be further improved by including a number of variables to capture geographical conditions that may influence growth. This is generally achieved by the use of a dummy variable to indicate whether a country is landlocked or not. Where countries are landlocked, access to shipping lanes is restricted and consequently trade may be influenced. The result of including such a dummy variable in the regression is reported in column (3). The coefficient of the landlocked dummy is , implying that long-term growth of countries without direct access to shipping lanes is affected negatively. The coefficient is statistically insignificant, and the inclusion of the landlocked dummy does not affect the implications of convergence, the per capita GDP coefficients being more or less the same and remaining significant. The results remain relatively unchanged when trade policy is accounted for. The model includes a measure of openness which captures how much trade countries are involved in. However, the emphasis of this chapter is to understand the effects of joint trade policies on convergence. To capture such effects, two dummy variables are included one for custom unions and the other for free trade areas. Within the sample, if a country belongs to a customs union, it receives a value of 1 otherwise 0. If a country is involved in a multilateral free trade area it receives a value of 1 otherwise 0. Countries that are not involved in either type of arrangement or subscribe to any other type of joint policy coordination, receive a value of 0 for each dummy. The most common third option is a preferential trade agreement. The results in column (4) of Table 5.4 show that the dummies for customs union and free trade area both contribute negatively to growth. The estimated coefficients are negative but not statistically significant. A sample including 86 countries implies that membership in a customs union or a free trade area does not affect growth in any meaningful manner. The explanatory power of the model is not affected, and nor are the convergence findings. Conditional convergence remains statistically significant. 5.5 Dispersion and Convergence An alternative method of defining convergence is the reduced dispersion of countries incomes. Currently, for example, per capita GDP varies significantly from the richest to the poorest countries. The dispersion of GDP per capita levels and growth 141

153 Number of Countries clearly indicate a lack of convergence between countries across the world. Over time, the dispersion between countries will increase or decrease because of many factors. These changes allow observers to infer trends of convergence or divergence across countries. A quick study of the global distribution of income per capita shows a distinct distribution where countries have not converged. A temporal distribution for 1970 and 2003 is illustrated in Figure 5.2. This is a smoothed histogram constructed using data on GDP per capita from the Penn World Table 6.2. The distributions are truncated at $US 30,000. Consider first the solid continuous line labelled The median per capita GDP is approximately $US 3,000, as indicated by the vertical line. About 40 countries fell below the median in Now consider the dashed line labelled It represents the distribution of incomes in Two features are evident, first the median income increases substantially to more than $US 7,000. Moreover, the number of countries below the median has increased. Another feature is the long tail of the distribution in This is prima facie evidence that different countries do not convergence. Figure 5.2 World s Distribution of Income Median Median ,000 10,000 15,000 20,000 25,000 30,000 Per capita GDP (PPP US$) Source: Penn World Table 6.2 The analysis of dispersion, and thus convergence, aims to understand the behaviour of a country s income over time vis-à-vis a group of countries. Figure 5.3 is a good starting point for such an analysis, as a number of possible scenarios, with respect to deviation, are possible. The perfect scenario for convergence is when the standard 142

154 deviation of growth within a group of countries is zero in other words, convergence of growth has taken place such that there are no deviations from the mean. Perfect convergence is unlikely to take place in the real world, however; thus, the alterative is a low standard deviation. Take the hypothetical example of the GCC. If on average the GCC s standard deviation is 2.3, and the six countries dispersion varies around this value over time. The GCC would maintain minium dispersion since it is not increasing or declining but stable at a low average. The other two conditions displayed in Figure 5.4 demonstrate the two opposite conditions of convergence and divergence. In the case of convergence curve, the standard deviation is expected to decline over time. This translates to reduced dispersion and volatility. In contrast, if the standard deviation increases over time, it implies that the economies in question are diverging. Figure 5.3 Dispersion and Convergence Standard Deviation Divergence Convergence GCC Converged Time The Case of the GCC Consider the case of the GCC s GDP growth measured by the annual percentage growth in GDP valued at PPP $US obtained from the International Monetary Fund s (IMF) World Economic Outlook Figure 5.4 illustrates the average growth of the GCC countries, indicated by the solid line. The figure also indicates the range of growth each year shown by the shaded area. In the first few years where a greater spread is observed, growth rates per annum varied by about 40%. This range declined during the 1980s but remained substantial. The pattern reversed in the early 1990s when the first Gulf War took place. The range of growth rates in the GCC increased significantly, with large fluctuations primarily as a result of the Iraqi invasion of Kuwait in 1990, when the latter s economy contracted significantly. These large fluctuations in growth rates declined towards the end of the period. With the exception of Qatar in 1997, the rest of 143

155 the GCC countries grew at comparable rates to the region s average. The range of growth rates remained at a minimum in 2005 and Figure 5.4 illustrates a number of important features of the GCC economies. The first is that the GCC countries do not necessarily grow at the same rate despite sharing similar vulnerabilities with respect to oil prices. Second, there appears to have been greater conformity of growth rate across the region in the past decade. However, inferences from Figure 5.4 are indicative at best. The ranges of growth rates shown there do not take into account the size of the six economies. For example, the large fluctuations noticed in the early 1990s can be primarily attributed to the third largest economy within the GCC: Kuwait. An economy such as Bahrain, the smallest of the six, can also show substantial effects on the range of growth. In other words, the six countries are weighted equally. A better practice is to control for both the size factor and the contribution of each country within the GCC. Figure 5.4 GCC Average Growth Dispersion (% p.a.) Source: IMF (2007) and Author s calculations Note: Solid line is unweighted mean of growth in GDP of country group. The width of the shaded region is the range of growth rates To account for the size of each country, each country s share of the total regional output is required. Let c = 1,, 6 denote the six GCC countries. Let Y ct be the GDP of country c at time t. The share of country c of the GCC s output is equivalent to: Yct (5.3) wct 6 Y c 1 ct. 144

156 The expression w ct defines the share of county c of the GCC s total output at time t. Taking the arithmetic mean of expression (5.3) for years t and t 1, produces: (5.4) wct wct wc, t Table 5.5 reports the derived weights for each of the GCC countries for five-year intervals from 1981 to Using these weights, an inter-temporal picture of the GCC can be obtained. To indicate the changes in the sizes of the GCC members, Figure 5.5 plots the values of w ct in 2006 against those in Figure 5.5 illustrates a number of points. Quite clearly Saudi Arabia dominates the region with its physical and economic size. In fact, in 1981 Saudi Arabia s share of the GCC s output was approximately 70%. It is followed distantly by the UAE, Kuwait, Oman, Qatar and Bahrain in that order. The shift in the share of total output can also be realised from Figure 5.5,where the 45 line indicates which countries share grew and which did not. In the case of Saudi Arabia, its output share fell from 1981 to This is indicated by its point being below the 45 line. The same is true for Kuwait. In contrast, the UAE, Oman, Qatar and Bahrain all experienced an increase in their share of GCC output over the past two decades. This is particularly true in the case of the UAE, where an increase from 15% to 21% was experienced. Evidently, the shares of each of the GCC countries are not constant and they shift through time. Thus, instead of using a simple (or unweighted) average and the range in Figure 5.4, the use of weighted averages and standard deviations is more informative. Given the weights from Equation (5.4) the weighted mean of real growth can be expressed as follows: (5.5) g 6 w g t ct ct c 1. The corresponding weighted standard deviation is as follows: (5.6) 6 V = w g - g. c=1 ct ct t 2 Figure 5.6 plots the weighted mean of GDP growth for the GCC countries in the solid line. The dashed vertical lines are the weighted standard deviation derived from Equation (5.6). The weighted mean of growth here declines in , but increases 145

157 2006 again in the following few years. It fluctuates during the 1980s and plateaus in the early 1990s. This trend continues until the late 1990s, when a decline begins to take place, most obviously in Table 5.5 Output Shares of the GCC Countries Bahrain Kuwait Oman Qatar Saudi Arabia United Arab Emirates Source: Author s calculations Figure 5.5 Relative Importance of GCC Members, 1980 and 2006 (Share of Total GDP) Saudi Arabi a UAE 10 0 Bahrai n Qatar Oman Kuwait Source: Author s calculations A short-lived respite takes place in 2000 before the slump of 2001 takes hold. Towards the end of the period, however, the weighted mean growth increases again and stabilises. Of interest here are the weighted standard deviation bands: throughout the period the bands are significantly large, especially during the 1990s. These large

158 deviations, significant during the last three years of the period as well, do not support the hypothesis of converging economies within the GCC in terms of growth. Figure 5.7 compares the two means. Panels A and B reproduce the mean and weighted mean of growth. The most striking difference is in the greater fluctuations in panel A compared to those in B. The simple mean detects the growth rates of change regardless of the size effect. Thus from 1990 to 1992 the mean changes dramatically. This effect is not present when the weighted mean is considered in panel B. During the first few years of the period, the large fluctuations in the simple mean are minimised. In the second half of the period, the weighted mean in panel B reflects similar movement as a simple mean. Figure 5.6 GCC Growth: Weighted Mean and Standard Deviation (% p.a) Source: Author calculations Note: Solid line is weighted mean of growth in GDP of country group The width of the dashed lines are the weighted deviations of growth in the six GCC countries Emphasising the significance of dispersion, Figure 5.8 illustrates the weighted standard deviation derived from Equation (5.6). The figure shows that the dispersion of growth declines sharply during the first year and then remains low during the 1980s. In 1990, the weighted standard deviation of growth more than triples as a response to political turmoil in the region. Growth declines sharply but spikes again in The late 1990s experience low dispersion as the weighted standard deviation remains low. However, towards the end of the period a marked increase in dispersion is evident. Based on the figure, periods of convergence in the 1980s and late 1990s can be observed, although the occurrences of divergence within the region are significantly greater. In 1990, and 1995, and from 2002 onwards, strong deviations persist where values of the weighted standard deviations increase three-fold at least. This confirms 147

159 suggestions of divergence between the GCC economies especially in the last few years. Figure 5.7 GCC s Real GDP Growth (% p.a) A. Unweighted Mean 20% B. Weighted Mean 15% 10% 5% 0% -5% Source: Author s calculations. 148

160 Figure 5.8 Weighted Standard Deviation of Growth (% p.a.) Source:Author s calculations Other Regional Integration Experiences Convergence amongst states or regions may be subject to the characteristics of the countries involved. Historically, nations and regions have come together to improve their economic, social and political circumstances. Economic initiatives such as the early customs unions between the German states in the 19 th century and, more recently, the European experience, are examples of policy integration and cooperation. Politically, the birth of the United States of America and the Commonwealth of Australia are other examples of unification and consolidation. In the Caribbean islands, states work together to maintain economic ties and cooperation. The GCC s experience can be compared with a number of regions that have undergone similar or more comprehensive integration, including the EU, the US and Australia. The EU represents the most successful economic integration process between countries. The US and Australia offer a situation where individual states and territories choose to form a federal union: they provide a case of fully-fledged integration, both economic and political. The EU provides a good example of economic integration that has proved operational. However, for the sake of comparison, only the EU15 will be referred to. These include the original six members, Belgium, France, Germany, Italy, Luxembourg and the Netherlands. Denmark, Finland, Greece, Ireland, Norway, Portugal, Spain, Sweden and the UK are later signatories. Figure 5.9 depicts the weighted mean of 149

161 growth for the EU15. The EU15 countries have grown significantly since the early 1980s, where average growth exceeded 3% in some years. This upward trend reversed in 1988 when growth declined until Another surge in growth was experienced in 1995, but then rapidly slowed until The EU15 enjoyed accelerated growth until 2000 following the global slowdown. In 2003 the EU15 s growth picked up again, but only briefly, and in 2006 the slowdown reversed. Figure 5.9 illustrates the fluctuating growth of the EU15, the cycles through which growth sped up and declined. The early 1990s mark the Exchange Rate Mechanism (better known as ERM) crisis and the departure of several of its members, notably the UK. The global slowdown in 2001 is also evident within the EU15 with a notable decline in growth. During most periods, growth dispersion ranges between 2% to 4%. While the dispersion increases during the slowdown years in the 1990s, it remains relatively constant during most of the years that follow. There are no indications of divergence or convergence based on the figure. The European experience indicates a minor decline in dispersion, however convergence is evident based on the low dispersion observed. Figure 5.9 Growth in EU15: Weighted Mean (% p.a.) Source: Author s calculations based on World Economic Outlook 2007 data. Australia presents another good benchmark against which to compare the GCC. Formed of a similar number of states as the GCC is, it demonstrates an example of full economic integration, so observing the convergence experience of Australia against that of the GCC is useful. Figure 5.10 shows the weighted mean of growth and its dispersion for Australia. From 1986 to 2005 the Australian states and territories experience 150

162 positive growth on average, except in 1990 where growth is negative. In 1992 growth rises substantially, and the states and territories grow by about 4%. During the 1990s growth remains positive, despite declining in Another notable decline in growth is observed from 1998 to This is very likely related to the Asian financial crisis and its contagion effect on other markets. Despite global slowdown in 2001, Australian states and territories are shown to enjoy positive growth that dips slightly in Dispersion on the other hand reveals quite a different picture. The dashed lines in Figure 5.10 indicate that the dispersion in states and territories is relatively large; in some years growth it exceeds 6% on average. This is especially true during the last few years displayed. Only in the middle years can a decline in dispersion be detected, notably in The dispersion remains relatively constant throughout the time period, however. The Australian states and territories do not show signs of convergence between 1986 and 2005 and deviations from the mean growth rate remain persistent. Figure 5.10 Growth in Australian States: Weighted Mean (% p.a.) Source: Author s calculations based Australian Bureau of Statistics data. Figure 5.11 illustrates the weighted mean of real growth for the 50 American states. After a recession at the beginning of the period, the growth rate recovered in 1981 and The spurt of growth was not sustained in the following years and growth declined dramatically. During the rest of the 1980s, it kept declining, bottoming in the recession. During the 1990s the states maintained positive growth fluctuating between 2% and 5%, but in 2000 and 2001, the economy experienced 151

163 another major slowdown, where growth fell below 1%. The states recovered to pre-2001 levels but in 2005 growth declined again, although remaining above 1%. Figure 5.11 also depicts the dispersion of the American states. Dispersion is shown as generally declining from the early 1980s, where differences are about 10% between states. The dispersion remains persistent throughout the time period. It starts to decline during the 1990s to reach about 4%. Overall, from 1981 to 2006 the states show a minor reduction in dispersion but no signs of divergence. The dispersion remains relatively constant over the second half of the period. This suggests that the US has converged considerably, in comparison to the GCC Figure 5.11 United States Weighted Mean of Growth (% p.a.) Mean 3.2 Std. Dev Source: Author s calculations based on Bureau of Economic Analysis and U.S. department of Commerce data. Figure 5.12 compares the GCC s weighted mean of real GDP growth with that of the EU15, Australia and the US. With the exception of Australia, the figures depict a weighted mean growth over the period Figure 5.12 illustrates the stark difference between the GCC and other comparable regions. Surprisingly, the EU15 is the most converged, followed closely by the US and Australia. These three show moderate fluctuations in growth generally, while in contrast the GCC s growth is more volatile on average and its dispersion significantly greater than in its counterparts. With respect to growth, the EU15 bloc provides the most suitable example of convergence; but whichever comparison is made, convergence has not taken place within the GCC countries. 152

164 GCC Figure 5.12 Weighted Mean of Real GDP Growth (% p.a.) EU15 Australia United States Source: Author s calculations 153

165 5.6 Deviations from the Mean Approach Much of the discussion above has dealt with the empirical estimation and descriptive statistics of convergence. In this section we investigate another approach to convergence, one that uses deviations from the mean of income. Instead of taking a purely statistical approach, this framework adds statistical inference while allowing for a non-parametric approach. The following analysis does not concern itself with the theoretical underpinnings of growth but focuses on the coexistence of economic integration and income convergence. Since the GCC is central to this analysis, methodologies that apply to small samples are particularly useful. As a very small sample, such as the GCC, restricts the usefulness of extensive parametric analysis, using descriptive statistics makes it possible as shown earlier to draw some inferences regarding the impact of economic integration on income convergence. This framework does have its shortcomings, however one being that it does not explicitly offer an economic understanding of the important relationship between economic integration and income convergence. Adopting the approach of Ben-David (1993, 1996), a small sample convergence estimate can be obtained. This allows for region-specific estimates of convergence based on dispersion measures. Using deviation from the group mean as a measure of convergence or divergence used, the following model is created: Let be the log of per capita GDP of country i in year t, where i = 1,2,,k. Future per capita GDP at t+1 is directly related to that in current year t. Therefore: (5.7) The unweighted mean of the group of countries k at year t is. Equally, the future mean of per capita GDP of group of countries k is directly related to that in the current year, t. Accordingly: (5.8) it 1 yit. Taking the deviation from the mean on both sides gives us: y y t 1 yt. y t y it (5.9) y y y y i, t 1 t 1 i, t t. 154

166 Differenced Deviations from the Mean Convergence is present when <1. The speed of convergence can also be deduced by modifying the model and differencing the deviations. 3 Subtracting sides gives: (5.10) from both The transformation of (5.10) allows for an estimation of the speed of convergence. Ben- David (1993) uses (5.10) to estimate the speed at which individual countries converge to the group mean. The coefficient of 1 gives the rate of convergence of country i s per capita GDP to the group s average. Larger values of 1 imply greater speeds of convergence. Figure 5.13 plots the relationship between the differenced deviations and the level deviations from the world mean for the period 1970 to The regression line is flat, which indicates that deviations from the mean are not declining over time. This supports earlier deviation analysis, of global convergence in section 5.5. A GCCspecific picture is shown in Figure Here, a negative slope is evident from 1970 to Figure 5.14 lends support to the argument that individual GCC countries are converging toward the group mean. The deviations of GCC incomes appear to be declining over time. 1.2 ( y y ) (1 ) y y. i, t 1 t 1 i, t t Figure 5.13 Income Convergence, y i, t y t Deviations from the Mean Source: Author s calculations based on Penn Tables 6.2 data 3 For details see Ben-David (1993, p666) 155

167 Differenced Deviations from the Mean Figure 5.14 GCC Income Convergence, Weighted Deviations from the Mean Source: Author s calculations based on Penn Tables 6.2 data Table 5.6 reports results of estimating Equation (5.10). In the case of the GCC, Table 5.6 reports the speed of convergence (1-ϕ) = The coefficient of ϕ is equal to 0.952, which is significant at the 5% level. This result complies with the condition noted earlier of convergence being present when ϕ <1. Column (2) of Table 5.6 reports the same regression using weighted deviations. Individual observations are weighted by GDP shares derived earlier, so deviations are now represented as follows: (5.11) wit yi, t 1 yt 1 wit yi, t yt ( ) (1 ). The weights w are derived from Equation (5.4). This is done to keep this it approach as comparable to the previous descriptive analysis as possible. The results do not differ in a meaningful way. The coefficient (1-ϕ) is still significant at the 1% level. In this case, ϕ = 0.946, which is very similar to the estimate from Equation (5.10). The results in Table 5.6 confirm that with respect to dispersion, the GCC shows signs of convergence within the period 1970 to Using these estimates can also help find the half-life of convergence. Ben-David (1993, 1996) shows that using the expression log0.5 log yields the half-life of convergence for the respective group in question. For the GCC, this yields 14 and 12.5 years respectively. Based on this, the GCC members should converge towards the group mean in 25 to 28 years. 156

168 Differenced Log Deviations from the Mean Table 5.6 Convergence The Deviations from the Mean Approach ( y y ) (1 ) y y Unweighted (1) GCC Weighted (2) EU15 (3) Speed of Convergence (0.018) (0.031) (0.006) Adjusted R Number of Observations i, t 1 t 1 i, t t Note: White cross-section heteroskedasticity consistent standard errors in parentheses The analysis conducted by Ben-David (1993) was based on the six founding members of the European Economic Community. Here the model is applied to the EU15 despite entry years. The assumption is made that the EU15 have been part of the integration process since 1970, which is not strictly true for all countries but does, however, allow for comparisons with the dispersion descriptive analysis presented earlier. Figure 5.15 plots the EU15 s differenced deviations from the mean against the levels of those deviations. Figure 5.15 EU15 Income Convergence, Log Deviations from the Mean Source: Author s calculations based on Penn Tables 6.2 data 157

169 Estimates of (1-ϕ) in Equation (5.10) for the EU15 are found to be statistically insignificant and the half-life is 62 years. Despite the long projected convergence timeframe, the result is not robust. However, a plot of the variables for the EU15 presents an interesting pattern. In comparison with the GCC s convergence pattern in Figure 5.14, the EU15 results show significantly less deviation. The tight cluster of points indicates that the EU15 has undergone substantial convergence. The regression line is marginally negative here, indicating possible further convergence. This is in line with analysis presented earlier with respect to weighted standard deviations. The statistical insignificance may be a manifestation of past convergence. 5.7 Summary and Conclusions The aim of this chapter was to examine the convergence of the six GCC countries with respect to their income and growth. The key hypothesis proposed was that economic integration amongst the GCC countries would influence growth and incomes within the region so much that they would converge. The implications of convergence are pronounced when considering the objective of creating a monetary union in the foreseeable future. The GCC countries rely on similar economic attributes, including similar growth cycles and incomes, to reinforce their integration efforts. The integration process aims to bring these economies closer through trade liberalisation and other joint policy coordination efforts. The chapter used three main approaches for the theory of convergence in the GCC: the β-convergence, dispersion of growth and deviation-from-the-mean approaches. Using the first approach, on a global scale based on the neoclassical approach to growth and convergence, evidence of countries arriving at the same income levels is not present. In fact, in some cases diversion of income is the situation. When the world is divided into homogeneous income groups, some support is lent to convergence. Such results are not robust however, with most estimates of β- convergence being insignificant. Since no evidence was found of convergence based on the original neoclassical approach, conditional convergence was introduced to try to capture the heterogeneity of the sample. Convergence is evident when growth is conditioned on country-specific characteristics, such as economic, social and political variables. The variables most effective for this sample included initial levels of income, fertility, inflation, openness, investment and political liberties. An assessment of the effect of regional trading blocs in the form of currency unions or free trade areas on 158

170 convergence suggested that there is no statistically significant effect of these arrangements on growth or convergence of incomes. The second approach to convergence used dispersion of growth. Focusing on the GCC countries, the dispersion of growth, using unweighted and weighted averages, indicated persistent differences. The GCC countries do not exhibit signs of convergence with respect to growth. The analysis compared the GCC countries with other regions: the EU, US and Australia. The GCC shows a greater variability of growth over time compared to these regions. The findings of dispersion analysis indicate no convergence between the GCC countries during the period 1980 to Finally, the chapter used deviations from the mean methodology, to estimate convergence based on a small sample. Using data from 1970 to 2003, the GCC countries showed signs of reduction in deviations from the mean. The findings of this approach indicate convergence within the GCC. This result should be viewed with caution as half-life estimates signify a long period of convergence; approximately two decades at least. The EU exhibits greater convergence between its member countries. The analysis of the effect of the GCC s economic integration on income and growth demonstrates the negligible convergence that has taken place within the region. Given the underlying conditions of trade liberalisation and standardisation of standards and policies, the impact of the process is not clear with respect to incomes and growth. The GCC has yet to reach its potential as an RTA. 159

171 Appendix A5.1 A Neoclassical Growth Model The neoclassical models of economic growth providing the foundations for extensive work on -convergence were led by Barro (1991), Barro and Sala-i-Martin (1992, 2004), and Barro et al. (1991). These studies link the neoclassical growth models developed by Ramsey (1928), Solow (1956), Cass (1965), and Koopmans (1965) to derive empirical estimates of per capita income convergence. Generally, the theoretical underpinnings of such empirical work are related to the Solow growth model, which will be used here for the purposes of developing a simple derivation. The Solow model assumes a production function based on two factors of production: (A5.1) where Y is the output of the economy and K and L are capital and labour inputs. The constant A attached to labour can be considered as knowledge of production. Thus the term AL is the effective labour input and not the absolute numbers per se. The neoclassical model also assumes constant returns to scale. Therefore if the inputs are doubled, the output will also be doubled. Using this feature, if both sides of the equation are divided by AL, an expression for output per effective worker can be derived such that: Y K F,1. AL AL Y K Using y and k the expression above can be reduced to: AL AL Y=F K,AL, (A5.2) y f k. The production function of this one sector and two input closed economy is assumed to take a Cobb-Douglas form; therefore: (A5.3) 1 Y t K t AL t 0< <1, where and 1 are the input share of capital and labour respectively. The model assumes that labour and knowledge (technology) growth rates are exogenous. Their growth rates are defined as: 160

172 (A5.4) (A5.5) dl t dt da t dt nl t, ga t. Expressions (A5.4) and (A5.5) imply that the rate of growth of labour and technology are constant at n and g respectively. Assuming both L and A grow exponentially over time, then: (A5.6) (A5.7) The growth rate of effective labour is thus (n + g). Two forces savings and depreciation affect the capital stock in the model. Within the neoclassical framework the capital stock increases over time subject to the savings rate s. Savings per effective worker, on the other hand, is a function of y defined above, so the saving is sf(k). On the other hand, the depreciation rate reduces the capital stock at a rate δ. Total depreciation of the capital stock within the economy is δk. Thus based on total savings sf(k), total depreciation δk and growth rate of labour, the growth of the capital stock is: dk dt (A5.8) sk t n g k t. The inclusion of (n+g+ ) into the equation is explained as follows: Investment (equal to saving) increases the capital stock for two purposes, firstly to increase the stock of capital available, and secondly to replace worn-out capital and match the growing demands of labour (growing at a rate of n). A steady state is reached when capital stops growing; in other words, the investment is sufficient to replace depreciated capital and to compensate for effective labour growth (n + g). Expression (A5.8) can also be modified to include consumption as part of the model where change in the stock of capital becomes: dk dt (A5.9) f k c n g k t, where c=c/al, consumption per effective worker. The consumption of households maximises the utility, (A5.10) Here c expression: 0 nt L t = L 0 e, gt A t = A 0 e. nt t U u c e e dt. C L, or consumption per worker; and is the rate of time preference. The 161

173 1 c 1 u(c) is 1 and >0, The first-order condition for maximising (A5.10) is: dc (A5.11) dt 1 f k. c At the steady state y, k and c grow at the rate of technological growth g. In this state, the level of capital labour ratio satisfies the following: (A5.12) f k ˆ g. Thus, the growth of output at the steady state is subject to the depreciation, rate of time preference and technological change rate. Reverting to the Cobb-Douglas form introduced earlier, it is assumed that technological progress behaves as follows: (A5.13) y f k Ak. Linearising Equations (A5.11) and (A5.13) and solving for gives the expression for the transition of output per effective worker towards the steady state. The solution is: (A5.14) ˆ ˆ t t y t y e yˆ e log log 0 log 1. The parameter determines the speed of convergence towards the steady state. On average the growth of y (per capita income) over a period of time T is determined as follows: (A5.15) T 1 yt 1 e yˆ log x log. T y(0) T yˆ 0 Equation (A5.15) implies that the growth of income per capita around the steady state, over a period of T years, is governed by the initial condition of y(0). Thus, lower initial per capita income produces greater speeds of convergence towards the steady state. For estimation purposes Equation (A5.15) can be reorganised as follows: y a e y g t u y it, 1 it (A5.16) log 1 log 1 here ˆ i i, t 1 it, a g 1 e log( y ), and u it is a random disturbance term. i i i log ˆy t Barro and Sala-i-Martin s (1992) analysis was based on the following equation: 162

174 y 1 1 y i, t t T T it T (A5.17) log c 1 e log yi, t T,, i t t T The dependent variable on the left hand side of Equation (A5.17) is the annualised growth rate of GDP per capita from time t to t+t. On the right-hand side, c is a constant and the yields the convergence speed. The expression y i,t is the initial GDP per capita at time t. Equation (A5.17) can be estimated under conditional convergence through the use of dummies. The inclusion of dummies allows for differentiated intercepts depending on the each region s steady state and growth. The constant in Equation (A5.17) represents the steady state level and its growth. Thus the constant is: T ˆ C x 1 e / T log y xt it. The model assumes that all countries (regions) have the same steady state such that yˆ i yˆ. The model also assumes that the growth rate towards the steady state is the same for all (regions), therefore x i = x. The regional dummies will isolate specific regional effects with respect to steady state gap and growth, as seen earlier. 163

175 Appendix A5.2 Sensitivity Analysis The results reported earlier in this chapter depended on cross-sectional estimation of the convergence effect. It is also possible to estimate the equations using a panel setting to take advantage of the combination of cross-sectional and time series attributes of the data. Furthermore, it is useful to introduce some relevant variables such as population and fixed capital as components of the neo-classical models of growth. Although the emphasis is on convergence per se, the effect of these variables can be significant on the estimates derived. While using panel estimation, the introduction of fixed effects may contribute in improving the explanatory power of the model. Table A5.2.1 reports the linear estimation of Equation (5.2) where the coefficient β indicates convergence or not. Generally, the panel estimates perform better than the cross-sectional coefficients reported above. The overall R 2 is substantially improved although still very low. The introduction of fixed effects of countries improves the β coefficient significantly in columns (2) and (3). The estimated coefficients are ten fold larger compared to those in Table 5.4. They are also statistically significant. This gives further support to the conditional convergence results. Since the GCC countries are of central interest in this study, one of the most important characteristics of these economies is considered in the convergence analysis. The GCC are known as some of the world s major oil producers; the inclusion of such a resource is of considerable interest. The neoclassical theory does not account for natural resources but emphasises the stock of capital at a given point in time. Here the fact that countries are oil producers is taken into account, along with the other conditioning variables to examine any effects such a resource may have on growth and, thus, convergence. Table A5.2.2 reports the cross-sectional estimates of a number of scenarios with respect to oil as a resource. In column (1), a dummy is assigned to countries that produce oil. These countries include developed and developing countries. The coefficient of implies that the intercept for oil-producing countries is different from the rest of the world. This result is statistically insignificant; therefore, oil producers do not have different intercepts compared to the rest of the world. Alternatively, column (2) reports the same regression, but using oil production at the beginning of the period, Oil production levels are positively-related to growth in this case. This may be an intuitive result; countries that produce significant volumes of 164

176 oil may experience greater growth as is the case in the GCC. This result is not statistically significant, however. Finally, the stock of oil measured in proven reserves is taken into account in the regression in column (3). Here, the larger stock of oil reserves improves growth. The coefficient is statistically insignificant for this sample. In the context of the GCC and other oil producers, there is no significant relationship between the stock of oil and growth. In terms of convergence, the change in the coefficient of per capita GDP is marginal when all three scenarios are compared. Thus it is not possible to infer any useful relationships between the natural resource and convergence in this case. 165

177 Table A5.2.1 Conditional Convergence Panel Estimates, Model Variable (1) (2) (3) (4) Constant (0.0336) (0.0848) (0.2531) (0.0463) Per capita GDP (0.0066) (0.0200) (0.0123) (0.0032) Fertility (0.0072) (0.0160) (0.0111) (0.0046) Government Spending (0.0021) (0.0061) (0.0055) (0.0010) Inflation (0.0013) (0.0021) (0.0013) (0.0013) Investment (0.0119) (0.0117) (0.0073) (0.0074) Life (0.0076) (0.0515) (0.0300) (0.0109) Openness (0.0072) (0.0069) (0.0055) (0.0034) Political Rights (0.0029) (0.0057) (0.0036) (0.0043) Terms of Trade (0.0078) (0.0122) Secondary School Enrolment (0.0008) (0.0007) Fixed Effects Country No Yes Yes No Time No No No Yes Adjusted R Number of Observations Note: White hetersokedasticty consistent standard errors in parentheses. 166

178 Table A5.2.2 Cross-Sectional Conditional Convergence Estimates Based On Oil Production Model Variable (1) (2) (3) Constant (0.0271) (0.0272) (0.0274) Per capita GDP (0.0019) (0.0018) (0.0019) Fertility (0.0056) (0.0055) (0.0055) Investment (0.0054) (0.0054) (0.0054) Political Rights (0.0027) (0.0027) (0.0027) Inflation (0.0018) (0.0018) (0.0018) Openness (0.0025) (0.0025) (0.0025) Oil Producing Country (0.0035) Oil Production (0.0003) Proven Oil Reserves (0.0004) Adjusted R Number of Observations Note: White heteroskedasticity consistent standard errors in parentheses. 167

179 Appendix A5.3 Data Descriptions and Sources This appendix describes the data description and sources are listed. Table A5.3.1 describes the variables used in this chapter, their units and where they were obtained from. Table A5.3.2 details the countries included in the sample. Table A5.3.1 Data Descriptions and Sources Variable Units of Measurement Source GDP per Capita PERCAP Population and Population Growth POP and POPGR Fertility FERT Gross Fixed Capital INV Inflation INFL Government Expenditures GOVT Domestic Credit DOMC Real $US (2000) Millions Births per Woman Percent of GDP Percent per Annum Percent of GDP Bank Sources Credit as percent of GDP International Monetary Fund (2007) World Bank Penn World Table 6.2 World Bank (2008b) World Bank (2008b) World Bank (2008b) International Monetary Fund (2007) World Bank (2008b) World Bank (2008b) World Bank (2008b) (Continued on next page) 168

180 Table A5.3.1 (Continued) Data Descriptions and Sources Variable Description Source Secondary School Enrolment SEC Scientific Publication PUB Terms of Trade TOT Political Rights PR Civil Liberties CL LANDLOCKED dummy RTA dummy Total number of students enrolled in secondary schools Number of scientific and technical articles published Net barter terms of trade Index based on survey ranges from 1 to 7, where 7 is most restrictive Normalised to range from 0 to 1 Index based on survey ranges from 1 to 7, where 7 is most restrictive Normalised to range from 0 to 1 Countries with no direct to sea shipping lanes or ports Takes the value of 1 for no access and 0 for otherwise Dummy variable indicate membership within a regional trade agreement Takes the value of 1 if a country is a member and 0 otherwise World Bank (2008b) UNESCO 2008 World Bank (2008b) World Bank (2008b) Freedom House 2008 Freedom House 2008 World Trade Organization

181 Table A5.3.2 Countries Included In Regressions Afghanistan Georgia Oman Albania Germany Pakistan Algeria Ghana Panama Angola Greece Papua New Guinea Antigua and Barbuda Grenada Paraguay Argentina Guatemala Peru Armenia Guinea Philippines Australia Guinea-Bissau Poland Austria Guyana Portugal Azerbaijan Haiti Qatar Bahamas, The Honduras Romania Bahrain Hong Kong SAR Russia Bangladesh Hungary Rwanda Barbados Iceland Samoa Belarus India Sao Tome and Principe Belgium Indonesia Saudi Arabia Belize Iran, Islamic Republic of Senegal Benin Ireland Serbia Bhutan Israel Seychelles Bolivia Italy Sierra Leone Bosnia and Herzegovina Jamaica Singapore Botswana Japan Slovak Republic Brazil Jordan Slovenia Brunei Darussalam Kazakhstan Solomon Islands Bulgaria Kenya South Africa Burkina Faso Kiribati Spain Burundi Korea Sri Lanka Cambodia Kuwait St. Kitts and Nevis Cameroon Kyrgyz Republic St. Lucia Canada Lao People s Democratic Republic St. Vincent and the Grenadines Cape Verde Latvia Sudan Central African Republic Lebanon Suriname Chad Lesotho Swaziland Chile Liberia Sweden China Libya Switzerland Colombia Lithuania Syrian Arab Republic Comoros Luxembourg Taiwan Province of China Congo, Democratic Republic of Macedonia Tajikistan Congo, Republic of Madagascar Tanzania Costa Rica Malawi Thailand C te d Ivoire Malaysia Timor-Leste, Dem. Rep. of Croatia Maldives Togo Cyprus Mali Tonga Czech Republic Malta Trinidad and Tobago Denmark Mauritania Tunisia Djibouti Mauritius Turkey Dominica Mexico Turkmenistan Dominican Republic Moldova Uganda Ecuador Mongolia Ukraine Egypt Morocco United Arab Emirates El Salvador Mozambique United Kingdom Equatorial Guinea Myanmar United States Eritrea Namibia Uruguay Estonia Nepal Uzbekistan Ethiopia Netherlands Vanuatu Fiji New Zealand Venezuela Finland Nicaragua Vietnam France Niger Yemen, Republic of Gabon Nigeria Zambia Gambia, The Norway Zimbabwe 170

182 CHAPTER Introduction PRICE CONVERGENCE This chapter is concerned with the effects of economic integration on prices and considers the effect of reduced trade barriers between countries. The chapter aims to determine if the GCC s economic integration has played a role in reducing price deviations across borders. Within this context we will examine the Law of One Price (LOP) to establish a better understanding of the GCC s experience. Just as effective economic integration may be thought of as an implication of price convergence, departures from LOP may indicate market segmentations that still exist within the region (Goldberg and Verboven 2005, Broda and Weinstein 2008). Such segmentation suggests the presence of barriers to trade and border effects, such as those explored by Engel and Rogers (2001). This chapter will consider the question of price convergence in two major ways. First, studying the dispersion of prices on an aggregate to disaggregate level and observing price behaviour over time will provide an overall picture of the region and its integration progress. We will observe prices from aggregate inflation data to micro prices. Second, the dispersion analysis will be complemented with statistical and econometric analyses of price behaviour. The aim is to test statistically any convergence within the region generally, and between country pairs specifically. The chapter will proceed as follows: Section 6.2 compares GCC inflation patterns over time with those of comparable regions. The comparison will be based on similarities in economic integration experiences and benchmarks. Section 6.3 presents purchasing power parity (PPP) theory, to emphasise the significance of exchange rate arrangements and price convergence. Observing trends in both the GCC and the G7 countries will help to contrast two extreme cases of fixed versus floating exchange rates and their relation to PPP. Section 6.4 examines the convergence of the GCC countries based on deviation from the mean. Here the difference-in-differences approach will be utilised. Section 6.5 disaggregates the inflation data for the years based on eight broad commodity groups. These data are tested for significant convergence (or divergence) with the objective of identifying possible categories of importance. Section 6.6 allows for further disaggregation using micro commodity prices. Using the 171

183 Economist Intelligence Unit (EIU) CityData we employ highly disaggregated prices of consumer goods and services. This section will test convergence within the region and investigate potential border effects. Section 6.7 presents the unique case of Coke, following this single good across different countries to verify if LOP holds within the region or not. The chapter concludes with a summary of findings. 6.2 The GCC Inflation Experience In this section the inflation experience of the GCC countries over last 30 years will be compared with other states and regions. The aim is to establish a benchmark by which GCC countries inflation rates can be compared. The European Union countries, Australian states and territories, and the states of the USA are considered analogous forms of economic integration and thus provide useful benchmarks Inflation within the GCC Region The GCC s economic integration has taken on a mood of urgency in the past few years. After a lull during the late 80s and 90s, significant progress has been made to achieve a fuller economic union. Having established a customs union in 2003 and implemented a common market in 2008, the GCC s next major milestone is monetary union. Successfully establishing a credible monetary union will eliminate exchange rate volatility in intra-gcc trade and commerce. Equally important, the financial sector will benefit from unified laws and regulations that have the potential to facilitate deeper financial markets within the region. The anticipated benefits of monetary union and a common currency may not be realised until their actual launch. This, however, does not mean that some convergence of prices has not already taken place. Trade liberalisation and common market initiatives allow for greater mobility of goods and factors of production, so that price dispersion may already have declined. This premise is supported by the fact that the GCC countries have maintained fixed exchange rates for a long period of time vis-à-vis the dollar. GCC inflation rates have been historically low compared to other developing and developed countries, perhaps the result of the US dollar peg in imposing monetary policy discipline. Figure 6.1 illustrates the weighted inflation rate and standard deviation for GCC countries for the period The solid line represents the weighted inflation and the vertical broken lines indicate the weighted standard deviations. The weights used here and below are proportional to nominal GDPs 172

184 expressed in terms of US dollars. The average inflation rate in the GCC remains in single digits for the period. Price changes of individual countries (not shown in the figure) fluctuate between -8% and 16%, so there has been considerable variability. At the beginning of the period the GCC inflation rates differ by about 5% and in they reach their highest levels. Political events in the region explain the significant difference in prices. Inflation declines until , in both the average and dispersion. This pattern reverses in the last few years of the period under study, where on average inflation increases. GCC countries also exhibit greater dispersion of price changes. The increasing weighted standard deviations of the last 10 years indicate that GCC inflation rates have not converged. Despite low average inflation rates over the whole period, the spread of inflation rates within the region does not decline sufficiently to suggest a significant downward trend. Moreover, price increases towards the end of the period have taken place simultaneously with an increase in dispersion Figure 6.1 GCC Weighted Inflation, (% p.a) Source: Author calculations based on World Economic Outlook Inflation in Other Economic Regions The European Union (EU) is a good benchmark for comparison with the GCC, given the similarities between the two integration initiatives. Figure 6.2 displays the weighted mean of inflation of the EU15 members (Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden, and United Kingdom) for the period The striking feature of Figure 173

185 6.2 is the declining weighted average of the EU15. The 1980s show a substantial decline of inflation within the EU15 from 12% to 4%. This is reversed in the late 1980s when inflation climbs to about 6%. However, during the 1990s and well into the 2000s inflation essentially remains low within the EU15. The post-2001 period shows relative stability in price levels within the region. While inflation declines during the 80s and 90s, dispersion remains large and persistent. The decline in the dispersion, as indicated by the vertical broken lines, is only evident in the second half of the period. However, in 1999, a few years preceding the Euro s official launch, a marked decline in dispersion can be observed. Dispersion remains at low levels for the rest of the period. Figure 6.2 confirms price convergence within the EU15, especially around the time of the Euro launch. Figure 6.2 EU15 Weighted Mean of Inflation (% p.a.) Source: Author s calculations based on World Economic Outlook Australia is a federation of eight states and territories, and is another example of complete economic integration. Figure 6.3 shows the weighted mean of inflation and standard deviations in the Australian states and territories for the period Immediately noticeable is the decline in inflation in the first part of the period. Weighted mean of inflation declines from 8% in 1986 to less than 2% in Prices increase during the mid-90s but decline again with the onset of the Asian financial crisis. Inflation accelerates again from , to peak just below 5%. After 2001, inflation declines. There are marginal increases in inflation until

186 The dispersion of inflation (the weighted standard deviation) between the states and territories tells a different story, however. Figure 6.3 shows minimal dispersion over the period. During the first years of the period dispersion is about 1%, then this declines during the middle years. Low levels of dispersion remain persistent until 2005, when negligible divergence can be detected. Figure 6.3 illustrates that Australians states and territories have converged with respect to inflation. Figure 6.3 Australian States and Territories Weighted Mean of Inflation (% p.a.) Source: Author s calculations and Australian Bureau of Statistics. Another possible comparison to the GCC is the US. Forming a complete political and economic union, the US is an example of a completely integrated region. Figure 6.4 illustrates the mean of inflation for four major US regions: Northeast, Midwest, South, and West. As can be seen from the figure, a sharp decline in the inflation takes place between 1981 and 1983, falling from 10% to about 3%. Inflation dips again in 1986 to 2%, and accelerates to 5% in After the recession of 1991, inflation continues to decline to its lowest levels in 1998, below 2%. Inflation increases again until the slowdown in This decline is temporary, before prices increase again until On average, inflation in the US has declined significantly between 1981 and However, the decline is not smooth. Dispersion, on the other hand, has been significantly low. Although the first few years see inflation dispersion of more than 2%, during the middle years dispersion falls notably. This is especially true in the 1990s during the major expansion period. During the 2000s dispersion increases to pre levels. Overall, the US inflation dispersion remains largely small. The US has partially converged with respect to inflation, and dispersion remains at a minimal level. 175

187 During the same time period, the EU15, the Australian states and territories and the US regions all show substantial price convergence. However, in the case of the GCC, inflation rate differences have remained persistent especially during the past few years. Further analysis of the trends of the GCC inflation rates needs to be undertaken to understand such patterns. Figure 6.4 United States Mean of Inflation (% p.a) Source: Author s calculations based on Bureau of Labour Statistics 6.3 Exchange Rates and Prices Open-economy macroeconomics emphasises an important link between exchange rates and prices in the form of purchasing power parity. PPP theory was briefly introduced in Chapter 2 as a prelude to empirical work reported later in this chapter. In what follows we elaborate further on PPP theory and measurement. The Law of One Price (LOP) postulates that the price of a commodity consumed in two different countries will be equalised when compared in a common currency, given no barriers to trade and ignoring transportation costs. Price differentials will be eliminated based on arbitrage. Purchasing power parity (PPP) represents a broader view of this price equalisation proposition. The PPP theory, in absolute form, proposes that a basket of goods typically consumed domestically will be equal in value to an identical basket consumed by consumers in a different country. One of the most common examples of testing this relationship is the Big Mac Index, introduced by The Economist magazine in the 1980s. Big Mac burger prices are compared across a number of 176

188 countries, with all prices converted into US dollars. Deviations from parity are assumed to be over- or under-valuation of the respective currencies. The groundwork for the theory behind PPP was laid by Gustav Cassel (1928). Balassa (1964) introduced a model of PPP that included productivity. Productivity differentials in his model help explain relative price changes between traded and nontraded goods within countries, and lead to the conclusion that those countries with higher productivity differentials between traded and non-traded sectors experience larger increases in the prices of non-traded goods. Consequently this causes real exchange appreciation of the currencies of rich (higher productivity) countries. In the same regard, Samuelson s (1964) work builds on productivity differences and their effects on relative prices. This relationship between productivity differentials relative prices/currency values is known as the Balassa-Samuelson effect. In what follows we describe PPP theory and provide a brief overview of empirical tests of the approach Versions of PPP The law of one price is the fundamental starting point of PPP. In its basic form, it says that the price of a good should be the same anywhere, once accounted for in a single currency. Thus: (6.1) * where P i and P are the prices of the good in question at home and abroad; and S is the i P SP, nominal exchange rate (the domestic currency cost of a unit of foreign currency). Equation (6.1) will hold true if transportation costs and trade barriers are assumed to be nonexistent. This is likely to be the case with gold, at least as an approximation; but for a number of other internationally traded commodities such as grain, minerals, or energy, transportation costs need to be taken into account. The LOP can be extended to price indexes as well as to individual commodity prices: Pi and P become i P and P, the costs of purchasing a basket of goods at home and abroad. The absolute version of PPP in this case is the same as that in Equation (6.1): P SP, or P (6.2) S. Equation (6.2) stipulates that the exchange rate is the ratio price levels. i P The quantity SP is the cost of the foreign basket measured in terms of domestic currency; under PPP, this cost coincides with the cost of the domestic basket, P. i 177

189 Accordingly, P SP measures the extent to which PPP does not hold. It is convenient to express this disparity logarithmically: P (6.3) q log. SP When PPP holds q = o, whereas when q > 0 the domestic currency is overvalued as prices at home are too high relative to those abroad; when q < 0 the currency is undervalued. It is for this reason that q of Equation (6.3) is known as the real exchange rate. The above discussion is based on the assumption of no transportation costs or barriers to trade. However, when such costs are involved, as is often the case in the real world, absolute PPP cannot hold. In this case Equation (6.2) becomes P (6.4) S, P where θ is a constant that represents the wedge between domestic and foreign prices. If we differentiate Equation (6.4) with respect to time and take logs, the constant θ drops out and we get (6.5) S ˆ P ˆ P ˆ, where a circumflex (^) denotes proportional change. Before concluding this overview of PPP theory, two further comments are in order. First, in most modern economies much more than one half of all goods and services produced do not enter international trade. Examples are construction and many services that are labour intensive; haircuts are the classic example. The prices of these non-traded goods can be linked to those traded via substitution in consumption and production. Thus, the equalisation across countries of traded good prices, via the arbitrage mechanism discussed above, can also lead to equalisation of the prices of notraded goods. In this case, then, the existence of non-traded goods in the market baskets does not necessarily invalidate PPP theory; but as the substitution in consumption and production may take substantial time, the existence of non-traded goods may give rise to long lags in the equalisation of prices across countries. Another way of thinking about the impact of non-traded goods is in terms of transportation costs. Gold is a good that has a high value-to-weight ratio, so its transport cost in terms of its value (the proportional transport costs) are low. It is for this reason that gold is easily traded internationally, and, via arbitrage, its price is equalised around the world. By contrast, many other goods have low value-to-weight ratios and have high transport costs; bricks are an example. When transport costs are sufficiently high, the 178

190 good in question is not traded internationally: that is, it becomes a non-traded good. This link between transport costs and non-traded goods can be used to deal with goods that do not enter into international trade via the value of parameter θ in Equation (6.4) above. The second point worth mentioning is what is known as the stochastic version of PPP (Clements et al. 2010). Here, random deviations from parity are allowed for. While prices are not exactly equalised, if they fall in a neutral band then the evidence can still be not inconsistent with this version of PPP Empirical Methods of Testing PPP The concept of PPP has been extensively tested. Froot and Rogoff (1995) provide an extensive survey on the subject of PPP and the empirical testing methods used. The following draws on this survey to summarise some methods and approaches. Earlier empirical developments took the form used in Frenkel (1978): (6.6) s p p, t t t t where s t is the log of the spot exchange rate at time t, 179 pt and p are the log price levels t of the domestic and foreign countries at time t, t is a random error, and α and β are parameters to be estimated. Testing for PPP validity was based on β=1. Results of such models generally rejected PPP when samples did not include hyperinflation episodes. These models ignored two main issues: endogeneity and stationarity of exchange rates. The former was addressed by using instrument variables such as time trends and lags of inflation and exchange rates. The latter was never adequately addressed in this type of empirical testing. An alternative approach, which Froot and Rogoff (1995) describe as the second stage, shifted focus to real exchange rates. Such models tested for a null hypothesis of random walk in real exchange rates. The alternative tested for stationarity of the real exchange rate as defined in Equation (6.3). One of the common techniques used here is the Dickey-Fuller (1979) test that regresses the real exchange rate on its own lag, a trend, and differenced polynomial representation of the real exchange rate. The third stage of empirical tests methodologies used to investigate PPP is based on cointegration type tests. Based on Engle and Granger s (1987) two-step process, cointegration theory proposes that linear combinations of nonstationary series can be stationary. An alternative approach is Johansen (1991), a single step cointegration test that addresses some of the inefficiencies of the two-step process (Froot and Rogoff 1995).

191 The above methodologies use consumer price indices (CPIs) or wholesale price indices (WPIs) to test for stationarity of real exchange rates. In contrast, Isard (1977) uses disaggregated prices. Comparing US and German export prices, he finds persistent deviations. Giovannini (1988) finds similar deviations when studying Japanese export prices. In both cases the nominal exchange rate is found to be correlated strongly with relative prices (Froot and Rogoff 1995). To avoid some of the pitfalls of using CPI data to study real exchange rates, in 1986 The Economist magazine developed the Big Mac Index (BMI). The BMI utilises the Big Mac burger as a unit of comparison, and the ratio of Big Mac prices in two countries, measured in US dollars and converted using spot exchange rates, indicates the real exchange rate. This is used to determine if currencies are over- or undervalued. The novelty in this approach is due to the comparison of an identical product that forms a basket. Unlike the CPI, where the basket of consumer goods varies from one country to another, the BMI has identical ingredients and components. Several studies have used the BMI index to examine real exchange rate behaviour over time (Ong (2003) and Clements et al. (2010)). A recent study by Hassanain (2004) of PPP with relevance to the GCC indicates that the relationship holds. Taking ten countries, including five GCC members, Hassanain (2004) tests the proposition that PPP holds for the period Using the real exchange rate, and using alternate numeraire base currencies, he is able to reject the existence of a unit root. Hassanain (2004) estimates the half-life to be 2.2 years for the five GCC countries included in the sample. The long period of mean reversion is explained by the firm pegs to the US dollar, noted above, and poor arbitrage opportunities in the goods market. He also shows that the volatility of the US dollar contributes to deviation from the mean. The results of Hassanain s proposition will be revisited here and its validity verified. While LOP and PPP remain the main theories to investigate price convergence, descriptive methods such as dispersion have also been used. A number of studies have addressed the question of price convergence with respect to economic integration areas, primarily the European Union. Rogers (2007), for example, compares Europe and the US by observing the dispersion of prices within major cities in both continents. He finds that price volatility declined within Europe generally and the EU11 specifically prior to the launch of the Euro. This was in contrast to US price volatility. Rogers (2007) finds that price convergence has taken place within the EU although it may not be attributable to the Euro per se. He cites contributory factors such as policy coordination and convergence, trade liberalisation, and a host of other potential effects. He does not 180

192 analyse these effects but merely points out that they may have contributed to reductions in price dispersion. Engel and Rogers (2001), on the other hand, use the variance of prices as a primary measure in their model to capture deviations from PPP. Their model includes more traditional variables such as distance and border dummies as well GCC verses the Group of Seven As the theory of PPP does not refer to the exchange rate regime, it is equally applicable to fixed or floating exchange rates. Under a fixed rate, the prices adjust to bring them in line with the exchange rate; under a floating regime, prices and rate adjust jointly. Under a fixed rate a tendency for the equalisation of inflation rates is to be expected, but floating rates imply that countries can choose their own rates of inflation, so these will in all likelihood differ. In this sub-section, we compare the experience of the GCC, where exchange rates are more or less fixed, with that of the Group of Seven (G7), where currencies have floated. Figure 6.5 demonstrates the stability of the GCC exchange rates during the period The figure displays two axes: on the right-hand side we can detect the exchange rates of Qatar, Saudi Arabia and the UAE. The left-hand axis measures the exchange rates of Bahrain, Kuwait and Oman. Apart from some early adjustment in the 1980s the period shows constant exchange rates. Consequently, the cross-exchange rates within the GCC have been stable over time. In effect, this near-fixed exchange rate regime means that these countries have a pseudo-currency-union; therefore, some degree of the benefits associated with monetary unions and common currencies may apply. One of these benefits is potential price convergence within the region. In contrast to the GCC s fixed exchange rates, the Group of Seven (G7) have liberalised their exchange rates to move freely against each other. Figure 6.6 illustrates the movement of the exchange rates of the G7 countries. The right-hand axis measures the exchange rates of France, Germany and Japan. The left-hand side axis measures the exchange rates of the other G7 countries and the Euro. The degree of volatility experienced by the G7 countries post-bretton Woods is clear. Shown in Figure 6.7 are the individual inflation rates of the six GCC countries. For clarity they are divided into two groups. Figure 6.7 does not indicate a tendency towards equalisation: inflation rate differences remain persistent within the GCC. Although the inflation rate differentials decline during some years, towards the end of the period there are significant deviations between the countries. 181

193 Figure 6.5 GCC Exchange Rates ($US cost of a unit of local currency) Source: World Economic Outlook 2009 Figure 6.6 G7 Countries Exchange Rates ($US cost of a unit of local currency) Source: World Economic Outlook