Trade agreements galore

Uitnodiging som | 2012

som

Tristan Kohl

Trade agreements galore This book identifies who the key actors are on the stage of international trade policy. It explores what governments regulate by taking an in-depth look at the legal provisions contained in hundreds of economic integration agreements. We line up the arguments for and against regionalism and show how international trade institutions affect world trade. Our results, based on augmented gravity equations and novel, extensive datasets, quantify how much countries gain (or lose) from participating in the World Trade Organisation, the Generalised System of Preferences and other trade agreements. We also illustrate where trade blocs emerge and discuss when and why they change.

Research school SOM Faculty of Economics and Business University of Groningen PO Box 800 9700 AV Groningen The Netherlands

Tristan Kohl Trade agreements galore

faculty of economics and business

Hierbij nodigt Tristan Kohl

Tristan Kohl Trade agreements galore Who, what, when, where, why, how and how much?

Trade agreements galore

Who, what, when, where, why, how and how much? De verdediging vindt plaats op donderdag 18 oktober 2012 om 12.45 uur precies in de aula van het Academiegebouw van de Rijksuniversiteit Groningen, Broerstraat 5, Groningen. Aansluitend is er een receptie in de Spiegelzaal.

Among the key findings are that WTO membership fosters international trade, although countries’ gains go hand in hand with the extent of their multilateral commitments. Regionalism also has favourable effects, yet individual agreements often have no discernible impact. Finally, quantification of the underlying legal provisions shows that comprehensive agreements are good for world trade, but this does not hold for their more modern legal provisions.

Paranimfen: Suzanne Kok 06 24 73 42 06 [email protected] Hans Smilde 06 48 97 19 84 [email protected]

www.rug.nl/feb/onderzoek ISBN 978-90-367-5709-6

9 789036 757096

u uit voor het bijwonen van de openbare verdediging van zijn proefschrift, getiteld

Theses in Economics and Business

Trade agreements galore Who, what, when, where, why, how and how much? Tristan Kohl

Publisher:

University of Groningen Groningen, The Netherlands

Printed by:

Ipskamp Drukkers B.V.

ISBN:

978-90-367-5709-6

eISBN:

978-90-367-5708-9

c 2012 Tristan Kohl

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system of any nature, or transmitted in any form or by any means, electronic, mechanical, now known or hereafter invented, including photocopying or recording, without prior written permission of the publisher.

Trade agreements galore Who, what, when, where, why, how and how much?

Proefschrift

ter verkrijging van het doctoraat in de Economie en Bedrijfskunde aan de Rijksuniversiteit Groningen op gezag van de Rector Magnificus, dr. E. Sterken, in het openbaar te verdedigen op donderdag 18 oktober 2012 om 12.45 uur

door

Tristan Kohl geboren op 4 mei 1983 te Pietersburg, Zuid-Afrika

Promotores:

Prof. dr. S. Brakman Prof. dr. J.H. Garretsen

Beoordelingscommissie:

Prof. dr. R.J.M. Alessie Prof. dr. J.H. Bergstrand Prof. dr. P.A.G. van Bergeijk

To my parents

Acknowledgements “. . . and then there are legal forces that may affect the company’s profitability, but we will not discuss them in detail because we are, after all, economists.” During a strategic management course, these very words convinced me that something had to be done about such a lack of interest in legal affairs. I took up reading law in my spare time and combined law and economics in my Master’s thesis. Bart van Ark and Marcel Timmer encouraged me to consider pursuing a career in academia. Ger Lanjouw introduced me to the world of international trade institutions. Thank you for getting the ball rolling. Steven Brakman challenged me to raise the bar and start the PhD project that has culminated in this dissertation. Harry Garretsen joined when he moved to Groningen. Working with Steven and Harry is a striking example of Proverbs 27:17 in action. I am deeply grateful for their expert honing of my academic skills and unwavering encouragement, every step of the way. In addition, I would like to thank Rob Alessie, Jeff Bergstrand and Peter van Bergeijk for finding the time to provide extensive and stimulating feedback on a previous version of the manuscript. Parts of this dissertation were presented at international conferences and/or inspired by participating in the WTO Summer Programme at the Graduate Institute, Geneva. I am grateful to Richard Baldwin, Gerrit Faber, Marco Fugazza, Michael ¨ Kotter, Xuepeng Liu, Will Martin, Phil McCann, Joseph Pelzman, Kevin Staub, Cees Sterks, Steve Suranovic, Charles van Marrewijk, Liane Voerman and Per Wijkman for fruitful discussions and helpful suggestions. Over the years, my roommates Hengki, Jianhong, Klaas, Robert and Roger have made going to work a real treat. Thanks for bearing with my (short-lived) in-door fountain, plants and doodles on the white board. Sharing an office with Aleid has led to the writing of a joint paper (with more to come) and, even better, an invaluable friendship.

iv A special word of appreciation goes to Ana, Beppo, Dirk A., Gaaitzen, Gijsbert, Janneke, Jutta, Maaike, Monique, Padma, Sathyajit, Shu and the fellowship of room 643 for good times and moral support. Thanks to the GEM gems and the SOM office for service with a smile, and my colleagues throughout the Duisenburg Building for creating an enjoyable working environment. Thanks also to all my students for their eagerness to learn and willingness to hear me go on (and on) about my research. I look forward to defending my thesis with Suzanne and Hans by my side. Suzanne is my academic comrade in arms with a contagious dose of ambition and joie de vivre. By now, Hans has probably learnt too much economics than is good for him. Attending one of my lectures “for fun” is a true sign of his support and friendship. Bert, Gideon, Ingmar, Marie-Eve, Marlies, Rina and Santanu are duly noted for all the ways they have found to keep my feet on the ground. Baie dankie and tige tank to my (extended) family for their undying love and support. I am used to medical conversations at the dinner table because of my parents’ professional background. My favourite topic was anatomy (for lack of understanding anything else) which probably explains why chapter 5 makes this book feel like a “dissectation”. My wife, Ammerens, is the one who keeps me going. She shows endless patience when there is “just one more calculation” left to do during weekends, puts a smile on my face no matter what the circumstances, and makes life all the sweeter when there is cause for celebration. I don’t need a PhD to know that marrying her is the best thing I ever did.

Tristan Groningen, August 2012

Contents 1

Introduction

1

1.1

Setting the stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2

1.2

The WTO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4

1.2.1

Why liberalise trade? . . . . . . . . . . . . . . . . . . . . . . . .

6

1.2.2

The cost of protectionism . . . . . . . . . . . . . . . . . . . . .

7

Regionalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

9

1.3

1.4 2

3

1.3.1

The rise of regionalism . . . . . . . . . . . . . . . . . . . . . . .

12

1.3.2

Welfare implications . . . . . . . . . . . . . . . . . . . . . . . .

17

1.3.3

Building vs. stumbling blocks . . . . . . . . . . . . . . . . . . .

19

Through the looking glass . . . . . . . . . . . . . . . . . . . . . . . . .

21

The WTO’s effect on trade: What you give is what you get

25

2.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

25

2.2

Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

26

2.3

Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

32

2.3.1

Regionalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

32

2.3.2

Econometric estimation . . . . . . . . . . . . . . . . . . . . . .

36

2.3.3

Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

41

2.4

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

45

2.5

Discussion & conclusion . . . . . . . . . . . . . . . . . . . . . . . . . .

52

The development of trade blocs in an era of globalisation

57

3.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

57

3.2

Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

60

3.2.1

Regionalisation of nations . . . . . . . . . . . . . . . . . . . . .

60

3.2.2

Importance of regionalisation . . . . . . . . . . . . . . . . . . .

61

Trade blocs in the world economy . . . . . . . . . . . . . . . . . . . .

64

3.3

vi

Contents

3.4

3.5 4

Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

64

3.3.2

Static vs. dynamic trade blocs . . . . . . . . . . . . . . . . . . .

66

3.3.3

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

68

What drives the development of trade blocs? . . . . . . . . . . . . . .

72

3.4.1

Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

72

3.4.2

Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

73

3.4.3

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

75

Discussion and conclusion . . . . . . . . . . . . . . . . . . . . . . . . .

81

Do we really know that trade agreements increase trade?

83

4.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

83

4.2

Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

86

4.2.1

Genesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

86

4.2.2

Old school . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

88

4.2.3

New school . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

89

Methodology and results . . . . . . . . . . . . . . . . . . . . . . . . . .

90

4.3.1

Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

90

4.3.2

Benchmark estimates . . . . . . . . . . . . . . . . . . . . . . . .

91

4.3.3

Individual EIA effects . . . . . . . . . . . . . . . . . . . . . . .

96

4.3

4.4 5

3.3.1

Discussion and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . 103

I just read 296 trade agreements

107

5.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

5.2

What is in an EIA? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

5.3

5.4

5.2.1

Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

5.2.2

Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

5.2.3

Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

5.2.4

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

What makes an EIA comprehensive? . . . . . . . . . . . . . . . . . . . 131 5.3.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

5.3.2

Enter: EIA index . . . . . . . . . . . . . . . . . . . . . . . . . . 132

5.3.3

Determinants . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

5.3.4

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

Are comprehensive EIAs good for trade? . . . . . . . . . . . . . . . . 143 5.4.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

5.4.2

Approach and results . . . . . . . . . . . . . . . . . . . . . . . 143

5.4.3

EIA effects revisited . . . . . . . . . . . . . . . . . . . . . . . . 152

Contents 5.5

vii

Discussion and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . 154

Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156 6

Conclusions

Samenvatting (Summary in Dutch)

161 181

List of Tables 1.1

Import tariffs on manufactured goods, % . . . . . . . . . . . . . . . .

3

1.2

Domestic cost of protection for selected US industries, 1990 US$ million

1.3

Number of enforced EIAs per country, 2011 . . . . . . . . . . . . . . .

14

2.1

Results in the WTO effect literature . . . . . . . . . . . . . . . . . . . .

27

2.2

Developments in the WTO effect literature . . . . . . . . . . . . . . .

31

2.3

EIAs by year of enforcement . . . . . . . . . . . . . . . . . . . . . . . .

34

2.4

Countries in the dataset . . . . . . . . . . . . . . . . . . . . . . . . . .

42

2.5

Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . .

43

2.6

Data sources in the WTO effect literature . . . . . . . . . . . . . . . .

44

2.7

Rose (2004) revisited . . . . . . . . . . . . . . . . . . . . . . . . . . . .

46

2.8

Model selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

47

2.9

(ZI)NB results, 1948-2007 . . . . . . . . . . . . . . . . . . . . . . . . . .

48

2.10 ZINB results by development status, 1948-2007 . . . . . . . . . . . . .

49

2.11 (ZI)NB results by income classification, 1948-2007 . . . . . . . . . . .

50

2.12 ZINB results by negotiation round . . . . . . . . . . . . . . . . . . . .

51

3.1

Predictive power of a static regional indicator . . . . . . . . . . . . . .

67

3.2

Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . .

74

3.3

Countries in dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

74

3.4

Pooled probit estimation results, 1950-2005 . . . . . . . . . . . . . . .

76

3.5

Cross-sectional probit estimation results, 1950-2005 . . . . . . . . . .

78

3.6

Probit estimation results based on random sampling . . . . . . . . . .

80

4.1

Countries in dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

91

4.2

Benchmark gravity estimates . . . . . . . . . . . . . . . . . . . . . . .

93

4.3

First-difference estimates, 1948-2007 . . . . . . . . . . . . . . . . . . .

95

4.4

Individual EIA effects . . . . . . . . . . . . . . . . . . . . . . . . . . .

97

8

x

List of Tables 5.1

Provisions in GPTAD: WTO+ . . . . . . . . . . . . . . . . . . . . . . . 112

5.2

Provisions in GPTAD: WTOX . . . . . . . . . . . . . . . . . . . . . . . 113

5.3

Provisions in GPTAD: Institutional Quality . . . . . . . . . . . . . . . 113

5.4

Coding examples: WTO+ . . . . . . . . . . . . . . . . . . . . . . . . . 115

5.5

Coding examples: WTOX . . . . . . . . . . . . . . . . . . . . . . . . . 120

5.6

EIAs by year of enforcement . . . . . . . . . . . . . . . . . . . . . . . . 124

5.7

Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

5.8

Descriptive statistics: EIA index . . . . . . . . . . . . . . . . . . . . . . 134

5.9

Comparison of group means . . . . . . . . . . . . . . . . . . . . . . . . 135

5.10 Descriptive statistics: Determinants . . . . . . . . . . . . . . . . . . . . 140 5.11 Determinants of EIA comprehensiveness indices . . . . . . . . . . . . 142 5.12 Basic gravity equation estimates . . . . . . . . . . . . . . . . . . . . . 145 5.13 Extended gravity equation estimates . . . . . . . . . . . . . . . . . . . 148 5.14 WTO+, WTOX and IQ provisions’ effect on trade . . . . . . . . . . . . 151 5.15 Overview of factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

List of Figures 1.1

Exports as a percentage of GDP, 1990 US$ . . . . . . . . . . . . . . . .

4

1.2

Number of WTO participants, 1948-2011 . . . . . . . . . . . . . . . . .

6

1.3

Regionalism in 1970 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

10

1.4

Regionalism in 2011 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

10

1.5

The proliferation of EIAs, 1948-2011 . . . . . . . . . . . . . . . . . . .

13

1.6

Number of countries with EIAs, 1948-2011 . . . . . . . . . . . . . . .

13

1.7

Number of EIAs by level of development, 1948-2011 . . . . . . . . . .

15

2.1

Distribution of missing, zero and positive values per year . . . . . . .

38

3.1

Trade bloc orientations in 1950 . . . . . . . . . . . . . . . . . . . . . .

69

3.2

Trade bloc orientations in 1960 . . . . . . . . . . . . . . . . . . . . . .

69

3.3

Trade bloc orientations in 1970 . . . . . . . . . . . . . . . . . . . . . .

69

3.4

Trade bloc orientations in 1980 . . . . . . . . . . . . . . . . . . . . . .

70

3.5

Trade bloc orientations in 1990 . . . . . . . . . . . . . . . . . . . . . .

70

3.6

Trade bloc orientations in 2000 . . . . . . . . . . . . . . . . . . . . . .

70

3.7

Trade bloc orientations in 2005 . . . . . . . . . . . . . . . . . . . . . .

71

4.1

Individual estimates by year of enforcement . . . . . . . . . . . . . . 101

4.2

Individual estimates by number of participants . . . . . . . . . . . . . 101

4.3

Individual estimates by development status . . . . . . . . . . . . . . . 102

4.4

Individual estimates by geographic scope . . . . . . . . . . . . . . . . 102

5.1

Benchmarks of EC and US EIAs . . . . . . . . . . . . . . . . . . . . . . 127

5.2

Descriptive statistics: Provision scores . . . . . . . . . . . . . . . . . . 130

5.3

EIA effects and I E . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

Chapter 1

Introduction I have never faltered, and I will never falter, in my belief that enduring peace and the welfare of nations are indissolubly connected with friendliness, fairness, equality and the maximum practicable degree of freedom in international trade. Cordell Hull US Secretary of State (1933-1944)1

It happened in 1947. Government officials from 23 countries gathered in Geneva to sign an international agreement known as the General Agreement on Trade and Tariffs (GATT). Its objective is to liberalise trade restrictions among all its participants on the basis of equal treatment.2 With its 157 members to date, the GATT—known as the World Trade Organisation (WTO) since 1995—has grown to become the world’s largest intergovernmental platform promoting free(r) international trade.3 The growth in the multilateral trade system’s membership base has gone hand in hand with a dramatic reduction in tariffs, thanks to its members’ ongoing efforts to liberalise trade in a non-discriminatory fashion. The WTO boasts that half of world trade is now subject to zero tariffs (WTO, 2011e).

1 Source:

Dam (1970, p. 12). a detailed account of the political forces involved with the GATT’s creation, see Hoekman & Kostecki (2001) and Barton, Goldstein, Josling & Steinberg (2006). 3 Countries that acceded to the GATT (1947-1994) are formally called contracting parties and countries acceding to the WTO (1995-present) are members. For simplicity, “GATT” and “WTO” are used interchangeably throughout this book, as are the terms “contracting party” and “member”. 2 For

2

Chapter 1 The significant achievements of the multilateral trade system instil a sense of

awe at the vision that its founders—among others Cordell Hull—had for the “new world order” in the postwar era. It also feeds the impression that non-discriminatory trade liberalisation promoted by the WTO has caused its members’ international trade to flourish. However, the WTO is not the only institution shaping international trade policy. It will be shown that the past five decades have witnessed a striking buildup of trade agreements that violate the WTO’s cardinal principle of non-discrimination. These agreements—called economic integration agreements (EIAs)—are discriminatory in nature, as their participants use them to grant preferential tariffs to selected partners only.4 Despite the WTO’s ongoing non-discriminatory trade reforms, the remarkable growth of EIAs casts doubt on the multilateral trade system’s ability to advance trade among its members. Are EIAs a viable alternative to the WTO, or is there room for both forms of institutional arrangements to coexist? The remainder of this chapter introduces the issues that are central to this thesis. The organisation is as follows. Section 1.1 describes the political economic climate that gave rise to the creation of the multilateral trade system. Section 1.2 explains the main principles upon which the WTO is based and explains the rationale of trade liberalisation. Section 1.3 explores the evolution of trade blocs in the world economy and their potential role in the multilateral trade system. Finally, section 1.4 develops the roadmap that will be used to explore the interplay between international trade and its institutions throughout this book.

1.1

Setting the stage

With the onset of the Industrial Revolution in the late nineteenth century, a number of western economies opened up to international trade. Table 1.1 shows how several major economies reduced their import tariffs on manufactured goods between 1820 and 1875. In abolishing prohibitive trade policies, France, Russia, Spain and Sweden experienced substantial tariff cuts. Tariffs were approximately halved in Denmark and the Netherlands, while they were virtually completely abolished by the UK. This led to an initial surge in trade which was dubbed the “first wave of globalisation” (Baldwin & Martin, 1999). 4 EIAs are used as an umbrella term to cover any trade agreement that is not part of the WTO’s multilateral system, including regional trade agreements, free trade areas, customs unions, etc.

Introduction

3

Table 1.1. Import tariffs on manufactured goods, % Country Denmark France Germany Italy Russia Spain Sweden Netherlands United Kingdom United States

1820 30 Prohibition N.A. N.A. Prohibition Prohibition Prohibition 7 50 45

1875 15-20 15-20 4-6 8-10 15 15-20 3-5 3-5 0 40

1913 14 20 13 18 84 41 20 4 0 4

1931 N.A. 30 21 46 Prohibition 63 21 N.A. N.A. 48

1950 3 18 26 25 Prohibition N.A. 9 11 23 14

1998 4 4 4 4 13 4 4 4 4 5

Note: N.A. means “not available”. Source: Brakman, Garretsen, van Marrewijk & van Witteloostuijn (2006).

However, world trade started to dwindle with tariffs gradually rising in a number of countries during World War I. The Great Depression caused the so-called “great trade collapse”. In the Federal Reserve’s attempt to restrain an overzealous US stock market with contractionary monetary policy, higher interest rates lowered investment and domestic demand. Countries committed to maintaining the gold standard followed suit, increasing pressure to implement protectionist trade policy (Eichengreen, 1998). A famous example of the US’ protectionism is the Smoot-Hawley Tariff Act, which was created to allegedly prevent unemployment in a sluggish American economy and to provide relief to farmers facing international competition. The bill sent US import tariffs soaring by an average of 20 percent and led to a 7 percent drop in US imports in its first two years (Irwin, 1998). Many trade partners retaliated, leading to widespread “beggar-thy-neighbour” policies aimed at shifting effective demand to domestic goods and barring imports by means of tariffs, quantitative restrictions and currency devaluations. Figure 1.1 depicts various economies’s sales (exports) to the rest of the world, divided by its annual value added (GDP) in order to correct for inflation and economic growth. The cataclysmic economic impact of the Great Depression and World War II sent governments scrambling to find solutions that could restore their countries to economic health. Some nations adopted more liberal trade policies again by the 1950s, as witnessed by Table 1.1, but world trade was only back at its level of the 1930s by the 1960s. Since then, however, the “second wave of globalisation” has witnessed progressive trade liberalisation.

4

Chapter 1

Figure 1.1. Exports as a percentage of GDP, 1990 US$

Source: Maddison (2001), Table F-5.

1.2

The WTO

Governments around the world acknowledged the benefits and importance of trade liberalisation in the post-war climate of the 1940s. Led by the US and UK, they set about creating international institutions to facilitate their economic recovery and cooperation. These efforts culminated in 23 nations signing the GATT in 1947,

. . . (r)ecognising that their relations in the field of trade and economic endeavour should be conducted with a view to raising standards of living, ensuring full employment, a large and steadily growing volume of real income and effective demand, developing the full use of the resources of the world and expanding the production and exchange of goods . . . (b)eing desirous of contributing to these objectives by entering into reciprocal and mutually advantageous arrangements directed to the substantial reduction of tariffs and other barriers to trade and to the elimination of discriminatory treatment in international commerce. Preamble of the GATT

Introduction

5

The WTO facilitates trade liberalisation by providing a platform for its participants to discuss trade reforms, based on the principles of (1) non-discrimination, (2) first-difference reciprocity and (3) transparency WTO (2011d). Ad 1. Non-discrimination is embodied by the most-favoured nation (MFN) clause (Article I GATT). It means that the most extensive trade concessions negotiated between two participants are extended to all participants. Another aspect of non-discrimination is national treatment, which requires members to provide equal treatment to domestic and foreign providers of goods, services, capital and intellectual property once they have entered the domestic market. Ad 2. First-difference reciprocity means that changes in the benefits (concessions) extended by one party to another should not be left unrewarded. This is meant to ensure mutually beneficial outcomes so that that all parties have an incentive to make concessions. Ad 3. A transparent, predictable trade system is reliable and should facilitate cross-border trade in goods, services and investment. Once trade instruments are “bound” by negotiations, the international community has certainty that WTO members will not suddenly raise trade barriers above the levels to which they have committed themselves. Transparency also involves the use of tariffs as the major instrument of trade restriction and discourages non-tariff barriers, which are considered vague and complex. Throughout the multilateral trade system’s existence, it has witnessed an increase both in the number of participants (see Figure 1.2) and in its scope.5 The WTO participants have set about fulfilling the GATT’s objective of gradual trade liberalisation by negotiating concessions during so-called trade rounds. The first five rounds occurred between the late 1940s and early 1960s and focused on tariff liberalisation. The Kennedy Round (1964-1967) included negotiations on anti-dumping measures. The Tokyo Round (1973-1979) did not only address further tariff reforms, but also included topics such as non-tariff reforms, technical barriers to trade, subsidies and import licensing procedures. The Uruguay Round (1986-1994) laid the foundation for the GATT to be incorporated in the WTO, thereby extending the scope of the multilateral trade system to, amongst others, a binding dispute settlement procedure, agreements on intellectual property and trade in agricultural products, services and textiles. The Doha Round, based on the Doha Development Agenda (DDA), has been underway since 2001 and aims at improving less-developed nations’ trade positions in the word economy. 5 The transition from GATT to WTO forced informal participants to either become formal members or to opt out, which explains the small drop in members after 1995. Details are provided in chapter 2.

6

Chapter 1

Figure 1.2. Number of WTO participants, 1948-2011

Sources: Tomz, Goldstein & Rivers (2007) and WTO (2011a,b).

1.2.1

Why liberalise trade?

The economic gains from free(r) trade can be classified as either static or dynamic. Static gains from trade are welfare improvements resulting from changes in relative prices, without altering the production base. Consumers gain because trade liberalisation entails a change in relative prices that enables them to purchase a greater number and/or variety of goods, called static consumer gains. This change in relative prices also leads to a redistribution of productive resources to comparative-advantage industries, which enables even more consumption and a higher welfare level, also known as static production gains. Dynamic gains are embodied by economic growth that results from an increase in the domestic production base. This may happen for several reasons. First, trade in productive resources such as capital and labor allows firms to develop a more efficient production base. The availability of and access to foreign inputs speeds up the process of economic growth. Second, trade liberalisation means (better) access to foreign markets, providing opportunities for firms to obtain economies of scale. As a result, trade induces more competition on domestic markets, challenging inefficient incumbents and those able to exploit market power to join the race for better performance or to leave the market. In the long run, resources are reallocated to be used more efficiently.

Introduction

7

Third, competitive pressure from domestic and foreign rivals alike induces a search for competitive advantage by stimulating research and innovation to further reap productivity gains. Finally, free(r) trade causes industries and nations to become more dependent on their foreign counterparts. Such interdependencies may reduce the probability of international hostilities and even facilitate international cooperation and security. In other words, the benefits of trade may be extended into realms other than that of pure economics. The benefits of open trade notwithstanding, there is not a single country in the world that does not restrict trade by means of one or the other protectionist policy. Why? For starters, trade policy can be used by price-setting countries to alter their terms of trade in such a way that they obtain a net welfare gain. However, the country’s trade partners pick up the price tab, which is (world) welfare decreasing. Another reason to restrict trade is to (temporarily) protect emerging domestic industries from foreign competition in order to acquire the skills needed to successfully compete on international markets. This so-called infant industry argument especially applies to developing countries. The argument can be extended to apply to any industry that a government considers to have positive externalities, i.e. the success of the protected industry is expected to cause growth in other domestic industries. Moreover, governments could choose to restrict trade in the interest of national defense, public health and safety, or environmental protection. But trade policy can also be considered to be an instrument in an attempt to coerce foreign foes (and allies) to change their behaviour to obtain a more “desirable” outcome. History abounds with examples, including the Organisation of Arab Petroleum Exporting Countries (OAPEC)’s response to US involvement with Israel in the Yom Kippur war that led to the 1973 oil crisis, the western world’s “disinvestment” embargo on South Africa during its apartheid regime, and American and European (arms) embargoes on nations such as Iran, Myanmar, North Korea and Sudan.

1.2.2

The cost of protectionism

What are the costs of protectionism? The economic impact of trade policy reforms is commonly measured by determining their welfare effects in partial equilibrium or computable general equilibrium models. These effects are decomposed in changes in consumer surplus, producer surplus, and government revenue (see, e.g., Brown et al., 2001; van Marrewijk, 2007).

8

Chapter 1

Table 1.2. Domestic cost of protection for selected US industries, 1990 US$ million Industry Frozen orange juice concentrate Rubber footwear Ceramic tiles Luggage Women’s purses Chinaware Glass and glassware Women’s shoes Costume jewellery

Tariff (%)

Consumer cost

Producer gain

Deadweight cost

Consumer cost per job

30.0 20.0 19.0 16.5 13.5 11.0 11.0 10.0 9.0

281.0 208.0 139.0 211.0 148.0 102.0 266.0 376.0 103.0

101.0 55.0 45.0 16.0 16.0 18.0 162.0 70.0 46.0

35.0 12.0 2.0 26.0 13.0 2.0 9.0 11.0 5.0

0.461 0.122 0.401 0.934 0.192 0.244 0.180 0.102 0.097

Source: Hufbauer & Elliott (1994).

An example of the welfare effects associated with protectionism in the US is provided in Table 1.2. It shows the welfare effects of tariffs and gives estimates for the cost of protecting jobs in 9 US industries. The deadweight costs and consumer cost per protected job demonstrate that protectionism involves economic inefficiencies that decrease social welfare. Similar work has been done for China and Europe (Shuguang, Zhang, Chang & Wan, 1999; Messerlin, 2001). Although protectionism has a price tag, standard measurement techniques reveal these costs to be relatively low. Brakman & van Marrewijk (1996) discuss how welfare costs of US protectionism range between only 0.01 and 0.1 percent of the US’ GDP. Emerson, Gros, Italianers, Pisani-Ferry & Reichenbach (1992) suggest that protectionism in the EU amounts to a mere 0.2 to 0.3 percent of the EU’s GNP. These seemingly low costs of protectionism are deceptive. This is because the methodology involved in measuring the welfare effects of trade reform typically only estimates static effects of trade liberalisation under perfect competition. Markheim (1994) found that ex ante studies on the welfare effects of trade liberalisation underestimate the increase in trade flows by 49.7 to 99.8 percent, implying an understatement of the welfare gains. Accounting for dynamic effects under imperfect competition—allowing for economies of scale and scope, the introduction of new product varieties, quality improvements and knowledge spillovers—may give rise to substantially higher welfare costs of protectionism (Brakman & van Marrewijk, 1996). Frey & Schneider (1978) also discuss how the apparently low cost of protectionism encourages protectionist behaviour by politicians who are under the impression that large benefits (votes) can be secured at a small cost (a fraction of GDP). The reality is that protectionism is much more costly in a world with im-

Introduction

9

perfect competition and potential dynamic gains that are deterred by protectionist policies. For now, it can be concluded that there are difficulties involved in estimating the precise cost of protectionism. However, it is also clear that it is trade liberalisation, not protectionism, that yields (long-run) welfare gains.

1.3

Regionalism (T)he provisions of this Agreement shall not prevent, as between the territories of contracting parties, the formation of a customs union or of a free-trade area or the adoption of an interim agreement necessary for the formation of a customs union or of a free-trade area. Article XXIV:5 GATT

Despite ongoing efforts to promote freer trade among an increasing number of countries in the multilateral, non-discriminatory setting of the WTO, trade policy is notoriously controversial. This is due, at least in part, to a phenomenon known as “regionalism”. The term is based on the notion that countries engage in discriminatory trade agreements to foster economic integration within a geographically confined space. Figure 1.3 reflects the regional nature of EIAs until the 1970s. Over the years, however, an increasing number of EIAs between countries from different regions has emerged, thereby stretching the term “regionalism” to include such interregional agreements. Figure 1.4 shows the “spaghetti bowl” of countries that were connected with each other through various EIAs in 2011.6 Adding to regionalism’s complexity is that no two trade agreements are alike. Clearly, regionalism’s discriminatory focus is in stark contrast to the WTO’s cardinal rule of non-discrimination. Members benefit from preferential market access to other members’ markets, while non-members run the risk of losing their foreign market share. However, Article XXIV (GATT) allows members to form EIAs, provided that they do not raise barriers vis-`a-vis other members with respect to trade in goods. Article V of the General Agreement on Trade in Services (GATS) extends this provision to cover trade in services. Moreover, the Enabling Clause allows developed 6 For

the complete movie, see Kohl (2011).

10

Figure 1.3. Regionalism in 1970

Figure 1.4. Regionalism in 2011

Chapter 1

Introduction

11

countries to discriminate between developed, developing and least developed countries (LDCs) in order to extend preferential treatment to poorer trade partners, without obliging beneficiaries to reciprocate.7 Several analysts, including Bhagwati (1993), criticise the legal ambiguities in these provisions. For example, although the provisions state that “duties and other regulations of commerce” are not to be “higher or more restrictive” than those previously used, nothing is said about the methodology that must be used to determine these instruments’ restrictiveness. In addition, the only condition is that trade policy after the EIA should not be more restrictive towards non-members than before, but nothing is specified about how potential losses undergone by non-members are to be minimised or compensated by the EIA’s members. Ambiguous rules have the advantage of flexibility, but make it difficult to be specific about the conditions that must be met to ensure trade-creating EIAs that complement the multilateral system. Why would politicians create an institution that permits such an important departure from its fundamental principle of non-discrimination? Two possible answers are proposed. First, consider the formative years of the GATT. The two principle negotiators were the US and UK. The US were reluctant to include provisions in the draft text that would allow any deviation from the multilateral system. The UK, however, were set upon keeping its Imperial Preferences intact. As John M. Keynes, at the time the UK Treasury’s representative, declared:

[I feel] so passionately that our hands must be free to make something new and better of the postwar world . . . (w)e must be free to work out new and better arrangements which will win in substance.8

The UK’s participation in the multilateral system proved to be an important bargaining chip to secure sufficient support for Article XXIV.9 Although this left the US somewhat disgruntled, others also favoured creating the Article, including the French Union and the relatively novel Benelux, a trade agreement consisting of Belgium, Luxembourg and the Netherlands. 7 Trade

agreements of this type do occur, they are to be viewed primarily in the context of economic development, rather than economic integration. 8 Source: US Department of State (1941, p. 16-17). 9 See Irwin, Mavroidis & Sykes (2008) for an excellent account of the political considerations that were involved.

12

Chapter 1 Second, it may be more (politically) feasible for nations to sign away far-reaching

trade concessions to a handful of like-minded partners instead of settling for less ambitious liberalisation at the multilateral level. It may be more desirable to have a smaller number of countries granting free market access within their group in an attempt to obtain free(r) trade, rather than facing higher levels of protectionism in a deadlocked multilateral system.

1.3.1

The rise of regionalism

Regionalism is no trivial matter. Although there was just a handful of regionallyoriented EIAs in the 1950-60s, their number has rapidly grown since then. Figure 1.5 shows that the number of EIAs in the world economy has sharply risen from two in the 1950s to more than 300 by 2011, with the number of interregional agreements also gradually increasing (see Bhagwati, 1991; Crawford & Fiorentino, 2005; World Bank, 2005). Fluctuations in the number of EIAs after 2000 can be mostly attributed to Eastern European countries’ pre-accession agreements with the European Community (EC) and among each other. These agreements are counted as individual agreements until formal accession in either 2004 or 2007 (see also Pomfret, 2007). The number of countries with at least one EIA has also grown substantially, resulting in hubs (countries with overlapping agreements) and spokes (countries of which the EIAs are not overlapping in terms of membership). Multi-membership may be costly because of overlapping regulations and ineffective due to a lack of unambiguous political commitment to one clear set of trade policy objectives. Alternatively, being active in several EIAs reduces a country’s probability of suffering from trade discrimination and could generate positive (non)-economic spill-over effects (X. Cheng, Wang & Lui, 2009; Baldwin, 2008b; van Bergeijk, 2009). Figure 1.6 shows that most countries were involved in only one agreement until the early 1990s but have since then moved to enforcing multiple agreements, suggesting that the benefits of multi-membership outweigh the cost.

Introduction Figure 1.5. The proliferation of EIAs, 1948-2011

Note: Only enforced EIAs are considered. Source: Author’s calculations.

Figure 1.6. Number of countries with EIAs, 1948-2011

Note: Only enforced EIAs are considered. Source: Author’s calculations.

13

14

Chapter 1

Table 1.3. Number of enforced EIAs per country, 2011 Number 0

1

2

3-5

6-10

11-15 16-20 21-25 25+

Country American Samoa, Anguilla, Bermuda, British Indian Ocean Territory, Cook Islands, French Southern Territories, Gibraltar, Guam, Kosovo, Mayotte, Mongolia, Niue, North Korea, Palau, Pitcairn, Sao Tom´e & Principe, Virgin Islands, Wallis & Futuna. Andorra, Aruba, Cape Verde, Cayman Islands, Equatorial Guinea, Ethiopia, Falkland Islands, French Polynesia, Gabon, Greenland, Iraq, Macao, Marshall Islands, Mauritania, Mozambique, Netherlands Antilles, New Caledonia, Republic of Congo, Seychelles, Somalia, St. Helena, St. Pierre-Miquelon, Yemen. Afghanistan, Angola, Bhutan, Burundi, Cameroon, Central African Republic, Chad, Comoros, Democratic Republic of Congo, Djibouti, Eritrea, Gambia, Ghana, Guinea, Hong Kong, Iran, Kiribati, Kuwait, Liberia, Madagascar, Malawi, Maldives, Micronesia, Nauru, Nepal, Nigeria, Qatar, Rwanda, Samoa, San Marino, Saudi Arabia, Sierra Leone, Tanzania, Tonga, Tuvalu, Uganda, United Arab Emirates, Zambia, Zimbabwe. Algeria, Azerbaijan, Bahrain, Bangladesh, Belarus, Benin, Botswana, Burkina Faso, Cuba, Dominican Republic, Ecuador, Faeroe Islands, Fiji, Guinea-Bissau, Ivory Coast, Kenya, Lebanon, Lesotho, Libya, Mali, Mauritius, Namibia, Nicaragua, Niger, Oman, Palestinian Authority, Papua New Guinea, Senegal, Solomon Islands, South Africa, Sri Lanka, Sudan, Swaziland, Syria, Taiwan, Tajikistan, Togo, Turkmenistan, Uzbekistan, Vanuatu. Albania, Antigua & Barbuda, Argentina, Australia, Bahamas, Barbados, Belize, Bolivia, Bosnia & Herzegovina, Brazil, Brunei, Cambodia, Canada, China, Costa Rica, Dominica, Egypt, El Salvador, Georgia, Grenada, Guatemala, Guyana, Haiti, Honduras, Indonesia, Israel, Jamaica, Jordan, Kazakhstan, Kyrgyz Republic, Laos, Liechtenstein, Macedonia, Malaysia, Moldova, Montserrat, Morocco, Myanmar, New Zealand, Pakistan, Panama, Paraguay, Philippines, Russia, Serbia & Montenegro, Saint Kitts & Nevis, Saint Lucia, Saint Vincent & Grenadines, South Korea, Suriname, Trinidad & Tobago, Tunisia, Uruguay, Vietnam. Armenia, Colombia, Croatia, Japan, Peru, Thailand, Ukraine, United States, Venezuela. India, Mexico, Singapore, Turkey. Chile, Iceland, Norway, Switzerland. EU-27.

Notes: BRIC and OECD countries are printed in bold. Source: Author’s calculations.

To illustrate how countries have become involved in multiple EIAs, consider Table 1.3. It lists the number of EIAs in which the countries of the world were involved in 2011. Notice that the nations without any EIAs tended to be poor island states or, unsurprisingly, North Korea. Countries with two to five EIAs were mostly from Africa and the Middle East, while those from Asia, the Caribbean and Latin America often had six to ten agreements. Several OECD countries and major emerging economies—Brazil, Russia, India and China, also known as the BRICs—had up to 20 EIAs in 2011, while European countries were involved in more than 20 agreements.

Introduction

15

Figure 1.7. Number of EIAs by level of development, 1948-2011

Source: Author’s calculations.

According to Figure 1.7, the majority of countries joining the scene of regionalism in the 1990s were developing nations and least developed countries (LDCs). Their participation in EIAs has set the stage for an explosive number of trade agreements involving developed and/or only developing countries. Interestingly, the sharp decrease in agreements between developed countries suggests that these agreements have been expanded to include poorer nations. However, plurilateral agreements with members from all levels of development are scarce, as are agreements between developed countries and LDCs. Then again, trade agreements between developing countries and LDCs have gained in popularity. Why have EIAs become so popular? Is the WTO not enough? Why do even WTO members opt for some degree of regionalism in their otherwise multilateral trade policies? Krugman (1993) provides three reasons that can be generalised as “dissatisfaction with the multilateral framework”. First, the cost of non-compliance is higher when there are fewer members. EIAs give a better sense of predictability about members’ intended compliance than the multilateral trade system. Put differently, free riding is less attractive when fewer participants are involved.

16

Chapter 1 Second, anti-regionalists argue that it is politically much more difficult to reach

an agreement in multilateral negotiations—also known as “slow multilateralism”— because the trade barriers and interests involved are too numerous and too complex. This is witnessed by, for example, the lack of progress in the Doha Round. However, Baldwin (1997) weakens this argument by arguing that most EIAs only cover basic tariff and quota structures, which are much less complicated than the complex issues that have been successfully discussed in the history of the multilateral trade system. The third reason originates from the theory of hegemonic stability, which argues that the multilateral system has become more difficult to maintain because of the US’ reduced economic dominance in the system. Some authors even propose that regional bloc formation has mainly resulted from the US’ alleged change from a multilateral to a regional trade policy in the 1990s (Bhagwati, 1993). This view is plausibly and factually contested by Baldwin (1997), who shows that regionalism in Latin America occurred due to the US’ unwillingness to pursue preferential trade relations with its southern neighbors in that era. Sager (1997) provides two additional arguments to explain regionalism. First, many EIAs are the result of geographic proximity. A shorter distance between countries implies lower transport and communication costs, which increases the likelihood of economic trade agreements being formed (see also Baier & Bergstrand, 2004). Second, regionalism can be a response to geopolitical circumstances, where political leaders attempt to enhance their countries’ cooperation by using an EIA to signal political commitment with regard to national security (e.g., the US-Israeli Free Trade Agreement) or economic development (e.g., Economic Community of West African States) (see also Barton et al., 2006; Afesorgbor & van Bergeijk, 2011). Baldwin (1993, 2006) and Baldwin & Jaimovich (2010) take a long-term perspective on trade liberalisation and trade bloc formation. Starting with domestic demand for protectionism being equal to domestic supply for protectionism, an idiosyncratic shock forces a country to reconsider this equilibrium. If the pro-membership forces outweigh the anti-membership forces, it becomes optimal to join an EIA or move towards even deeper integration. The new equilibrium has knock-on effects on non-members, as non-member exporters can now expect a higher benefit from membership, generating pressure for non-members to obtain a new equilibrium. If they cannot join an existing trade bloc, they may be able to join another or form one with other excluded countries (see also Hufbauer, 1989; Freund, 2000b).

Introduction

1.3.2

17

Welfare implications

Does regionalism lead to suboptimal welfare levels? In general, the argument is that these agreements may create lock-in effects that prevent their members from extending concessions to non-bloc members in the multilateral system. As a result, global free trade becomes unattainable and prevents an optimal level of world welfare from being obtained. For decades, trade economists have worked on ways to determine the economic consequences of trade reforms. Is regionalism beneficial to world welfare? Viner (1950) was among the first to analyse the possible welfare impact of trade bloc formation by considering the case of European integration. He demonstrated that countries granting one another preferential market access does not strictly lead to a welfare improvement. He arrived at this conclusion by introducing the concepts of trade creation and diversion. When countries trade with each other, such that the imported good is produced more efficiently abroad than domestically, social welfare is improved (i.e. more can be imported for the same amount of money). This is called trade creation. In contrast, trade diversion occurs when countries form a trade area and obtain their imports from a member country rather than a cheaper non-member, thereby decreasing social welfare. It follows that the net welfare effect of EIAs depends on the extent of trade creation and trade diversion, which in turn is determined by changes in the terms of trade and demand and supply elasticities. This outcome, called Viner’s ambiguity, shows that regionalism may be welfare improving, but not necessarily. Meade (1955) introduced what later became known as the Kemp-Wan (1976) theorem. If countries form a trade bloc that eliminates all internal tariffs while changing external tariffs to keep the level of imports from non-bloc members constant, this leads to a Pareto improvement for all countries involved. Other demonstrations of the Kemp-Wan theorem also show that regionalism leads to Pareto improvements of world welfare (see Dixit & Norman, 1980; Krishna & Panagariya, 2000). The critical assumption for the Kemp-Wan outcome to hold is that the level of trade with non-bloc members is fixed after the trade bloc has been formed. On the other hand, the WTO regulations stipulate that the average level of tariffs levied on goods from non-bloc members should not be higher after the EIA has taken effect. Feenstra (2004, p. 193) argues that the proper criterion to prevent regionalism from leading to welfare losses is that of the Kemp-Wan theorem (i.e. trade remains unaffected) and not that of the WTO (i.e. tariffs are not changed). So, although

18

Chapter 1

the authors specify the exact conditions under which the presumption of the WTO would hold, the WTO rules do not fulfil these conditions in reality. One might therefore conclude that under the WTO, EIAs do not increase welfare. However, Lipsey (1957) uses a three-country, two-commodity model to demonstrate that trade diversion is not necessarily welfare decreasing. A trade diverting customs union may increase social welfare when trade shifts from the lowest cost non-member supplier to a higher cost supplier belonging to the union. This outcome depends on the terms of trade among the custom union’s members. Yu (1982) extends this analysis and shows that the effect of trade creation and diversion on welfare resulting from customs union formation depends on how employment responds to changes in the tariff rate, the terms of trade and the factor intensity of the importable good. Just like Lipsey, Yu concludes that under special circumstances, trade creating (diverting) customs union formation could theoretically be welfare reducing (increasing). These are important qualifications to Viner and Meade’s work that do not make it easier to determine the welfare impact of regionalism. For the time being, it can be concluded that welfare impacts of regionalism depend on more than the extent of trade creation and diversion alone. To add insult to injury, Nobel-prize winner Paul Krugman constructed a model to determine whether world welfare depends on the number of trade blocs. In one contribution, Krugman (1991a) shows that countries forming a trade bloc use market power to turn the terms of trade in their favor. Other blocs retaliate and world welfare suffers as a result. In the extreme case, countries either form onecountry trade blocs (i.e. autarky with low welfare levels) or belong to a worldwide trade bloc (i.e. global free trade with a high welfare level). Even more discouraging is Krugman’s prediction that world welfare will reach its lowest level with three trade blocs. This is somewhat troubling because the global economy is typically characterised as a triad of the US, European Union and Japan (see Ohmae, 1985). This outcome is based on the assumptions of a singlegood national economy, symmetric nations and trade blocs, and that changes in the terms of trade resulting from a change in domestic tariffs are offset by changes in the foreign tariffs. Although it is based on a simplified model, the main result is that theoretically, world welfare is least favoured by a small number of trade blocs. In subsequent work, however, Krugman (1991b) extends his earlier model by introducing transport costs. The predicted outcome then changes considerably. If inter-bloc transportation costs are sufficiently high, regionalism is found to raise world welfare. In effect, high transportation costs imply that trade blocs become

Introduction

19

relatively isolated, independent regions in which free trade is attainable. Such blocs arise when neighbouring countries (with low transportation costs) trade and form a trade bloc. Krugman shows that these “natural” trade blocs go a long way in promoting free trade within their blocs and raising global welfare (see also Wonnacott & Lutz, 1989; Summers, 1991). Summarising, the theoretical ambiguities render the Vinerian concepts of trade creation and diversion problematic when assessing the welfare effects of trade bloc formation. However, approaches such as Krugman (1991b) and Frankel (1997) account for the distance between trade partners and between trade blocs, revealing that regionalism has potentially positive welfare effects. Still, others (employing, e.g., a Kemp-Wan approach) remain concerned that regionalism undermines the multilateral trade system. The toolkit used to analyse welfare impacts of regionalism remains a complicated and ambiguous one.

1.3.3

Building vs. stumbling blocks

Apart from what Bhagwati & Panagariya (1996) call the short-run, static effects of regionalism, considerable attention has also been given to its long-run dynamic effect. Bhagwati, an influential proponent of the multilateral system, introduced a new strand of literature when he asked whether trade blocs should be perceived as “building blocs” or “stumbling blocs” on the road to global free trade in Bhagwati (1991, 2008). In a critical survey of the literature, Baldwin (2008a) identifies three major stumbling bloc mechanisms, namely (1) the preference-erosion/exploitation stumbling bloc, (2) the goodies-bag stumbling bloc and (3) the cherry-picking stumbling bloc. Ad 1. Suppose that all countries start off with MFN preferences. A select group of countries with sufficient buyer power could form a trade bloc, exploiting their market power by raising its external tariffs vis-`a-vis non-members. This ensures welfare gains for the trade bloc, while outsiders lose. This kind of regionalism is a stumbling bloc to world free trade because the bloc would have to sacrifice their welfare gains if they had to abolish their trade bloc preferences and return to MFN preferences (see Kennan & Riezman, 1990; Krishna, 1998; Freund, 2000a,b). Ad 2. Extending preferences to a member of the trade bloc means that the beneficiary’s producers have better market access than their foreign (non trade-bloc) rivals. These rents can be thought of as a “bag of goodies” that the trade bloc partners use to buy non-economic benefits from one another. The larger the margin of preferences granted, the richer the “bag of goodies”. In this way, large countries ex-

20

Chapter 1

tend preferences to small (and often poor) countries, which offer their cooperation in terms of, e.g., anti-drug and anti-terror policies in return for improved market access. Such trade blocs can be stumbling blocs as long as the margins offered by the trade bloc exceed those attainable in a MFN setting (see Lim˜ao, 2007). Ad 3. Levy (1997) illustrates a setting in which countries have much to gain and little to lose from bloc formation, while at the same time having much to lose and little to gain if they were to switch from a trade bloc to global free trade. This is due to relatively large Stolper-Samuelson losses (i.e. scarce factors lose more than abundant factors gain) and a smaller Krugman variety effect (i.e. consumers gain due to better access to more varieties of goods). In sum, the main idea behind stumbling bloc logics is that regionalism accrues benefits to its participants while being (potentially) disadvantageous to non-members. Countries benefiting from the formation of a trade bloc will be reluctant to sacrifice these gains as long as they exceed the gains from multilateral liberalisation. However, regionalism may very well serve as a stimulus for further multilateral liberalisation. Baldwin (2006) calls this the juggernaut building bloc logic (see also Grossman & Helpman, 1994; Imai, Katayama & Krishna, 2009; Baldwin & Jaimovich, 2010). Suppose that there is a national political market for protectionism (e.g., tariffs). Import-competing firms demand protection, which is supplied by their national welfare-maximising government. Once reciprocal negotiations to multilaterally liberalise trade begin, importers and exporters will behave differently. Importers continue lobbying for more protectionism. In contrast, exporters become anti-protectionists due to the fact that they benefit from increased market access if foreign tariffs, and therefore domestic tariffs, are low. With lower tariffs in place, pro-liberalisation exporting industries grow and anti-liberalisation industries decline, thereby reducing demand for protectionism. One could argue that this juggernaut logic critically depends on the reciprocal nature of multilateral trade liberalisation and not on regionalism per se. Although true, regionalism only gives pro-liberalisation exporters improved access to a limited (regional) market. Since worldwide market access is the ultimate objective, these exporters will continue lobbying for multilateral liberalisation. In that sense, regionalism functions as a building bloc. The literature review in this section illustrates that the remarkable proliferation of regionalism has attracted much debate about its causes, its welfare effects and possible implications for the multilateral trade system. As discussed, economic

Introduction

21

theory gives arguments both in favour of and against regionalism. In light of its rapid growth over the past decades, the question is particularly whether regionalism does a better job at promoting free trade than the WTO? From a theoretical perspective, the jury is still out. However, empirical evidence may shed more light on the economic impact of the WTO and regionalism. How this discussion proceeds is laid out in the following section.

1.4

Through the looking glass The proof of the pudding is in the eating. Medieval proverb

The discussion has so far illustrated the WTO’s impressive efforts to dramatically liberalise international trade among its members in a non-discriminatory fashion. However, the equally remarkable proliferation of EIAs calls into question whether the multilateral trade system has indeed had the anticipated tradepromoting effects. Therefore, the aim of this thesis is to assess the impact that these international trade institutions have had on world trade since the creation of the new world order. Using the gravity equation, which is the “empirical workhorse” in international trade studies, the impact of WTO and EIA membership on international trade flows can be quantified. Our findings shed more light on the role that international trade institutions have in shaping the global economy. Does the WTO promote trade? Surprisingly, the widely-held expectation that trade is fostered by WTO membership was challenged by Rose (2004). In seminal empirical work on this topic, Rose found that being a WTO member is “mysteriously” not conducive to international trade. This has sparked a new literature in which trade economists, political scientists and policy makers have addressed this apparent ineffectiveness of the multilateral trade system, which is also the focus of chapter 2. An extensive review of the “WTO effect” literature will reveal two major methodological flaws that bias the empirical results that have been obtained so far. Specifically, Rose’s approach is revisited and shown to have serious limitations. Once these drawbacks are overcome, the chapter shows that nations gain from their WTO membership. An in-depth analysis of WTO members’ degree of participation reveals that the largest increase in trade flows can be ascribed to countries that have

22

Chapter 1

implemented the most extensive trade liberalisation schemes, while countries with weak commitments benefit the least. Another important empirical finding from chapter 2 is that countries engaged in EIAs also experience trade-promoting effects, regardless of possible gains arising from WTO membership. The evidence on the proliferation of regionalism in chapter 1 suggests that the configuration of trade blocs may be subject to change. Chapter 3 therefore investigates the interplay between regionalism and trade blocs in the postwar period. Traditionally, trade blocs have been identified based on their geopolitical characteristics. The dominant example in the literature is that of the “triad,” the notion that the world consists of three trade blocs based on North America, the European Union and Japan. The development of international trade would then be explored by studying changes in trade flows between these three, exogenously defined trade blocs. Instead, chapter 3 introduces a new trade bloc variable that is based on the intramax hierarchical clustering technique, which defines dynamic trade blocs based on actual trade intensities. So, instead of taking the traditional approach of exogenously defining the trade blocs and then studying the changes in their respective trade flows, we first identify the trade blocs endogenously, based on their current trade flows. Our approach yields a more intuitive view of how trade blocs have developed in the past 60 years. Moreover, multivariate analysis will be applied with explanatory variables that are based on the gravity equation of international trade. This is done in order to determine which mechanisms of globalisation are key in explaining the configuration of trade blocs in the world economy. The chapter shows that, in addition to geographic proximity, regionalism is also closely associated with the probability that nations are clustered in trade blocs. Despite a large body of empirical evidence suggesting that regionalism fosters trade, the issue of causality has remained relatively unexplored, leaving the door open for erroneous trade-policy recommendations. Do nations trade more because their governments implement trade agreements, or do policy-makers simply affirm existing economic realities? This is the central issue that is considered in chapter 4. In addition, studies on the effects of regionalism vary in their methodological approaches and sample coverage, which renders comparison of individual agreements’ effects impossible. It also remains unclear whether countries trade more in anticipation of announced liberalisation or, instead, EIAs only foster trade once they have been allowed sufficient time to be phased in. Chapter 4 considers these

Introduction

23

issues by applying first-differencing techniques to a large panel dataset. Correcting for endogeneity bias confirms that EIAs have trade-promoting effects, but these turn out to be considerably smaller than expected. Moreover, the individual effects of 89 EIAs are determined. As will be shown, there is considerable variation in the impact that EIAs have on their members’ international trade flows. Can these results be explained? New insights from the empirical literature suggest that the discussion should perhaps no longer be focused on preferential tariffs alone. Work by Carpenter & Lendle (2010) and Keck & Lendle (2011) suggests that, compared to MFN tariffs, the preferential margins on those goods that are also included in EIAs are relatively small. The authors estimate that 50 percent of world trade occurs between countries that have preferential agreements in place, but that less than 20 percent of it is actually eligible for preferential treatment. Even more strikingly, the global trade-weighted preference margin is estimated to be a mere 1 percent, while only 2 percent of world trade is eligible for margins exceeding 10 percentage points. These findings coincide with work by Damuri (2009), who demonstrates that commodities with high MFN rates are typically also excluded from EIAs. Until now, conventional studies of EIAs and international trade have assumed all EIAs to be equal. This is witnessed by the use of binary variables to account for the presence of these trade agreements. Although convenient, this instrument fails to account for potential differences between EIAs that have been identified in chapter 4. Understanding what drives these differences may be crucial to understanding why some EIAs are more effective in promoting trade than others. Therefore, chapter 5 opens the proverbial black box and takes stock of the provisions covered by almost all EIAs in the world economy since World War II. Our approach provides a new means to quantify the heterogeneity of trade agreements and to study their effects on international trade. The possibility of using EIAs’ content to construct an index of trade regulation is explored, as well as an investigation of the determinants of comprehensive trade agreements. The chapter also uncovers an important finding in the debate on whether EIAs threaten to undermine or may actually complement the WTO. It reveals that virtually all EIAs build, to a large extent, on existing WTO policies. A minority of agreements contains provisions that are not currently part of the WTO’s mandate. As such, an analysis of what is written in EIAs shows that these agreements are—being firmly rooted in WTO policy—complements, not threats, to the multilateral trade system. However, opening this black box also reveals that not all provisions are good for trade.

24

Chapter 1

In fact, some provisions are found to actually decrease trade, which stresses the importance of addressing both the purpose and context of individual provisions and agreements in EIA-related research. Finally, chapter 6 reviews the main research findings and concludes with recommendations for future research.

Chapter 2

The WTO’s effect on trade: What you give is what you get 2.1

Introduction

The previous chapter introduced the World Trade Organisation (WTO) as a forum for intergovernmental negotiations on trade regulation and non-discriminatory liberalisation. The multilateral trade system’s ever-expanding membership base, combined with the steady growth of world trade has led to the conventional belief that WTO participation plays an important role in advancing international trade. Curiously, an initial contribution by Rose (2004) does not support the notion that WTO membership stimulates world trade.1 This unexpected finding is central to the inquiry presented in this chapter: Does participation in the WTO promote world trade? An extensive survey of the literature in section 2.2 will be used to argue that Rose’s results are biased and that there is a need to improve (1) the quality of the data and (2) the empirical strategy. Ad 1. Several data improvements have already been put forward in various contributions in the literature. However, a strikingly large number of economic integration agreements (EIAs) is excluded from analysis. For example, Rose (2004) only uses ten EIAs out of a possible 200+. It will be shown that this omission introduces selection bias in the EIA parameter estimates and leads to incorrect inferences about the effect of policy on international trade.

1 The change in international trade volumes induced by WTO participation is what will be called the “WTO effect” throughout this book.

26

Chapter 2 Ad 2. Typical datasets on international trade include a large number of zeros.

This poses a methodological problem for studies such as Rose (2004) which estimate a log-linear specification of the gravity equation of international trade with ordinary least squares (OLS) regression. In doing so, a substantial number of observations is ignored. More recent contributions to the WTO effect literature have adopted Tobit or Poisson maximum likelihood estimation (MLE) to include these zero trade “flows”. However, these methods are vulnerable to estimation bias and it will be argued that negative binomial (NB) and zero-inflated negative binomial (ZINB) MLE are more suitable alternatives. Details on how these two shortcomings are solved are addressed in section 2.3, along with a presentation of the updated dataset. Section 2.4 revisits Rose’s findings and shows that his remarkable results are not robust. Next, a full-fledged examination of the WTO’s effect on trade is performed with the empirical tools called for in this study. The findings suggest that developed and developing nations have gained from WTO membership, while least developed countries (LDCs) have benefited from preferential market access under the Generalised System of Preferences (GSP).2 Regionalism is also found to play a significant role in fostering international trade, although its effect is less extreme than suggested by earlier studies. Finally, section 2.5 discusses the main findings in greater detail and concludes the chapter.

2.2

Literature

Studies on the effect of WTO membership on world trade have only emerged after almost 60 years of multilateralism. Table 2.1 provides an overview of the main publications and their findings. Rose (2004) performed the first comprehensive econometric study of the WTO effect, using a standard gravity model to specifically account for the effect of formal WTO membership and participation in GSP schemes on average trade. The outcome of Rose’s study is a negative, not significant relationship between trade and formal WTO membership. Although the inclusion of country fixed effects yields a small, positive WTO effect, the paper does not provide robust statistical evidence that WTO membership has played a significant role in creating trade (column 1). 2 Following the WTO’s practice, this chapter considers LDCs and developing countries to be two distinct groups of countries.

The WTO’s effect on trade

27

Table 2.1. Results in the WTO effect literature Variable

(1) Rose (2004)

Both in WTO

−0.04 (0.05) −0.06 (0.05)

One in WTO - industrial

(2) Subramanian & Wei (2007)

(3) Tomz et al. (2007)

(4) Felbermayr & Kohler (2006)

(5) Liu (2009)

(6) Herz & Wagner (2011)

0.54 (0.06) 0.27 (0.05)

0.17/0.05 (0.06)/(0.02) 0.13/0.06 (0.09)/(0.02)

0.47 (0.04) 0.21 (0.03)

0.62 (0.02)

1.87 (0.14) −0.31 (0.04)

- developing – importer

0.27 (0.02) 0.39 (0.02)

– exporter Country granting GSP - importer 0.86 (0.03) - exporter EIA

Observations R2 Fixed effects

Method

2.15 (0.15)

0.19 (0.03)

1.20 (0.11)

1.65 (0.11)

0.77 (0.07)

0.14/0.11 (0.06)/(0.02)

234,597 0.65 Time

76,094 0.75 Importer, Exporter & Year OLS

234,597 0.85 Dyad & Year

42,542 0.45 Importer, Exporter & Year Tobit (IM/EM)

OLS

OLS

0.06 (0.02) 0.22 (0.02) 0.32 (0.01)

−0.04 (0.01) 0.01 (0.01) 0.34 (0.00)

1,119,372 N.A. Dyad & Year

697,223 N.A. Dyad & Year

Poisson

Poisson

Notes: Levels of significance are not reported due to missing data. Robust standard errors in brackets. IM means “intensive margin”, EM means “extensive margin” and N.A. means “not applicable”. Other estimates are omitted to save space.

Subramanian & Wei (2007) use a gravity equation with time, importer and exporter fixed effects. Contrary to Rose (2004), these authors also take the following three asymmetries into account: (1) asymmetries between developed and developing countries (including LDCs); (2) asymmetries between developing countries that joined the WTO before and after the Uruguay Round; and (3) asymmetries between sectors in which the WTO has (not) liberalised trade barriers. Ad 1: The WTO was designed as a two-tier organisation, with more liberalisation obligations for developed than developing country members. Non-tariff barriers (NTBs, especially quantitative restrictions) were not allowed. However, following the principle of special and differential treatment (S&D), developing nations

28

Chapter 2

had fewer obligations to liberalise. Furthermore, article XVIII:B (GATT) permitted developing nations to use NTBs and quotas for balance of payments reasons. These instruments were frequently used and resulted in diverging trade policies between developed and developing member states. It should be noted here that, compared to developed and developing nations, LDCs enjoy the most generous exemptions in terms of trade liberalisation commitments in the multilateral trade system. Ad 2: Developing countries joining the WTO after the Uruguay Round received fewer exemptions than incumbent developing countries in order to close the gap between developed and developing countries in terms of their respective obligations to liberalise trade. Hence, there is an asymmetry between the effects of the WTO on (developing country) members acceding before and after the Uruguay Round. Ad 3: This asymmetry arises from the fact that developed countries were not required to liberalise sectors such as agriculture, textiles and clothing. These countries were allowed to maintain higher tariffs and to implement quotas, amongst others, in the Multi-Fibre Arrangement (MFA). Subramanian & Wei (2007) find that the WTO strongly increases merchandise imports for industrialised countries and developing nations that joined the multilateral trade system after the Uruguay Round by e1.87 − 1 ≈ 550 percent (column 2). However, developing nations that joined the WTO before the Uruguay Round experienced small to no gains from membership due to the lack of their liberalisation commitments. The authors also argue that the WTO does not by design aim at liberalising all industries. Membership of the WTO has the smallest impact on trade in protected industries such as agriculture and textiles, as opposed to larger gains in more liberalised industries.3 They conclude that dealing with institutional asymmetries and using a better econometric specification of the gravity equation solves Rose’s “mystery” of not finding evidence that the WTO increases trade. Tomz, Goldstein & Rivers (2007) point out that formal, de jure WTO membership is not the same as informal, de facto participation. In other words, Rose (2004)’s data on WTO membership assume that multilateral rules only applied to formal (de jure) members that were listed on the WTO membership roster. However, several de facto participants—whose names were not registered on that roster—also had 3 Unfortunately, the aggregate nature of the trade data in the present dataset does not allow for industry-specific analyses. However, studies focusing on a select number of countries, years and industries have found supporting evidence for WTO membership with respect to capital-intensive commodities (Engelbrecht & Pearce, 2007) and trade excluding agriculture, textiles and oil (Kim, 2010).

The WTO’s effect on trade

29

rights and obligations under these rules and therefore need to be included in the WTO membership variable as well. These participants can be divided into three categories: colonies, newly independent states (NIS) and provisional members.4 The first category of de facto participation involves colonies. According to Article XXVI:5(a) GATT, every contracting party signed the Agreement on behalf of itself and the territories for which it had international responsibility (although exemptions were possible). Belgium, Denmark, the Netherlands, Portugal, Spain, the United Kingdom and the United States all applied GATT to virtually all their colonies upon signing the Agreement (Tomz et al., 2007, p. 2007). The second category of de facto participation has bearing on decolonisation. What would happen with a colony’s de facto participation upon becoming a NIS? The GATT provided two options: either become a formal member under the terms agreed to by the NIS’ former metropolitan government, or attempt to become a formal member by negotiating new terms. Pending their decision, the GATT participants abode by the existing Agreement as long the NIS did the same. It was only upon the creation of the WTO in 1995 that NISs were forced to choose between acceding as formal members (e.g., Burkina Faso) or withdrawing (e.g., Kiribati and Tonga) (Tomz et al., 2007, p. 2008). The final category of de facto participation involves provisional members. A number of states were treated as contracting parties even though negotiations for full accession had not yet been concluded. Provisional members who enjoyed this special arrangement before the transition to the WTO include Argentina, Colombia, Egypt, Iceland, Israel, Japan, the Philippines, Switzerland, Tunisia and Yugoslavia (Tomz et al., 2007, p. 2008). Tomz et al. (2007) create a new WTO dummy variable that includes de facto participation and also corrects for a number of incorrect de jure specifications in Rose (2004). A standard gravity equation with time and country-pair (dyad) fixed effects is then estimated using OLS. The authors find a positive WTO effect. Specifically, the increase in trade is at least as large for non-member participants as it is for formal members. This result mainly stems from the fact that most nonmember participants have historically had strong colonial ties with other WTO members/participants. Column 3 shows that if both trade partners are WTO participants, their trade is e0.54 − 1 ≈ 70 percent higher than if neither country participates, all else constant. 4 The terms membership and participation are used interchangeably throughout this chapter, but the context will be sufficiently specific when distinctions between de jure, formal membership and de facto, informal participation are necessary.

30

Chapter 2 An issue that had thus far been ignored is the incorporation of zero trade “flows”.

OLS estimates require a log-linear specification of the gravity equation, which means that the log of the dependent (trade) variable is also needed. However, as will be shown in section 2.3.2, a significant number of trade observations is zero. The loglinear specification is problematic because ln(0) = n.d., so these observations are dropped from the regression analysis and thereby introduce estimation bias. The first study in the WTO effect literature to specifically address this problem is by Felbermayr & Kohler (2006), who use a corner-solutions Tobit method. They find a WTO effect at the intensive (extensive) margin of 19 (5) percent if both trading partners are WTO members and 14 (6) percent if only one is a member (column 4).5 Liu (2009) and Herz & Wagner (2011) use Poisson quasi-MLE to determine the WTO effect. Liu (2009) disentangles the WTO effect in terms of the intensive and extensive margins of trade and finds that if both trade partners are WTO members, their trade is 60 percent (39 percent intensive, 21 percent extensive) higher than if neither is a member, all else constant. If only one of them is a WTO member, trade still is 23 percent (8 percent intensive, 15 percent extensive) higher than if both are non-members (column 5). Herz & Wagner (2011) are the first to follow Subramanian & Wei (2007) in distinguishing between importers and exporters. Moreover, they incorporate Tomz et al. (2007)’s improved WTO dummy variable and adopt Liu (2009)’s approach in using Poisson quasi-MLE to include the zeros in their analysis. They find that a country-pair’s trade volume rises by 86 percent if both countries are WTO participants. If the importer is a WTO participant and the exporter is not, trade increases by 48 percent. Non-participants import 31 percent more if their exporters are WTO participants, all else constant (column 6). Despite these contributions to the literature, it will be shown that two issues still lead to potentially biased estimates. Table 2.2 presents an overview of the methodological developments in the WTO effect literature. Almost all studies use datasets with extensive country coverage and annual data for most of the multilateral trade system’s history, with the most recent datasets being the least fragmented (rows “Year” and “Interval”).

5 Trade at the intensive margin studies the WTO effect on country-pairs that already have a trade relationship. However, WTO membership may also lead to the creation of trade relationships that countries would otherwise not have had, which is referred to as trade at the extensive margin.

The WTO’s effect on trade

31

Table 2.2. Developments in the WTO effect literature Issue

(1) Rose (2004)

(2) Subramanian & Wei (2007)

(3) Tomz et al. (2007)

(4) Felbermayr & Kohler (2006)

(5) Liu (2009)

(6) Herz & Wagner (2011)

1948-1999 Annual Arithmetic average

1950-2000 5 Years Unidirectional imports

1965-2004 5 Years Unidirectional exports

1948-2003 Annual Unidirectional imports

1953-2006 Annual Unidirectional imports

Zero trade WTO variable

Ignored Only de jure

Ignored Only de jure

Included Only de jure

Included Only de jure

EIA variable Method

Limited OLS

Limited OLS

1948-1999 Annual Arithmetic average, exports and imports Ignored De jure and de facto Limited OLS

Limited Tobit

Limited Poisson

Included De jure and de facto Limited Poisson

Year Interval Trade specification

Note: Shaded cells indicate problematic issues that may yield biased parameter estimates.

However, the earliest studies fail to deal with the drawbacks associated with how the dependent variable is defined. This gives rise to potential estimation bias due to misspecification of the dependent variable (row “Trade: Specification”). Rose (2004) defines average trade as an arithmetic average or log of sums. Nevertheless, Baldwin & Taglioni (2006) point out that the correct specification must be a geometric average because Rose’s estimated model is derived from a multiplicative specification.6 Moreover, J. Anderson & van Wincoop (2003)’s and Feenstra (2004)’s theoretical derivations of the gravity equation show that the dependent variable must be unilateral trade (imports or exports) and not aggregate trade (i.e., imports plus exports, as in Rose’s study). Compared to aggregate trade flows, the advantage of unidirectional trade flows is that the researcher can identify potential asymmetries between importers and exporters.

6 The

former is calculated as ln[ 14 (exportij + export ji + importij + import ji )], the latter as ln[(exportij × 1

export ji × importij × import ji ) 4 ].

32

Chapter 2 The first few studies also fail to incorporate zero trade “flows”, which is also

likely to yield biased results (row “Zero trade”). Moreover, Tomz et al. (2007)’s improvement of the WTO participation variable has only been incorporated by Herz & Wagner (2011). As such, all other studies suffer from biased WTO effect estimates because the independent variable of interest is misspecified (row “WTO variable”). This study incorporates the piecemeal improvements that are scattered over various studies in the literature. Moreover, it addresses the two remaining issues that have so far been ignored.7 As indicated in the lower rows of Table 2.2, the proposed improvements deal with (1) the specification of the EIA dummy variable and (2) the econometric method used to estimate the gravity equation with zero trade “flows”. The reasons why these issues warrant closer inspection and how they improve upon the existing literature are discussed in sections 2.3.1 and 2.3.2, respectively.

2.3 2.3.1

Methodology Regionalism

The WTO effect literature acknowledges that regionalism, i.e. economic integration governed by means of EIAs, may have trade-creating effects. Gravity equation specifications usually allow for EIAs to be included as a binary variable that is 1 if trade between two countries is regulated by an EIA and 0 otherwise. So far, the literature’s evidence on regionalism’s effect on trade has been mixed. It can be argue that these estimates are biased for at least two reasons. First, Rose (2004) only uses ten EIAs to construct his EIA dummy variable, which is limited to ANZCERTA, ASEAN, CACM, CARICOM, EEC/EC/EU, Mercosur, NAFTA, PATCRA, SPARTECA and US-Israel FTA. These EIAs do not qualify as an appropriate representation of all the EIAs that actually exist. For an overview of the 200+ EIAs that could and should have been included, see Table 2.3. Using such a small sample of EIAs to represent the entire population is certain to introduce bias in the EIA point estimates. This problem is persistent in the WTO effect literature, especially because a number of papers re-use Rose’s EIA variable (see Table 2.6 on p. 44). Second, the literature has almost exclusively relied on the WTO website as its only source of information about EIAs. Although WTO members have an obliga7 Chapter

4 considers potential issues with endogeneity bias and how they may be addressed.

The WTO’s effect on trade

33

tion to notify all existing and newly ratified EIAs to the WTO Secretariat, in practice it may take several months, or even years, before the Secretariat receives the required information and publishes it online. The problem is that the researcher who only looks at this source of information does not get the most up to date information on all EIAs that are or have been active on the world trade scene. A related complication is that the WTO website only has data on the most recent version of an EIA. For example, the EC and Middle Eastern countries had established several bilateral trade relationships by the 1970s. These treaties underwent some changes after the turn of the millennium and their renewed content was notified to the WTO Secretariat. Since new treaties typically replace older versions, only data regarding the latest version, including the year of ratification, are reported online. In the case of EC-Middle Eastern trade agreements, the naive researcher will wrongly infer from the WTO website that these agreements (signed after 2000) are first of their kind and that their predecessors simply did not exist. These forms of selection bias are solved by exploiting several databases that provide information on EIAs. This has the advantage that data on EIAs can be included even if they are no longer listed on the WTO’s website because they have been replaced by newer versions or have ceased to exist. There are also EIAs that have not been notified to the WTO, but which have been enforced by their respective governments nonetheless.8 In all cases, the EIAs concerned are comprehensive agreements that cover trade in (almost) all goods, and not just a limited number of sectors. In addition, every effort was made to verify each EIA’s date of ratification, the participating countries and when they joined or left the EIA. This involved several additional sources such as the text of the original treaties, government records documenting the parliamentary discussions and final ratification of the treaties, newspapers and professional magazines. By including EIAs that were ignored by Rose (2004) and others, more than 40 percent of the EIA dummy variables in the present dataset come from sources other than the WTO website. The number of country-pairs registered as having an EIA increases by more than 220 percent compared to Rose’s original dataset. The next section discusses developments in the methodology used in estimating the WTO effect and argues why a (ZI)NB model is more appropriate than OLS and Poisson (quasi-)MLE when zero trade “flows” are to be included in the analysis. 8 There are no provisions in international law that require governments to notify their EIAs to the WTO before they can be enforced.

34

Chapter 2

Table 2.3. EIAs by year of enforcement Year

EIA

0, there is evidence of overdispersion in the data and the negative binomial is preferred to Poisson. If α > 0, the next step is to determine whether the negative binomial or 10 Zero-truncated models are not considered because they only allow for positive trade flows, which does not help to address the zero trade flow problem.

40

Chapter 2

its zero-inflated variant must be used. This is determined by using a Vuong (1989) test. Negative values favour the negative binomial, while positive values favour its zero-inflated variant. As will be shown in section 2.4, the α test statistics repeatedly indicate a preference for NB to Poisson. These test criteria suggest that it is more appropriate to estimate the WTO effect using (ZI)NB MLE than OLS or standard Poisson models previously employed in the literature. Moreover, positive Vuong z-statistics are often found, implying that ZINB MLE is preferable to the non-inflated variant. If the Vuong z-statistic is negative, the NB’s results are automatically calculated and reported. The gravity model that is estimated with this (ZI)NB MLE procedure is:

Tijt

=

β

β

β

β

β

β

β

β 0 × Yit 1 × Yjt 2 × Pit 3 × Pjt 4 × Dij5 × Bothijt6 × Oneijt7

(2.5)

δ

× GSPijt × EI Aijt × zijt × Fiγi × Fj j × Ftζ t × eijt , β8

β9

β10

where P is population and Bothijt is a dummy variable that is 1 (0) if both (none of the) countries in the country-pair are WTO participants. Oneijt is a dummy that is 1 if only one country in the country-pair is a WTO participant and 0 otherwise. GSPijt is a binary variable that is 1 if the GSP applied to a dyad’s trade relationship and 0 if not. EI Aijt is a binary variable that is 1 if both countries in a dyad belong to the same EIA and 0 otherwise. zijt is a vector of additional controls that are described below. The multilateral resistance term is incorporated by using dummies that represent individual country effects (Fi and Fj ), while common shocks and trends are captured by the year dummies (Ft ). This estimation strategy is recommended to obtain reliable estimates with modified Poisson models (Allison & Waterman, 2002). We note here that the literature’s conventional strategy to code WTO, EIA and GSP participation as a binary variable (complete participation or none whatsoever) has the advantage of simplicity. One disadvantage is that it fails to acknowledge that the members’ degree of participation may change over time as a result of, for example, gradual liberalisation. Chapter 4 considers this issue in more detail. Another disadvantage is that the agreement itself may change over time. European integration, for example, started with a customs union and has gradually evolved to a monetary union. It is not impossible to acknowledge these changes in gravity equations, but it proves to be a complex task that is considered in chapter 5.

The WTO’s effect on trade

2.3.3

41

Data

The panel dataset covers 181 countries and contains observations for the period 1948 (the GATT’s founding year) to 2007. Table 2.4 lists the countries included in the dataset. The panel is arranged by country-pair and year, regardless of missing or zero values. Each country-pair is represented twice, once as ij and once as ji. This is done because bilateral imports are used as the dependent variable, which avoids Baldwin & Taglioni (2006)’s critique on the specification of the dependent variable (see p. 31). There is a maximum of 181 × 180 × 60 = 1, 954, 800 observations and approximately 98 percent of world trade is covered by the present dataset. Bilateral trade data (imports c.i.f. and exports f.o.b. in US$ millions) are from IMF (1995, 2008). The dependent variable of choice is bilateral imports. In case of missing values, the country’s trade partner’s bilateral exports are used as a proxy of that country’s bilateral imports. Following Liu (2009), a 10 percent c.i.f. rate is assumed when exports are used to replace missing imports. Baldwin & Taglioni (2006) argue that deflating trade data with a common price index may bias the regression estimates, but that time fixed effects may sufficiently address this issue. Since time fixed effects are included in the regression estimates, there is no problem with deflating trade by the US Consumer Price Index (All Consumer Goods, 1983-4 = 100) obtained from Bureau of Labor Statistics (2008). Data on GDP (in 1990 international dollars) are from Maddison (2007). Additional data are from World Bank (2011b) using the GDP in 2000 international dollars series, which was reconverted to be consistent with Maddison’s data. Data on population were also obtained from Maddison (2007). Population data for 1948-49 are from World Bank (1951). Several variables are from CEPII (2008): simple geodesic distance (in kilometres), country size (in square kilometres), whether countries share a common major/official language, a border, the number of countries in the dyad that are islands or landlocked, whether the countries in the dyad used to be one country, and details on their colonial history. Details on the EIA variable are provided in section 2.3.1. Data on countries’ WTO participation status are from Tomz et al. (2007). A number of updates were necessary, mostly for a number of countries that became formal WTO members in the period 2000-2007. Updates were obtained from the WTO website.

42

Chapter 2

Table 2.4. Countries in the dataset Afghanistan∗, Albania, Algeria, American Samoa, Angola∗, Antigua & Barbuda, Argentina, Aruba, Australia, Austria, Bahamas, Bahrain, Bangladesh∗, Barbados, Belgium, Belize, Benin∗, Bermuda, Bhutan∗, Bolivia, ˆ d’Ivoire, Cambodia∗, Cameroon, Canada, Botswana∗, Brazil, Brunei, Bulgaria, Burkina Faso∗, Burundi∗, Cote Cape Verde∗, Cayman Islands, Central African Republic∗, Chad∗, Chile, China, Colombia, Comoros∗, Costa Rica, Cuba, Cyprus, D.R. Congo∗, Denmark, Djibouti∗, Dominica, Dominican Republic, Ecuador, Egypt, El Salvador, Equatorial Guinea∗, Eritrea∗, Ethiopia∗, Faeroe Islands, Falkland Islands, Fiji, Finland, France, French Polynesia, Gabon, Gambia∗, Germany, Ghana, Gibraltar, Greece, Greenland, Grenada, Guam, Guatemala, Guinea∗, Guinea-Bissau∗, Guyana, Haiti∗, Honduras, Hong Kong, Hungary, Iceland, India, Indonesia, Iran, Iraq, Ireland, Israel, Italy, Jamaica, Japan, Jordan, Kenya, Kiribati∗, Kuwait, Laos∗, Lebanon, Lesotho∗, Liberia∗, Libya, Luxembourg, Macao, Madagascar∗, Malawi∗, Malaysia, Maldives∗, Mali∗, Malta, Marshall Islands, Mauritania∗, Mauritius, Mexico, Micronesia, Mongolia, Montserrat, Morocco, Mozambique∗, Myanmar∗, Namibia, Nauru, Nepal∗, Netherlands, Netherlands Antilles, New Caledonia, New Zealand, Nicaragua, Niger∗, Nigeria, North Korea, Norway, Oman, Pakistan, Palau, Panama, Papua New Guinea, Paraguay, Peru, Philippines, Poland, Portugal, Qatar, Republic of Congo, Romania, Rwanda∗, St. Helena, St. Kitts & Nevis, St. Lucia, St. Pierre-Miquelon, St. Vincent & Grenadines, Samoa∗, S˜ao Tom´e & Pr´ıncipe∗, Saudi Arabia, Senegal∗, Seychelles, Sierra Leone∗, Singapore, Solomon Islands∗, Somalia∗, South Africa, South Korea, Spain, Sri Lanka, Sudan∗, Suriname, Swaziland, Sweden, Switzerland, Syria, Tanzania∗, Thailand, Togo∗, Tonga, Trinidad & Tobago, Tunisia, Turkey, Tuvalu∗, Uganda∗, United Arab Emirates, United Kingdom, United States, Uruguay, Vanuatu∗, Venezuela, Vietnam, Yemen, Zambia∗, Zimbabwe. Notes: Nations marked ∗ are LDCs.

GSP was initiated in 1969 to allow developed nations to unilaterally grant preferential market access to exporters from developing countries. Rose’s GSP dummy is only based on GSP decisions published in 1974, 1979 and 1984. In addition to using data from these years, data have also been obtained data for several others (UNCTAD, 1974, 1975, 1979, 1981, 1985, 2005, 2006, 2008). The published preferences are assumed to have remained valid during years for which data could not be obtained. Thus, data for 1973 also apply to preceding years, 1974 also applies to 1975-6, 1977 also applies to 1978, 1979 also applies to 1980-3, 1984 also applies to 1985-94, 2004 also applies to 1995-2003 and 2005 also applies to 2006. 1995 is considered a “break” year in the long gap between data for 1984 and 2004 because GSP schemes are likely to have been altered during the Uruguay Round. Descriptive statistics of the dataset are provided in Table 2.5, while a summary of the data sources used in previous studies and in the present dataset is provided in Table 2.6.

The WTO’s effect on trade

43

Table 2.5. Descriptive statistics Variable Imports Both in WTO One in WTO GSP EIA ln GDP importer ln GDP exporter ln Population importer ln Population exporter ln Land area importer ln Land area exporter ln Distance Number landlocked Number of islands Common land border Common language Ever colony Common coloniser Currently colonised Common country

Observations

Mean

Std. deviation

Minimum

Maximum

1,054,520 1,954,800 1,954,800 1,954,800 1,954,800 1,494,000 1,494,000 1,483,380 1,483,380 1,911,600 1,911,600 1,869,120 1,869,120 1,869,120 1,954,800 1,869,120 1,954,800 1,954,800 1,954,800 1,954,800

119.06 0.33 0.43 0.12 0.04 9.52 9.52 1.66 1.66 10.99 10.99 8.84 0.29 0.68 0.01 0.21 0.01 0.14 0.00 0.01

1,589.26 0.47 0.50 0.32 0.19 2.19 2.19 1.89 1.89 3.04 3.04 0.75 0.50 0.67 0.11 0.41 0.10 0.34 0.03 0.09

0.00 0.00 0.00 0.00 0.00 3.53 3.53 -3.69 -3.69 1.95 1.95 2.35 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

214,440 1.00 1.00 1.00 1.00 16.04 16.04 7.18 7.18 16.12 16.12 9.90 2.00 2.00 1.00 1.00 1.00 1.00 1.00 1.00

WTO website

UNCTAD 1974, 1979, 1984

EIA variable

GSP variable

Rose (2004)

Rose (2004)

Rose (2004)

Rose (2004)

Rose (2004)

Rose (2004)

1950-2000 5 years IMF DOTS

Subramanian & Wei (2007)

Rose (2004)

Rose (2004)

Original WTO documentation Rose (2004)

Rose (2004)

Rose (2004)

1948-1999 Annual Rose (2004)

Tomz et al. (2007)

N/A

N/A

Baier & Bergstrand (2007)

N/A

IMF IFS

N/A

1965-2004 5 years IMF DOTS

Felbermayr & Kohler (2006)

UNCTAD 1973-5, 1977, 1979, 1984, 2001 CIA WFB

WTO website

WTO website

IMF IFS, Maddison (2007), PWT, WBDI

1948-2003 Annual IMF DOTS, WED, WTF US CPI

Liu (2009)

UNCTAD 1973-1986, 2001, 2005 PWT

WTO website

Tomz et al. (2007)

IMF IFS, Maddison (2007), PWT, WBDI

IMF IFS, WBDI

1953-2006 Annual IMF DOTS

Herz & Wagner (2011)

US CPI all urban consumers, Bureau of Labor Statistics (2008) Maddison (2007), World Bank (1951), World Bank (2011b) Tomz et al. (2007) & WTO website McGill (2009); Tuck (2009); WorldTradeLaw.net (2009); World Bank (2011c); WTO (2011c) UNCTAD 1973-4, 1977, 1979, 1984, 2005-6, 2008 CEPII (2008)

1948-2007 Annual IMF DOTS

This study

Notes: The longest available time-series were used. Gaps were filled using trends obtained in alternative data sources. CIA WFB is CIA World Factbook, IMF DOTS is IMF Direction of Trade Statistics, IMF IFS is International Financial Statistics, PWT is Penn World Tables, WBDI is World Bank Development Indicators, WED is World Export Dataset and WTF is World Trade Flows.

CIA WFB

WTO website

WTO variable

Control variables

IMF IFS, PWT, WBDI

Population & income

US CPI all urban consumers, freelunch.com

1948-1999 Annual IMF DOTS

Year Interval Trade

CPI

Rose (2004)

Variable

Table 2.6. Data sources in the WTO effect literature

44 Chapter 2

The WTO’s effect on trade

2.4

45

Results

First, Rose (2004)’s calculations are re-examined in Table 2.7. The results in columns 1-3 only include year effects, while those in columns 4-6 include country and year effects. The original results are repeated in columns 1 (with year effects) and 4 (with country and year effects).11 Results in column 2 and 5 use data from the new dataset, but are restricted to using the countries and years in Rose’s sample. Note that the number of observations is substantially larger. The present dataset is arranged such that one country features twice a year in each country-pair, once as importer and once as exporter, which yields a balanced panel. Rose’s data, on the other hand, constitute an unbalanced panel. This means that a country is sometimes featured as an importer, sometimes as an exporter, without any explanation as to why this is the case. This difference in how the data are organised largely explains why the number of observations differ between columns 1-2 and 4-5, respectively. The results in column 3 and 6 provide an updated version of Rose’s estimates. These results are updated in the sense that they are based on (1) the full dataset, with a larger number of countries and years, (2) imports instead of the arithmetic average of trade as dependent variable, which follows Baldwin & Taglioni (2006), (3) Tomz et al. (2007)’s unbiased WTO participated data and (4) this chapter’s EIA variable with extensive coverage instead of Rose’s biased indicator. Remarkably, following Rose’s approach does not yield his mysterious “noneffect”. According to the estimates presented in columns 2 and 5, the WTO has a positive effect on the trade flows between its members. Why do the results differ from the benchmarks? The only possible explanation is that there are minor differences in the (versions of) data sources employed in the construction of both datasets (see Table 2.6). These findings indicate that Rose’s results are not robust to replication with different data. Even though the effect of WTO membership for both trade partners turns out to be positive, WTO membership of only one trade partner is found to have a negative effect on trade in the replicated columns. Then again, positive WTO effects for country-pairs where either one or both partners are WTO members are found in the updated columns. These findings stress the relevance of correcting for the problems posed by the biased trade, WTO and EIA dummies used in earlier work.

11 Rose’s model specification is slightly different from equation 2.5. In particular, he uses the natural logarithm of the product of the importer and exporter’s real GDP. In the remainder of this chapter, these variables are included separately.

46

Chapter 2

Table 2.7. Rose (2004) revisited Variable

(1) Original

(2) Replicated

(3) Updated

(4) Original

(5) Replicated

(6) Updated

Both in WTO

-0.039 (0.05) -0.058 (0.05) 0.852*** (0.03) 1.112*** (0.11) -1.110*** (0.02) 0.916*** (0.01)

0.420*** (0.04) -0.137*** (0.02) 1.091*** (0.03) 1.294*** (0.09) -1.129*** (0.02) 1.121*** (0.01)

0.405*** (0.05) 0.104* (0.05) 0.420*** (0.03) 0.665*** (0.04) -0.895*** (0.02) 0.875*** (0.01)

0.130** (0.05) 0.034 (0.04) 0.519*** (0.03) 1.332*** (0.10) -1.306*** (0.02) 0.219*** (0.05)

0.149*** (0.04) -0.162*** (0.02) 0.749*** (0.02) 0.728*** (0.10) -1.300*** (0.02) 1.000*** (0.03)

0.231*** (0.05) -0.016 (0.05) 0.307*** (0.02) 0.624*** (0.03) -1.026*** (0.02) 0.468*** (0.04)

234,597 0.65 Year

418,008 0.67 Year

436,775 0.60 Year

234,597 0.72 Importer, Exporter & Year

418,008 0.75 Importer, Exporter & Year

436,775 0.66 Importer, Exporter & Year

One in WTO GSP EIA ln Distance ln GDP

Observations R2 Fixed effects

Notes: OLS with country and/or time fixed effects. Regressand (“Original” and “Replicated”): natural logarithm of the arithmetic average of import plus export. Regressand (“Updated”): natural logarithm of imports. Estimates marked ***/**/* are significant at the 1/5/10 percent level. Robust standard errors (clustered by country-pair) are in brackets. Estimates for intercepts, other controls and fixed effects are omitted to save space.

Now that we have shown the importance of using the appropriate measures of WTO and EIA participation, the question that remains is how the (updated and improved) data should be modelled in the presence of zero trade “flows”. Analysis of the improved trade variable reveals reveals that it is overdispersed, given that the conditional mean and variance are not equal. Moreover, goodnessof-fit measures provide further evidence in favour of (ZI)NB MLE. Indeed, the average predicted probabilities for the Poisson model fare worse than those provided by (ZI)NB models. Table 2.8 displays the penalised log-likelihood-based statistics (AIC and BIC), where the BIC penalises model complexity (i.e., the number of estimated parameters) more severely than the AIC. Both indicate a strong preference for NB over Poisson. Additional evidence favouring (ZI)NB MLE are given by the α’s and Vuong z-statistics, which are provided in the tables showing the parameter estimates obtained with (ZI)NB MLE (Tables 2.9-2.12).

The WTO’s effect on trade

47

Table 2.8. Model selection Model

Statistic

Poisson

BIC AIC

7.77e+7 112.16

Poisson versus NB

BIC AIC LRX2

-6.92e+6 4.79 8.46e+7

Difference

Prefer

Over

Evidence

8.46e+7 107.38

NB NB NB

Poisson Poisson Poisson

Very strong p = 0.00

Source: Stata output generated with the countfit package.

Using the updated dataset and improved EIA variable, the WTO effect is estimated with (ZI)NB MLE.12 The results are presented in Table 2.9. The results for the (ZI)NB analyses are subdivided into two columns, one for the “active” group (using negative binomial regression) and one for the “passive” group (using logit). A variable’s coefficient in the “passive” column shows the extent to which that variable contributes to the chance of being in the passive group. For example, the EIA variable in Table 2.9 has a negative coefficient in the “passive” column, which means that if a pair of countries participates in an EIA, this reduces (increases) the probability that the pair belongs to the passive (active) group. In Table 2.9, column 1, results are shown for the following variables of interest: both countries in a dyad participate in the WTO, only one country in a dyad participates in the WTO, GSP applies to the trade relationship, and both countries belong to the same EIA. In column 2, it is investigated whether asymmetries arise between importers and exporters in case either is the sole WTO participant in the country-pair. Differences between GSP beneficiaries and GSP benefactors are also accounted for. Finally, Tomz et al. (2007) are followed by disentangling formal WTO members from informal participants so as to investigate potential asymmetries between these two groups in column 3. Interestingly, all three columns indicate that the WTO has net trade-promoting effects for country-pairs where both the importer and exporter are members. Positive WTO effects are also found when only one of the trade partners is a member. What is most striking about these findings is that trade between members increases more compared to trade between a member and non-member. Moreover, formal members gain more than informal participants. So, the results suggest that there is a relation between the experienced gains and the extent to which nations have made multilateral commitments. 12 MLE

is performed using the DFP algorithm (Gould, Pitblado & Sribney, 2006, chapter 1).

48

Chapter 2

Table 2.9. (ZI)NB results, 1948-2007 Variable Both in WTO - formal & formal - formal & informal - informal & informal One in WTO - importer - exporter - formal & outsider - informal & outsider GSP - beneficiary - benefactor EIA ln Distance ln GDP - importer - exporter ln Population - importer - exporter ln Land area - importer - exporter Number landlocked - one - both Number of islands - one - both Common land border Common language Ever colony Common coloniser Currently colonised Common country Log pseudo-likelihood Observations Zeros α Vuong Method

(1)

(2)

(3)

Active

Passive

Active

Passive

0.643*** (0.05)

-0.40

0.581*** (0.05)

-0.40

0.342*** (0.05)

Passive

0.759*** (0.07) 0.757*** (0.06) 0.607*** (0.05)

-0.24 -0.12 -0.04

0.493*** (0.06) -0.209*** (0.05)

0.03 0.10

0.03 0.225*** (0.05) 0.370*** (0.05)

0.229*** (0.02)

Active

0.04 -0.02

-0.36

0.216*** (0.04) -1.108*** (0.02)

-0.08 -2.12

0.283*** (0.03) 0.177*** (0.04) 0.218*** (0.04) -1.109*** (0.02)

0.754*** (0.03) 1.008*** (0.03)

-3.96 -4.94

0.757*** (0.03) 1.006*** (0.03)

-3.96 -4.94

0.754*** (0.03) 1.005*** (0.03)

-3.97 -4.94

0.160*** (0.06) 0.385*** (0.05)

-1.12 -1.87

0.130*** (0.06) 0.400*** (0.06)

-1.12 -1.87

0.122*** (0.06) 0.406*** (0.06)

-1.12 -1.87

0.127 (0.07) -0.368*** (0.07)

-3.46 -4.02

0.160* (0.07) -0.385*** (0.07)

-3.46 -4.02

0.150* (0.07) -0.402*** (0.07)

-3.46 -4.02

1.077* (0.49) 2.484* (0.98)

0.03 0.01

1.158* (0.49) 2.649** (0.98)

0.03 0.01

1.223* (0.49) 2.775** (0.98)

0.03 0.01

0.569* (0.29) 1.306* (0.58) 0.634*** (0.09) 0.433*** (0.04) 1.503*** (0.10) 0.667*** (0.06) -0.384 (0.80) 0.887*** (0.11)

-0.07 -0.02 -0.03 -0.07 -0.04 0.01 0.00 -0.01

0.594* (0.29) 1.355* (0.58) 0.635*** (0.09) 0.429*** (0.04) 1.502*** (0.10) 0.668*** (0.06) -0.378 (0.80) 0.880*** (0.11)

0.00 -0.02 -0.03 -0.07 -0.04 0.01 0.00 -0.01

0.597* (0.29) 1.364* (0.58) 0.634*** (0.09) 0.437*** (0.04) 1.502*** (0.10) 0.667*** (0.06) -0.385 (0.80) 0.886*** (0.11)

-0.07 -0.02 -0.03 -0.07 -0.04 0.01 0.00 -0.01

-1,885,759 787,840 351,065 3.04*** -186.37*** NB

-1,885,763 787,840 351,065 3.04*** 184.79*** ZINB

-0.20 -0.17 -0.08 -2.12

0.268*** (0.03) 0.176*** (0.04) 0.205*** (0.04) -1.109*** (0.02)

-0.20 -0.17 -0.08 -2.12

-1,885,594 78,7840 351,065 3.04*** -85.92*** NB

Notes: (ZI)NB regression with bilateral imports as dependent variable. Estimates marked ***/**/* are significant at the 1/5/10 percent level. Robust standard errors (clustered by country-pair) are in brackets. Estimates for intercepts and fixed effects are omitted to save space.

The WTO’s effect on trade

49

Table 2.10. ZINB results by development status, 1948-2007 Variable

(1) Developed

(2) Developing

(3) LDC

Both in WTO One in WTO GSP EIA ln Distance ln GDP - importer - exporter ln Population - importer - exporter ln Land area - importer - exporter Number landlocked - one - both Number of islands - one - both Common land border Common language Ever colony Common coloniser Currently colonised Common country

0.836*** (0.10) 0.440*** (0.09) -0.340*** (0.04) 0.183** (0.06) -0.921*** (0.04)

0.646*** (0.07) 0.331*** (0.06) 0.147*** (0.04) 0.264*** (0.05) -1.281*** (0.03)

0.273 (0.15) 0.117 (0.12) 0.266*** (0.07) 0.275** (0.08) -1.116*** (0.06)

1.085*** (0.08) 1.153*** (0.05)

0.777*** (0.04) 1.005*** (0.05)

0.548*** (0.06) 0.822*** (0.11)

-0.339 (0.23) -0.087 (0.09)

0.115 (0.09) 0.529*** (0.08)

-1.273*** (0.26) 0.389* (0.18)

-0.227* (0.10) 0.066 (0.10)

0.146 (0.08) -0.174* (0.07)

2.277*** (0.43) -0.075 (0.13)

0.574 (0.47) 1.763 (0.96)

2.074** (0.08) 4.650** (1.52)

1.344 (0.87) 2.938 (1.75)

1.837*** (0.22) 3.937*** (0.44) -0.039 (0.15) 0.295*** (0.08) 1.452*** (0.12) -0.149 (0.32) -0.493 (0.48) 0.028 (0.50)

1.946*** (0.31) 3.996*** (0.63) 0.241* (0.10) 0.365*** (0.05) 0.990*** (0.10) 0.708*** (0.08) 0.161 (0.85) 0.832*** (0.13)

0.621 (0.47) 1.491 (0.96) 1.111*** (0.16) 0.515*** (0.08) 2.317*** (0.21) 0.615*** (0.10) (omitted) 0.790*** (0.21)

Log pseudo-likelihood Observations Zeros α Vuong Method

-682,458 173,956 43,485 2.05**** 1.34* ZINB

-940,570 437,107 209,364 3.22** 31.17*** ZINB

-233,698 176,777 98,216 2.93* 32.51*** ZINB

Notes: ZINB regression with bilateral imports as dependent variable. Estimates marked ***/**/* are significant at the 1/5/10 percent level. Robust standard errors (clustered by country-pair) are in brackets. Estimates for intercepts, fixed effects and the “passive” group are omitted to save space. Variables are omitted when not applicable. Development status is based on World Bank (2011a).

Robustness checks are performed to ensure the consistency of these findings. (ZI)NB MLE is first performed for countries that have been grouped by their development status. The results are presented in Table 2.10. The second robustness check groups nations by their income level, with the results displayed in Table 2.11. Finally, the effects vary by WTO negotiation round, as shown in Table 2.12.

50

Chapter 2

Table 2.11. (ZI)NB results by income classification, 1948-2007 Variable

(1) HI

(2) UMI

(3) LMI

(4) LI

Both in WTO One in WTO GSP EIA ln Distance ln GDP - importer - exporter ln Population - importer - exporter ln Land area - importer - exporter Number landlocked - one - both Number of islands - one - both Common land border Common language Ever colony Common coloniser Currently colonised Common country

0.728*** (0.09) 0.352*** (0.09) 0.015 (0.03) 0.069 (0.06) -1.013*** (0.04)

0.328*** (0.10) 0.171* (0.09) 0.147** (0.06) 0.304*** (0.07) -1.271*** (0.04)

0.803*** (0.08) 0.412*** (0.07) 0.148* (0.06) 0.087 (0.07) -1.378*** (0.04)

0.635*** (0.15) 0.409*** (0.12) 0.202** (0.07) 0.215** (0.08) -1.199*** (0.06)

0.670*** (0.04) 1.119*** (0.05)

1.268*** (0.06) 0.993*** (0.07)

0.831*** (0.05) 0.922*** (0.06)

0.729*** (0.08) 0.945*** (0.08)

0.03 (0.07) -0.322*** (0.09)

-0.491** (0.16) 0.599*** (0.11)

0.298 (0.18) 0.910*** (0.10)

-1.190*** (0.24) 0.737*** (0.11)

-0.114 (0.12) 0.335*** (0.09)

-0.744*** (0.18) -0.179 (0.11)

5.755*** (1.25) -0.460*** (0.10)

-0.516* (0.25) -0.326** (0.11)

-0.752 (0.45) -0.762 (0.92)

2.915*** (0.85) 3.262*** (1.75)

0.927 (0.96) 2.568 (1.95)

3.310*** (0.74) 6.792*** (1.47)

1.551*** (0.21) 3.217*** (0.42) -0.198 (0.18) 0.361*** (0.06) 1.341*** (0.12) -0.083 (0.11) 0.053 (0.76) 0.937 (0.57)

1.627*** (0.38) 3.615*** (0.78) 0.121 (0.12) 0.462*** (0.08) 0.821*** (0.18) 0.998*** (0.14) -3.663*** (0.55) 0.872*** (0.19)

2.183*** (0.44) 4.272*** (0.87) 0.599*** (0.15) 0.354*** (0.07) 1.745*** (0.22) 0.896*** (0.10) (omitted) 0.468* (0.19)

1.214* (0.39) 2.540** (0.81) 1.003*** (0.16) 0.398*** (0.08) 2.262*** (0.26) 0.694*** (0.09) (omitted) 0.787*** (0.21)

Log pseudo-likelihood Observations Zeros α Vuong Method

-861,716 244,680 74,111 2.37** -48.67*** NB

-353,309 167,183 79,838 3.13* 63.68*** ZINB

-407,549 205,546 104,005 3.08* -106.29*** NB

-236,838 170,431 93,111 3.12* -3.08* NB

Notes: (ZI)NB regression with bilateral imports as dependent variable. Estimates marked ***/**/* are significant at the 1/5/10 percent level. Robust standard errors (clustered by country-pair) are in brackets. Estimates for intercepts, fixed effects and the “passive” group are omitted to save space. Variables are omitted when not applicable. Income classifications are based on World Bank (2011a). HI is high income, UMI is upper middle income, LMI is lower middle income and LI is low income.

1.366*** (0.16) 2.119*** (0.20) -0.440 (0.36) -2.220*** (0.42) -0.567** (0.20) 0.524 (0.32) -22.581*** (0.14) -45.24 (·) 1.422*** (0.44) 3.204*** (0.88) 0.335* (0.13) 0.282** (0.09) 1.770*** (0.15) 1.294*** (0.18) (omitted) 0.641*** (0.23)

0.901 (0.50) 2.903*** (0.71) 1.287 (0.94) -1.961 (1.08) 4.837*** (0.57) -0.632 (1.03) -0.273 (35.94) 0.053 (71.95) 1.067 (1.04) 2.480 (2.07) 0.086 (0.13) -0.119 (0.09) 1.908*** (0.16) 1.023*** (0.19) (omitted) 0.474 (0.26) -22,564 12,790 7,976 1.11** 48.60*** ZINB

Both in WTO One in WTO GSP EIA ln Distance ln GDP - importer - exporter ln Population - importer - exporter ln Land area - importer - exporter Number landlocked - one - both Number of islands - one - both Common land border Common language Ever colony Common coloniser Currently colonised Common country

Log pseudo-likelihood Observations Zeros α Vuong Method

-178,165 108,395 70,671 1.39** 105.64*** ZINB

1.036*** (0.38) 2.214*** (0.75) 0.155 (0.11) 0.165* (0.08) 1.827*** (0.14) 1.085*** (0.16) (omitted) 0.244 (0.20)

-0.375 (1.289) -0.591 (2.59)

-0.518** (0.19) 0.440* (0.21)

0.008 (0.21) -2.205 (0.25)

1.500*** (0.12) 2.066*** (0.13)

(3) 1948-61, until Dillon Round -0.171** (0.06) -0.127** (0.05) (omitted) -0.096 (0.11) -0.812*** (0.03)

-298,206 170,437 107,599 1.41** 114.87*** ZINB

1.477*** (0.37) 3.049*** (0.73) 0.169 (0.10) 0.243*** (0.06) 1.771*** (0.13) 1.027*** (0.11) (omitted) 0.382*** (0.16)

-1.546 (1.28) -2.978 (2.55)

-0.108 (0.13) 0.501*** (0.15)

-0.310* (0.16) -1.518*** (0.16)

1.263*** (0.09) 1.614*** (0.10)

(4) 1948-67, until Kennedy Round -0.223*** (0.05) -0.143*** (0.04) (omitted) 0.409*** (0.12) -0.800*** (0.03)

-655,659 318,048 175,715 1.73** 150.61*** ZINB

1.419*** (0.31) 3.131*** (0.61) 0.181* (0.09) 0.380*** (0.05) 1.683*** (0.11) 0.749*** (0.08) -0.073 (0.84) 0.953*** (0.13)

-0.02 (0.73) -0.04 (1.49)

-0.221* (0.10) 0.105 (0.09)

-0.175 (0.11) -0.787*** (0.11)

1.193*** (0.07) 1.526*** (0.06)

(5) 1948-79, until Tokyo Round -0.055 (0.06) -0.044 (0.05) 0.244*** (0.03) 0.363*** (0.08) -0.885*** (0.03)

-1,283,244 570,699 282,803 2.70** 115.14*** ZINB

1.842*** (0.28) 3.918*** (0.56) 0.463*** (0.09) 0.457*** (0.05) 1.650*** (0.10) 0.802*** (0.07) -0.638 (0.806) 0.926*** (0.12)

2.750*** (0.66) 5.681*** (1.34)

0.08 (0.08) -0.278*** (0.07)

0.198*** (0.07) 0.511*** (0.07)

0.905*** (0.08) 1.071*** (0.04)

(6) 1948-94, until Uruguay Round 0.810*** (0.06) 0.430*** (0.05) 0.375*** (0.03) 0.182** (0.06) -1.185*** (0.02)

Notes: ZINB regression with bilateral imports as dependent variable. Estimates marked ***/**/* are significant at the 1/5/10 percent level. Robust standard errors (clustered by country-pair) are in brackets. Estimates for intercepts, fixed effects and the “passive” group are omitted to save space. Variables are omitted when not applicable.

-103,927 63,947 42,292 1.67** 74.43*** ZINB

(2) 1948-56, until Geneva Round 0.028 (0.07) -0.048 (0.055) (omitted) (omitted) -0.830*** (0.04)

(1) 1948-51, until Torquay Round -0.083(0.09) -0.038 (0.06) (omitted) (omitted) -0.851*** (0.04)

Variable

Table 2.12. ZINB results by negotiation round

The WTO’s effect on trade 51

52

Chapter 2

2.5

Discussion & conclusion

The aim of this chapter was to inspect Rose (2004)’s surprising finding that WTO membership has no effect on international trade. It has been argued that the specification of the EIA variable and failure to account for zero trade “flows” are potential sources of biased results. What happens when these issues are addressed? The first step was to replicate Rose’s findings with different data to confirm the robustness of his unexpected “non-effect” of WTO membership on trade. As already indicated, doing so yields positive WTO, GSP and EIA effects. Rose’s “noneffect” is not robust. Interestingly, the obtained EIA effect is remarkably lower than prior OLS estimates in the WTO effect literature suggest. Where Rose (2004), Subramanian & Wei (2007) and Tomz et al. (2007) find that EIAs increase trade by 230, 420 and 115 percent respectively, the present results indicate that this effect is e0.62 − 1 ≈ 90 percent. These more nuanced findings are line with other major studies in the field of EIAs (see, e.g., Baier & Bergstrand, 2007). In comparison, GSP schemes boost trade by over 40 percent and joint WTO participation by about 30 percent, all else constant. This confirms that EIA selection bias gives rise to overestimated EIA parameter estimates, but that—beside WTO membership and GSP schemes—regionalism has an influential role in advance international trade. As discussed above, merely updating the data is not enough to obtain intuitive results. Choosing an appropriate modelling strategy that incorporates zero trade “flows” appears to be of significant importance in order to obtain sensible and robust results. Having selected (ZI)NB MLE as preferred empirical strategy— compared to OLS and other Poisson models—to incorporate zero trade “flows,” the next step is to re-estimate the WTO effect with the updated dataset presented in this study. Table 2.9 shows positive and significant α’s for all estimates. Additional goodness-of-fit tests indicate overdispersion, suggesting that a Poisson model is not suitable to estimate the gravity model. If the Vuong z-statistic is negative, the first step (logit) in the ZINB MLE procedure is automatically skipped and only results from the second step (NB MLE) are reported. In these cases, NB MLE is preferable to the zero-inflated version. Overall, the goodness-of-fit statistics provided in Table 2.12 and the α and Vuong z-statistics provide strong support for the (ZI)NB MLE procedure applied in this study. The results in Table 2.9 (column 1) show that trade increases by around 90 percent if both trade partners are WTO members and by 40 percent if only one of them is a member. GSP schemes have an overall trade-increasing effect. Just as with the OLS estimates, the results also ascribe an important role to EIAs in stimulat-

The WTO’s effect on trade

53

ing trade, with EIAs boosting trade by 25 percent. Interestingly, this is less than a third of the OLS estimate, which was already considerably lower than those obtained in previous studies. Hence, the improved EIA variable and methodology reduce the bias introduced by earlier estimates of the effect that regionalism has on trade. Note that most of the other gravity equation coefficients are in line with conventional estimates in the literature. In Table 2.9 (column 2), a distinction is made between importers and exporters in the WTO and GSP variables. Non-members gain from trading with WTO members, regardless of the member’s status as importer or exporter. The intuition is that members may extend the scope of their import and export liberalisation to members of anon-members alike, thereby allowing non-members to indirectly benefit from other countries’ participation in the multilateral trade system (for a comparison, see Herz & Wagner, 2011, p. 1022). The results again confirm that GSP schemes are trade-promoting. Tomz et al. (2007)’s disaggregated WTO participation data is used to estimate how the WTO effect may differ among formal members and informal participants. The results in Table 2.9 (column 3) suggest that informal participation yields smaller gains than formal membership. This empirical finding makes intuitive sense: according to the historical narrative of the GATT/WTO discussed above, informal participants usually were least developed countries (LDCs) or developing countries. Throughout the multilateral trade system’s evolution, LDCs were exempted from most obligations on the trade agenda. Compared to LDCs, developing nations had to make more commitments, though not as much as developed countries (Barton et al., 2006; Subramanian & Wei, 2007). For the first time in the WTO effect literature, this institutional reality is clearly reflected by the present empirical results and confirmed by additional robustness checks. Upon taking WTO participants’ development status into account, developed and developing WTO participants are found to gain from trade, although the size of these gains depends on the extent of the countries’ liberalisation. This means that developed nations gain more than developing countries, while LDCs only seem to gain from GSP schemes (see Table 2.11). This outcome is only slightly different when countries are analysed by their income classifications. In this case, LDCs are the main beneficiaries of GSP schemes, but the results also suggest gains from WTO participation (see Table 2.10). Overall, the empirical results suggest that the more extensive a WTO participant’s commitments are to multilateral trade liberalisation, the more it will experience the resulting trade-creating effects.

54

Chapter 2 Allowing the results to vary across time periods reveals that the multilateral

trade system did not promote trade in its younger years (see Table 2.12). This may be explained, on the one hand, by the increasing complexity of negotiations and on the other by the substantial number of informal participants that were not required to implement liberalisation, thereby offsetting potential gains created by formal members’ liberalisation efforts (WTO, 2011d).13 However, the Kennedy and Tokyo Rounds have been especially influential in creating a positive WTO effect. Indeed, Preeg (1970) notes that the first five postwar multilateral trade conferences were not as important as the Kennedy Round, during which average tariffs on industrial products were cut by 35 to 40 percent, with two-thirds of the cuts exceeding 50 percent. Although this is not immediately borne out by the results in column 5, the positive WTO effect that participants experienced after the Tokyo Round suggests that significant long-run gains have materialised. As discussed in chapter 1, the arena of international trade policy has been largely shaped by the multilateral trade system on the one hand and by regionalism on the other. Countries engaged in the WTO have managed to gain better access to foreign markets and restrict their protective measures through consecutive rounds of trade negotiations. The empirical evidence presented in this chapter highlights three important findings. First, Rose (2004)’s initial “non-effect” of WTO membership on trade is not robust. In fact, it is difficult not to obtain favourable WTO effects. However, merely updating the data is not enough to obtain robust and intuitive results. Using an appropriate modelling strategy is equally relevant. It has been argued that the large proportion of zeros in international trade datasets creates problems for both traditional log-linear and recent approaches using Poisson QMLE alike. (ZI)NB models successfully deal with the problems of overdispersion and excessive zeros. In all cases that have been examined in this chapter, the test statistics indicate that (ZI)NB MLE is more suitable than Poisson QMLE when estimating gravity equations with large panel datasets involving zero trade “flows”. Second, WTO membership has a strong trade-promoting effect among trade partners, even if not all of them are WTO participants. However, “what you give is what you get”: formal members gain more than non-member participants, suggesting that the extent of WTO participants’ experienced trade gains goes hand in hand 13 Observing variation in countries that have been members since the GATT’s inception is not possible due to data limitations. The trade liberalisation efforts of these countries may have been even more extensive than those of countries that joined later because, being GATT founders, they showed the greatest formal initiative to liberalise trade. This may imply that the obtained estimates of the WTO effect are underestimated, although this seems difficult to verify directly.

The WTO’s effect on trade

55

with their multilateral liberalisation commitments. This result is shown to be surprisingly robust and perfectly reflects the institutional evolution of the multilateral trade system described in chapters 1, 2 and 5. Third, countries with EIAs trade more compared to those without EIAs, all else constant. Using the extensive and up-to-date dataset on trade and EIAs that corrects for selection bias, this chapter confirms that regionalism fosters international trade, although the obtained estimates are considerably lower than those previously suggested in the literature. This chapter has established the WTO’s effect on international trade since its creation. The coming chapters focus on regionalism’s role in the world economy. Specifically, chapter 3 explains how trade blocs have developed throughout the post-war period and that their configuration depends on geographical proximity and regionalism. Chapter 4 demonstrates how individual EIAs affect international trade, followed by an in-depth analysis of the provisions that are actually covered by these agreements in chapter 5.

Chapter 3

The development of trade blocs in an era of globalisation∗ 3.1

Introduction

The discussion in chapter 1 explains how two institutional forces—the World Trade Organisation (WTO) and regionalism—have set the stage for international trade policy in the post-war period. Having dealt with the implications of WTO membership on world trade in chapter 2, we now turn to the issue of regionalism. A well-documented fact in the international trade literature is that the amount of countries’ trade crucially depends on the transport costs involved in shipping the goods from exporter to importer. According to the gravity equation of international trade, the greater the geographic distance between two countries, the higher these costs and the smaller the amount of trade between them will be, all else held constant. Distance matters for bilateral trade. This implies that cross-border trade is mainly regional in nature and is reflected by regionally-oriented trade blocs. In contrast, one of the take-away messages of the “spaghetti bowl” of economic integration agreements (EIAs)—as illustrated on p. 10—is that regionalism is no longer regional, suggesting that a simple relation with respect to distance does not hold. Clearly, there is an apparent change in the direction of trade policy towards interregional cooperation. This begs the question whether the trade blocs underlying this global network of trade agreements has also evolved to become less geographically oriented. ∗ Based

on Kohl & Brouwer (2012).

58

Chapter 3 A characteristic for regionalism until roughly the 1980s is that nations used

regionally-focused trade agreements to strengthen their economic ties with their geographic neighbour. The combination of transport costs and economic policy, both favouring intraregional integration, can then be expected to enhance the formation of trade blocs that are truly regional in nature. However, an increasing number of EIAs have been enforced between countries from different regions since the late 1980s. This development has been accompanied with economic globalisation, featuring technological innovation, decreasing transport costs, the expanding presence of multinational enterprises (MNEs) and the development of global value chains. Do these events still favour intraregional integration, or have the trade blocs in the world economy become increasingly global? Geographic space can be defined by its administrative, economic, cultural, historic, socio-political and/or legal context, but this does not imply that business and economic activities strictly occur along the same lines. In describing trade blocs, economic studies typically divide space into time-invariant geographic regions such as the so-called triad of North America, the EU and Japan (see, e.g., Ohmae, 1985; O’Loughlin & Anselin, 1996). As such, these trade blocs remain static over time and do not reflect possible changes in their configuration—their geographic focus—in times of increasingly globally-oriented trade policy. Instead of relying on exogenously defined trade blocs, the focus of this chapter is how they can be identified endogenously. And it is still geographically relevant because trade among countries is imbalanced and influenced by changes in political and institutional factors, their (shared) history, colonial ties, language and culture, also known as the centripetal and centrifugal forces of regionalisation and globalisation (compare K. Anderson & Norheim, 1993; Shin, 2002; Andresen, 2009b). How can this be done? The intramax clustering technique, which is a hierarchical clustering method that is common in spatial studies and new to the international trade literature, can be used to group countries into trade blocs. It groups predefined areas (e.g., at the district or country level) based on the level of interaction between them. One of its popular applications is in the use of commuter flows to study urban regions and their catchment areas. The resulting clusters are also known as functional areas and henceforth called “trade blocs”.

The development of trade blocs in an era of globalisation

59

In this chapter, the intramax method is applied to annual bilateral trade data to identify trade blocs in the world economy. Note that although an initial amount of space is identified (i.e., national borders), intramax does not require neighbouring countries to belong to the same trade bloc. As will be explained below, the question whether countries belong to the same trade bloc according to this method solely depends on their aggregate bilateral merchandise trade flows and not on any other characteristics. The dataset used in this chapter covers the period 1950 to 2005 so that each country’s trade bloc can be identified per year and changes in trade bloc orientations be monitored over time. This is a more expansive dataset than those used in previous studies on trade clusters (compare Poon & Pandit, 1996; Poon, Thompson & Kelly, 2000). Here, the dynamic nature of trade is emphasised, which gives rise to very different trade blocs than those identified by static classifications that are based on geographic or political factors. Taking a long-term perspective, insight is obtained as to how data-driven, dynamic trade blocs develop over time. For example, Brazil, Russia, India and China (BRICs) are found to have been drivers of their respective trade clusters long before being labelled as “emerging markets” in the literature. If one studies longitudinal data, the first impression is that trade and the direction of trade is not stable. How can one investigate such fluctuations and do they affect the configuration of the intramax-based trade clusters? How stable are these clusters over time? How can possible differences and temporal changes be explained? Building on the work of Poon & Pandit (1996), this study investigates a number of variables associated with these trade clusters—geographic, economic, political, historical and cultural—and how they change over time. Specifically, gravity-equation variables from chapter 2 are applied in a multivariate analysis to investigate the probability that countries are grouped into one and the same trade cluster (compare Ravenstein, 1985; Frankel, Stein & Wei, 1995, 1996). Is geographic proximity the main reason? Or are there additional factors that contribute to the probability that countries belong to the same trade cluster? It is expected that in addition to geodesic distance, trade partners’ economic mass, cultural similarity, shared colonial history and border(s), their degree of geographic isolation and their involvement in EIAs play are associated with how trade blocs are configured.

60

Chapter 3 Interestingly, our approach complements the traditional applications of the grav-

ity equation, where distance is used as a determinant for the level of cross-border trade between two countries. Here, distance is associated with the probability that trade partners belong to the same trade cluster. The present study adds the existing gravity-equation based literature by using countries’ probability of being a trade bloc—as opposed to their level of bilateral trade—as unit of analysis. The remainder of this chapter is organised as follows. The literature is reviewed in section 3.2, which elaborates on the issues raised in this introduction. Section 3.3 focuses on how the intramax technique is used to obtain trade blocs that are entirely data-driven, such that their identification is independent of predefined geographic characteristics. It then illustrates that predefined trade blocs are poor predictors of the trade blocs identified with the intramax procedure, supporting the notion that trade blocs are dynamic in nature. The section concludes by visualising the trade clusters that have been identified for the period 1950-2005 and discussing the main developments. Next, section 3.4 presents an empirical analysis of the geographic, economic, political, historical and cultural variables that may be associated with the probability that countries form dynamic trade clusters. Finally, section 3.5 discusses the main findings and concludes.

3.2

Literature

What drives the (geographic) clustering of trading nations? The literature indicates that the regionalisation of nations is increasing and gives rise to a realistic division of economic spaces in the world economy. As discussed below, it is suggested that the (geographic) clustering of nations occurs due to forces of natural and institutional regionalisation (see, amongst others, Poon & Pandit, 1996; Poon, 1997a,b; Poon et al., 2000; MacLeod & Jones, 2007; Glenn, 2008; Andresen, 2009a,b).

3.2.1

Regionalisation of nations

Economic activity increasingly occurs within supranational spaces or regional states, resulting in functionally interconnected transnational spaces. The “space” in which trade takes place is defined by the flows of economic activity, rather than by political boundaries (Glenn, 2008). Political boundaries can be identified as being either physical (e.g., oceans or mountains) or as a (real or imagined) line in the sand that defines the boundary of a nation, state, city or other jurisdiction, separating the

The development of trade blocs in an era of globalisation

61

rights and laws of one from the other’s (Gregory, Johnston, Pratt, Watts & Whatmore, 2009). This study uses political boundaries as the division between nation states (compare Shin, 2002). Emerging from economic activities, this regional state therefore comprises the whole of two or more nation states, based on the outcome of global trade and multinational activity (Poon & Pandit, 1996). Due to intensifying forces of globalisation in the 1990s, the number and volume of linkages between countries has also been strengthened and forced integration of otherwise spatially separate economic activities (Poon et al., 2000). Glenn (2008) finds similar evidence for an increasing number of countries to be more integrated into a common economic system, where he defines regionalisation as “. . . economic activity [that] is becoming ever more concentrated within clearly identifiable geographic regions” (p. 80). Poon (1997b) argues that this regionalisation of international trade is a “natural phenomenon”, emphasising that the governmental promotion of linkages between countries by joining EIAs, including free trade agreements and customs unions, can spur the natural process of regionalisation. Frankel et al. (1995, 1996) reflect that close historical and geographic ties between the countries drive this “natural” process of regionalisation. Poon (1997a,b) adds that regionalisation very often happens without these explicit aims or measures and underlines that the regionalisation of countries is often only driven by market forces. These “natural trading regions” consist of countries having high trading intensities with one another due to geographic proximity, lower transaction costs and cultural affinities creating spatial biases (compare Frankel et al., 1995, 1996). It must be noted that this regionalisation process based on trade differs significantly from closed trade blocs. The latter is basically formed by political institutions and decisions, whereas the former reveals the workings of global markets (Poon & Pandit, 1996).

3.2.2

Importance of regionalisation

Now that we are in an era of globalisation, an increasing number of countries has become integrated in the global economic system. The “end of geography” was consequently declared by several authors (see, among others, Greig, 2002; Friedman, 2005). This spurred the powerful counterargument that “the world is not flat” (for an excellent overview, see Christopherson, Garretsen & Martin, 2008), stating that globalisation sometimes makes geodesic distance less important due to, e.g., improved mobile and electronic communication methods. At the same time, how-

62

Chapter 3

ever, distance continues to be relevant, as witnessed by the growing role of trade in knowledge-intensive sectors, in which face-to-face contact is required to facilitate transferring specific sectoral knowledge (Dicken, 2007). This remains important even in the presence of decreasing costs to accessing foreign markets (Andresen, 2009b). Most countries have a growing quantity of their national economic activities in some sort of a “relationship” with a growing number of economic partners (Andresen, 2009a,b). Do these global flows of trade increase the probability of regionalisation with a larger number of countries? Are fewer trade clusters the result of this growing number of relationships? According to Glenn (2008), the answer is “No”. Even though trade might increase at a global level, the distance to trade partners influences the volume of trade and therefore the regionalisation processes. In a conceptual framework, MacLeod & Jones (2007) agree that distance is the most dominant determinant for generating economic regions, but underline the influence of culture, politics and history on relative distance (“spatial flows”) as discontinuous or strengthening forces. The territories (or economic regions, rather) that are created by these spatial flows have strong, yet breakable ties because they are victim to continuous “territorial restructuring” (p. 1182) in a world of political and economic turbulence. This study provides an empirical application to this conceptual idea. Poon et al. (2000) argue that although the forces of globalisation have led to a decrease in the number of trade blocs since 1985, they find no evidence of a triadisation of the world economy. They demonstrate that the changing shape and nature of trade clusters is largely ascribed to continental lines and at the same time strengthened by FDI, which does appear to have a “network” shape and does not need continuous continental regions. Andresen (2009a) agrees with these findings. According to Poon et al. (2000), the fluctuation of trade clusters is “operated by centrifugal and centripetal forces operating simultaneously, resulting in constellations of relationships (e.g., trade clusters) where space is both sticky (important) and fluid (flexible) at the same time” (p. 440, emphasis added). Economic flows do not only follow absolute space, but are also led by relative distance, formed by historical ties, shared cultures and (changing) political systems. Countries trade with their “natural partners” (K. Anderson & Norheim, 1993; Frankel et al., 1995, 1996; Michalak & Gibb, 1997).

The development of trade blocs in an era of globalisation

63

The importance of spatial structure in the process of regionalization was investigated by Poon & Pandit (1996). They find, rather than a triad structure of global trade, evidence for six trade clusters focused around six core market countries— the US, Germany, France, the UK and the former USSR—underlining that these clusters are not geographically contiguous regions but “functional units” defined by the volume of trade interactions and explained by the intensity of bilateral trade between the members. They argue that “scale economies and large efficient markets are instrumental in shaping emerging regional configuration” (p. 284), which is in accordance with the New Trade Theory. This evidence was indicated by Michalak & Gibb (1997, p. 266) as a strong case for “regionalism as one of the most influential factors determining world trade flows.” Andresen (2009a) finds that over time, trade clusters become more regionally focused. This is because the importance of distance (as a proxy for transportation costs) is increasing. Although nations in trade clusters do not need to be neighbouring per se, geographic proximity does improve the chance of being in the same trade cluster. Similar results for the importance of proximity to increasing trade intensification are reported by Hanink & Cromley (2005). As Andresen (2009a,b) states, the arrangements of nations in economic space originate from a trading networks, which in turn are strongly influenced by historical and political ties (see also Shin, 2002; Lee & Park, 2005). Having historical similarities and political ties decreases the relative distance between nations and increases the intensity of trade between the countries involved. According to Andresen (2009a)’s analysis, historical ties significantly influence regionalisation until 1981, especially if the historical ties have led to a shared institutional context. For example, a shared religious majority could create similar cultures for nations, as Yamazaki (1996) finds that Christianity has provided a unifying framework for Europe. Other cultural ties such as language can also improve trade relationships, especially because a shared language is often combined with historical colonial ties (see, among others, Shin, 2002; Lee & Park, 2005; Head, Mayer & Ries, 2010). Our reading of the literature suggests that trade clusters emerge not only due to proximity, but also on account of other geographic, economic, political, historical and cultural characteristics. As discussed above, it is of particular interest whether trade blocs have been regionally focused in the period leading up to the 1980s, owing to geographic proximity and the regionally-focused nature of trade policies embedded in EIAs. Have trade blocs become more inter-regionally oriented in the wake of globalisation and interregional trade agreements? These questions cannot

64

Chapter 3

be adequately addressed when countries are grouped into trade blocs a` priori. The next section therefore explains how trade blocs can be obtained endogenously and used in the empirical investigation that follows.

3.3

Trade blocs in the world economy

This section is structured as follows. Section 3.3.1 discusses how trade clusters are identified using the intramax hierarchical clustering technique. section 3.3.2 shows that traditional, static regional classifications are poor predictors of the dynamic, trade-based functional areas identified by the intramax procedure. Finally, results for the period 1950-2005 are presented in section 3.3.3.

3.3.1

Method

The intramax hierarchical clustering technique identifies functional areas using flow data. Technically, the procedure works as follows1 All trade flow data are arranged in a square contingency table, or an origin-destination matrix, for a given year. The origins (exporting countries) are in rows and the destinations (importing, reporting countries) in columns. The intramax algorithm maximises the proportion of the total interaction which takes place within the aggregation of basic data units—the amount of within-group interaction (I)—while minimising the number of cross-boundary movements (see Masser & Brown, 1975). The objective function applied by the intramax procedure is based on the differences between the observed flows (aij ) and their expected probabilities, as indicated by the marginal row and column totals, (aij∗ ):

max I = ( aij − aij∗ ) + ( a ji − a∗ji ), i 6= j.

(3.1)

In applying this function, a transformed matrix is calculated which measures the largest total interactions between pairs of countries in excess of the total of the expected values derived from the row and column totals. The expected value of each element is the product of the column sum times the ratio of the row sum to the total interaction. For example, the expected flow out of Country 2 to Country 1, 1 The FlowMap software package can be used to perform the intramax procedure. See http://flowmap .geog.uu.nl for more information.

The development of trade blocs in an era of globalisation

65

a21 , where aij is the element in row i and column j of the contingency table, is given as:

∗ a21 =

∑ j a2j = i ∑ j aij

∑ ai1 ∑ i

∑ ai1 i

∑ j a2j . n

(3.2)

The row sum of the contingency table is ai∗ = ∑ j aij , the column sum is a j∗ = ∑i aij and the total interaction is the sum of row sums, n = ∑i ∑ j aij . Normalising the flows such that n = 1 and aij∗ = ai∗ a j∗ means that “the difference between observed and expected values aij − aij∗ for the flow between zone i and j may be taken as a measure of the extent to which the observed flow exceeds (or falls below) the flow that would have been expected, simply on the basis of the size of the row and column marginal totals” (Masser & Brown, 1975, p. 512). Based on this procedure, the pair of countries with the highest absolute level of interaction are merged and henceforth considered to belong to the same functional area. Their interaction (i.e., cross-border trade) becomes intrazonal. This merger causes the matrix to be reduced by one column and one row. The objective function can then be reapplied and the modified matrix recalculated to yield the next pair with the highest level of interaction. This procedure is repeated until the (n − 1)th iteration has been completed. Intuitively, consider a world of 5 countries: A, B, C, D, and E. On the outset, if the country-pair with the largest interaction is D and E, these countries are merged into functional area DE. We then consider the flows between A, B, C, and bloc DE. If the largest interaction is then between A and B, they form functional area AB. Next, consider the flows between AB, C, and DE. When the largest interaction occurs between bloc AB and country C, they are merged, which yields a newlyformed bloc ABC and bloc DE. The merging process continues until all blocs have been joined into one, single functional area. The output is conveniently drawn in dendograms that show how countries have been clustered, based on the degree of intrazonal interaction. If the degree of intrazonal interaction is 0 percent, each country is taken as its own unique functional area. If this number is 100 percent, it means that all countries have been merged into one cluster. At what degree of intrazonal interaction should a functional area be identified? The literature does not provide a uniform answer. Some authors draw the proverbial line at a level of clustering where homogeneity within a cluster is lost (Goetgeluk, 2006), but it is unclear how homogeneity should then be defined. One way

66

Chapter 3

is by choosing clusters if there is a large increase in the intrazonal flows. However, a large increase in the intrazonal flows during the fusion process does not generally indicate “a merger of two rather homogeneous zones.” Still, the most practical “stop criterion” is one that uses the functional areas found “just before the high increase in intrazonal flows” (Goetgeluk, 2006, p. 11), although some degree of freedom may need to be maintained to identify realistic clusters. For each year of output, the dendograms are used to determine if countries belong to the same functional area. Following Poon et al. (2000), a 10 percent cut-off is applied to identify large breaks in intra-cluster interactions. Countries that are in the same part of the dendogram after such a break has occurred belong to the same trade bloc. The results are illustrated in section 3.3.3. Note that the intramax method assumes that every country ultimately belongs to a bloc of countries that trade intensively among themselves. So, although the algorithm facilitates the endogenous identification of trade clusters in the sense that regions are not defined up front, it does not allow for the possibility that countries may not belong to any cluster (see Andresen, 2009a, p. 29).

3.3.2

Static vs. dynamic trade blocs

As discussed in the literature review, analyses of world trade flows are typically conducted by identifying economic clusters on the predefined basis of political borders and/or geographic characteristics, such as studies on the effectiveness of the “triad”. As such, these regional classifications tend to be static, in the sense that they do not change over time. Consider, for example, the World Bank (2011a)’s classification of economies by region. It divides the worlds economies into 9 regions: East and Southern Africa, West Africa, East Asia and Pacific, South Asia, Eastern Europe and Central Asia, Rest of Europe, Middle East, North Africa, and the Americas. This regional classification is simply based on static geographic characteristics. Alternatively, the identification of these regions can be entirely data-driven. The advantage of this approach is that functional areas are identified by actual economic activity—in this case trade flows. Given the dynamic nature of the data, it means that functional regions can be identified not only for different time periods, but that the composition of these clusters is also subject to change. In short, the intramax procedure yields a way to identify dynamic, data-driven trade blocs that are not subject to predefined political or geographic characteristics.

The development of trade blocs in an era of globalisation

67

Table 3.1. Predictive power of a static regional indicator Period

Correlation

Static estimate

Observations

Wald χ2

Log likelihood

1950-2005 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

0.262 0.200 0.133 0.146 0.132 0.261 0.198 0.259 0.211 0.326 0.347 0.368 0.485

-0.280*** (0.00) -0.469*** (0.01) -0.454*** (0.01) -0.428*** (0.01) -0.412*** (0.01) -0.283*** (0.01) -0.314*** (0.01) -0.275*** (0.01) -0.294*** (0.01) -0.155*** (0.01) -0.135*** (0.01) -0.118*** (0.01) -0.141*** (0.01)

390,960 32,580 32,580 32,580 32,580 32,580 32,580 32,580 32,580 32,580 32,580 32,580 32,580

23,801 4,048 3,947 3,736 3,579 2,023 2,422 1,922 2,163 657 497 385 543

-258,106 -19,841 -19,982 -20,221 -20,373 -21,485 -21,237 -21,546 -21,400 -22,247 -22,330 -22,388 -22,306

Note: The dependent variable is the intramax bloc variable. The independent variable is a dummy that is 1 if both countries in a dyad belong to the same region according to World Bank (2011a) and 0 otherwise. Correlations are displayed in the column labelled “Correlation”. Probit estimates (marginal effects) are labelled “Static estimate”. Estimates marked ***/**/* are significant at the 1/5/10 percent level. Robust standard errors (clustered by country-pair) are in brackets.

Are these dynamic blocs any different from standard, static regional classification systems? The answer clearly is affirmative. As discussed below, the intramax technique yields trade blocs that change in terms of their number, size, orientation and composition from year to year. Moreover, visual inspection shows that the functional areas are considerably different compared to the ones suggested by, e.g., the World Bank’s regional classification system. This result is quantitatively strengthened by the results in Table 3.1. A probit model is used to estimate the probability that the static regional indicator correctly predicts the dynamic intramax cluster variable. Hence, the dependent variable is 1 if a country-pair belongs to the same dynamic trade cluster in year t and 0 otherwise. The independent variable is 1 if a country-pair belongs to the same static cluster based on World Bank (2011a). The correlation coefficients indicate a weak correlation between these variables. Moreover, the statistically significant, yet negative probit estimates indicate that static country-pair groupings incorrectly predict the probability that these countries in fact belong to the same dynamic trade bloc. In other words, static region classifications are poor predictors of dynamic, data-driven trade clusters obtained from the intramax technique.

68

Chapter 3

3.3.3

Results

The trade clusters obtained from the intramax procedure are mapped for the period 1950-2005 in Figures 3.1-3.7.2 For ease of comparison, each cluster is labeled as being oriented towards the major economy (country) in that cluster. This orientation is simply based on the dominant economy in that particular cluster based on its economic size, as measured in terms of GDP. Visual inspection of the figures indicates that there is a decrease in the number of clusters over time and that they have become more geographically focussed. According to Figure 3.1, some clusters have the expected colonial ties with African countries or India in the 1950s. These ties become weaker in the 1960s. Colonial ties with France and the UK are still visible, but the northeastern part of Africa is clustered with the USSR. In this period, the European cluster splits into different clusters and the US dominance in Latin America decreases (see Figure 3.2). This trend continues into the 1970s. Colonial ties decrease in importance, although—with the exception of Australia—the link with the Commonwealth remains strong. Two clusters emerge in Europe, with one focusing on Germany and the other cluster in the eastern part focussing on the USSR. The cluster around Japan increases strongly and the cluster around India makes its debut (see Figure 3.3). The US regains importance in Latin America in the 1980s (see Figure 3.4). The largest cluster in Africa has links with most of Europe. The trade cluster around Japan gets even larger and extends into east Africa. In general, the clusters start becoming more geographically connected. The most noticeable change in the 1990s is the emergence of the South-African cluster and the strong increase of the Indian cluster (see Figure 3.5). Furthermore, this is the only period in which the UK has no trade cluster dominance and is included in the one oriented towards Germany. The US dominance in Latin America disappears. By the turn of the millennium, the Indian and South-African clusters merge and include China (see Figure 3.6). This is the only time China is in a different cluster. France has a strong cluster with Africa and the UK returns, only this time not in the context of the Commonwealth but as the centre of a north-European cluster. Europe is split in three clusters: one around the northern European countries, one around east Europe and Russia, and the third cluster is southern Europe and Africa. The clusters appear to be even more geographically continuous around 2005 (see Figure 3.7). Africa seems to have settled mostly in one cluster with India and a strong geographic focus is evident in the case of Europe and the Asian-Pacific. 2 The

underlying trade data are described on p. 73.

The development of trade blocs in an era of globalisation

Figure 3.1. Trade bloc orientations in 1950

Figure 3.2. Trade bloc orientations in 1960

Figure 3.3. Trade bloc orientations in 1970

69

70

Figure 3.4. Trade bloc orientations in 1980

Figure 3.5. Trade bloc orientations in 1990

Figure 3.6. Trade bloc orientations in 2000

Chapter 3

The development of trade blocs in an era of globalisation

71

Figure 3.7. Trade bloc orientations in 2005 Three conclusions can be drawn from these results. First, as discussed in the introduction, the rise of interregional EIAs and economic globalisation in the past three decades may have altered the scope of the world’s trade blocs to be less geographically oriented. Nevertheless, the trade blocs obtained from the intramax analysis do not reflect this development. Despite regionalism’s increasingly inter-regional coverage, the configuration of the underlying trade blocs has remained remarkably regionally-focused. Second, the intramax method yields no evidence of triadisation. Although there are clusters with an orientation towards the US, Germany and Japan for the period 1950-200, the latter disappears in 2005. In fact, the results show strong evidence of other trade clusters next to the so-called triad. Surprisingly, the clusters orientated towards Brazil, Russia, India, and Japan and/or China have been important for their regional economies for many decades, although some of these countries have only recently attracted the literatures attention as “emerging markets”. Third, the North and Latin American clusters are much more stable over time in terms of the number of countries and the trade bloc orientation compared to the rest of the world. Most “turbulence” seems to occur in Europe, Africa and Asia. It is also clearly seen that the importance of historical/colonial ties fade after the 1980s (compare Poon et al., 2000; Andresen, 2009a,b; Head et al., 2010). The next section explores how countries may be clustered into trade blocs. This will be done by showing how geographic, economic, political, historical and cultural variables derived from the gravity equation are associated with the probability that trade partners belong to the same trade bloc obtained with the intramax technique.

72

Chapter 3

3.4

What drives the development of trade blocs?

Having constructed the intramax-based trade blocs in the previous section, we now proceed to explore how various representations of absolute and relative distance are associated with their development. The empirical model is introduced in section 3.4.1, followed by a brief description of the dataset in section 3.4.2. Results are presented in section 3.4.3.

3.4.1

Method

The intramax hierarchical clustering technique has been used to determine if pairs of countries belong to the same trade bloc in a given year. But what drives this trade bloc formation? This section discusses the modelling strategy that is used to obtain insight into how geographic, economic, political, historical and cultural factors are associated with the trade blocs identified by the intramax procedure. A probit model is estimated according to: 0 Pr ( Bijt = 1| Xijt ) = Φ( Xijt β ),

(3.3)

where Bijt is a binary variable that is 1 if both countries in a country-pair, with importer i and exporter j, belong to the same functional area obtained from the intramax technique in year t.3 Xijt is a vector of regressors, with: Xijt = (ln( Distanceij ), EI Aijt , ln( GDPijt ), Borderij , Landlockedij , Languageij , Colonyij , Coloniserijt , Countryijt , Fi , Fj , Ft )0 .

− −−

(3.4)

The time-varying regressors are as follows. EI Aijt is a binary variable that is 1 if the country-pair has an economic integration agreement and 0 otherwise. GDPijt is the average real GDP of both trade partners. Ft represents year dummies in the pooled regressions to account for unobserved time-varying phenomena. The remaining regressors are time-invariant. Distanceij is the geographic distance between countries i and j in kilometres and Borderij is a binary variable that is 1 if the countries share a land border and 0 otherwise. Landlockedij accounts for countries’ access to the ocean and takes on the values 0, 1 or 2, depending on the number of countries in the dyad that are landlocked. Languageij is a binary variable that is 1 if the country-pair shares a common language and 0 otherwise. Colonyij 3 An interesting alternative is to examine the probability that a country-pair belongs to the same EIA. This avenue is explored in Baier & Bergstrand (2004).

The development of trade blocs in an era of globalisation

73

is a binary variable that is 1 if the country-pair has ever been in a colonial relationship and 0 otherwise and Coloniserijt is a binary variable that is 1 if the country-pair has a common coloniser after 1945. Countryijt is a binary variable that is 1 if both countries are the same country and 0 otherwise, which accounts for cases such as Czechoslovakia breaking into two nations at some point in the dataset. Fi and Fj represent dummies for countries i and j, respectively, which allow for variations in the import prices faced by trade partners vis-`a-vis the prices relative to all their other trade partners. These country-level effects are known as the “multilateral resistance term” (see J. Anderson & van Wincoop, 2003; Feenstra, 2004). As discussed in chapter 2, bilateral trade flows are influenced by the MRT. In the current empirical setting, the MRT also needs to be controlled for because the dependent (intramax clustering) variable is heavily reliant on these trade flow data.

3.4.2

Data

A description of the panel dataset that is used throughout this study is provided in section 2.3.3, which discusses its construction and the data sources. Descriptive statistics of the variables in the dataset are provided in Table 3.2. Note that there are two differences between the datasets employed in chapters 2 and 3. First, fewer variables are needed in this chapter, which lowers the probability that observations are dropped due to missing data. A larger number of countries can consequently be included in the final analysis. For a complete list of countries, see Table 3.3. Second, the panel is restricted to 5-year intervals and covers the period 1950 to 2005. Although annual data on intramax-driven trade blocs may provide for a richer analysis, their construction is a labour-intensive task. Moreover, the objective is to study the long-term evolution of these trade blocs and not year-by-year changes by themselves.

74

Chapter 3

Table 3.2. Descriptive statistics Variable B ln Distance EIA ln GDP Common border Number landlocked Common language Ever colony Common coloniser Common country

Observations

Mean

Std. deviation

Minimum

Maximum

283,862 281,212 283,862 216,086 281,212 283,862 281,212 281,212 281,212 281,212

0.12 8.81 0.05 10.41 0.02 0.27 0.19 0.01 0.13 0.01

0.33 0.75 0.21 1.73 0.13 0.49 0.39 0.11 0.34 0.09

0.00 2.35 0.00 3.91 0.00 0.00 0.00 0.00 0.00 0.00

1.00 9.90 1.00 15.91 1.00 2.00 1.00 1.00 1.00 1.00

Table 3.3. Countries in dataset Afghanistan, Albania, Algeria, American Samoa, Angola, Antigua & Barbuda, Argentina, Armenia, Aruba, Australia, Austria, Azerbaijan, Bahamas, Bahrain, Bangladesh, Barbados, Belarus, Belgium, BelgiumLuxembourg, Belize, Benin, Bermuda, Bhutan, Bolivia, Bosnia & Herzegovina, Botswana, Brazil, Brunei, Bulgaria, Burkina Faso, Burundi, Cambodia, Cameroon, Canada, Cape Verde, Cayman Islands, Central African Republic, Chad, Chile, China, Colombia, Comoros, Costa Rica, Croatia, Cuba, Cyprus, Czech Republic, Czechoslovakia, D.R. Congo, Denmark, Djibouti, Dominica, Dominican Republic, East Germany, Ecuador, Egypt, El Salvador, Equatorial Guinea, Eritrea, Estonia, Ethiopia, Faeroe Islands, Falkland Islands, Fiji, Finland, France, French Guiana, French Polynesia, Gabon, Gambia, Georgia, Germany, Ghana, Gibraltar, Greece, Greenland, Grenada, Guadeloupe, Guam, Guatemala, Guinea, Guinea-Bissau, Guyana, Haiti, Honduras, Hong Kong, Hungary, Iceland, India, Indonesia, Iran, Iraq, Ireland, Israel, Italy, Ivory Coast, Jamaica, Japan, Jordan, Kazakhstan, Kenya, Kiribati, Kuwait, Kyrgyzstan, Laos, Latvia, Lebanon, Lesotho, Liberia, Libya, Lithuania, Luxembourg, Macao, Macedonia, Madagascar, Malawi, Malaysia, Maldives, Mali, Malta, Marshall Islands, Martinique, Mauritania, Mauritius, Mexico, Micronesia, Moldova, Mongolia, Montserrat, Morocco, Mozambique, Myanmar, Namibia, Nauru, Nepal, Netherlands Antilles, Netherlands, New Caledonia, New Zealand, Nicaragua, Niger, Nigeria, North Korea, Norway, Oman, Pakistan, Palau, Panama, Papua New Guinea, Paraguay, Peru, Philippines, Poland, Portugal, Qatar, Republic of Congo, Reunion, Romania, Russia, Rwanda, St. Helena, St. Kitts & Nevis, St. Lucia, St. Pierre-Miquelon, St. Vincent & Grenadines, Samoa, S˜ao Tom´e & Pr´ıncipe, Saudi Arabia, Senegal, Serbia & Montenegro, Seychelles, Sierra Leone, Singapore, Slovak Republic, Slovenia, Solomon Islands, Somalia, South Africa, South Korea, Spain, Sri Lanka, Sudan, Suriname, Swaziland, Sweden, Switzerland, Syria, Tajikistan, Tanzania, Thailand, Togo, Tonga, Trinidad & Tobago, Tunisia, Turkey, Turkmenistan, Tuvalu, Uganda, Ukraine, United Arab Emirates, United Kingdom, United States, Uruguay, USSR, Uzbekistan, Vanuatu, Venezuela, Vietnam, Yemen, Zambia, Zimbabwe.

The development of trade blocs in an era of globalisation

3.4.3

75

Results

Estimation results for the pooled dataset spanning 1950-2005 are displayed in Table 3.4. Columns 1-3 are based on a limited number of (time-varying) independent variables and columns 4-6 include all regressors listed in equation 3.3. Comparison of the goodness of fit between these two sets reveals that the variables responsible for most of the variation in the data are of an economic and/or time-varying nature, i.e., GDP and EIAs. The dependent variable is lagged by 5 years and included as regressor in columns 2 and 5. An additional lag of 10 years is included in columns 3 and 6. Doing so increases the goodness of fit. Moreover, the positive coefficients suggest that the evolution of trade-bloc formation is pathdependent. As expected, the results indicate that the probability that countries belong to the same trade cluster is very much influenced by distance in both absolute and relative terms. The negative sign of the distance coefficient means that if the physical distance between two countries increases, the probability of being in the same trade bloc declines. Although related, there is a slight difference between how the effect of distance is interpreted in the standard gravity equation and how it should be interpreted in the present study. The standard interpretation is that countries trade less over greater distances. Here, the findings reflect that countries are likely to be clustered in the same trade bloc, the greater their proximity. Indeed, the one implies the other: the negative effect of distance on trade volumes between import and exporter also decreases their probability of being clustered in the same trade bloc. However, the contribution of the present approach lies in its explicit focus on the clustering of trade partners. Strikingly, participation in EIAs is found to increase a country-pair’s likelihood of being in a trade bloc. This result is robust across the different model specifications in Table 3.4 and for most of the cross-sections in Table 3.5. It supports one of the key findings that already became apparent in the previous section, namely, that trade blocs have remained surprisingly regional despite the increasingly global focus of the web of trade agreements active in the world economy. The positive coefficient for GDP means that poorer countries have a lower probability of being in a trade cluster than rich countries due to the purchasing power of the (extended) market. With respect to the geographic controls, contiguity (having a common border) is a complicated variable. Instinctively, sharing a border should increase the probability of being in a cluster (compare Andresen, 2009a), but the na-

76

Chapter 3

Table 3.4. Pooled probit estimation results, 1950-2005 Variable ln Distance EIA ln GDP

(1)

(2)

(3)

(4)

(5)

(6)

-0.868*** (0.02) 0.386*** (0.02) 0.091*** (0.01)

-0.732*** (0.01) 0.279*** (0.02) 0.070*** (0.01) 0.899*** (0.02)

-0.683*** (0.02) 0.236*** (0.02) 0.055*** (0.01) 0.757*** (0.01) 0.565*** (0.01)

-0.860*** (0.02) 0.370*** (0.02) 0.098*** (0.01)

-0.740*** (0.01) 0.273*** (0.02) 0.073*** (0.01) 0.867*** (0.02)

-0.238*** (0.05)

-0.226*** (0.04)

-0.701*** (0.01) 0.239*** (0.02) 0.057*** (0.01) 0.741*** (0.01) 0.546*** (0.01) -0.219*** (0.04)

0.880*** (0.16) 1.809*** (0.33) 0.172*** (0.02) 0.678*** (0.07) 0.324*** (0.02) 0.087 (0.08)

0.789*** (0.14) 1.624*** (0.29) 0.098*** (0.02) 0.480*** (0.06) 0.225*** (0.02) 0.065 (0.07)

0.748*** (0.14) 1.538*** (0.27) 0.025 (0.02) 0.402*** (0.05) 0.160*** (0.02) 0.052 (0.07)

-58,102 13,098 216,086 0.27

-49,605 27,475 197,004 0.32

-43,142 30,714 177,250 0.35

Bt−1 Bt−2 Common border Number landlocked - one - both Common language Ever colony Common coloniser Common country

Log-likelihood χ2 Observations Pseudo R2

-58,906 10,860 216,086 0.26

-49,912 26,606 197,004 0.32

-43,270 30,113 177,250 0.35

Notes: Marginal effects. The dependent variable is binary, where a country-pair in the same functional cluster scores 1 and 0 otherwise. Estimates marked ***/**/* are significant at the 1/5/10 percent level. Robust standard errors (clustered by country-pair) are in brackets. Intercepts and fixed effects are omitted to save space.

tions in the trade blocs are often in a clusters with more than 16 nations with which they obviously cannot all share a border. This has a negative bearing on the estimated parameter value. The geographic position of countries or the set of countries also has a strong influence on the probability of being in a trade cluster. Being landlocked means less access to seaports and hence an increase in transportations costs (compare Frankel et al., 1996). If one of the countries in a dyad is not landlocked, it is less likely to be part of a landlocked country’s trade cluster than when both are landlocked. A common language (as a proxy for cultural similarity) has a positive effect on the probability of being in a trade bloc, which confirms the discussed literature.

The development of trade blocs in an era of globalisation

77

This result is not robust to the inclusion of a lagged dependent variable. When one of the two countries has been a colony of the other country in the past (such as Senegal being colonised by France, Angola by Belgium and India by the United Kingdom), this is reflected in a positive effect on the probability of being in the same trade cluster. Similar effects are found for two countries in a dyad that have a common former coloniser (such as Senegal and Nigeria do with respect to France and India and Myanmar (Burma) with respect to the UK). Not unexpectedly, having been the same country (such as the Czech Republic and Slovakia) has positive effects on the probability of being in the same trade cluster, although this is not robust to the inclusion of a lagged dependent variable. Next, we investigate the variation of these findings across time. This is done by estimating equation 3.3 cross-sectionally at each of the 5-year intervals. The parameter estimates are presented in Table 3.5. The effects for proximity reflect our discussion of the literature. The declining importance of distance on trade bloc formation until the 1970s can perhaps be ascribed to the notion that proximity becomes less restrictive to the geographic distribution of economic activity due to the forces of globalisation. However, the need for personal information exchanges and the location-bound nature of multinational enterprises’ firm-specific advantages may have contributed to the importance of proximate countries being in the same trade cluster (see also Rugman & Verbeke, 2004). Higher GDP levels and the presence of EIAs increase the probability of trade bloc formation, although the coefficients show some fluctuation across time periods. This may be explained by the changing “North-South” orientation of trade, i.e., more market extension rather than cheap imports (Poon & Pandit, 1996). Contiguity has a negative effect on clustering. As explained before, this has to do with the number of countries in a trade bloc that do not share a border due to their geographic layout. Again, being landlocked increases the probability of belonging to the same trade cluster. The language variable is not robust across all time periods, but tends to suggest that cultural familiarity promotes trade bloc formation. Having common colonial ties has a positive effect on the probability of being in the same cluster (compare Head et al., 2010). Note that the colonial effect might be overtaken by shared language, shared cultural background and institutional similarities (compare Glenn, 2008; Andresen, 2009b). The results for common coloniser are similar to the longtimespan estimations. The effect of two countries having been one country in the past, tends not to be significant.

1950

-1,212 1,375 6,006 0.46

0.978*** (0.36) 1.996** (0.75) 0.852*** (0.09) 1.841*** (0.24) 1.360*** (0.16) 0.415 (0.31)

0.359*** (0.08) -0.501** (0.19)

-1.085*** (0.05)

1955

-1,928 1,867 7,482 0.42

-0.050 (0.45) -0.001 (0.98) 0.867*** (0.07) 0.718*** (0.15) 1.014*** (0.11) 0.365 (0.28)

0.352*** (0.06) -0.535*** (0.16)

-1.000*** (0.04)

1960

-2,667 2,019 13,572 0.31

1.517*** (0.32) 3.046*** (0.65) 0.574*** (0.05) 0.883*** (0.15) 0.553*** (0.07) -0.051 (0.15)

-0.833*** (0.03) 0.739** (0.23) 0.230*** (0.04) -0.256* (0.11)

1965

-2,890 2,644 14,520 0.33

1.615*** (0.35) 3.417*** (0.71) 1.085*** (0.05) 1.313*** (0.14) 0.655*** (0.06) -0.140 (0.14)

-0.543*** (0.03) 1.479*** (0.21) 0.261*** (0.04) -0.008 (0.11)

1970

-3,762 2,303 16,256 0.31

0.616 (0.37) 1.523* (0.75) 0.584*** (0.05) 1.279*** (0.12) 0.496*** (0.06) -0.129 (0.14)

-0.693*** (0.03) 0.442** (0.16) 0.205*** (0.04) 0.031 (0.10)

1975

-4,819 2,649 19,460 0.29

0.923* (0.37) 1.711* (0.75) 0.207*** (0.04) 0.628*** (0.12) 0.266*** (0.05) 0.185 (0.14)

-0.700*** (0.03) 0.405** (0.13) 0.059* (0.02) 0.058 (0.10)

1980

-5,008 2,766 20,880 0.32

0.925** (0.29) 1.736** (0.59) 0.241*** (0.04) 0.519*** (0.11) 0.459*** (0.05) 0.086 (0.14)

-0.875*** (0.02) -0.024 (0.10) 0.153*** (0.02) 0.151 (0.10)

1985

-5,052 2,753 22,350 0.35

1.047** (0.39) 2.590** (0.80) -0.041 (0.04) 0.728*** (0.12) 0.603*** (0.05) 0.278 (0.16)

-0.954*** (0.03) -0.092 (0.08) 0.028 (0.02) 0.139 (0.11)

1990

-5,130 2,643 22,350 0.46

3.912*** (0.57) 7.920*** (1.15) -0.012 (0.04) 0.172 (0.15) 0.267*** (0.05) 0.007 (0.23)

-1.510*** (0.04) -0.255** (0.09) 0.287*** (0.03) -0.157 (0.14)

1995

-4,394 2,966 22,350 0.56

0.635 (0.58) 1.265 (1.17) -0.216*** (0.05) 0.762*** (0.17) 0.435*** (0.06) 0.855* (0.34)

-1.495*** (0.06) 0.839*** (0.08) 0.301*** (0.03) -0.307* (0.15)

2000

-5,305 3,816 22,952 0.46

2.053*** (0.53) 3.994*** (1.07) -0.319*** (0.04) 0.842*** (0.17) 0.394*** (0.05) 0.210 (0.20)

-1.299*** (0.05) 0.457*** (0.06) 0.168*** (0.03) -0.467*** (0.12)

2005

-3,872 2,985 19,182 0.47

0.035 (0.46) 0.528 (0.93) 0.187*** (0.04) 0.614*** (0.16) -0.037 (0.05) 0.110 (0.21)

-1.284*** (0.05) 0.274*** (0.07) -0.069* (0.03) -0.350** (0.11)

Notes: Marginal effects. The dependent variable is binary, where a country-pair in the same functional cluster scores 1 and 0 otherwise. Estimates marked ***/**/* are significant at the 1/5/10 percent level. Robust standard errors (clustered by country-pair) are in brackets. Estimates for intercepts and fixed effects are omitted to save space.

Log-likelihood χ2 Observations Pseudo R2

Common country

Common coloniser

Ever colony

Common language

- both

Number landlocked - one

Common border

ln GDP

EIA

ln Distance

Variable

Table 3.5. Cross-sectional probit estimation results, 1950-2005

78 Chapter 3

The development of trade blocs in an era of globalisation

79

Sensitivity analysis Until now, the results have accounted for unobserved timevarying (annual) phenomena and time-invarying importer and exporter effects. As argued above, these terms represent J. Anderson & van Wincoop (2003)’s “multilateral resistance term” (MRT). According to Baier & Bergstrand (2007), a potential drawback of the the current specification is that it does not allow for the countrylevel effects to be time-varying, which ideally should be the case in the presence of time-varying price levels. Consequently, Fi and Fj may have to be replaced by time-varying importer and exporter effects, Fit and Fjt , respectively. Moreover, unobserved factors influencing trade flows (or in this case trade bloc formation) may be correlated with those influencing the presence of EIAs. This may be controlled for by including dyad effects Fij . Equation 3.4 is accordingly modified and is specified as:

Xijt = (ln( Distanceij ), EI Aijt , ln( GDPijt ), Borderij , Landlockedij , Languageij , Colonyij , Coloniserijt , Countryijt , Fit , Fjt , Fij )0 .

(3.5)

The extensive nature of the present dataset and, as a result, the great number of time-varying and dyad fixed effects, make it computationally infeasible to estimate the model on the full dataset at once. Fortunately, estimation is still feasible when using random sampling techniques. Table 3.6 shows the parameter estimates of ten (non-overlapping) samples of 20,000 observations that were randomly drawn from the dataset.4 What happens when a time-varying MRT is used? All samples consistently indicate that the probability that countries are clustered decreases as the distance between them increases. Notice that the parameter estimates for the EIA variable are positive and slightly larger than those obtained with the time-invariant MRT, suggesting that dyad fixed effects do eliminate unobserved factors that are likely to affect both trade and regionalism. The variables controlling for cultural similarity and colonial history are positive in all the samples. Then again, the results for economic size, contiguity, being landlocked and having a shared history show are not systematically significant. Overall, the sensitivity analysis reinforces the main finding that geographic proximity and regionalism are associated with trade bloc formation. 4 Estimates based on random samples of countries confirm these findings. All these results are available upon request.

-4,395 1,630 15,210 0.22

-4,367 1,614 15,102 0.21

-0.074* (0.04) 0.102 (0.10) 0.229*** (0.04) 0.567*** (0.12) 0.253*** (0.05) 0.231 (0.14)

-0.692*** (0.02) 0.387*** (0.06) -0.037*** (0.01) -0.126 (0.11)

Sample 2

-4,380 1,313 15,286 0.22

-0.013 (0.04) 0.178 (0.10) 0.204*** (0.04) 0.564*** (0.11) 0.362*** (0.05) 0.370* (0.18)

-0.694*** (0.03) 0.468*** (0.06) -0.017 (0.01) -0.244* (0.11)

Sample 3

-4,457 1,868 15,266 0.22

-0.049 (0.04) -0.167 (0.10) 0.212*** (0.04) 0.552*** (0.11) 0.306*** (0.05) 0.146 (0.15)

-0.722*** (0.02) 0.379*** (0.06) -0.031** (0.01) 0.004 (0.11)

Sample 4

-4,350 1,254 15,155 0.22

-0.033 (0.04) -0.097 (0.12) 0.274*** (0.04) 0.272* (0.13) 0.359*** (0.05) 0.103 (0.16)

-0.709*** (0.03) 0.298*** (0.06) -0.015 (0.01) -0.114 (0.11)

Sample 5

-4,464 1,677 15,180 0.20

-0.063 (0.04) -0.219 (0.11) 0.213*** (0.04) 0.614*** (0.11) 0.322*** (0.05) 0.162 (0.13)

-0.656*** (0.02) 0.426*** (0.06) -0.040*** (0.01) -0.112 (0.11)

Sample 6

-4,479 1,313 15,241 0.20

-0.052 (0.04) 0.088 (0.10) 0.197*** (0.04) 0.311* (0.13) 0.344*** (0.05) 0.350* (0.14)

-0.650*** (0.02) 0.380*** (0.06) -0.030** (0.01) -0.074 (0.10)

Sample 7

-4,385 1,708 15,153 0.21

0.034 (0.04) -0.089 (0.10) 0.215*** (0.04) 0.553*** (0.12) 0.324*** (0.05) 0.263 (0.15)

-0.693*** (0.02) 0.386*** (0.06) -0.034*** (0.01) -0.145 (0.10)

Sample 8

-4,436 1,656 15,232 0.21

-0.045 (0.04) -0.089 (0.11) 0.201*** (0.04) 0.414** (0.14) 0.315*** (0.05) 0.303* (0.13)

-0.713*** (0.02) 0.304*** (0.06) -0.017 (0.01) -0.124 (0.10)

Sample 9

-4,359 1,629 15,225 0.21

-0.064 (0.04) -0.105 (0.10) 0.235*** (0.04) 0.514*** (0.11) 0.271*** (0.05) 0.501** (0.17)

-0.687*** (0.02) 0.436*** (0.06) -0.013 (0.01) -0.206 (0.11)

Sample 10

Notes: Marginal effects. Random samples, based on 20,000 observations. The dependent variable is binary, where a country-pair in the same functional cluster scores 1 and 0 otherwise. Estimates marked ***/**/* are significant at he 1/5/10 percent level. Robust standard errors (clustered by country-pair) are in brackets. Estimates for intercepts and fixed effects are omitted to save space.

Log-likelihood χ2 Observations Pseudo R2

Common country

Common coloniser

Ever colony

Common language

- both

Number landlocked - one

Common border

ln GDP

-0.094** (0.04) -0.013 (0.10) 0.273*** (0.04) 0.331** (0.12) 0.369*** (0.05) 0.439** (0.15)

-0.649*** (0.02) 0.461*** (0.06) -0.034*** (0.01) 0.042 (0.10)

ln Distance

EIA

Sample 1

Variable

Table 3.6. Probit estimation results based on random sampling

80 Chapter 3

The development of trade blocs in an era of globalisation

3.5

81

Discussion and conclusion

This chapter contributes to the literature by providing estimates of dynamic trade bloc formation for the period 1950-2005 with a significantly larger dataset than its predecessors (compare Poon & Pandit, 1996; Poon, 1997a,b; Poon et al., 2000). It assesses how trade blocs have evolved and shows how geographic, economic, political, historical and cultural factors are associated with their composition. The results can be summarised as follows: First, the intramax technique provides strong evidence that trade blocs are highly dynamic and different to those obtained by static classifications based on countries geographic or political characteristics. These results oppose the notion of triadisation in the international business literature, which suggests that the world economy can simply be divided in three static regions. Our visual results also indicate that emerging economies such as Brazil, Russia, India and China have played a central role in trade bloc formation long before attracting attention as “BRICs” in the recent literature. Second, increasing distance between countries reduces the probability that they belong to the same trade bloc. At the same time, countries do not have to be strict neighbours in order to belong to the same trade cluster. However, the negative distance coefficient makes clear that even though countries do not have to be geographic neighbours, they are only likely to belong to the same trade cluster if the distance between them is not too large. These results complement the gravityequation literature by showing that distance not only affects the level of trade between nations, but that it also influences their probability of belonging to the same trade cluster. Third, cultural similarities promote trade-bloc formation. Geographic positioning is also relevant in the sense that the probability that countries belong to different trade blocs decreases with their access to seaports. Countries that used to be colonial dependencies are more likely to trade with their colonisers than with other countries. Unsurprisingly, having a common coloniser increases the probability that nations are in the same trade cluster. Contrary to the expectations raised in the beginning of this chapter, the trade blocs obtained with the intramax technique show that their orientation is still surprisingly regional, despite the prevalence of an interregional “spaghetti bowl” of inter-regional trade agreements. The overall findings discussed in the previous and present chapter indicate that regionalism is positively and robustly associated with trade promotion and the formation of (regional) trade blocs.

82

Chapter 3 Until now, however, an underlying issue that has yet to be fully addressed in

this thesis is that of endogeneity: does the legal enforcement of a new trade agreement result in more trade, or do governments sign trade agreements only to reflect the economic realities in which their economies are already involved? The answer to the question of endogeneity has important implications for trade policy. Indeed, policy makers may rest assured that their trade agreements will bear the envisioned fruits of trade creation when EIAs withstand the test of potential endogeneity. Conversely, the economic merit of these agreements is doubtful if they only appear to be a day after the fair. This matter is the focus of following chapter.

Chapter 4

Do we really know that trade agreements increase trade? 4.1

Introduction

Chapter 1 illustrates the remarkable proliferation of regionalism since the 1990s. This development has triggered the “building vs. stumbling block” debate, which considers whether preferential trade liberalisation poses a threat to the continuity of non-discriminatory trade policies implemented through the multilateral trade system of the World Trade Organisation (WTO). The increasingly more important role of regionalism in the international trade policy arena gives the distinct impression that governments are all too eager to intensify discriminatory trade relationships with a select number trade partners, while non-discriminatory negotiations at the multilateral level seem to struggle along. Using the gravity equation to study determinants of international trade flows, the literature commonly finds that regionalism fosters trade (see also chapter 2). In addition, chapter 3 suggests that, in addition to proximity, regionalism is strongly associated with the formation of regional trade blocs. Nevertheless, at least three features in the applied international trade literature warrant additional attention. First, the prevalent approach assumes that trade agreements bring about trade liberalisation that will change the amount of cross-border trade (henceforth called the “EIA effect”). But what if the reverse is true? If governments sign EIAs only after substantial changes in their trade levels have already materialised, then factors other than trade policy are affecting trade and possibly the decision to sign trade

84

Chapter 4

agreements in the first place. The result is that conventional EIA effects may suffer from endogeneity bias. With only the exception of a few studies that will be noted below, it is a relatively unexplored problem in the literature. The present chapter aims to provide new, unbiased estimates of regionalism’s effect on trade. When it comes to empirical evidence on EIA effects, a trade-off is usually made between a “generalist” and a “specialist” approach. The former involves aggregation of numerous EIAs in large datasets to measure the impact of economic integration on world trade “at large” (see, e.g., Rose, 2004; Baier & Bergstrand, 2007). The empirical outcome of the generalist strategy usually is that regionalism promotes world trade, ceteris paribus. While these studies have the appeal of generalisability, not much else can or is normally said about the agreement-specific economic consequences of the individual, underlying EIAs. As such, the insightfulness of empirical findings based on this approach for trade policy specialists is limited. In contrast, a large literature pursues a specialist approach in which the effects of individual EIAs are examined. Its advantage is that it goes beyond generalising statements about trade policy. In acknowledging that EIAs are different in nature, their heterogeneous effects on international trade flows may be explored in greater detail. Consider, for example, the European Community (EC) and the Israel-US Free Trade Agreement (FTA). As will be shown in chapter 5, the former is far more extensive in terms of trade liberalisation than the latter. It could be argued that they differ in economic and political motive, with the emphasis of the EC being on regional integration in Europe. The main objective of the Israel-US FTA, however, is to serve the US as an instrument to pursue foreign policy goals in the Middle East (Rosen, 2004). Conceptual differences between EIAs may lead to different economic outcomes, adding to the relevance of disaggregating regionalism’s “overall” effect to illustrate EIA specific outcomes. Unfortunately, studies following the specialist strategy are usually restricted to one geographic area, a small number of EIAs and shorter time horizons. Most importantly, methodological improvements applied at the generalist level do not tend to be systematically applied when investigating heterogeneous EIA effects. So, while the “specialists” certainly provide details on select subsets of EIAs, methodological differences make it challenging to compare findings across studies.

Do we really know that trade agreements increase trade?

85

Finally, most studies are conducted in such as way that they only explore the effect of present trade liberalisation on present trade flows. In doing so, intertemporal effects are ignored. Consider the possibility that nations trade more in anticipation of their governments enforcing a new trade agreement. McLaren (1997) argues that infrastructure and delivery systems involving sunk costs may be redirected to increase trade, in anticipation of an EIA to become effective. This is what is called the “anticipation effect” in this chapter. Alternatively, trade increases may be (temporarily) delayed while trade liberalisation is gradually enforced. A reading of the trade agreements covered in this study suggests that it is not uncommon for the tariff schedules to specify gradual liberalisation over the course of several years. For example, an agreement may specify tariff cuts of 20 percent upon enforcement of the agreement, 50 percent after 5 years, 80 percent after 7 years and 100 percent after 10 years. Clearly, not capturing these potential “phase-in effects” may underestimate the true impact of EIAs. The added value of this chapter is that it addresses the shortcomings of the generalist and specialist approaches by considering both sides of the proverbial coin. Using Baier & Bergstrand (2007)’s first-differencing technique that controls for endogeneity bias and inter-temporal effects, we estimate the gravity equation for an up-to-date dataset that covers 165 countries, 60 years and 165 EIAs. In addition to providing novel estimates at an aggregate level, a key feature of this chapter is that it systematically applies the same methodology across a large number of EIA to obtain agreement-specific EIA effects. It will be shown that traditional estimates of the EIA effect tend to be exaggerated if the possibility of endogeneity remains unaccounted for. Phase-in effects are more often confirmed than are anticipation effects. Surprisingly, estimates of the agreements’ individual effect on trade are often zero when endogeneity bias has been eliminated. Taken together, the results suggest that EIAs tend to be a product, rather than a cause of intensified cross-border activity. However, there are no apparent reasons why some EIAs are better at promoting trade than others, even when concerns regarding endogeneity bias have been alleviated. The remainder of this chapter is structured as follows. Section 4.2 reviews the empirical literature on EIA effects. Section 4.3 presents the methodology and results in a step-wise fashion. Section 4.4 discusses the main findings and concludes.

86

4.2

Chapter 4

Literature

Studies on the EIA effect date back to the years when the gravity equation was in its infancy. The degree to which EIAs affect international trade, given the trading nations’ economic and geographic properties, is commonly estimated using a loglinear form of the basic gravity equation of international trade: ln( Mijt ) = β 0 + β 1 ln( GDPit ) + β 2 ln( GDPjt ) + β 3 ln( Distanceij )

+ β 4 EI Aijt + β 5 zijt + eijt ,

(4.1)

where Mijt indicates country i’s aggregate imports from exporter j in year t, β 0 is a constant, and GDP is a proxy for the respective country’s economic size. Distanceij is the physical distance between i and j, proxied by the distance between their capitals in kilometres. EI Aijt is a binary variable that is 1 if both countries in a countrypair belong to the same EIA and 0 otherwise. Such binary variables are typically included for each and every EIA included in the dataset in order to determine the extent to which EIAs affect trade among countries belonging to the same trade bloc (also known as “intra-bloc bias”). zijt represents a host of control variables such as physical size, the number of landlocked and island states in the dyad, whether countries share a common border, language, colonial heritage, etc.

4.2.1

Genesis

Using a simple gravity equation specified as equation 4.1, Tinbergen (1962) shows that Commonwealth and Benelux preferences had trade-creating properties, while Linnemann (1966) confirms this for French and Belgian preferences. Aitken (1973) finds that the European Economic Community (EEC) and European Free Trade Association (EFTA) had a negative effect on exports in the 1950s, but that both of these trade agreements had a positive effect in the 1960s (see also Balassa, 1967). Sattinger (1978) and Abrams (1980) argue that these positive effects may also be ascribed to the EEC and EFTA during most of the 1970s and, according to Bikker (1987), the 1980s, although Bergstrand (1985)’s results are less compelling. Trading preferences in the Council for Mutual Economic Assistance (CMEA) are also found to be trade-creating during 1954-1970 (see Hewett, 1976; Pelzman, 1977) and between the EEC and Central and Eastern European Countries (CEEC-4) during 1987-2005 (see Caporale, Rault, Sova & Sova, 2009).

Do we really know that trade agreements increase trade?

87

Extending the scope of preferential treatments beyond European borders, Aitken & Obutelewicz (1976) show that trade between the EEC and African Associate Countries (AAC) also gradually benefited from preferential market access during 1956-1971 (see also Oguledo & MacPhee, 1994). Sapir (1981) confirms that the EEC and 10 major developing nations experienced increased trade as a result of preferential treatment during 1967-1978. Economic integration efforts in the Americas were also frequently discussed, with scholars providing mostly qualitative reasons for the Latin American Free Trade Association (LAFTA) and Central American Common Market (CACM)’s failure and speculating about the viability of the relatively new Andean Pact (see Elkan, 1975; Lizano & Willmore, 1975; Ffrench-Davis, 1977; Vaitsos, 1978). Brada & Mendez (1983, 1985) are among the first who provide more extensive studies, with a dataset covering 1954-1977 and 46 countries. They estimate the individual effects of 5 major EIAs on trade flows by including separate dummy variables for each EIA. The authors find positive effects for the EEC, EFTA and CACM, an insignificant effect for the Andean Pact and a negative one for LAFTA. Subsequent studies by Havrylyshyn & Pritchett (1991) show positive effects for the EEC, LAFTA, CACM and a negative one for the Association of Southeast Asian Nations (ASEAN) (1980-1982, 35 countries). In contrast, Wang & Winters (1991) and Hamilton & Winters (1992) find trade-creating effects for all of the above, including ASEAN (1984-1986, 76 countries). Soloaga & Winters (2001) find positive effects for Latin-American countries involved in the Andean Pact, CACM, LAFTA and/or Mercosur (1980-1996, 58 countries). Frankel, Stein & Wei (1995)’s estimates yield trade-creating effects for the Asia-Pacific Economic Cooperation (APEC), Mercosur, the Andean Pact, and the EEC as of the 1980s, but not for the EFTA or NAFTA (North American Free Trade Agreement, 1965-1990, 63 countries). Frankel & Wei (1997) use a slightly different approach to the one described thus far. They account for the possibility that EIAs may also have trade-diverting effects by including additional “extra-bloc openness” binary variables: ln( Mijt ) = β 0 + β 1 ln(Yit ) + β 2 ln(Yjt ) + β 3 ln( Distanceij ) I E + β 6 zijt + eijt , + β 4 EI Aijt + β 5 EI Aijt

(4.2)

I is 1 when one EIA member trades with another EIA member and 0 where EI Aijt E is 1 when an EIA member trades with a non-member and otherwise, and EI Aijt

0 otherwise. The former measures the extent to which EIA members focus their trade among themselves, in addition to what can be predicted based on their eco-

88

Chapter 4

nomic and geographic properties (i.e., preference-induced trade creation or intrabloc bias). The latter is an indicator of the degree to which EIA members trade with non-members, below or beyond what can be expected within the gravity model (i.e., preference-induced trade diversion or extra-bloc openness). Frankel & Wei (1997)’s find evidence of trade creation for the Andean Pact, ASEAN, EEC and Mercosur. The EFTA and US-Canada FTA (the predecessor of NAFTA) show signs of trade diversion, while open bloc trade creation is ascribed to the Andean Pact, ASEAN and EEC (1965-1990, 63 countries).1

4.2.2

Old school

Rose (2000), Feenstra, Markusen & Rose (2001) and Frankel & Rose (2002) use the gravity equation to show that EIAs generally lead to net trade creation. However, these studies are biased because they fail to account for unobserved price indices, which, as shown by J. Anderson & van Wincoop (2003), are a crucial addition to the gravity equation. Their argument is that so-called “multilateral resistance terms” (MRT) must be included in gravity equations to take into account that trade between two trading countries is also affected by their bilateral trade barrier relative to their average trade barriers vis-`a-vis all of their other trade partners. Feenstra (2004, chapter 5) demonstrates that these unobserved price indices can be conveniently incorporated in estimates of the gravity equation by adding importer and exporter fixed effects. This typically yields: ln( Mijt ) = β 0 + β 1 ln( GDPit ) + β 2 ln( GDPjt ) + β 3 ln( Distanceij ) + β 4 EI Aijt

+ β 5 zijt + γiFi + δj Fj + ζ t Ft + eijt ,

(4.3)

where Fi represents fixed effects for the importing country and Fj for the exporting country. Year effects (Ft ) are also commonly included to correct for common trends and shocks. Numerous studies have subsequently incorporated these methodological improvements, but results remain mixed. For example, Ghosh & Yamarik (2004) provide an extensive robustness study with data covering 1970-1995 (5-year intervals), 186 countries and 12 major EIAs. Using extreme bounds analysis, the authors show that none of the EIAs are trade creating. However, Eicher, Henn & Papageorgiou (2008) use the same data and find strong evidence for EIA induced trade 1 Open bloc trade creation occurs when countries liberalise trade barriers among both bloc members and non-members at the same time.

Do we really know that trade agreements increase trade?

89

creation after accounting for model uncertainty with Bayesian model averaging. Trade-creating effects were found for the EU, diverting effects were detected for NAFTA, and open bloc trade creation was found among EIAs in Asia and Europe. These findings suggest that after five decades of research, empirical evidence on the effectiveness of EIAs in advancing trade is still mixed.

4.2.3

New school

As stated above, an important issue that has yet to receive more attention in the applied gravity-equation literature is the fact that trade policy is not strictly exogenous. This gives rise to endogeneity bias in the estimated effect of EIAs on trade levels because countries self-select into EIAs as a result of pre-existing trade levels. In other words, economic and/or political forces not fully captured lead nations to trade more, even though a trade agreement is absent. Only later may the need and/or opportunity arise for the governments of these trade partners to formalise their relationship with trade agreements covering issues such as bilateral liberalisation, economic cooperation, harmonisation of standards, etc. In this case, trade facilitates EIA formation. As explained, this goes against the intuition normally used when interpreting gravity-equation estimates, including the results presented in chapters 2 and 3. Trefler (1993) employs an instrumental variable (IV) approach to account for possible endogeneity in U.S. trade policy. He shows that the impact of policy on trade is ten times larger when trade policy is treated endogenously. Baier & Bergstrand (2002) use IV and Heckman control functions to deal with endogeneity issues in a larger dataset. They find that correcting for endogeneity quadruples the estimated overall effect of EIAs on trade, although the results are mixed for individual major EIAs. Magee (2003) finds similar but unstable results with an IV approach. However, Baier & Bergstrand (2007) argue that panel regression techniques are more suitable to account for endogeneity and yield more stable results with countrypair fixed effects (I.-H. Cheng & Wall, 2005), first-differencing (Bayoumi & Eichengreen, 1998) and matching methods (Baier & Bergstrand, 2009). The estimated EIA effects are subject to endogeneity bias if any of the right-hand side variables are correlated with the error term. Wooldridge (2002, chapter 10) explains that omitted variables, selection bias, simultaneity and measurement error are potential sources of endogeneity bias. Baier & Bergstrand (2007) explain that the error term could represent unobservable policy-related barriers (not accounted for by

90

Chapter 4

typical RHS gravity variables) that are correlated with governments’ decision to sign an EIA. Moreover, J. Anderson & van Wincoop (2003)’s multilateral resistance terms are actually time-varying, which means that the importer and exporter fixed effects also need to be time-varying to fully capture the MRT. Therefore, Baier & Bergstrand (2007) estimate EIA’s collective treatment effect on world trade using a first-differences approach (1960-2000 (in 5-year intervals), 96 countries and 52 EIAs). By taking a 10-year phase-in period into account, the authors find a treatment effect of 68 percent. Moreover, they obtain a negative and statistically insignificant anticipation effect, suggesting that “firms delay trade temporarily in anticipation of an impending agreement” (p. 90). For more applied work following this approach, see Baier et al. (2007) for EIAs in Latin America and Baier et al. (2008) for trade agreements in Europe. The present study builds on Baier & Bergstrand (2007)’s analysis by providing new and refined estimates using a dataset that covers a greater number of countries, years and EIAs. As shown in this literature review, all contributions to date have only considered the role of EIAs as a collective or that of a handful of individual EIAs. The consistent methodological approach followed in the present study, combined with the large and up-to-date dataset, makes it possible to determine the effect of a large number of EIAs on international trade, both collectively and individually. As such, this study makes it possible to compare EIA effects across a large number of trade agreements. The next section discusses the estimation strategy and results in greater detail.

4.3 4.3.1

Methodology and results Data

The dataset is based on the one discussed in section 2.3.3. It includes 165 countries and contains observations for the period 1948-2007 and yields a maximum of 165 × 164 × 60 = 1, 623, 600 observations. Table 4.1 lists the countries included in the dataset, while the EIAs are the same as those included in chapter 2. Note that when it comes to using the first-differencing technique with a 10-year phase-in period, only those agreements that been enforced before 1998 are considered. This amounts to 89 agreements that can be estimated individually, which are listed in Table 4.4 on p. 97.

Do we really know that trade agreements increase trade?

91

Table 4.1. Countries in dataset Afghanistan, Algeria, Angola, Antigua & Barbuda, Argentina, Aruba, Australia, Austria, Bahamas, Bahrain, Bangladesh, Barbados, Belgium, Belize, Benin, Bhutan, Bolivia, Botswana, Brazil, Brunei, Bulgaria, Burkina Faso, Burundi, Cambodia, Cameroon, Canada, Cape Verde, Cayman Islands, Central African Republic, Chad, Chile, China, Colombia, Comoros, Costa Rica, Cuba, Cyprus, Democratic Republic of Congo, Denmark, Djibouti, Dominica, Dominican Republic, Ecuador, Egypt, El Salvador, Equatorial Guinea, Eritrea, Ethiopia, Faeroe Islands, Falkland Islands, Fiji, Finland, France, French Polynesia, Gabon, Gambia, Germany, Ghana, Greece, Greenland, Grenada, Guatemala, Guinea, Guinea-Bissau, Guyana, Haiti, Honduras, Hungary, Iceland, India, Indonesia, Iran, Ireland, Israel, Italy, Ivory Coast, Jamaica, Kenya, Kiribati, Kuwait, Laos, Lesotho, Liberia, Libya, Luxembourg, Madagascar, Malawi, Malaysia, Maldives, Mali, Malta, Marshall Islands, Mauritania, Mauritius, Mexico, Micronesia, Montserrat, Morocco, Mozambique, Myanmar, Namibia, Nauru, Nepal, Netherlands, Netherlands Antilles, New Caledonia, New Zealand, Nicaragua, Niger, Nigeria, Norway, Oman, Pakistan, Panama, Papua New Guinea, Paraguay, Peru, Philippines, Poland, Portugal, Qatar, Republic of Congo, Romania, Rwanda, St. Helena, St. Kitts & Nevis, St. Lucia, St. Pierre-Miquelon, St. Vincent & Grenadines, Samoa, S˜ao Tom´e & Pr´ıncipe, Saudi Arabia, Senegal, Seychelles, Sierra Leone, Singapore, Solomon Islands, South Africa, South Korea, Spain, Sri Lanka, Sudan, Suriname, Swaziland, Sweden, Switzerland, Syria, Tanzania, Thailand, Togo, Tonga, Trinidad & Tobago, Tunisia, Turkey, Tuvalu, Uganda, United Arab Emirates, United Kingdom, United States, Uruguay, Vanuatu, Venezuela, Vietnam, Zambia, Zimbabwe.

4.3.2

Benchmark estimates

Following the literature, this chapter uses the gravity equation with GDP, geodesic distance and trade agreements—specifically, EIAs and unilateral schemes introduced under the Generalised System of Preferences (GSP)—as key determinants of the level of cross-border trade flows between two nations. The purpose of this section is to explore the “true” effect of economic integration agreements on world trade. First, we estimate the gravity equation based on its “old school” variant with J. Anderson & van Wincoop (2003)’s original, time-invariant multilateral resistance terms (MRT) to get a sense of the parameter estimates that permeate the literature. The model is:

ln( Mijt ) = β 0 + β 1 ln( GDPit ) + β 2 ln( GDPjt ) + β 3 ln( Distanceij ) + β 4 GSPijt

+ β 5 EI Aijt + γiFi + δj Fj + ζ t Ft + eijt ,

(4.4)

where Fi and Fj are importer and exporter effects, i.e, the time-invariant MRT, and Ft controls for unobserved time-varying phenomena.

92

Chapter 4 In order to explicitly account for potential concerns about endogeneity bias, we

then proceed with a “new school”’ version of the gravity equation with Baier & Bergstrand (2007)’s time-varying MRT. The improved specification is: ln( Mijt ) = β 0 + β 1 ln( GDPit ) + β 2 ln( GDPjt ) + β 3 ln( Distanceij ) + β 4 GSPijt

+ β 5 EI Aijt + γitFit + δjtFjt + ζijFij + eijt ,

(4.5)

where Fit and Fjt are importer-year and exporter-year effects, respectively, that account for a time-varying version of J. Anderson & van Wincoop (2003)’s MRT. Fij controls for unobserved phenomena that may be correlated with both a countrypair’s level of trade (left-hand side variable) and factors related to the dyad having an EIA (right-hand side variable). How do these specifications perform empirically? Regression estimates are provided in Table 4.2. Results based on the “old school” equation 4.4 are shown in the top panel and those from the “new school” equation 4.5 are in the lower panel. Column 1 estimates the full model with annual data and behaves as expected: trade decreases over distance and increases with income and trade agreements, all else constant. Notice that the large number of country-time effects usually makes it impossible to use annual data. In these cases, we resort to 5-year averages (1948-1952, 1953-1957,. . ., 2003-2007) in column 2. Alternatively, 5-year intervals (1950, 1955,. . ., 2007) are used to verify the robustness of the results in column 3. The “old school” estimates are robust. However, the “new school” EIA parameter estimates are remarkably lower and GSP becomes insignificant. Notice that estimates for bilateral distance and income are not reported because these variables are collinear with the dyad-fixed effects and country-time effects, respectively. Columns 4-9 rerun the gravity equations on annual data for different time periods. Interestingly, the “old school” EIA effects are positive in all decades except the 1950s. A possible explanation is that there were only very few EIAs active in this timeframe. In contrast, the “new school” EIA effects are remarkably lower and not always statistically significant. This suggests that the conclusions drawn about regionalism’s impact on trade throughout the past six decades crucially depend on allowances for endogeneity bias. A similar conclusion can be drawn with respect to GSP. “New school” results are not robust across the specified timeframes, while “old school” results are positive in all time periods except the 1950s (because it predates GSP) and the 2000s, perhaps reflecting that GSP schemes nowadays tend to be dissolved and replaced by EIAs.

(1) 1948-2007 Annual

(2) 1948-2007 Averages

(3) 1948-2007 Intervals

(4) 1950s Annual

442,141 0.65

0.753*** (0.02) 1.043*** (0.03) 0.710*** (0.04) 0.374*** (0.02)

-1.184*** (0.02)

120,425 0.69

0.662*** (0.02) 0.907*** (0.03) 0.817*** (0.04) 0.398*** (0.03)

-1.172*** (0.02)

92,710 0.69

0.751*** (0.03) 1.052*** (0.03) 0.876*** (0.04) 0.483*** (0.03)

-1.270*** (0.02)

28,480 0.67

1.134*** (0.11) 1.420*** (0.14) -0.832*** (0.14) (omitted)

-0.773*** (0.03)

43,186 0.68

1.038*** (0.08) 1.095*** (0.12) 0.516*** (0.12) 0.470*** (0.04)

-0.879*** (0.02)

(5) 1960s Annual

70,120 0.70

0.869*** (0.07) 0.751*** (0.09) 0.399*** (0.10) 0.487*** (0.04)

-1.213*** (0.03)

(6) 1970s Annual

141,996 0.04 0.35

0.503*** (0.03) -0.003 (0.02) 106,750 0.02 0.35

0.593*** (0.03) -0.041 (0.03) 32,052 0.00 0.25

0.025 (0.07) (omitted)

51,100 0.01 0.22

1.117*** (0.21) 0.058* (0.03)

82,609 0.00 0.22

0.096 (0.05) 0.054 (0.04)

102,521 0.02 0.13

0.073 (0.06) -0.001 (0.04)

92,296 0.70

0.550*** (0.06) 0.587*** (0.07) 0.548*** (0.07) 0.412*** (0.03)

-1.353*** (0.03)

(7) 1980s Annual

133,914 0.00 0.12

0.178*** (0.04) 0.036 (0.03)

117,527 0.73

0.918*** (0.06) 1.071*** (0.08) 0.630*** (0.05) 0.299*** (0.03)

-1.443*** (0.02)

(8) 1990s Annual

123,593 0.00 0.31

0.203** (0.06) -0.696*** (0.07)

88,075 0.58

-0.124 (0.14) 0.480** (0.15) 0.676*** (0.04) -0.195*** (0.03)

-1.025*** (0.02)

(9) 2000s Annual

Notes: With the exception of columns 2 and 3, estimates are based on annual data. Column 2 uses 5-year averages and column 3 uses 5-year intervals. Estimates marked ***/**/* are significant at the 1/5/10 percent level. Robust standard errors (clustered by country-pair) are in brackets. ln Distance and ln GDP are collinear with the “new school” fixed effects. Other estimates are omitted when not applicable or to save space.

Observations Between R2 Within R2

GSP

EIA

Equation 4.5: “New school” specification with importer-year, exporter-year and country-pair effects

Observations R2

GSP

EIA

- exporter

ln GDP - importer

ln Distance

Equation 4.4:“Old school” specification with importer, exporter and year effects

Variable

Table 4.2. Benchmark gravity estimates

Do we really know that trade agreements increase trade? 93

94

Chapter 4 Finally, equation 4.5 can be rewritten in first-differences. Doing so is of partic-

ular interest in applied work because first-differencing eliminates the country-pair fixed effects and thereby significantly reduces the right-hand side variables and computation time. The lagged EIA variable is now also added to account for the trade agreements’ potential phase-in effects. As above, the model is estimated using 5-year averages and 5-year interval data for robustness. Adding a lag gives:

dlnMij,t−(t−1) = β 1 dEI Aij,t−(t−1) + β 2 dEI Aij,(t−1)−(t−2) + β 3 dGSPij,t−(t−1)

+ dFi,t−(t−1) + dFj,t−(t−1) + vij,t−(t−1) ,

(4.6)

where vij,t−(t−1) = eijt − eij,t−1 is assumed to be white noise. The coefficients of interest are β 1 , which measures the effect of EIA formation on trade 5 years after the agreement has been enforced, and β 2 , which measures this effect for the 5 years thereafter. The overall treatment effect is the sum of the obtained estimates for β 1 and β 2 , provided that these are significant at the 5 percent level. As above, GDP is collinear with the country-year effects and subsequently dropped from the estimation procedure. Note that the obtained parameter estimates and standard errors are unbiased and consistently estimated, provided that (1) the error terms are not serially correlated and (2) the EIA variable can be assumed to be strictly exogenous. The former can easily be taken into account by computing cluster robust standard errors, a procedure applied throughout this study (Wooldridge, 2002, p. 283). The latter can be tested by including a future value of the dependent variable of interest (EIA), as demonstrated by Wooldridge (2002, p. 285). The assumption of strict exogeneity is violated and the estimated treatment effect biased when anticipation effects are found, i.e., when the estimated coefficient is not equal to zero. Equation 4.6 can easily be augmented to include a future value of EI Aijt and this is done as a robustness check. The results from estimating equation 4.6 are reported in Table 4.3. Columns 1-3 show parameter estimates for data based on 5-year averages and columns 4-6 for data arranged at 5-year intervals. Columns 1 and 4 do not include a lagged variable. Columns 2 and 5 are based entirely on equation 4.6, featuring a 10-year phase-in period. Columns 3 and 6 also include a future EIA variable to control for anticipation effects.

Do we really know that trade agreements increase trade?

95

Table 4.3. First-difference estimates, 1948-2007 Variable

(1) Averages

(2) Averages

(3) Averages

(4) Intervals

(5) Intervals

(6) Intervals

0.136*** (0.04)

0.104*** (0.04) 0.203*** (0.04)

0.089** (0.03)

0.097*** (0.03) 0.196*** (0.03)

GSP

-0.053** (0.03)

-0.050* (0.03)

0.038 (0.04) -0.130*** (0.05) 0.011 (0.04) 0.044* (0.03)

0.018 (0.03)

0.020 (0.03)

0.108*** (0.04) 0.207*** (0.04) 0.104*** (0.04) 0.068** (0.03)

Observations R2

108,542 0.204

105,174 0.204

86,925 0.194

79,363 0.167

77,038 0.165

63,634 0.176

EIA EIA (lagged) EIA (lead)

Notes: Estimates in columns 1-3 are based on 5-year averages and those in columns 4-6 are based on 5-year intervals. Estimates marked ***/**/* are significant at the 1/5/10 percent level. Robust standard errors (clustered by country-pair) are in brackets. Estimates for fixed effects are omitted to save space.

Two conclusions can be drawn from Table 4.3. First, notice that the anticipation effect is insignificant when 5-year averages are used. This outcome is consistent with Baier & Bergstrand (2007), suggesting that there are no anticipation effects overall and that the assumption of strict exogeneity is not violated. However, the anticipation effect is significant when 5-year intervals are used, which may bias the parameter estimates. The remainder of this chapter continues estimating the EIA effects with data based on 5-year averages, with robustness checks with the interval data. Second, correcting for endogeneity bias and phase-in effects turns out to be of remarkable importance. The EIA effect is e(0.136) − 1 ≈ 15 percent without a lag (column 1). Allowing for phase-in effects increases the EIA effect to e(0.104+0.203) − 1 ≈ 35 percent. The interval data provide a similar outcome. We have established that our concerns about endogeneity are warranted and that Baier & Bergstrand (2007)’s methodology is appropriate to obtain unbiased EIA effects. Following the “generalist” approach, in which a large dataset on EIAs is used to determine their aggregate effect on world trade, the results confirm that endogeneity bias overestimates traditional EIA effects, while not accounting for phase-in effects underestimates regionalism’s effect on trade. The following section adopts the methodological improvements that have so far been implemented at an aggregate level and applies them at the level of the individual trade agreement. In doing so, the chapter provides estimates for a large number of trade agreements while consistently using the most recent advances in the literature.

96

Chapter 4

4.3.3

Individual EIA effects

How may individual EIA effects be determined? Traditionally, an “old school” gravity equation without time-varying MRT would include a separate dummy variable for every EIA. As discussed in the previous section, the obtained parameter estimates are biased without a time-varying MRT and dyad fixed effects. Unfortunately, it becomes computationally cumbersome to estimate a gravity equation that includes all the appropriate fixed effects as well as a large number of dummies for individual trade agreements. However, the first-differencing technique can still be used in such a way that unbiased (in terms of endogeneity) EIA effects are obtained for an individual trade agreement S:

S S O dlnMij,t−(t−1) = β 1 dEI Aij,t −(t−1) + β 2 dEI Aij,(t−1)−(t−2) + β 3 dEI Aij,t−(t−1)

(4.7)

+ β 4 dGSPij,t−(t−1) + dFi,t−(t−1) + dFj,t−(t−1) + vij,t−(t−1) , where EIA dummies with superscript S are 1 if agreement S is enforced and 0 otherwise. EIA variables with superscript O are 1 if any agreement other than S is enforced. Say that we want to estimate the impact of the Australia-New Zealand Closer Economic Agreement (ANZCERTA). β 1 and β 2 will then measure the 10year phase-in effect of this particular EIA. This is also known as the average treatment effect. It is obtained by adding the (statistically significant) coefficients. As in the previous section, we also control for the presence of all other EIAs and GSP arrangements in the dataset, which are captured by β 3 and β 4 , respectively. The estimates based on equation 4.7 are provided in Table 4.4. Columns 1-4 are based on estimates with 5-year averaged data. Column 1 shows the estimated EIA effect after the first phase-in period of 5 years (β 1 ) and column 2 for the second phase-in period of 5 years (β 2 ). Finally, column 3 shows the total treatment effect for each individual agreement. This value is the sum of the statistically significant coefficients in the first two columns.

Do we really know that trade agreements increase trade?

97

Table 4.4. Individual EIA effects Agreement Andean Community (Cartanega) Arab Maghreb Union (AMU) Asia Pacific Trade Agreement (APTA) Association of Caribbean States (ACS) Association of South East Asian Nations (ASEAN) Australia-New Zealand (ANZCERTA) Australia-Papua New Guinea (PATCRA) Bolivia-Chile Bolivia-Mexico Canada-Chile Canada-Israel Canada-US(-Mexico) FTA (CUSFTA/NAFTA) Caribbean Community (CARICOM) CARICOM-Colombia CARICOM-Venezuela Central American Common Market (CACM, 1966-1970) Central American Common Market (CACM, 1990-present) Central American Economic Integration Agreement (CAEIA) Central European Free Trade Agreement (CEFTA) Chile-Venezuela China-India Colombia-Mexico-Venezuela Common Market for Eastern and Southern Africa (COMESA) Costa Rica-Mexico Costa Rica-Panama Economic Community of Central African States (ECCAS/CEMAC) Economic Community of West African States (ECOWAS) European Community (EC) EC-Algeria EC-Andorra EC-Bulgaria EC-Cyprus EC-Czech Republic EC-Egypt EC-Estonia EC-Faroe Islands EC-Hungary EC-Iceland EC-Latvia EC-Lithuania EC-Malta EC-Norway EC-Overseas Countries and Territories EC-PLO EC-Poland EC-Romania EC-Russia

(1) β1

(2) β2

(3) Total

0.557** (0.20) 1.303*** (0.34) 0.158 (0.28) -0.506*** (0.14) -0.172 (0.21) 0.270 (0.15) 0.282 (0.38) 0.475 (0.33) 0.244 (0.29) 0.755 (1.12) 0.723 (0.68) 0.193 (0.14) 0.430* (0.20) -0.269 (0.20) 0.461*** (0.12) 2.267*** (0.21) 0.917** (0.29) 0.473 (0.25) -0.773* (0.32) 0.317 (0.18) -0.245 (0.65) -0.649 (0.62) 0.328* (0.15) -0.082 (0.27) -0.150 (0.16) -0.159 (0.26)

0.476** (0.18) 0.100 (0.30) 0.106 (0.23) 0.441*** (0.13) 0.357 (0.24) 0.255 (0.15) -0.162 (0.28) -1.203 (0.72) -0.019 (0.26) 0.087 (0.71) -0.719 (0.48) -0.344 (0.41) 0.049 (0.17) 0.110 (0.18) -1.020*** (0.16) 0.847*** (0.21) -1.051*** (0.29) 1.920*** (0.20) 0.648 (0.37) -0.501 (0.69) 0.113 (0.96) -0.027 (0.28) -0.077 (0.18) 0.091 (0.25) -0.280 (0.19) -0.274 (0.35)

1.033 1.303 0.000 -0.065 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 -0.559 3.114 -0.134 1.920 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

0.040 (0.15) 0.758*** (0.08) 0.785*** (0.09) 0.884*** (0.09) 0.432*** (0.07) -0.214*** (0.06) -0.257*** (0.07) 0.713*** (0.07) -0.165* (0.08) 0.508*** (0.09) -0.218** (0.07) 0.755*** (0.08) -0.315** (0.10) -0.315** (0.10) -0.150* (0.06) 0.714*** (0.07) 0.458*** (0.09) 0.615*** (0.09) -0.154* (0.08) 0.468*** (0.07) 0.615*** (0.09)

-0.746*** (0.15) 0.094* (0.04) -0.057 (0.07) -0.180* (0.09) 0.158* (0.07) 0.190*** (0.04) -0.036 (0.09) 0.042 (0.05) 0.367*** (0.07) 0.475*** (0.09) -0.071 (0.09) 0.050 (0.05) 0.261*** (0.08) 0.261*** (0.08) 0.290*** (0.05) 0.082 (0.04) -0.057 (0.08) 0.472*** (0.08) -0.040 (0.08) 0.171* (0.07) 0.472*** (0.08)

-0.746 0.758 0.785 0.884 0.432 -0.024 -0.257 0.713 0.367 0.983 -0.218 0.755 -0.054 -0.054 0.290 0.714 0.458 1.087 0.000 0.468 1.087

Continued on next page

98

Chapter 4 Table 4.4 (continued)

Agreement EC-Slovak Republic EC-Slovenia EC-Switzerland-Liechtenstein EC-Syria EC-Turkey Economic Cooperation Organisation (ECO) European Economic Area (EEA) European Free Trade Association (EFTA) EFTA-Bulgaria EFTA-Czech Republic EFTA-Estonia EFTA-Hungary EFTA-Israel EFTA-Latvia EFTA-Lithuania EFTA-Poland EFTA-Romania EFTA-Slovak Republic EFTA-Slovenia EFTA-Turkey Faroe Islands-Norway Faroe Islands-Switzerland Gulf Cooperation Council (GCC) Honduras-Panama Hungary-Israel Hungary-Turkey India-Maldives India-Nepal Israel-Poland Israel-Turkey Israel-US Laos-Thailand Latin American Integration Association (LAIA) Melanesian Spearhead Group (MSG) Mercado Comn del Sur (MERCOSUR) MERCOSUR-Bolivia Mexico-Nicaragua Organisation of Eastern Caribbean States (OECS) Romania-Turkey South Asian PTA/FTA (SAPTA/SAFTA) South Pacific Regional Trade and Economic Cooperation Agreement (SPARTECA/PACER) West African Economic and Monetary Union (WAEMU)

(1) β1

(2) β2

(3) Total

-0.257*** (0.07) -0.095 (0.10) 0.700*** (0.07) 0.729*** (0.08) 0.529*** (0.10) 0.394 (0.45) 0.735*** (0.08) 0.185*** (0.05) 0.165 (0.14) 0.152 (0.16) -0.589* (0.27) -0.026 (0.16) 0.188 (0.12) -0.589* (0.27) -0.589* (0.27) -0.073 (0.15) 0.211 (0.14) 0.152 (0.16) -0.526* (0.23) 0.408** (0.16) 0.721** (0.28) -1.389*** (0.38) 0.102 (0.35) -0.066 (0.19) -0.733** (0.24) -0.309 (0.47) -1.212* (0.53) -1.419*** (0.35) 0.303* (0.13) -1.348*** (0.26) 0.633* (0.25) 1.421* (0.69) -0.160 (0.18) 0.498 (0.35) 0.449 (0.41) -0.385 (0.38) -1.029 (0.68) 0.277 (0.32) -2.555*** (0.77) -0.792* (0.33) 0.572* (0.25)

-0.036 (0.09) 0.452*** (0.08) 0.042 (0.04) 0.049 (0.07) 0.447*** (0.09) -0.646 (0.48) -0.026 (0.07) 0.023 (0.05) -0.352 (0.22) -0.936** (0.29) 0.314 (0.28) -0.984*** (0.20) -0.522** (0.20) 0.314 (0.28) 0.314 (0.28) -0.726*** (0.21) -0.389 (0.22) -0.936** (0.29) 0.015 (0.29) -1.013*** (0.26) -0.816 (0.78) 0.508 (0.29) 1.105*** (0.29) 0.162 (0.21) 0.396* (0.18) 0.673*** (0.16) 0.789 (0.63) 0.439 (0.73) 0.079 (0.35) 1.021*** (0.30) -0.374** (0.14) -1.700* (0.80) 0.409* (0.16) -1.343*** (0.37) -0.622 (0.56) 0.776* (0.39) 1.902*** (0.40) -0.082 (0.31) 2.791*** (0.45) 0.088 (0.27) 0.522* (0.26)

-0.257 0.452 0.700 0.729 0.976 0.000 0.735 0.185 0.000 -0.936 0.000 -0.984 -0.522 0.000 0.000 -0.726 0.000 -0.936 0.000 -0.605 0.721 -1.389 1.105 0.000 -0.733 0.673 0.000 -1.419 0.000 -0.327 -0.374 0.000 0.000 -1.343 0.000 0.000 1.902 0.000 0.236 0.000 0.000

-0.567* (0.27)

0.008 (0.36)

0.000

Notes: Based on 5-year averages. The total treatment effect is the sum of statistically significant EIA parameter estimates at a 5 percent level. Estimates marked ***/**/* are significant at the 1/5/10 percent level. Robust standard errors (clustered by country-pair) are in brackets. Other estimates are omitted to save space.

Do we really know that trade agreements increase trade?

99

The results from column 3 indicate that unique treatment effects could be estimated for 89 agreements. Surprisingly, a majority of 38 EIAs (42.7 percent) have a net effect of zero, while the effect of 29 others (32.6 percent) is positive and the remaining 22 agreements (24.7 percent) have a positive effect on trade. Agreements for which the overall effect is zero include ASEAN, CEFTA, MERCOSUR and NAFTA. Trade agreements with an overall positive treatment effect include AMU, CACM, EC, EEA and GGC. In contrast, negative treatment effects can be observed for ACS, several EC accession treaties, ECOWAS, a number of EFTA agreements with Eastern-European partners and the Israel-US FTA. Unfortunately, there seems to be no clear explanation why some EIAs have had trade-promoting effects while others have not. The variation in these individual effects certainly suggests that one EIA is not be the same as the other. As will become apparent in the following pages, the issue of EIA heterogeneity may be so complicated that all of chapter 5 is exclusively devoted to studying this topic in greater detail. For now, the remainder of this study explores whether our individual EIA effects can be categorised by various characteristics such as their year of enforcement, the number of participants, the members’ development status and the geographic scope of the agreements.

Year of enforcement

Starting with the year of enforcement, note that the oldest

agreements (including the EC) are not visible in Figure 4.1 because there is, at most, a single observation per year. The bar-and-whisker plots can be interpreted as follows. The bars indicate the 25th to the 75th percentiles. The vertical line in the bar indicates the median. The horizontal lines, i.e., the whiskers, run from the 5th to the 95th percentiles. Interestingly, we find that agreements enforced before the 1990s have been generally trade enhancing, while several agreements signed afterwards have performed more poorly. One explanation may be found in the relatively less developed and hence uncompetitive state of the Eastern European, former USSR and small Latin American economies that typically signed EIAs in this period (Pomfret, 2007).

Membership base One dimension of interest is the number of countries participating in the agreement. It has been suggested that one of the important reasons why countries have turned to bilateral agreements is because it is easier for a small number of like-minded governments to negotiate an ambitious trade agreement with significant liberalisation, as opposed to “weak” agreements negotiated by a

100

Chapter 4

large group of governments with a larger variety of diverging interests (see Bhagwati, 2008). On the other hand, of course, it is also possible that a large number of parties at the negotiating table, envisioning the potential gains (e.g., market size), have a larger incentive to successfully reach a deal than would be the case in a bilateral situation. To our knowledge, there is no empirical literature that investigates whether “large” trade agreements perform differently from “small” ones. The dots in Figure 4.2 suggest that agreements between two countries vary widely in their effectiveness, although the bar and whisker plots indicate that agreements of up to 6 participants tend to have a positive effect on trade. The evidence is mixed for agreements with 7 to 15 members, although larger agreements fare better. Further research may explore why this would be the case. Development status Countries may be categorised as developed, developing or least developed country (LDC, based on World Bank, 2011a). This is roughly similar to the (developed) North-(developing) South paradigm familiar in the international trade and economic growth literature. The supposition is that developed countries are better integrated in the global economy and have superior access to the relevant resources and capabilities to produce commodities that can be sold on the international market. These nations are more likely to further gain from trade agreements offering preferential market access than underdeveloped counterparts that have poor access to the international market. Figure 4.3 shows that the development status of an EIA’s members does have some bearing on its effectiveness. As far as the OLS estimates are concerned, EIAs only run the risk of having a negative effect on trade when the membership base consists entirely of LDCs or a mix of developing nations and LDCs. This would seem to suggest that EIAs increase their chance of success when at least one developed nation participates in the agreement. However, correcting for endogeneity prevents any straightforward interpretation of the results. The FD estimates indicate that agreements with mixed membership tend to fare better than agreements among “equals”. Geographic scope

Finally, one could differentiate between intraregional and in-

terregional agreements. As explained with the example of the EC and Israel-US FTA in the introduction, the political and economic motives to enforce EIAs may differ and be reflected in their geographic scope. Nevertheless, Figure 4.4 does not show any remarkable differences between intra- and interregional agreements.

Do we really know that trade agreements increase trade? Figure 4.1. Individual estimates by year of enforcement

Figure 4.2. Individual estimates by number of participants

101

102 Figure 4.3. Individual estimates by development status

Figure 4.4. Individual estimates by geographic scope

Chapter 4

Do we really know that trade agreements increase trade?

4.4

103

Discussion and conclusion

Answering the question posed by the title of this chapter, the answer is a qualified “Yes”. Strikingly, traditional estimates of the EIA effect are exaggerated when possible endogeneity bias is not taken into account. A standard application of the gravity equation suggest that overall, EIAs boost trade by 140 percent. However, accounting for endogeneity with first-differencing yields substantially lower treatment effects. Then again, the findings also suggest that EIAs become increasingly more effective while their liberalisation schedules become gradually phased in. This chapter shows that EIAs increase trade by “just” 40 percent once endogeneity, anticipation effects and phase-in effects are taken into consideration. Taken together, it can be concluded, on the one hand, that researchers should be weary of overstating EIA’s effects on trade when they ignore endogeneity bias; on the other hand, disregarding the fact that trade liberalisation schemes are gradually enforced over a long period of time may underestimate the EIA effects. This is also found to be true at the level of individual agreements. Moreover, significant variation is found in the effectiveness of individual EIAs. This supports the idea that—contrary to gravity equation’s binary approach—EIAs are heterogeneous in nature. Sensitivity analyses suggest that the overall effectiveness of EIAs do not seem to be highly motivated by characteristics such as trade partners’ development status, the number of countries participating in the agreement or its geographic scope.2 What explains the remarkably lower EIA effects reported in this chapter? Three ideas are offered. The first explanation is the presence of unilateral trade liberalisation. Consider the fact that unilateral trade liberalisation is not captured by an agreement, but solely based on a policy decision single-handedly imposed by a government. This could be in force long before any economic integration agreement has been signed. In the presence of unilateral liberalisation, countries signing EIAs may just be confirming the liberalisation already enforced under their unilateral policies. Even if the agreement introduces new venues for trade liberalisation, the bulk of it may have already been enforced before the agreement was conceived. For more on the prevalence of unilateral trade policy, see Baldwin (2010). 2 This finding raises the idea that EIAs serve a network function, connecting a broad range of countries and commodities. As the global value chain is continually sliced into smaller pieces, it is not the separate parts comprised of EIAs or regions that matter as much for international trade as the sum of these parts in an intertwined, interconnected network of economic integration. This avenue is further explored in Alba, Hur & Park (2010).

104

Chapter 4 Participation in the World Trade Organisation (WTO) may be a second explan-

ation. With most economies now being WTO members, it is a possibility that the bulk of trade liberalisation is effectuated by means of the Most Favoured Nations (MFN) tariffs, which leaves little room for further substantial liberalisation with EIAs. As discussed at length in chapter 2, a large literature confirms that both WTO and regionalism fosters trade. A third explanation has to do with limitations to the scope of the individual trade agreement itself. It may well be the case that agreements do radically liberalise trade among participants, but only for a small number of goods that do not significantly impact aggregate trade. Hence, an EIA may well bring about substantial liberalisation, but only in a limited amount of product categories. This is an obvious drawback in using aggregate trade data. A growing body of literature is focusing on trade patterns in disaggregate trade flows. How the product-level liberalisation details affect these patterns is interesting for future research, which may provide a better understanding of the detailed (economic) mechanisms in EIAs. The results presented in this chapter suggest that trade agreements cannot be viewed as homogeneous vehicles that can be systematically used by governments to implement trade liberalisation, gain market access and stimulate economic integration. Some work, many do not. A number of explanations have been offered why many agreements are less effective than anticipated by the literature, arguing for a deeper understanding of the political-economy forces at work. Economists may benefit from not only knowing about trade policies, but also their effectiveness in the ongoing debate on the merits and threats of regionalism. At the same time, policy makers need to be aware that what looks good on paper may not necessarily result in the economic benefits they envision. The notion of EIA heterogeneity introduced in this chapter raises the following question: Do EIAs have varying outcomes on international trade flows because they differ by design? As argued above, trade agreements may vary in terms of their scope (trade in goods versus trade in goods and services) or the product groups included in the liberalisation schedules. The extent of the liberalisation, e.g., the degree to which applied tariffs are cut or abolished, may not be the same across all EIAs. The regulation of so-called “behind-the-border” non-tariff barriers such as technical barriers to trade, sanitary and phytosanitary regulations, and safety and labelling requirements may also differ. Dissimilarities may also arise from trade partners’ ability to settle disputes and the quality of the institutional framework needed to implement and maintain the agreements.

Do we really know that trade agreements increase trade?

105

If trade agreements differ by design and implementation, they may also have a heterogeneous impacts on cross-border trade. Chapter 5 takes a first step towards the issue of trade agreements’ design. Essentially, it looks “under the bonnet” of almost 300 EIAs and takes stock of the key legal provisions contained in each agreement. A measure—called the EIA comprehensives index—is developed to account for the number of provisions contained in these EIAs. Armed with quantified information on EIAs’ differences in design, the chapter then re-examines whether regionalism’s effect on trade can be related to its contents.

Chapter 5

I just read 296 trade agreements 5.1

Introduction

The expansion of economic integration agreements (EIAs) since the 1990s—which was illustrated in chapter 1—is a cause of concern for trade policy makers and economists. The issue at hand is whether these agreements could complement or undermine commitments made at the multilateral level of the World Trade Organisation (WTO). One view is that countries may strengthen their existing WTO commitments with a select number of partners who are willing and able to implement more aggressive liberalisation than multilaterally feasible. These trade-promoting forces are thought to ultimately seep into the multilateral trade system and stimulate liberalisation favouring other WTO members (see Baldwin, 1997). An opposing view is that countries’ frustration with lengthy and complex multilateral negotiations may provide incentives for governments to abandon or stall the process. Instead, they will prioritise on capitalising on faster and more flexible trade concessions with other like-minded partners. It is argued that by strengthening discriminatory trade policies with EIAs, potentially greater benefits that could have been extended to all through the WTO’s non-discriminatory MFN principle is are lost (see Bhagwati, 1993, 2008). Despite its merits, the empirical literature suggests that the discussion about whether regionalism curses or complements the multilateral trade system should no longer be focused on preferential tariffs alone. Why? First, the WTO’s World Trade Report (WTO, 2011e) finds that non-discriminatory (most-favoured nation, MFN) tariff bindings are continually decreasing lower levels.

108

Chapter 5

Half of world trade is even subject to zero MFN tariff rates. As discussed in chapter 2, this is a testimony to the extensive trade liberalisation that the multilateral trade system has fostered throughout its existence. But there is only so much tariffcutting that can be done, as tariff margins grow increasingly thinner. Strikingly, Keck & Lendle (2011) find that this is true for both the WTO and EIAs. Of course, the Report also acknowledges that there are still ample product categories that enjoy high MFN rates. The “protection for sale” literature (discussed on p. 20) suggests that domestic import-opposing lobbies oppose trade liberalisation. Interestingly, Damuri (2009) shows that commodities with a high degree of protection at the level of the WTO are typically also exempted from liberalisation efforts enforced through regionalism. Taken together, these new facts give the distinct impression that, while tariffs have been largely cut multilaterally, industries that do remain protected in the WTO are also protected from EIAs. The original focus of the “building vs. stumbling block” debate is therefore running out of steam: it is no longer just about tariffs. But then, the first question that arises is: “Why do countries sign EIAs?” One study provides useful clues. Baier & Bergstrand (2004) investigate the economic determinants of trade agreements. The authors demonstrate that similar economic conditions in terms of GDP levels and capital per worker, geographic proximity and geographical remoteness contribute to the likelihood that countries enforce a common EIA. The second question is: “What are they signing?” Chapter 4 extensively shows that there is considerable variation in the degree to which individual EIAs affect international trade. It argues that trade agreements may differ by design, i.e., the undertakings to which the trade partners are legally bound will vary, depending on the intention and outcome of the negotiations. But does the common usage of binary variables, which indicate the lack or presence of an EIA between a pair of countries, do justice to the heterogeneous nature of the underlying trade agreements? Although it is certainly convenient to make the simplifying assumption that trade agreements are either “fully present” or “fully absent” in regression analyses, insightful information stemming from inherent differences between EIAs is ignored. In chapter 4, we argued that EIAs may differ by design and, in turn, have different outcomes on trade. Paraphrasing George Orwell, “All EIAs are equal, but some are more equal than others”. As such, the central aim of this chapter is to assess differences between EIAs in terms of the provisions that they contain. This

I just read 296 trade agreements

109

information can then be used to construct a quantifiable measure of EIA’s comprehensiveness. In doing so, this chapter explores the value-added of explicitly accounting for EIA heterogeneity in empirical studies of international trade. It will be shown, for example, that identifying what countries agree to in their EIAs may also explain why they sign them. The contributions and organisation of this chapter are as follows. The first contribution is in terms of improved data coverage. In seminal work, Horn et al. (2010) study the contents of 31 EIAs involving either the European Community (EC) or US. In contrast, this chapter extends the sample to contain no less than 296 trade agreements among WTO members and non-members alike, making it the most exhaustive survey to date. Section 5.2 takes stock of the provisions contained in these EIAs and identifies the policy domains contributing to the most prominent similarities and differences among the agreements. In doing so, it addresses the question what nations are signing. The second contribution is quantification. Section 5.3 uses the information on the EIAs’ provisions to quantify the extent to which the agreements’ contents are comprehensive. This measure, called the EIA comprehensiveness index, is subsequently used to shed light on the question as to why nations sign EIAs. Strikingly, WTO membership will be shown to be of significant importance. Finally, we explore whether the EIA dummy can successfully be replaced by the “smarter” EIA index in a typical setting of the gravity equation. This is done in section 5.4 to determine how an EIA’s comprehensiveness is associated with its estimated effect on international trade. Remarkably, not all provisions contained in EIAs are trade-promoting. Provisions that are in line with WTO regulations are found to be trade-promoting, while measures that go beyond the WTO’s mandate actually decrease trade. Section 5.5 discusses these outcomes and concludes.

5.2 5.2.1

What is in an EIA? Literature

Horn, Mavroidis & Sapir (2010) provide the first systematic study of 17 EIAs involving the EC and 14 involving the US. The authors take stock of the various policy areas that are covered by the undertakings laid out in these agreements, paying attention to (1) the legal enforceability of the provisions and (2) the extent to which the undertakings are included in the WTO’s mandate.

110

Chapter 5 Ad 1. For each provision identified by Horn et al. (2010), they did not only

account for a provision being covered by an EIA, but also for its legal enforceability. This is because a policy area could be covered, but the undertaking may be too imprecisely formulated to give rise to a legal obligation that would be enforceable in the event of a dispute settlement proceeding. Provisions are considered to be legally enforceable only if the undertaking “specified at least some obligation that is clearly defined and that is likely to effectively bind the parties” (Horn et al., 2010, p. 1572). It may also be the case that undertakings are not legally enforceable because they are explicitly excluded from the EIA’s dispute settlement procedures. Ad 2. Provisions that confirm countries’ existing multilateral obligations and that may also deepen such commitments are categorised as “WTO+” provisions. Examples of WTO+ provisions are measures on anti-dumping, restrictions on state aid and the liberalisation of trade in services. In contrast, “WTOX” provisions involve policy areas that are not covered by the WTO’s current mandate and may compromise the WTO’s ability to expand into these legal territories with binding, non-discriminatory policy. Examples range from anti-terrorism to environmental and labour market regulations. Horn et al. find that both the EC and US are strongly committed to legally enforceable WTO+ undertakings, although the EC emphasises obligations on state trading enterprises (STEs) more than the US. In turn, the US focuses on traderelated investment measures (TRIMs), technical barriers to trade (TBT) and trade in services (GATS). WTOX provisions feature more prominently in the EC’s agreements, but often lack enforceability. However, both trade powers also have credible WTOX commitments. The World Trade Report (WTO, 2011e) extends Horn et al.’s coverage to 96 EIAs and shows that traditional WTO+ provisions on tariff liberalisation are abundant and legally enforceable. This also applies to the newer WTO+ policy areas such as intellectual property rights and investment and WTOX areas on competition policy and capital mobility. The contribution of this section is that it takes stock of the content of 296 EIAs, thereby including almost all EIAs that have been enforced to date. It builds on Horn et al. (2010) and WTO (2011e) by examining the coverage and legal enforceability of 13 WTO+ and 4 WTOX policy areas and introduces 9 indicators of institutional quality (IQ). The remainder of this section is structured as follows. The coding approach employed in this study is described at length in section 5.2.2, followed by an overview of the underlying EIAs in section 5.2.3. Section 5.2.4 provides two sets of results.

I just read 296 trade agreements

111

First, it compares the methodological approaches by benchmarks the present dataset with that of Horn et al. (2010). Second, descriptive statistics of the main trends in the data are presented.

5.2.2

Approach

This study draws on the Global Preferential Trade Agreements Database (GPTAD), which was developed by the World Bank and the Tuck Centre for International Business World Bank (2011c). GPTAD is an extensive database that contains the legal texts of virtually all trade agreements that have been enforced in the postwar period. Its unique feature is that it makes the agreements text searchable for a large number of keywords. The provisions contained in each agreement have been classified according to WTO criteria, which allows the user to compare provisions across agreements. So, a researcher interested in measures on anti-dumping and countervailing measures may search the database with these keywords. All agreements containing provisions on this topic will then be listed, along with the relevant chapters, titles and/or articles for each agreement. As discussed in WTO (2011e), the number of policy areas depends on the identification strategy. Horn et al. (2010), for instance, use chapter and article headings of the agreements in their sample to reach a total of 52 policy areas. An alternative would be to compile a detailed list of each and every single policy area that could conceptually be included in an EIA. Although this approach has the merit of exhaustiveness and precision, which is arguably a preferred route when analysing a limited set of agreements, it introduces even more complexity when the objective of a study such as this one is to identify the key areas of importance for a substantial number of EIAs. The provisions identified in the present study are as follows. First, GPTAD features 13 WTO+ policy areas. These provisions are part of the WTO’s current mandate and are listed in Table 5.1, along with a brief intuition of they are related to trade. Table 5.2 describes the purpose of an additional 4 WTOX policy domains that extend beyond the scope of the WTO. Finally, details on 9 indicators of the agreement’s institutional quality (IQ) are provided in Table 5.3.

112

Chapter 5

Table 5.1. Provisions in GPTAD: WTO+ Provision

Description

Agriculture

Agreement to liberalise trade in agricultural commodities by reducing/abolishing barriers to trade such as tariffs, quotas and subsidies. Agreement to harmonise agricultural policies may also be included. Undertakings may be in line with, deepen and/or broaden the scope of provisions specified in the GATT 1994/WTO Agriculture Agreement.

Anti-Dumping & Countervailing Measures (AD & CVM)

Agreement with rules on anti-dumping and countervailing measures that specify the conditions under which parties may deviate from their liberalisation commitments to offset injury caused by dumping. Undertakings may be in line with, deepen and/or broaden the scope of provisions specified in the GATT 1994/WTO Agreement on Subsidies and Countervailing Measures (SCM Agreement).

Customs Administration

Agreement to reduce administrative barriers to trade by simplifying customs administration with respect to issues such as import licensing requirements, valuation and nomenclature. Undertakings may be in line with, deepen and/or broaden the scope of provisions specified in the GATT 1994/WTO Agreement on Import Licensing Procedures.

Export Restrictions

Agreement to liberalise duties, charges and/or quantitative restrictions on exported goods. Undertakings may be in line with, deepen and/or broaden the scope of provisions specified in the GATT 1994.

Import Restrictions

Agreement to liberalise duties, charges and/or quantitative restrictions on imported goods. Undertakings may be in line with, deepen and/or broaden the scope of provisions specified in the GATT 1994.

Intellectual Property Rights (IPR)

Agreement on the protection of intellectual property rights (copyrights, patents, trademarks, etc.) in foreign markets. Undertakings may be in line with, deepen and/or broaden the scope of provisions specified in the WTO Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPS Agreement).

Investment

Agreement to prohibit discriminatory trade-related investment practices such as local content requirements, trade balancing requirements and foreign exchange restrictions. Undertakings may be in line with, deepen and/or broaden the scope of provisions specified in the WTO Agreement on Trade-Related Investment Measures (TRIMS).

Public Procurement

Agreement to grant access to foreign parties and further liberalise the market for public procurement. Undertakings may be in line with, deepen and/or broaden the scope of provisions specified in the WTO Agreement on Government Procurement (GPA).

Sanitary & Phytosanitary Measures (SPS)

Agreement to simplify and/or harmonise import requirements with respect to food safety and animal and plant health. Undertakings may be in line with, deepen and/or broaden the scope of provisions specified in the WTO Sanitary and Phytosanitary (SPS) Agreement.

Services

Agreement to liberalise trade in services. Undertakings may be in line with, deepen and/or broaden the scope of provisions specified in the General Agreement on Trade in Services (GATS).

State Aid

Agreement to restrict any form of aid that could give rise to unfair competitive advantages. Undertakings may be in line with, deepen and/or broaden the scope of provisions specified in the GATT 1994/WTO Agreement on Subsidies and Countervailing Measures (SCM Agreement).

State Trading Enterprises (STE)

Agreements to ensure market access and non-discriminatory behaviour by governmental enterprises. Undertakings may be in line with, deepen and/or broaden the scope of provisions specified in the GATT 1994.

Technical Barriers to Trade (TBT)

Agreements to reduce barriers to trade by simplifying and harmonising standards and technical barriers such as testing and certification procedures. Undertakings may be in line with, deepen and/or broaden the scope of provisions specified in the WTO Agreement on TBT.

Source: World Bank (2011c).

I just read 296 trade agreements

113

Table 5.2. Provisions in GPTAD: WTOX Provision

Description

Capital Mobility

Agreement to improve capital mobility by relaxing restrictions on foreign capital and facilitating cross-border financial transfers.

Competition

Agreements on competition policy to restrict or prohibit monopolies’ activities to promote undistorted competition.

Environment

Agreement to uphold environmental laws, provided that they are not used as disguised barriers to trade. Commitments to enforce environmental laws so as not to attract (foreign) business activity that would exploit environmental resources

Labour

Agreement to uphold labour laws so as not to attract (foreign) business activity that would exploit employees and/or to facilitate labour mobility.

Source: World Bank (2011c).

Table 5.3. Provisions in GPTAD: Institutional Quality Provision

Description

Consultations

Signatories wishing to address issues arising from the implementation of the EIA, or their broader economic partnership in general, may engage in a diplomatic dialogue known as consultations “with a view to finding a mutually satisfactory solution”. When specified, consultation procedures provide details on when and where consultations are to be held, which parties (e.g. non-governmental organisations, external advisors, etc.) may be allowed to attend, and the issues that may be addressed. In most cases, signatories must first attempt to solve disputes according to consultation procedures before having access to the EIA’s dispute settlement mechanism.

Definition

By providing definitions of key concepts, signatories increase the clarity, scope and certainty of their commitments.

Dispute Settlement

By agreeing on dispute settlement procedures, signatories reduce ambiguity and create a judicially binding mechanism that ensures the implementation of the EIA.

Duration & Termination

Signatories reduce ambiguity about their commitments by specifying the duration of the EIA and the means by which it can be terminated.

Evolutionary Clause

Signatories commit themselves to a built-in periodic review mechanism that facilitates amendments and improvements to the original EIA.

Institutional Framework

The signatories provide details on the institutional framework that will be used to oversee the implementation of the EIA.

Objectives

The signatories enhance the clarity and context of their commitments by specifying the objectives they envision by signing the EIA.

Plan & Schedule

The signatories commit themselves to a specific timetable by detailing the schedule according to which the EIA is to be implemented.

Transparency

The signatories commit themselves to creating greater institutional transparency, e.g. by agreeing on how and when information on economic policy will be shared.

Source: World Bank (2011c).

114

Chapter 5 Having identified the provisions that can be extracted from GPTAD, every EIA

can now be coded. GPTAD is used to assign a binary variable to each policy area that is covered by the EIA under investigation. Policy areas for which the agreement contains a provision are coded 1 and 0 otherwise. In order for a provision to be considered “covered” (C) and scored 1, all that is needed is for the provision to reflect agreement by both parties to somehow cooperate with a view of trade liberalisation. The issue of legal enforceability is not relevant at this stage. Hence, a provision calling for an exchange of Parties’ information on their environmental policies would score a 1, but so would provisions that prohibit lax enforcement of environmental policies aimed at attracting foreign investment. The odd provisions that only state that Parties reserve the right to protect their natural resources are scored 0 because such measures do not require any form of agreement or cooperation. A provision that is also deemed to be legally enforceable scores 1 for “enforceability” (E). Our criteria reflect and build on those laid down by Horn et al. (2010). Provisions that are legally enforceable typically use the word “shall”. For example: “Parties shall grant service providers treatment no less favourable than that accorded to their own.” Timing is also important. A provision calling for gradual liberalisation of government procurement policies, without indicating the date by which the liberalisation must be complete, scores 0. This is because it is unclear when the Party must be able to meet that particular requirement. Provisions stating that Parties “shall negotiate,” “shall consider” or “shall cooperate” are also difficult to be enforced. Negotiations may still fail and not abolish trade barriers. It also seems very unlikely that it would be able to prove that Parties have not given due consideration to a matter or that they have not cooperated.1 For clarity and ease of replication, several excerpts from actual trade agreements are provided below. Table 5.4 shows examples of WTO+ provisions and classifies them as being either enforceable or non-enforceable, based on the criteria discussed above. The same is done for WTOX provisions in Table 5.5.

1 All IQ provisions enjoy a 100 percent legal enforceability because they provide the underlying organisational mechanism that are needed to implement the agreed upon commitments, including consultations and dispute settlement.

I just read 296 trade agreements

115

Table 5.4. Coding examples: WTO+ Provision

Covered (1), Enforceable (0)

Covered (1), Enforceable (1)

AD & CVM

(. . . ) The provisions of this Article shall not be subject to the dispute settlement provisions of this Agreement.

Each Party retains its rights and obligations under Article VI of GATT 1994 and the WTO Agreement, and their successors, with regard to the application of antidumping and countervailing duties.

Customs Administration

The Member States recognise that the objectives of this Agreement may be promoted by harmonisation of customs policies and procedures in particular cases. Accordingly the Member States shall consult at the written request of either to determine any harmonisation which may be appropriate.

The Parties shall apply the provisions of Article VII of GATT 1994 and the WTO Agreement on the Implementation of Article VII of GATT 1994 for the purposes of determining the customs value of goods traded between the Parties.

IPR

Each Party, recognising the importance of protecting intellectual property in further improving the business environment in the Party, shall: (a) endeavour to improve its intellectual property protection system; (b) comply with the obligations set out in the international agreements relating to intellectual property to which it is a party; (c) endeavour to become a party to international agreements relating to intellectual property to which it is not a party; (d) endeavour to ensure transparent and streamlined administrative procedures concerning intellectual property; (e) endeavour to ensure adequate and effective enforcement of intellectual property rights; and (f) endeavour to further promote public awareness of protection of intellectual property.

The Parties agree that the WTO Agreement on Trade-Related Aspects of Intellectual Property Rights shall govern and apply to all intellectual property issues arising from this Agreement. Each Party affirms its rights and obligations with respect to each other Party under the TRIPS Agreement. Each Party shall accord to the nationals of each other Party treatment no less favourable than it accords to its own nationals with regard to the protection1 of intellectual property, subject to the exceptions provided in the TRIPS Agreement and in those multilateral agreements concluded under the auspices of WIPO. The Parties shall grant and ensure adequate and effective protection of intellectual property rights on a non-discriminatory basis, including effective measures for enforcing such rights against infringement, and particularly against counterfeiting and piracy.

Continued on next page

116

Chapter 5 Table 5.4 (continued)

Provision

Covered (1), Enforceable (0)

Covered (1), Enforceable (1)

Investment

To promote investments, the Parties agree to enter into negotiations in order to progressively liberalise the investment regime. To promote investments and to create a liberal, facilitative, transparent and competitive investment regime, the Parties agree to enter into negotiations in order to progressively liberalise their investment regimes, strengthen cooperation in investment, facilitate investment and improve transparency of investment rules and regulations, and provide for the protection of investments. If a Party grants to a non-Party, after the entry into force of this Agreement, a more favourable investment framework than under this Agreement, it shall afford adequate opportunity to the other Parties to seek to obtain, including through possible negotiations, comparable conditions, on a mutually beneficial basis.

The Sides will not: - impose local taxes or charges, directly or indirectly on goods, covered by the present agreement, of another Side, at the rate that exceeds the level of relevant taxes or charges imposed on analogous goods of the local production or those produced in third countries; - introduce special restrictions or demands towards export and import of goods, covered by the present agreement, that in similar cases are not used towards analogous goods of the local production or those produced in third countries;- use different rules towards warehousing, unloading, storage, shipment of goods, originated from another country to the agreement, as well as towards repayments and remittances, with the exception of rules that in similar cases are used towards domestic goods or those originated from third countries.

Public Procurement

The Parties will progressively develop their respective rules, conditions and practices on public procurement and shall grant suppliers of the other Party access to contract award procedures on their respective public procurement markets not less favourable than that accorded to companies of any third country. The Parties consider the liberalisation of their respective public procurement markets as an objective of this Agreement. The Parties aim at opening up of the award of public contracts on the basis of non-discrimination and reciprocity. The Parties shall, subject to their laws, regulations and policies, exchange information in respect of their government procurement policies and practices.

The Parties consider the opening up of the award of public contracts on the basis of nondiscrimination and reciprocity, to be a desirable objective. 2. As of the entry into force of this Agreement, both Parties shall grant each others companies access to contract award procedures a treatment no less favourable than that accorded to companies of any other country.

Continued on next page

I just read 296 trade agreements

117 Table 5.4 (continued)

Provision

Covered (1), Enforceable (0)

Covered (1), Enforceable (1)

SPS

The Parties shall aim to reduce differences in standardisation and conformity assessment. To this end the Parties shall conclude where appropriate agreements on mutual recognition in the field of conformity assessment.

Each party affirms its rights and obligations with respect to each other Party under the SPS Agreement. Each Party undertakes not to adopt or maintain any prohibition or quantitative restriction on the importation of any goods of the other Parties or on the exportation of any goods destined for the territory of the other Parties, except in accordance with its WTO rights and obligations or other provisions in this Agreement. The Parties reaffirm the rights and obligations relating to SPS measures under the SPS Agreement among those Parties that are parties to the said Agreement. The Parties shall apply their regulations in sanitary and phytosanitary matters in a nondiscriminatory fashion and shall not introduce any measures that have the effect of unduly obstructing trade.

Services

The Parties agree to enter into negotiations to progressively liberalise trade in services with substantial sectorial coverage. Each Party shall provide free transit over the territory of its country for goods originated within the customs territory of the other Party or having originated in third countries and destined for the customs territory of the other Party or any third country, and shall supply the exporters, importers, and shipping companies involved in such transit operations with all the available resources and services required for the execution of these transit operations on terms (including financial) that are not worse than the terms for providing the same resources and services to exporters, importers, and national shipping companies of any other third country. Contracting Parties shall conclude a special agreement on transit.

Each Party shall accord services and service suppliers of any other Party treatment no less favourable than that provided by those of the Party. There shall be free movement of services.

Continued on next page

118

Chapter 5 Table 5.4 (continued)

Provision

Covered (1), Enforceable (0)

Covered (1), Enforceable (1)

State Aid

The Parties shall review the issue of disciplines on subsidies related to trade in services in the light of any disciplines agreed under Article XV of GATS with a view to their incorporation into this Agreement.

Each Party agrees to eliminate and not reintroduce all forms of export subsidies for agricultural goods destined for the other Parties. The following are incompatible with the proper functioning of this Agreement in so far as it affects trade between the Contracting Parties: any state aid which distorts or threatens to distort competition by favouring certain undertakings or the production of certain goods. Contracting Parties shall not use state aid in the form of subsidies to enterprises or in any other form if the result of such state aid would be the distortion of normal economic conditions in the territory of the other Contracting Party. The Parties confirm their rights and obligations arising from the WTO Agreement on Subsidies and Countervailing Measures.

STE

The Contracting Parties shall adjust progressively any state monopoly of a commercial character so as to ensure that no discrimination regarding the conditions under which goods are procured and marketed exists between nationals of the Contracting Parties.

The Parties shall adjust progressively any state monopoly of a commercial character so as to ensure that by the date of entry into force of this Agreement, no discrimination regarding the conditions under which goods are procured and marketed exists between nationals of the Parties. Each Party shall ensure that any state monopoly supplier of a service in its Area does not, in the supply of the monopoly service in the relevant market, act in a manner inconsistent with the Partys commitments under this Chapter. The States Parties to this Agreement shall ensure that any state monopoly of a commercial character be adjusted, subject to the provisions laid down in Protocol D, so that no discrimination regarding the conditions under which goods are procured and marketed will exist between nationals of Party 1 and of Party 2.

Continued on next page

I just read 296 trade agreements

119 Table 5.4 (continued)

Provision

Covered (1), Enforceable (0)

Covered (1), Enforceable (1)

TBT

The parties agree to strengthen their cooperation in measures including technical barriers to trade/non-tariff measures. The Member States shall:(a) examine the scope for taking action to harmonise requirements relating to such matters as standards, technical specifications and testing procedures, domestic labelling and restrictive trade practices; and (b) where appropriate, encourage government bodies and other organisations and institutions to work towards the harmonisation of such requirements.

Member States shall eliminate other non-tariff barriers on a gradual basis within a period of five years after the enjoyment of concessions applicable to those products. Each Party undertakes not to adopt or maintain any prohibition or quantitative restriction on the importation of any goods of the other Parties or on the exportation of any goods destined for the territory of the other Parties, except in accordance with its WTO rights and obligations or other provisions in this Agreement. The Parties reaffirm the rights and obligations relating to standards, technical regulations and conformity assessment procedures under the TBT Agreement among those Parties that are parties to the said Agreement. The rights and obligations of the Parties, relating to technical barriers to trade (technical regulations, standards and conformity assessment procedures) and the respective measures, shall be governed by the WTO Agreement on Technical Barriers to Trade.

Note: Excerpts from various EIAs obtained via World Bank (2011c).

120

Chapter 5

Table 5.5. Coding examples: WTOX Provision

Covered (1), Enforceable (0)

Covered (1), Enforceable (1)

Capital Mobility

Not available

Each Party shall permit all transfers relating to a covered investment to be made freely and without delay into and out of its territory. Such transfers include: (a) contributions to capital; (b) profits, dividends, interest, capital gains, royalty payments,management fees, and technical assistance and other fees; (c) proceeds from the sale of all or any part of the covered investment or from the partial or complete liquidation of the covered investment; (d) payments made under a contract entered into by the investor, or the covered investment, including payments made pursuant to a loan agreement; (e) payments made pursuant to paragraphs 1 and 2 of Article 10.6 and Article 10.11; and (f) payments arising under Section B. 2. Each Party shall permit returns in kind relating to a covered investment to be made as authorised or specified in a written agreement between the Party and a covered investment or an investor of the other Party. 3. Each Party shall permit transfers relating to a covered investment to be made in a freely usable currency at the market rate of exchange prevailing on the date of transfer.

Competition

The Commission shall adopt, at the General Secretariats proposal, the rules which are needed to guard against or correct practices which may distort competition within the Subregion, such as dumping, improper price manipulations, manoeuvres made to upset the normal supply of raw materials and others with a like effect. In this respect, the Commission shall consider the problems that could derive from the imposition of levies and other restrictions on exports.

Where a Partys monopoly supplier competes, either directly or through an affiliated company, in the supply of a service outside the scope of its monopoly rights and which is subject to that Partys specific commitments, the Party shall ensure that such a supplier does not abuse its monopoly position to act in its territory in a manner inconsistent with such commitments.

Continued on next page

I just read 296 trade agreements

121 Table 5.5 (continued)

Provision

Covered (1), Enforceable (0)

Covered (1), Enforceable (1)

Environment

Member Countries shall undertake joint policies that enable a better use of their renewable and non-renewable natural resources and the preservation and improvement of the environment.

A Party shall not fail to effectively enforce its environmental laws, through a sustained or recurring course of action or inaction, in a manner affecting trade between the Parties, after the date of entry into force of this Agreement. Subject to the requirement that such measures are not applied in a manner which would constitute a means of arbitrary or unjustifiable discrimination between the Parties where the same conditions prevail, or a disguised restriction on international trade, nothing in this Chapter shall be construed to prevent the adoption or enforcement by a Party of measures: (a) necessary to protect public morals; (b) necessary to protect human, animal or plant life or health. Each Party recognises that it is inappropriate to encourage investments by investors of the other Party by relaxing its environmental measures. To this effect each Party should not waive or otherwise derogate from such environmental measures as an encouragement for establishment, acquisition or expansion of investments in its Area.

Labour

Cooperation between the Parties will complement the cooperation set out in other Chapters of this Agreement. Areas of cooperation may include but should not be limited to: science, agriculture including the wine industry, food production and processing, mining, energy, environment, small and medium enterprises, tourism, education, labour, human capital development and cultural collaboration. Cooperation on labour and employment matters of mutual interest and benefit will be based on the concept of decent work.

Neither Party shall require labour market testing, labour certification tests or other procedures of similar effect as a condition for temporary entry in respect of natural persons on whom the benefits of this Chapter are conferred. Each Party shall grant entry and temporary stay to nationals of the other Party in accordance with this Chapter including the provisions of Annex 13.

Note: Excerpts from various EIAs obtained via World Bank (2011c).

122

Chapter 5 It is useful to reflect on some of the comments on Horn et al. (2010)’s notion of

legal enforceability that are discussed in WTO (2011e, p. 129-130). Firstly, the question whether an undertaking is sufficiently formulated to give rise to an obligation is a matter of treaty interpretation. For example, dispute settlement rulings of the WTO Appellate Body have shown that obligations may arise from statements using the word “should” instead of “shall”. Secondly, provisions that are excluded from the agreements dispute settlement system may still be subject to dispute settlement flowing from commitments that the Parties may have elsewhere. This argument applies not only to WTO+ provisions related to other commitments at the WTO, but also to WTOX commitments arising from, for example, international treaties on labour standards and environmental protection. Thirdly, the legal enforceability of a provision that allows the use of countermeasures to enforce rights or obligations may be limited by commitments stemming from other agreements. Finally, provisions not subject to dispute settlement may still be enforceable through political and diplomatic channels, but the reverse could also hold: it may not at all times be possible to enforce provisions which are subject to dispute settlement, due to political, non-legal and/or resource considerations. Indeed, it is important to acknowledge that there are limitations to the extent to which the legal enforceability of an undertaking can be determined with absolute certainty. However, it is still useful to differentiate, in one way or another, between those undertakings that give the reader some sense of concrete and imminent policy liberalisation and those that merely reflect a loosely defined agreement to explore possible avenues of future cooperation.

I just read 296 trade agreements

5.2.3

123

Data

Table 5.6 lists the 344 EIAs that are or have been enforced in the period 1948-2011. 296 of these agreements have been classified by GPTAD and are included in this study.2 As discussed above, this is the first study to date which quantifies the provisions contained in such a large number of EIAs. Earlier studies by Horn et al. (2010) and WTO (2011e) rely on restricted samples of 31 and 96 EIAs, respectively. Contrary to Horn et al. (2010) and following WTO (2011e), agreements with non-WTO members are included. This is useful to investigate possible differences or similarities between the nature and number of provisions contained in (non) WTO members’ trade policy commitments. Moreover, the sample is not restricted to only those agreements that have been notified to the WTO. The reason for this is because notification is neither a legal prerequisite for governments to be able to enforce an EIA, nor some form of WTO endorsement that it is a real EIA. Of the 296 EIAs in the sample, 193 have been notified to the WTO. Finally, agreements are included even if they have already expired. This is because these agreements also contain information about the areas for which their governments (at some point in time) enforced the specific trade policies that are of interest in this study.

2 The

complete dataset is available at http://www.tristankohl.org.

124

Chapter 5

Table 5.6. EIAs by year of enforcement Year

EIA

0.9) include the Australia-Singapore, EC, Japan-Switzerland and NAFTA. In contrast, GGC and PAFTA have low scores (